Astronomical Science

Boötes-I, Segue 1, the Orphan Stream and CEMP-no Stars: Extreme Systems Quantifying Feedback and Chemical Evolution in the Oldest and Smallest Galaxies

Gerard Gilmore1 function has been extended down to most enigmatic of the ultra-faint systems. Sergey Koposov1 luminosities three orders of magnitude These targets include: the lowest luminos- John E. Norris2 below previous limits in recent years. ity system Segue 1 (Belokurov, 2007a; Lorenzo Monaco3 Remarkably, these extremely low lumi- see Simon et al. [2011] for a detailed Keck David Yong2 nosity objects, the dwarf spheroidal study) and the Orphan stream (Belokurov Rosemary Wyse4 (dSph) galaxies with total luminosities as et al., 2007b), both of which are at similar 1 Vasily Belokurov low as 1000 LA, comparable to a moder- distances to the bifurcated tidal tail of Doug Geisler5 ate star cluster, are quite unlike star Sgr (Ibata, Gilmore & Irwin, 1994); the N. Wyn Evans1 ­clusters — they are real galaxies. Fortu- common-distance and similar-velocity Michael Fellhauer5 nately, with their very few red giants but pair Leo-IV and Leo-V (Belokurov et al., Wolfgang Gieren5 more populous main-sequence turn- 2008); and the inner regions of Boötes-I Mike Irwin1 off stars, they are within range of detailed (Belokurov et al., 2006b), a surviving ex­ Matthew Walker1, 6 study, with considerable efforts currently ample of one of the first bound objects Mark Wilkinson7 underway at the Very Large Telescope to form in the Universe, providing a touch- Daniel Zucker8 (VLT) and Keck. stone to test the chemical evolution of the earliest low-mass stars. These lowest-luminosity galaxies provide 1 Institute of Astronomy, Cambridge, a unique opportunity to quantify the Our FLAMES GIRAFFE and UVES spec- United Kingdom ­formation and chemical enrichment of the tra are beginning to quantify the kine­ 2 Mount Stromlo Observatory, Australian first bound structures in the Universe. matics and abundance distribution func- National University, Canberra, Australia At present, there are no convincing mod- tions in these systems, including several 3 ESO els for the origin and evolution of these element ratios, providing the first quan­ 4 Johns Hopkins University, Baltimore, extreme objects — observations lead the titative study of what we find to be survi- USA way. The ultra-faint dSphs certainly have vors of truly primordial systems that 5 Departamento de Astronomia, extreme properties. They are clustered in ap ­parently formed and evolved before the ­Universidad de Concepcion, Chile groups on the sky, their velocity disper- time of reionisation. Our target fields are 6 Harvard University, Cambridge, USA sions are tiny — no more than 3–4 km/s summarised in Figure 1. 7 University of Leicester, Leicester, United at the lowest luminosities — yet the ob­ Kingdom jects themselves are very extended, with 8 Macquarie University, Sydney, Australia half-light radii of hundreds of parsecs, Segue 1 and the Orphan Stream implying extreme dark matter dominance. Their chemical abundances are also Segue 1 is the lowest luminosity galaxy Galactic satellite galaxies provide a extreme, with dispersions of several dex, known. It has a wide abundance range, unique opportunity to map the history and containing stars down to [Fe/H] = –4. including hosting a very carbon-enhanced of early star formation and chemical There are hints that they are associated metal-poor star with no enhancement evolution, the baryonic feedback on gas with, or possibly entangled in, kinematic of heavy neutron-capture elements over and dark matter, and the structure of streams and superimposed on — or in — Solar ratios (called a CEMP-no star: c.f., low-mass dark matter halos, in surviv- the tidal tails of the more luminous Sagit- Norris et al. [2010a] for our VLT study and ing examples of the first galaxies. We tarius dSph (Sgr) galaxy. Beers & Christlieb [2005] for the defini- are using VLT–FLAMES spectroscopy tions of metal-poor star subclasses). Such to map the kinematics and chemical The lowest luminosity dSphs do not look stars, which are like those in the Milky abundances of stars in several ultra- like the tidal debris of larger systems, Way Halo field, are suspected to be suc- faint dwarf spheroidal galaxies and the they look like the most primordial galax- cessors of the very first supernovae — enigmatic Orphan Stream in the Halo. ies of all. Arguably even more interesting see below. The velocity distribution Two paths of early chemical enrichment than their relevance as galaxy building measured in Segue 1, which is consistent at very low iron abundance are observed blocks is to understand the objects them- with the Keck study of Simon et al. (2011), directly: one rapid and carbon-rich selves. Are they the first objects? Did they shows a very narrow distribution, consist- (CEMP-no), one slow and carbon-nor- contribute significantly to reionisation? ent with a dispersion of less than 4 km/s. mal. We deduce long-lived, low-rate What do they tell us of the first stars? In spite of this low dispersion, Segue 1 is star formation in Boötes-I, implying What was the stellar initial mass function completely dark-matter dominated, with insignificant dynamical feedback on the (IMF) at near-zero metallicity? How are a mass-to-light ratio in excess of 1000. structure of its dark matter halo, and the faintest dSphs related to more lumi- The velocity distribution function, in which find remarkably similar kinematics in the nous dSphs and the Milky Way galaxy? Segue 1 has its radial velocity near apparently discrete systems Segue 1 V = 200 km/s, also indicates the pres- and the Orphan Stream. In order to address these questions, we ence of stars from the Sagittarius tidal are obtaining VLT FLAMES observations stream (at V = 0 km/s), which is at a very using both the spectrographs GIRAFFE similar distance, and stars in a cold With the discoveries from the Sloan and UVES, with exposures of up to more ­kinematic structure with radial velocity ­Digital Sky Survey, the galaxy luminosity than 15 hours, of member stars of the V = 300 km/s. This cold highest-velocity

