Mon.Not.R.Astron.Soc.000,1–6(2013) Printed1February2013 (MNLATEXstylefilev2.2) Galaxy Pairs in the Local Group Azadeh Fattahi1⋆, Julio F. Navarro1, Else Starkenburg1, Christopher R. Barber1, 3 and Alan W. McConnachie2 1 0 1Department of Physics and Astronomy, University of Victoria, PO Box 3055 STN CSC, Victoria, BC, V8W 3P6, Canada 2 2NRC Herzberg Institute of Astrophysics, 5071 West Saanich Road, Victoria, BC, V9E 2E7, Canada n a J 0 1February2013 3 ] ABSTRACT O Current models of galaxy formation predict that galaxy pairs of comparable mag- C nitudes should become increasingly rare with decreasing luminosity. This seems at . odds with the relatively high frequency of pairingsamong dwarf galaxiesin the Local h Group. We use literature data to show that ∼ 30% of all satellites of the Milky Way p - and Andromeda galaxies brighter than MV =−8 are found in likely physical pairs of o comparable luminosity. Besides the previously recognised pairings of the Magellanic r Clouds and of NGC 147/NGC 185,other candidate pairs include the Ursa Minor and t s Dracodwarfspheroidals,aswellastheAndI/AndIIIsatellitesofM31.Thesepairsare a much closer than expected by chance if the radial and angular distributions of satel- [ lites were uncorrelated;in addition, they have very similar line-of-sight velocities and 2 luminosities that differ by less than three magnitudes. In contrast, the same criteria v pairfewerthan4%ofsatellitesinN-body/semi-analyticmodelsthatmatchtheradial 1 distribution and luminosity function of LocalGroup satellites. If confirmed in studies 6 oflargersamples,the highfrequencyofdwarfgalaxypairingsmayprovideinteresting 1 clues to the formation of faint galaxies in the current cosmologicalparadigm. 3 1. Key words: galaxies:dwarf-galaxies:formation-galaxies:evolution-LocalGroup 1 2 1 : v 1 INTRODUCTION habithalosofsimilarvirialmass.Inaddition,extrapolating i such models to the faintest galaxies known indicate that X The shapes of the galaxy and dark halo mass functions few, if any, galaxies more massive than a few million solar r differ substantially in the ΛCDM paradigm (see, e.g., massesareexpectedtoforminhaloswithvirialmassbelow a Benson et al. 2003). This is usually interpreted to imply 1010M⊙. that the “efficiency” of galaxy formation, as measured by Recent work has highlighted potential disagreements the ratio between the stellar mass of a galaxy (M ) and gal thevirial1 massofitshosthalo(M200),variesstronglywith between these model predictions and observations, includ- ing the lack of a characteristic velocity at the faint-end of virialmass.Inparticular,Mgal/M200shoulddecreasesteeply blind HI surveys (Zwaan et al. 2010); and the low virial toward low halo masses in order to match the shallow faint endofthegalaxyluminosityfunction(see,e.g.,Moster et al. mass(substantiallybelow1010M⊙)inferredfromdynamical dataforthedwarfspheroidalcompanionsoftheMilkyWay 2010;Behroozi et al. 2010;Guo et al. 2011). (Boylan-Kolchin et al.2012)andfornearbydwarfirregulars On the scale of dwarf galaxies, which we define con- ventionally here as those with Mgal < 109.5M⊙, simple (Ferrero et al. 2012). It could be argued, however, that the evidenceforsubstantialdisagreementisunconvincing,given abundance-matchingmodelssuggestadependencenearlyas steepasM ∝M3 inthedwarfgalaxyregime(Guo et al. that theinferences are eitherindirect (inthecase of theHI gal 200 velocityfunction)orbasedonsmallandheterogeneoussam- 2010). Such steep scaling would imply that dwarfs span- ples(inthecaseofthenearbydwarfs;seeWang et al.2012; ning several decades in stellar mass should nevertheless in- Vera-Ciro et al. 2012). It is therefore important to consider further tests of ⋆ Email:[email protected] themodelpredictions.WeexploreherehowthesteepMgal- 1 Wedefineallvirialquantitiesasthosecorrespondingtoasphere M200 relation predicted for dwarfs affects the