Photometric characterization of multiple populations in star clusters: The impact of the first dredge-up
Maurizio Salaris, Chris Usher, Silvia Martocchia, Emanuele Dalessandro, Nate Bastian, Sara Saracino, Santi Cassisi, Ivan Cabrera-Ziri, Carmela Lardo
aa r X i v : . [ a s t r o - ph . GA ] J a n MNRAS , 1–7 (2015) Preprint 14 January 2020 Compiled using MNRAS L A TEX style file v3.0
Photometric characterization of multiple populations instar clusters: The impact of the first dredge-up
Maurizio Salaris, ⋆ Chris Usher, Silvia Martocchia, , Emanuele Dalessandro, Nate Bastian, Sara Saracino, Santi Cassisi, , Ivan Cabrera-Ziri, † and Carmela Lardo Astrophysics Research Institute, Liverpool John Moores University, 146 Brownlow Hill, Liverpool L3 5RF, UK European Southern Observatory, Karl-Schwarzschild-Strasse 2, D-85748 Garching bei M¨unchen, Germany INAF-Osservatorio di Astrofisica e Scienza dello Spazio di Bologna, Via Gobetti 93/3, Bologna, II-40129, Italy INAF-Osservatorio Astronomico d’Abruzzo, via M. Maggini, sn. 64100, Teramo, Italy INFN - Sezione di Pisa, Largo Pontecorvo 3, 56127 Pisa, Italy Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA Laboratoire d’astrophysique, Ecole Polytechnique F´ed´erale de Lausanne (EPFL), Observatoire de Sauverny, CH-1290 Versoix, Switzerland
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
The existence of star-to-star light-element abundance variations (multiple populations,MPs) in massive Galactic and extragalactic star clusters older than about 2 Gyr is bynow well established. Photometry of red giant branch (RGB) stars has been and stillis instrumental in enabling the detection and characterization of cluster MPs, throughthe appropriate choices of filters, colours and colour combinations, that are mainlysensitive to N and –to a lesser degree– C stellar surface abundances. An importantissue not yet properly addressed is that the translation of the observed widths of thecluster RGBs to abundance spreads must account for the effect of the first dredge-upon the surface chemical patterns, hence on the spectral energy distributions of starsbelonging to the various MPs. We have filled this gap by studying theoretically theimpact of the dredge-up on the predicted widths of RGBs in clusters hosting MPs.We find that for a given initial range of N abundances, the first dredge up reducesthe predicted RGB widths in N-sensitive filters compared to the case when its effecton the stellar spectral energy distributions is not accounted for. This reduction is astrong function of age and has also a dependence on metallicity. The net effect is anunderestimate of the initial N-abundance ranges from RGB photometry if the firstdredge-up is not accounted for in the modelling, and also the potential determinationof spurious trends of N-abundance spreads with age.
Key words: convection – galaxies: star clusters: general – stars: abundances –Hertzsprung-Russell and colour-magnitude diagrams
During the last 10-15 years both spectroscopic and photo-metric observations have definitely established that individ-ual Galactic globular clusters (GCs) host multiple popula-tions (MPs) of stars, characterised by C-N, O-Na (and also,but not always, Mg-Al) anticorrelations and He abundancespreads (see, e.g. Gratton et al. 2012; Milone et al. 2017,2018; Bastian & Lardo 2018; Gratton et al. 2019). Scenar-ios for the origin of MPs (reviewed, e.g, by Renzini et al.2015; Bastian & Lardo 2018) generally invoke more than oneepisode of star formation, envisaging that stars with CNONa ⋆ E-mail: [email protected] † Hubble fellow (and He) abundance ratios similar to those observed in halofield stars are the first objects to form (P1 stars), whilststars enriched in N and Na (and He) and depleted in C andO formed later (P2 stars). These P2 stars are supposed toform out of chemically processed material ejected by someclass of massive P1 stars, usually denoted as polluters. Todate, none of the proposed polluters can explain quantita-tively the full ensemble of chemical patterns observed in in-dividual GCs (Renzini et al. 2015; Bastian & Lardo 2018).Photometric (see, e.g., Larsen et al. 2014;Dalessandro et al. 2016; Gilligan et al. 2019; Lagioia et