Cold, old and metal-poor: New stellar substructures in the Milky Way's dwarf spheroidals
aa r X i v : . [ a s t r o - ph . GA ] A p r Draft version April 25, 2019
Preprint typeset using L A TEX style emulateapj v. 12/16/11
COLD, OLD AND METAL-POOR: NEW STELLAR SUBSTRUCTURESIN THE MILKY WAY’S DWARF SPHEROIDALS
V. Lora , E. K. Grebel , S. Schmeja , and A. Koch Instituto de Radioastronom´ıa y Astrof´ısica,Antigua Carretera a P´atzcuaro 8701, 58089 Morelia, M´exico Astronomisches Rechen-Institut, Zentrum f¨ur Astronomie der Universit¨at Heidelberg,M¨onchhofstr. 12-14, 69120 Heidelberg, Germany and Technische Informationsbibliothek,Welfengarten 1b, 30167 Hannover, Germany
Draft version April 25, 2019
ABSTRACTDwarf spheroidal galaxies (dSph) orbiting the Milky Way are complex objects often with complicatedstar formation histories and internal dynamics. In this work, we search for stellar substructures infour of the classical dSph satellites of the Milky Way: Sextans, Carina, Leo I, and Leo II. We applytwo methods to search for stellar substructure: the minimum spanning tree method, which helps us tofind and quantify spatially connected structures, and the “brute-force” method, which is able to findelongated stellar substructures. We detected the previously known substructure in Sextans, and alsofound a new stellar substructure within Sextans. Furthermore, we identified a new stellar substructureclose to the core radius of the Carina dwarf galaxy. We report a detection of one substructure in Leo Iand two in Leo II, but we note that we are dealing with a low number of stars in the samples used.Such old stellar substructures in dSph galaxies could help us to shed light on the nature of the darkmatter halos, within which such structures form, evolve, and survive.
Subject headings: galaxies: dwarf — galaxies: evolution — astronomical databases: miscellaneous —methods: data analysis INTRODUCTION
Dwarf spheroidal galaxies (dSph) are the most com-mon type of galaxies in the Universe and often consid-ered to be the building blocks of more massive galax-ies in hierarchical formation scenarios (Dekel & Silk1986; Bullock & Johnston 2005; Cooper et al. 2013;Pillepich et al. 2015). The dSph satellites of the MilkyWay (MW) are the best studied dwarf galaxies, since in-dividual stars can be resolved and evolutionary historiescan be derived in great detail. A common property of alldSph galaxies is the presence of an old population (starswith ages of 10 Gyr and more), which in many casesturns out to be the dominant population e.g. (Grebel2000; Grebel & Gallagher 2004). There is growing evi-dence, that some dSphs (e.g., Carina, Fornax, Sculptor,and Sextans) experienced extended star formation his-tories or multiple episodes of star formation The gener-ally spatially extended old, metal-poor population, andspatially concentrated young, metal-rich population in-dicate extended star formation episodes in their centers(Harbeck et al. 2001).Some dSph galaxies also present dynamically cold stel-lar substructures. For example, Ursa Minor (UMi) showstwo distinct density peaks (Kleyna et al. 1998). One rep-resents the underlying galaxy’s field population with aradial velocity dispersion of σ = 8 . − . The seconddensity peak appears with a radial velocity dispersion of σ = 0 . − . This second peak is located on thenorth-eastern side of the major axis of UMi at a distanceof ∼ . [email protected] sub-population, possibly a disrupted cluster, i.e., a dy-namically cold stellar clump.These clumpy substructures indicate that there is ahigh degree of complexity already on small scales, whichit is also witnessed in other systems (such as in the dis-tribution of the globular clusters in the Fornax dSph),mergers between dwarf galaxies in the local Universe(Amorisco et al. 2014; del Pino et al. 2015) and a pos-sible merger event in the Large Magellanic Cloud (LMC)(Mackey et al. 2016).Another example is found in Sextans. Kleyna et al.(2004) reported the line-of-sight radial velocity disper-sion profile of Sextans based on a sample of 88 starsextending to about 1 . < −
2) component with a ve-locity dispersion of ∼
10 km s − , and a more concen-trated metal-rich component of extremely low velocitydispersion. On the other hand, Simon & Geha (2007)obtained 214 CVn I member stars that did not reveal anytrace of different populations as reported by Ibata et al.(2006). Ural et al. (2010) obtained spectroscopic datafor 26 stars in the dSph and investigated whether theirdata exhibit any evidence of multiple populations as pro-posed by Ibata et al. (2006) (under the assumption thateach population was Gaussian), but no clear signature ofdistinct sub-populations was found. They argue that thepossible detection of a sub-population rather depends onthe total number of stars in the data set, the fraction ofstars in the sub-population, the difference in velocity dis-persion between the populations, and the observationalerrors.A final example is the Andromeda II (And II) dSph.McConnachie et al. (2007) identified different stellarpopulations in And II for which they constructed ra-dial profiles and found that the horizontal branch hasa constant spatial distribution out to a large radius.In contrast, they found that the reddest stars on thered giant branch in And II are more centrally concen-trated. The latter stellar component has an average ageof ∼ ∼ − L V ∼ × L ⊙ and L V ∼ × L ⊙ , respectively) the formation of large coresvia stellar feedback is not obviously expected (Amorisco2017). If stellar feedback in faint galaxies is found to beinefficient, then alternative candidates for DM should be seriously considered.The detection of stellar substructures in dSphs, and theimplications of their survival in the core/cusp problem,prompted us to test other dwarf galaxies, and to inves-tigate if we could find stellar substructures within them.From velocities, metallicities, and positions of red giantstars in the Sextans, Carina, Leo I, and Leo II dSphs, weanalyze whether or not these dSphs could host dynami-cally cold debris.This article is organized as follows: In § §
3. The results are presented in §
4. Finally, we presentour conclusions in § THE FOUR DWARF SPHEROIDAL GALAXIES
Sextans
The Sextans dSph is located at a distance of 86 ± M V = − . ± . µ V = 27 . ± . P A =57 ° and an ellipticity e = 0 .
30 (Mu˜noz et al. 2018).Battaglia et al. (2011) report the systemic velocity inthe heliocentric system of Sextans to be v sys = 226 ± . − .Irwin & Hatzdimitriou (1995) computed a core radius r c = 16 . ± . r t = 160 ±
50 ar-cmin. More recently, Roderick et al. (2016) fitted a Kingmodel to the radial distribution of Sextans computing acore radius r c = 26 . ± . r t = 83 . ± . . ± .