frequency of of mean density equal to 200 times the critical density for clo- galaxy pairs of comparable luminosity. Such pairs, when sure. Furthermore, weshall hereafter refer to the stellar mass of closeenoughtoinhabitthesamedarkmatterhalo(referred agalaxyas“galaxymass”,forbrevity. tohereafter as “physical pairs”), are expected to berare at 2 Fattahi et al. allluminosities,butespeciallysointhescaleofdwarfs.This 2 DATASETS isbecausethefaintercompanionsinphysicalpairstracethe We use the recent compilation by McConnachie (2012) as halo substructure, and subhalos are, by and large, far less the source of the positions, distances, line-of-sight veloci- massivethanthemainhalo:themostmassivesubhalotypi- ties, and magnitudes of Local Group dwarfs that we use in callyhasamassonlyonehundredththatofthemainsystem ouranalysis. Allvelocities are heliocentric and corrected to (see, e.g., Springel et al. 2008; Wang et al. 2012). therest frameoftheGalaxy.Inordertopreventbiases due Pairsofcomparableluminosityarethereforemorelikely to incompleteness, we consider only satellites brighter than to form in fairly massive halos, where galaxy formation ef- MV =−8 located within 300 kpc of the Milky Way or An- ficiency decreases with increasing halo mass, partly com- dromeda (M31) galaxies. The sample consists of 29 dwarfs, pensating the mass difference between the main halo and 17 of which orbit around M31; the rest are satellites of the its most massive subhalo. On dwarf galaxy scales the situ- Milky Way. ation is reversed, and the precipitous decline in galaxy for- For comparison, we have identified analogous samples mationefficiencywithdecreasinghalomassshouldcurbthe of simulated satellites in the six ∼ 1012M⊙ halos of the formation of physical pairs of comparable luminosity. More Aquarius Project (Springelet al. 2008) using the model of generally speaking, isolated associations of dwarf galaxies Starkenburget al. (2012). This is a semi-analytic model should be rare. They are known to exist (e.g., Tully et al. grafted ontothelevel-2Aquariusruns,which simulateeach 2006; Soares 2007), but their cosmological abundance and halo with several hundredmillion particles, thusensuring a dependenceon luminosity havenot yet been adequatelyes- resolution high enough to track the formation of all halos tablished (see Sales et al. 2012, for a recent attempt). and subhalos that might plausibly host the dwarf galaxies brighter than MV =−8 we use in our analysis. The model The Local Group offers an interesting environment to satellites of each Aquarius halo have luminosity and radial testtheseideas.Advantagesincludethefactthat,awayfrom distributionsthatarebroadlyconsistentwiththeMilkyWay the “zone of avoidance” caused by Galactic dust, the cen- and M31 and therefore provide a useful testbed of the sig- sus of Milky Way (MW) satellites brighter than MV ∼ −8 nificance of our results for ΛCDM dwarf galaxy formation is complete (see, e.g., Whitinget al. 2007), and that accu- models. rate magnitudes, positions, distances, and line-of-sight ve- There are a total of 175 simulated satellites brighter locities are known for all. Many fainter systems in the Lo- than MV = −8 within 300 kpc of the primary galaxies of cal Group still remain undiscovered, as demonstrated by all six Aquarius halos (on average 29 per halo). The model recent discoveries both by the Sloan Digital Sky Survey provides not only the full 3D position and velocity infor- in the Milky Way (Koposov et al. 2008), and by the Pan- mation for all of them, but also allows us to track their Andromeda Archaeological Survey around M31 (PAndAS; evolution. Our simulated sample does not include satellites McConnachie et al. 2009). In the latter the census is likely whosedarkmatterhaloshavebeenfullydisruptedbytides, complete within the PAndAS survey area down to a mag- since their fate is uncertain. We refer the interested reader nitude limit of MV ∼ −6.5, although some brighter dwarfs toStarkenburget al.