al.2019; Nardiello et al. 2019, and references therein) andto a lesser extent spectroscopic (Mucciarelli et al. 2009)observations have also shown that MPs are not confinedonly to Galactic GCs, but are present also in old clusters © M. Salaris et al. of the Magellanic Clouds, Fornax and M31. Integratedspectroscopy of old extragalactic clusters in M31 alsoconfirms the signature of MPs amongst old massive stellarclusters (see, e.g., Schiavon et al. 2013)A further recent development has seen the detectionof MPs in resolved massive extragalactic clusters downto ages of ∼ pseudocolours )are sensitive to the abundance of mainly nitrogen (pluscarbon and oxygen to a much lesser extent) in the atmo-spheres of the target stars, and can clearly detect the pres-ence of MPs (see, e.g., Monelli et al. 2013; Piotto et al. 2015;Milone et al. 2017; Niederhofer et al. 2017b; Salaris et al.2019). Due to the distance of the targets, red giant branch(RGB) stars are typically observed for photometric (and alsospectroscopic) MP detection.By analyzing results for a number of MagellanicCloud (MC) clusters covering a large range of ages (2-8 Gyr), Martocchia et al. (2018) and Martocchia et al.(2019) – hereinafter M19– found in their sample a gen-eral trend between the measured width of a cluster RGBand the cluster age. They considered the pseudocolours C F N , F W , F W ≡ ( F N − F W ) − ( F W − F W ) and C F W , F W , F N ≡ ( F W − F W ) − ( F W − F N ) in the filter systems of the WFC3 and ACS (for F W ) cameras on board the Hubble Space Telescope (HST) –both sensitive to the abundance of N in the stel-lar spectra– and found that the RGB width shows a generalincrease with increasing age in their cluster sample. A nat-ural explanation for this occurrence is that the N spread inthese massive clusters increases with increasing cluster age.However, there is an important additional phenomenonto consider when translating the observed RGB widths toN abundance spreads, that has been so far largely un-explored. The samples of cluster stars considered in M19are distributed between the base of the RGB and approx-imately the RGB bump. This range covers almost the en-tire evolution through the first dredge-up (FDU – see, e.g.,Karakas & Lattanzio 2014; Salaris et al. 2015, and refer-ences therein), that starts on the subgiant branch and endsbelow the RGB bump level. During the FDU the surfaceN abundance increases compared to the initial value, andfrom basic stellar physics we expect this increase to de-pend on the initial nitrogen abundances. The variation of the surface abundances due to the FDU impacts the stellarspectral energy distributions, hence the predicted coloursand pseudocolours sensitive to this element. The upshotis that the observed RGB widths in M19 are determinedby a combination of the initial N spreads plus the effectof the dredge up. This is true for any colour or pseudo-colour sensitive to the surface N abundance in RGB stars,like C F W , F W , F W ≡ ( F W − F W ) − ( F W − F W ) (see, e.g., Milone et al. 2017; Saracino et al. 2019),or C F W , F N , F W ≡ ( F W − F N ) − ( F N − F W ) (Zennaro et al. 2019), and C F W , F N , F W ≡( F W − F N ) − ( F N − F W ) (Milone et al. 2019),devised to detect MPs in GCs and younger MC massiveclusters by means of the so-called ‘chromosome maps’.In this paper we explore this issue showing quali-tatively, and in case of M19 results also quantitatively,the important role played by the FDU when translat-ing to N abundance spreads the observed RGB widths inmagnitude-pseudocolour diagrams and chromosome maps.Section 2 presents our theoretical calculations and the ef-fect of the FDU on the surface abundances of MP stars,followed in Sect. 3 by an analysis of the impact of theFDU on the C F W , F W , F W , C F W , F N , F W and C F W , F N , F W pseudocolours used in the chromosomemaps to disentangle cluster MPs, plus a quantitative esti-mate of the effect on the C F N , F W , F W pseudocolouremployed by M19. Conclusions follow in Sect. 4. We have employed the BaSTI isochrones (Pietrinferni et al.2004) for two P1 solar scaled chemical compositions with[Fe/H]= − − − − MNRAS , 1–7 (2015) irst dredge-up and multiple populations Figure 1.
