The metallicity of the old open cluster NGC 67911 Raffaele Gratton2, Angela Bragaglia3, Eugenio Carretta3 and Monica Tosi3 [email protected], [email protected], [email protected], [email protected] 6 0 0 2 ABSTRACT n We haveobservedfourredclumpstarsinthe veryoldandmetal-richopenclusterNGC 6791 a to derive its metallicity, using the high resolution spectrograph SARG mounted on the TNG. J Using a spectrum synthesis technique we obtain an average value of [Fe/H] = +0.47 (±0.04, 2 r.m.s. = 0.08) dex. Our method was tested on µ Leo, a well studied metal-rich field giant. 1 We also derive average oxygen and carbon abundances for NGC 6791 from synthesis of [O i] at v 6300 ˚A and C2 at 5086 ˚A, finding [O/Fe] ≃−0.3 and [C/Fe] ≃−0.2. 7 2 Subject headings: Stars: abundances – Galaxy: disk – Galaxy: open clusters – Galaxy: open clusters: 0 individual: NGC 6791 1 0 1. Introduction color-magnitude diagrams technique (Tosi et al. 6 1991), and will be combined with metal abun- 0 The determination of the abundances in stars dances from high resolution spectroscopy to pro- / ofdifferent ageandGalactic locationis oneof the h vide robust constraints on the current and past p basic tools to interpret the chemical evolution of disk properties. - the Milky Way disk. Galactic open clusters are o We have already presented the detailed chem- particularly well suited to this purpose (e.g., Friel r ical abundances of a few old open clusters (Bra- t 1995)sincetheyreachagesasoldasthediskitself, s gaglia et al. 2001; Carretta et al. 2004, 2005). To a cover a large range in metallicities and ages, are oursamplewenowaddNGC 6791which,withan : observedinregionsoftheGalacticdisklikelychar- v age of about 9-10 Gyr, is one of the oldest open i acterizedbydifferentstarformationhistories,and X clusters of our Galaxy and has super solar metal- their distances and ages can be determined with licity (e.g., Peterson & Green 1998; Chaboyer et r a precisionnotreachablefor fieldstars,exceptfor a al. 1999; Stetson et al. 2003; Carney et al. 2005; the nearest ones. King et al. 2005). This cluster, almost as old as We are presently studying in an accurate and the Galactic disk, is of paramount importance to homogeneous way a significant sample of open study the time evolution of the disk properties. clusters (Bragaglia & Tosi 2006 and references Apart from its age, NGC 6791 is interesting for therein): reliable distances, reddenings and ages hispeculiarHorizontalBranch(HB),mostlycom- are derived from photometry with the synthetic posed by red stars, but with a (unusual) blue tail (Kaluzny&Udalski1992;Liebertetal.1994). Its 1 Basedonobservations madewiththeItalianTelesco- studycouldberelevantforanumberofissues,like pioNazionaleGalileo(TNG)operated onthe islandofLa PalmabytheFundacio´nGalileoGalileioftheINAF(Isti- e.g., the UV upturn in elliptical galaxies, the HB tutoNazionalediAstrofisica)attheSpanishObservatorio morphology and its connection with mass loss in delRoquedelosMuchachosoftheInstitutodeAstrofisica globular clusters. deCanarias 2INAF-OsservatorioAstronomicodiPadova,vicoloOs- Despite its peculiarities and its importance for servatorio5,I-35122Padova, Italy Galacticformationstudies,onlyonedetailedwork 3 INAF-OsservatorioAstronomicodiBologna,viaRan- dealing with the chemical pattern in NGC 6791, zani1,I-40127Bologna,Italy basedonmodernfineabundanceanalysisandhigh 1 resolutionspectroscopy,hasbeenpublishedsofar: correction,spectraextraction,andwavelengthcal- Peterson&Green(1998)analyzedthecoolestand ibration; the individual RVs were derived and the brightest (at V = 15.0, B −V = 0.48) blue HB spectrawereshiftedtozeroRVandaveraged. Ta- star. Besides this case, really high S/N, high res- ble1liststheidentifications,coordinatesandmag- olution spectra can be obtained in acceptable ex- nitudes for the stars, together with the heliocen- posure times only for the brightest, hence cooler tricRVsobtainedaveragingallindividualonesfor giantsinthiscluster. Attheverylargemetallicity each stars. Also shown is the S/N of the summed of NGC 6791 these pose a great challenge to the spectra measured near 6000 ˚A by comparing the observers(see next Section) and great care has to spectrum of the starswith those of µ Leo (see be- betakentoensurethereliabilityofthe analysisof low),thatcanbeconsideredvirtuallynoiselessfor their extremely crowded spectra. the purposes of this comparison. This paper is organized as follows: Sect. 2 Tocross-checkourmethodofabundancederiva- presents our data; atmospheric parameters and tion, we also analyzed a spectrum of µ Leo with iron abundances from spectrum synthesis are de- the same procedure. This star was selected be- scribedinSect. 