(2012)fordetailsonthesemi-analytic at larger radius likely have yet to be identified (as exem- model. plifiedbythediscoveriesofAndromedaXXVIIIandXXIX; Bell et al. 2011; Slater et al. 2011). Further,we know that at least some of thesatellites of 3 ANALYSIS AND RESULTS the Milky Way and M31 are very likely physically associ- It has long been noticed that the spatial distribution of ated and bound to each other. An obvious pairing is that Milky Way satellites is highly anisotropic (Lynden-Bell of the Magellanic Clouds (see, e.g., Kallivayalil et al. 2006, 1976),and isoften described as apolarplanewhose signifi- and references therein). Around M31, there have been sug- cance has been the matter of much recent debate(see, e.g., gestionsthatNGC147andNGC185alsoformaboundpair Kroupaet al. 2005; Zentneret al. 2005; Libeskind et al. (van den Bergh 1998). If these pairs are bound,their vicin- 2005; Metz et al. 2007). Around M31, 14 out of 17 satel- ity to their primary galaxy suggests that we are observing lites in our sample are in the hemisphere nearer the Milky them just before they are separated by the tidal field of Way(McConnachie & Irwin2006).Further,severalhavere- the main galaxy (Besla et al. 2007; Sales et al. 2011). This cently been shown to delineate a flattened structure that impliesveryrecentaccretionandindicatesthattheiroccur- in total comprises at least half of all known M31 satellites rence should not beuncommon amongst isolated systems. (Ibataet al. 2013). These are unlikely configurations for a Taken at face value, the existence of these two pairs of virialized population and hint strongly at recent accretion. dwarfsseemsatoddswiththeexpectedrarityofsuchassoci- Our pairing procedure begins by identifying satellites ations. Weusethisasmotivation tosearch,usingliterature whose nearest neighbour is unusually close when compared data,forotherdwarfgalaxypairsintheLocalGroup.Wede- with the probability distribution of nearest-neighbour dis- scribeinSec.2theobservationaldataset andthesimulated tances, dnn, obtained by Monte Carlo sampling a random satellite dataset we use for comparison. In Sec. 3 we intro- isotropicpopulationofsatelliteswiththesametotalnumber duce the pairing procedure we have adopted and compare andradialdistribution.WeillustratethisinFig.1,wherewe the results with those obtained when the same procedure show the dnn distribution expected for two satellites of the isapplied toahybridN-body/semi-analyticmodelof satel- Milky Way (Draco and Sagittarius), and two of M31 (And lite galaxy formation applied to N-body simulations from III and And IX). theAquariusProject.Weconcludewithabriefsummaryin The bottom left panel of Fig. 1 shows that the near- Sec. 4. estsatellitetoSagittarius(theLMC,52kpcaway)isabout Galaxy Pairs in the Local Group 3 Figure2.Distributionoftheprobability,P,ofhavinganearest neighbourascloseasorcloserthanobservedifthesatelliteswere isotropicallydistributed around each primaryand had the same radial distribution as that of the Milky Way (solid thick line) Figure 1. Distribution of nearest-neighbour distances, dnn, to andM31(soliddashedline).Only20%ofsatellitesareexpected twoMilkyWaysatellites(panelsontheleft)andtwoM31satel- to have P < 0.2, with a very weak dependence on the number lites (panels on the right) expected if satellites were distributed of satellites and the shape of the radial profile. The probability isotropicallyabouteachprimarywitharadialdistributionconsis- distribution obtained for a semi-analytic model applied to the tent withthe observedone. Theprobability,P,that asatellite’s six Galaxy-sized halos of the Aquarius Project is shown by the nearest neighbour lies, by chance, as close as or closer than ob- dashed blue histogram; 48 out of 175 satellites