Stellar populations in star clusters
RRAA
Vol. No. , R esearch in A stronomy and A strophysics Key words: galaxies: star clusters: general — galaxies: star formation — stars: rotation
Stellar populations in star clusters
Cheng-Yuan Li , , Richard de Grijs , and Li-Cai Deng Department of Physics and Astronomy, Macquarie University, North Ryde, NSW 2109,Australia; [email protected] Purple Mountain Observatory, Chinese Academy of Sciences, Nanjing 210008, China Kavli Institute for Astronomy & Astrophysics and Department of Astronomy, Peking University,Yi He Yuan Lu 5, Hai Dian District, Beijing 100871, China International Space Science Institute–Beijing, 1 Nanertiao, Zhongguancun, Hai Dian District,Beijing 100190, China Key Laboratory for Optical Astronomy, National Astronomical Observatories, Chinese Academyof Sciences, 20A Datun Road, Chaoyang District, Beijing 100012, China
Abstract
Stellar populations contain the most important information about star clus-ter formation and evolution. Until several decades ago, star clusters were believed tobe ideal laboratories for studies of simple stellar populations (SSPs). However, dis-coveries of multiple stellar populations in Galactic globular clusters have expandedour view on stellar populations in star clusters. They have simultaneously generateda number of controversies, particularly as to whether young star clusters may havethe same origin as old globular clusters. In addition, extensive studies have revealedthat the SSP scenario does not seem to hold for some intermediate-age and young starclusters either, thus making the origin of multiple stellar populations in star clusterseven more complicated. Stellar population anomalies in numerous star clusters arewell-documented, implying that the notion of star clusters as true SSPs faces seriouschallenges. In this review, we focus on stellar populations in massive clusters with dif-ferent ages. We present the history and progress of research in this active field, as wellas some of the most recent improvements, including observational results and scenar-ios that have been proposed to explain the observations. Although our current abilityto determine the origin of multiple stellar populations in star clusters is unsatisfac-tory, we propose a number of promising projects that may contribute to a significantlyimproved understanding of this subject.
Star clusters are the basic units of star formation (Lada & Lada 2003): almost all stars form in clus-tered environments. Current consensus on the formation of star clusters suggests that most stars formtracing the turbulent structure of the interstellar medium and in an initially supervirial state. Withinan extremely short period (about one crossing time), the initially turbulent, ‘fractal’ structures willcollapse into bound clusters (Bonnell et al. 2008; Allison et al. 2009; Girichidis et al. 2012; Moeckel& Burkert 2015). Subsequently, a large proportion will gradually dissipate into the galactic field a r X i v : . [ a s t r o - ph . S R ] S e p C.-Y Li et al. (de Grijs 2010). Understanding the stellar populations of star clusters is, therefore, of fundamentalimportance for understanding many astrophysical processes, including star cluster formation andevolution, the chemical evolution of Galactic stellar populations, as well as the stellar dynamics instar clusters.The ‘simple stellar population’ (SSP) scenario is the assumption that stars in a star cluster alloriginate from a common progenitor giant molecular cloud (GMC), during the same era, and thusthey would share similar metallicities. It has been confirmed that the initial star-forming process instar clusters approximately resembles a single burst (Cabrera-Ziri et al. 2014). The combination ofgas expulsion owing to energetic photons ejected by the most massive first-generation stars and thestrong stellar winds triggered by the first batch of Type II supernovae will quickly exhaust all ofthe gas in the GMC, thus quenching the star- and cluster-forming process (Bastian & Strader 2014).The nature of most open clusters (OCs) and young massive clusters (YMCs) has been confirmed asresembling SSPs.During the last few decades, observations have revealed the presence of multiple stellar popula-tions in Galactic globular clusters (GCs). The observational evidence can be classified into photomet-ric and spectroscopic evidence. The former refers to the fact that the photometric color–magnitudediagrams (CMDs) of some GCs display multiple distinct features in or along their main sequences(MSs; e.g., NGC 2808: Piotto et al. 2007), their subgiant branches (SGBs; e.g., 47 Tuc: Anderson etal. 2009), their red-giant branches (RGBs; e.g., NGC 288: Piotto et al. 2013), or even in their hori-zontal branches (HBs; e.g., NGC 2808: Bedin et al. 2000). Sometimes, individual GCs even displaya combination of such multiple features. In the photometric CMDs, these features can be explainedas the result of a diversity of ages, helium abundances, and metallicities. Since SSP stars formed atapproximately the same time from a common molecular cloud, variations in age, helium, and metalabundance hence unambiguously reflect the occurrence of more than a single star-forming episodeduring their host clusters’ evolution.GCs have also been subject to intense scrutiny based on spectroscopic analyses, which have re-vealed star-to-star chemical dispersions of specific elements. Cohen et al. (1978) first noted that theNa abundance variations of RGB stars in the GCs M3 and M13 exceeded the observational errors.Her pioneering work has stood the test of time: subsequent measurements of Galactic GCs haveuncovered the well-known Na–O anticorrelation, where oxygen-depleted stars have higher sodiumabundance (Carretta et al. 2004a; Gratton et al. 2004; Carretta et al. 2009). Another well-studiedGalactic GC relationship is the Mg–Al anticorrelation, which is most easily seen in intermediate-metallicity clusters (e.g., M13; Shetrone 1996). The large scatter in the abundances of some specificelements is not expected if these GCs were SSPs. A straightforward explanation of the scatter prop-erties is that there has been secondary star formation, fueled by abundance-enhanced material. In thisreview we will not discuss the chemical evolution of stellar populations in depth, since this aspectwas introduced in great detail in the review of Gratton et al. (2004).A key problem associated with the stellar populations in star clusters relates to an importantopen question: do GCs have the same origin as OCs and YMCs? It seems that the latter, very youngobjects may encounter significant difficulties to survive to the cosmic age ( ≥
10 Gyr) owing toa range of effects, including internal two-body relaxation, which causes most stars to ‘evaporate’from the cluster (Spitzer 1987). Only those clusters that initially contained at least 10 stars canavoid disruption within 10 Gyr (Portegies Zwart et al. 2010). On the other hand, in terms of theirinitial masses, only young star clusters with masses between M (cid:12) and M (cid:12) have the capacityto capture their initial runaway gas. Because the subsequent Type II supernovae explosions wouldfurther accelerate the runway gas flows, the mass threshold for a YMC to capture its initial runawaygas would increase dramatically. If a star cluster can not marshall additional gas reserves, its star-forming process will cease rapidly. As regards the stellar populations in star clusters, almost allobserved OCs and YMCs will fail to generate multiple stellar populations, which, however, is acommon feature of most observed GCs (see Li 2015, his Chapter 4). tellar populations in star clusters 3 This review is organized as follows. In Section 2 we show that the SSP approximation is nota tentative assumption but based on convincing evidence. Readers will realize why the presence ofmultiple stellar populations in GCs (and also in some young and intermediate-age clusters) is a prob-lem. In Section 3 we discuss observational results of stellar populations in star clusters with differentages. In Section 4 we compare and contrast currently popular scenarios that aim at explaining theorigin of multiple stellar populations; their advantages and disadvantages are also discussed. We nextdiscuss, in Section 5, how one can go about examining some of these compelling and controversialscenarios. A brief summary is given in the concluding section (Section 6).
