The Chemical Compositions and Evolutionary Status of Red Giants in the Open Cluster NGC 752
aa r X i v : . [ a s t r o - ph . S R ] N ov Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 29 August 2018 (MN L A TEX style file v2.2)
THE CHEMICAL COMPOSITIONS ANDEVOLUTIONARY STATUS OF RED GIANTS IN THEOPEN CLUSTER NGC 752
G. B¨ocek Topcu , ⋆ , M. Af¸sar , , M. Schaeuble , C. Sneden , Department of Astronomy and Space Sciences, Ege University, 35100 Bornova, ˙Izmir, Turkey; Department of Astronomy and McDonald Observatory, The University of Texas, Austin, TX 78712
Accepted 2014 November 7
ABSTRACT
We present detailed chemical compositions of 10 red giant star members of the Galactic(open) cluster NGC 752, derived from high-resolution (R ≈ S/N ≥ T eff , log g , [Fe/H] and ξ t ) from equivalent widths of Fe i , Fe ii , Ti i , and Ti ii lines. The metallicity we obtainedfor NGC 752 ([Fe/H] = − ± α (Mg, Si, Ca), light odd-Z (Na, Al), Fe-group (Sc, Ti, V,Cr, Mn, Fe, Co, Ni, Cu, Zn), n -capture (Y, La, Nd, Eu), and p -capture (Li, C, N, O)species for each star. Furthermore, we also derived abundances of the LiCNO p -captureelement group and carbon isotopic ratios, using synthetic spectrum analyses of the Li i i ] line at 6300 ˚A, the CH G-band features near 4311and 4325 ˚A, the C bandheads at 5160 and 5631 ˚A, and , CN red system linesin the 7995 − C/ C ratios, we suggest that the 10 observed red giants can be separated into threefirst-ascent, six red-clump and one red horizontal branch star.
Key words: stars: abundances – stars: atmospheres. Galaxy: open clusters andassociations: individual: NGC 752
Open clusters (OCs) are excellent probes to investigate bothstellar and Galactic disk evolution. According to classicaltheories, OCs members were formed from the same proto-cluster cloud at the same time and the same distance. Thislends particular importance to the morphology of an OCcolor-magnitude diagram (CMD), as it allows determinationof the temperatures, luminosities, and evolutionary states ofcluster members more accurately than can be done for fieldstars.All members of individual OCs should have comparableinitial chemical compositions, differing only in their initialmasses. As predicted by the models, the individual nucle-osynthesis and mixing histories of the members are responsi-ble for any surface abundance differences among evolved OCstars. In classical stellar evolution, the so-called first dredge- ⋆ E-mail: [email protected] (GBT); [email protected] (MA); [email protected] (MS);[email protected] (CS) up, which occurs at the base of the red giant branch (RGB) ,is the main mechanism that changes the surface abundances.The best indicators of this mixing are the observed changesin abundances of the elements susceptible to proton-capturemechanisms. In stars with masses and metallicities similar tothe sun, photospheric abundances of C and C/ C willdecrease while the abundances of He (unobservable), Cand N will increase. Surface Li abundances vary greatlyfrom one star to the next, and on average will also greatlydecrease during the star’s evolution from main sequence tothe RGB phase.The quantitative changes in LiCNO abundancesdepend on the initial masses and metallicities ofthe stars (e.g. Sweigart et al. 1989; Charbonnel 1994;Boothroyd & Sackmann 1999; Marigo 2001). For example,standard predictions suggest that C/ C = 20 to 30 is typ- In this paper we will use the term red giant (RG) generically,meaning all cool and luminous stars; RGB will designate starson the first ascent of the giant branch; and RC will designateHe-burning red clump stars.c (cid:13)
G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden ical for RGB stars (e.g. Iben 1967; Dearborn et al. 1976),though observations yield a more complex answer. In par-ticular, early spectroscopic studies (e.g. Tomkin et al. 1975,1976; Lambert & Ries 1981; Sneden et al. 1986) showed thatRGB and RC stars often exhibit C/ C <
15, much lowerthan expected.Gilroy (1989, hereafter Gil89) found a strong anti-correlation between stellar mass and C/ C ratios in clus-ters with turnoff masses M < ⊙ . This trend could notbe explained by canonical stellar evolution models withoutinvoking some sort of extra convective envelope mixing. Re-cently, Af¸sar et al. (2012) have discovered anomalously low C/ C ratios among metal-rich thin-disk field red horizon-tal branch stars, indicating that their carbon isotope ratiosmay have been altered during RGB evolution. Several extra-mixing processes have been suggested to explain these lowisotopic ratios, e.g., Sweigart & Mengel (1979); Charbonnel(1994); Charbonnel et al. (1998); Boothroyd & Sackmann(1999); Charbonnel & Lagarde (2010). For first-ascent RGBstars with M < ⊙ , especially at low metallicities,both C/ C and C/N ratios sharply drop at the so-called“luminosity function bump” (Gratton et al. 2000). At thisevolutionary stage, the outward-advancing H-burning shellcancels out the chemical discontinuity left by the convec-tive envelope, allowing further mixing to take place (e.g.Charbonnel 1994).The observational challenge is to discern the resultsof the extra-mixing processes through the abundances instars with well-determined masses and luminosities. Thisis difficult for field stars, since it is not easy to determinetheir masses. However, mass estimation is easily done foropen and globular cluster RGs. Detailed studies of LiCNOabundances of evolved members of individual OC’s areincreasing (e.g., Gilroy 1989; Gilroy & Brown 1991; Luck1994; Tautvaiˇsiene et al. 2000; Tautvaiˇsien˙e et al. 2005;Smiljanic et al. 2009; Mikolaitis et al. 2010, 2011a,b, 2012).In this paper we aim to add to this growing literature by pre-senting high-resolution spectral analyses of NGC 752 RGmember stars. We report atmospheric parameters, [Fe/H]metallicities , relative abundances and ratios of elements be-longing to the α , light odd-Z, Fe-peak, and neutron-capturegroups, with a particular focus on the LiCNO group. Weemploy new, comprehensive laboratory transition studies ofCH, C , and CN molecular bands, which have materially in-creased the reliability of proton-capture abundances derivedin RG stars.The structure of this paper is as follows: in § §
3. We discuss compilation ofatomic/molecular line lists and equivalent width measure-ments in §
4. The derivation of model atmospheric parame-ters is described in §
5, followed by abundance analysis in § § For elements A and B, [A/B] = log ( N A /N B ) ⋆ – log( N A /N B ) ⊙ and log ǫ (A) = log ( N A /N H ) + 12.0 . Also, metal-licity will be taken to be the [Fe/H] value. Table 1.
NGC 752 cluster parameters.Quantity Value Ref.Right Ascension (2000) 01 57 41 WEBDADeclination (2000) +37 47 06 WEBDAGalactic longitude 137.125 WEBDAGalactic latitude − E ( B − V ) 0.035 Dan94( m − M ) ⊙ Bartaˇsi¯ut˙e et al. (2007)
NGC 752 is one of the closest ( ≃
447 pc, Daniel et al. 1994,hereafter Dan94) intermediate age (1 − i c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752
77 and 208 yielded strong Li features with abundances oflog ǫ (Li) = 1.1 and 1.4, respectively. These stars were thus in-terpreted as first-ascent giants, while the rest of the observedmembers were identified as evolved He-burning clump stars.As mentioned in §
1, Gil89 reported high-resolution spectralanalyses of C isotopic ratios from CN features near 8004 ˚Aand new Li abundances in selected RGB members of 19 clus-ters. In the lower turnoff mass OCs of her sample, a clearcorrelation between cluster turnoff mass and C/ Cwasfound. The prior results for NGC 7789 (Pilachowski 1986,Sneden & Pilachowski 1986) fit in with the general trend;see Gil89 Figure 9.Interpretations of NGC 752 RG evolutionary states de-pend on fundamental cluster parameters such as distancemodulus, reddening, age, main-sequence turnoff mass, andmetallicity. Here we summarize the literature values of thesequantities, noting that estimates of ( m − M ) , E ( B − V ),age, M TO , and [Fe/H] have not significantly changed fromearliest investigations to the present time. Reddening and Distance Modulus:
In the publicationhistory of NGC 752, most of the methods for derivingthese quantities have been applied to its main sequencemembers. Dan94 collected the data of six photometric sys-tems and transformed them to the Johnson UBV system.They derived E ( B − V ) = 0.035 ± .
005 and ( m − M ) =8 . ± .
10 (447 ±
10 pc). Bartaˇsi¯ut˙e et al. (2007) conductedseven-color Vilnius photometry for NGC 752, and from thebest isochrone match to the CMD they found ( m − M ) =8 . ± .
03 (409 ± V ∼ = 18.5. Using the data of 70 photometric members, theyderived ( m − M ) = 8 . ± .
32 (472 ±
72 pc), consistent withtheir prior estimate due to its larger uncertainty.
Cluster Age and Main Sequence Turnoff Mass:
Thesequantities have been determined by comparing the NGC 752photometry to isochrones. Dan94 derived an age of 1 . ± . . ± . . ± .
04 Gyr (Bartaˇsi¯ut˙e et al. 2007)and 1 .
41 Gyr (Bartaˇsi¯ut˙e et al. 2011). the cluster turnoffmass Bartaˇsi¯ut˙e et al. (2007) suggest M TO ≈ . ⊙ , whileGil89 estimated a turn off mass M TO ≈ . ± .