have P <0.2, or servedishighlightedbytheshadedregionofeachhistogramand 27%ofthetotal.ThecorrespondingdistributionforLocalGroup quoted in each panel’s legend. A downward arrow indicates the satellites is shown by the solid red histogram (the contribution distance to the primary galaxy. The top panels illustrate cases of Milky Way satellites is highlighted by the shaded area of the where the probability is rather small, indicative of a potential histogram).IntheLocalGroup,morethan40%ofsatelliteshave physicalassociation.Thebottompanels,ontheotherhand,illus- P <0.2,aresultexpectedtohappenbychanceinfewerthanone tratetwocaseswherethenearestneighboursarenotsignificantly in100randomrealizations.Theangularandradialdistributions closerthanexpected atrandom. ofsatellitesthusseemhighlycorrelatedandsuggestthepresence ofphysically-associatedpairs. twiceasfarasthedistanceatwhichtheprobabilitydistribu- tion of nearest-neighbour distances peaks. If the radial and change almost imperceptibly if we confine the Monte Carlo angular distribution of Milky Way satellites were uncorre- samplestoathree-dimensionalstructureasflatasobserved lated then Sagittarius would be expected to have a nearest for the Milky Way, i.e., roughly 3:1 in its major-to-minor neighbour as close or closer than observed in 99 out of 100 axis ratio. randomrealizations(P =0.99).Sagittariusisthusrelatively On the other hand, these distributions differ markedly isolated and unlikely to be a memberof a physical pair. from the results for the Milky Way (shaded histogram in Thesituation reversesforDraco:itsnearestneighbour, Fig. 2) or the combined M31+MW satellites (labelled “Lo- UrsaMinor,liesonly23kpcaway.Thisismuchcloserthan calGroup”inFig.2).AK-Stestyieldsaprobabilityofless expectedatrandom;anearestneighbourthatcloseoccursin than 0.3% that the Local Group P distribution is statisti- fewerthan4outof100randomtrials(P =0.04).Theright- callyconsistentwiththatoftherandomsamples.Thereisa hand panels of Fig. 1 show as well two analogous examples clearexcessofsmaller-than-expectednearest-neighbourdis- for the M31 satellite population. In this case, And IX is tances in the Local Group that is difficult to account just unlikely to bea memberof a pair (P =0.90), whereas And bychance.Forexample,45% ofLocal Groupsatellites have III is unusually close to And I (P = 0.10), hinting at a P <0.2comparedwiththe20%expectedifthedistribution possible physical association. was isotropic. The distribution of the probability, P, that dnn is as Our pairing procedure therefore retains all P < 0.2 smallorsmallerthanobservediftheradialandangulardis- pairs (listed horizontally in the labels of Fig. 2) for further tributionofMilkyWaysatelliteswereuncorrelatedisshown scrutiny. A true physical pair must also differ little in ve- by the solid and dotted lines in Fig. 2 for the Milky Way locity, so we impose a maximum difference of 75 km/s in andM31satellites,respectively.Bothcurvesarerathersim- the line-of-sight velocity difference of the likely members. ilar, indicating that the P distribution is insensitive to the Thisthresholdismotivatedbythevelocitydifferenceofthe total number of satellites or their radial distribution. It is Magellanic Clouds, where there is little doubt about their also insensitive to assuming that the satellite distribution physicalassociation.Weassumeforsimplicitythatthesame is isotropic. Indeed, the solid and dotted curves in Fig. 2 threshold applies regardless of the luminosity of the pair; 4 Fattahi et al. Figure3.Leftpanel:GalactocentricvelocityversusdistanceforLocalGroupsatellitesbrighterthanMV =−8.Distancesaremeasured from the center of each primary: MW satellites are shown as magenta squares; M31’s as green triangles. Velocities for the former are Galactocentric radial velocities; for the latter they refer to line-of-sight velocities relative to the systemic velocity of M31. Dotted curves indicate, for reference, the escape velocity from an NFW halo with virial velocity V200 = 250 km/s and concentration c = 10 (Navarroetal. 