The first issue confronting any fundamental discussion of stellar populations is the star-formationmode in star clusters. Theoretically, except for the most massive and most compact star clusters,stars in a star cluster should form in single-burst mode: the star-formation process in a star clusterwill cease rapidly after formation of the first-generation stars owing to the quick exhaustion of theinitial gas, which is mainly caused by stellar wind-induced mass loss (e.g., energetic photons ejectedby the most massive stars; Longmore et al. 2014) and supernova explosions. The typical escapevelocity should be comparable to the sound speed in ionized hydrogen (H II ) gas ( ∼
10 km s − ;Banerjee & Kroupa 2015). For some massive clusters, gas expulsion may be dominated by radiationpressure, and therefore their escape velocities could even exceed 10 km s − (Krumholz & Matzner2009). The timescale of gas expulsion driven by these initial stellar winds is very short, usually lessthan or comparable to the cluster’s crossing time ( ∼ ≤ ∼ M (cid:12) to 25 M (cid:12) ) evolve off the MS. A15 M (cid:12) star could eject 3 × erg into the interstellar medium over the period of 0.1 Myr, which issufficient to unbind 10 M (cid:12) gas within a 1 pc region (Baumgardt & Kroupa 2007). Therefore, a largefraction of residual gas in a cluster with an age between 3 Myr an 10 Myr is not expected. Duringthis phase, Type II supernova explosions will occur frequently. These multiple supernova explosionswill be the ‘death blow’ to any residual gas remaining in the cluster. Supernova explosions will expelall stellar material at a velocity of up to × km s − (10% of the speed of light), driving a shockwave through the interstellar medium of their host star clusters. Such a shock wave will accelerateall remaining gas to a velocity of several 100 km s − . Calculations have shown that only clustersthat have formed from the initial condensations with masses of 10 M (cid:12) to 10 M (cid:12) can survive thefirst series of supernova explosions (Shustov & Wiebe 2000).Because the initial star-formation process ceases very quickly, the age range of the first-generation stars would thus be constrained to a timescale of several million years. This timescale,compared with the typical ages of most GCs ( ∼
10 Gyr) and YMCs (10 Myr – 100 Myr), is indeedvery short.Stars that are less massive than ∼ M (cid:12) will undergo the post-RGB and asymptotic giant branch(AGB) phases, when they evolve to their final evolutionary stages. The ages of their host star clus-ters must be at least
30 Myr, which is already equivalent to the typical age of most YMCs. Theseintermediate-mass AGB stars will deposit most of their stellar material into the interstellar medium,which is thought to be important for secondary star formation: ejected stellar material from AGBstellar winds forms the building materials of newly born stars, possibly leading to the formation ofmultiple stellar populations with enhanced chemical abundances, at least compared with the initialstellar generation in the cluster. Many models aim at explaining the observed multiple-stellar popu-lations in GCs are based on this type of ‘self-enrichment’ scenario (i.e., Ventura & D’Antona 2009;Valcarce & Catelan 2011). However, since in most GCs the observed secondary stellar generation
C.-Y Li et al. has a comparable total mass to the first stellar generation, to explain the observed high fractionsof secondary stellar generations, these scenarios have to assume extremely high cluster masses atearly times (about 10 to 100 times their current masses). In addition, the velocity of the AGB stellarwinds is comparable to the initial O-type stars’ stellar winds, which are ∼
10 km s − to 100 km s − .Because initial gas expulsion will cause expansion of the cluster to a less compact configuration andleads to the loss of a fraction of the initial cluster mass, an exposed cluster is unlikely able to retainthese intermediate-mass AGB stellar winds.Mass loss caused by evaporation will further complicate the estimates (McLaughlin & Fall2008). The amount of mass lost through this mechanism can be calculated as ˙ m ≈ (cid:18) ρ h M (cid:12) pc − (cid:19) − / M (cid:12) Gyr − , (1)where ρ h is the gas density inside the cluster’s half-mass radius, r h . Assuming a typical GC age of12 Gyr, most GCs should initially have been 10 to 15 times more massive than their current mass,indicating a strong capacity to retain their initial runaway gas. However, for OCs and YMCs, mostof which are younger than ∼
100 Myr, the expected amount of mass loss is much smaller. Therefore,their initial masses do not reach the mass threshold for capturing runaway gas.Assuming that the average half-mass radius of young clusters is ∼ v esc ( t ) ≈ . (cid:115) M cl ( t ) r h ( t ) km s − . (2)It follows immediately that only for clusters with masses of M cl = 10 M (cid:12) to M (cid:12) , the initialrunaway stellar material will be retained in their gravitational potential wells. In Fig. 1 we presentthe minimum mass for clusters to retain their initial runaway material as a function of their currentmass for two different escape velocities: 10 km s − corresponds to the gas-expulsion velocity andthe slowest AGB ejecta (top panel), while 100 km s − corresponds to the lowest escape velocitydriven by Type II supernovae (bottom panel). Figure 1 shows that the OCs will fail to capture theirinitial runaway gas in both cases, while about half of the observed YMCs in the Milky Way will failto capture the slowest AGB ejecta. For an escape velocity of 100 km s − , all YMCs will lose theirinitial gas. This also holds for most GCs. Note that the escape velocity of 100 km s − is only a lowerlimit to the velocities generated by supernova explosions, so our estimates are conservative.This is why star clusters are thought to approximate SSPs: during the initial phase ( ≤ ≤ tellar populations in star clusters 5 log M tot [ M ⊙ ] l og M c r i [ M ⊙ ] l og M c r i [ M ⊙ ] log M tot [ M ⊙ ] OCsYMCsGCsCritical line
Fig. 1
Critical mass, log M cri , to capture the initial stellar winds of AGB stars for starclusters as a function of their actual mass, log M tot − ; (bottom) 100 km s − .Davidge 2012; Lima et al. 2014), are indeed observed. Combined with observations of YMCs withages of ∼
100 Myr, several lines of evidence imply that there will be little residual gas for starformation in YMCs after the gas-expulsion period (Bastian et al. 2013; Bastian & Strader 2014;Cabrera-Ziri et al. 2015).As discussed in Section 1, star clusters will lose stars through dynamical two-body relaxation.This disruption process is further accelerated by external tidal shocks. As a consequence, a star clus-ter’s mass will continue to decrease during its entire evolution (if no additional merger or accretion
C.-Y Li et al. events occur), which means that clusters can retain their initial gas only during the early stages oftheir evolution. However, detailed studies have confirmed that OCs and YMCs are SSPs, with rareexceptions, such as the cluster NGC 6791 (Geisler et al. 2012).The ubiquitous multiple stellar populations in GCs immediately pose a conundrum: How dothese clusters trigger secondary star formation if most of their initial material would have been lostat an early stage? In fact, this issue relates to another key open question: Do GCs have the same originas young star clusters? In Section 3 we introduce some relevant background to acquaint readers withthe details of stellar populations in various types of star clusters.
Spectroscopic analyses investigating star-to-star chemical variations provide evidence that OCs arechemically homogeneous (e.g., Shen et al. 2005; D’Orazi & Randich 2009; Ting et al. 2012)—thisseems to also hold for old OCs like NGC 188, see Norris & Smith (1985)—thus confirming thattheir stars indeed formed coevally from a common GMC. This conclusion is also valid for stellarassociations (e.g., De Silva et al. 2013) and star-forming regions (e.g., Nieva & Sim´on-D´ıaz 2011),indicating a high degree of chemical homogeneity in the primordial GMCs.The most direct method to examine if a group of stars in a star cluster are coeval is by ex-ploring their CMD, since the CMD of an SSP can be described very well by a single isochrone.However, because of Galactic extinction, OC CMDs are usually messy, which renders explorationsof the age distributions of their member stars difficult. Currently, measuring a single star’s age is stillchallenging. No single method works well for a broad range of stellar types or for the full range inages, but this may change in the next few years with improvements resulting from the technique ofasteroseismology (Soderblom 2010).Another promising approach to studying stellar ages is by measuring their surface lithium abun-dances, since the lithium abundance is suggested to decrease with increasing stellar age. Studiesbased on the Li clock surprisingly reveal that the ages of some stars seem to exceed their host OCs’isochronal ages (Duncan & Jones 1983; Palla et al. 2007). However, it is unclear whether theseexceptions can be made to resemble large-scale ongoing star formation. For instance, Sacco et al.(2007) showed that only three of their 59 stars analyzed in the young σ Orionis cluster show Lidepletion, and among the former only one star has a nuclear age that exceeds the isochronal age. Biket al. (2014) analyzed the ages of the most massive O-type stars in the extremely young, massivecluster W3 Main (with an age of approximately 3–5 Myr). They found that star formation in W3Main has lasted for about 2–3 Myr and is still ongoing.In addition to the Li clock, which only works for individual stars, exploring the luminositydistribution of main-sequence turn-off (MSTO) stars provides an independent approach to test theSSP scenario. Although Galactic extinction makes studying the intrinsic magnitude spread of MSTOstars in OCs difficult, some exceptions exhibit robust statistical results. Based on a comparison ofstellar isochrones and the observed stellar distribution in the CMD, Eggen (1998) found that theage ranges of the Hyades and Praesepe OCs seem to span several hundred million years, as impliedby the apparent luminosity spread of MSTO stars. A similar luminosity dispersion has also beenfound in the Orion Nebula Cluster, for which an internal age spread on the order of 10 Myr wasinferred (Palla & Stahler 2000). More recent studies have found that NGC 1856, a 300 Myr-oldLarge Magellanic Cloud (LMC) cluster, displays an apparently extended MSTO (eMSTO) region(Milone et al. 2015a), which may indicate prolonged star formation, lasting ∼