The Messenger 151 – March 2013 25 Astronomical Science Gilmore G. et al., Boötes-I, Segue 1, the Orphan Stream and CEMP-no Stars

 Figure 1. The background shows the “Field of Streams” (Belokurov et al., 2006a), the distribution of turn-off  stars observed by the Sloan Digital Sky Survey, statistically colour-coded  by distance with the bluer colour showing more nearby, the redder more distant streams. The fields  observed for this project by the VLT are marked in grey, with representative  velocity distributions indicated. The @S HNM velocity distribution for Boötes-I is  marked by its extreme narrowness.

#DBKHM The low surface brightness Orphan V 6WUHDP Stream is very evident in velocity  space, again with very low velocity 6HJXH %RÓWHV, dispersion. Segue 1 shows three dis-  tinct velocity peaks, corresponding 0RQRFHUR to the Sgr Stream, Segue 1 itself, and 6DJLWWDULXV6WUHDP the “300 km/s” stream, evident only in  /HR,9 2USKDQ 6WUHDP velocity space.

l      1HFGS@RBDMRHNM stream is not yet well defined, and remains pens to be passing through a busy part 240 pc), at 60 kpc Galactocentric dis- under study. of the outer Galaxy. The metal-poor tance, whose special features are its ([Fe/H] = –3.5) CEMP-no star Segue 1-7, low surface brightness and low total lumi-