13 M ⊙ usingphotometric data then available. Metallicity:
Since metallicity is one of the most impor-tant parameters in the evolution of OCs, several NGC 752studies have been devoted to this parameter. Like mostOCs, its metallicity is approximately solar. Nevertheless, avariety of photometric and spectroscopic techniques havebeen applied to derive the [Fe/H] value more accurately.Dan94 used both spectroscopic and photometric approachesto derive a mean cluster metallicity of [Fe/H] = − . ± .
05. From Vilnius photometry, Bartaˇsi¯ut˙e et al. (2007)derived h [Fe/H] i = − . ± .
03, and Bartaˇsi¯ut˙e et al.(2011) revised this value to h [Fe/H] i = +0 . ± . h [Fe/H] i = +0.05.Paunzen et al. (2010) derived h [Fe/H] i = − . ± . Figure 1.
The CMD of the OC NGC 752 with PARSECisochrones (Bressan et al. 2012). of NGC 752, and from 10 non-binary members they de-rived [Fe/H] = − ± h [Fe/H] i = − . ± . h [Fe/H] i = 0 . ± . h [Fe/H] i = − . ± .
05 also from four RGBs, but only oneis common with the sample of Car11. We conclude that acluster metallicity of [Fe/H] ∼ The CMD of NGC 752 has been studied by several authors(e.g. Dan94, Bartaˇsi¯ut˙e et al. 2007 and Bartaˇsi¯ut˙e et al.2011). We used the combined photometric data fromDan94, since the data from Bartaˇsi¯ut˙e et al. (2007) andBartaˇsi¯ut˙e et al. (2011) do not include all the NGC 752 RGmembers present in our target list (Table 2).Figure 1 shows a comparison of the photometric dataobtained by Dan94 to the latest set of PARSEC isochrones(Bressan et al. (2012)). Earlier studies of NGC 752 em-ployed isochrones published by VandenBerg et al. (2006),Meynet et al. (1993) and Castellani et al. (1992) to obtainmetallicity and age estimates of this cluster. However, sev-eral important physical considerations, such as convectiveovershooting, which might be important for age estimationin intermediate-age OCs such as NGC 752, have been ne-glected in these older calculations. Additionally, old opacity c (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden
Table 2.
Observed NGC 752 RGs. Coordinates, K magnitudes and proper motions are taken from Simbaddatabase. B and V magnitudes are from Dan94.Star Other Name RA DEC
B V K µ α µ δ (2000) (2000) (mas/yr) (mas/yr)Cluster MembersMMU 1 BD+37 407 01 55 12.616 +37 50 14.54 10.31 9.39 7.23 3.4 − − − − − − − − − − − − − data (Rogers & Iglesias 1992; Maeder & Meynet 1991) wasused in the calculations of the final isochrone tracks.PARSEC isochrones are calculated using not only twodifferent types of overshooting (core convective & enve-lope), but also two different types of opacities. For thehigh temperature regime, OPAL 1996 (Iglesias & Rogers1996) data is used in the computation, while the low tem-perature opacities are obtained from the AESOPUS code(Marigo & Aringer 2009). This ensures that isochrone tracksfor any given input chemical composition can be calculated.In addition to these improvements, the equations of statewere calculated from the lastest version of the freely avail-able FREEEOS package. The combination of these updatedparameters and input physics should yield more reliableisochrones and thus also age and metallicity estimates.To obtain our final estimates of the age and metallicityof this cluster, an iterative fit of the PARSEC isochrones tothe photometric data was done. The final results can be seenin Figure 1. Our results are in good agreement with thoseobtained by previous studies, indicating that the choice ofthe isochrone source is not important. The final input pa-rameters of our isochrone are Z = 0.014 and an age of 1.6Gyrs. These lead to a turnoff mass of M TO ≈ . M ⊙ , againin good agreement with all previous studies. To assemble the NGC 752 sample we began with the the OCdatabase WEBDA , selecting stars labeled as RGs in its V versus B − V CMD. We checked the membership status ofthese stars using the OC RG radial velocity (RV) surveyof Mer08. In that paper, velocities were reported from theCOROVEL “spectrovelocimeter” observations of 1309 po-tential members of 166 clusters. Mer08 surveyed 30 possiblemembers of NGC 752, finding just 10 stars without obviousspectroscopic binary companions and with RVs close to thecluster mean of h RV i = +5.04 ± − (standarddeviation of the sample σ = 0.32 km s − ). We gathered http://freeeos.sourceforge.net/ spectra of these 10 stars. We also observed two suspectednon-member stars from their list to test whether we couldrule out their NGC 752 membership from our data alone.Our target RGs are listed in Table 2, giving identificationsin the MMU (Mer08) and BD systems. Table 2 also lists RA,DEC, and proper motions values taken from the SIMBAD database, B and V magnitudes from Dan94, and K magni-tudes from the 2MASS survey (Skrutskie et al. 2006). Allof our programme stars are bright ( V <
High resolution spectra of the NGC 752 targets were gath-ered with the Robert G. Tull Cross-Dispersed Echelle spec-trograph (Tull et al. 1995) on the 2.7 m Harlan J. SmithTelescope at McDonald Observatory. The wavelength rangeof the spectra was 4000 to 8000˚A with a resolving power of R ≡ λ/ ∆ λ ≈ , λ ≃ tasks inthe ccdred package, including bias subtraction, flat-fielding,and scattered light subtraction. Then, the spectra were ex-tracted by using tasks in the echelle package. Th-Ar com-parison lamp exposures taken at the beginning and end ofeach night were used for wavelength calibration. To filter outcosmic-ray and other single-exposure anomalies, we com-bined all available integrations for each star. In combinationwith the IRAF telluric task, we used the spectrum of a hot,rapidly rotating star to remove telluric line contamination(O and H O) from our target spectra.Our target stars are solar-metallicity RGs and, as such, http://simbad.u-strasbg.fr/simbad/ http://iraf.noao.edu/ c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Table 3.
Observing logs and radial velocities of the observed stars.Star Obs. Date Exp. S/N
RV a RV b RV c RV d ( s ) ( kms − ) ( kms − ) ( kms − ) ( kms − )Cluster MembersMMU 1 2012 November 3600 160 4 . ± .
20 5 . ± .
15 5 . ± . . ± .
20 4 . ± . . ± .
19 4 . ± . . ± .
19 5 . ± . . ± .
19 4 . ± . . ± .
20 5 . ± .
09 6 . ± . . ± .
20 5 . ± .
09 5 . ± . . ± .
23 5 . ± .
09 6 . ± . . ± .
19 5 . ± .
09 6 . ± .
30 6 . ± . . ± .
19 4 . ± . − . ± . − . ± . . ± .
24 9 . ± . a This study. b Mermilliod et al. (2008). c Carrera & Pancino (2011). d Reddy et al. (2012). have very line-rich spectra. Pure continuum regions stretch-ing more than about half an ˚Angstrom are very difficult toidentify even in the yellow-red spectral regions; this compro-mises any attempts to make signal-to-noise (
S/N ) estimatesdirectly from the reduced spectra. After performing somenumerical tests, we chose to use the photon counts of thefinal co-added one-dimensional spectral orders as primary
S/N indicators. We made these calculations at λ ≃ S/N values computedin this manner are in reasonable agreement with estimatesthat we made of point-to-point spectrum flux scatter in tinyspectral regions that appear to be free of line absorption.We also matched the observed spectra with synthetic spec-tra (see § S/N estimates from the scatter inthe observed minus computed spectrum differences. Thesevalues also were in reasonable agreement with the primaryphoton-count statistics given in Table 3.We measured the apparent RV shifts of our targetswith the cross-correlation method provided in the fxcor (Fitzpatrick 1993) routine of IRAF. The cross-correlationtechnique requires a template spectrum. In order to avoid er-ror contributions that can arise from rest-frame wavelengthcorrections made to a stellar standard-star spectrum, weopted against selecting the spectrum of another observedstar as the template. Instead we created an artificial spec-trum (see §
6) with model atmospheric parameters similar tothose of the NGC 752 programme stars. This spectrum wascomputed for the wavelength range from 5020 to 5990 ˚A.We then used the task rvcorrect given in IRAF’s rv pack-age to transform the geocentric into heliocentric RVs. Wemeasured the RV for each star after combining individualexposures. However, we tested the RV scatter on individualexposures for a couple of stars, finding the scatter to be verysmall, σ ≃ − .Basic observational parameters of our programme starsalong with measured RVs their associated errors fromIRAF’s fxcor task are listed in Table 3. In this table we also give the RVs reported in Mer08, Car11 and Red12.From these data, for NGC 752 we obtain h RV i = 4 . ± .
20 km s − ( σ = 0.63 km s − ), which is in good agreementwith 5 . ± .
08 km s − derived by Mer08.The RVs of NGC 752 RG members as determinedby Mer08 and confirmed in this study are not sharedby the suggested non-members. For MMU 39, we derived RV = − − , about 26 km s − away from the clus-ter mean, clearly ruling out membership. For MMU 215,our RV = 9.5 km s − differs only by 4.7 km s − from thecluster mean, but this still represents a 7 σ deviation. Thusstars MMU 39 and MMU 215 do not belong to NGC 752by this criterion. Nevertheless, we kept them in our sampleto see if their chemical compositions would provide furtherinformation to address the question of cluster membership. An important task in any spectroscopic abundance analysisis to create a list of relatively unblended lines that have re-liable transition probabilities. This is especially importantamong stars that are cooler than G spectral type, becausetheir spectra are composed of overlapping atomic and molec-ular transitions, which adversely affect many potentially use-ful lines.
