Probing the Dark Matter Content of Local Group Dwarf Spheroidal Galaxies with FLAMES

Reports from Observers

Mark I. Wilkinson1 radial length scale) which are inferred to field stars (Shetrone et al. 2001; Tolstoy Jan T. Kleyna 2 contain dynamically significant quantities et al. 2006). Thus, the Galactic halo was Gerard F. Gilmore1 of dark matter. not predominantly formed by disrupt- N. Wyn Evans1 ing galaxies like the present-day dSphs. Andreas Koch 3 The proximity of the Milky Way dSph sat- Nevertheless, the chemical properties Eva K. Grebel 3 ellites makes it feasible to obtain spectra of dSphs provide valuable insights into Rosemary F. G. Wyse 4 for large numbers of individual stars in the processes of galaxy formation and Daniel R. Harbeck 5 them, facilitating more detailed studies of evolution on the smallest scales. their properties than are possible for galaxies beyond the Local Group. These It is the large apparent mass-to-light ra- 1 Institute of Astronomy, Cambridge spectra provide radial velocities and tios of dSph galaxies, and the conclu- University, Cambridge, United Kingdom chemical abundance determinations sion that they contain significant amounts 2 Institute for Astronomy, Honolulu, USA which, respectively, enable us to to probe of dark matter, that has caused them 3 Astronomical Institute of the University the stellar velocity distribution function to attract much attention in recent years. of Basel, Department of Physics and and halo potential within a dSph and pro- The first evidence of this emerged in Astronomy, Switzerland vide detailed information about the star- 1983, when Aaronson published velocity 4 The John Hopkins University, Baltimore, formation history of these systems. measurements of three carbon stars in USA Two recent Messenger articles have dis- the Draco dSph. These data implied a 5 University of Wisconsin, Madison, USA cussed in detail the insights into the velocity dispersion for Draco of 6.5 km/s, chemical evolution of the stellar popula- from which Aaronson cautiously inferred tions in dSphs which have been gained a mass-to-slight ratio of 30 solar units. We present preliminary kinematic re- by recent observations with the FLAMES As this was significantly higher than the sults from our VLT programme of spec- spectrograph on the VLT (Koch et al. values typical of globular star clusters, troscopic observations in the Carina 2006a; Tolstoy et al. 2006). In this article this provided the first hint that the dSphs dwarf spheroidal galaxy using the we will focus on the implications of the were a class of stellar system distinct FLAMES multi-object spectrograph. kinematic data in the Carina dSph for our from the globular clusters, despite the These new data suggest that the dark understanding of the dark matter. fact that in many cases their stellar mass matter halo of this galaxy has a uniform was similar. Subsequent observations density core. The implications for our The dSphs are the lowest-luminosity sat- have borne out these early estimates and understanding of the nature of the dark ellites of the Milky Way and M31, charac- all dSphs observed to date have velocity matter are discussed. terised both by their low stellar luminosi- dispersions in excess of 6.5 km/s. ties (up to 2 × 107 solar luminosities, with the majority in the range 105–106 solar By 1998, due to the dedicated efforts of Dark matter in dSphs luminosities) and the apparent absence of several teams, velocity dispersions had gas and on-going star formation. The been measured for eight dSphs (Mateo According to the current cosmological recent discovery of the Ursa Major dSph 1998). However, dSph velocity distribu- paradigm, non-luminous, non-baryonic (Willman et al. 2005) and the Canes tions were still represented solely by the matter contributes approximately 20 % of Venatici (Zucker et al. 2006) and Bootes central velocity dispersion, which con- the overall mass-energy budget of the (Belokurov et al. 2006) dSph candidates tains only limited information about the Universe. Locally, on the scale of galaxies brings to twelve the number of dSphs nature of the system. The situation like the Milky Way, it constitutes about around the Milky Way. A similar number changed in 1997, with the publication of 90 % of the gravitating mass. Determining have been identified orbiting M31. At the first velocity dispersion profile for the nature of this dark matter is a key least one dSph is clearly being disrupted a dSph (Mateo 1997). Using 215 individual goal of contemporary astronomy. The by the tidal field of the Milky Way – the stellar velocities in Fornax, Mateo showed dwarf spheroidal galaxies (dSphs) of the Sagittarius dSph has extensive debris that the velocity dispersion remains ap- Local Group provide a particularly valua- trails which have been observed over proximately flat almost to the edge of the ble window on the properties of dark much of the sky. The stellar distributions light distribution. This profile is inconsist- matter on small (~ 1 kpc) scales. Several of most dSphs exhibit some level of dis- ent with the simplest model of Fornax of these low-luminosity galaxies have turbance in their outer regions, although in which the mass is distributed in the been found to have extremely large whether this is due to their proximity same way as the light and the stars have mass-to-light (M/L) ratios (up to 1000 in to their parent galaxies remains contro- an isotropic velocity distribution. solar units) which suggests that these versial. Although the dSphs were once are the most dark-matter-dominated stel- thought to be the primordial building The past decade has seen a rapid in- lar systems known in the Universe (e.g. blocks whose disruption contributed to crease in the size of dSph kinematic data Kleyna et al. 2001, Wilkinson et al. 2004, the hierarchical build-up of the stellar sets, driven by the availability of multi- Mateo et al. 1998). Given the apparent halo of the Milky Way, recent studies object spectrographs on 4-m to 10-m- absence of dark matter in globular star have found that the chemical composi- class telescopes which allow the simulta- clusters, they are also the smallest stellar tion of the stars in the present-day neous acquisition of spectra for large systems (in terms of stellar mass and dSphs differs substantially from that of numbers of stars. The first dSph disper-