The Orphan Stream provides another which we have studied with the VLT, nosity (Mv = –6). Boötes-I has low mean example of a Halo stream, although in lies almost four half-light radii from the metallicity (with a range from [Fe/H] = this case it has a sufficiently high surface centre of Segue 1, while all the other well- –3.7 to –1.9) and hosts a CEMP-no star brightness to allow it to be traced over studied members are inside 2.3 half- that we have studied with the VLT more than 60 degrees of arc. The peri- light radii (70 pc). Does this hint at tidal ([Fe/H] = –3.3; Norris et al., 2010b). Our galacticon of the Orphan Stream orbit truncation of an earlier, much larger (and GIRAFFE s­ pectra have been analysed for passes close to Segue 1, and hence to more luminous?) predecessor? Segue 1 a kinematic study (see Koposov et al., the Sagittarius tidal tail. As Figure 1 illus- is very deep inside the Galactic tidal 2011). Using careful data reduction tech- trates, the internal velocity dispersion field, well inside the (disrupting) Sgr dSph, niques, Koposov et al. (2011) showed in the Orphan Stream is unresolved at the and the Large Magellanic Cloud–Small that GIRAFFE spectra obtained in single resolution of the observations, being Magellanic Cloud pair, with their gaseous one-hour observing blocks over several less that 3–4 km/s. The Orphan Stream tidal stream. All the dSph galaxies that years can be combined to deliver radial and Segue 1 kinematics and distance, are not deep in the Galactic tidal field velocities with an accuracy floor ap­ the latter determined from main-sequence have much larger half-light radii (Gilmore proaching 0.1 km/s. From this study we turn-off fitting, provide an interesting et al., 2007). We also have the remarka- demonstrated that Boötes-I has a smaller ­illustration of the complexity of the outer ble similarity (indeed, near identity) of the velocity dispersion than suggested Galactic Halo. Both Segue 1 and the ­distances and radial velocities of Segue 1 by previous studies. We showed the kine- Orphan Stream have similarly low internal and the Orphan Stream as further clues. matics to be best de­s­cribed by a two- velocity dispersions. Remarkably, at the In spite of searching, we have not (as yet) component system, a majority with dis- Orphan Stream’s closest approach to found any trace of any extra-tidal Segue 1/ persion 2.5 km/s, and a minority with Segue 1, both have the same Galacto- Orphan Stream member stars, or a dispersion as high as 9 km/s, or by a sin- centric distance (within uncertainties), ­physical link between the systems: there gle dispersion of 4.6 km/s. Our pre- which is also the same as the local Sagit- are no identified stars with appropriate ferred interpretation is that the apparent tarius tail, being separated by only a few velocities in the spatial region between multi-component kinematic structure kiloparsecs (kpc) at most. Even more the Orphan Stream peri-galacticon and may reflect orbital anisotropy inside bizarrely, at the point of closest ap­proach, Segue 1. Are they just ships passing Boötes-I. Our observations to date are the orbit of the Orphan Stream has in the dark night? The hunt for enlighten- limited to the central regions, so ­further exactly the same Galactocentric radial ment continues. study at larger radii is required to clarify velocity as does Segue 1. Is this coinci- the situation. dence, or evidence of a common history? Boötes-I There are key issues in early galaxy The analysis of Simon et al. (2011) sug- ­evolution which can be resolved by the gests that Segue 1 is contained inside Boötes-I provides an example of a comp­ ­analysis of chemical element distribu- its tidal radius, and is a robust, albeit lementary case: an apparently normal, tions. These include the early stellar high- small and faint, galaxy, which just hap- extended dSph galaxy (half-light radius mass IMF, star formation rates at very

26 The Messenger 151 – March 2013 a) b) Figure 2. A schematic diagram of the two star for- 1 - FIRST GENERATION SUPERNOVAE 2 - SECOND GENERATION STARS Carbon-rich branch mation and evolutionary paths evident at the lowest First SNe First SNe abundances. The very first stars, apparently of high +3 Carbon-rich +3 Carbon-rich Efficient cooling mass, generate two patterns of chemical enrichment Star formation & SNe ejecta o ejecta — carbon-enhanced and carbon-normal (Figure 2a). o ati ati This difference may correspond to different progeni- n r n r → tors, or may simply be spatial inhomogeneity from ro ro /i

n/i a single SN. The carbon-enhanced material very rapidly cools, apparently forming stars with a wide

Carbo range of masses, including SN progenitors, which First SNe Carbon First SNe 0 C-normal 0 C-normal generate standard Galactic Population ii abundances ejecta ejecta (Figure 2b). This process continues until the carbon abundance in the ISM is diluted to the Solar value, Iron/hydrogen ratio Iron/hydrogen ratio when [Fe/H] ~ –3. The lower carbon abundance ISM –5 –3 –1 –5 –3 –1 does not take part in this enrichment process, but on a slower timescale stars are formed, again with a wide range of masses and an apparently standard IMF (Figure 2c). This enrichment path also eventually c) d) 3 - SECOND GENERATION STARS 4 - THIRD GENERATION STARS reaches [Fe/H] = –3. The whole ISM is now suffi- Carbon-poor branch ciently enriched for efficient cooling, so that chemi- First SNe First SNe cal evidence of the evolutionary path is now lost +3 Carbon-rich +3 Carbon-rich (Figure 2d). Should either path correspond to suffi- ejecta ejecta o o ciently slow star formation, SNe Type i a will generate ati rati low [ /Fe] at this stage. The correspondence of this n α n r ro ro model with data from stars in Boötes-I (circles) and /i /i Segue 1 (triangles) is shown in Figure 2e. Efficient cooling AGB carbon; Carbon Carbon First SNe First SNe Inefficient slow cooling; SN I a Fe production? 0 C-normal 0 C-normal late star formation and SNe ejecta ejecta significant in our present sample. The

