When does the onset of multiple stellar populations in star clusters occur? Detection of enriched stellar populations in NGC 2121
aa r X i v : . [ a s t r o - ph . S R ] M a r Draft version April 2, 2019
Typeset using L A TEX twocolumn style in AASTeX62
When does the onset of multiple stellar populations in star clusters occur? Detection of enriched stellar populationsin NGC 2121
Chengyuan Li
1, 2,3 and Richard de Grijs Department of Physics and Astronomy, Macquarie University, Sydney, NSW 2109, Australia Centre for Astronomy, Astrophysics and Astrophotonics, Macquarie University, Sydney, NSW 2109, Australia Key Laboratory for Optical Astronomy, National Astronomical Observatories, Chinese Academy of Sciences, 20A Datun Road International Space Science Institute–Beijing, 1 Nanertiao, Zhongguancun, Hai Dian District, Beijing 100190, China
Submitted to ApJABSTRACTStar-to-star light-element abundance variations, know as multiple stellar populations (MPs), arecommon in almost all Galactic globular clusters. Recently, MPs have also been detected in a numberof massive clusters with ages in excess of 2 Gyr in the Large Magellanic Cloud (LMC), thus indicatingthat age is likely a control parameter for the presence of MPs. However, to conclusively confirm thisnotion, additional studies of star clusters in the LMC’s ‘age gap’ of 3–6 Gyr are required. Here, weuse
Hubble Space Telescope observations to study the 3 Gyr-old cluster NGC 2121. Compared withso-called ‘simple’ stellar population models, the cluster’s red-giant branch exhibits an apparent spreadin a specific color index that is sensitive to intrinsic chemical spreads. The observed spread can beexplained by an intrinsic spread in nitrogen abundance of ∼ Keywords: globular clusters: individual: NGC 2121 – Hertzsprung-Russell and C-M diagrams INTRODUCTIONThe notion that all star clusters are simple stel-lar populations (SSPs), with color–magnitude diagrams(CMDs) described by a single isochrone of fixed age andmetallicity, is a view of the past. Current observationsshow that almost all Galactic globular clusters (GCs)are composed of multiple stellar populations (MPs), ex-hibiting multiple subgiant branches (SGBs; Piotto et al.2012), red-giant branches (RGBs; Piotto et al. 2015),main sequences (MSs; Piotto et al. 2007), and some-times combinations of these features. Spectroscopicstudies have demonstrated conclusively that such mul-tiple features can be explained by star-to-star chemi-cal variations, including spreads in C, N, O, Na, Mg,Al, and in some cases iron content (Cannon et al. 1998;Carretta et al. 2009; Marino et al. 2009; Pancino et al.
Corresponding author: Chengyuan [email protected]
Li & de Grijs
Although this correlation between the significance ofthe MPs and cluster mass has been confirmed for oldGalactic GCs (Milone et al. 2017), it is unclear if thefraction of the enriched population also correlates withcluster mass in other galaxies. First, the fractions of en-riched stars in extragalactic clusters are highly uncertain(Niederhofer et al. 2017). In addition, to date only thestellar populations in clusters of satellite galaxies (e.g.,in the Large and Small Magellanic Clouds (LMC andSMC) or the Fornax dwarf; Martocchia et al. 2018) havebeen resolved; most of these clusters are younger thanthe Galactic GCs (Larsen et al. 2014; Dalessandro et al.2016; Martocchia et al. 2018). Since clusters will losemass during their evolution, younger clusters with com-parable masses to those of the Galactic GCs will eventu-ally evolve to become less massive than most GCs by thetime they reach similar ages. The 2.2 Gyr-old LMC clus-ter NGC 1978 is, thus far, the youngest cluster knownto exhibit MPs (Martocchia et al. 