LETTERdoi:10.1038/nature11717

A vast, thin plane of corotating dwarf galaxiesorbiting the Andromeda galaxyRodrigo A. Ibata1, Geraint F. Lewis2, Anthony R. Conn3, Michael J. Irwin4, Alan W. McConnachie5, Scott C. Chapman6,Michelle L. Collins7, Mark Fardal8, Annette M. N. Ferguson9, Neil G. Ibata10, A. Dougal Mackey11, Nicolas F. Martin1,7,Julio Navarro12, R. Michael Rich13, David Valls-Gabaud14 & Lawrence M. Widrow15

Dwarf satellite galaxies are thought to be the remnants of the popu-lation of primordial structures that coalesced to form giant galaxieslike the Milky Way1. It has previously been suspected2 that dwarfgalaxies may not be isotropically distributed around our Galaxy,because several are correlated with streams of H I emission, and mayform coplanar groups3. These suspicions are supported by recentanalyses4–7. It has been claimed7 that the apparently planar distri-bution of satellites is not predicted within standard cosmology8, andcannot simply represent a memory of past coherent accretion. How-ever, other studies dispute this conclusion9–11. Here we report theexistence of a planar subgroup of satellites in the Andromeda galaxy(M 31), comprising about half of the population. The structure is atleast 400 kiloparsecs in diameter, but also extremely thin, with aperpendicular scatter of less than 14.1 kiloparsecs. Radial velocitymeasurements12–15 reveal that the satellites in this structure havethe same sense of rotation about their host. This shows conclusivelythat substantial numbers of dwarf satellite galaxies share the samedynamical orbital properties and direction of angular momentum.Intriguingly, the plane we identify is approximately aligned with thepole of the Milky Way’s disk and with the vector between the MilkyWay and Andromeda.

We undertook the Pan-Andromeda Archaeological Survey16

(PAndAS) to obtain a large-scale panorama of the halo of the Andro-meda galaxy (M 31), a view that is not available to us for the MilkyWay. This Canada–France–Hawaii Telescope survey imaged about400 square degrees around M 31, which is the only giant galaxy inthe Local Group besides our Milky Way. Stellar objects are detectedout to a projected distance of about 150 kiloparsecs (kpc) from M 31,and about 50 kpc from M 33, the most massive satellite of M 31. Thedata reveal a substantial population of dwarf spheroidal galaxies thataccompany Andromeda17.

The distances to the dwarf galaxies can be estimated by measuringthe magnitude of the tip of the red-giant branch18. Improving onearlier methods, we have developed a Bayesian approach that yieldsthe probability distribution function for the distance to each individualsatellite19. In this way we now have access to homogeneous distancemeasurements (typical uncertainties 20–50 kpc) to the 27 dwarf gala-xies (filled circles in Fig. 1) visible within the PAndAS survey area20

that lie beyond the central 2.5u.In Fig. 2 these distance measurements are used to calculate the sky

positions of the homogeneous sample of 27 dwarf galaxies as theywould appear from the centre of the Andromeda galaxy. Visually, thereappears to be a correlation close to a particular great circle (red line):

this suggests that there is a plane, centred on M 31, around which asubsample of the satellites have very little scatter. This is confirmed bythe Monte Carlo analysis presented in the Supplementary Informa-tion, where we show that the probability of the alignment of the sub-sample of nsub 5 15 satellites marked red in Figs 1 and 2 occurring atrandom is 0.13% (see Supplementary Fig. 1).

Following this discovery, we sought to investigate whether the sub-sample displayed any kinematic coherence. The radial velocity of eachsatellite is shown in Fig. 3, corrected for the bulk motion of theAndromeda system towards us: what is immediately striking is that13 out of the 15 satellites possess coherent rotational motion, such thatthe southern satellites are approaching us with respect to M 31, while thenorthern satellites recede away from us with respect to their host galaxy.