The Messenger 124 – June 2006 25 Reports from Observers Wilkinson M. I. et al., Probing the Dark Matter Content of Local Group dSphs

sion profile based on multi-fibre observa- 25 Figure 1: Spatial distribution of Carina tions was that of Draco, based on veloc- members observed in our survey. 20 Stars approaching and receding (re- ities for 159 stars obtained using the 15 lative to Carina) are shown as blue WYFFOS spectrograph on the William and red symbols, respectively. The Herschel Telescope (Kleyna et al. 2001). 10 size of the symbol is proportional to the magnitude of the radial velocity. Since that time, dispersion profiles 5 n) The ellipse indicates the nominal tidal have been measured for all the Milky Way mi

rc 0 limit of Carina with semi-major axis dSphs. For most Milky Way dSphs, the of 28.8 arcmin. large (25 arcmin diameter) field of view of Y (a −5 the FLAMES spectrograph on the VLT −10 is ideally matched to their angular size −15 facilitating the efficient coverage of the full −20 area of each object. −25 −30 −20 −10 0 10 20 30 The properties of dSph haloes are largely X (arcmin) determined by the physical properties of the dark matter. For example, the number of low-mass dark-matter haloes in global stellar kinematics of all the dSphs errors are 1.2 km/s for the entire data the vicinity of a Milky Way-type galaxy is have not yet been defined sufficiently set and 1.5 km/s for the faintest stars a strong function of the nature of the dark well to yield definitive results on the inner (V = 20–20.5). matter – cold dark matter (CDM) simu- slope of their dark matter distributions. lations typically predict hundreds of low- Our Carina data set will be sufficiently A total of 1257 targets in Carina were ob- mass haloes around the Milky Way while large to place robust constraints on the served. Of these, 535 are probable mem- warm dark-matter models contain sig- profile of a dSph dark-matter halo, while bers of Carina (based on their radial nificantly fewer. However, it is currently our modelling will further the analyses velocities). The remainder are foreground not clear which objects in the simulations of the data from all the dSphs. Milky Way stars which can be used to correspond to the observed dSphs. A investigate the properties of the velocity number of authors have claimed that the distribution of our own Galaxy (see properties of dSphs are consistent with Observations of the Carina dSph below). Of the probable members, some their residing in haloes of mass > 109 437 stars have spectra with sufficient solar masses. In this case, the numbers The Carina dSph has previously attracted signal-to-noise to enable chemical abun- of dSphs might be consistent with attention due to its unusual, bursty star- dances to be determined using the Cal- CDM simulations. One of the key goals of formation history, with the most recent cium triplet estimator (Koch et al. 2006b). our observations in Carina is to test in peak occurring around 3 Gyr ago (e.g. The spatial distribution of Carina mem- detail whether its properties are consist- Monelli et al. 2003). Its central velocity bers from our survey is shown in Figure 1. ent with those of an object with a mas- dispersion is 6.8 km/s (Mateo et al. 1993), As our primary goal in this survey is to sive, extended dark halo. implying a mass-to-light ratio of about map out the velocity structure of Carina, 30–40 (solar units). Majewski et al. (2005) it is particularly important to observe The shapes of the dark-matter density provide evidence of irregularity in the stars at large projected radii as these profiles of dSphs are also sensitive to the outer regions of the light profile, perhaps place the strongest constraints on the type of dark matter. Galaxy haloes com- associated with tidal disturbance. Our mass and extent of the galaxy. As can be posed of cold dark matter are predicted Carina data were obtained as part of seen in the figure, our targets probe to to be cusped, with the