18 kpc) Mu˜noz et al.(2018) reported a core radius r c = 20 . ± .
05 arcminand a tidal radius r t = 60 . ± . − .
93 (Battaglia et al. 2011). Lokas (2009) computed ahigh mass-to-light-ratio of
M/L ≈
97. Battaglia et al.(2011) computed a
M/L ≈ − r ∼
16 arcmin(400 pc) from Sextans’ center, corresponding to the coreradius reported by Irwin & Hatzdimitriou (1995).Later, Battaglia et al. (2011) reported the detection ofa cold substructure consisting of nine very metal-poorstars close to the center of Sextans.These stars have verysimilar distances, kinematics, and metallicities. The av-erage metallicity of this 9-star group is [Fe/H]= − . .
15 dex 1 σ -scatter. This group of stars was takenfrom the most metal-poor stars, which have a low veloc-ity dispersion ( ∼ . − ) and whose average radialvelocity is 72 . ± . − . Battaglia et al. (2011) as-sume that the ratio of the stars in the substructure isrepresentative with respect to their total number of Sex-ld stellar substructures in MW’s dSphs 3 TABLE 1Parameters of Sextans, Carina, Leo I and Leo II dSphs.
Parameter Sextans Carina Leo I Leo II Ref. a ( α J , δ J ) 10 h m s , -01 ° ′ ′′ h m s , -50 ° ′ ′′ h m s , 12 ° ′ ′′ h m s , 22 ° ′ ′′ ± ± +16 − kpc 233 ±
15 kpc 2,2,2,2M V − . ± . − . ± . e core ± .
05 arcmin 7.97 ± .
16 arcmin 3.6 ± . ± . tidal ± . ± .
98 arcmin 13.5 ± . ± . sys ± . − − − − 〈 [Fe/H] 〉 -1.93 ± . ± . ± . ± . a References: (1) Mateo (1998), (2) McConnachie (2012), (3) Bellazzini et al. (2004), (4) Bellazzini et al. (2005) (5) Irwin & Hatzdimitriou(1995), (6) Battaglia et al. (2011), (7) Koch et al. (2006), (8) Koch et al. (2007a), (9) Koch et al. (2007b), Lokas (2009) & (11) Mu˜noz et al.(2018). tans members (174 stars). Thus the substructure wouldaccount for 5% of Sextans stellar population.Finally, Roderick et al. (2016) reported on photomet-ric evidence of stellar substructures associated with Sex-tans. The stellar substructures that they find extend outto a distance of 82 arcmin (2 kpc) from Sextans’ centre.The existence of such stellar substructures in the outerregions of Sextans might indicate that Sextans is under-going tidal disruption. However, Roderick et al. (2016)found that the substructures surrounding Sextans appearto be both aligned with, and perpendicular to Sextans’major axis. The latter suggests that Sextans is not neces-sarily undergoing a strong tidal disruption. Mu˜noz et al.(2018) reported a fairly regular morphology in Sextans,with no obvious signs of tidal features.In contrast, recently Cicuendez et al. (2018) reportedsigns of past accretion/merger events in Sextans: a ‘ring-like’ feature. They claim that the kinematically detectedring in Sextans bears a morphological resemblance to thestellar stream in the And II dSph (Amorisco et al. 2014),which probably merged with another dwarf galaxy.
Carina
Carina is located at a heliocentric distance of 105 ± M V = − . ± . µ V = 25 . ± . P A = 60 ° , a core radius r c = 7 . ± .
16 arcmin, atidal radius r t = 58 . ± .
98 arcmin, an ellipticity e = 0 . v sys = 223 . − (Koch et al. 2006).The large velocity dispersion of Carina is often in-terpreted as evidence of a large mass-to-light ratio(Mateo et al. 1993). Lokas (2009) found a mass-to-lightratio for Carina of 66, from where they inferred that Ca-rina’s kinematics is dominated by its DM halo.Spectroscopic observations of Carina indicate that ithas two dominant stellar populations that should be inequilibrium in the same DM halo, and that such a DMhalo has a less cuspy inner density profile than previouslythought (Hayashi et al. 2018).The Carina dSph has experienced a complex starformation history. It is the only dSph to exhibitclearly episodic star formation interrupted by long qui-escent periods (de Boer et al. 2014; Santana et al. 2016;Hurley-Keller et al. 1998; Monelli 2003). Its color-magnitude diagram shows three different stellar popu- lations corresponding to 11, 5, and 0 . > − ∼
12 Gyr and 4 − ∼
200 pc of Carina’s center. Fabrizio et al. (2016) com-pared observations of the radial velocity distribution ofold and intermediate-age stars in the Carina dSph withN-body simulations. They found a good agreement withthe V rot /σ ratio in the central regions of the dwarf. Thelatter indicates that Carina might have been a diskydwarf galaxy that experienced several strong tidal in-teractions with the MW. Leo I
The Leo I dSph is one of the most remote dSphs asso-ciated with the MW (e.g. Grebel et al. 2003). It hasa Galactocentric distance of 254 kpc (Bellazzini et al.2004), a position angle
P A = 78 ° , a core radius r c =3 . ± . r t = 13 . ± . e = 0 .