150 Myr. A recentstudy found that the 150 Myr-old YMC NGC 1850 also seems to harbor an eMSTO region (Bastianet al. 2016). Because both NGC 1850 and NGC 1856 reside in the LMC, they are not affected bysignificant extinction due to dust in the Galactic disk. The morphologies of the eMSTO regions in tellar populations in star clusters 7 these YMCs may be a ‘smoking gun,’ suggesting that stars in star clusters may form continuouslyrather than in a single starburst event.Such eMSTOs seem to be ordinary features of intermediate-age (1–2 Gyr-old) star clusters in theLMC and the Small Magellanic Cloud (SMC). For instance, Mackey & Broby Nielsen (2007) foundthat the LMC cluster NGC 1846 shows a split in its MSTO region, which can be well describedby two isochrones with ages of 1.5 Gyr and 1.8 Gyr. Similar results have been found for the LMCclusters NGC 1783 and NGC 1806, with indications that their age dispersion may be at least 300 Myr(Mackey et al. 2008). Milone et al. (2009) studied 16 intermediate-age LMC clusters by exploringtheir MSTO regions, concluding that 70% have MSTOs that are inconsistent with the expectationsfrom SSPs. Girardi et al. (2013) studied the SMC clusters NGC 411 and NGC 419. They foundthat their eMSTO regions are consistent with age spreads of ∼
700 Myr, which may be the largestapparent age spreads known among all intermediate-age star clusters (but see Wu et al. 2016). Similarresults have also been found by Rubele et al. (2010, 2011). Girardi et al. (2009) even claimed thepresence of dual red clumps (RCs) in NGC 419. Li et al. (2014a) analyzed the MSTO regions in theLMC clusters NGC 1831 and NGC 1868. They found that their implied internal age spreads are 280Myr and 320 Myr, respectively, although the compact RC of NGC 1831 is inconsistent with such alarge age spread.The discovery of eMSTO regions in intermediate-age star clusters strongly challenges our un-derstanding of star cluster formation and evolution. Various scenarios have been proposed to ex-plain the observations. Goudfrooij et al. (2011, 2014) found a correlation between the widths ofthe MSTO regions and the early escape velocities in intermediate-age star clusters. They claimeda velocity threshold of 12–15 km s − for all clusters featuring eMSTOs, which is consistent withthe escape velocity of gas expulsion and slow AGB stellar winds. This implies that the gas used forsecondary star formation may have originated from intermediate-mass AGB stars or massive binarystars. However, since most of these star clusters are more massive than 10 M (cid:12) , their initial stellarpopulations should have contained many massive OB stars. It is then concerning that these OB starsdid not expel the interstellar medium at very early stages through stellar feedback and supernovaexplosions. This scenario was challenged by Cabrera-Ziri et al. (2015), who carefully studied thestellar populations of NGC 7252-W3, a possible YMC candidate with an age of ∼
570 Myr and anescape velocity of 193 km s − . However, no evidence of an extended star-formation history (eSFH)was found. Bastian & Strader (2014) searched for residual gas in 13 LMC and SMC YMCs withages ≤
300 Myr. Their conclusion was, once again, negative: No residual gas was detected. In sum-mary, most clusters are unable to accrete large gas reservoirs to support ongoing star formation.However, this does not mean that they will not be able to capture external gas after the initial gas-expulsion phase. Li et al. (2016b) found the clear presence of younger stellar populations in threeintermediate-age star clusters, NGC 1783, NGC 1806, and NGC 411, which seem to have an externalorigin. These young stellar populations are tightly associated with SSP isochrones, which cannot beexplained by contamination due to field stars spanning a range of ages and metallicities. In addition,the radial profiles of these stellar populations clearly peak in the clusters’ central regions (for a moredetailed rebuttal, see Section 4). They speculated that these star clusters may have accreted externalgas to form new stars at a later stage, although the newly born stars only represent a minor fractionin mass.As regards the eSFH problem, numerous debates now focus on whether other features ofintermediate-age star clusters are also consistent with an eSFH. Li et al. (2014b) found that in the1.7 Gyr-old cluster NGC 1651, despite exhibiting an eMSTO consistent with a 450 Myr age spread,the SGB morphology can only be reconciled with an age spread of ≤
160 Myr. The same conclusionwas reached by Bastian & Niederhofer (2015) for NGC 1806 and NGC 1846. However, their resultswere criticized by Goudfrooij et al. (2015), who performed numerical simulations based on theseclusters’ physical parameters to show that the morphologies of their SGBs may still be consistentwith significant age spreads. To explain the apparent inconsistencies between the widths of the SGBs
C.-Y Li et al. and the stellar evolutionary models, they suggested stellar convective overshooting as a viable alter-native solution. However, Li et al. (2016a) found a more conspicuously narrow SGB in NGC 411,which can only be explained by an SSP. This is in obvious contrast with the extent of its eMSTOregion. They concluded that NGC 411 is more likely composed of a coeval stellar population ratherthan of multiple-aged stellar populations. Girardi et al. (2013) claimed that NGC 411 appears to haveexperienced an eSFH lasting ∼
700 Myr, which immediately introduces an apparent discrepancy withrespect to the observed SGB. On the other hand, Mucciarelli et al. (2014) studied eight giant stars inthe cluster NGC 1806 and did not find any evidence of a possible Na–O anticorrelation, common toalmost all Galactic GCs. The discovery of Mucciarelli et al. (2014) indeed stands in marked contrastto the chemical variations present in GCs. In Fig. 2 we present the Na–O diagram pertaining to theNGC 1806 stars (pink), compared with that of GCs (gray). −0.5 0 0.5−0.200.20.40.60.8 [O/Fe] [ N a/ F e ] Fig. 2
Na–O diagram of stars in NGC 1806 (pink) and Galactic GCs (gray). The NGC1806 data were obtained from Mucciarelli et al. (2014); GC data are from Carretta et al.(2009).
Old GCs appear to represent a separate population of star clusters compared with OCs, YMCs, andalso intermediate-age star clusters. Multiple stellar populations are thought to be a common featureof GCs in the Milky Way (Gratton et al. 2012). The first evidence of multiple stellar populations inGCs may date back to the 1980s: Hesser & Bell (1980) found a dispersion in the CN-band strengths tellar populations in star clusters 9 among seven MS stars in the GC 47 Tucanae (47 Tuc), indicating that a process leading to N en-hancement must have occurred in some of those stars. Since MS stars in 47 Tuc lack the capacity ofmixing N to the stellar surface, such a CN variation must have a primordial origin. Similar resultswere also confirmed in other GCs, e.g., in NGC 362 (Smith 1983), NGC 6171 (Smith 1988), M71(NGC 6838; Penny et al. 1992), M3 (NGC 5272; Smith 2002), and M13 (NGC 6205; Smith & Briley2006)). N-enhanced stellar populations exhibit a strong absorption line at ∼ ∼ × K) and the Na proton-capture process( ∼ × K). The central temperatures of the MSTO stars in GCs ( ∼ M (cid:12) ) do not reach thethreshold temperature for initiation of the proton-capture process required for the production of Na(Decressin et al. 2007). Therefore, the Na–O anticorrelation is unlikely the result of stellar evolution(e.g., processes affected by the first dredge-up or meridional mixing; Sweigart & Mengel 1979),but it must instead have a primordial origin intrinsic to the nature of multiple stellar populations.A similar Mg–Al anticorrelation has also been found in GCs (e.g., Gratton et al. 2001; Yong et al.2003); for more details, see Gratton et al. (2004, 2012). It seems that all Galactic GCs harbor Na–Oanticorrelations, while light-element variations are also present in some LMC GCs (Johnson et al.2006).The iron abundance in a star cluster can only be enhanced through the ejecta of massive, core-collapse supernovae (Cayrel 1986; Parmentier et al. 1999). If a cluster exhibits an iron dispersion,this indicates that its SFH has lived through supernova explosions. Such clusters have been observed:Marino et al. (2009) explored 17 giant stars in the GC M22 and identified two groups with differentiron and s-process abundances. A similar result, for M2, was obtained by Yong et al. (2014), whodetermined a metallicity dispersion exhibiting three peaks, at [Fe/H] ∼ − . , − . , and − . dex.Such a metallicity dispersion has also been found in the GC NGC 1851 (Carretta et al. 2010). Theexamples of M2, M22, and NGC 1851 strongly suggest that at least some GCs have experiencedmuch more complex SFHs than provided by individual starburst episodes.More impressive evidence comes from precise photometry: the multiple stellar groups that showdifferent properties in the CMD have also been found to possess different chemical compositions.For example, NGC 1851 was found to harbor dual SGBs; its two branches were also found to havedifferent CNO contents (Cassisi et al. 2008). D’Antona et al. (2005) studied the color distributionof MS stars in the GC NGC 2808 based on Hubble Space Telescope ( HST )/Wide Field PlanetaryCamera-2 (WFPC2) observations. They found that the width of its MS is inconsistent with an SSP,which requires a helium abundance ranging from Y = 0 . to Y = 0 . . This has been confirmed byPiotto et al. (2007), who obtained accurate photometry of NGC 2808 based on observations with the HST /Advanced Camera for Surveys (ACS). They found that the broadened MS is actually composedof three distinct subsequences, indicating three stellar populations with helium-abundance peaks at Y = 0 . , Y = 0 . , and Y = 0 . . Spectroscopic evidence was provided by Marino et al. (2014),who analyzed 96 HB stars in NGC 2808 and found that their Na content depends on their position inthe CMD, with blue HB stars having higher Na abundances than their red counterparts. The multiplefeatures in GCs are more apparent in ultraviolet observations (Piotto et al. 2015). A recent study ofNGC 7089 (M2) even revealed seven stellar populations (Milone et al. 2015b).Photometric analyses also enable one to investigate the kinematics and dynamics of the differ-ent stellar populations. Milone et al. (2012) found dualities in the MS, SGB, and RGB of 47 Tuc,which can be explained by a CN-weak, O-rich, and Na-poor first stellar generation with a normal helium abundance ( Y ∼ . ) and a secondary stellar generation with CN-strong, O-poor, Na-richelemental abundances as well as an enhanced helium abundance ( Y ∼ . ). The radial behav-iors of both stellar generations are different: the second stellar generation is more concentrated thanthe first generation. Li et al. (2014c) placed their results on a firmer footing through near-infraredobservations. They determined more obvious radial stellar population gradients, concluding that thecentral region of 47 Tuc is almost entirely dominated by the secondary stellar generation, while thecluster’s periphery is fully composed of first-generation stars. They also found that the secondarystellar generation has enhanced helium abundance and metallicity. Similar results were also foundfor NGC 362 (Carretta et al. 2013) and several other GCs (Lardo et al. 2011).The observational status of stellar populations in star clusters can be summarized as follows:(a) Despite small numbers of stellar members, extremely young, massive clusters still embeddedin their GMCs (usually with ages ≤ The presence of multiple stellar populations in GCs is undisputed, making it the greatest openchallenge to the SSP scenario for star clusters.