density in the in- VLT Large Programme 171.B-0520 (PI: the nominal tidal radius of Carina (around ner regions rising as 1/rn, where n lies in Gilmore). The targets were selected to lie 28.8 arcmin). Our sample increases the the range 1–1.5. Alternatively, if the dark on the red giant branch of the colour- number of measured stellar velocities for matter were warm, i.e. composed of par- magnitude diagram of Carina and span Carina members by almost an order of ticles which had a non-negligible intrin- the magnitude range V = 17.5 to 20 magnitude and extends to radii at which sic velocity dispersion, then this would (see Figure 2 of Koch et al. 2006a). The both the effects of an extended halo as lead to a shallow or uniform density core, photometric and astrometric data for well as the tidal effects of the Milky Way whose size depends on the actual ve- the input catalogue were obtained by the may be observable. locity dispersion of the particles. The pres- ESO Imaging Survey (EIS) as part of ence or absence of a cusp, as well as the Wide-Field Imager Pre-FLAMES sur- The histogram in Figure 2 shows the dis- its precise steepness is one of the most vey. The spectroscopic data were taken tribution of velocities for our full sample direct measurements of the nature of in visitor mode and reduced using the of stars along the line of sight to Carina. dark matter. There is already evidence of standard GIRAFFE reduction software, The members of Carina are clearly visible cored haloes in some dSphs (e.g. Kleyna with sky subtraction performed using our as the narrow peak centred at about et al. 2003; Goerdt et al. 2006) but the own in-house software (see Wilkinson 223.8 km/s. The systemic velocity of identification of a property of the dark et al. 2006 [in prep.] for more details on Carina is not well separated from the ve- matter requires a universal result. The the data processing). Median velocity locities of Milky Way stars along the line

26 The Messenger 124 – June 2006 350 Figure 2: Velocity distribution of all profiles for the metal-rich and metal-poor stars observed along the line of sight stars in Carina based on the above defini- to Carina. The peak due to Carina 300 members at around 223.8 km/s is tion. The amplitude of the dispersion clearly visible. profile of the metal-rich stars appears to 250 be systematically lower than that of the metal-poor stars. This is consistent with 200 their relative spatial distributions, since

N the more extended metal-poor population 150 would be expected to exhibit a larger velocity dispersion. The observation by 100 Harbeck et al. (2001) that the younger stellar populations in Carina are more 50 centrally concentrated than the older stars, combined with the age-metallicity 0 −50 0 50 100 150 200 250 300 350 relation for Carina stars identified by

vlos (km/s) Koch et al. (2006b), would then suggest an interpretation of the kinematic data in 15 ity dispersion with position in the dSph. In terms of a secondary, more recent burst Figure 3 we present the dispersion profile of star formation concentrated towards for Carina, plotted as a function of pro- the centre of Carina. s) 10 jected radius. To calculate the dispersion

(km/ in each radial bin of this profile, we as- The flatness of the velocity dispersion sume that the projected velocity distribu- profile has important implications for the

(R) 5

σ tion in each bin is a Gaussian convolved dark-matter density profile of Carina. with the stellar velocity errors (which For a stellar system in virial equilibrium, 0 are also assumed to be Gaussian). We the Jeans equations provide a simple 15 also introduce a power-law interloper estimator of the halo mass distribution velocity distribution to take account of the once the spatial distribution of the stars s) 10 presence of foreground Milky Way stars and their velocity-dispersion profile are

(km/ in our velocity samples. known. Figure 4 shows the density profile inferred from our Carina data set.