The probability that 13 or more out of 15 objects should share thesame sense of rotation is 1.4% (allowing for right-handed or left-handed rotation). Thus the kinematic information confirms the spatialcorrelation initially suspected from a visual inspection of Fig. 2. Thetotal significance of the planar structure is approximately 99.998%.

Thus we conclude that we have detected, with very high confidence,a coherent planar structure of 13 satellites with a root-mean-squarethickness of 12.6 6 0.6 kpc (,14.1 kpc at 99% confidence), that coro-tate around M 31 with a (right-handed) axis of rotation that points ap-proximately east. The three-dimensional configuration can be assessedvisually in Fig. 3. The extent of the structure is gigantic, over 400 kpcalong the line of sight and nearly 300 kpc north-to-south. Indeed, sinceAndromeda XIV and Cassiopeia II lie at the southern and northernlimits of the PAndAS survey, respectively, it is quite probable thatadditional (faint) satellites belonging to this structure are waiting tobe found just outside the PAndAS footprint. Although huge in extent,the structure appears to be lopsided, with most of the satellites popu-lating the side of the halo of M 31 that is nearest to the Milky Way. Thecompleteness analysis we have undertaken shows that this configura-tion is not due to a lowered detection sensitivity at large distance, butreflects a true paucity of satellites in the more distant halo hemisphere21.

The existence of this structure had been hinted at in earlier work22,thanks to the planar alignment we report on here being viewed nearlyedge-on, although the grouping could not be shown to be statisticallysignificant from the information available at that time. Other previ-ously claimed alignments do not match the present plane, althoughthey share some of the member galaxies: the most significant align-ment of ref. 23 has a pole 45u away from that found here, and the(tentative) poles of the configuration in ref. 4 are 23.4u away, whereasthose of a later contribution6 have poles 34.1u and 25.4u distant from

1Observatoire Astronomique de Strasbourg, 11 rue de l’Universite, F-67000 Strasbourg, France. 2Sydney Institute for Astronomy, School of Physics, A28, The University of Sydney, New South Wales 2006,Australia. 3Department of Physics and Astronomy, Macquarie University, New South Wales 2109, Australia. 4Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK.5NRCHerzberg Institute of Astrophysics, 5071West Saanich Road, Victoria, British Columbia, V9E2E7,Canada. 6Departmentof Physics and AtmosphericScience,DalhousieUniversity, 6310Coburg Road,Halifax, Nova Scotia, B3H 4R2, Canada. 7Max-Planck-Institut fur Astronomie, Konigstuhl 17, 69117 Heidelberg, Germany. 8University of Massachusetts, Department of Astronomy, LGRT 619-E, 710 NorthPleasant Street, Amherst, Massachusetts 01003-9305, USA. 9Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK. 10Lycee International desPontonniers, 1 rue des Pontonniers, F-67000 Strasbourg, France. 11The Australian National University, Mount Stromlo Observatory, Cotter Road, Weston Creek, Australian Capital Province 2611, Australia.12Department of Physics and Astronomy, University of Victoria, 3800 Finnerty Road, Victoria, British Columbia, V8P 5C2, Canada. 13Department of Physics and Astronomy, University of California, LosAngeles, PAB, 430 Portola Plaza, Los Angeles, California 90095-1547, USA. 14LERMA, UMR CNRS 8112, Observatoire de Paris, 61 Avenue de l’Observatoire, 75014 Paris, France. 15Department of Physics,Engineering Physics, and Astronomy, Queen’s University, Kingston, Ontario, K7L 3N, Canada.

