Astronomical Science

New Surprises in Old Stellar Clusters

Ivo Saviane1 first light on 8 December 1908, and eight A few historical notes Enrico V. Held 2 years later it was used by Harlow Shapley Gary S. Da Costa 3 to publish the first paper of his series The practice of using GCs to test the Veronica Sommariva 4 “Studies Based On The Colors And Mag- resolving and light-collecting power of Marco Gullieuszik 2 nitudes In Stellar Clusters”. In the intro- new telescopes and instrumentation con- Beatriz Barbuy 5 duction, he recalled that, “No serious tinued until well after Shapley, and these Sergio Ortolani 6 attempts have been made to determine beautiful objects are still often featured accurate magnitudes, chiefly because in press releases today when new facili- of the lack of dependable magnitude ties are inaugurated. So it is not surpris- 1 ESO scales for the fainter .” Such attempts ing that just a handful of years after its 2 INAF – Osservatorio Astronomico di could finally be made thanks to the avail- first light in 1948, the 200-inch telescope Padova, Italy ability of that superb telescope, which at Palomar was used by Arp, Baum & 3 Australian National University, Mt. could collect photometric data of such Sandage (1953) to resolve the faint stars Stromlo Observatory, Australia distant objects as Galactic globular clus- defining the turnoff feature in the colour– 4 INAF – Osservatorio Astrofisico di ters (GCs). magnitude (CM) diagram of M92. It was ­Arcetri, Italy a breakthrough that suggested the con- 5 University of Sao Paulo, Brazil Among the motivations to undertake his nection between the main sequence 6 University of Padova, Italy work, Shapley quoted the possibility of in the CM diagram of young clusters, and solving the problem of “the order of stel- the later evolutionary phases seen in lar evolution; that is, the probable charac- diagrams. Galactic globular clusters have a long ter of the progression of spectral type and distinguished history as conspicu- (color) with age”. Almost one hundred The theoretical interpretation of such ous providers of simple populations years later, this motivation still underlies a ­diagrams came almost immediately with where models of low-mass stars can be large fraction of stellar cluster studies. the work of Hoyle and Schwarzschild, tested. However their simplicity has In particular globular clusters have been, who, in 1955, were able to calculate a stel- been challenged several times and a and still are, crucial in testing theories lar age of 6.2 Gyr by comparing the track

few peculiar objects have been identi- of low-mass stellar evolution, because of a 1.1 MA to the CM diagram of the fied. Only recently has the case for ex­­ their stars were born in a single episode, clusters M92 and M3. To be able to per- tended star formation histories become thus representing the simplest conceiv­ form the analytic calculations, some sim- really compelling, with the unprece- able population. Or at least this has been plifications had to be introduced, so the dented quantity and quality of data from the common wisdom for many years. imagers and But as observations and data analysis multiplexing spectrographs on large tel- techniques have improved, this view has Figure 1. The fuzzy object above the ESO 3.6-metre escopes. With ages close to that of started to change. This change has been telescope (and the NTT in the foregound) is , the most massive globular cluster in the the Universe, globular clusters can also helped in part by our work, as we illus- . This object was perhaps the nucleus of a help us uncover the earliest phases trate here, after setting things in context. dwarf that was disrupted several Gyr ago. in the formation of the Milky Way. In both areas, one of the major challenges is to have an abundance ranking of all clusters based on the same metallic- ity index, and for as many stars as ­possible inside each cluster; a homo­ geneous compilation is only available for about half of all globular clusters. A few years ago we began a project to help close this gap and we report on a few surprising results that are emerging.

In his book Reflecting Telescope Optics Ray Wilson (Wilson, 2004) writes that the 60-inch telescope on Mt. Wilson was, in its time, “arguably the greatest relative advance in astronomical observing potential ever achieved [together with W. Herschel’s 20-foot focus telescope]”, made possible by the technical genius of George Ritchey. The telescope had its