31 (Mu˜noz et al. 2018), and a systemic Lora et al. (a) (b)(c) (d)
Fig. 1.—
Brute-force method: 2D Histograms of metallicity and radial velocity per elliptical annulus bin. As an example, usingBattaglia et al.’s 2011 data, we show the 2D histograms for four elliptical-radial bins for the Sextans dwarf; (a) [0-10] arcmin, (b) (10-20] arcmin, (c) (20,30] arcmin, and (d) (30,40] arcmin. The typical size of the uncertainty in metallicity is − . . − . velocity v sys = 284 . − (Koch et al. 2006).Because of Leo I’s high radial velocity and itslarge Galactocentric distance, Koch et al. (2007a) ar-gue that Leo I might be an isolated system, whichis currently not affected by Galactic tides. However,Boylan-Kolchin et al. (2013) argue that it is very unlikelythat Leo I is not bound to the MW galaxy, under thepremise that Galactic satellites are associated with DMsubhalos. They used high resolution numerical simula-tions of a MW like DM halo and found that 99.9% of thesubhalos in the simulations are bound to their host halo.On the other hand, if Leo I passed very close to theGalactic center (around 1 Gyr ago), then the observedkinematics and population segregation in Leo I, alongwith its distorted structural parameters, its age, its lastprominent burst of star formation, and its large radial ve-locity relative to the Galactic Center, can be explained(Mateo et al. 2008). Because Leo I presents no tidalarms, the latter scenario could result from the interac-tion with a third body, which placed Leo I into its presenthighly-elliptical orbit.Sohn et al. (2013) studied the detailed orbital historyof Leo I. They found that Leo I entered the MW virialradius 2 . D GC = 91 kpc) around one Gyr ago.However, Koch et al. (2012b) measured the proper mo-tion of Leo I and found that Leo I might not be boundto the MW. Furthermore, they say that it is likely that Leo I was formed and evolved in isolation, and it is nowapproaching its first encounter with the Galactic halo.Gaia Collaboration, Helmi, et al. (2018) estimatedthat Leo I has a period greater than 5 Gyr, and theypredict that its last pericentric passage (at a distanceof ∼
100 kpc) took place around 1 Gyr ago. The Leo IdSph has been used to derive a limit on the mass of theMW DM halo. If Leo I is indeed bound to the MW, itsets constrains on the MW mass (Boylan-Kolchin et al.2013). Gaia Collaboration, Helmi, et al. (2018) seta lower limit on the mass of the MW DM halo of9 . +6 . − . × M ⊙ , based on the assumption that Leo Iis indeed bound to the MW. Recentely, Fritz et al.(2018) found that Leo I is “back-splashing” if oneconsiders the (preferred) heavy MW DM halo model( M MW = 1 . × M ⊙ ).Interestingly, stellar substructure has been detected inLeo I. Mateo et al. (2008) reported six stars uniformlydistributed and kinematically distinct from the mainLeo I stellar component. They claim that these starsmight represent a tidal feature, but they warn that thestatistics are too poor and that further members of thiskinematic structure would need to be identified in orderto conclude whether the substructure is real or not. Leo II
The Leo II dSph is located at a Galactocentric distanceof 233 kpc (Bellazzini et al. 2005). It has a position an-gle
P A = 12 ° , a core radius r c = 2 . ± . r t = 9 . ± . e = 0 . v sys = 223 . − (Koch et al. 2006).Leo II is believed not to be experiencing strong Galac-tic tides, and its proper motion indicates that it mightnot even be a bound satellite to the MW (L´epine et al.2011). Piatek et al. (2016) measured the proper mo-tion of Leo II. They found that the motions they mea-sured support the idea that Leo II fell into the MWDM halo as a part of a group. On the other hand,Gaia Collaboration, Helmi, et al. (2018) concluded thatthe infall of Leo II as a group is unlikely.Vogt et al. (1995) calculated a M/L ≈
7, which indi-cates that Leo II must be embedded in a DM halo, butit is not an extreme case. Koch et al. (2007b) obtaineda large data set of radial velocity measurements out toLeo II’s limiting radius. They found (depending on thetotal luminosity adopted) a
M/L = 25 −
50. They con-cluded that this
M/L ratio together with the flatnessof its dispersion profile indicate that Leo II is a DM-dominated system.Komiyama et al. (2007) reported the existence of asmall stellar substructure beyond Leo II’s tidal radius.The substructure’s luminosity compares to that of a glob-ular cluster ( M V < − . METHODS “Brute-force” method
We search for stellar substructures in the four selecteddSph galaxies. For this purpose, we first constructed2-dimensional histograms on a grid of metallicity vs. ve-locity (see Figure 1) for every elliptical annuli of constantellipticity and position PA in the selected galaxy.We adopt an initial metallicity met i (the minimummetallicity value of the studied data) and a final metallic-ity met f (the maximum metallicity in the data), with aninterval in metallicity ∆ met and an interval in velocityof ∆ v , see Figure 1.Subsequently, with both metallicity ( met ) and velocity( v ) fixed, we count the stars in our data, that satisfybeing in the interval met < met ∗ met + ∆ met , (1)where met ∗ is the metallicity of the star that is beinganalyzed. We repeat the procedure with the velocity,such that v < v ∗ v + ∆ v , (2)where v ∗ corresponds to the velocity of the star thatis being analyzed. With the resulting number of starsthat satisfy the conditions 1 and 2 (per elliptical annu-lus width), we build histograms of the number of stars tofind peaks in the counts of stars that could be interpretedas stellar substructures.The peaks in the counts of stars that are interestingto us are the ones with low metallicities, and that aredynamically cold, meaning that the velocity dispersionof such group of stars should be significantly lower thatthat of the complete stellar sample (per elliptical annulusbin).In order to declare whether a peak in counts is signifi-cant or not, we realize Monte Carlo tests of normal dis- tributions for both metallicity and velocity (in the sameelliptical annulus bin where the peak was found) and findthe average and standard deviation ( σ ) of these distribu-tions. If the number of counts in the peak (which is foundin a defined metallicity-velocity cell) is ≥ . σ , or inother words, the probability of finding a peak > . σ is ≤
5% we take such a count peak to be significant.
Minimum spanning tree method
Another method to investigate whether stars of a cer-tain metallicity range are in some way kinematicallyclumped (i.e., spatially more concentrated) we make useof a minimum spanning tree (MST, e.g., Schmeja 2011).The MST is the unique set of straight lines (“edges”) con-necting a given set of points (“vortices”) without closedloops, such that the sum of the edge lengths is mini-mum. This construct from graph theory has been widelyapplied in astronomy to cluster and structure analysis,from the large-scale distribution of galaxies to the inter-nal structure of star clusters (Schmeja 2011, and refer-ences therein). We apply an approach similar to the oneintroduced by Allison et al. (2009) to identify and quan-tify mass segregation in star clusters. We construct theMST for the stars of a given metallicity range and deter-mine the mean edge length γ mp for those stars. We usethe geometric mean rather than the arithmetic mean inorder to minimize the influence of outliers (Olczak et al.2011). Then we construct the MST of the same numberof randomly selected stars from the entire sample anddetermine the mean edge length γ rand . This is done 200times in order to obtain the mean value h γ rand i . Theratio R = h γ rand i γ mp , (3)is a measure for the concentration of the stars of the sub-sample relative to the entire stellar population. A valueof R ≈ R ≫ RESULTS
The case of Sextans
For the case of Sextans, we used Battaglia et al.’s(2011) data. They obtained VLT/FLAMESintermediate-resolution spectroscopic observationsin the near-infrared CaII triplet (CaT) region for 1036distinct targets along the line-of-sight to Sextans. Themagnitudes and colors of those targets are broadlyconsistent with red giant branch (RGB) stars. Fromthat sample they obtained 789 stars with S/N and errorin velocity that produce reliable line-of-sight velocitiesand CaT equivalent widths. A subset of 174 starsfrom those are RGB stars that are probable memberswith line-of-sight velocities accurate to ± ± . σ kinematic cut.Finally, in order to refine their membership criteria, theyused the Mg I line at 8806 . (a) (b) Fig. 2.— (a) The Sextans dSph members reported by Battaglia et al. (2011) are shown as filled-black circles. The white small circle (inthe center) represents the center of Sextans reported by Mateo (1998). The stellar substructure previously found by Battaglia et al. (2011)(called sxt1 in the text), is shown with filled blue squares. The new stellar substructure, reported in this work ( sxt2 ) is shown with filledlight blue squares (and two extra violet squares, see text in Section 2.1). The solid black stars, in both substructures ( sxt1 and sxt2 )correspond to the centroid of each stellar substructure. The solid ellipse correspond to the core radius reported by Irwin & Hatzdimitriou(1995). The dotted ellipse is placed at the center of sxt2 substructure, which corresponds to 43 . . . sxt2 is shown as black filled squares. Our first target of study is the stellar clump found inSextans by Battaglia et al. (2011). We will refer to thisclump of stars as sxt1 . With our brute-force method(see Section 3.1), we should be able to reproduce previ-ous findings. Using the Battaglia et al. (2011) data, wefollow the method of Section 3.1. We set met i = − met f = − .