The atmospheric parameter and chemical compositionderivations in this work were conducted with the currentversion of the local thermodynamic equilibrium (LTE) lineanalysis and synthetic spectrum code MOOG (Sneden1973). Here we discuss the input atomic/molecular line data. (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden
The transitions employed in this study fell into three cate-gories. (1)
Unblended Ti and Fe neutral and ionized specieslines that were used for model atmospheric parameter de-terminations. Abundances from these lines were calculatedfrom their measured equivalent widths ( EW ). (2) Selectedadditional unblended atomic species lines that were used forrelative abundance ratio determinations. Again,
EW s wereused to calculate their abundances. (3)
Other atomic andmolecular species lines that are heavily blended, or have sig-nificant hyperfine/isotopic substructure, were analyzed us-ing spectral synthesis techniques.Several sources were used to identify unblended lines:solar spectrum atlases (Delbouille et al. 1973, Kurucz et al.1984, Wallace et al. 2011), solar spectrum identifica-tions (Moore et al. 1966), the Arcturus spectrum atlas(Hinkle & Wallace 2005), and the interactive database ofhigh-resolution standard star spectra SpectroWeb (Lobel2011) . Since it becomes increasingly difficult to determinethe continuum in bluer spectral regions, we did not includethe lines at wavelengths shorter than 4500 ˚A. We also triedto bypass the regions for which we applied telluric line re-moval. As a result, the overall spectral region we used for EW measurements ranges from 4500 ˚A to 7250 ˚A. Addi-tionally, we discarded absorption lines with noticeable asym-metrical structure, even if we could not find any informationabout line contamination from the solar and Arcturus spec-tral resources.As mentioned above, model atmospheric parameterswere determined using transitions of Fe i , Fe ii , Ti i , andTi ii . For these species, we imposed additional line strengthlimits, based on EW measurements and initial analyses ofprogramme star MMU 77. Very strong lines in MMU 77,those with EW s higher than 150 m˚A (reduced widths RW = log ( EW/λ ) > − λ ≃ EW <
10 m˚A or
RW < − i , 12 Fe ii , 12 Ti i , and 5Ti ii lines. These line strength cutoffs were not applied tospecies with only a few available transitions (e.g., Na i andLa ii ). We used various references for transition probabili-ties. The line list for all species, ordered by atomic numberand ionization state, is shown in Table 4 (full version avail-able online), where we list the wavelengths, lower excitationenergies, log gf s, and the references for the adopted tran-sition probabilities and hyperfine/isotopic splitting. When-ever possible, we used a single laboratory-based homoge-neous transition probability study for a species. In partic-ular, for Ti i , Ti ii , Ni i , La ii , Nd ii , and Eu ii we used labdata obtained by the University of Wisconsin atomic physicsgroup. Frustratingly, there are no recent and comprehensivelab studies for Fe i and Fe ii , the most crucial elements in anyspectroscopic study of stars. Therefore, we have taken theirtransition probabilities mostly from O’Brian et al. (1991),NIST and VALD (Kupka et al. 2000) (see online versionof Table 4). For the other species that we analyzed usingtheir measured EW s (Si i , Ca i , Cr i , and Cr ii ), transition http://spectra.freeshell.org/spectroweb.html http://physics.nist.gov/PhysRefData/ASD/lines form.html http://vald.inasan.ru/˜vald3/php/vald.php Table 4.
Line list of species.The machine-readable versionof the entire table is available in the online journal.Species Wave. LEP log gf EW / syn Ref.(˚A) (eV)Li i .
17 syn KuruczCH 4310 syn Mas14 a CH 4325 syn Mas14C b C c O i − .
72 syn Caf08 d Na i − .
70 syn NISTNa i − .
56 syn NISTNa i − .
26 syn NISTMg i − .
62 syn KuruczMg i − .
83 syn KuruczMg i − .
95 syn KuruczAl i − .
35 syn KuruczAl i − .
58 syn KuruczAl i − .
64 syn KuruczAl i − .
65 syn KuruczSi i − .
90 EW Lob11 e Si i − .
61 EW VALDSi i − .
04 EW NIST a Masseron et al. (2014) b Brooke et al. (2014) c Sneden et al. (2014) d Caffau et al. (2008) e Lobel (2011) probabilities were taken from sources noted in Table 4. Upto 164 lines are potentially available for EW measurements.Fe-group species V i , Co i and Sc ii transitions have sig-nificant hyperfine substructure, and their transitions shouldnot be treated as single lines. We still derived their abun-dances from EW measurement, but with the blended-lineanalysis option in the MOOG code. For our work, weadopted the hyperfine substructure wavelengths and rela-tive gf s from the Kurucz (2011) line compendium. Thesespecies also lack good recent laboratory data. Therefore, wedetermined empirical log gf values for these species from areverse solar analysis. We started the reverse analysis of Sc ii with a transition probability taken from Lawler & Dakin(1989), and the initial transition probabilities of V i and Co i were taken from the Kurucz database. For this task, we mea-sured EW s from a very high-resolution solar flux spectrum(Kurucz et al. 1984), and forced the total gf values to repro-duce the solar abundances recommended by Asplund et al.(2009) (see § i , Na i ,Mg i , Cr i , Sc i , Cu i , Zn i , Y ii , La ii , Nd ii , Eu ii ) have com-plex transitions, caused by their own substructures and/orby blending with other absorbers. For these species, we de-rived the abundances by spectrum synthesis, using recenttransition probabilities whenever possible. Special mentionshould be made here of the availability of new, very exten-sive and accurate laboratory data for molecules that ap- http://kurucz.harvard.edu/linelists.htmlc (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Figure 2.
Comparison of our EWs with the ones given by Red12and Car11 for MMU 311 (top panel). Differences in EWs are de-fined as ∆ EW = EW literature − EW this study (bottom panel). pear in almost every spectral region of RG stars: the C and C C Swan system (Brooke et al. 2013; Ram et al.2014), the CH and CH (Masseron et al. 2014) G-band, CN and CN (Brooke et al. 2014; Sneden et al. 2014), redand blue systems, and MgH (Hinkle et al. 2013). With thenew gf line data for these molecules, the accuracy of C andN abundance determinations can be significantly improved.Furthermore, syntheses of various atomic transitions thatare surrounded by these molecular lines can now be accom-plished with greater confidence. We measured
EW s in a semi-automated manner, using amodified version of the Interactive Data Language (IDL)code that was introduced by Roederer et al. (2010) and re-fined by Brugamyer et al. (2011). The observed line pro-files were matched interactively with theoretical Gaussianline profiles in most cases, or Voigt profiles for some of thestronger lines. Central line depths were also recorded for usein estimating initial values of effective temperature ( T eff )for the programme stars via the line depth ratio methoddiscussed below. In Table 5 (also available online) we havelisted EW s measurements of all our target stars.We tested the accuracy of our
EW s measurementsin several ways. First, we re-measured the
EW s of somelines using the Gaussian fit approximation in IRAF’s splot task; good agreement was found with our IDL code results.Second, we compared our
EW s with those measured byCar11 and Red12. In Figure 2 we show EW correlationsfor MMU 311, the NGC 752 RG shared by all three studies.For 54 lines used by us and Red12, we find ∆ EW Red =0.58 ± σ = 4.69 m˚A), and for 59 lines shared withCar11 we find ∆ EW Car = 0.60 ± σ = 4.64 m˚A).Considering the differences in spectroscopic data and mea-surement methods among these three studies, we regard the EW agreement as satisfactory. We determined parameters for the programme stars with astandard approach of calculating abundances from the
EW s of Fe and Ti neutral and ionized lines, and requiring of theseabundances:(i) for T eff , that there be no difference, on average, in the Fe i abundances of low and high excitation ( χ ) potential lines;(ii) for ξ t , that there be no difference on average between Fe i and Ti i abundances of weak and strong lines (no apparenttrend with reduced width, RW = log ( EW/λ ));(iii) for log g , that the mean abundances of neutral and ion-ized Fe and Ti lines agree;(iv) for [Fe/H], that the value employed in creating themodel atmosphere agrees with the derived value.These four atmospheric parameters are somewhat cou-pled, e.g., the lowest excitation lines are often the strongestones. Since we have a larger excitation potential range forTi i than the other species, and more Fe i than the otherspecies, and more Fe i and Fe ii lines than Ti i and Ti ii lines,we gave extra weight to Fe in deriving model atmosphereparameters. After several trials we adopted a uniform 0.35weight for Ti lines. The parameter results proved to be in-sensitive to the exact weight that was employed here.We were mindful of the potential for undesired corre-lations among the atmospheric parameters as discussed byTorres et al. (2012). Therefore, we used a semi-automatediterative approach in model derivation, one that allowed ex-amination of abundance changes caused by each alterationin T eff , ξ t , log g , and metallicity values. To accomplish theiterations more efficiently, we used a code that altered theinput model parameters in response to abundance slopeswith χ and RW , and mismatches between neutral and ion-ized species or input and output metallicities. This code is amodified version of one that has been employed in previouslarge-sample abundance analyses by Hollek et al. (2011) andRoederer et al. (2014). The abundance trends in text andgraphical form are available for inspection in each iteration,so that user judgment can be applied to parameter changesattempted by the code. Implementation of this scheme isdiscussed in § Gray & Johanson (1991) demonstrated that precise T eff val-ues for F − K dwarf stars could be obtained from calibra-tions of the central depth ratios of absorption line pairs se-lected to have different responses to changes in temperature.Their basic line depth ratio (LDR) method was expandedin T eff , log g , and metallicity space in subsequent studies(e.g., Strassmeier & Schordan 2000, Gray & Brown 2001).In Af¸sar et al. (2012) we employed several LDRs to confirmthe T eff values derived from Fe i excitation equilibria for asample of red horizontal-branch stars.A comprehensive examination and re-evaluation of theLDR technique was done by Biazzo et al. (2007a,b). Thoseauthors identified 15 pairs formed from 28 total lines in the6199 − T eff , but relatively insensitive to log g and c (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden
Table 5.