6 2 | N A T U R E | V O L 4 9 3 | 3 J A N U A R Y 2 0 1 3

Macmillan Publishers Limited. All rights reserved©2013

Galactic longitude

Gal

actic

latit

ude

A7IC10

A28

A6

IC10IC10IC10

LGS3 IC1613

A26A26147

C2

185

A25

A27

A17

A9

A1

A3

A13A16

A14

A5

A24

A11

A10

A23

A15

A2

M33

A22

A19A20

A21

A18

205

M32M32M32

205205205

A29

A12

Figure 1 | Map of the Andromeda satellite system. The homogeneousPAndAS survey (irregular polygon) provides the source catalogue for thedetection and distance measurements of the 27 satellite galaxies20 (filled circles)used in this study. Near M 31 (blue ellipse), the high background hampers thedetection of new satellites and precludes reliable distance measurements forM 32 and NGC 205 (labelled black open circles); we therefore exclude theregion inside 2.5u (dashed circle) from the analysis. The seven satellites knownoutside the PandAS area (green circles and arrows) constitute a heterogeneous

sample, discovered in various surveys with non-uniform spatial coverage, andtheir distances are not measured in the same homogeneous way. A reliablespatial analysis requires a data set with homogeneous selection criteria, so we donot include these objects in the sample either. The analysis shows that thesatellites marked red are confined to a highly planar structure. We note that thisstructure is approximately perpendicular to lines of constant Galactic latitude,so it is therefore aligned approximately perpendicular to the Milky Way’s disk(the grid squares are 4u3 4u).

0

100

200

300

400

500

600

M 31-centric galactic longitude

M 3

1-ce

ntric

gal

actic

latit

ude

A21

A18

A19A20

A22

M33

A2

A15

A23

A10

A5A24

A26

A25

147

A30A17

185

A9

A16A3

A1

A13

A11A14

A12

A27

MW

Figure 2 | Satellite galaxy positions as viewed from Andromeda. The Aitoff–Hammer projection shows the sample of 27 satellites20 (filled circles from Fig. 1)as they would be seen from the centre of the Andromeda galaxy. In thesecoordinates the disk of Andromeda lies along the equator. ‘M 31-centricgalactic latitude’ means what a fictitous observer in the M31 galaxy wouldcall ‘galactic latitude’. The background image represents the probabilitydensity function of the poles derived from 105 iterations of resampling the 27satellites from their distance probability density functions, and finding the

plane of lowest root mean square from a subsample of 15 (the colour scale onthe right shows the relative probability of the poles, and is dimensionless). Aclear narrow peak at (lM 31 5 100.9u6 0.9u, bM 31 5 238.2u6 1.4u)highlights the small uncertainty in the best-fit plane. The solid red line,which passes within less than 1u of the position of the Milky Way(yellow circle labelled ‘MW’), represents the plane corresponding to this bestpole location.

LETTER RESEARCH

3 J A N U A R Y 2 0 1 3 | V O L 4 9 3 | N A T U R E | 6 3

Macmillan Publishers Limited. All rights reserved©2013

ours. Without the increased sample size, reliable three-dimensionalpositions and radial velocities, and most importantly, a spatiallyunbiased selection function resulting from the homogeneous pano-ramic coverage of PAndAS, the nature, properties and conclusivestatistical significance of the present structure could not be inferred.

The present detection proves that in some giant galaxies, a signifi-cant fraction of the population of dwarf satellite galaxies, in this casearound 50% (13 out of 27 over the homogeneously surveyed PAndASarea), are aligned in coherent planar structures, sharing the samedirection of angular momentum. The Milky Way is the only othergiant galaxy where we have access to high-quality three-dimensionalpositional data, and the existence of a similar structure around ourGalaxy is strongly suggested by current data2,5,7. The implications forthe origin and dynamical history of dwarf galaxies are profound. It alsohas a strong bearing on the analyses of dark matter in these darkest of

galaxies, because one cannot now justifiably assume such objects tohave evolved in dynamical isolation.