28 dex and ∆ met = 0 . v i = −
20 km s − , v f = 20 km s − , and a∆ v = 10 km s − .Then we construct histograms of star counts per el-liptical radius. These histograms help us to identify theaccumulation of stars in a particular radius bin. The el-liptical radial bins in the case of Sextans have a size of10 arcmin. As expected, we find sxt1 .The metallicity of this group of stars ( sxt1 ) is in therange − . ≤ met ≤ − .
43 dex, with an average metal-licity [Fe/H]= − .
64, and the velocity is in the range − ≤ v ≤ −
12 km s − . The fact that we are able torecover the sxt1 clump is an indicator that our methodis working well.In Figure 2a we plot Sextans’ member stars (black cir-cles) taken from the Battaglia et al. (2011) data set. Thecenter of the Sextans dSph (Mateo 1998) is shown as awhite circle. The sxt1 clump stars are shown as filledblue squares, and the solid black star shows the centroidof sxt1 .We did not only recover the sxt1 clump reported byBattaglia et al. (2011), but we also found a second oldcold-stellar substructure. We will refer to this new sub-structure as sxt2 . The sxt2 stellar substructure consistsof eight stars with metallicities ranging from − .
99 to ≤ − .
63. The average metallicity is [Fe/H]= -2.78, and the velocities range from 3 . ≤ v ≤ .
08 km s − . Wecomputed the centroid (relative to the center of Sextans,see Table 2) of sxt2 to be located at (4 . − . sxt2 ) as filled cyan squares, and the respective centroidplotted as a solid black star. It has to be noted that if onerelaxes the velocity constraint and lets the velocity cover3 . ≤ v .
16 km s − , we can add two new metal-poorstars. Therefore, we end with a 10-star substructure withan average metallicity of − .
76. If we include these twostars, the centroid of the 10-star substructure changes to(4 . − . sxt1 and sxt2 andconfirms them as highly concentrated clumps with R =3 . ± .
20 and R = 3 . ± .
31, respectively (see Equa-tion 3).In order to demonstrate that random groupings of starswith matching metallicities and velocities are in fact veryrare, we realize Monte Carlo tests of normal distributionsfor both metallicity and velocity (see Section 3.1). Wefind that the probability of finding sxt1 is ∼
1% (i.e., ∼ . σ ). We also find that the probability of finding sxt2 is ∼
5% (i.e., 2 . σ ).The latter supports the suggestion that in fact the newclump sxt2 reported in this work is likely physical.Interestingly, Roderick et al. (2016) reported clear ev-idence of a stellar substructure distributed evenly aboutthe center of Sextans. The substructure extends upto a distance of 2 kpc. In order to compare our find-ings to Roderick et al. (2016), we built contours of starcounts using Roderick et al.’s (2016) substructure data,ld stellar substructures in MW’s dSphs 7 (a) (b) Fig. 3.— (a) In this panel we show the Carina dSph members reported by Koch et al. (2008) as filled black circles. The center of Carinais represented with a filled red circle. The new stellar substructure found in this work ( car1 ) is shown with filled yellow squares.The whitestar corresponds to the centroid of the stellar substructure (relative to Carina’s center, see Table 2). The two solid black ellipses show thecore radius at 8 . . car1 are overplotted as filled yellow squares.The black solid ellipse is located at the distance of the core radius (8 . shown in Figure 2b. We over-plotted their core (whitedashed-line) and tidal radius (black line) (26 . . sxt2 substructure lies in a densesubstructure region, giving a hint that sxt2 might be amember of the annular substructure within Sextans.Battaglia et al. (2011) transformed the heliocentricline-of-sight velocities into line-of-sight velocities in aframe at rest with respect to the Galactic center, v GSR (where GSR stands for Galactic standard of rest, v sys,GSR = 78 . ± . − ). They reported a velocitydispersion for the six innermost stars making up theirclump of 1 . ± . − , and an average GSR velocityof 72 . ± . − . They also computed an averagemetallicity for their clump of [Fe/H]= -2.6.We follow Battaglia et al. (2011) and compute a veloc-ity dispersion and a mean GSR velocity for eight stars of sxt2 (see the cyan squares in Figure 2a) of 2 . − and 87 . − , respectively.Since the velocity dispersion could be a strong functionof the galactocentric radius, we compute the velocity dis-persion of all the stars in Battaglia et al.’s (2011) databelonging to the same elliptical annulus as sxt2 . Wefound that the velocity dispersion of the elliptical annu-lus associated with sxt2 is 9 km s − .The velocity dispersion of sxt2 is very cold (comparedto the velocity dispersion of all the stars belonging tothe same elliptical annulus) and its average metallicity([Fe/H]= -2.78) is even lower than that of sxt1 . The lowmetallicity of sxt2 , together with its cold kinematics,suggests that the stars belonged to a stellar (globular)cluster. If we add the two extra stars shown as magentasquares in Figure 2a (where we have relaxed the velocitycondition, see text above), we obtain a velocity dispersionof 4 . − and an average metallicity of [Fe/H]=- 2.76. The velocity dispersion of the 10-star clump is abit higher than the one computed with eight stars, butit remains colder than the velocity dispersion of sxt2 ’sassociated elliptical annulus (9 km s − ).As done by Battaglia et al. (2011), if the 174 stars inthe Sextans sample are representative, then sxt2 wouldaccount for 4 .