Equivalent width measurements (in m˚A) of the NGC 752 RG members. Themachine-readable version of the entire table is available in the online journal.Species Wavelength MMU(˚A) 1 3 11 24 27 77 137 295 311 1367Si i i i i i Table 6.
Photometric and spectroscopic atmospheric parameters.Star T eff,(B-V) T eff,(V-K) T eff,(LDR) log g ,phot T eff,spec log g ,spec ξ spec [ M/H ](K) (K) (K) (cm s − ) (K) (cm s − ) (km s − )Cluster MembersMMU 1 4979 4888 5038 ±
18 2.81 5005 2.95 1.07 − . ±
21 2.79 4886 2.76 1.10 − . ±
29 2.74 4988 2.80 1.14 − . ±
22 2.54 4839 2.42 1.23 − . ±
26 2.64 4966 2.73 1.16 − . ±
21 2.67 4874 2.80 1.15 − . ±
19 2.60 4832 2.51 1.29 − . ±
29 2.74 5039 2.88 1.10 − . ±
20 2.63 4874 2.68 1.24 − . ±
32 2.52 4831 2.42 1.22 − . ±
39 4811 2.20 1.22 − . ±
61 4350 1.81 1.29 +0 . [Fe/H], at least for the metallicities of typical disk stars.Biazzo et al. (2007a) developed cubic polynomials to express T eff as a function of a depth ratio for each line pair, both forsharp-lined stars and ones with significant rotational broad-ening. As discussed in § T eff (LDR) estimates. Thesevalues are tabulated in Table 6, along with their standarddeviations and the number of participating LDR pairs.We also estimated photometric effective temperaturesof our programme stars from the B , V , and K magni-tudes in Table 2. For this purpose we used metallicity-dependent T eff versus color calibrations for giant stars givenby Ram´ırez & Mel´endez (2005). The temperatures that wecalculated for ( B − V ) and ( V − K ) colors are givenin Table 6. Metallicity is an important consideration incolor-temperature calibrations, affecting ( B − V ) more than( V − K ) because the B band suffers a large amount ofline blanketing. Therefore, since ( V − K ) colors are nearlymetallicity independent, for the initial T eff (phot) we onlyused the computed ( V − K ) temperature. We then calcu-lated a non-weighted mean of T eff (phot) and T eff (LDR) to The K magnitude is from 2MASS, and Ram´ırez & Mel´endez(2005) label the color ( V − K ). form the initial T eff estimate for the programme stars; thesealso are given in Table 6.To calculate physical (cluster) gravities we used thestandard equation,log g ⋆ = 0 . M V ⋆ + BC − M Bol ⊙ ) + log g ⊙ +4 log( T eff ⋆ T eff ⊙ ) + log( m ⋆ m ⊙ ) . (1)The adopted solar parameters were M bol =4 . , log g = 4 .
44 and T eff = 5780 K. Temperature-dependent bolometric corrections were calculated with thepolynomial formula and coefficients given in Table 1 ofTorres (2010). For NGC 752, the absolute magnitudes M V were computed with the V magnitudes of Table 2, and thecluster distance modulus, reddening and turnoff mass wereused as given in Table 1.To calculate the physical gravities we varied the stel-lar masses from 1.5 to 1.95 M ⊙ , which is a range from theminimum turnoff mass suggested in the literature to RGmass provided by the PARSEC isochrone we applied. Theresults of these experiments always led us to the spectro-scopic gravities that were expected for the RG stars and theyalways remained in our uncertainty limits (see § . M ⊙ , which isthe turnoff mass of NGC 752 suggested in Bartaˇsi¯ut˙e et al.(2007). Previous studies ( §
2) suggest about solar metallicityfor NGC 752, so we used an initial value of [M/H] = 0. Fi-nally, we adopted a microturbulent velocity that is typicalfor solar-metallicity RG stars: ξ =1.2 km s − . c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Figure 3.
An example model atmosphere determination usingboth ionized and neutral species of Fe and Ti for MMU 77.
We fed the Ti and Fe
EW s ( § T eff , log g , and[M/H] with software developed by Andy McWilliam and In-ese Ivans, kindly made available for our use.After establishing initial model atmosphere parameters,we conducted the final parameter search with the code andthe model convergence criteria described in § EP and RW val-ues; there is no apparent trend between the abundances andthese two parameters.The initial and final model atmosphere parameters forall of our targets are given in Table 6. We also summarize inTable 7 the atmospheric parameters of our programme starsthat were obtained by Gil89, Car11, and Red12.Comparisons of the final iterated effective temperaturesfrom our spectra, T eff( spec ), and the initial T eff estimatesfrom photometry and LDR measurements are shown in Fig-ure 4. On average, values are in reasonable accord, with themean difference being h T eff ( initial ) − T eff ( final ) i = −
18 K. However, offsets can be seen among the individual T eff de-termination methods. For both photometric temperatures,uncertainties arise from reddening and distance parameters,and the overall offsets can come from the color- T eff cali-brations. We derive h T eff ( B − V ) − T eff ( spec ) i = − ± h T eff ( V − K ) − T eff ( spec ) i = − ±
16. For tempera-tures calculated with LDR method, we find h T eff ( LDR ) − T eff ( spec ) i = 81 ±
18. Continuum placement uncertaintiescan be of concern here, so we made numerical experimentsin which the continuum choices were changed to substan-tially lower and higher values. These resulted in T eff (LDR)changes ≪
25 K because the LDR method compares depthsof lines situated very close in wavelength, a feature central tothe original LDR method design (Gray & Johanson 1991).In Figure 4, one can see that the three coolest stars areoffset in the comparisons by ∼ −
70 K compared to theother seven stars. This small offset does not substantiallyeffect metallicities and relative abundance ratios of thesestars with respect to the majority of our sample.A comparison of calculated gravities (log g phot ) withspectroscopic gravities (log g spec ) is given in Figure 5. Theyagree well with each other: h ∆log g i = − . ± .
03. There isa small trend that is mostly due to the three coolest, lowestgravity RGs in our sample. Differences in the evolutionarystatus of the members (see §
7) may create such a fluctuation.The metallicities we derived for the NGC 752 RGs fromthe model atmosphere analysis have a slight scatter aroundthe solar metallicity (Table 6). The mean metallicity of thecluster calculated from these 10 members is < [M/H] > = − . ± .
04. The metallicities of all NGC 752 targets arein agreement except for MMU 137, which has a metallicityof [M/H] = − .
16. However, this star also has one of thelowest log g values in our sample.We also applied these atmospheric parameter determi-nation methods to the suspected non-member stars MMU 39and MMU 215. Since no parallax, reddening, and massdata are available for these stars, we could not calculatetheir T eff (B-V), T eff (V-K) and log g phot values. Thereforewe used the LDR temperatures for T eff and adopted typicalRG log g values for the initial parameter guesses. We found[ M/H ] = − .
33 for MMU 39 ( ∼ σ from the cluser mean),and [ M/H ] = +0 .
09 ( ∼ σ from the mean) for MMU 215.Clearly NGC 752 membership is ruled out for these twostars from our RV and metallicity computations, in agree-ment with the results of Mer08. These stars are eliminatedfrom further discussion in this paper. We estimated the internal uncertainties in atmospheric pa-rameters by running a series of analyses on the spectral dataof MMU 77. First we changed the effective temperature in50 K steps while keeping the other atmospheric parame-ters fixed. This process was repeated until the magnitudeof the difference between high and low excitation potentialFe i lines exceeded the σ value of the initial line abundances.This method led to an average T eff uncertainty of ∼
100 K.We applied a similar method to estimate the internal un-certainty for microturbulence velocity. This time we onlychanged the velocity in 0.1 km s − steps and focused on theabundance changes in the elements that have both neutraland ionized lines: Ti i , Ti ii , Fe i , Fe ii , Cr i . and Cr ii . The c (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden
Figure 4.
Comparison of ∆ T eff = T eff ( x ) − T eff (spec) with spec-troscopic T eff . T eff ( x ) stands for the temperatures of T eff ( B − V ), T eff ( V − K ) and T eff ( LDR ). Figure 5.
Comparison of ∆log g=log g ( phot ) − log g ( spec ) withderived spectroscopic log g. typical average uncertainty achieved for ξ t was 0.3 kms − .The internal uncertainty level for log g was also derived bytaking into account the abundance differences in both neu-tral and ionized species. The highest abundance difference(over ± σ level) between Cr i and Cr ii was found to be0 .
06 dex, which corresponds to an uncertainty of 0 .
16 dexin log g .To estimate the external uncertainty in our T eff val-ues, we compared them with the available T eff estimates for Table 7.