Initial value ( ∆ [ N / Fe ] ini =0.8 – dashed line) of the dif-ference in the surface [N/Fe] ratio for a set of bimodal populationswith [Fe/H]= − ∆ [ N / Fe ] FDU in the surface abundances after the FDU (solidline – see text for details). towards the surface, leaving behind a chemical discontinuity.This discontinuity is eventually crossed by the advancing (inmass) H-burning shell, causing the appearance of the RGBbump in the luminosity function of old stellar populations(see, e.g., Cassisi & Salaris 2013). The change of surface ni-trogen abundance (and carbon, whilst the oxygen abundanceis essentially never altered in the age range investigated here)during the FDU depends on the mass of the star (hence thepopulation age, see e.g. Salaris et al. 2015, and referencestherein), but also on the initial abundance pattern. The rea-son is that during the FDU the convective envelope reacheslayers where the abundances of C and N attained the equi-librium values of the CN cycle during the main sequence.The equilibrium abundance of N is typically higher (andthe C abundance lower) than the standard solar scaled (or α -enhanced) counterpart for a given total metallicity, hencethe FDU causes an increase of surface N (and decrease of C).When the initial metal mixture is N-enhanced (and carbondepleted) the equilibrium abundance of N (and C) becomesmore comparable to the initial one, and the effect of theFDU is much less appreciable or negligible.Figure 1 shows the run with age (ages between 1.0 and13.5 Gyr) of ∆ [N/Fe], defined as the difference of surface[N/Fe] between a population with with N-enhanced P2 com-position and a coeval one with P1 composition ([Fe/H]= − ∆ [ N / Fe ] ini (the same for all ages) and thecorresponding surface abundance differences at the comple-tion of the FDU ( ∆ [ N / Fe ] FDU ).The values of ∆ [ N / Fe ] FDU are lower than ∆ [ N / Fe ] ini ,and display a clear trend with age, despite the fact that ∆ [ N / Fe ] ini is the same at all ages. For the younger pop-ulations ∆ [ N / Fe ] FDU is much smaller than ∆ [ N / Fe ] ini , but with increasing age it comes progressively closer to its ini-tial value. The reason is that in RGB models with P1 initialN abundance, the surface [N/Fe] at the end of the FDUincreases with decreasing age (see, e.g. Salaris et al. 2015),whilst the impact of the FDU is much smaller or negligi-ble in N-enhanced populations. The effect of the FDU onthe surface carbon abundance is also small or negligible inP2 models with initial C-depleted abundances, whilst in P1models the surface [C/Fe] at the end of the FDU gets pro-gressively lower with decreasing age.Summarizing, for bimodal coeval populations with fixed ∆ [ N / Fe ] ini (independent of age) and ages between 1.0 and13.5 Gyr, the difference ∆ [ N / Fe ] FDU measured at the endof the FDU is predicted to be lower than the initial value,showing a trend with age – ∆ [ N / Fe ] FDU decreasing for de-creasing age. This general behaviour is true irrespective ofthe exact value of [Fe/H].
We start here discussing the impact of the FDUon the representative N-sensitive C F W , F W , F W , C F W , F N , F W and C F W , F N , F W pseudo-colours used in the chromosome maps to detect MPs fromcluster photometry. In a chromosome map the total width ofthe cluster RGB in one of these pseudocolors is normalizedto the value taken two magnitudes above the main sequenceturnoff in the F W filter (Milone et al. 2017). This is alevel where typically the FDU has either already started oris essentially completed, but still below the RGB bump, be-yond which extra mixing processes that further affect thesurface C and N abundances appear to be efficient, at leastin low mass stars (see, e.g., Lagarde et al. 2019, and refer-ences therein).As long as cluster ages are of the order of 10-13 Gyr,the effect of the FDU on the surface abundances –hencethe RGB C F W , F W , F W , C F W , F N , F W and C F W , F N , F W values– is basically negligible, espe-cially at low metallicity. The situation is however differ-ent for intermediate-age clusters, as shown in Figs. 2, 3and 4 that display M F W - C F W , F W , F W , M F W − C F W , F N , F W and M F W − C F W , F N , F W di-agrams for two P1-P2 bimodal populations (at a represen-tative [Fe/H]= − F W filter) is also marked.At the beginning of the FDU the pseudocolour dif-ference between P1 and P2 RGBs is essentially the samefor these two ages, in all three diagrams; when moving tobrighter magnitudes the FDU progresses, causing a reduced[N/Fe] (and [C/Fe]) difference between P1 and P2 stars ata given brightness, and a decreased separation of the se-quences. The effect is stronger at younger ages, because ofthe increasing impact of the FDU on the surface abundances(see Fig. 1). Clearly, any interpretation in terms of ∆ [ N / Fe ] ini of the RGB widths in the chromosome maps must account MNRAS000