6% (or ∼
6% if one includes the extratwo stars) of Sextans’ stellar population. If we take thevalue for the luminosity of Sextans of 4 . × L ⊙ (Irwin & Hatzdimitriou 1995; Lokas 2009) and a typi-cal value for the mass-to-light ratio for globular clusters, M/L ∼ sxt2 are4 × M ⊙ and 2 × L ⊙ (5 × M ⊙ and 2 . × L ⊙ ,if we include the two extra stars).The fact that the MST analysis confirmed that sxt2 is concentrated with respect to the overall Sextans sam-ple, along with its metal-poor, dynamically cold nature,reasserts the idea that Sextans has not experienced astrong tidal encounter, indicating that it could have amore circular orbit around the Galactic Center. The lowmetallicity of sxt2 suggests that it might be very old(i.e., >
10 Gyr) and its progenitor (as well as the progen-itor of sxt1 , Battaglia et al. 2011) would be most likelya globular cluster, and such a globular cluster would thenbe among the most metal-poor globular clusters known.The detection of cold old stellar substructures indwarf spheroidal galaxies has been helpful to shed somelight on the core/cusp DM problem at galactic scales.For example, a cored DM halo profile is preferred whenexplaining the existence of the stellar substructures inUMi and Sextans (Kleyna et al. 2003; Lora et al. 2009,2013). The stellar substructure sxt2 gives a new test tocorroborate if (in particular) Sextans is embedded in acored DM halo. Lora et al.
The case of Carina
For the case of Carina, we used Koch et al.’s (2006)data. Their targets in Carina were chosen from pho-tometry and astrometry obtained by the ESO ImagingSurvey (Nonino et al. 1999). Koch et al. (2006) selectedtheir targets to cover magnitudes from the tip of the RGBdown to 3 mag below the RGB tip (20 . −
300 km s − range), andthen rejecting 3 σ outliers. Koch et al. (2006) determinedthe metallicities of the RGB stars sample through theequivalent widths-CaT method.We apply the same methods described in Section 3 tothe Carina dSph. We set the metallicity limits met i = − met f = − . met = 0 . v i = −
20 km s − and v f = 20 km s − with a ∆ v = 10 km s − . We find a group of nine starswith metallicities in the range − . ≤ met ≤ − . . ≤ v ≤ .
31 km s − , within ellipticalannuli widths of 5 arcmin. We refer to this group ofstars as car1 . We plot car1 in Figure 3a as filled yellowsquares. car1 ’s centroid is shown with a white star. Wenotice that the centroid of car1 ( − . , .
153 arcmin)is quite close to the center of Carina (just ∼
14 pc away),represented with a filled red circle.Applying the MST analysis, car1 is found as a promi-nent clump with a ratio R = 3 . ± .
30 (see Equation 3).We demonstrate that random groupings of stars withmatching metallicities and velocities in Carina’s dataare very rare. We realize Monte Carlo tests of normaldistributions for both metallicity and velocity (see Sec-tion 3.1). We find that the probability of finding car1 is ∼ .
5% (i.e., 2 . σ ) supporting the suggestion that thesubstructure CAR1 is physical. The stellar substructurefound in Carina could be an unbound object in the throesof destruction, which could have started out as a typicalcluster with a radius of ∼ − sxt1 .For the nine stars group ( car1 ) we computed avelocity dispersion of 4 km s − , a mean velocity of230 . − , and an average metallicity of [Fe/H] ∼− . car1 . We found that the velocity dispersion ofthe elliptical-annuli associated to CAR1 is 7.2 km s − .If the 437 star-sample is representative of Carina’ sstellar component, then car1 would account for 2%of Carina’s stellar population. If we adopt a mass-to-luminosity-ratio of 2 (typical for globular clusters) and aluminosity of 3 . × ( Lokas 2009), then we can roughlyestimate that car1 ’s mass is ∼ . × , very similar to the mass estimate of sxt1 and sxt2 .As we mentioned before, Fabrizio et al. (2011) foundevidence of a secondary maximum in radial velocity at adistance ∼
200 pc from Carina’s center, which they in-terpreted as reminiscent of a substructure with transitionproperties (i.e., a transition between a bulge-like and/ordisk-like structure). In Figure 3b we took the sky distri-bution of the stars in Fabrizio et al.’s (2011) photometriccatalog (their Figure 13), where they plot isodensity con-tour levels from 5% to 95%. We over plotted the car1 clump of stars. All the (nine) stars lie inside the 45%isodensity contour level, which coincides with the coreradius of Carina (see the black ellipse in Figure 3b), andit also coincides with the location where Fabrizio et al.(2011) reported their substructure.
The case of Leo I
For the case of Leo I, we used Koch et al.’s (2007a)(hereafter K07) data together with Bosler et al.’s (2007)(hereafter B07) data.The targets in Koch et al. (2007a) were selected fromphotometry obtained with the framework of the Cam-bridge Astronomical Survey Unit (Irwin & Lewis 2001)at the 2 . I ∼
18 mag) down to 1 . I . . ∼
5, sufficient to obtain accurate ra-dial velocity measurements. They derived radial veloci-ties from their final reduced spectra by cross-correlatingthe three strong Ca lines (8498, 8542, and 8662) with asynthetic template spectrum of the CaT region. Theyestimated the metallicities of their RGB stars sample,from the widely used method of relating the equivalentwidths of the CaT to metallicity.On the other hand, Bosler et al. (2007) obtained low-dispersion spectra of red giants in Leo I (and Leo II)using the Keck I 10-m telescope and LRIS. They ob-tained a mean S/N of 18 (for Leo I stars). They verifiedthe membership of their RGB stars by deriving their ra-dial velocities. From their 121 stars observed in Leo I,a number of 90 have heliocentric velocities within 3 σ ofthe average velocity, and thus were selected as membersof Leo I. They estimated the metallicities of their samplerelating the equivalent widths of the CaT (as done byKoch et al. 2007a).It has to be noted that the K07 and B07 data sets donot overlap.We took both data sets, and applied the brute forcemethod (Section 3.1) to scrutinize the possible existenceof stellar substructure. We set the ranges for the metal-licity met i = − . met f = − . met =0 . v i = −
20 km s − and v f = 20 km s − with ∆ v = 10 km s − .We only found one group of stars that was relevant,using elliptical annuli widths of 4 arcmin. This group ofstars consists of six stars (hereafter, li1 ). The metallicityof li1 is in the range − . ≤ met ≤ − .
52, with anaverage metallicity [Fe/H]= − .