2018). Intriguingly,its slightly younger, 1–2 Gyr-old counterparts appearto be fully chemically homogeneous (e.g., NGC 1806;Mucciarelli et al. 2014). These results suggest that agemight be another important factor controlling the ap-pearance of MPs.One popular hypothesis suggests that the observedMPs may have formed through non-standard stellarevolutionary effects associated with stellar rotation(e.g., Bastian & Lardo 2018). Most clusters youngerthan 2 Gyr exhibit extended main-sequence turn-off re-gions (eMSTOs) in their CMDs (e.g., Milone et al. 2009;Cordoni et al. 2018), which are likely caused by differ-ential stellar rotation (Cordoni et al. 2018). Note, how-ever, that MPs seem to occur at ages where the eMSTOshave already disappeared. Martocchia et al. (2018) sug-gest that magnetic fields may play a role in generatingchemical anomalies. They suggest that MPs might bea specific feature of low-mass stars featuring strongmagnetic fields. Such stars cannot be rapidly rotat-ing because of magnetic braking (Cardini & Cassatella2007). This hypothesis, although as yet speculative, im-plies that older clusters without eMSTOs should exhibitMPs.To date, NGC 1978 is the only known cluster exhibit-ing MPs that is younger than 3 Gyr. The stellar pop-ulations of clusters with ages between 3 and 6 Gyr arestill poorly studied, however. To determine the exactage of the onset of MPs, it is important to study clus-ters with ages in the LMC’s so-called ‘age gap’ between3 and 6 Gyr (Piatti et al. 2002). NGC 2121 is one ofthe few suitable clusters in this context. The most re-cent dedicated study of NGC 2121 dates from 18 yearsago (Rich et al. 2001). Its authors derived a cluster age of 3 . ± . − . ± . Hubble Space Telescope ( HST ), we will show thatMPs are indeed present along the cluster’s RGB. Thispaper is organized as follows. In Section 2 we intro-duce our data reduction. In Section 3 we summarizeour analysis approach and present our main results. Wealso compare our results with SSP models. In Section 4we present a discussion and our conclusions. DATA REDUCTIONWe use observations obtained with both the
HST ’sUltraviolet and Visual Channel of the Wide Field Cam-era 3 (UVIS/WFC3) and the Wide Field and PlanetaryCamera (WFPC2). The UVIS/WFC3 images obtainedfrom the
HST
Data Archive were observed through theF343N and F438W passbands (program ID: GO-15062,PI: N. Bastian), while the WFPC2 images provide thecorresponding observations in the F555W and F814Wpassbands (program ID: GO-8141, PI: R. M. Rich). TheUVIS/WFC3 data set is composed of three frames takenthrough the F343N passband, with exposure times of540 s and 1060 s (twice), as well as three frames takenthrough the F438W passband, with exposure times of120 s and 550 s (twice). For both the F555W and F814Wpassbands, the WFPC2 data set contains four frameseach, for each frame, the exposure time is 400 s.Similarly to our previous papers (e.g., Li et al. 2017),we applied point-spread-function (PSF) photometryto the ‘ flt ’ and ‘ c0f ’ frames based on the stan-dard recipes recommended by the
Dolphot2.0 package(Dolphin. 2011a,b, 2013).
Dolphot2.0 is a photomet-ric package specifically designed for
HST photometricanalysis. We use its WFC3 and WFPC2 modules todeal with the relevant observational data. They in-clude built-in charge-transfer efficiency corrections andphotometric calibration routines such as aperture andzeropoint corrections.
Dolphot2.0 has been validated ultiple populations in NGC 2121
HST photometric analyses (e.g., Monelli et al. 2010).In our photometry, only objects meeting the follow-ing criteria were selected as ‘good’ stars: (1) flaggedby
Dolphot2.0 as a ‘bright star.’ (2) Not centrallysaturated. (3) Sharpness is between − < .