Intriguingly, the Milky Way lies within 1u of the plane reported here,the pole of the plane and the pole of the Milky Way’s disk are approxi-mately perpendicular (81u), and furthermore this plane is approximatelyperpendicular to the plane of satellites that has been proposed to sur-round the Galaxy (given that its pole points approximately towardsAndromeda7 within the uncertainties). Although these alignments maybe chance occurrences, it is nevertheless essential information about thestructure of the nearby Universe that must be taken into account infuture simulations aimed at modelling the dynamical formation historyof the Local Group.

The formation of this structure around M 31 poses a puzzle. Fordiscussion, we envisage two broad classes of possible explanations:accretion or in situ formation. In either type of model, the small scatter

A15

A9

NGC185 C2NGC147

A25A26

A17

A1

A12

A14

A11A2

A24

A21

A19

A22

M33

A23

A20A3

A10

M31

Click figure to activate

–200 0.00 200

z – zM 31 (kpc)

–200

0.00

200

y (k

pc)

2000.00

–200

y (kpc)

2000.00–200

z – zM 31 (kpc)

A16

A18

Figure 3 | Three-dimensional view (online only in the PDF version) or two-dimensional screenshot (in the print version) of the planar, rotatingstructure. The coordinate system is such that the z direction is parallel to thevector pointing from the Milky Way to M 31, x increases eastwards and ynorthwards. Only the radial component of the velocity of each satellite ismeasured, and these velocities are shown as vectors pointing either towards oraway from the Milky Way. As in Figs 1 and 2, red spheres mark the planarsatellites, and blue spheres represent the ‘normal’ population. The coherentkinematic behaviour of the spatially very thin structure (red) is clearly apparentviewed from the y–z plane. With the exception of And XIII and And XXVII, thesatellites in the planar structure that lie to the north of M 31 recede from us,whereas those to the South approach us; this property strongly suggests

rotation. Our velocity measurements15 (supplemented by values from theliterature14), have very small uncertainties, typically ,5 km s21. The irregulargreen polygon (visible only in the x–y plane of the three-dimensional onlineversion) shows the PAndAS survey area, the white circle (visible only in the x–yplane of the three-dimensional online version) indicates a projected radius of150 kpc at the distance of M 31, and the white arrow (visible only in the three-dimensional online version) marks a velocity scale of 100 km s21. (And XXVIIis not shown in this diagram because its most likely distance is 476 kpc behindM 31). This figure is three-dimensionally interactive in the online version(allowing the reader to change the magnification and viewing angle), and wasconstructed with the S2PLOT programming library26.

RESEARCH LETTER

6 4 | N A T U R E | V O L 4 9 3 | 3 J A N U A R Y 2 0 1 3

Macmillan Publishers Limited. All rights reserved©2013

of the satellites out of the plane is difficult to explain, even though theorbital timescales for the satellites are long (around 5 Gyr for satellitesat 150 kpc). All the galaxies in the plane are known to have old, evolved,stellar populations, and so in situ formation would additionally implythat the structure is ancient.

In an accretion scenario, the dynamical coherence points to anorigin in a single accretion of a group of dwarf galaxies. However,the spatial extent of the progenitor group would have to be broadlyequal to or smaller than the current plane thickness (,14.1 kpc), yet nosuch groups are known. Interpreting the coherent rotation as a resultof our viewing perspective24 requires a bulk tangential velocity for thein-falling group of the order of 1,000 km s21, which seems unphysi-cally high. A further possibility is that we are witnessing accretionalong filamentary structures that are fortuitously aligned. In situ for-mation may be possible if the planar satellite galaxies formed like tidal-dwarf galaxies in an ancient gas-rich galaxy merger7, but then thedwarf galaxies should be essentially devoid of dark matter. If the planarM 31 dwarfs are dynamically relaxed, the absence of dark matter wouldbe greatly at odds with inferences from detailed observations25 of MilkyWay satellites, assuming the standard theory of gravity. An alternativepossibility is that gas was accreted preferentially onto dark matter sub-halos that were already orbiting in this particular plane, but then theorigin of the plane of sub-haloes would still require explanation. Weconclude that it remains to be seen whether galaxy formation modelswithin the current cosmological framework can explain the existenceof this vast, thin, rotating structure of dwarf galaxies within the halo ofour nearest giant galactic neighbour.