1997). Filled symbols highlight satellites with a nearest neighbour much closer than expected by chance (P < 0.2, see Fig. 2). Pairs satisfying additional proximity criteria in velocity (∆V < 75 km/s) and magnitude (∆MV < 3) are joined together by ellipsestoindicatethattheyarelikelyphysicalpairs.Theseconstitute28%ofthetotalandinclude(i)theMagellanicClouds;(ii)NGC 147 and NGC 185; (iii)UrsaMinor and Draco; and (iv) And I and And III. Right panel: Same as left panel, but for the semi-analytic satellitepopulationofthesixAquariushalos.Differentcolorscorrespondtodifferenthalos.Notethatthesamecriteriathatpair28%of LocalGroupsatelliteslinkonlysixsatellitesintheAquariussimulations,orjust3%ofthetotal. thisisalsoconsistentwiththeideathatmostdwarfsshould The blue dotted histogram in Fig. 2 shows that the P inhabit halos of similar mass (see Sec. 1) (see Sales et al. distribution for Aquarius satellites differs little from that 2012,for a more thorough discussion). Finally, since we are expected from an isotropic distribution. Only 27% of the mainlyinterestedinpairsofcomparableluminositieswere- 175 satellites haveP <0.2; of thoseonly3 pairs(i.e., fewer tain aslikely pairsonly thosewheretheirmagnitudesdiffer than 4%) pass as well the velocity and magnitude criteria. by ∆MV <3. The three Aquarius pairs (out of six halos) singled out by Four pairs remain after applying these constraints, as the analysis are shown in the right-hand panel of Fig. 3. shown in Fig. 3. Around the Milky Way, aside from the Trackingtheirorbitsbackintimerevealsthatnoneofthem MagellanicClouds,theUrsaMinorandDracopairissingled are actually physically related but that they result simply out: they are 23 kpc apart, and their velocities differ by from chance, transient associations in position and veloc- only 12 km/s. Further, they have both approximately the ity space. Note that this does not imply that all satellites sameluminosity,whichmakesthemespeciallysingular.The havebeenaccretedinisolation.AsdiscussedbyWang et al. procedure also joins together two pairs of satellites around (2012), a few of the bright satellites in Aquarius were ac- M31:NGC147andNGC185, aswell asAndIandAndIII. creted in groups, but the accretion happened early and the Thelatterareseparatedby33kpc;differinvelocitybyjust groups have long been disrupted by the tidal field of the 32 km/s; and in magnitude by 1.7. mainhalo.Thisconfirmsthemodelexpectationthatphysi- This analysis suggests that nearly 30% of Local Group calassociationsamongstsatellitesshouldbeextremelyrare. satellites are in likely pairs (8 out of 29). Of the four pairs, TherelativelyhighfrequencyofsatellitepairingsintheLo- two are almost indisputably associated (the Magellanic calGroupindicatesthattheradialandangulardistributions Clouds and NGC 147/NGC 185; see, however, Geha et al. ofsatellitesarecorrelated,afactthatisnoteasilyaccounted 2010 for an alternate view of the latter) but the other two forbycurrentdwarfgalaxyformation modelsintheΛCDM might in principle result from chance close encounters be- paradigm. tweenunrelatedsatelliteswhereprojectioneffectsreducethe line-of-sight velocity difference. In order to quantify these effects we have applied the same pairing procedure to the satellite populations of thesix Aquariushalos, as identified by thesemi-analytic model of Starkenburget al. (2012). Galaxy Pairs in the Local Group 5 4 SUMMARY AND CONCLUSIONS for relic evidence of past interactions between likely pairs (suchastheMagellanicStreamfortheLMC/SMC(see,e.g., We have studied possible pairings amongst the satellites of Mathewson et al. 1974; Putman et al. 1998); and, finally, the Milky Way and of M31. Our procedure, which iden- (iv) by extending