A combination of spectral and photometric evi-dence supports the presence of multiple stellar populations in GCs, which also differ in both theirkinematics and dynamics. This may indicate primordial chemical inhomogeneities. However, aswe will show in Section 4, stellar evolution is probably still a straightforward solution to explainthe observations of some of these clusters.
An intuitive explanation of the eMSTO regions observed in intermediate-age star clusters is thattheir member stars may have formed during a time span of several hundred million years. Since themorphology of the MSTO is very sensitive to the age distribution of its stellar population, manyauthors have suggested that eMSTO regions in intermediate-age clusters might imply eSFHs of atleast
300 Myr. Goudfrooij et al. (2014) proposed a model for massive clusters characterized by aprolonged round of star formation: the central escape velocities of their eMSTO model star clusterswere sufficiently high to accrete a significant amount of pristine gas from their surroundings orto retain the gas from the ejecta of intermediate-mass AGB stars and massive binaries. Since theremnant gas does not fully escape from the infant clusters, most of this gas reservoir would beaccumulated in the central regions of the clusters, subsequently forming secondary populations ofstars. This process would last several million years until all gas is exhausted by star formation ormultiple supernova explosions. However, the lack of residual gas in YMCs (Bastian et al. 2013) allbut invalidates this scenario. A second proposed origin of such an age spread is the merger betweentwo clusters with different ages (Mackey & Broby Nielsen 2007) or between a cluster and a star-forming GMC (Bekki & Mackey 2009). If this scenario turns out to work, it would imply that mostLMC clusters actually belonged to binary systems in the past, since 70% of intermediate-age starclusters in the LMC show such eMSTOs (Milone et al. 2009). However, presently binary clustersonly account for about 10% of the current cluster population of the Magellanic Clouds (Dieball &Grebel 2000). tellar populations in star clusters 11
An alternative scenario to explain eMSTOs draws on a range in stellar rotation rates. Stellarrotation can alter the morphology of MSTO regions in two ways: (i) the centrifugal force resultingfrom rotation reduces the stellar self-gravity, thus decreasing the surface effective temperature, whichin turn renders its observed color redder than that of the non-rotating counterparts. The reductionin stellar self-gravity also decreases the stellar core pressure, hence reducing the central nuclearreaction rate (Faulkner et al. 1968), which causes a dimming of fast rotators compared with slowlyor non-rotating stars. The observed stellar luminosities and colors also depend on the inclinationangles of individual stars: a rapidly rotating star will look redder around its equatorial region thannear its poles. This effect is known as ‘gravity darkening’ (see Georgy et al. 2014, their Fig. 1). (ii)Rotation enlarges the stellar convective cores, transporting hydrogen from the outer layers into thecentral region, in turn replenishing the central fuel for hydrogen burning, which may thus prolongthe stellar MS lifetime. This is called ‘rotational mixing.’ It has been suggested that stellar rotationof O-, early-B-, and F-type stars may lead to lifetime increases by about 25% (Maeder & Meynet2000; Girardi et al. 2011).Less massive stars ( ≤ . M (cid:12) ) are not expected to become fast rotators because of magneticbraking (Schatzman 1962; Mestel & Spruit 1987), i.e., the stellar magnetic field exerts a torqueon the ejected matter during stellar evolution, resulting in a steady transfer of angular momentumaway from the star. For example, stars earlier than F0-type can easily reach an average rotationalvelocity of 100–200 km s − , while the typical rotational velocity of G0-type stars is only 12 km s − (McNally 1965). A more recent study of a large number of B–F-type solar-neighborhood stars wascarried out by Royer et al. (2007), who confirmed that most of these stars are fast rotators.Bastian & de Mink (2009) suggested that rapid stellar rotation of F-type stars may lead to themisconception that intermediate-age star clusters harbor significant age spreads. In their scenario,the dominant effect that is responsible for the extent of the (e)MSTO region is gravity darkening.However, Girardi et al. (2011) countered that their model is unrealistic, because it ignores rota-tional mixing. The latter authors showed that rotational mixing will mitigate the broadening causedby gravity darkening. As a result, the combination of these two effects may still produce a nar-row MSTO, and therefore, they argued, stellar rotation cannot be the (only) cause of the eMSTO.However, Yang et al. (2013) pointed out that the conclusion of Girardi et al. (2011) is tenable only inthe presence of a high convective mixing efficiency. Yang et al. (2013) studied the collective effectsof gravity darkening and rotational mixing with a moderate mixing efficiency. They successfully re-produced the eMSTO regions for 1–2 Gyr-old clusters. A corollary of their results is that the MSTOarea in a cluster depends on its actual age (see also Brandt & Huang 2015). Niederhofer et al. (2016)found a correlation between the full width at half maximum (FWHM) of the postulated age spreadsin clusters and their actual ages (based on isochrone fitting to the blue boundary of their eMSTOregions) for 12 intermediate-age LMC star clusters, which is consistent with the predictions of thestellar rotation scenario. In Fig. 3 we present the predicted correlation between the FWHM of theclusters’ age spreads and their typical ages for coeval stellar populations (reproduced from Fig. 8f ofYang et al. 2013); observational data are also included (Goudfrooij et al. 2014; Correnti et al. 2015;Niederhofer et al. 2015b).Figure 3 shows that the MSTO regions of clusters with ages of up to 2.5 Gyr are more extendedthan expected for SSPs. The maximum age spread derived from the eMSTO region occurs for agesranging from ∼ Age (Gyr) F W H M M S T O ( M y r ) Goudfrooij et al.(2014)Niederhofer et al.(2015)Correnti et al.(2015)Yang et al.(2013)
Fig. 3
Widths of implied cluster age spreads as a function of isochronal age. Black dashedline: Predicted FWHM of cluster age spreads that would be derived from their eMSTO re-gions as a function of cluster age. A positive (negative) FWHM means that the populationsof fast rotators are redder (bluer) than the non- or slowly rotating populations. Blue, red,and orange rectangles: star cluster data from Goudfrooij et al. (2014), Niederhofer et al.(2015b), and Correnti et al. (2015), respectively.than the non- or slow rotators, predominantly owing to rotational mixing. The maximum age spreadderived from their MSTO regions peaks at t ∈ [1.0,1.5] Gyr.If the apparently significant age spreads of several hundred million years are indeed valid forYMCs ( ≤
300 Myr), one should expect pre-MS stars to appear in YMC CMDs, i.e., the age distribu-tions of their member stars would scatter to zero age. However, as shown in Fig. 3, the internal agespreads of YMCs derived from their eMSTO regions only span small fractions of their ages. Onceagain, this supports the rapid stellar rotation scenario, since it will only partially broaden the MSTOregion.The relative importance of gravity darkening and rotational mixing in star clusters is as yet un-clear owing to a lack of direct observational evidence. The main difference between these processesresides in the loci of the rapidly rotating population. Gravity darkening will produce an eMSTOwhere most fast rotators reside toward the red side of the MSTO, while rotational mixing will resultin fast rotators being distributed on the blue side of the MSTO region. However, measuring stellarrotation in dense star clusters at the distance of the LMC is difficult with current facilities. Since bothstellar rotation and age spreads will produce eMSTOs in star clusters, important differences whichmay allow us to distinguish between these processes may appear on the SGB. Because gravity dark-ening does not produce a mass spread among MSTO stars, once the MSTO stars have evolved off theMS, they will slow down rapidly as a result of the conservation of angular momentum. Subsequently,the coeval MSTO population characterized by different stellar rotation rates will converge into a nar-row SGB (Wu et al. 2016). However, rotational mixing will produce a mass spread among evolvedstars, which will still broaden both the MSTO and SGB regions. In Table 1 we summarize the fea-tures of the MSTO region and the SGB for four cases: (1) an SSP without stellar rotation; (2) a tellar populations in star clusters 13
Table 1
MSTO and SGB features corresponding to (1) an SSP, (2) an age spread (eSFH),(3) gravity darkening, and (4) rotational mixing.