The Messenger 149 – September 2012 23 Astronomical Science Saviane I. et al., New Surprises in Old Stellar Clusters

track was only in qualitative agreement Table 1. Major metallicity compilations. For each reference, NP and NC are the number of programme clus- with the observations. However com­ ters, and the number of clusters with [Fe/H] homogenised on the same scale. The fifth column lists whether (r)esolved stars or (i)ntegrated light was studied. puters soon became powerful enough to allow numerical calculations, which Reference Acronym NP NC i/r Method were pioneered by Icko Iben and collabo- Searle, L. & Zinn, R. 1978, ApJ, SZ78 19 – r Mean of reddening-free rators. Iben & Rood (1970) were able 225, 357 magnitudes for seven bands Zinn, R. 1980, ApJS, 42, 19 Z80 79 84 i Reddening-independent colour to follow the evolution of metal-poor stars quotient Q39 for objects of different mass, metallicity, Zinn, R. & West, M. J. 1984, ApJS, ZW84 60 121 i EW of Ca ii K line, G-band, Mg i and helium abundance, thus yielding 55, 45 triplet isochrones that could be compared Armandroff, T. E. & Zinn, R. 1988, AZ88 27 – i EW of Ca ii infrared (IR) lines AJ, 96, 92 quantitatively to observed CM diagrams. Rutledge, G. A. et al. 1997, PASP, RHS97 52 70 r EW of Ca ii IR lines The comparison exercise was immedi- 109, 907 ately performed by Sandage (1970) with Carretta, E. et al. 2009, A&A, C09 19 133 r EW of Fe lines in medium-­ four globular clusters, and his investiga- 508, 695 resolution spectra Harris, W. E. 2010, arXiv, 1012.3224 H10 157 – – Average of literature tion set the foundations for the classic research lines that extend into our times. controversy arose in considerations of which find that up to 80% of the Milky Based on the colour and luminosity of the whether there was an age spread, or a Way could have been formed in situ (e.g., clusters’ turnoffs, Sandage established majority of coeval clusters plus a few de Rossi et al., 2009). an average age of 11.5 Gyr, and cluster- younger ones with possible extragalactic to-cluster age differences formally not origin. To settle the question, a large, Notwithstanding the success of relative greater than ~ 2% (although photometric homogeneous and high quality photo- age studies, they still suffer from one errors allowed an age spread as large metric database of CM diagrams was major problem: in addition to its age, the as one Gyr). The data were not in conflict needed, a task that was becoming possi- luminosity and colour of a star at the with contemporary estimates of the ble at that time thanks to the introduc- turnoff depend on its metallicity, so a key ­Hubble time, and they were also consist- tion of charge coupled devices (CCDs) as input to relative age studies is a metallic- ent with the rapid-collapse (~ 108 yr) for- detectors into astronomical instrumen­ ity compilation that is also homogeneous. mation of the Milky Way halo, as pro- tation. Rosenberg et al. (1999) could take their posed by Eggen, Lynden-Bell & Sandage [Fe/H] values from Rutledge et al. (1997), (1962). Globular clusters had thus ac­­ With these detectors, errors better than which was based on homogeneous quired a prominent role both for cosmol- 0.01 magnitudes could be achieved measurements of the equivalent widths ogy and for theories of galaxy formation. at the cluster turnoff, which opened the (EW) of the Ca ii infrared triplet lines. The new perspective also brought ques- way to comparing the position of that The photometric data were collected in tions about cluster and star formation ­feature — relative to the red giant branch the V- and I-bands because it had been in the primordial Universe, and about sur- or the horizontal branch — in many shown empirically that the V−I colour vival mechanisms against internal and ­different clusters, with sufficient accuracy ­difference is much less sensitive to [Fe/H] external disruption processes. to detect age differences of the order than the B−V one (Saviane et al., 1997). of 0.5 Gyr. Because of the small format of However homogeneous [Fe/H] values are those early CCD chips, the fast, post- available for only a fraction of clusters, a A uniform database of turnoff phases of stellar evolution (the fact that was one of the main motivations horizontal branch in particular) could not for our project. The determination of absolute ages be well sampled, so the photometric col- requires knowledge of cluster distances, our of the turnoff relative to the giant The current state of [Fe/H] data is sum- which are difficult to obtain because branch became the easiest parameter to marised in Figure 2 and in Table 1. The of the paucity in the solar vicinity of pop- be measured. The combination of all largest homogeneous spectroscopic ulation ii standard candles that can be these ideas was behind the Rosenberg et sample of individual stars is still that of calibrated via parallax. Relative ages are al. (1999) study, which was based on a Rutledge et al. (1997) and if we add the comparatively easier to obtain, and there- homogeneous photometric database for clusters of Saviane et al. (2012) then we fore many studies after Sandage tried to 35 clusters, assembled with the 91-centi- find that ~ 55% of objects have a metal- uncover the age distribution of Galactic metre Dutch telescope at La Silla and licity measurement based on the same globular clusters, with mixed results. The the 1-metre Johannes Kapteyn Telescope index (the “reduced” equivalent width of situation was reviewed by Stetson, van (JKT) at La Palma. The conclusion was the Ca ii infrared triplet). We then have den Bergh & Bolte (1996), who concluded that most globular clusters were formed to move to integrated-light spectroscopy, that “as of the current date the state of in a single epoch, but a few objects were to the samples of ZW84 and AZ88 (see the field is still somewhat muddled”, but clearly formed at later stages. Meanwhile Table 1). They are based on the equiva- also hoped that “data now being col- Ortolani et al. (1995) had also found that lent widths of various metallic lines (Ca ii lected by numerous groups in various the Bulge is coeval with the oldest Halo K line, G-band and Mg i triplet), and of sub-disciplines may resolve the remain- clusters. This is consistent with recent the Ca ii infrared triplet, respectively. After ing controversy within a few years”. The simulations of the formation of the Galaxy excluding objects in common with the