Spectroscopic atmospheric parameters of NGC 752members studied by Gil89, Car11 and Red12.Star T eff,spec logg ,spec ξ spec [ M/H ](K) (cm s − ) (km s − )Gil89MMU 1 5000 2.85 1.90 0 . . . . . . − . − . − . − . these stars that were also investigated by Gil89, Car11, andRed12.: Gil89, Car11 and Red12. The atmospheric parame-ters obtained in these studies are also listed in Table 7. Wealso calculated the photometric and LDR temperatures andinvestigated the differences between these two temperaturesand our spectroscopically derived T eff values (Figure 4). Anoverall comparison of the differences among these tempera-tures yielded an average external uncertainty of about 150K. The external uncertainty in log g was estimated by com-paring our log g values with the ones gathered from the lit-erature for shared NGC 752 stars. We have made use ofthe same set of studies that were used to estimate the un-certainty in T eff . Our results are usually in good agreementwith the published ones, with an average scatter of ∼ g value of 3.2.Since this result is significantly different from the averageprovided by other studies, we did not include it in our un-certainty estimations. In Figure 5, we plot the spectroscopiclog g versus calculated log g values. The standard deviationof the differences between these two log g values was foundto be ∼ ±
150 K in T eff , ± g and ± − in ξ t . We also in-vestigated the uncertainties in elemental abundances causedby these adopted uncertainties, and have listed them in Ta-ble 8. The sensitivity of [X/Fe] to the uncertainties in modelatmosphere parameters is typically much smaller than thevalues of Table 8 because of correlations in the ionizationbalances of most species of interest here.We also investigated the uncertainty limits for C/ C ratios. Adopting different model atmospheres wasnot very effective in changing the isotopic ratios since the CN and CN molecular lines are essentially the same inexcitation energies, thus not really sensitive to the changesin model atmosphere parameters. To determine the uncer-tainties, we fit synthetic spectra with varying C/ C ratiosto the observed CN and CN features. The resulting un-certainty limits we derived using this method are listed inTable 8.
Abundances of species with non-blended transitions thatcould be treated as single absorbers were derived from their EW measurements. Other species exhibiting complex hy-perfine splitting and/or isotope shifts, and those with linesthat suffer significant blending of lines by other species,were treated either to blended-line analyses or full synthetic-observed spectrum matching. Several figures in this sectionwill illustrate the comparisons of different synthetic spectrawith observed data points.To normalize our abundances, we re-measured all of ourstellar lines in the integrated solar flux atlas of Kurucz et al.(1984). The solar model atmosphere was calculated usingthe Castelli & Kurucz (2003) grid, assuming T eff = 5777K, log g = 4 .
44 cgs, ξ = 0 .
85 km − . Abundances foundfrom this analysis are listed in Table 9 alongside the solarabundances recommended by Asplund et al. (2009). Sincewe assumed solar abundances given by Asplund et al. for c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Table 8.
Sensitivity ( σ ) of derived abundances to the model atmo-sphere changes within uncertainty limits for the star MMU 77.Species ∆ T eff (K) ∆log g ∆ ξ t (kms − ) − / + 150 − / +0.25 − / +0.3Li i − / +0.26 +0.06 / +0.09 +0.09 / +0.11C +0.08 / +0.11 +0.08 / +0.09 +0.09 / − − / +0.10 0.00 / +0.05 +0.03 / +0.01O − / +0.09 − / +0.19 − / +0.09Na i − / +0.07 − / − / − i − / +0.05 0.00 / − / − i − / − / − / − i +0.08 / − − / +0.05 +0.04 / − i − / +0.14 +0.04 / − / − i − / +0.23 0.00 / / − i − / +0.16 +0.02 / / − ii +0.13 / − − / +0.12 +0.13 / − ii +0.06 / − / +0.13 +0.11 / − i − / +0.19 0.00 / / − ii +0.07 / − / +0.13 +0.12 / − i − / +0.13 − / − / − i − / +0.10 0.00 / +0.02 +0.15 / − ii +0.19 / − − / +0.15 +0.09 / − i − / +0.12 − / +0.07 +0.02 / − i − / +0.10 − / +0.06 +0.16 / − i − / +0.13 0.00 / +0.08 +0.13 / − i − / − − / +0.01 0.00 / − ii − / − − / − / − ii − / +0.01 − / +0.04 0.00 / − ii − / − − / +0.07 +0.07 / − ii − / − / +0.01 +0.01 / − C/ C 2 / / − − / Sc ii , V i , and Co i , no independent solar abundance was de-rived for these elements. Estimated sigmas given in paren-thesis were determined by taking into account the contin-uum placement, goodness of the fit and smoothing of thesynthetic spectrum. For most species, Of course, we employstandard LTE analyses, while more detailed physics (e.g.,accounting for solar granulation, multi-stream atmosphericmodels, non-LTE corrections) of Asplund et al. is being ne-glected. Even so, our results are mostly in agreement withthose of Asplund et al.. Our [X/H] and [X/Fe] results forNGC 752 will be quoted differentially with respect to ourown solar analysis abundances.We derived abundances for 26 species of 23 elements.We will organize the discussion of these abundances by ele-ment groups: α (Mg, Si, Ca); light odd-Z (Na, Al); Fe-group(Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn); n -capture (Y, La,Nd, Eu); and p -capture (Li, C, N, O). In Table 10 (also avail-able online ), we list the derived abundances for individualRG members of NGC 752 and the mean abundances forthe cluster, < [X/Fe] > . The mean abundances from Red12and Car11 are also given in the last two columns of the tablefor comparison. The general distribution of solar normalizedelemental abundances can be seen in Figure 6. The meanabundances of most species (20 out of 26) fluctuate within ± Online version of this table will also include the number of linesused for the species and the scatter of the abundance obtainedfrom each species.
Table 9.
Solar abundances.Species log ǫ ⊙ log ǫ ⊙ (this study) (Asplund et al. 2009)Li i ± (0 .
05) 1.05 ± .
10C 8.43 ± (0 .
05) 8.43 ± .
05N 8.13 ± (0 .
05) 7.83 ± .
05O 8.69 ± (0 .
05) 8.69 ± . i ± (0 .
10) 6.24 ± . i ± .
16 7.6 ± . i ± .
18 6.45 ± . i ± .
05 7.51 ± . i ± .
03 6.34 ± . ii ± . i ± .
06 4.95 ± . ii ± . i ± . i ± .
04 5.64 ± . i ± . i ± .
06 5.43 ± . i ± .
04 7.50 ± . ii ± . i ± . i ± .
07 6.22 ± . i ± .
10 4.19 ± . i ± (0 .
05) 4.56 ± . ii ± .
04 2.21 ± . ii ± .
06 1.10 ± . ii ± (0 .
05) 1.42 ± . ii ± .
08 0.52 ± . Figure 6. [X/Fe] values of the species studied for the entire RGsample. Dotted line represents the solar values. Abundances ofCNO, odd-Z, alpha, iron-group and n-capture elements are shownby stars, squares, triangles, crosses and empty circles, respectively. and Nd depart from the solar abundance mix by ≥ α , Odd-Z, Fe-group and Neutron-captureelements α and Odd-Z elements : The analyzed α elements are Mg,Si, and Ca. We did not include Ti because even though Tihas been labeled as an α element in metal poor stars, itsdominant isotope Ti is not an even multiple of α parti- c (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden
Table 10.
Elemental abundances of individual stars and average abundances ( < [X/Fe] > ) for NGC 752. The average abundances fromRed12 and Car11 are also given. The machine-readable version of the entire table is available in the online journal.Species MMU1 3 11 24 27 77 137 295 311 1367 < [X/Fe] > Red12 Car11[C / Fe] − − − − − − − − − − − ± / Fe] 0.15 0.07 0.10 0.14 0.18 0.12 0.14 0.14 0.13 0.11 0.13 ± / Fe] − − − − − − − − − − − ± i /Fe] 0.04 − ± i /Fe] − − − − − − − − − − − ± − i /Fe] − − − ± − i /Fe] 0.06 0.08 0.08 0.15 0.09 0.13 0.19 0.09 0.10 0.15 0.11 ± i /Fe] 0.10 0.12 0.08 0.13 0.11 0.13 0.11 0.09 0.09 0.12 0.11 ± − ii /Fe] 0.00 − − − − − − − − ± i /Fe] − − − − − − − − ± − − a [Ti ii /Fe] 0.02 − − − − − − − − − ± − i /Fe] − − − − − − − − − − − ± i /Fe] 0.04 0.05 − − − ± − ii /Fe] 0.01 0.03 0.02 0.09 0.00 0.05 0.02 0.08 0.02 0.01 0.03 ± i /Fe] − − − − − − − − − − − ± − i /H] 0.04 − − − − ± − a [Fe ii /H] − − − − − − − − − − ± − i /Fe] − − − − − − − − − − − ± − i /Fe] 0.01 − − − ± − − i /Fe] − − − − − − − − − − − ± − i /Fe] − − − − ± − ii /Fe] − − − − − − − − − ± − a [La ii /Fe] 0.08 0.05 0.05 0.08 0.07 0.20 0.12 0.05 0.06 0.04 0.08 ± ii /Fe] 0.16 0.02 0.21 0.18 0.25 0.28 0.22 0.15 0.16 0.13 0.18 ± ii /Fe] 0.01 0.13 − − ± a Car11 list the average abundances of the same species, e.g. [Fe i /H] and [Fe ii /H] abundances are given as an average: [Fe/H]. Here, welist the abundances from Car11 according to species with the majority of the analyzed lines. cles. We therefore treat Ti as a Fe-peak. Si and Ca abun-dances were derived from EW measurements of their neutralspecies lines, while we applied synthetic spectra to determinethe Mg abundances. Note that three Mg i lines were used,but the feature at 7811.1 ˚A has very large damping wingsthat cannot be modeled with standard line parameters; theline wings are far too broad. To compensate, we arbitrar-ily increased the damping constants for this line based onmatching it to the solar spectrum. We also analyzed the 5170˚A region that encompasses two of the very saturated Mg i “b” lines, but we did not include their derived abundancesin the mean Mg abundance. We employed spectrum synthe-ses to determine the abundances of odd-Z light elements ofinterest, Na and Al.The overall comparison of the average abundances ofall species with recent NGC 752 high-resolution abun-dance studies by Red12 and Car11 is given in Table 10.Here and throughout the rest of the paper, our definitionof comparison is the difference between the mean clusterabundances from these studies and our results: for Red12,∆[X / Fe] R ≡ [X / Fe]
Red12 − [X / Fe] this study , and for Car11,∆[X / Fe] C ≡ [X / Fe]
Car11 − [X / Fe] this study . In Red12, themean abundances were determined from four cluster mem-bers. Car11 also derived the abundances of four members,but two of them have been reported as spectroscopic bi-naries by Mer08. Our results are mostly in good agree-ment with Red12 within error limits: ∆[Mg / Fe] R = 0 . / Fe] R = − .