Scenario MSTO Region SGB RegionSSP Narrow NarroweSFH Extended in color and luminosity Extended in luminosityGravity darkening Extended in color and luminosity NarrowRotational mixing Extended in color and luminosity Extended in luminosity significant age spread (eSFH); (3) an SSP with fast rotators affected by gravity darkening; and (4)an SSP with fast rotators affected by rotational mixing.The Geneva stellar evolutionary code includes the effects of initial stellar rotation, (Ekstr¨om etal. 2012; Georgy et al. 2013a,b; Yusof et al. 2013) but it currently only covers a limited stellar massrange and a small number of fixed initial rotation rates. (Observationally, we only have access tocurrent stellar rotation rates). In Fig. 4 we present synthetic CMDs of SSPs with Ω ini / Ω crit ≤ . (left) and Ω ini / Ω crit ≥ . (right) evolved to an age of 1 Gyr and for a metallicity of Z = 0 . , i.e.,the value closest to that typical of most intermediate-age LMC and SMC star clusters. Applicationof the Geneva code shows that for initial stellar rotation rates of Ω ini / Ω crit ≤ Ω ini / Ω crit ≥ Ω ini / Ω crit ≥ without convective overshooting results in Fe abundances that most closely match the loci of starsin clusters with ages between 0.05 Gyr and 3 Gyr. Another striking contrast between the widths ofthe eMSTO region and the SGB is exhibited by the SMC cluster NGC 411, which hosts an eMSTOregion that implies an apparent age spread of ∼
800 Myr, but it harbors a extremely tight SGB, whichcan only be described by a single-aged isochrone (Li et al. 2016a). In Fig. 5 we present the CMDs of(left) the LMC cluster NGC 1651 and (right) the SMC cluster NGC 411. The best-fitting isochronesthat cover their eMSTO regions (indicated by the purple outlines) are also shown; the SGB regionsare indicated by regions traced by black dashed lines. One can immediately see that, compared withtheir eMSTO regions, the clusters’ SGBs do not show the expected corresponding broadening orbifurcation. This combined analysis of cluster eMSTO regions and tight SGBs indeed disfavors theage-spread scenario. Instead, only an extremely rapidly rotating population will produce an eMSTOcombined with a tight SGB (see the comparison in Fig. 4), which may indicate that stellar rotationin star clusters occurs at much higher velocities than previously expected (Wu et al. 2016). The Geneva models are only available for nine different initial rotation rates, Ω ini / Ω crit = 0.0, 0.1, 0.3, 0.5, 0.6, 0.7,0.8, 0.9, and 0.95, where Ω ini is the initial stellar angular rotation rate and Ω crit is the critical, ‘break-up’ value. The stellar V – I (mag) V ( m ag ) V – I (mag) V ( m ag ) Fig. 4
Synthetic (
V, V − I ) CMD for a coeval stellar population (based on the Genevamodels) evolved to an age of 1 Gyr and with a metallicity of Z = 0 . , for stars withinitial rotation rates of (left) Ω ini / Ω crit ≤ Ω ini / Ω crit ≥ masses included range from . M (cid:12) to M (cid:12) . In addition, for the mass range between . M (cid:12) and M (cid:12) , the model suiteis defined for two initial rotation rates, Ω ini / Ω crit = 0 . and 0.568. B – I (mag) B ( m ag ) B – I (mag) B ( m ag ) Fig. 5
CMDs of the intermediate-age star clusters NGC 1651 (left) and NGC 411 (right).Both harbor eMSTO regions (purple outlines) and well-populated SGBs (black dashedoutlines). The best-fitting isochrones (blue, red solid lines: young, old isochrones) are alsoshown, determined based on the extents of their eMSTO regions. tellar populations in star clusters 15
The discovery of eMSTOs in YMCs further confirms this. The eMSTO of NGC 1856, a 150Myr-old YMC, suggests an ∼
80 Myr age spread (Correnti et al. 2015). Correnti et al. (2015) con-cluded that if the observed eMSTO is caused by stellar rotation instead, then rotational mixing willreproduce the observations if the stellar population is composed of a combination of two-thirdsrapidly rotating stars and one-third slowly/non-rotating stars. The same conclusion was reached byMilone et al. (2016), who discovered a split main sequence in the YMC NGC 1755. This featurefavors the variable stellar rotation scenario rather than an age spread. They considered whether allstars in NGC 1755 were born rapidly rotating. All these studies were based on the premise that ro-tational mixing is the dominant effect in determining the extent of an eMSTO. If, on the other hand,stellar rotation were responsible for the eMSTOs of YMCs, rotational mixing should be of signifi-cant importance. Thus, to further constrain the internal age distributions and the relative importanceof rotational effects, future investigations of the SGBs or RCs of intermediate-age star clusters areurgently required.In summary, the apparent eMSTO regions of intermediate-age star clusters and YMCs in theLMC and SMC can be explained by scenarios postulating either an age spread or rapid stellar rota-tion. The field still has a long way to go before reaching a unanimous conclusion.
Unlike for intermediate-age star clusters, the presence of multiple stellar populations in old GCsis well-established. Scenarios that have been proposed to explain the formation of multiple stellarpopulations in old GCs are diverse, but most can be grouped into two categories. The first type ofhypothesis suggests a primordial origin, in the sense that GCs may have been born with chemicalinhomogeneities, i.e., GCs were formed with a primordial chemical dispersion. The second type ofscenario favors the SSP origin, and the multiple stellar populations in GCs are then the result ofstellar evolution (see also Kraft 1994).Most primordial scenarios draw on self-pollution of intra-cluster gas, occurring during the earlystages of cluster evolution. Various polluters have been proposed, including those originating fromthe ejecta of rapidly rotating massive stars (Decressin et al. 2007), massive binaries (de Mink et al.2009), or evolved post-giant-branch stars (Ventura & D’Antona 2009). Valcarce & Catelan (2011)proposed a scenario that divides GCs into three groups, where the most massive GCs, with initialmasses ≥ M (cid:12) (e.g., M22), can retain the ejecta of all types of massive stars, including thoseof their core-collapse supernovae. Intermediate-mass GCs, with masses of several × M (cid:12) (likeNGC 2808) would only be able to retain a fraction of the fast winds from massive stars. The leastmassive GCs would not be able to retain any gas ejected by massive stars (or their products), but theycould form new stars from the slow winds of intermediate-mass stars of the first stellar generation.Self-enrichment scenarios usually postulate multiple episodes of star formation. If confirmed, thismeans that we may have underestimated the capacity of GCs to collect and retain gas. Recently,D’Antona et al. (2016) proposed a similar scenario. They claimed that a temporal sequence of AGBgenerations may be responsible for the observed multiple stellar populations in GCs: the pollutionprocess occurs after the Type II supernovae epoch, lasting until the third dredge-up associated withthe AGB population. They ascribe the cluster-to-cluster abundance variations to differences in manyprocesses and gas sources which were involved in the formation of the secondary generation. Basedon this scenario, after the Type II supernovae epoch ( ∼
20 Myr), GCs should still have been 100times more massive then their current masses; otherwise the self-pollution scenario would be hardpressed to explain the high fraction of secondary stellar generations observed.Bastian et al. (2015) carefully studied all current self-enrichment scenarios by comparing themto observations. They found that an intrinsic problem among all these scenarios is that they areunable to produce consistent abundance trends among He, Na, and O. To form a secondary stellargeneration, clusters should be able to retain or accrete material from their surroundings (Conroy &