24 The Messenger 149 – September 2012 Figure 2. Sources of metallicity for the bility of FORS2: the maximum slit density 157 globular clusters included in the can be reached with the mask exchange Harris (2010) catalogue. Note that more clusters are being discovered by unit, however the lengthy operations of the SDSS and VISTA/VVV surveys. mask manufacture and insertion meant The acronyms are given in Table 1, and that only a few objects per night could in addition: MR = metallicity indices have been observed. Therefore we opted based on medium resolution spectros- copy; HR = [Fe/H] from high resolution for the configurable slits solution, and studies; CM = broadband imaging (CM left the masks for the most compact diagram). The four colour groups indi- objects. cate homogeneous resolved MR spec- troscopy (green shades), homogene- ous integrated light MR spectroscopy All data were reduced with the FORS2 (blue shades), individual resolved MR pipeline (Izzo et al., 2010) which delivers or HR studies (yellow shades), and no wavelength-calibrated, sky­-subtracted spectroscopy or no data (red shades). spectra in an extremely efficient manner; Five clusters in common between AZ88 and ZW84 have been assigned organising the data took more time than to AZ88. Four of the AZ88 and one of reducing them. On the other hand be- the ZW84 clusters have [Fe/H] values cause of the high crowding typical of measured with high-resolution spectra globular clusters, often more than one as well. star per slit was extracted, so a major RHS S12 AZ88 AZ84 part of the work was to cross-check the MR HR CM No data identification of the ~ 600 spectra. Clus- ter members were identified as objects cited studies, these two datasets com- When the first CM diagrams of main with both radial velocity and reduced prise 8% and 12% of clusters, respec- sequence stars based on CCD imaging equivalent widths not deviating signifi- tively. No other homogeneous datasets appeared, they showed a virtually zero- cantly from the average, taking into exist: 12% of clusters have medium or width locus, confirming the visual impres- account that uncertainties in our mean high-resolution spectroscopic data of sion of a smooth, single-age stellar popu- velocities are of order 5–6 km/s. The resolved stars from a variety of sources, lation. However as more and more data ­fraction of stars that were eventually con- and for another ~ 8% the metallicity accumulated, things started to look less firmed as members varies significantly ­estimate comes from their CM diagram. simple. In particular, abundance from cluster to cluster, with a median Finally there are a handful of objects for ­variations have been found for light ele- value of 53% and being always better which no [Fe/H] data exist. ments in all clusters that have been than 20%. Especially for bulge clusters searched so far (see the very recent where field contamination is high, the Rutledge et al. (1997) obtained their spec- review of Gratton et al. [2012]). In addi- catalogues of member stars for follow-up tra with the modest 2.5-metre Dupont tion, a spread in one or more of iron- high-resolution studies is another useful ­telescope at Las Campanas, so homoge- peak, n-capture, and α elements has also byproduct of our work. neous [Fe/H] data are missing mostly been found for ω Cen, M54, M22, for outer halo or heavily extincted clus- NGC 1851, NGC 2419, ­Terzan 5 and The equivalent widths of calcium lines in ters. As relative age studies include more NGC 5824. This is particularly interesting red giant stars depend on the Ca abun- and more outer halo objects, we are because it is direct evidence for ex- dance and also on their luminosity, so forced to take metallicities from a variety tended star formation in these clusters. this dependency was removed by adding of sources. For example Marín-Franch et In three of these seven objects the a linear term in V−VHB and obtaining the al. (2009) based their study on the largest spread was discovered or confirmed from “reduced” equivalent width. A calibration homogeneous photometric sample of our FORS2 project. relation from such reduced equivalent Galactic globular clusters (64 objects), widths to the scale (Carretta et al., 2009) but for as many as 25% of them they had was then obtained, with 14 clusters hav- to estimate [Fe/H] from the Zinn & West Improving the situation ing well-determined [Fe/H] values by (1984) compilation of the reddening-inde- high-resolution studies. It was then pendent Q39 photometric index. To sum- In May 2006 we had two observing applied to convert the reduced equivalent marise, for almost half of GCs, the metal- nights assigned to us at the VLT, to col- widths of programme clusters into metal- licity values are based on data that are lect medium-resolution FORS2 spectra licity values. These new [Fe/H] values either not spectroscopic, not homogene- for globular clusters whose distance are significantly different from literature ous or not of individual stars. Further- modulus is too large for high-resolution values for about half of our programme more, even when all these conditions are studies, either because of distance or clusters, as graphically illustrated by satisfied, the number of stars measured high extinction. The weather did not help ­Figure 3, which shows a type of diagram might be too small to look for metallicity us very much, so of the 49 planned tar- first introduced by Zinn (1993). Our new dispersions, precisely the area where gets we could observe only twenty, plus average abundances are lower than liter- globular clusters started to show the first eight calibrators. To increase observing ature values for six clusters (Pyxis, Terzan surprises. efficiency we used the multi-object capa- 3, HP1, NGC 7006, NGC 6569 and