08. The discrepancies with Car11 are larger: ∆[Mg / Fe] C = 0 .
23 and ∆[Ca / Fe] C = − .
20. Our Si i abundances agree well with those of Red12 but ∆[Si / Fe] C = − .
09. The differences between the abundances obtainedfrom Na i lines in our study and the others are small,but for the Al i , the differences are ∆[Al / Fe] R = 0 .
12 and∆[Al / Fe] C = − .
15. The reason for these discrepancies mostlikely stem from different analysis methods employed inthese studies. For instance, while we used the spectral syn-thesis method to obtain Al i abundances, Red12 and Car11used EW s . Fe-group elements : The abundances of Ti i , Ti ii , Cr i ,Cr ii and Ni i were determined from EW s in single-line anal-yses. As discussed in § ii ,V i , Co i ) have significant hyperfine substructure and no re-cent comprehensive transition probability studies. There-fore, we conducted reverse solar analyses to derive their tran-sition probabilities. Hyperfine component parameters fromthe Kurucz database were adopted for these transitions. Weapplied blended-line EW analyses (see § gf ’s, and then used these to calculate the abun-dances.We employed full synthetic spectrum computations forother Fe-group elements in our list: Mn i , Cu i , and Zn i .We used three absorption lines of Mn i located near ∼ i to obtainits abundance. Although the Cu i line at 5782.1 ˚A is near c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Figure 7.
Synthetic and observed spectra comparison of MMU 77for the 6345.1 ˚A Eu ii and 6262.3 ˚A La ii lines. The observedspectrum is represented by black points. Assumed abundancesfor the three synthetic spectra are given in the figure legend. Thebest fit to the observed spectrum is given with a red solid line. The0.3 dex lower and upper deviations from the adopted abundanceare represented with green and blue solid lines, respectively. the spectral gap between the echelle orders, it was takeninto account whenever it was available. The scatter in Znabundances, determined from only the Zn i line at 6362.3 ˚A,is mainly caused by CN and other atomic contaminants inthis region.Most of our Fe-group abundances are in good agree-ment with those of Red12 and Car11, except for the tworeverse solar analysis species, V i and Co i . Although therelative deficiencies that we found in Mn and Cu may bereal, we repeat our cautions that recent lab studies con-cerning these elements are unavailable. For V, the dis-crepancies are ∆[V / Fe] R ≡ .
13, ∆[V / Fe] C ≡ .
11 andfor Co, ∆[Co / Fe] R ≡ .
10, ∆[Co / Fe] C ≡ .
13. For Mn,∆[Mn / Fe] R ≡ .
10. And for Zn, ∆[Zn / Fe] R ≡ .
12. Themean Sc abundance is in good agreement with both Red12and Car11.
Neutron-capture elements : We applied syntheticspectrum analyses to three so-called s -process (slowneutron-capture) elements Y, La, Nd, and one r -process(rapid neutron-capture) element Eu. We also tried to deriveBa abundances in our stars. Unfortunately, all Ba ii linesare saturated in most of our target stars. Abundances de-rived from Ba lines are very sensitive to damping, microtur-bulence, and the outer atmosphere structures of our stars.Given these difficulties, we decided to discard Ba from ourelement list. Three Eu ii and four La ii lines were used to de-termine their abundances. In Figure 7, one of the regions forEu ii and one for La ii are illustrated for MMU 77. We showthree different syntheses with assumed abundances that areseparated by 0.3 dex, with the middle (red color) syntheticspectrum having the abundance that best fits the observedfeature. The synthetic line profiles of these features clearlymatch the details of the observed lines.All n -capture element abundances are in good agree-ment with those of Red12 and Car11 within the mu-tual uncertainties, except for ∆[Nd II / Fe] R ≡ − .
12 and∆[Nd II / Fe] C ≡ .
16. The difference may be caused by thedifferent analytical methods used in each study.
Figure 8.
Synthetic and observed spectrum comparison ofMMU 77 for the [O i ] 6300.3 ˚A line. Lines and symbols are areequivalent to those in Fig. 7. We employed full synthetic spectrum computations to studyall features of the LiCNO abundance group. Using the newmolecular laboratory data summarized in § i ] 6300.3 ˚A line. Then,C abundances were determined from CH features locatedaround 4311 ˚A and 4325 ˚A, and C bands at 5160 ˚A and5631 ˚A region. Finally for the N abundances, we used CNand CN red system lines in the 7995 − i ] (6300.30 ˚A) line. The feature is blended with bothNi i (6300.34 ˚A) and CN (6300.27 ˚A). We tried to accountfor these contaminants carefully (see Sneden et al. 2014 fordetails). Transition probabilities of Johansson et al. (2003)were used for the Ni i line. We derived C abundances fromseveral molecular band regions: the CH G-band, and the C (0 −
0) bandhead at 5155 ˚A and the (0 −
1) bandhead at 5635˚A. In Figure 9, we give a portion of CH G-band in panel(a). The two bandheads for C are illustrated in panels (b)and (c). The complex structures of these molecular regionslead to fluctuating C abundances from star to star, but theystill remained within error limits. Final C abundances werecalculated by taking an average of the values obtained fromthese molecular regions. Figure 10 shows an example of Nabundance synthesis from the CN region. An average wastaken to obtain final N abundances.To our knowledge, the C and N abundances derivedhere are the first for NGC 752 members. In standard stel-lar evolution, the mean N abundance is expected to in-crease and the C abundance decrease by the RGB evolu-tionary phase, Our derived C and N abundances follow thistrend. However, our O abundances exhibit a subsolar abun- c (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden
Figure 9.
Synthetic and observed spectra of three wavelengthregions used for C abundance determination. A portion of theCH G-band (panel a), C (0 −
0) bandhead at 5155 ˚A (panel b)and C (0 −
1) bandhead at 5635 ˚A (panel c) are illustrated. Linesand symbols are as in Figure 7. dance. This is in disagreement with the invariance of so-lar normalized O abundances suggested by classical stel-lar evolution. Car11 determined O abundances from thesame forbidden line and found [O/Fe]=0 . ± .
04. How-ever, unlike our study, they did not take into account theinterdependence of C, N and O abundances while determin-ing the oxygen abundances. Furthermore, they found some-what higher surface gravities (see § − . ± .
04. Recently, Maderak et al. (2013) re-ported an average [O/Fe]= − − O i triplet region. Taking into ac-count the non-LTE abundance corrections for oxygen triplet(up to almost 0.2 dex, e.g., Ecuvillon et al. 2006), Maderak’snon-LTE [O/Fe] abundance is found to be around − T eff for our 10 NGC 752 programme stars. Thisplot reveals no obvious trend with T eff , and there are noneto be found with the other model atmosphere parameters. C/ C : CN and CN features that can be used for C/ C determination are located throughout the red spec-
Figure 10.
A portion of synthetic and observed spectral regionswith CN absorption features used for N abundance determination.Lines and symbols are as in Figure 7.
Figure 11.
Distribution of C, N and O abundances plotted versusthe effective temperatures for RG members. tral domain. The most useful ones for our spectral coverageare in the ∼ C/ C valuesfrom the ratio of CN at 8003.5 ˚A to CN at 8004.6 ˚Adue to severe line blending in other features. In Figure 12,we give an example of different assumed C/ C ratiosmatched to this feature for two members of NGC 752. Welist all C/ C ratios for our sample and the ratios gath-ered from the literature in Table 11. The C/ C ratiosdetermined for our targets range between 13 and 25. Gil89derived C/ C for six members of NGC 752, but later twoof them were found to be spectroscopic binaries by Mer08.The main differences between Gil89 and our work are thatwe have higher spectral resolution and C/ C ratios weredetermined during the iterative process that we obtainedCNO abundances, and therefore they should be more reli-able. Our results indicate that two of the RGs, MMU 1 andMMU 77 have C/ C = 25 while Gil89 found lower values: C/ C=17 for MMU 1 and C/ C=16 for MMU 77. Forthe other shared RGs, MMU 295 and MMU 311, there is bet-ter agreement: we found C/ C=20 and 15, respectively,while Gilroy reported C/ C=16 for both members. Li : We derived Li abundances from the neutral Li6707.8 resonance doublet, which is somewhat blendedwith Fe i line at 6707.4 ˚A. Hyper-fine structure of the res-onance doublet was also taken into account. An example c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Figure 12.