Iron/hydrogen ratio Iron/hydrogen ratio observed lack of scatter in these ele- –5 –3 –1 –5 –3 –1 ment ratios at a given [Fe/H] requires that: (i) the stars formed from gas that was enriched by ejecta sampling the e) Bootes-1 STARS mass range of the progenitors of core- collapse supernovae (SNe); (ii) the super- nova progenitor stars formed with an 2 +3 IMF similar to that of the Solar neighbour- hood today; and (iii) the ejecta from all o C-rich fast cooling ati 1 SNe were efficiently well-mixed. Both the e] n r first and last points set an upper limit [C /F on how rapidly star formation could have

bon/i ro 0 proceeded, since: the star formation re­ gions need to populate the entire massive- Car C-poor slow cooling 0 star IMF, the stars need sufficient time –1 to all explode, and the gas needs time to –4 –3.5 –3 –2.5 –2 –1.5 [Fe/H] mix the ejected enriched material. All Iron/hydrogen ratio these steps must occur before substan- –5 –3 –1 tial numbers of low-mass stars form.

The observed lack of scatter in the early times, and their consequences, and Enhanced ratios of α-element/Fe above α-element abundance ratios requires that feedback on baryonic gas and the dark the Solar values are expected in stars the well-sampled IMF of core-collapse matter potential well. Our UVES observa- formed from gas that is predominantly supernova progenitors is invariant over tions of Boötes-I address these points enriched by these endpoints of massive the range of time concerned and/or the directly (Gilmore et al., 2013). stars. Thus chemical abundances in iron abundance. This can be expressed the stars formed within the first 0.5 Gyr as a constraint, from the scatter, on the after star formation began will reflect the variation in slope of the massive star α-elements and the IMF products of predominantly core-collapse IMF, assuming that the ratios reflect IMF-­ supernovae. averaged yields. A scatter of 0.02 dex The α-elements, together with a small constrains the variation in IMF slope to be amount of iron, are created and ejected Although we see hints of a declining 0.2. The overall agreement between the by core-collapse supernovae, on time- α-element abundance, suggestive of a values of the elemental abundances in scales of less than 108 years after the for- resolution of the duration of the chemical Boötes-I stars and in the field of the Halo mation of the supernova progenitors. evolution of Boötes-I, this is not formally implies the same value of the massive

The Messenger 151 – March 2013 27 Astronomical Science Gilmore G. et al., Boötes-I, Segue 1, the Orphan Stream and CEMP-no Stars