5. Our photometric ap-proach resulted in identification of 19,507 and 10,915stars in the UVIS/WFC3 and WFPC2 frames, respec-tively. We carefully combined both output catalogs bycross-matching the stars in common. Our final, com-bined stellar catalog contains 6856 stars. MAIN RESULTSThe most salient feature of the MPs in most GCs andintermediate-age clusters is the star-to-star variations inlight elements (e.g., C, N, O, Na). Stars in these clus-ters usually have different C, N, and He abundances,and variations in these elements could broaden or splitthe clusters’ RGBs. Specifically, an N spread wouldstrengthen the NH molecular feature at ∼ HST ’s F438W passband includes the CH absorptionfeature ( ∼ − F438W colors owingto their different N (and, to a lesser extent, C) abun-dances. N-rich stars will appear redder than N-poorstars. On the other hand, since N-rich stars should, inprinciple, also be He-rich, they are generally hotter thanHe-poor stars of the same luminosity. Helium variationsamong cluster member stars could thus also be revealedby examination of their F438W − F814W colors. Thisis illustrated in Figure 1: in the top panel we show twomodel spectra for stars at the base of the RGB, with nor-mal and enriched nitrogen abundances ([Fe/H]= − Spectrum2.77 package (Gray & Corbally 1994) and based on theATLAS9 stellar atmosphere models (Kurucz 1970, 1993)We note that the RGB of NGC 2121 is severely con-taminated by a young field-star population in CMDsinvolving the F343N and F438W filters. This problemcan only be ameliorated by introducing additional ob-servations in the F814W passband (i.e., the WFPC2observations). Figure 2 shows three CMDs involvingthe F343N, F438W, and F555W passbands. At first ∼ grayro/spectrum/spectrum.html glance, we find that although the cluster’s RGB is tightin the F438W–F814W vs F438W (middle) and F555W–F814W vs F555W (right) CMDs, it shows a moderatebroadening in the F343N–F814W vs F343N CMD (left).To quantify any broadening of the RGB caused bychemical variations, we simulated multi-band photom-etry of our target cluster. We first used the MESAIsochrone and Stellar Tracks (MIST; Paxton et al. 2011,2013, 2015; Choi et al. 2016; Dotter 2016) models togenerate the best-fitting isochrones representative ofthe observations. Recently, Barker & Paust (2018)reported a problem associated with isochrone fitting to HST photometry in UV–optical–IR passbands. Thisproblem has also been recognized by the community atlarge (e.g., Gontcharov et al. 2019; Howes et al. 2019).Specifically, model parameters determined throughisochrone fitting to the optical–IR CMD cannot be usedto adequately describe photometric measurements in-volving UV passbands. In this paper, we encounteredthe same problem. We found that we were unable toidentify a set of model parameters that allow us tosimultaneously fit all three CMDs in Figure 2. Ourbest-fitting age, metallicity, and distance modulus weredetermined based on the CMD involving the F555W andF814W passbands, resulting in log ( t yr − ) = 9 . ± . − . ± .
10 dex, and ( m − M ) = 18 . ± . E ( B − V ) = 0.09, 0.10, and 0.12 mag (for the CMDsfrom the left to the right in Figure 2). Since this isunphysical, it is more likely that these offsets may havebeen caused by some unknown calibration limitation ofthe MIST models across different photometric systems.We agree with Barker & Paust (2018), who warned that“until the models are fixed, they should not be used forfitting or determining stellar populations in the UV.”Therefore, in this paper we only focus on the width ofthe cluster’s RGB, which should not be affected by themodel’s limitations. We also suggest that the extinctionderived from the optical–IR CMD likely represents themost accurate value, E ( B − V ) = 0.12 mag.Our best-fitting age and metallicity are consistentwith the values obtained by Rich et al. (2001). How- Li & de Grijs
Figure 1. (top) Model spectra for different CNO abundances. The blue spectrum represents stars with ‘normal’ abundances,while the red spectrum represents stars with enhanced nitrogen and depleted carbon and oxygen abundances (∆[N/Fe] = 1.0dex, ∆[C/Fe] = ∆[O/Fe] = − − g values pertain to stars at the base of the RGB.Our model spectra were smoothed with a Gaussian kernel defined by σ = 10˚ A . -1 0 1 2 3181920212223242526 0 1 2181920212223242526 0 0.5 11819202122232425 Figure 2.