Received 18 September; accepted 23 October 2012.

1. Klypin, A., Kravtsov, A. V., Valenzuela, O. & Prada, F. Where are the missing galacticsatellites? Astrophys. J. 522, 82–92 (1999).

2. Lynden-Bell, D. Dwarf galaxies and globular clusters in high velocity hydrogenstreams. Mon. Not. R. Astron. Soc. 174, 695–710 (1976).

3. Lynden-Bell, D. & Lynden-Bell, R. M. Ghostly streams from the formation of theGalaxy’s halo. Mon. Not. R. Astron. Soc. 275, 429–442 (1995).

4. Metz, M., Kroupa, P. & Jerjen, H. The spatial distribution of the Milky Way andAndromeda satellite galaxies. Mon. Not. R. Astron. Soc. 374, 1125–1145 (2007).

5. Metz,M.,Kroupa,P.&Libeskind,N. I.TheorbitalpolesofMilkyWaysatellitegalaxies:a rotationally supported disk of satellites. Astrophys. J. 680, 287–294 (2008).

6. Metz, M., Kroupa, P. & Jerjen, H. Discs of satellites: the newdwarf spheroidals. Mon.Not. R. Astron. Soc. 394, 2223–2228 (2009).

7. Pawlowski, M. S., Pflamm-Altenburg, J. & Kroupa, P. The VPOS: a vast polarstructure of satellite galaxies, globular clusters and streams around the Milky Way.Mon. Not. R. Astron. Soc. 423, 1109–1126 (2012).

8. Komatsu, E. et al. Seven-year Wilkinson Microwave Anisotropy Probe (WMAP)observations: cosmological interpretation. Astrophys. J. 192 (Suppl.), 18 (2011).

9. Zentner, A. R., Kravtsov, A. V., Gnedin, O. Y. & Klypin, A. A. The anisotropicdistribution of galactic satellites. Astrophys. J. 629, 219–232 (2005).

10. Lovell, M. R., Eke, V. R., Frenk, C. S. & Jenkins, A. The link between galactic satelliteorbits and subhalo accretion. Mon. Not. R. Astron. Soc. 413, 3013–3021 (2011).

11. Wang, J., Frenk, C. S., Navarro, J. F., Gao, L. & Sawala, T. The missing massivesatellites of the Milky Way. Mon. Not. R. Astron. Soc. 424, 2715–2721 (2012).

12. Collins, M. L. M. et al. The scatter about the ’Universal’ dwarf spheroidal massprofile: a kinematic study of the M31 satellites And V and And VI. Mon. Not. R.Astron. Soc. 417, 1170–1182 (2011).

13. Tollerud, E. J. et al. The SPLASH survey: spectroscopy of 15 M31 dwarf spheroidalsatellite galaxies. Astrophys. J. 752, 45 (2012).

14. McConnachie, A. W. The observed properties of dwarf galaxies in and around theLocal Group. Astron. J. 144, 4 (2012).

15. Collins, M. L. et al. The non-universal dSpH mass profile? A kinematic study of theAndromeda dwarf spheroidal system. Mon. Not. R. Astron. Soc. (submitted).

16. McConnachie, A. W. et al. The remnants of galaxy formation from a panoramicsurvey of the region around M 31. Nature 461, 66–69 (2009).

17. Richardson, J. C. et al. PAndAS’ progeny: extending the M31 dwarf galaxy cabal.Astrophys. J. 732, 76 (2011).

18. Lee, M. G., Freedman, W. L. & Madore, B. F. The tip of the red giant branch as adistance indicator for resolved galaxies. Astrophys. J. 417, 553–559 (1993).