We start here discussing the impact of the FDUon the representative N-sensitive C F W , F W , F W , C F W , F N , F W and C F W , F N , F W pseudo-colours used in the chromosome maps to detect MPs fromcluster photometry. In a chromosome map the total width ofthe cluster RGB in one of these pseudocolors is normalizedto the value taken two magnitudes above the main sequenceturnoff in the F W filter (Milone et al. 2017). This is alevel where typically the FDU has either already started oris essentially completed, but still below the RGB bump, be-yond which extra mixing processes that further affect thesurface C and N abundances appear to be efficient, at leastin low mass stars (see, e.g., Lagarde et al. 2019, and refer-ences therein).As long as cluster ages are of the order of 10-13 Gyr,the effect of the FDU on the surface abundances –hencethe RGB C F W , F W , F W , C F W , F N , F W and C F W , F N , F W values– is basically negligible, espe-cially at low metallicity. The situation is however differ-ent for intermediate-age clusters, as shown in Figs. 2, 3and 4 that display M F W - C F W , F W , F W , M F W − C F W , F N , F W and M F W − C F W , F N , F W di-agrams for two P1-P2 bimodal populations (at a represen-tative [Fe/H]= − F W filter) is also marked.At the beginning of the FDU the pseudocolour dif-ference between P1 and P2 RGBs is essentially the samefor these two ages, in all three diagrams; when moving tobrighter magnitudes the FDU progresses, causing a reduced[N/Fe] (and [C/Fe]) difference between P1 and P2 stars ata given brightness, and a decreased separation of the se-quences. The effect is stronger at younger ages, because ofthe increasing impact of the FDU on the surface abundances(see Fig. 1). Clearly, any interpretation in terms of ∆ [ N / Fe ] ini of the RGB widths in the chromosome maps must account MNRAS000 , 1–7 (2015)
M. Salaris et al.
Figure 2.
P1 (solid lines) and P2 (dashed lines) RGBs (at[Fe/H]= − M F W − C F W , F W , F W diagramfor ages equal to 3 (upper panel) and 6 (lower panel) Gyr, ∆ [ N / Fe ] ini =0.8, accounting for the effect of FDU on the spectralenergy distributions. The dotted lines display P1 RGBs calculatedwithout considering the variation of N (and C) due to the FDU.The horizontal thin lines mark the level corresponding to twomagnitudes above the main sequence turnoff, where the width ofthe RGB is taken in the chromosome maps (see text for details). for the effect of the FDU on the surface [N/Fe] and pseudo-colours, for ages lower than typical GC ages.To give at least one quantitative example of therole played by the FDU when inferring and/or com-paring ∆ [ N / Fe ] ini values amongst different clusters us-ing N-sensitive photometric properties, we consider the C F N , F W , F W pseudocolour employed by M19. Fig-ure 5 shows theoretical RGBs for two bimodal P1-P2 pop-ulations with ages equal to 6 and 13.5 Gyr at a repre-sentative [Fe/H]= − ∆ [ N / Fe ] ini =0.8, plotted in the M F W − C F N , F W , F W diagram used by M19. Thedisplayed range of M F W magnitudes corresponds approx-imately to the range employed in M19 analysis, that roughlyencompasses the entire development of the FDU.As for the cases discussed before, at a given age the C F N , F W , F W separation between P1 and P2 RGBsdecreases with decreasing magnitude, due to the effect of theFDU on the surface N abundances. The variation (decrease)of C plays a much smaller role, but has the same qualitativeeffect of the increase of N, that is to shift the RGB to largervalues of C F N , F W , F W .The dotted lines display the P1 RGBs that do not ac-count for the effect of the FDU on the spectral energy dis-tributions due to the change of the surface abundances. At13.5 Gyr the no-FDU RGB is almost coincident with theFDU case, because at this metallicity and age its effect onthe surface abundances is very small also for the P1 compo-sition. For the 6 Gyr case, the no-FDU P1 RGB runs parallel Figure 3.