Comparison of synthetic and observed spectra ofMMU 77 and 24 for the observed CN features around 8004 ˚A.These two stars represent the highest and lowest C/ C ratiosin our sample. Lines and symbols are as in Figure 7. of synthetic/observed spectrum comparison is given in Fig-ure 13. As seen in this figure, Li abundances vary greatlyfrom star to star. We were able to measure Li abundancesfor RGs such as MMU 77 and MMU 311 (top two panelsof Figure 13), while RGs such as MMU 137 (bottom panel)does not have a detectable Li feature (log ǫ < ǫ = 1.34. For the samemember, Pil86 and Gil89 also determined Li abundancesusing the same resonance line and found log ǫ = 1.1 and1.4, respectively. We could not detect Li in MMU 295, andneither could Pil86 (log ǫ < +0 .
5) but Gil89 reported a de-tection and with log ǫ = 0 .
45. Our non-detection is to bepreferred as it is based on higher-quality spectra.
This NGC 752 study is the first of a series of papers present-ing the results of chemical abundance analysis and investi-gating the evolutionary status of the RG members of OCs.Our main focus was the abundances of the key elements ofLiCNO p -capture group to help us understand the evolu-tionary stages of the stars and to reveal the discrepanciesamong RG members.Before starting with the analysis process, we firstconfirmed cluster membership of the 10 selected RGs inNGC 752 by deriving new RV values. Since important pa-rameters such as reddening, distance modulus, and turnoffmass are readily available for clusters such as NGC 752, wewere able to calculate relatively accurate initial parametersfor each star. To derive the most accurate atmospheric pa-rameters, we made a significant effort to establish the initialparameters prior to detailed spectrum analysis. Initial effec-tive temperatures were determined by taking the average ofthree different temperatures calculated from the ( B − V ), Figure 13.
Comparison of synthetic and observed spectra ofMMU 77, 311 and 137 for 6707.8 ˚A Li line. These three stars rep-resent the diversity in Li abundances among the RGs of NGC 752.Lines and symbols are as in Figure 7. ( V − K ) colors and line depth ratios of the temperaturesensitive species. We calculated initial log g values usingthe formula described in § T eff , log g , ξ t and [Fe/H] by using bothneutral and ionized species of Fe and Ti. The model atmo-spheres of the individual RG members (Table 6) yielded anaverage cluster metallicity of about solar, < [M/H] > = − α (Mg,Si, Ca), Fe-group (Ti, Cr, Ni, Mn, Cu, Zn) and n-captureelements (Y, La, Nd, Eu). We derived C/ C ratios ofthe RGs by employing the synthetic spectrum fitting to the CN and CN futures located around 8004 ˚A. The elemen-tal abundances are mostly scattered around solar values asseen in Figure 6. From that figure, it is obvious that speciessuch as Cu, Mn, and Nd exhibit more scatter due to ei-ther lack of updated gf values or complex hyper-fine struc-ture. However, the overall abundance consistency among allRG members of NGC 752 indicates that they all share thesame origin. Recently, Carraro et al. (2014) have reportedthe abundances of seven RC members in the OC NGC 4337,which is nearly identical to NGC 752 in age and metallic-ity. The CMDs of these clusters are very similar except thatNGC 752 has fewer main sequence stars. NGC 4337 andNGC 752 appear to have differences in the abundances of c (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden
Table 11.
Lithium abundances, carbon isotopic ratios and probable evolutionary status of NGC 752 RG members.Star log ǫ (Li) a log ǫ (Li) b C/ C b log ǫ (Li) c C/ C c Evol. PhaseMMU 1 < +0.5 0.45 17 0.15 25 RGBMMU 3 1.25 25 RGBMMU 11 1.00 25 RCMMU 24 < +0.5 < < +0.3 < < +0.5 0.25 16 < < +0.3 0.77 16 0.78 20 RCMMU 1367 < a Pilachowski et al. (1988) b Gilroy (1989) c This study some species. For example, Carraro et al. report that Na,V and Co are overabundant and Zn is underabundant inNGC 4337, while we find that they have almost solar abun-dances in NGC 752. CNO abundances in both OCs indicatethat the observed members are certainly evolved. However,C and O abundances are deficient in NGC 752, while N ismore abundant in NGC 4337. These differences suggest thatboth clusters might have somewhat different chemical origin.Further study of these species in both clusters is needed toexplain and explore these differences in greater detail.One of the main goals of our study was to derive reliableabundances of the p -capture LiCNO abundance group. Toour knowledge, our study is the first that reports the C andN abundance patterns of evolved stars in NGC 752. Theresults show the agreement among the CNO abundancesof the individual members, which also indicates a commonorigin for the NGC 752 cluster member RGs.The C under-abundances and N over-abundances con-form to expectations from classical theory of stellar evo-lution. p -capture reactions are the main mechanisms thatgovern He-production in the core of a star during the main-sequence and later in the H-burning shell during the RGBevolution. When a star evolves along subgiant branch andbecomes a first-ascent RGB star, its outer convective en-velope broadens and reaches down to the H-burning shell,eventually dredging the H-processed material up to its sur-face. This so-called first dredge-up phenomenon carries thesignatures of internal mixing up and shows itself via alteredabundances in mainly C and N on the stellar surface(e.g. Iben 1967; Iben & Renzini 1984).In NGC 752, O is underabundant by about 0.18 dex.However, it is not uncommon to find subsolar oxygenabundances in OCs. We refer the reader to Figure 19of Yong et al. (2005), who investigate the distribution of[O/Fe] abundances versus age and Galactocentric distances(R GC ) of different OC populations in the Galaxy. NGC 752,with an age of 1.6 Gyr and R GC = 8.3 kpc (Red12), is simi-lar to many other other OCs with subsolar metallicities andsimilar Galactocentric distances. We also note the small dis-persion among the O abundances of our cluster members.Having the lowest abundance of [O/Fe]=-0.23, MMU 3 hasa unique spectroscopic feature for our sample: it shows Ca ii HK emission lines. MMU 77, which has almost the sametemperature and gravity as MMU 3, does not carry any sig-nature of Ca ii HK emission, and its [O/Fe] = − i ] line at 6300.3 ˚A. Furtherinvestigation of this unusual behavior is beyond the scopeof this paper.According to canonical stellar evolution theories, sur-face C/ C and Li content can be altered by the end ofthe first dredge-up after the main sequence and subgiantphases. They are often the main indicators of stellar evolu-tion. We found an unexpectedly large range in both of theseabundances in NGC 752 RGs. In our sample, which occu-pies a relatively small domain of the HR diagram, there areexamples that have high C/ C and high Li abundances,indicative of being first-ascent RGBs. There are also caseswith low C/ C and Li abundances, indicative of much fur-ther evolution, perhaps the post-He-flash RC evolutionarystage. We have only one RG, MMU 295, with no Li featurebut C/ C = 20. Low C/ C ratios have been used asan extra-mixing indicator for some decades. This explana-tion dates back at least to Gil89, who found anomalouslylow isotopic ratios in many RG OC members, and showedan anti-correlation between C/ C and the turnoff massof a cluster.Our investigation of NGC 752 light elements benefitsfrom higher spectral resolution, larger sample size, and newmolecular data that yields CNO abundances along with Liand C/ C. We found low C/ C ratios for only fourRGs in in our sample (Table 11). Three of these four alsoshow no detection of Li: MMU 24, 137 and 1367. Such starsare prime suspects for having undergone some extra-mixingprocesses. However, Li is a fragile element and very sensitiveto the details of main sequence stellar envelope conditions.Sestito et al. (2004) have investigated the Li vs. T eff distri-bution in NGC 752 by studying 18 solar-type members andshowed that Li abundance decreases as the temperature ofthe cluster members decreases. This indicates that the mem-bers with slightly different temperatures may have slightlydifferent initial Li abundances. c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Figure 14.
Zoom-in version of the RG region from the CMDgiven in Figure 1. The RGB, RC and AGB parts of the isochroneare illustrated with (red) solid, dashed and dotted lines, respec-tively. The points are labeled with the star names.
Figure 15.