Spergel 2011), but current investigations have shown that clusters are almost gas-free at a very earlystage (2–3 Myr), independent of their mass (Bastian et al. 2014; Hollyhead et al. 2015). While theseresults suggest that clusters may not be able to retain their gas during the initial gas-expulsion phase,this scenario does not prevent a cluster from accreting external gas onto its core potential whilemoving through a background medium (Naiman et al. 2011).An unexpected discovery that relatively young, 1–2 Gyr-old star clusters can harbor secondarystellar populations was made by Li et al. (2016b), who found that the LMC clusters NGC 1783 andNGC 1806, as well as the SMC cluster NGC 411, host secondary stellar populations that are gen-uinely younger than their first-generation stars by about 500 Myr to 1 Gyr. The observed, differentyoung stellar populations are tightly associated with younger isochrones, indicating that they arestrictly coeval. However, these young stellar populations are somewhat less centrally concentratedcompared with the first-generation RGB stars in the same clusters. These authors speculated that theyounger stars may have an external origin, which renders the gas-accretion scenario promising onceagain. However, because the observed young stellar populations in those extragalactic star clustersonly occupy minor mass fractions, additional numerical simulations are required to further study theimplications of these observations. In Fig. 6 we present the observed CMD of NGC 1783 as an exam-ple. Although the young stars have a slightly more extended nature compared to the approximatelyequal-luminosity RGB stars, they are firmly concentrated within 1 to 2 core radii. The observedyoung stellar population stars have a number-density profile that clearly peaks in the clusters’ coreregions. This cannot be explained by invoking background contamination (see Fig. 7), which impliesthat these newly born stellar populations indeed physically belong to their host star clusters.Another primordial formation scenario is based on cluster mergers, which is the prevailingmodel for some of the most massive GCs, such as NGC 1851 (Carretta et al. 2010). An advantage ofthis hierarchical formation scenario is that it solves the ‘initial exhaustion’ problem. This scenariopreferentially favors clusters that are located in crowded environments, such as the so-called ‘clustersof clusters’ (e.g., cluster pairs in the Antennae interacting system; Bastian et al. 2009). Cluster mem-bers in these environments are expected to frequently collide and merge (Amaro-Seoane et al. 2013).However, most GCs in our Milky Way are located in the Galactic halo, where frequent mergers arenot expected to occur.The evolutionary scenario attributes the observed abundance variations to the dredge-up of ma-terial that has been processed through the CNO cycle in the cluster stars themselves. Because theobserved chemical dispersions of different elements in GC stars only represent variations in the stel-lar surface abundances, any process that can somehow transfer core material to the stellar surfacewould produce star-by-star variations in elemental abundances. Because the evolutionary scenariostrongly depends on the efficiency of stellar convective mixing, various mechanisms that may af-fect the mixing of stellar material have been proposed. The most promising scenario suggests thatstars will undergo deep mixing when they evolve off the MS. Lee (2010) carefully studied the Na–Odistributions in Galactic GCs. They found a dependence of the Na–O anticorrelation on RGB lumi-nosity. This is probably owing to the internal deep mixing of evolved stars during their ascent of theRGB. Their results thus partially support the evolutionary scenario.Another possible mechanism considers stellar magnetic fields: Nucci & Busso (2014) proposeda scenario involving the advection of thermonuclear ashes by magnetized domains emerging nearthe H shell to explain AGB-star abundances. Based on magnetohydrodynamics calculations, theyverified that the stellar-envelope crossing times are sufficient to facilitate chemical dispersion in ahuge convective shell. They claimed that magnetic advection is a promising mechanism for deepmixing, which may explain the observed abundance anomalies in GCs. However, because stars inGCs are usually late-G- or K-type stars, the central temperatures of individual low-mass stars do notreach the threshold for producing the observed Na–O anticorrelation. Jiang et al. (2014) proposeda model based on fast rotators that are produced by binary mergers or interactions. They found forbinary merger products which are normal, rapidly rotating stars earlier than F-type, that these rapidly tellar populations in star clusters 17 B – I (mag) B ( m ag ) BA Fig. 6
CMD of the LMC cluster NGC 1783 (1.4 Gyr old). NGC 1783 hosts two apparentlyyounger stellar populations, which are tightly associated with two isochrones (shown asthe blue dashed and the red solid lines), indicated as populations A and B. Similar featuresare also found in the LMC cluster NGC 1806 and the SMC cluster NGC 411.rotating stars can produce a Na–O anticorrelation with high significance. They claimed that binaryinteractions are a possible solution to the chemical anomalies in GCs.The evolutionary scenarios do not require an extremely massive origin for clusters to generateabundance anomalies. Jiang et al. (2014)’s scenario only depends on a cluster’s binary fraction.This implies that OCs would also display Na–O anticorrelations if they are sufficiently old for largenumbers of their binaries to have merged. However, so far we do not have any conclusive evidencethat OCs may harbor multiple stellar populations. Cantat-Gaudin et al. (2014) studied the chemicalhomogeneity of RC stars in the OC NGC 6751, but no obvious correlations or anticorrelations amongAl, Mg, Si, and Na were found. A similar conclusion was also reached by Bragaglia et al. (2014),who studied 35 evolved stars in an old OC, NGC 6791 (8–9 Gyr old; Brogaard et al. 2012). Theydid not detect any significant star-to-star chemical dispersions in C, N, O, or Na either. In Fig. 8we present the performance of the scenario of Jiang et al. (2014). In the top panel, we show thetheoretical CMD of an SSP following binary mergers; in the bottom panel, we show the Na–Oanticorrelation produced by this model compared with observations made in GCs. log R [arcsec] l og ρ [ a r c s ec ] Fig. 7
Number-density profile of young stellar population stars in NGC 1783 (see also theCMD in Fig. 6).Figure 8 shows that there is a clear correlation between the stellar Na (as well as the O) abun-dance and luminosity. The binary products are, in fact, ‘blue straggler stars’ (BSSs). For more detailsabout the primordial and evolutionary scenarios for the origin of multiple stellar populations in GCs,we recommend the review by Kraft (1994); see also Denissenkov et al. (2015).
The origin of multiple stellar populations in star clusters is a subject of significant current debate.To reach closure on this question, the most straightforward approach would consist of observing a‘young GC’ and resolving its initial stellar population(s). Most scenarios suggest that a cluster’s massplays a key role in the existence of multiple populations. The most popular scenarios suggest thatGCs should initially have been 10 to 100 times more massive than their current masses. Therefore,a ‘good’ candidate for follow-up observations should be a young cluster with a mass of at least M (cid:12) . To date, no such supermassive young clusters have been observed in either the Milky Way orits satellites.Strictly speaking, young GCs should reside at a redshift of z ∼ , which is even beyond theability of the James Webb Space Telescope to resolve. Therefore, resolving such young clusterswill not be feasible in the near future. However, some candidate extremely massive young GCshave already been detected in nearby starburst galaxies. Smith et al. (2006) found five possiblesupermassive cluster candidates in the galaxy M82. The mass of one of these ‘clusters,’ M82-A1, is . +0 . − . × M (cid:12) , at an age of only . ± . Myr. If this object truly is an individual YMC, itmay be the best candidate GC predecessor known. However, our current facilities are not yet able tellar populations in star clusters 19 log T eff l og ( L / L ⊙ ) −1 −0.5 0 0.5 1−0.500.51 [O / Fe] [ N a / F e ] Fig. 8
The importance of binary systems in realistic stellar populations. Top: CMD of anSSP including binary interactions. Color bar: stellar Na abundance. Bottom: Na–O rela-tionships produced by the model (blue) compared with observations (gray bullets). Eachsequence of blue dots represents an SSP, characterized by different initial abundances.to resolve individual stars in such supermassive young star clusters, and only synthetic spectra areavailable to study their stellar populations (e.g., Cabrera-Ziri et al. 2015).Once next-generation telescopes have been developed, a top priority to improve our understand-ing in this field will be to resolve individual stars in these candidate young GCs. If these supermassiveyoung objects are confirmed as individual star clusters (and not blended pairs or groups of clusters),then studying their stellar populations will significantly help us constrain the origin of GCs. Usingthe CMDs of these supermassive young star clusters, we can determine their ages and any internalage dispersions, which is of fundamental importance for studying the star-formation mode in youngGCs. If the multiple stellar populations observed in GCs represent multiple episodes of star forma-tion, then one should detect dispersed age distributions in all supermassive young star clusters. Thismeans that large fractions of pre-MS stars are expected to reside in those clusters, and their residualgas should still represent a large mass fraction.