NGC 121 CMDs. (left) F343N − F814W vs F343N; (middle) F438W − F814W vs F438W; (right) F555W − F814W vsF555W. The red lines are the best-fitting isochrones. In each panel, the errorbar is on the left side. ultiple populations in NGC 2121 E ( B − V ) = 0 .
07 mag to 0.14 mag (Udalski 1998;Kerber et al. 2007). Our newly derived distance modu-lus to NGC 2121 is slightly smaller than the canonicalLMC value (( m − M ) = 18 .
50 mag; e.g., de Grijs et al.2014). We visually confirmed that our fits adequatelydescribe most of the CMD sequences. The best-fittingisochrones to each of the CMDs are presented in Figure2. To quantify whether NGC 2121 has a broadened RGBcaused by a chemical spread, we need to compare theobserved CMD with that of a simulated SSP. In princi-ple, in addition to chemical spreads, several other fac-tors may also cause a broadening of the RGB, including(i) photometric uncertainties, (ii) photometric artefacts(cosmic rays, bad or hot pixels, etc.), (iii) differencesin distances to individual stars, (iv) differential redden-ing, and (v) field-star contamination. Any broadeningcaused by distance differences to the cluster stars is neg-ligible because of the large distance to the LMC. Photo-metric uncertainties and artefacts can be assessed basedon artificial-star tests. For the images observed witheach camera, we generated 28,000 artificial stars locatedon the best-fitting isochrone between the onset of theSGB and the upper part of the RGB. Their spatial distri-butions were homogeneous. To avoid a situation whereartificial stars dominate the background and crowdinglevels, we only added 100 artificial stars to the raw im-ages at any one time. We used the same photometricmethod to measure these input stars and applied thesame data reduction as used for the observations to theartificial stellar catalog. Finally, we recovered 24,943artificial stars from the WFPC2 frames, correspondingto a completeness level of 89% for the SGB and RGBstars. From the UVIS/WFC3 observations, we recov-ered 27,792 artificial stars, indicating a completenesslevel of close to 99%.The recovered artificial stellar population should beaffected by the same photometric uncertainties and arte-facts as the observations. However, it cannot revealthe level of internal differential reddening. Milone et al.(2012) developed a statistical method to study the red-dening distribution in small areas (such as the core re-gions of GCs). However, this method is not applicable toour target cluster, because the number of stars is small(i.e., less than 10% of the numbers observed for mostGCs; see Milone et al. 2012)). Fortunately, NGC 2121exhibits a tight SGB. In the CMD based on the F555Wand F814W filters, any broadening caused by chemicalinhomogeneities is negligible. Therefore, any additionalbroadening that cannot be reproduced by photometricuncertainties or artefacts must be caused by differential
Figure 3.
Observed SGB of NGC 2121 (top) comparedwith simulated SGBs characterized by different degrees ofdifferential reddening. From the second to the bottom,the amounts of reddening adopted correspond to ∆ A V =0 . , .
04, and 0.08 mag. reddening. As such, we compared the width of the simu-lated SGB to the observations and determined the best-fitting differential reddening level. We conclude that thedegree of differential reddening in NGC 2121 is likely oforder ∆ A V = 0 . ± .
01 (∆ E ( B − V ) = 0 . ± . Li & de Grijs
Figure 4.