(R) 5

σ The dispersion profile of Carina based on We have used the surface-density pro- the complete data is remarkably flat, file determined by Majewski et al. (2005) 0 with an amplitude of about 7–8 km/s out to determine the stellar-density distribu- 0 10 20 30 to the edge of the data at a radius of tion and have assumed that the gravita- R (arcmin) approximately 45 arcmin. Koch et al. tional potential is everywhere dominated (2006b) showed that if the stellar pop- by the dark matter. Spherical symmetry Figure 3: Top: Velocity dispersion profile for all stars ulation is divided at a metallicity of is assumed and we have simplified the in Carina. Bottom: Velocity dispersion profiles for [Fe/H] = –1.68, the ‘metal-rich’ sample is form of the Jeans equations by assuming stars with [Fe/H] < –1.68 (blue) and [Fe/H] > –1.68 (red). slightly more centrally concentrated than that the velocity distribution is every- the ‘metal-poor’ population. The lower where isotropic. The black line in the den- panel of Figure 3 shows the dispersion sity plot shows the form of the density of sight. This makes it more difficult to establish membership, as simple criteria 109 (e.g. the standard velocity cut of three times the internal velocity dispersion) may include a non-negligible level of foreground contamination. 108 ) –3 c kp Dark matter in Carina 0 (M

The large central velocity dispersions in 107 dSphs are usually interpreted as evidence Figure 4: Mass density profile of of significant amounts of dark matter. Carina as inferred from the velocity Much stronger constraints on the amount dispersion profile using Jeans Carina of dark matter in a dSph, as well as its equations and assuming velocity 1/r isotropy. The black curve shows 106 spatial distribution, can be obtained from 10–2 10–1 100 the expected relation for a standard knowledge of the variation of the veloc- r (kpc) cusped halo.

The Messenger 124 – June 2006 27 Reports from Observers Wilkinson M. I. et al., Probing the Dark Matter Content of Local Group dSphs