As Fig. 2 but for the M F W − C F W , F N , F W diagram. Figure 4.
As Fig. 2 but for the M F W − C F W , F N , F W diagram. to the P2 one, and diverges steadily from the calculationsthat include the FDU.To assess quantitatively the impact of the FDU onthe interpretation of M19 results, we have considered P1-P2 pairs of RGBs with ∆ [ N / Fe ] ini =0.8, ages between 3 and13.5 Gyr for [Fe/H]= − − MNRAS , 1–7 (2015) irst dredge-up and multiple populations Figure 5.
P1 and P2 RGBs (at [Fe/H]= − M F W − C F N , F W , F W diagram for 6 (solid lines) and 13.5 (dashedlines) Gyr, ∆ [ N / Fe ] ini =0.8, accounting for the FDU (that has anegligible effect on P2 RGBs). The dotted and dash-dotted bluelines display P1 RGBs calculated without considering the varia-tion of N (and C) due to the FDU, for 6 and 13.5 Gyr, respectively(see text for details). row mass range involved– in the M F W range of Fig. 5.We have then added to each synthetic star photometrya typical Gaussian photometric error taken from the ar-tificial star tests, that is approximately the same for allstars of the cluster sample, in this representative magni-tude range. The number of synthetic stars is roughly thesame as the number of stars employed in M19 analysis.For each pair of P1-P2 stars we have then calculated the C F N , F W , F W distribution and determined the 1 σ dispersion σ ( C F N , F W , F W ) RGB –as in M19– repeat-ing the procedure 100 times to determine its average value.The top panel of Fig. 6 displays our theoretical σ ( C F N , F W , F W ) RGB as a function of age, togetherwith σ ( C F N , F W , F W ) RGB measured by M19 for theirsample of Magellanic Cloud clusters that show MPs, plusM19 determinations for the Milky Way GCs 47 Tuc, M15and NGC2419. These bimodal MPs with 1:1 ratio and aconstant ∆ [ N / Fe ] ini =0.8 display a trend with age qualita-tively similar to the observations. This is due to the effectof the FDU on the surface N abundances discussed above.However, a constant ∆ [ N / Fe ] ini with age does not matchthe observed average slope, for a linear fit to the observed σ ( C F N , F W , F W ) RGB values gives a slope equal to0.0060 ± − − ∆ [ N / Fe ] ini , and a general increase of ∆ [ N / Fe ] ini withage. To clarify this point, the same figure displays also Figure 6.
Top panel:
Theoretical σ ( C F N , F W , F W ) RGB as a function of age for [Fe/H]= − − ∆ [ N / Fe ] ini =0.8. The dash-dotted line displaysthe results for ∆ [ N / Fe ] ini =1.1 and [Fe/H]= − σ ( C F N , F W , F W ) RGB values de-termined by M19 for a sample of MC clusters with photomet-ric detection of MPs (filled red triangles for clusters with [Fe/H]around − − Bottom panel:
The same as the toppanel but with theoretical σ ( C F N , F W , F W ) RGB valuescalculated for ∆ [ N / Fe ] ini =0.8 without accounting for the effect ofthe FDU on the spectral energy distributions. the σ ( C F N , F W , F W ) RGB values for the same typeof bimodal MPs (1:1 ratio between P1 and P2 stars),but calculated considering P2 models with ∆ [ N / Fe ] ini =1.1(and [C/Fe]=[O/Fe]= − ∆ [ N / Fe ] ini =1.1 that roughlyagrees with the range measured spectroscopically in 47 Tuc(Carretta et al. 2005; Marino et al. 2016).The theoretical σ ( C F N , F W , F W ) RGB values atfixed age, [Fe/H] and ∆ [ N / Fe ] ini depend of course on thestatistical distribution of the [N/Fe] abundances withinthe prescribed range. Still considering a bimodal distri-bution in terms of [N/Fe], when changing the P1/P2 ra-tio from 1:1 to a population with 70% P1 stars and30% P2 stars with ∆ [ N / Fe ] ini either 0.8 or 1.1 dex, the σ ( C F N , F W , F W ) RGB dispersions are reduced by onlya few 0.001 mag. A larger reduction (by up to 0.03 mag in thecase of ∆ [ N / Fe ] ini =1.1) is found when the [N/Fe] abundancesare distributed uniformly within the prescribed ∆ [ N / Fe ] ini range, a sort of extreme opposite case compared to a 1:1 bi-modal [N/Fe] distribution. However, even in this case the in-crease of σ ( C F N , F W , F W ) RGB with age is preserved.Equal-weight multiple subpopulations quantized in terms of[N/Fe] –depending on their number– lead to intermediateresults between the bimodal 1:1 population, and the case ofuniform [N/Fe] distribution.