63. The velocities are inthe range 288 ≤ v ≤ . − , the average velocityld stellar substructures in MW’s dSphs 9 (a) (b) Fig. 4.— (a) Leo I dSph members reported by Koch et al. (2007a) are shown as black open circles. The red giant stars reported byBosler et al. (2007) are represented with black open triangles. The white plus symbols represent Mateo et al. (2008)’s data. Overplottedare the group of stars reported in this work li1 (-filled-magenta squares); the group of stars reported by Koch et al. (2007a) (filled greensquares), and the group of stars reported by Mateo et al. (2008) (filled blue squares). The inner solid black ellipse corresponds to the coreradius and the outer solid black ellipse corresponds to the tidal radius (Mu˜noz et al. 2018).(b) Heliocentric velocities as a function of the elliptical radius. The dotted vertical lines show the core (3 . ′ ) and tidal (12 . ′ ) radius. Thehorizontal dotted line shows v Helio ≈
96 km s − , the average velocity of Mateo et al. (2008)’s sample. is 286 km s − , and the velocity dispersion is 1 . − .We computed the velocity dispersion of all the stars inthe K07 and B07 data belonging to the same ellipticalannulus as LI1 . We found that the velocity dispersion ofthe elliptical annulus associated with
LI1 is 13 km s − .We show the stars belonging to li1 as filled magentasquares in Figure 4. The open black circles are the redgiants reported by K07 and the open black triangles cor-respond to B07’s stars.The MST method only finds few rather weak concen-trations with R . . li1 is a realsubstructure.K07 detected a minor significant rise in the radial dis-persion profile, around 220 pc from the center of Leo I(corresponding to the core radius), which they not findto be associated with any real localized kinematical sub-structure (we will refer to this group of stars as k1 ). Weplot this group of five stars as filled green squares in Fig-ure 4. We observe that the clump reported in this workand k1 only have one star in common. Adding to K07’sand B07’s data, additional data of Leo I star members,Mateo et al. (2008) (hereafter M08) presented kinematicresults of stars in the Leo I dSph (see white plus sym-bols in Figure 4). They found a group of six stars (seefilled blue squares in Figure 4) with velocities in a nar-row range, with a dispersion of ∼ − and a meanvelocity of ∼
96 km s − (see horizontal dotted line inFigure 4b). They argue that this group of stars mightbe a kinematically cold group, but they warn that thestatistics are poor. They found this group of stars byplotting the heliocentric velocity as a function of the el-liptical radius (see our Figure 4b and Figure 9 of M08).For comparison, we over plotted the clump of stars re-ported in this work, li1 (magenta filled squares); and the group of five stars reported by K07 (filled green squares).We perform a statistical study to show whether ran-dom groupings of stars in K07 and B07 data are rare ornot. We realize Monte Carlo tests of normal distribu-tions for both metallicity and velocity (see Section 3.1).We find that the probability of finding li1 is ∼
75% (i.e.,0 . σ ). Such a finding reflects the fact that, in the caseof Leo I, we are dealing with low number statistics, andthus li1 is not statistically significant.The velocity dispersion of li1 is a factor of ∼ σ v,LI = 1 . − ) than the dispersion found inM08’s group of stars. This, together with the low av-erage metallicity found in li1 , indicates that li1 (if real)might be a dynamically cold stellar substructure.Five of the six stars in li1 are contained in the B08data. Then, if we suppose that the B07’s data are repre-sentative of the Leo I main stellar population, li1 wouldaccount for ∼
5% of it. Depending on the luminosityadopted (McConnachie 2012), a zero order mass estimatefor li1 would be M LI ≈ (3 . − × M ⊙ . The calcu-lated mass for li1 is higher than for the stellar substruc-tures in Sextans and Carina (see Section 4.1 and 4.2, re-spectively), but it is comparable to the mass of the mostmassive globular cluster in the Fornax dSph and Sagit-tarius dSph, ∼ (2 − . × M ⊙ (Mackey & Gilmore2003). The case of Leo II
For the case of Leo II, we used Koch et al.’s (2007b)(hereafter K207) data, together with B07 data.The targets in Koch et al. (2007b) were selected fromphotometry that was obtained by the Cambridge Astro-nomical Survey Unit (Irwin & Lewis 2001) at the 2 . Fig. 5.—
We show the sky position of K207 and B07’s Leo II data as open circles, and S17 data as plus symbols. The two substructuresfound in the S17 data are shown as filled cyan squares, and filled purple squares. The (cyan/purple) stars show the position of the centroidof each of the substructures. The two sold ellipses show the core (2.9’) and tidal (8.7’) radius. The black star shows the position of thecenter of the substructure reported by Komiyama et al. (2007), the white region shows the extent of the substructure (4 × . ). ing members of Leo II’s RGB stars. Koch et al. (2007b)obtained VLT/FLAMES spectroscopic observations inthe near-infrared CaT region, using the GIRAFFE mul-tifiber spectrograph in low-resolution mode, centered atthe near-infrared CaT at 8550 ˚A. To eliminate Galac-tic contaminants, Koch et al. (2007b) determined indi-vidual radial velocities by means of cross-correlation ofthe three CaT lines against synthetic Gaussian templatespectra. The templates were synthesized adopting rep-resentative equivalent widths of the CaT in RGB stars.The spectroscopic metallicity and age distributions werederived using the near-infrared CaT calibration method(Koch et al. 2007c).On the other hand, B07 obtained CaT abundances andradial velocities for 74 RGB stars in Leo II, using thelow-resolution spectrograph LRIS. They obtained a meanS/N of 23. They verified the membership of their RGBstars by deriving their radial velocities. From their 90stars observed in Leo II, 83 have heliocentric velocitieswithin 3 σ of the average velocity, and thus are designatedas members of Leo II. B07 estimated the metallicities oftheir sample relating the equivalent widths of the CaT.It has to be noted that K207 and B07 data sets do notoverlap.In addition, Spencer et al. (2017a) (hereafter S17) ob-tained a large data set of RGB member candidates ofLeo II, which were separately analyzed. S17 performedspectroscopic observations with the Multiple Mirror Tele-scope using Hectochelle, a multifiber, single-order echellespectrograph. They obtained simultaneous estimatesof radial velocity, effective temperature, surface gravity,and metallicity by fitting a library of smoothed, syn-thetic stellar spectra to each Hectochelle spectrum inpixel space (Walker et al. 2015). In order to separatestellar members from nonmembers they employed a ve-locity cut. Again, stars with radial velocities larger than 3 σ were taken to be Galactic foreground stars. After per-forming the kinematic cut, they still had 11 stars withvelocities and positions similar to those of Leo II stars.Therefore, they applied an extra cut in the data basedon stellar surface gravities. From both criteria, they endwith a total of 175 Leo II members.We look for stellar substructures in the Leo II dSph,combining the RGB star data of K207 and B07. Weapplied the brute-force method setting the ranges forthe metallicity met i = − . met f = − . dexand ∆ met = 0 . v i = −
20 km s − and v f = 20 km s − , with ∆ v = 4 km s − .The typical measurement errors for the metallicity andthe velocity (as in Leo I) are 0 .