Observed MS and MSTO region of NGC 2121(left) compared with the model MS and MSTO includingphotometric uncertainties and differential reddening (right). an eMSTO. In addition, its apparently tight SGB, shownin Figure 3, is also known to represent a coeval stellarpopulation (e.g., Li et al. 2014, 2016). In Figure 4 weshow an example comparison between the observationsand the simulation in the F438W − F814W vs F438WCMD.Because of the large distance to the LMC, using stellarproper motions to reduce field contamination is not pos-sible. In addition, the field of view of the combined ob-servations is very small, rendering selection of an appro-priate field region for reference purposes troublesome.To minimize the impact of field-star contamination, weselected RGB stars according to their distribution intwo of our CMDs (i.e., F555W–F814W vs F555W andF438W–F814W vs F438W). We did not select RGBstars from the CMD involving F343N photometry, be-cause in that section of parameter space RGB stars mayhave been affected by star-to-star variations in N. Weused the simulated CMDs to determine the typical re-gions occupied by the majority of RGB stars and subse-quently used those regions to select RGB stars from ourobservational parameter space: see Figure 5. Only starslocated in the RGB selection boxes in both CMDs wereconsidered RGB stars. As shown by Martocchia et al.(2017), selecting RGB stars from multiple CMDs can beused to effectively reduce field-star contamination.We used a similar method as Monelli et al. (2013) toquantify any broadening of the RGB caused by MPs.We constructed the color index C F343N , F438W , F814W = (F343N − F438W) − (F438W − F814W). This pseudo-color is an effective index to uncover MPs (Monelli et al.2013; Martocchia et al. 2017). We next compared theobservational and simulated (artificial) distributionsof stars in the F438W versus C F343N , F438W , F814W di-agrams: see Figure 6. The simulated diagram on theright includes all 25,000 stars. We also randomly se-lected a subsample of our artificial stars composed of
Figure 5.
Selection of RGB stars from the observations.Only stars located in both selection boxes are consideredcluster member stars. The selection boxes were defined basedon the simulated CMDs (bottom). the same number of stars as in our observations, i.e.,only containing 120 RGB stars (middle panel; red cir-cles). Figure 6 shows that the observed RGB exhibits alarger spread than the simulated SSP RGB. To quantifythis broadening, we adopted the best-fitting isochroneas our fiducial line and calculated the correspondingdeviations of the pseudo-color indices, as illustrated inFigure 7. The distributions of the pseudo-color indicesof the observed and simulated RGB stars are presentedin Figure 7 as well.From inspection of Figure 7, we found that the peakof the observed distribution of the pseudo-colors for thesimulated SSP is not consistent with the observations.Again, this is likely owing to the fitting problem encoun-tered when UV observations are involved. Therefore, weonly focus on the internal spread of the pseudo-color dis-tribution rather than the actual pseudo-color values. Wefound that the distribution of the pseudo-colors for thesimulated SSP can be adequately described by a singleGaussian profile, P (∆ C ) = 0 . e − ( ∆ C − . . ) , (1)with a standard deviation of σ = 0 .
054 mag. The ob-served pseudo-color distribution of the red-giant stars,however, is not well-described by a single Gaussian dis-tribution; it is much broader than the distribution re-sulting from the simulation. If we were to force a singleGaussian profile to fit the distribution, the ‘best-fitting’ ultiple populations in NGC 2121 -2 -1.5 -12020.52121.52222.5 -2 -1.5 -12020.52121.52222.5 -2 -1.5 -12020.52121.52222.5 Figure 6.
F438W vs C F343N , F438W , F814W diagram for the observations (left panel, with the errorbar on the left side) andthe simulations (middle and right). The middle panel shows a simulated subsample (with the same number of stars as theobservations), while the right-hand panel includes all simulated stars.
Figure 7.