profile expected from cosmological simu- gation of our Carina data, and a velocity is probed by kinematic observations. Fur- lations of galaxy formation in the cold sample towards M31, Martin et al. (2006) ther, several of the dSph mass estimates dark-matter paradigm. In contrast to the provided an alternative interpretation, presented in this plot are based only on profiles of the haloes seen in simula- namely that they detect the velocity sig- the central velocity dispersion – the mass tions, in the central regions of Carina it nature of the Monoceros ring, another scale may therefore be somewhat dif- appears that the dark-matter density is large stellar structure which is suggested ferent once detailed dynamical models close to uniform. to surround the Milky Way disc at low are completed for all dSphs. However, latitude. this plot demonstrates that it is possible We emphasise, however, that the dark- to interpret the velocity dispersions of matter density profile presented here is In order to gain a better understanding of dSphs in terms of the intrinsic properties based on strong assumptions about the what the internal kinematics of dSphs of the dark matter. stellar spatial and velocity distributions can tell us about the nature of the dark which may not be justified, particularly at matter, it is important to consider the very large radii. By assuming velocity properties of these objects as a class. In Future work isotropy, we are prey to the strongest de- Figure 5, we plot the mass-to-light ratios generacy of this problem, namely that a of the local group dSphs against their Now that velocities have been measured flat dispersion profile in an isolated dSph absolute V-band magnitudes. The figure as far out as the nominal tidal radii in can be produced either by the presence is based on that given in Mateo et al. most dSphs, the next major advance of large amounts of mass at large radii, or (1998) but includes more recent mass de- will be to obtain usefully large samples of by radially varying anisotropy in the veloc- terminations where these are available – stars at much larger radii. Although ity distribution. In our modelling of the the mass for Carina is based on the Wilkinson et al. (2004), and Munoz et al. Draco dSph (Wilkinson et al. 2002), we simple dynamical modelling presented (2005) have taken the first steps in this constructed a family of dynamical models above. The solid curve shows the ex- direction, the results are still controversial which included dark matter haloes of pected relation for a population of objects and larger data sets are needed. The varying extent, as well as radially varying with a stellar mass-to-light ratio of unity key difficulties are the large areas of sky velocity anisotropy. We are currently ex- (solar units) and a total halo mass of which must be surveyed, and the pre- tending this methodology to new dynami- 4 × 107 solar masses. In this case, the selection of targets in regions where the cal models appropriate for the (multi- total mass-to-light ratio, (M/L)Tot is simply expected surface density of dSph mem- population) case of Carina (Wilkinson et given by (M/L)Tot = (M/L)Stars + MDM/LV, bers is extremely low. However, stars in al. 2006, in prep.). These models will en- where MDM is the mass in dark matter these regions will provide extremely use- able us to quantify the properties of and LV is the total V-band luminosity. Not- ful information about dSph haloes. We the dark-matter density profile of Carina. withstanding the significant scatter about expect that at some level all dSphs must Our models incorporate multiple trac- this relation, it appears that most obser- be affected by the tidal field of the Milky er populations, to enable us to use the vations to date are consistent with dSphs Way. However, thus far there is no de- metal-rich and metal-poor sub-popula- having a common halo mass scale of finitive kinematic evidence of this process tions of Carina (and other dSphs) as around 4 × 107 solar masses, generally in any dSph. Constraining the radii at independent tracers of the underlying cored mass distributions, and a low cen- which the Milky Way begins to disturb dark-matter potential. tral dark-matter density. It is important dSphs will make it possible to derive the to note that this mass scale refers to the current total mass and extent of their In addition to their value in constraining mass interior to the edge of the stellar dark haloes. Of course, identifying funda- models of the dark matter, it has become distribution, since this is the region which mental properties of the dark matter re- apparent that our kinematic data can also be used to investigate the properties of the Milky Way velocity distribution 3.5 along the line of sight to Carina. Two recent papers have investigated this novel 3 possibility. Wyse et al. (2006) compared UMa Draco the velocity distributions along the line 2.5 UMi ] of sight to three dSphs (Carina, Draco V AndIX 2 Carina tot, and Ursa Minor) and found an excess of Sextans

stars with an azimuthal velocity of about M/L) 1.5 AndII

100 km/s relative to the predictions of g [( LeoII LeoI

lo Fornax smooth Galaxy models. They conclude Figure 5: Mass-to-light ratios versus 1 absolute V magnitude for Local Group that this kinematic structure, which is dSphs. The solid curve shows the seen along widely separated lines of sight, 0.5 Sculptor relation expected if all dSph haloes may be the remains of an object which contain about 4 × 107 solar masses of fell into the Milky Way at early times, and 0 dark matter interior to their stellar distributions. Arrows indicate mass may even have contributed to the for- −8 −10 −12 −14 −16 estimates which are lower limits based on central velocity dispersions only. mation of the Thick Disc. In their investi- Mv