11 dex and 5 . − ,respectively.Using the brute-force algorithm and the MST method,we did not find any significant stellar substructure inthe K207 + B07 data set. The non-detection of stellarsubstructures in these data might be related to the smallsample size (128 with K207 and B07 combined), and/orto the large Galactocentric distance of Leo II.On the other hand, analyzing the S17 data with the brute-force algorithm we found two considerable groupsof stars. In Figure 5 we show K207 combined with B07data as open circles; and K17 data as plus symbols. Thetwo groups of stars are shown as filled cyan squares (here-after lii1 ), and filled purple squares (hereafter lii2 ).The substructure lii1 is the most metal-poor groupthat we find. Five stars make up lii1 . It has a meanmetallicity of − .
35, a mean velocity of 72 . − , anda velocity dispersion of 1 . − , in a elliptical annuluswidth of 5 arcmin. The substructure lii2 is constitutedof seven stars. The mean metallicity is −
2, the meanvelocity is 78 . − , and it has a velocity dispersionof 1 . − , in a elliptical annuli width of 6 arcmin.We compute the velocity dispersion of all the stars inld stellar substructures in MW’s dSphs 11the S17 data belonging to the same annulus as LII1 and
LII2 . We found that the velocity dispersion of the ellip-tical annulus associated with
LII1 is 7 km s − , and thevelocity dispersion associated with LII2 is 7.6 km s − .In Figure 6a we show the systemic velocity as a func-tion of the elliptical radius, where we can clearly observethat the substructures appear elongated (about 4 to 6arcmin). The average velocity of lii1 is very similar tothe systemic velocity of Leo II ( δ = 0 . − ). Fig-ure 6b shows the metallicity as a function of the ellipticalradius. 25% of the stars in the S17 data have metallici-ties lower than ∼ −
2. Moreover, only 8.6% of the starsin the S17 sample have metallicities lower than ∼ − . lii1 .The velocity as a function of metallicity is shown inFigure 6c. In the velocity-metallicity space, one canclearly observe both substructures clumped together.The mean velocity of lii1 is 1 σ away from Leo II’s aver-age velocity (shown as a shaded region in Figure 6c). Thelow metallicity of lii1 together with its velocity makesit a thought-provoking stellar substructure.It is worthwhile mentioning that Komiyama et al.(2007) carried out wide-field V, I imaging of Leo II ex-tending far beyond Leo II’s tidal radius. They reportedthe existence of a substructure in the eastern part of thegalaxy (containing four bright RGB stars) with a physi-cal size of 270 ×
170 pc located beyond the tidal radius(8 . ′ ), with a luminosity close to that of a globular clus-ter (see black star in Figure 5). They suggest that thissubstructure is a disrupted globular cluster that is merg-ing with the main stellar component of Leo II.Interestingly, two of the stars in lii1 are located be-yond the tidal radius, and the interpretation could besimilar to that of the knotty stellar structure reportedby Komiyama et al. (2007). We computed a first orderapproximation on the stellar mass of lii1 making the as-sumption that the number of RGB stars in the S17 datais representative of Leo II. Then, lii1 would account for ∼
3% of Leo II’s total stellar mass. Taking a luminosityof L V = 7 . × L ⊙ (Coleman et al. 2007) and a typi-cal value (for globular clusters) of the mass-to-light ratioequal to two, then lii1 ’s mass would be ∼ . × M ⊙ .We perform a statistical study to investigate whetherrandom groupings of stars in the S17 data are rare. Wecarried out Monte Carlo tests of normal distributions forboth metallicity and velocity (see Section 3.1). We findthat the probability of finding LII1 is ∼
5% (i.e., 2 . σ ).On the other hand, the probability of finding LII2 isonly 1 . σ . Therefore, we conclude that LII2 is not sta-tistically significant. Indeed,
LII1 could be real stellardebris, but we have to keep in mind that we are dealingwith a low number of Leo II members.One must keep in mind that a high fraction of binarystars could falsify radial velocity measurements. Re-cently, Spencer et al. (2017b) determined a binary frac-tion of 0 . − .
34 for the Leo II dSph galaxy. CONCLUSIONS
In this paper we searched for stellar substructures infour dSph galaxies that are satellites of the MW. Wewere able to find the stellar substructure reported byBattaglia et al. (2011), and a new substructure in Sex-tans, sxt2 . The latter stellar substructure consists ofeight stars with metallicities from − .
99 to − .
63 dex. Moreover, if we relax the constraint on the velocity(3 . −
16 km s − ) we can add two more stars to theclump maintaining the same low metallicity range. Thedistance from the center of Sextans to the center of theten-star clump is ∼
751 pc, and the velocity dispersionis cold ( σ ≃ .
01 km s − ). If the stars of the sxt2 clump belong to a disrupted globular cluster then theselow metallicities would suggest that it would be one ofthe most metal-poor globular clusters known. It is veryencouraging to see that sxt2 lies in the densest regionreported by Roderick et al. (2016).We also find a cold stellar substructure close to thecore (240 pc) of the Carina dSph. This substructure car1 , consists of nine stars with metallicities rangingfrom − .
89 to − .
44 dex and a velocity dispersion of σ ≃ .
88 km s − . Such a substructure resembles a dis-rupted globular cluster, very similar to that found byBattaglia et al. (2011) in Sextans ( sxt1 ). It has to benoted that the distance from the center of Carina to thecenter of Carina’s substructure is only ∼ . car1 substructure could possibly be related to the sub-structure reported by Fabrizio et al. (2011), since bothlie close to the core of Carina.Analyzing Leo I, we find a new substructure, besidesthe one reported by K07 and M08. The li1 substructurehas six stars with metallicities ranging from − .
78 to − .