star IMF that enriched the stars in each of Given the amplitude of the [C/Fe] and essentially primordial initial abundances. the two samples, although our formal [Mg/Fe] values in CEMP-no stars, one This allows us uniquely to investigate the limit on this IMF slope is only agreement must also explain why stars are not found place of CEMP-no stars in a chemically within a slope range of 1. with intermediate C and Mg excesses at evolving system, as well as to limit the higher [Fe/H]. Apparently the highly C- timescale of star formation in this dSph. The Galactic Halo and Boötes-I (and and Mg-enriched ISM does not survive to The low elemental abundance scatter Segue 1) display a large range in carbon mix with “normal” enriched ISM and form requires low star formation rates, allowing abundance at low metallicity. For iron more stars with moderate CEMP-no time for SNe ejecta to be created and abundances greater than about enrichment. Rather, the cooling efficiency mixed over the large spatial scales rele- [Fe/H] = –3.2, excess carbon enhance- of the highly carbon-enriched material vant. This is further evidence that B­ oötes-I ments above the solar [C/Fe] ratio are must be sufficiently great that all of it survived as a self-enriching star-forming consistent with carbon production in cools and forms (the surviving) low-mass system from very early times. It also asymptotic giant branch (AGB) compan- stars mixing with “normal” SNe ejecta im ­plies that only unimportant amounts of ions (called CEMP-r/s stars, which have before [Fe/H] reaches –3 dex. That is, our dynamical feedback between the star apparent contributions of both rapid (r) Boötes-I data provide direct evidence for ­formation in Boötes-I and its dark matter and slow (s) nucleosynthetic processes; two discrete channels of chemical enrich- halo can have occurred. Boötes-I is Beers & Christlieb, 2005). At lower values ment at very low iron abundances. indeed a surviving primordial system, of [Fe/H], excess carbon is commonly ideal to investigate the earliest stages of seen, but is inconsistent with AGB pro- With our current knowledge of stars in star formation, chemical enrichment and duction: rather the CEMP-no stars are Boötes-I, there is no direct chemical dark matter properties. more likely to have formed from gas ­evolution track (assuming standard yields enriched by non-standard supernovae of carbon and iron) between the CEMP- (such as “mixing and fallback” super­ no stars and the carbon-normal stars Acknowledgements novae), or by the winds from rapidly rotat- with [Fe/H] < –3. This means that at very Based on data obtained under ESO programmes: ing massive stars. In both cases the low iron abundance there is no one-to- P182.B-0372; P383.B-0038; P383.B-0093; supernova progenitors were massive one relationship between [Fe/H] and the P185.B-0946. Data-taking completed in P89. D. G. stars formed from primordial material — time since the first SNe. This conclusion & W. G. gratefully acknowledge support from the the first stars. is summarised in Figure 2, which shows Chilean BASAL Centro de Excelencia en Astrophys- ica y Technologias Afines (CATA) grant PFB-06/2007. the chemical evolutionary enrichment J. E. N. and D. Y. acknowledge support by Australian sequence deduced from our UVES study, Research Council grants DP0663562 and CEMP-no stars and its consistency with observations. DP0984924. R. F. G. W. acknowledges partial sup- The CEMP-no stars form rapidly out of port from the US National Science Foundation through grants AST-0908326 and CDI-1124403, and Our discovery of CEMP-no stars in the gas enriched by only one generation of thanks the Aspen Center for Physics (supported by two dwarf galaxies Segue 1 and Boötes-I SNe and most likely prior to the onset of NSF grant PHY-1066293) for hospitality while this is strong evidence for their self-enrichment effective mixing. This results in a small work was completed. from primordial material. The carbon mixing length, spatial inhomogeneity and over-abundance reflects the yields of the a large scatter in elemental abundance References very first generation of supernovae or ratios. massive stars. This provides an oppor­ Beers, T. C. & Christlieb, N. 2005, ARAA, 43, 531 tunity to consider the evolutionary history Such a scenario requires that the gas Belokurov, V. et al. 2006a, ApJ, 642, L137 Belokurov, V. et al. 2006b, ApJ, 647, L111 of the extremely carbon-enriched, iron- within which the CEMP-no stars form can Belokurov, V. et al. 2007a, ApJ, 654, 897 poor interstellar medium gas in these gal- cool and be locked up in low-mass stars Belokurov, V. et al. 2007b, ApJ, 658, 337 axies. very rapidly, and with high efficiency, so Belokurov, V. et al. 2008, ApJ, 686, L83 that material with this abundance pattern Caffau, E. et al. 2011, Nature, 477, 67 Gilmore, G. et al. 2007, ApJ, 663, 948 A key piece of information is that the is removed from the system at early times. Gilmore, G. et al. 2013, ApJ, 793, 61 most iron-poor star currently known This picture is consistent with models Ibata, R., Gilmore, G. & Irwin, M. 1994, Nature, in Boötes-I is not carbon-enhanced. of the formation of very metal-poor low- 370, 194 ­Carbon-enhanced and carbon-normal mass stars which appeal to enhanced Koposov, S. et al. 2011, ApJ, 736, 146 Norris, J. E. et al. 2010a, ApJ, 722, L104 stars co-exist at the same low iron cooling due to carbon. It may well be that Norris, J. E. et al. 2010b, ApJ, 711, 350 ­abundance within the same system (and the CEMP-no material resulted from a Norris, J. E. et al. 2013, ApJ, 762, 28 in the Galactic field Halo; c.f. Caffau et very small number of (Population III?) Simon, J. et al. 2011, ApJ, 733, 46 al., 2011). This provides direct evidence supernovae, possibly only one. that carbon enhancement is not required for very low-iron abundance gas to cool and form low-mass stars. The additional A surviving primordial galaxy information we consider here is that CEMP-no stars are not found at [Fe/H] Our metallicity and elemental abundance greater than –2.5, either in the field Halo data show that Boötes-I has evolved as or in dwarf spheroidal galaxies. a self-enriching star-forming system, from

28 The Messenger 151 – March 2013