this kind of analysis to a volume-limited tifies unusually close associations in position and velocity surveyofdwarfgalaxyassociationsinthelocaluniverse(see, space, suggests that 8 out of the 29 satellites brighter than e.g., Karachentsev & Makarov 2008). If confirmed, the en- MV =−8(i.e., nearly∼30%) form 4likelypairsofcompa- hancedclusteringofdwarfsmayofferimportantcluestothe rable luminosity (∆MV <3). These include the Magellanic formationoffaintgalaxiesthathaveyettobeidentifiedand Clouds; UrsaMinorandDraco; NGC147and NGC185; as fully incorporated into galaxy formation models. well as AndI and And III. The same pairing procedure applied to a semi-analytic model of the satellite population in the six halos of the Aquarius Project yields a likely pair fraction of fewer than ACKNOWLEDGMENTS 4%,eventhoughthemodelsatelliteshaveluminosityandra- dialdistributionsthatmatchcloselythatoftheLocalGroup We are very grateful to Gabriella De Lucia, Amina Helmi spirals. Asexpected,noneoftheAquariuspairscorrespond and Yang-Shyang Li for their role in developing the semi- totruebinarysystems;rather,theyresultfromtransientas- analytic model of galaxy formation used in this paper. We sociations between otherwise unrelated satellites. The high would like to thank the anonymous referee for comments pair frequency of the Local Group is unlikely to be just a whichledtotheimprovementofthemanuscript.ESissup- statisticalfluke:thelikelypairfractionofAquariussatellites ported by the Canadian Institute for Advanced Research neverexceeds12%inthousandsofrandomtrialswheretheir (CIfAR) Junior Academy and by a Canadian Institute for magnitudes,angulardirectionsandvelocitiesarereshuffled. Theoretical Astrophysics (CITA) National Fellowship. We interpret these results as indicative of signifi- cant clustering in the dwarf galaxy population of the Lo- cal Group. Although our analysis only considers satel- REFERENCES lites brighter than MV = −8 due to incompleteness con- cerns and in order to allow comparison with simulations, Alvarez M. A., Busha M., Abel T., Wechsler R. H., 2009, there have also been suggestions that some of the fainter ApJ,703, L167 Galactic satellites are found in associations. In particular, Barnes D.G., deBlok W. J. G., 2001, AJ, 122, 825 Belokurov et al. (2008) show that Leo IV and Leo V are BehrooziP.S.,ConroyC.,WechslerR.H.,2010,ApJ,717, close to each other spatially and differ little in their line- 379 of-sight velocities. Further, the association of galaxies clos- BellE.F.,SlaterC.T.,MartinN.F.,2011, ApJ,742,L15 est to the Local Group, sometimes referred to as the loose Belokurov V., Walker M. G., Evans N. W., Faria D. C., NGC 3109 group (van den Bergh 1999), consists of four GilmoreG.,IrwinM.J.,KoposovS.,MateoM.,Olszewski dwarfgalaxies(NGC3109,Antlia,SextansAandB),twoof E., ZuckerD.B., 2008, ApJ,686, L83 which appear to be interacting (NGC3019 and Antlia; see Ben´ıtez-Llambay A.,NavarroJ.F.,AbadiM.G.,Gottloe- Barnes & deBlok2001).Theseassociationsareexpectedto berS.,YepesG.,HoffmanY.,SteinmetzM.,2012,ArXiv berare accordingtocurrentΛCDMgalaxy formation mod- e-prints els,andmaysignalthepresenceofamechanismthatboosts BensonA.J.,BowerR.G.,FrenkC.S.,LaceyC.G.,Baugh the likelihood of forming dwarf galaxies near other dwarfs, C. M., Cole S., 2003, ApJ, 599, 38 favouring some environments over others. No such effect is Besla G., Kallivayalil N., HernquistL., Robertson B., Cox present in current models of dwarf galaxy formation, which T.J.,vanderMarelR.P.,AlcockC.,2007,ApJ,668,949 relylargelyonthemassaccretionhistoryofindividualhalos Boylan-Kolchin M., Bullock J. S., Kaplinghat M., 2012, to set theproperties of a dwarf. MNRAS,422, 1203 Candidatemechanismsthatmaybiastheregionswhere Busha M. T., Alvarez M. A., Wechsler R. H., Abel T., dwarf galaxies form in a way that enhances their clustering Strigari L. E., 2010, ApJ, 710, 408 include patchy reionization (see, e.g., Lunnan et al. 2012); Ferrero I., Abadi M. G., Navarro J. F., Sales L. V., large-scale feedback effects on the surroundings of massive Gurovich S., 2012, MNRAS,425, 2817 galaxies (Alvarez et al. 2009; Busha et al. 2010; Font et al. FontA.S.,BensonA.J.,BowerR.G.,FrenkC.S.,Cooper 2011);andtheinteractionofdwarfgalaxieswiththecosmic A.,DeLuciaG.,HellyJ.C.,HelmiA.,LiY.