F438W vs C F343N , F438W , F814W diagram for theobservations (top left) and our simulations (top right), withtheir ∆( C F343N , F438W , F814W ) color-coded. The solid, dashed,and dash-dotted lines indicate the best-fitting isochrone(∆[C/Fe] = ∆[N/Fe] = ∆[O/Fe] = 0.00 dex) and the locifor ∆[N/Fe] = 0.50 dex (∆[C/Fe] = ∆[O/Fe] = − − C F343N , F438W , F814W ) prob-ability distributions for the observations (red histogram)and the simulation (blue histogram). The distribution of∆( C F343N , F438W , F814W ) for the simulated SSP RGB can bewell described by a singe Gaussian profile (red curve). function would be P (∆ C ) = 0 . e − ( ∆ C − . . ) . (2)The corresponding standard deviation would be σ =0 .
090 mag, about five-thirds that of the simulated SSP. To examine if the more broadened pseudo-color dis-tribution of the observed red-giant stars could simplyhave been caused by small-number statistics, we selected10 subsamples from the simulated stars containing thesame numbers of stars as the observation. All subsam-ples exhibit more dispersed pseudo-color distributionsthan the observations (Figure 8).Using ATLAS9 atmosphere models (Kurucz 1970,1993), we calculated 16 model spectra with atmosphereparameters ([Fe/H], T , log g ) that were the same asthose adopted by the best-fitting isochrones. Thesemodel spectra represent stars between the base and themiddle of the RGB, i.e., they are for ‘standard’ starswith ‘normal’ abundances (i.e., ∆[C/Fe] = ∆[N/Fe] =∆[O/Fe] = 0.00 dex). We then calculated a set of N-enhanced models with ∆[N/Fe] = 0.50 dex, ∆[C/Fe]= ∆[O/Fe] = − − C F343N , F438W , F814W diagramcan be explained by invoking chemical variations fromstandard abundances up to ∆[N/Fe] = 1.0 dex (with∆[C/Fe] = ∆[O/Fe] = − Li & de Grijs
Figure 8.
Age–mass plane for young and intermediate-age clusters and GCs with and without MPs; data havebeen derived from McLaughlin & van der Marel (2005);Baumgardt et al. (2013); Krause et al. (2016). Blue squaresare clusters without MPs (or where the presence of MPs isunclear). Red circles are clusters with MPs. The red penta-gram indicates the locus of NGC 2121. standard isochrone to meet the peak of the pseudo-colordistribution, an internal abundance spread of ∆[N/Fe]= 0.5 dex would still be required. We suggest that theactual N abundance spread among the RGB stars inNGC 2121 might be in the range 0.5–1.0 dex. Since theobserved pseudo-color dispersion of the red-giant starscannot be explained by photometric uncertainties, arte-facts, differential reddening, or small-number statistics,the only viable explanation is light-element star-to-starvariations, i.e., the RGB of NGC 2121 appears to becomposed of MPs.In Figure 8 we have added NGC 2121 to the age–mass plane for clusters with and without MPs, whichrepresents a summary of results from the literature asto whether MPs occur in GCs and their younger mas-sive counterparts of different ages and masses. Almostall old GCs show MPs, while their younger counter-parts (younger than 2 Gyr) do not. In the age rangeof 2–10 Gyr, clusters with and without MPs overlap inboth age and mass. The mass of NGC 2121, ∼ M ⊙ (McLaughlin & van der Marel 2005), is only half that ofNGC 1978 ( ∼ × M ⊙ ; Baumgardt et al. 2013), andits mass is comparable to most of its younger counter-parts. DISCUSSION AND SUMMARYIn this paper, we have analyzed the photometric ap-pearance of the NGC 2121 RGB in the F438W ver-sus C F343N , F438W , F814W diagram, which has been estab-lished as a diagnostic plot suitable for uncovering thepresence of MPs. A broadening of the cluster’s RGB isapparent when