19. Conn, A. R. et al. A Bayesian approach to locating the red giant branch tipmagnitude. I. Astrophys. J. 740, 69 (2011).

20. Conn, A. R. et al. A Bayesian approach to locating the red giant branch tipmagnitude. II. Distances to the satellites of M31. Astrophys. J. 758, 11 (2012).1209.4952.

21. McConnachie, A. W. & Irwin, M. J. The satellite distribution of M31. Mon. Not. R.Astron. Soc. 365, 902–914 (2006).

22. Majewski, S. R. et al. Discovery of Andromeda XIV: a dwarf spheroidal dynamicalrogue in the Local Group? Astrophys. J. Lett. 670, L9–L12 (2007).

23. Koch, A. & Grebel, E. K. The anisotropic distribution of M31 satellite galaxies: apolar great plane of early-type companions. Astron. J. 131, 1405–1415 (2006).

24. van der Marel, R. P. & Guhathakurta, P. M31 transverse velocity and local groupmass from satellite kinematics. Astrophys. J. 678, 187–199 (2008).

25. Walker, M. G. Dark matter in the Milky Way’s dwarf spheroidal satellites. Preprint athttp://arxiv.org/abs/1205.0311 (2012).

26. Barnes, D. G., Fluke, C. J., Bourke, P. D. & Parry, O. T. An advanced, three-dimensional plotting library for astronomy. Publ. Astron. Soc. Aust. 23, 82–93(2006).

Supplementary Information is available in the online version of the paper.

Acknowledgements We thank the staff of the Canada-France-Hawaii Telescope fortaking the PAndAS data, and for their continued support throughout the project. Wethankoneof our referees,B. Tully, for pointingout that IC1613couldalso beassociatedto the planar structure. R.A.I. and D.V.G. gratefully acknowledge support from theAgence Nationale de la Recherche though the grant POMMME, and would like to thankB. Famaey for discussions. G.F.L. thanks the Australian Research Council for supportthrough his Future Fellowship and Discovery Project. This work is based onobservations obtained with MegaPrime/MegaCam, a joint project of CFHT and CEA/DAPNIA, at the Canada–France–Hawaii Telescope, which is operated by the NationalResearch Council (NRC) of Canada, the Institut National des Sciences de l’Univers ofthe Centre National de la Recherche Scientifique (CNRS) of France, and the Universityof Hawaii. Some of the data presented here were obtained at the W.M. KeckObservatory,which isoperatedasascientificpartnershipamong theCalifornia Instituteof Technology, the University of California and the National Aeronautics and SpaceAdministration. The Observatory was made possible by the generous financial supportof the W.M. Keck Foundation.

Author Contributions All authors assisted in the development and writing of the paper.In addition, the structural and kinematic properties of the dwarf population, and thesignificance of the Andromeda planewere determined by R.A.I., G.F.L. and A.R.C., basedon distances determined by the same group (as part of the PhD research of A.R.C.). Inaddition, A.W.M. is the Principal Investigator of PAndAS; M.J.I. and R.A.I. led the dataprocessing effort; R.A.I. was the Principal Investigator of an earlier CFHT MegaPrime/MegaCam survey, which PAndAS builds on (which included S.C.C., A.M.N.F., M.J.I.,G.F.L., N.F.M. and A.W.M.). R.M.R. is Principal Investigator of the spectroscopic follow-upwith the Keck Telescope. M.L.C. and S.C.C. led the analysis of the kinematicdetermination of the dwarf population, and N.F.M. led the detection of the dwarfpopulation from PAndAS data. N.G.I. performed the initial analysis of the satellitekinematics.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of the paper. Correspondenceand requests for materials should be addressed to R.A.I.(rodrigo.ibata@astro.unistra.fr).

LETTER RESEARCH

3 J A N U A R Y 2 0 1 3 | V O L 4 9 3 | N A T U R E | 6 5

Macmillan Publishers Limited. All rights reserved©2013