The ‘eMSTO problem’ associated with intermediate-age star clusters has shown that rapid stellarrotation is an important process that may confuse our understanding of the SFHs of star clusters.Although measuring stellar rotation rates of individual stars in compact star clusters at the distanceof the LMC is challenging, some current instruments can already obtain direct measurements ofstellar rotation in LMC star clusters. Using the Multi Unit Spectroscopic Explorer (MUSE) at theEuropean Southern Observatory’s Very Large Telescope (VLT), one can in principle obtain low-resolution spectra of all stars resolvable by MUSE in a 1 arcmin field covering the central regionsof intermediate-age LMC star clusters, including of their F-type eMSTO stars. If rotation is the causeof the eMSTO, one will immediately obtain a clear and unequivocal signature in support of this.One can quantify stellar rotation rates by measuring line broadening in eMSTO stars. The Mg I triplet around 5175 ˚A is an ideal indicator of the stellar rotation rate. Adopting the rapid stellar ro-tation scenario, a stellar rotation rate from zero to a maximum of 70% of the break-up rate will beresponsible for the observed eMSTO in intermediate-age star clusters. This corresponds to equato-rial velocities of ∼
300 km s − . The MUSE spectral resolution will allow us to resolve differencesbetween no rotation and rotation rates as low as (50–)100 km s − , with a signal-to-noise ratio of 30.In Fig. 9 we present the simulated morphology of the Mg I triplet of F2V stars for different rotationvelocities, as they would be observed with MUSE. As long as there are no nearby stars in the sameresolution element with intensities within 1 mag of the target stars, one will be able to derive a sam-ple of stellar radial velocities, velocity dispersions, and the overall distribution of (eMSTO and other)rotation rates in such clusters. Such a project will provide a definitive answer to the importance ofthe stellar rotation scenario with respect to the internal age-spread model as regards the origin ofeMSTOs in intermediate-age star clusters. If the scenario of Jiang et al. (2014) is correct, a correlation between elemental abundances and theluminosities of BSSs is expected. BSSs are rejuvenated MS stars, formed through either stellar col-lisions (Hills & Day 1976) or binary mass transfer (McCrea 1964; Carney et al. 2001). The relativeimportance of these two formation channels during different evolutionary stages in star clusters hasnot yet been determined unequivocally from an observational perspective. However, numerical sim-ulations have shown that the number of BSSs of binary origin will continue to dominate the totalBSS sample over cosmic timescales ( ∼
10 Gyr). The number of collisional BSSs becomes compa-rable to the number of BSSs originating from binary interactions on timescales exceeding a Hubbletime (Hypki & Giersz 2013).Jiang et al. (2014)’s scenario draws on binary mergers to generate a population of fast rotators,producing surface abundance anomalies through evolution of the latter. A promising approach toexamine this scenario is by studying the correlation between luminosity and elemental abundancesas pertaining to BSSs. In Fig. 10 we present the relationship between the expected stellar oxygenand sodium abundances ([O/Fe], [Na/Fe]) and their absolute F555W-band magnitudes, calculatedusing the PARSEC stellar evolutionary models (Bressan et al. 2012) for a metallicity of Z = 0 . .A clear correlation between luminosities and elemental abundances appears, especially for stars thatare brighter than the typical magnitude of the MSTO region. To measure a significant number ofBSS magnitudes and chemical abundances, relatively old OCs would be good targets. Since theirmember stars have already evolved to old ages, there should have been sufficient opportunities forbinary mergers to occur. An ideal target is the 6.5 Gyr-old OC NGC 188 (Demarque et al. 1992).Mathieu & Geller (2009) found that approximately 76% of its BSSs are members of binary systems,indicating a large fraction of interacting binary stars. tellar populations in star clusters 21 Fig. 9
Simulations of the Mg I triplet of an F2V star for different rotation velocities of 0 kms − (black), 100 km s − (red), 200 km s − (blue), and 300 (green) km s − . The rotationaxis is oriented perpendicularly to the line of sight. From top to bottom, the signal-to-noiseratio (S/N) = 50, 30, 20; the adopted spectral resolution is R ∼ . If the mechanismresponsible for the eMSTOs in intermediate-age star clusters is rapid rotation, then it willbe detected at the minimum S/N = 30 for most of the lines. Significantly more work is required before we can connect young GCs to genuine GCs (which areusually defined as more massive than M (cid:12) —see Fig. 1—and older than 10 Gyr). Numerousstudies have addressed this issue. In general, YMCs should contain sufficient numbers of memberstars initially to survive for a Hubble time (Portegies Zwart et al. 2010). Another key parameter thatdetermines the survival time of YMCs is the slope of its stellar initial mass function (de Grijs &Parmentier 2007). Current consensus on long-term cluster survival implies that the combination ofinternal and external cluster dynamics must play a important role (de Grijs 2010).The discovery of young stellar populations in intermediate-age LMC star clusters is exciting. Itindicates that star clusters may have access to various channels to form secondary stellar populations −0.8 −0.6 −0.4 −0.2 00246 [O/Fe] (dex) F W ( m ag ) [Na/Fe] (dex) F W ( m ag ) Z = 0.01MSTOMSTO Fig. 10
Stellar F555W magnitude as a function of (top) [O/Fe] and (bottom) [Na/Fe]abundance, based on Jiang et al. (2014)’s model. The arrows indicate the typical magnitudeof the MSTO region. The metallicity adopted is Z = 0 . .(Li et al. 2016b). A key difference between the observed secondary stellar populations in theseyoung GCs compared with old stellar populations in genuinely old GCs relates to their dynamics.The young stellar populations in the young clusters are less centrally concentrated than the evolvedstars of the first generation. In old GCs the secondary stellar generations are usually more denselydistributed than the first-generation stars. Therefore, it is still unclear if the observed secondarystellar generations in these relatively young clusters are similar to the second-generation stars in oldGCs.More evidence for more evolved star clusters (e.g., with ages between 3 Gyr and 10 Gyr) isindeed required. Unfortunately, there is an age gap between 3 Gyr and 10 Gyr in the LMC clus-ter sample (Balbinot et al. 2010), so one has to turn to numerical simulations to study this aspect.Li et al. (2016b) proposed that ambient gas accretion may be a possible solution to the observedyoung stellar populations. They speculated that GMCs are possible sources of accreted gas. If thisis correct, then accretion should be a common process for all massive star clusters (irrespective oftheir ages) that are located in gas-rich galactic disks. However, the nature of the slightly less con-centrated spatial distributions of the young stellar populations compared with their first-generation tellar populations in star clusters 23 RGB counterparts is unresolved. Cabrera-Ziri et al. (2016) argued that the observed features mayhave been caused by misunderstood issues in field-star decontamination, but it is highly unlikelythat field-star contamination could artificially produce such tightly constrained, well-populated stel-lar populations (see Fig. 6), which also exhibit clearly centrally concentrated radial profiles (see Fig.7). Additional studies of the dynamics of these younger stellar populations are required: indeed, oneneeds to combine N -body simulations with up-to-date insights into stellar evolution to explore theseissues. More accurate models should also place clusters in an external gravitational field, whichwill affect the ‘evaporation’ processes in different stellar populations (for more details, see Spitzer1987). More detailed calculations should also use smoothed-particle hydrodynamics simulations totrace accretion flows between star clusters and GMCs. In this review, we have introduced the community’s up-to-date insights into the physics governingstellar populations in star clusters. We have shown that because of initial gas expulsion, most starclusters are not expected to exhibit multiple episodes of star formation. This renders the origin of thecommon observation of multiple stellar populations in star clusters an intriguing open question. Theobservational status of stellar populations in star clusters can be summarized as follows: – For star clusters younger than ∼
100 Myr, no conclusive evidence exists that they may harbormultiple stellar populations or ongoing star formation; no residual gas has been detected inextremely young clusters (Bastian & Strader 2014). – The eMSTO morphologies of some YMCs with ages older than 100 Myr (e.g., NGC 1850 andNGC 1856; Milone et al. 2015a; Bastian et al. 2016) are inconsistent with the expectations fromSSPs. It appears that such eMSTOs are a common feature of intermediate-age star clusters inthe LMC and SMC. However, spectroscopic analyses have shown that their member stars do noexhibit abundance anomalies (Mucciarelli et al. 2014). – The presence of multiple stellar populations in old GCs is irrefutable: both the morphologiesof their photometric features and star-to-star chemical abundance variations challenge the SSPscenario.Various scenarios have been proposed to explain these observed deviations from genuine SSPs.For young and intermediate-age star clusters, age spreads (which favor eSFHs) and rapid stellarrotation (which suggests that clusters are SSPs) are in competition. For old GCs, all scenarios canbe classified as either primordial or evolutionary.We have proposed a number of projects that seem feasible in the near future and which mayshed light on our understanding of stellar population problems in star clusters. These include directmeasurements of stellar rotation rates in compact star clusters, studies of the elemental abundancesof BSSs, and the use of numerical simulations to study gas accretion. Since a number of possiblecandidate young GCs have been identified in nearby starburst galaxies, employing next-generationtelescopes to study these objects will significantly contribute to an improved understanding of theorigin of stellar populations in star clusters.
Acknowledgements
C. L. is partially supported by a Macquarie Research Fellowship and byStrategic Priority Program ‘The Emergence of Cosmological Structures’ of the Chinese Academy ofSciences (grant XDB09000000). R. d. G. and L. D. acknowledge research support from the NationalNatural Science Foundation of China through grants 11073001, 11373010, and 11473037.