compared with that of a simulated SSP.It is consistent with a model characterized by differentCNO abundances, implying star-to-star chemical varia-tions among the cluster’s red-giant stars. We have determined that F343N is indeed a key pass-band for use to unveil the presence of different stellarpopulations. We divided our selected RGB stars intotwo subsamples based on the dividing line halfway be-tween the standard isochrone and the locus pertainingto ∆[N/Fe] = 0.5 dex. We explored their color distri-butions in other CMDs, as shown in Figure 9. Indeed,both subsamples are fully mixed in the F555W − F814Wvs F555W CMD, but they exhibit distinct color dif-ferences in the CMD involving the F343N and F814Wpassbands. Stars with larger C F343N , F438W , F814W indiceshave redder F343N − F814W colors, indicating that theirtotal fluxes in the F343N passband are much lower thanthose of their blue counterparts, i.e., they are N-enrichedstars. Our detection of chemical variations among thered-giant stars in NGC 2121 also makes it the second-youngest cluster with MPs (after the 2.2 Gyr-old clusterNGC 1978).It would be interesting to examine whether the pris-tine and enriched stellar populations in the cluster havedifferent central concentrations. However, the spatialdistribution of stars across the WF3 chip of the WFPC2camera exhibits numerous areas with little or no stars,indicating that source confusion caused by the spikes as-sociated with bright stars in this region is severe. There-fore, the corresponding stellar completeness varies sig-nificantly at different radii. High-quality WFC3 obser-vations of a larger field of view, through the F814Wpassband, are required to resolve this problem.It is useful to compare the physical properties ofNGC 2121 with those of its younger counterpartswithout MPs, i.e., NGC 1783, NGC 1806, and NGC1846 (Mucciarelli et al. 2014; Martocchia et al. 2018;Zhang et al. 2018). The masses of NGC 1783, NGC1806, NGC 1846, and NGC 2121 are all similar( ∼ . × M ⊙ , 10 M ⊙ , 1 . × M ⊙ , and 10 M ⊙ , respectively; McLaughlin & van der Marel 2005;Baumgardt et al. 2013), as are their internal struc-tural parameters (such as their core, half-mass, andtidal radii; McLaughlin & van der Marel 2005; Li et al.2018). These clusters also have similar metallicitiesand distances to the LMC’s bar region (Li et al. 2018).Thus, both the external environments and the inter-nal dynamical properties of NGC 2121 and its youngercounterparts are similar, which implies that the pres-ence of MPs in NGC 2121 is unlikely caused by anyspecific formation environment or internal dynamics.A noticeable difference with NGC 1783, NGC 1806,and NGC 1846 is that these younger clusters exhibitapparent eMSTO regions while NGC 2121 and NGC1978 Martocchia et al. (2018) do not. Martocchia et al.(2018) proposed that the apparent chemical anomalies ultiple populations in NGC 2121 -2 -1.8 -1.620.52121.52222.5 0.9 1 1.119.52020.52121.5 1.8 2 2.2 2.42121.522 Figure 9.
NGC 2121 RGB in the F438W vs C F343N , F814W , F814W diagram (left). Stars are divided into two subsamples basedon the dividing line halfway between the standard isochrone and the locus for ∆[N/Fe] = 0.5 dex (see the pink and blue circles).The middle and right-hand panels present the same red-giant stars in the other CMDs. might be a specific feature of stars with masses below1.5 M ⊙ , which is roughly the mass of main-sequenceturnoff stars at an age of ∼ M ⊙ will exhibit evidenceof rapid rotation in their cluster CMD, as manifested byeMSTO regions. Although the details are still unclear,strong magnetic fields