The Messenger 149 – September 2012 25 Astronomical Science Saviane I. et al., New Surprises in Old Stellar Clusters

Figure 3. The metallicity of our pro- 3DQ  -&" gramme clusters on the Carretta et al. -&" (2009) scale is plotted here against -&" horizontal branch (HB) type. Isochro- -&" nes are from Rey et al. (2001), and l  are separated by 1.1 Gyr, with age +XMF@ decreasing from top to bottom. The -&" -&" oldest isochrone gives the age of clus- -&" '/ l ters at R < 8 kpc (Rey et al., 2001). 3DQ -&" The arrows connect the position of the /XWHR cluster, if [Fe/H] from Harris (2010) %D' -&" is used, to the position given by our 1TO -&" l  , metallicity value. -&"

l -&"

l  l l      !l1! 5 1

Figure 4. The 216 useful spectra of NGC 5824 and foreground stars taken with FORS2 in the calcium triplet region (840–870 nm). The three most promi- nent lines belong to the Ca ii ion, with wavelengths of 849.8, 854.2 and 866.2 nm.

26 The Messenger 149 – September 2012 NGC 6715), so to retain their horizontal could only have been achieved with of this process have been erased in the branch (HB) morphology they must be the Hubble Space Telescope (HST). By inner parts of the Galaxy, in its halo younger. The higher abundances of measuring stellar magnitudes on archival where dynamical times are long, several Lynga 7, NGC 6558 and NGC 6380 sug- Wide Field Planetary Camera 2 (WFPC2) stellar streams have been identified in gest instead older ages. Zinn had intro- images with his specialised software, the past. There is however still an order of duced the diagram of Figure 3 to see Anderson (2002) discovered that the magnitude discrepancy between the whether the age distribution of Galactic main sequence of ω Cen is split in two. large number of predicted satellites in the clusters would support the extended Multiple stellar populations were subse- Milky Way and the observed ones. halo formation proposed by Searle & Zinn quently discovered in many other clusters, Any addition of a new remnant is there- (1978). In this scenario, our [Fe/H] revi- including those having metallicity disper- fore important to confirm the paradigm. sion would bring more clusters into the sions. The latter are also among the “young halo” class, however the HB mor- most massive in the Galaxy, and some of Under the working hypothesis that clus- phology depends also on the mass loss them appear to be associated with stellar ters with metallicity spreads signpost along the red giant branch (RGB) and streams: M54 (MV = −10.0) is clearly the halo substructures, we have been the helium abundance, so the conclusion nuclear of the ­Sagittarius assigned more than 13 hours of addi- will have to be confirmed with precise dwarf galaxy, which is currently being tid- tional FORS2 time in period 89 to extend turnoff photometry. ally disrupted by the Milky Way, and our sample, and potentially to find still