References
Allison, R. J., Goodwin, S. P., Parker, R. J., et al. 2009, ApJ, 700, L994 C.-Y Li et al.Amaro-Seoane, P., Konstantinidis, S., Brem, P., & Catelan, M. 2013, MNRAS, 435, 809Anderson, J., Piotto, G., King, I. R., Bedin, L. R., & Guhathakurta, P. 2009, ApJ, 697, L58Balbinot, E., Santiago, B. X., Kerber, L. O., Barbuy, B., & Dias, B. M. S. 2010, MNRAS, 404, 1625Banerjee, S., & Kroupa, P. 2015, in:
The Origin of Stellar Clusters , ed. S. Stahler (Springer), in press(arXiv:1512.03074)Bastian, N., & Goodwin, S. P. 2006, MNRAS, 369, L9Bastian, N., & de Mink, S. E. 2009, MNRAS, 398, L11Bastian, N., Trancho, G., Konstantopoulos, I. S., & Miller, B. W. 2009, ApJ, 701, 607Bastian, N., Cabrera-Ziri, I., Davies, B., & Larsen, S. S. 2013, MNRAS, 436, 2852Bastian, N., & Strader, J. 2014, MNRAS, 443, 3594Bastian, N., Hollyhead, K., & Cabrera-Ziri, I. 2014, MNRAS, 445, 378Bastian, N., & Niederhofer, F. 2015, MNRAS, 448, 1863Bastian, N., Cabrera-Ziri, I., & Salaris, M. 2015, MNRAS, 449, 3333Bastian, N., Niederhofer, F., Kozhurina-Platais, V., et al. 2016, MNRAS, 460, 206Bate, M. R., Bonnell, I. A., & Bromm, V. 2003, MNRAS, 339, 577Baumgardt, H., & Kroupa, P. 2007, MNRAS, 380, 1589Bedin, L. R., Piotto, G., Zoccali, M., et al. 2000, A&A, 363, 159Bekki, K., & Mackey, A. D. 2009, MNRAS, 394, 124Bellini, A., Piotto, G., Bedin, L. R., et al. 2009, A&A, 507, 1393Bik, A., Stolte, A., Gennaro, M., et al. 2014, A&A, 561, A12Bonnell, I. A., Clark, P., & Bate, M. R. 2008, MNRAS, 389, 1556Bragaglia, A., Sneden, C., Carretta, E., et al. 2014, ApJ, 796, 68Brandl, B., Sams, B. J., Bertoldi, F., et al. 1996, ApJ, 466, 254Brandt, T. D., & Huang, C. X. 2015, ApJ, 807, 25Bressan, A., Marigo, P., Girardi, L., et al. 2012, MNRAS, 427, 127Brogaard, K., VandenBerg, D. A., Bruntt, H., et al. 2012, A&A, 543, A106Br¨uns, R. C., Kroupa, P., & Fellhauer, M. 2009, ApJ, 702, 1268Cabrera-Ziri, I., Bastian, N., Davies, B., Magris, G., Bruzual, F., & Schweizer, F. 2014, MNRAS, 441, 2754Cabrera-Ziri, I., Bastian, N., Longmore, S. N., et al. 2015, MNRAS, 448, 2224Cabrera-Ziri, I., Bastian, N., Hilker, M., et al. 2016, MNRAS, 457, 809Cabrera-Ziri, I., Niederhofer, F., Bastian, N., et al. 2016, MNRAS, 459, 4218Cantat-Gaudin, T., Vallenari, A., Zaggia, S., et al. 2014, A&A, 569, A17Carney, B. W., Latham, D. W., Laird, J. B., Grant, C. E., & Morse, J. A. 2001, AJ, 122, 3419Carretta, E., Bragaglia, A., & Cacciari, C. 2004a, ApJ, 610, L25Carretta, E., Gratton, R. G., Bragaglia, A., Bonifacio, P., & Pasquini, L. 2004b, A&A, 416, 925Carretta, E., Bragaglia, A., Gratton, R. G., et al. 2009, A&A, 505, 117Carretta, E., Gratton, R. G., Lucatello, S., et al. 2010, ApJ, 722, L1Carretta, E., Bragaglia, A., Gratton, R. G., et al. 2013, A&A, 557, A138Cassisi, S., Salaris, M., Pietrinferni, A., et al. 2008, ApJ, 672, L115Cayrel, R. 1986, A&A, 168, 81Chilingarian, I. V., & Mamon, G. A. 2008, MNRAS, 385, L83Cohen, J. G., Persson, S. E., & Frogel, J. A. 1978, ApJ, 222, 165Colucci, J. E., & Bernstein, R. A. 2012, ApJ, 749, 124Correnti, M., Goudfrooij, P., Puzia, T. H., & de Mink, S. E. 2015, MNRAS, 450, 3054Conroy, C., & Spergel, D. N. 2011, ApJ, 726, 36D’Antona, F., Bellazzini, M., Caloi, V., et al. 2005, ApJ, 631, 868D’Antona, F., Di Criscienzo, M., Decressin, T., et al. 2015, MNRAS, 453, 2637D’Antona, F., Vesperini, E., D’Ercole, A., et al. 2016, MNRAS, 458, 2122tellar populations in star clusters 25Davidge, T. J. 2012, ApJ, 761, 155Decressin, T., Meynet, G., Charbonnel, C., Prantzos, N., & Ekstr¨om, S. 2007, A&A, 464, 1029de Grijs, R., & Parmentier, G. 2007, ChJAA (Chin. J. Astron. Astrophys.), 7, 155de Grijs, R. 2010, RSPTA, 368, 693Demarque, P., Green, E. M., & Guenther, D. B. 1992, AJ, 103, 151de Mink, S. E., Pols, O. R., Langer, N., & Izzard, R. G. 2009, A&A, 507, L1Denissenkov, P. A., VandenBerg, D. A., Hartwick, F. D. A., et al. 2015, MNRAS, 448, 3314De Silva, G. M., D’Orazi, V., Melo, C., et al. 2013, MNRAS, 431, 1005Dieball, A., & Grebel, E. K. 2000, A&A, 358, 897Djorgovski, S., & Meylan, G. 1994, AJ, 108, 1292D’Orazi, V., & Randich, S. 2009, A&A, 501, 553Drinkwater, M. J., Gregg, M. D., Hilker, M., et al. 2003, Nature, 423, 519Duncan, D. K., & Jones, B. F. 1983, ApJ, 271, 663Eggen, O. J. 1998, AJ, 116, 284Ekstr¨om, S., Georgy, C., Eggenberger, P., et al. 2012, A&A, 537, A146Farias, J. P., Smith, R., Fellhauer, M., et al. 2015, MNRAS, 450, 2451Faulkner, J., Roxburgh, I. W., & Strittmatter, P. A. 1968, ApJ, 151, 203Geisler, D., Villanova, S., Carraro, G., et al. 2012, ApJ, 756, L40Georgiev, I. Y., Hilker, M., Puzia, T. H., Goudfrooij, P., & Baumgardt, H. 2009, MNRAS, 396, 1075Georgy, C., Ekstr¨om, S., Granada, A., et al. 2013a, A&A, 553, A24Georgy, C., Ekstr¨om, S., Eggenberger, P., et al. 2013b, A&A, 558, A103Georgy, C., Granada, A., Ekstr¨om, S., et al. 2014, A&A, 566, A21Girardi, L., Rubele, S., & Kerber, L. 2009, MNRAS, 394, L74Girardi, L., Eggenberger, P., & Miglio, A. 2011, MNRAS, 412, L103Girardi, L., Goudfrooij, P., Kalirai, J. S., et al. 2013, MNRAS, 431, 3501Girichidis, P., Federrath, C., Allison, R., Banerjee, R., & Klessen, R. S. 2012, MNRAS, 420, 3264Goudfrooij, P., Puzia, T. H., Chandar, R., & Kozhurina-Platais, V. 2011, ApJ, 737, 4Goudfrooij, P., Girardi, L., Kozhurina-Platais, V., et al. 2014, ApJ, 797, 35Goudfrooij, P., Girardi, L., Rosenfield, P., et al. 2015, MNRAS, 450, 1693Gratton, R. G., Bonifacio, P., Bragaglia, A., et al. 2001, A&A, 369, 87Gratton, R., Sneden, C., & Carretta, E. 2004, ARA&A, 42, 385Gratton, R. G., Carretta, E., & Bragaglia, A. 2012, A&A Rev., 20, 50Hansen, C. J., & Kawaler, S. D. 1994,
Stellar Interiors. Physical Principles, Structure, and Evolution , XIII,445 pp. Springer-Verlag: Berlin, Heidelberg, New York.Hesser, J. E., & Bell, R. A. 1980, ApJ, 238, L149Hills, J. G., & Day, C. A. 1976, Astrophys. Lett., 17, 87Hypki, A., & Giersz, M. 2013, MNRAS, 429, 1221Hollyhead, K., Bastian, N., Adamo, A., et al. 2015, MNRAS, 449, 1106Jiang, D., Han, Z., & Li, L. 2014, ApJ, 789, 88Johnson, J. A., Ivans, I. I., & Stetson, P. B. 2006, ApJ, 640, 801Johnson, C. I., & Pilachowski, C. A. 2010, ApJ, 722, 1373Kraft, R. P. 1994, PASP, 106, 553Krumholz, M. R., & Matzner, C. D. 2009, ApJ, 703, 1352Lada, C. J., & Lada, E. A. 2003, ARA&A, 41, 57Lardo, C., Bellazzini, M., Pancino, E., et al. 2011, A&A, 525, A114Lee, J.-W. 2010, MNRAS, 405, L36Lee, Y.-W., Joo, J.-M., Sohn, Y.-J., et al. 1999, Nature, 402, 55Lee, Y.-W., Han, S.-I., Joo, S.-J., et al. 2013, ApJ, 778, L136 C.-Y Li et al.Li, C., de Grijs, R., & Deng, L. 2013, MNRAS, 436, 149Li, C., de Grijs, R., & Deng, L. 2014a, ApJ, 784, 157Li, C., de Grijs, R., & Deng, L. 2014b, Nature, 516, 367Li, C., de Grijs, R., Deng, L., et al. 2014c, ApJ, 790, 35Li, C.,