may play a role in the appearanceof star-to-star chemical variations. This notion is sup-ported by our results, since the masses of the RGB andMSTO stars in NGC 2121 have decreased to below thiscritical mass. All of these stars should possess strongmagnetic fields.The detection of MPs in NGC 2121 underpins the hy-pothesis that age may be an important factor control-ling the presence of MPs. The transition period betweenclusters with and without MPs should occur at an ageof 2–3 Gyr. The mass of NGC 2121 is comparable tothose of most Galactic GCs exhibiting MPs. It is unclearwhether cluster mass also plays a role in the appearanceof MPs. However, if so, it should not be the only factorof importance, because otherwise the detection of MPsin younger massive clusters should also be expected.This conclusion is consistent with that of Zhang et al. (2018). It is important to search for MPs in other clus-ters of similar ages but lower masses than NGC 2121,such as NGC 2193 and ESO-56-SC40 (Baumgardt et al.2013). If these clusters exhibit similar CMD features asNGC 1978 and NGC 2121, this would lead to the con-clusion that mass is not a crucial parameter determiningthe presence of MPs. Future studies should then focuson which intrinsic transitions may have led to chemicalstar-to-star variations at ages between 2 and 3 Gyr. Onthe other hand, if such clusters do not exhibit MPs, thiswould imply that mass may still be a secondary param-eter controlling the presence of MPs.C. L. was supported by the Macquarie Research Fel-lowship Scheme. This work was also partly supportedby the National Natural Science Foundation of Chinathrough grants U1631102, 11373010, 11633005, and11803048. Facilities:
Hubble Space Telescope (UVIS/WFC3and WFPC2)
Software: dolphot2.0 (Dolphin. 2011a,b, 2013)Spectrum v2.77 (Gray & Corbally 1994)
REFERENCES
Bastian, N., & Lardo, C. 2018, ARA&A, 56, 83Barker, H., & Paust, N. E. Q. 2018, PASP, 130, 034204Baumgardt, H., Parmentier, G., Anders, P., & Grebel,E. K. 2013, MNRAS, 430, 676Bekki, K. 2017, MNRAS, 469, 2933Cannon, R. D., Croke, B. F. W., Bell, R. A., Hesser, J. E.,& Stathakis, R. A. 1998, MNRAS, 298, 601 Cardini, D., & Cassatella, A. 2007, ApJ, 666, 393Carretta, E., Bragaglia, A., Gratton, R. G., et al. 2009,A&A, 505, 117Choi, J., Dotter, A., Conroy, C., et al. 2016, ApJ, 823, 102Cordoni, G., Milone, A. P., Marino, A. F., et al. 2018, ApJ,869, 139 Li & de Grijs
Decressin, T., Meynet, G., Charbonnel, C., Prantzos, N., &Ekstr¨om, S. 2007, A&A, 464, 1029de Grijs, R., Wicker, J. E., & Bono, G. 2014, AJ, 147, 122Dalessandro, E., Lapenna, E., Mucciarelli, A., et al. 2016,ApJ, 829, 77Denissenkov, P. A., & Hartwick, F. D. A. 2014, MNRAS,437, L21D’Ercole, A., Vesperini, E., D’Antona, F., McMillan,S. L. W., & Recchi, S. 2008, MNRAS, 391, 825Dolphin A., DOLPHOT/WFC3 user’s guide, version 2.0.http://americano.dolphinsim.com/dolphin/dolphotWFC3.pdfDolphin A., DOLPHOT/WFPC2 user’s guide, version 2.0.http://americano.dolphinsim.com/dolphot/dolphotWFPC2.pdfDolphin A., DOLPHOT user’s guide, version 2.0.http://americano.dolphinsim.com/dolphot/dolphot.pdfDotter, A. 2016, ApJS, 222, 8Gray, R. O., & Corbally, C. J. 1994, AJ, 107, 742Gontcharov, G. A., Mosenkov, A. V., & Khovritchev, M. Y.2019, MNRAS, 483, 4949Howes, L. M., Lindegren, L., Feltzing, S., Church, R. P., &Bensby, T. 2019, A&A, 622, A27Kerber, L. O., Santiago, B. X., & Brocato, E. 2007, A&A,462, 139Krause, M. G. H., Charbonnel, C., Bastian, N., & Diehl, R.2016, A&A, 587, A53Kurucz, R. L. 1970, SAO Special Rep., 309, 309Kurucz, R. 1993, ATLAS9 Stellar Atmosphere Programsand 2 km s −1