-S.,McCarthy web (Ben´ıtez-Llambay et al. 2012). However, the effects of I. G., Navarro J. F., Springel V., Starkenburg E., Wang these models on dwarf galaxies have yet to be developed J., White S.D. M., 2011, MNRAS,417, 1260 fullyandasaconsequencetheimportanceofsucheffectson Geha M., van der Marel R. P., Guhathakurta P., Gilbert thedwarf galaxy population at large is still unknown. K.M., Kalirai J., KirbyE. N.,2010, ApJ, 711, 361 It is therefore important to firm up these findings (i) GuoQ.,ColeS.,EkeV.,FrenkC.,2011,MNRAS,417,370 by extendingtheanalysis tofainter satellites, which should Guo Q., White S., Li C., Boylan-Kolchin M., 2010, MN- be possible once photometric surveys of the northern and RAS,404, 1111 southern sky extend the complete catalog of Milky Way Ibata R. A., Lewis G. F., Conn A. R., Irwin M. J., Mc- satellites to fainter magnitudes; (ii) by verifying, through Connachie A. W., Chapman S. C., Collins M. L., Fardal accuratepropermotionstudies,thattheassociations inpo- M.,FergusonA.M.N.,IbataN.G.,MackeyA.D.,Martin sition and line-of-sight velocities remain once the full 6D N.F., Navarro J., Rich R. M., Valls-Gabaud D., Widrow phase space information is considered; (iii) by searching L.M., 2013, Nature,493, 62 6 Fattahi et al. KallivayalilN.,vanderMarelR.P.,AlcockC.,2006,ApJ, 652, 1213 KarachentsevI.D.,MakarovD.I.,2008,AstrophysicalBul- letin, 63, 299 Koposov S., Belokurov V.,EvansN. W., Hewett P. C., Ir- winM.J.,GilmoreG.,ZuckerD.B.,RixH.-W.,Fellhauer M., Bell E. F., GlushkovaE. V., 2008, ApJ, 686, 279 Kroupa P., Theis C., Boily C. M., 2005, A&A,431, 517 Libeskind N. I., Frenk C. S., Cole S., Helly J. C., Jenkins A., NavarroJ. F., Power C., 2005, MNRAS,363, 146 LunnanR.,VogelsbergerM.,FrebelA.,HernquistL.,Lidz A., Boylan-Kolchin M., 2012, ApJ,746, 109 Lynden-BellD., 1976, MNRAS,174, 695 Mathewson D.S.,Cleary M.N.,MurrayJ. D.,1974, ApJ, 190, 291 McConnachie A. W., 2012, AJ, 144, 4 McConnachie A.W.,IrwinM.J.,2006, MNRAS,365,902 McConnachie et al. 2009, Nature, 461, 66 Metz M., Kroupa P., Jerjen H., 2007, MNRAS,374, 1125 Moster B. P., Somerville R. S., Maulbetsch C., van den Bosch F.C., Maccio` A.V., Naab T., Oser L.,2010, ApJ, 710, 903 NavarroJ.F.,FrenkC.S.,WhiteS.D.M.,1997,ApJ,490, 493 PutmanM.E.,GibsonB.K.,etal.1998, Nature,394,752 Sales L. V., Navarro J. F., Cooper A. P., White S. D. M., FrenkC. S., Helmi A., 2011, MNRAS,418, 648 SalesL.V.,WangW.,WhiteS.D.M.,NavarroJ.F.,2012, MNRAS,p. 31 SlaterC.T.,BellE.F.,MartinN.F.,2011, ApJ,742,L14 Soares D.S. L., 2007, AJ, 134, 71 SpringelV.,WangJ.,VogelsbergerM.,LudlowA.,Jenkins A.,HelmiA.,NavarroJ.F.,FrenkC.S.,WhiteS.D.M., 2008, MNRAS,391, 1685 Starkenburg E., Helmi A., De Lucia G., Li Y.-S., Navarro J.F.,FontA.S.,FrenkC.S.,SpringelV.,Vera-CiroC.A., WhiteS. D.M., 2012, ArXive-prints Tully R. B., Rizzi L., Dolphin A. E., Karachentsev I. D., Karachentseva V. E., Makarov D. I., Makarova L., Sakai S., ShayaE. J., 2006, AJ, 132, 729 van den Bergh S., 1998, AJ, 116, 1688 van den Bergh S., 1999, ApJ, 517, L97 Vera-CiroC.A.,HelmiA.,StarkenburgE.,BreddelsM.A., 2012, MNRAS,p. 125 WangJ.,FrenkC.S.,CooperA.P.,2012,MNRAS,p.366 Wang J., Frenk C. S., Navarro J. F., Gao L., Sawala T., 2012, MNRAS,424, 2715 WhitingA.B.,HauG.K.T.,IrwinM.,VerdugoM.,2007, AJ, 133, 715 ZentnerA.R.,KravtsovA.V.,GnedinO.Y.,KlypinA.A., 2005, ApJ, 629, 219 Zwaan M. A., Meyer M. J., Staveley-Smith L., 2010, MN- RAS,403, 1969 ThispaperhasbeentypesetfromaTEX/LATEXfileprepared by theauthor.