NGC 1851 (MV = −8.3) is surrounded by more unknown examples of these pecu- The second finding of our project is that, an extensive stellar halo that may have liar stellar systems. for three clusters, the dispersion of resulted from the destruction of the dwarf reduced equivalent widths is larger than galaxy in which the cluster was once the measurement error. We thus con- embedded (Olszewski et al., 2009). Acknowledgements firmed the [Fe/H] dispersion of M54 and The completion of this project was significantly fos- discovered a metallicity dispersion in Theoreticians have been able to simulate tered by visits of two members of the collaboration M22 (in parallel with Marino et al. [2009]) ω Cen as the nuclear remnant of a dis- to Padova and ESO. We warmly thank the ESO and probably in NGC 5824. The metallic- rupted dwarf galaxy (Bekki & Freeman, Director General Discretionary Fund for supporting ity distribution of M22 (NGC 6656) has 2003) and Newberg et al. (2009) have them. been discussed in Da Costa et al. (2009), suggested that NGC 5824 (MV = −8.8) where it was found to range from −2.2 may be associated with the newly dis­ References to −1.2 dex, and to share many proper- covered Cetus Polar Stream (CPS). The ties with those of Cen, including a fast nuclear cluster scenario is quite appeal- Anderson, J. 2002, ASP Conf. Series, 265, eds. ω F. van Leeuwen, J. D. Hughes, G. Piotto, 87 rise to a peak and a broad tail to higher ing because it would offer an explanation Arp, H. C., Baum, W. A. & Sandage, A. R. 1953, AJ, abundances. M22 could be well charac- for the peculiarities illustrated above. If 58, 4 terised because three FORS2 masks a cluster is located in a deep potential Bekki, K. & Freeman, K. C. 2003, MNRAS, 346, L11 were dedicated to this cluster (41 mem- well inside a dwarf galaxy, it might be able Da Costa, G. S. et al. 2009, ApJ, 705, 1481 de Rossi, M. E. et al. 2009, MNRAS, 395, 210 ber stars), while the small number of to retain or to accrete gas, thus forming Eggen, O. J., Lynden-Bell, D. & Sandage, A. R. 1962, stars did not allow us to perform a similar multiple generations of stars or develop- ApJ, 136, 748 detailed analysis for NGC 5824; we could ing chemical anomalies. It is also interest- Gratton, R. G., Carretta, E. & Bragaglia, A. 2012, only estimate a value ≈ 0.11–0.14 dex for ing to note that both M54 and NGC 5824 A&A Rev., 20, 50 Izzo, C. et al. 2010, Proc. SPIE, 7737, 773729 its dispersion. However with observations lie on the same young isochrone in Fig- Marín-Franch, A. et al. 2009, ApJ, 694, 1498 carried out both at Gemini and at the ure 3, and that stars in the nuclei of early- Marino, A. F. et al. 2009, A&A, 505, 1099 VLT in period 87, we were able to collect type dwarf tend to be younger Monaco, L. et al. 2009, A&A, 502, L9 spectra for a couple of hundred stars, of than the underlying population (e.g., Newberg, H. J., Yanny, B. & Willett, B. A. 2009, ApJ, 700, L61 which more than a hundred were con- Monaco et al. 2009; Paudel et al., 2011). Olszewski, E. W. et al. 2009, AJ, 138, 1570 firmed as cluster members. The data It should be added that metallicity spreads Ortolani, S. et al. 1995, Nature, 377, 701 analysis is in progress, but the new data might be explained by other mechanisms Paudel, S., Lisker, T. & Kuntschner, H. 2011, MNRAS, seem to confirm the presence of a met­ (e.g., merging clusters with different 413, 1764 Rey, S.-C. et al. 2001, AJ, 122, 3219 allicity spread. Figure 4 shows a montage­ [Fe/H]), and for these we refer to Gratton Rosenberg, A. et al. 1999, AJ, 118, 2306 of all useful spectra extracted with the et al. (2012). Sandage, A. 1970, ApJ, 162, 841 FORS2 pipeline. Saviane, I. et al. 2012, A&A, 540, A27 In a bigger picture, finding evidence of Saviane, I., Rosenberg, A. & Piotto, G. 1997, in Stellar Ecology: Advances in Stellar Evolution, disrupting satellites in the eds. R. T. Rood & A. Renzini, 65 Outlook has great significance for the current Searle, L. & Zinn, R. 1978, ApJ, 225, 357 Lambda Cold Dark Matter galaxy forma- Stetson, P. B., van den Bergh, S. & Bolte, M. 1996, The discovery of metallicity dispersions tion paradigm. In this scheme, the Gal- PASP, 108, 560 Wilson, R. N. 2004, Reflecting Telescope Optics .I was greatly advanced by the advent of axy is built up through the merger and Basic Design Theory and its Historical Develop- multiplexing spectrographs at 8–10-metre accretion of lower mass systems, pre- ment, 2nd edn. (Springer) class telescopes, but the other important dominantly at early epochs. Indeed, while breakthrough in globular cluster research the spatial and kinematic signatures

The Messenger 149 – September 2012 27