Locations of the RGs in spectroscopic T eff - log g dia-gram. Same PARSEC isochrone is used as described in Figure 1.Symbols and labels are the same as described in Figure 14. To analyze the light elements among NGC 752 RGs inmore detail, we decided to explore the evolutionary statusof our RG sample. We re-examined the CMD for NGC 752.We used the photometric data of Dan94, and applied PAR-SEC isochrones (Bressan et al. 2012), see Figure 1. The bestisochrone match to the photometry was found for an age of1.60 Gyr and metallicity Z = 0.014. Figure 14 is a zoomed-inversion of the CMD Figure 1, centered on the RGB and RCregions more in detail.We then used the same isochrone to compare our spec-troscopically derived values in the T eff - log g plane for theNGC 752 RGs of this study. This comparison, which is in-dependent from the photometric parameters, is displayed inFigure 15. The locations of the targets in the T eff - log g diagram are in good agreement with the ones in Figure 14.By taking into account the locations of our RGs inboth Figures 14 and 15, along with the Li abundances and C/ C ratios derived from the spectral analysis, we tried toestimate the probable evolutionary stages of our programmestars. Our assignments are given in Table 11. The readershould be warned that we do not indicate T eff and log g uncertainties in Figure 15 for plotting clarity, but they arelarge enough to blur the distinction between RGB and RCmembership in some cases. Therefore, the evolutionary as-signments here are not firm conclusions, but rather plausibleestimates.Our reasoning is based on the following observations: (1) : MMU 77 lies closer to the RGB than other RGs inboth Figure 14 and 15. It also has high Li abundance andcarbon isotopic ratio (Table 11). Abundances and CMD lo-cation indicate that this star is a regular first-ascent RGB. (2): MMU 24 and MMU 137 present the most obvious ex-amples of post-He-flash RC (or even AGB) stars, as theycombine highly evolved positions in Figures 14 and 15 withlow C/ C and no Li detection. (3):
The CMD locationof MMU 311 appears to indicate RC status. However, theintermediate abundance values of C/ C and Li slightlynegate this interpretation. We label this star as an RC, butcaution is advised in this case. (4):
MMU 1 is not easily un-derstood. This star seems to be close to RGB in Figure 15while it is somewhat more luminous than a true RGB star inFigure 14. It has low Li but high C/ C. We believe thatMMU 1 is an RGB star, but this assignment is not easyto defend. (5):
MMU 295 is the hottest RG in our sam-ple. Therefore, it could possibly be a red horizontal branch(RHB) star. The case for this is strengthened by the star’slack of Li detection and moderate C/ C value of 20. Wetentatively suggest that our 10-star NGC 752 RG samplecontains three first-ascent RGBs, six RCs and one RHB.Spectroscopically-derived accurate temperatures, sur-face gravities, Li abundances and carbon isotopic ratios areamong the most important parameters to gain better un-derstanding about the evolutionary stages of the evolvedstars. However, matching observed HR Diagram locationsof NGC 752 RGs to isochrones is a challenge without know-ing stellar masses. We have a good estimate of the NGC 752turnoff mass due to comparisons of the observed data toboth the log g - T eff isochrone and the CMD. However, in-dividual cluster members might have slightly different ini-tial masses, which could lead us to observe some diversityamong the elemental abundances and C/ C ratios. Inter-nal mixing may have taken place with different efficiencies inthese stars. As a result, some of the RGs, especially with low C/ C ratios, may have experienced extra-mixing on theRGB after the 1st dredge-up, which takes place when low-mass stars reach the so-called luminosity-function bump.(Gratton et al. 2000; Charbonnel & Lagarde 2010).Consequently, RGs are important for understandinghow chemical evolution effects cluster stars. Additionally,they also provide excellent opportunities to investigate ef-fective mixing processes via their C/ C ratios. We willpresent more results that will investigate the chemical evo-lution of the RG members of several OCs in future papers.
ACKNOWLEDGMENTS
Major parts of this research occurred during several ex-change visits among the authors at the Department of As- c (cid:13) , 000–000 G. B¨ocek Topcu, M. Af¸sar, M. Schaeuble and C. Sneden tronomy and Space Sciences of Ege University and the De-partment of Astronomy of the University of Texas at Austin.We thank the people of both departments for their hospital-ity and encouragement of this study. Our work has been sup-ported by The Scientific and Technological Research Coun-cil of Turkey (T ¨UB˙ITAK, project No. 112T929), by the USNational Science Foundation (NSF, grants AST 09-08978and AST 12-11585), and by the University of Texas Rex G.Baker, Jr. Centennial Research Endowment. This researchhas made use of NASA’s Astrophysics Data System Bibli-ographic Services; the SIMBAD database and the VizieRservice, both operated at CDS, Strasbourg, France. This re-search has made use of the WEBDA database, operated atthe Department of Theoretical Physics and Astrophysics ofthe Masaryk University and of the VALD database, operatedat Uppsala University, the Institute of Astronomy RAS inMoscow, and the University of Vienna. We would like tothank to anonymous referee for his/her helpful discussionsand careful review of the manuscript.
REFERENCES
Af¸sar M., Sneden C., For B. Q., 2012, AJ, 144, 20Anthony-Twarog B. J., Twarog B. A., 2006, PASP, 118,358Asplund M., Grevesse N., Sauval A. J., Scott P., 2009,ARA&A, 47, 481Bartaˇsi¯ut˙e S., Deveikis V., Straiˇzys V., Bogdanoviˇcius A.,2007, Balt. Astron., 16, 199Bartaˇsi¯ut˙e S., Janusz R., Boyle R. P., Philip A. G. D., 2011,Balt. Astron., 20, 27Biazzo K., Frasca A., Catalano S., Marilli E., 2007a, As-tron. Nachr., 328, 938Biazzo K. et al., 2007b, A&A, 475, 981Boothroyd A. I., Sackmann I. J., 1999, ApJ, 510, 232Bressan A., Marigo P., Girardi L., Salasnich B., Dal CeroC., Rubele S., Nanni A., 2012, MNRAS, 427, 127Brooke J. S. A., Bernath P. F., Schmidt T. W., BacskayG. B., 2013, JQSRTJ. Quant. Spectrosc. Radiat. Transfer,124, 11Brooke J. S. A., Ram R. S., Western C. M., Li G., SchwenkeD. W., Bernath P. F., 2014, ApJS, 210, 23Brugamyer E., Dodson-Robinson S. E., Cochran W. D.,Sneden C., 2011, ApJ, 738, 97Caffau E., Ludwig H. G., Steffen M., Ayres T. R., BonifacioP., Cayrel R., Freytag B., Plez B., 2008, A&A, 488, 1031Carraro G., Monaco L., Villanova S., 2014, A&A, 568, A86Carrera R., Pancino E., 2011, A&A, 535, A30 (Car11)Castellani V., Chieffi A., Straniero O., 1992, ApJS, 78, 517Castelli F., Kurucz R. L., 2003, in N. Piskunov, W.W.Weiss, D.F. Gray, eds, Modelling of Stellar Atmospheres.IAU Symposium, Vol. 210, p. 20PCharbonnel C., 1994, A&A, 282, 811Charbonnel C., Lagarde N., 2010, A&A, 522, A10Charbonnel C., Brown J. A., Wallerstein G., 1998, A&A,332, 204Daniel S. A., Latham D. W., Mathieu R. D., Twarog B. A.,1994, PASP, 106, 281 (Dan94)Dearborn D. S. P., Eggleton P. P., Schramm D. N., 1976,ApJ, 203, 455 Delbouille L., Roland G., Neven L., 1973, Atlas pho-tometrique DU spectre solaire de λ λ c (cid:13) , 000–000 hemical compositions of RGs in the NGC 752 Mikolaitis ˇS., Tautvaiˇsien˙e G., Gratton R., Bragaglia A.,Carretta E., 2011b, MNRAS, 416, 1092Mikolaitis ˇS., Tautvaiˇsien˙e G., Gratton R., Bragaglia A.,Carretta E., 2012, A&A, 541, A137Moore C. E., Minnaert M. G. J., Houtgast J., 1966, Thesolar spectrum 2935 A to 8770 AO’Brian T. R., Wickliffe M. E., Lawler J. E., Whaling W.,Brault J. W., 1991, J. Opt. Soc. Am. B,, 8, 1185Paunzen E., Heiter U., Netopil M., Soubiran C., 2010,A&A, 517, A32Pilachowski C., 1986, ApJ, 300, 289Pilachowski C., Saha A., Hobbs L. M., 1988, PASP, 100,474P¨ohnl H., Paunzen E., 2010, A&A, 514, A81Ram R. S., Brooke J. S. A., Bernath P. F., Sneden C.,Lucatello S., 2014, ApJS, 211, 5Ram´ırez I., Mel´endez J., 2005, ApJ, 626, 465Reddy A. B. S., Giridhar S., Lambert D. L., 2012, MNRAS,419, 1350 (Red12)Roederer I. U., Sneden C., Thompson I. B., Preston G. W.,Shectman S. A., 2010, ApJ, 711, 573Roederer I. U., Preston G. W., Thompson I. B., ShectmanS. A., Sneden C., Burley G. S., Kelson D. D., 2014, AJ,147, 136Rogers F. J., Iglesias C. A., 1992, ApJ, 401, 361Sestito P., Randich S., Pallavicini R., 2004, A&A, 426, 809Shchukina N. G., Trujillo Bueno J., Asplund M., 2005, ApJ,618, 939Skrutskie M. F. et al., 2006, AJ, 131, 1163Smiljanic R., Gauderon R., North P., Barbuy B., Charbon-nel C., Mowlavi N., 2009, A&A, 502, 267Sneden C., 1973, ApJ, 184, 839Sneden C., Pilachowski C. A., 1986, ApJ, 301, 860Sneden C., Pilachowski C. A., Vandenberg D. A., 1986,ApJ, 311, 826Sneden C., Lucatello S., Ram R. S., Brooke J. S. A.,Bernath P., 2014, ApJS, 214, 26Strassmeier K. G., Schordan P., 2000, Astron. Nachr., 321,277Sweigart A. V., Mengel J. G., 1979, ApJ, 229, 624Sweigart A. V., Greggio L., Renzini A., 1989, in Bulletinof the American Astronomical Society. Vol. 21, p. 789Tautvaiˇsiene G., Edvardsson B., Tuominen I., Ilyin I., 2000,A&A, 360, 499Tautvaiˇsien˙e G., Edvardsson B., Puzeras E., Ilyin I., 2005,A&A, 431, 933Tomkin J., Lambert D. L., Luck R. E., 1975, ApJ, 199, 436Tomkin J., Luck R. E., Lambert D. L., 1976, ApJ, 210, 694Torres G., 2010, AJ, 140, 1158Torres G., Fischer D. A., Sozzetti A., Buchhave L. A., WinnJ. N., Holman M. J., Carter J. A., 2012, ApJ, 757, 161Tull R. G., MacQueen P. J., Sneden C., Lambert D. L.,1995, PASP, 107, 251VandenBerg D. A., Bergbusch P. A., Dowler P. D., 2006,ApJS, 162, 375Wallace L., Hinkle K. H., Livingston W. C., Davis S. P.,2011, ApJS, 195, 6Yong D., Carney B. W., Teixera de Almeida M. L., 2005,AJ, 130, 597 c (cid:13)000