52 dex, an average velocity of 286 km s − , and a ve-locity dispersion of σ ≃ . − . After the statisticalanalysis, we concluded that li1 is not significant.In the dSph galaxy Leo II, we found two significantgroups of stars, lii1 and lii2 . From those two groups,we found that only lii1 is statistically significant. Evenif the probability of finding lii1 is very low (0 . N -body simulations,one can study the survival of old cold substructures inthe DM halo of dSphs against phase mixing, and com-pare the evolution of the stellar substructures when thedark halo has a core and when the dark halo has a cuspyprofile (Kleyna et al. 2003; Lora et al. 2009, 2012, 2013;Contenta et al. 2017; Amorisco 2017). The existence ofthe old stellar clump in UMi (Kleyna et al. 1998) is inagreement with a cored DM profile rather than a cuspyNFW one. If the old stellar clump in UMi is dropped ina NFW cuspy profile, the stellar clump gets disrupted inthe first Gyr (Kleyna et al. 2003; Lora et al. 2009, 2012,2013).Lora et al. (2013) also found that a cored DM profileis needed in order to guarantee the survival of the oldstellar substructure found in Sextans by Battaglia et al.(2011), and the one found by Walker et al. (2006). Then,the new stellar substructures found in this work will becrucial to further investigate the core/cusp problem inother dSphs. Therefore, we will perform N-body simula-tions of the four dSph studied in this paper and the newstellar substructures to investigate their evolution withintheir parent DM halo.2 Lora et al. Fig. 6.—
Combined Leo II data of K07 and B07 as black plus symbols. The data of S17 shown as open black circles. The two substructuresfound are plotted as filled cyan squares ( lii1 ), and filled purple squares ( lii2 ). In panel (a) we show the radial velocity as a function ofthe elliptical radius. In panel (b) we show the metallicity as a function of the elliptical radius. The vertical dashed lines in panels (a)and (b) indicate the core and tidal radius of Leo II. In panel (c), we show the radial velocity as a function of the metallicity. The twovertical/horizontal solid lines represent the mean metallicity/velocity for each of the substructures. The dashed vertical and horizontallines indicate the mean value of the metallicity and velocity of S17 sample. The shaded zone indicates the 1 σ velocity region. Given the very small number of stars found to be as-sociated with the substructures in the current study andin others, and given how difficult it is to destroy GCs inthe absence of strong tidal fields, one might also suggestthat we are looking here at the possible remnants of veryold, metal-poor, low-mass clusters akin to open clusters.Either way, another interesting question is how such clus-ters were able to form to begin with in systems that wesee nowadays as very low stellar density objects. More-over, one could argue that these discoveries may add tothe recent discoveries of old clusters in low-mass dSphs,making them a more common occurrence than previouslythought.It is worthwhile to mention that Amorisco et al. (2014)point out that mergers of low-mass galaxies are expectedwithin the hierarchical model of galaxy formation. More-over, they report the kinematic detection of a stellarstream in the dSph Andromeda II, which they suggestcould be the remnant of a merger between two dwarfgalaxies (see also Koch et al. 2012a for low-mass range).Thus, further study on the properties of the stellar sub- structures reported in this work might shed light on theway galaxies assemble through mergers at very smallscales.V.L. gratefully acknowledges support from theCONACyT Research Fellowships program. V.L. thanksGiuseppina Battaglia, Mathew Walker and YutakaKomiyama for making their data available. V.L. thanksAlejandro Raga, Gustavo Bruzual and Sundar Srinivasanfor very helpful comments, suggestions and discussionswhich resulted in an improved version of this paper.S.S. was supported by Sonderforschungsbereich SFB881 “The Milky Way System” (subproject B5, fundingperiod 2011-2014) of the German Research Foundation(DFG) during part of this work.EKG and AK were supported by Sonderforschungs-bereich SFB 881 ”The Milky Way System” (subpro-jects A02 and A08) of the German Research Foundation(DFG).We thank the anonymous referee for very kind anduseful comments that improved the presentation of thispaper.
REFERENCESAmorisco, N. C., Evans, N. W. & van de Venn, G., 2014, Nature,Volume 507, Issue 7492, 335Amorisco, N. C., 2017, ApJ, 844, 64Allison, R. J., Goodwin, S. P., Parker, R. J., et al. 2009, MNRAS,395, 1449Battaglia, G., Tolstoy, E., Helmi, A., Irwin, M., Parisi, P., Hill, V.& Jablonka, P., 2011, MNRAS, 411, 1013Bellazzini, M., Gennari, N., Ferraro, F. R. & Sollima, A., 2004,MNRAS, 354, 708Bellazzini, M., Gennari, N. & Ferraro, F. R., 2005, MNRAS, 360,185Bono, G., Stetson, P. B., Walker, A. R., et al. 2010, PASP, 122,651Bosler, T. L., Smecker-Hane, T. A. & Stetson, 2007, MNRAS,378, 318Boylan-Kolchin, M., Bullock, J. S., Sohn, S. T., et al. 2013, ApJ,768, 140Bullock, J. S. & Johnston, K. V., 2005, ApJ, 635, 931Chen, J., Bryan, G. L. & Salem, M., 2016, MNRAS, 460, 3335Cicuendez, L. et al. 2018, A&A, 609, 53Coleman, M. G., Jordi, K., Rix, H. W., Grebel, E. K. & Koch, A.2007, AJ, 134, 1938 Contenta, F., Gieles, M., Balbinot, E. & Collins, M. L. M., 2017,MNRAS, 466, 1741Cooper, A. P., D’Souza, R., Kauffmann, G., et al. 2013, MNRAS,434, 3348de Boer, T. J. L., Tolstoy, E., Lemasle, B., Saha, A., Olszewski,E. W., Mateo, M., Irwin, M. J. & Battaglia, G. 2014, A&A572, A10Dekel, A. & Silk, J., 1986, ApJ, 303, 39del Pino, A., Aparicio, A. & Hidalgo, S. L., 2015, MNRAS, 454,3996Fabrizio M., Nonino, M., Bono, G., et al. 2011, PASP, 123, 384Fabrizio M. et al. 2016, ApJ, 830, 126Fritz, T. K., Battaglia, G., Pawlowsky, M. S., Kallivayalil, N., vander Marel, R., Sohn, S. T., Brook, C. & Besla, G. 2018, A&A619, A103Gaia Collaboration, Helmi, A. et al. 2018, A&A, 616, A12Grebel, E. K., 2000, in Star Formation from the Small to theLarge Scale, ed. F. Favata, A. A. Kaas, & A. Wilson (ESASP-445; Noordwijk: ESA), 87Grebel, E. K., Gallagher, J. S. & Harbeck, D., 2003, AJ, 125,1924Grebel, E. K. & Gallagher, J. S., 2004, ApJ, 610L, 89Harbeck, D., et al. 2001, AJ, 122, 3092 ld stellar substructures in MW’s dSphs 13ld stellar substructures in MW’s dSphs 13