28 The Messenger 124 – June 2006 quires that common properties in all the Acknowledgements Koch A. et al. 2006a, The Messenger 123, 38 dSphs be identified – we must therefore Koch A. et al. 2006b, AJ 131, 895 Mark I. Wilkinson acknowledges the Particle Physics Majewski S. R. et al. 2005, AJ 130, 2677 carry out similar studies of all dSphs. and Astronomy Research Council of the United Martin N. et al. 2006, MNRAS 367, L69 Kingdom for financial support. Andreas Koch and Mateo M. et al. 1993, AJ 105, 510 Finally, in the area of dynamical model- Eva K. Grebel thank the Swiss National Science Mateo M. 1997, ASP Conf. Ser. 116, 259 ling, identifying correlations between the Foundation for financial support. Mateo M. et al. 1998, AJ 116, 2315 Monelli M. et al. 2003, AJ 126, 218 kinematics and abundances of the stel- Munoz R. R. et al. 2005, ApJ 631, L137 lar populations in dSphs (e.g. Tolstoy et References Shetrone M. D. et al. 2001, ApJ 548, 592 al. 2006) is likely to provide important Tolstoy E. et al. 2006, The Messenger 123, 33 new information about the formation and Aaronson M. 1983, ApJ 266, L11 Wilkinson M. I. et al. 2002, MNRAS 330, 778 Belokurov V. et al. 2006, ApJL, submitted, Wilkinson M. I. et al. 2004, MNRAS 611, L21 evolution of these objects, which in turn astro-ph/0604355 Wilkinson M. I. et al. 2006, in proceedings of XXIst will further constrain models of any astro- Goerdt T. et al. 2006, MNNRAS 368, 1073 IAP meeting, EDP sciences, astro-ph/0602186 physical feedback on their dark matter. Harbeck D. et al. 2001, AJ 122, 3092 Willman B. et al. 2005, ApJ 626, L85 Kleyna J. T. et al. 2001, ApJ 564, L115 Wyse R. F. G. et al. 2006, ApJ 639, L13 Kleyna J. T. et al. 2003, ApJ 588, L21 Zucker D. B. et al. 2006, ApJ 643, L103

A Three-Planet Extrasolar System

Using the ultra-precise HARPS spectro- Planetary System graph on ESO’s 3.6-m telescope at around HD 69830 1 (Artist’s Impression). La Silla, a team of astronomers has discovered that a nearby star is host to three Neptune-mass planets. The in- nermost planet is most probably rocky, while the outermost is the first known Neptune-mass planet to reside in the habitable zone. This unique system is likely further enriched by an asteroid belt.

Over more than two years, the team carefully studied HD 69830, a rather in- conspicuous nearby star slightly less orbiting their parent star with periods The outer planet has probably accreted massive than the Sun. Located 41 light of 8.67, 31.6 and 197 days. some ice during its formation, and is years away towards the constellation likely to be made of a rocky/icy core sur- of Puppis, it is, with a visual magnitude of “Only ESO’s HARPS instrument installed rounded by a quite massive envelope. 5.95, just visible with the unaided eye. at the La Silla Observatory, Chile, made Further calculations have also shown that The team’s precise radial-velocity meas- it possible to uncover these planets”, said the system is in a dynamically stable con- urements allowed them to discover Michel Mayor, from Geneva Observatory, figuration. the presence of three tiny companions and HARPS Principal Investigator. “With- out any doubt, it is presently the world’s The outer planet also appears to be lo- 1 Lovis et al. 2006, Nature 441, 305. The team is most precise planet-hunting machine.” cated near the inner edge of the habit- composed of Christophe Lovis, Michel Mayor, able zone, where liquid water can exist at Francesco Pepe, Didier Queloz, and Stéphane The detected velocity variations are be- the surface of rocky/icy bodies. Although Udry (Observatoire de l’Université de Genève, Swit- tween two and three metres per sec- this planet is probably not Earth-like due zerland), Nuno C. Santos (Observatoire de l’Uni- versité de Genève, Switzerland, Centro de Astro- ond. Such small signals could not have to its heavy mass, its discovery opens the nomia e Astrofisica da Universidade de Lisboa and been distinguished from noise by most way to exciting perspectives. Centro de Geofisica de Evora, Portugal), Yann of today’s available spectrographs. Alibert, Willy Benz, Christoph Mordasini (Physika- With three roughly equal-mass planets, lisches Institut der Universität Bern, Switzerland), François Bouchy (Observatoire de Haute-Provence The newly found planets have minimum one being in the habitable zone, and a and IAP, France), Alexandre C. M. Correia (Uni- masses between 10 and 18 times the possible asteroid belt, this planetary sys- versidade de Aveiro, Portugal), Jacques Laskar mass of the Earth. Extensive theoretical tem shares many properties with our own (IMCCE-CNRS, Paris, France), Jean-Loup Bertaux simulations favour an essentially rocky Solar System. (Service d’Aéronomie du CNRS, France), and Jean-Pierre Sivan (Laboratoire d’Astrophysique de composition for the inner planet, and a Marseille, France). rocky/gas structure for the middle one. (Based on ESO Press Release 18/06)

The Messenger 124 – June 2006 29