MNRAS , 1–7 (2015)
M. Salaris et al.
The lower panel of Fig. 6 displays the theoretical σ ( C F N , F W , F W ) RGB values for the same bimodalMPs calculated with ∆ [ N / Fe ] ini =0.8 and a 1:1 ratio betweenP1 and P2 stars, but without FDU. The absolute values andtrends of σ ( C F N , F W , F W ) RGB with age are differentfrom the ‘correct ’ ones that include the effect of the FDU.The σ ( C F N , F W , F W ) RGB values are higher than theFDU case at the younger ages, showing a general anticorre-lation with age. This of course impacts the determination of ∆ [ N / Fe ] ini from the measured σ ( C F N , F W , F W ) RGB values in a sample of clusters.
The impact of the FDU on the observed width of RGBsin old and intermediate-age clusters hosting MPs has beenso far largely unexplored. We have addressed this issue byconsidering several N-sensitive pseudocolours employed todetect MPs in clusters of various ages. In all cases, for agiven initial difference ∆ [ N / Fe ] ini between coeval P1 and P2populations, the effect of the FDU is to reduce the predictedRGB width compared to the case when the effect of theFDU on the spectral energy distribution is neglected. Thereduction is a function of age, and has also some dependenceon [Fe/H]. These effects stem from the dependence of theFDU variations of the surface N and, to a lesser degree,C abundances, on age and metallicity in models with P1compositions.In the specific case of the pseudocolour employed byM19, when the FDU is accounted for, a constant ∆ [ N / Fe ] ini produces a general increase with age of the predicted σ ( C F N , F W , F W ) RGB , qualitatively similar with whatis observed, but with a shallower slope. The observed trendwith cluster age can be matched only by a combination ofboth the effect of the FDU at constant ∆ [ N / Fe ] ini , and ageneral increase of ∆ [ N / Fe ] ini with age. When the theoreti-cal spectral energy distributions do not account for the FDU,the values of ∆ [ N / Fe ] ini required to match the observed RGBwidths will be smaller. This effect only becomes significantfor ages below ∼
10 Gyr, when the FDU starts to alter moresignificantly the surface chemical composition– see, e.g., thecase of NGC1978 in Fig. 6.The FDU affects also the quantitative interpretation,in terms N-abundance spreads, of the chromosome maps,when employed to identify MPs in intermediate-age clus-ters. In the case of GCs, due to their old ages, the FDUis largely unable to affect appreciably the surface N andC abundances, hence its effect on the C F W , F W , F W , C F W , F N , F W , or C F W , F N , F W pseudocoloursis negligible. At younger ages the FDU can affect the pre-dicted values much more appreciably. Hence, also in caseof the chromosome maps, neglecting the FDU abundancechanges can lead to an underestimate of the initial [N/Fe]spread for a given measured value of the RGB width.The surface chemical changes due to the FDU playtherefore an important role in the interpretation of the ob-served width of RGBs in intermediate age clusters. It needsto be properly accounted for when determining initial N-abundance spreads from photometry, and potential correla-tions with age or other cluster parameters. ACKNOWLEDGEMENTS
CU, NB, and SM gratefully acknowledge financial supportfrom the European Research Council (ERC-CoG-646928,Multi-Pop). NB also acknowledges financial support fromthe Royal Society (University Research Fellowship). Sup-port for ICZ was provided by NASA through Hubble Fel-lowship grant HST-HF2-51387.001-A awarded by the SpaceTelescope Science Institute, which is operated by the As-sociation of Universities for Research in Astronomy, Inc.,for NASA, under contract NAS5-26555. SC acknowledgessupport from Premiale INAF MITiC, from INFN (Inizia-tiva specifica TAsP), and grant AYA2013- 42781P from theMinistry of Economy and Competitiveness of Spain.
REFERENCES
Bastian N., Lardo C., 2018, ARA&A, 56, 83Bastian N., et al., 2019, MNRAS, 489, L80Bertelli Motta C., Salaris M., Pasquali A., Grebel E. K., 2017,MNRAS, 466, 2161Carretta E., Gratton R. G., Lucatello S., Bragaglia A., BonifacioP., 2005, A&A, 433, 597Cassisi S., Salaris M., 2013, Old Stellar Populations: How to Studythe Fossil Record of Galaxy FormationCassisi S., Mucciarelli A., Pietrinferni A., Salaris M., Ferguson J.,2013, A&A, 554, A19Dalessandro E., Lapenna E., Mucciarelli A., Origlia L., FerraroF. R., Lanzoni B., 2016, ApJ, 829, 77Dalessandro E., et al., 2018, A&A, 618, A131Gilligan C. K., et al., 2019, MNRAS, 486, 5581Gratton R. G., Carretta E., Bragaglia A., 2012, A&ARv, 20, 50Gratton R., Bragaglia A., Carretta E., D’Orazi V., Lucatello S.,Sollima A., 2019, A&ARv, 27, 8Hollyhead K., et al., 2017, MNRAS, 465, L39Hollyhead K., et al., 2019, MNRAS, 484, 4718Karakas A. I., Lattanzio J. C., 2014,Publ. Astron. Soc. Australia, 31, e030Kurucz R. L., 1970, SAO Special Report, 309Kurucz R. L., 2005, Memorie della Societa Astronomica ItalianaSupplementi, 8, 14Kurucz R. L., Avrett E. H., 1981, SAO Special Report, 391Kurucz R. L., Furenlid I., 1979, SAO Special Report, 387Lagarde N., et al., 2019, A&A, 621, A24Lagioia E. P., Milone A. P., Marino A. F., Dotter A., 2019, ApJ,871, 140Larsen S. S., Brodie J. P., Grundahl F., Strader J., 2014, ApJ,797, 15Marino A. F., Villanova S., Piotto G., Milone A. P., Momany Y.,Bedin L. R., Medling A. M., 2008, A&A, 490, 625Marino A. F., et al., 2016, MNRAS, 459, 610Martocchia S., et al., 2017, MNRAS, 468, 3150Martocchia S., et al., 2018, MNRAS, 473, 2688Martocchia S., et al., 2019, MNRAS,Milone A. P., et al., 2017, MNRAS, 464, 3636Milone A. P., et al., 2018, MNRAS, 481, 5098Milone A. P., et al., 2019, arXiv e-prints, p. arXiv:1910.09683Monelli M., et al., 2013, MNRAS, 431, 2126Mucciarelli A., Origlia L., Ferraro F. R., Pancino E., 2009, ApJ,695, L134Mucciarelli A., et al., 2016, ApJ, 824, 73Nardiello D., Piotto G., Milone A. P., Rich R. M., Cassisi S.,Bedin L. R., Bellini A., Renzini A., 2019, MNRAS, 485, 3076Niederhofer F., et al., 2017a, MNRAS, 464, 94Niederhofer F., et al., 2017b, MNRAS, 465, 4159MNRAS , 1–7 (2015) irst dredge-up and multiple populations Pietrinferni A., Cassisi S., Salaris M., Castelli F., 2004, ApJ,612, 168Piotto G., et al., 2015, AJ, 149, 91Renzini A., et al., 2015, MNRAS, 454, 4197Salaris M., Weiss A., Ferguson J. W., Fusilier D. J., 2006, ApJ,645, 1131Salaris M., Pietrinferni A., Piersimoni A. M., Cassisi S., 2015,A&A, 583, A87Salaris M., Cassisi S., Mucciarelli A., Nardiello D., 2019, A&A,629, A40Saracino S., et al., 2019, MNRAS, 489, L97Sbordone L., Salaris M., Weiss A., Cassisi S., 2011, A&A, 534, A9Schiavon R. P., Caldwell N., Conroy C., Graves G. J., StraderJ., MacArthur L. A., Courteau S., Harding P., 2013, ApJ,776, L7Yong D., Grundahl F., Johnson J. A., Asplund M., 2008, ApJ,684, 1159Zennaro M., Milone A. P., Marino A. F., Cordoni G., LagioiaE. P., Tailo M., 2019, MNRAS, 487, 3239This paper has been typeset from a TEX/L A TEX file prepared bythe author.MNRAS000