3;Sect. 4providestheabundances cause it has stellar parameters and iron abun- ofcarbonandoxygen;adiscussionandasummary dancessimilartothoseexpectedintheNGC 6791 are given in Sect. 5. sample. The spectrum of µ Leo was acquired us- ing the FEROS spectrograph at the ESO 1.5 m 2. Observational material telescope at La Silla. The original spectrum has a resolution of R ∼ 48,000; however, in order to The spectra of the giants of NGC 6791 are ex- make the analysis as similar as possible to that tremely rich of lines, due to the rather cool tem- of the stars of NGC 6791, this spectrum was de- perature and the high metal content. To alleviate graded at the same resolution of our SARG spec- theanalysisproblems,wefocusedourattentionon tra. As evident from Figure 1, at this resolution (fainter)starsontheredclump,whicharewarmer the spectrum of µ Leo is indeed very similar to than the red giants. Even these spectra are actu- those of the stars of NGC 6791, except for the allyatleastasrichinlinesasthatofthecanonical higher S/N,and verysubtle differences in the line very metal-rich giant µ Leo (see Figure 1). The strengths. The fact that we are dealing with a long debate on the appropriate abundance to be different, lower resolution spectrum than the one attributedtothisstar(seee.g. Gratton&Sneden analyzedbyGratton&Sneden(1990)andadopta 1990) emphasizes the difficulties of the derivation different solar model explain the slight differences ofcorrectabundancesfromtheveryline-richspec- in the results with that paper (see Sect. 3.4). tra of cool giants. We chose our targets among the stars listed 3. Abundances in NGC6791 from synthe- as cluster members by Friel et al. (1989) on the sis of Fe lines basis of their radial velocities (RVs). They are allconfirmedmembersbyourownmeasurements. In NGC 6791 even the spectra of red clump All spectra were obtained with the high resolu- stars are so crowded that we deemed the tradi- tion spectrograph SARG mounted at the Italian tionalanalysisbasedonequivalentwidthsmeasure National Telescope Galileo (TNG) on Canary Is- not entirely reliable,due to the large inherent un- lands. TheresolutionisR=29,000andthewave- certainties in both the location of the continuum length coverageis 4620 - 7920˚A. Individual spec- level and the presence of blends (see Gratton & tra have been obtained (mostly in service mode) Sneden 1990 and Smith & Ruck 2000 for a simi- from November 2001to September 2005,with ex- larapproachinthe caseofspectraofmuchhigher posures ranging from 1 to 1.5 hours each. For resolution and S/N of µ Leo). eachstar,the total exposuretime rangesfrom4.5 We then derived iron abundances for our pro- to 8 hours. All spectra have been reduced using a gram stars by comparing the observed profiles for standard IRAF1 procedure for bias and flat field Observatory,whichareoperatedbytheAssociationofUni- 1IRAFisdistributedbytheNationalOpticalAstronomical versities for Research in Astronomy, under contract with theNationalScienceFoundation 2 anumberofFelinestosynthesesofsmallspectral modelatmospheresextractedfromthegridofKu- regions (typical width ∼ 2.5 ˚A) around the cho- rucz (1995a). For consistency with other papers sen lines. This procedure allows to take fully into analyzing stars in open clusters (see e.g. Carretta account the presence of blends. Furthermore, the et al. 2004), the models considered in this paper correctpositioningofthe continuumlevelis much are those with the overshooting option included. lessaproblem,sincewemaycomparedirectly the The fitting of the synthetic spectra to the ob- highest points of the spectrum in the observedre- served ones was done by eye. There is of course gions with those present in the synthetic spectra, some arbitrariness in the eye fitting, since differ- insofar we trust the line lists used in our analysis. ent weights can be attributed to the line cores or Theselistswerebuilttakingappropriatelyintoac- wings. However, we found that our eye fitting count both the inclusion of all relevant lines and givesthe sameaverageabundances,butmuchless the quality ofthe gf values,carefully discussedin line-to-line scatter, than a fitting based on more Gratton et al. (2003). objective criterions, such as a least square fitting These (extensive) line lists were obtained after to the data. The reason for the smaller scatter carefulcomparisonswithboththespectrumofthe likely reflects a better estimate of the appropriate Sun and of HR 3627. This is a cool (∼ 4200 K), levelofthelocalcontinuum,whichisafreeparam- very metal-rich ([Fe/H]∼ +0.3) star. With this eter in the fitting, and it is a critical issue when combinationofparametersthe strengthoflines in determining abundances. HR 3627 is typically similar to (or even stronger Thelineparameters(oscillatorstrengths,damp- than) that of the red clump stars of NGC 6791. ingbroadening)wereobtainedusingthesamepre- Wearefullyconfidentthatalinelistwellmatching cepts adopted in Gratton et al. (2003). The same the spectrumofthisstaralsoprovidessensiblere- line parameters and microturbulent velocity were sultsforourprogramstars. Thislengthyprepara- adopted for all stars. tory work on HR 3627 was based on excellent ob- servational material: its spectrum, acquired with 3.1. Atmospheric Parameters SARG, has both very high S/N (>400)and reso- lution(R∼150,000),sothatevenextremelyfaint Effective temperatures (Teff) and surface grav- ities (logg) were obtained from the photometry, possible contaminants could be detected and in- 2 usingB,V valuesfromMontgomeryetal.(1994) cludedinthelinelists. WeselectedanumberofFe and K magnitudes from 2MASS (Cutri et al. i andFe iilines thatwere free fromnearbystrong features. We restrictedto the spectral rangefrom 2003). WederivedtheTeff’sfromtheV −K colors about5500to7000˚A,toavoidthelowresponseof assuming the calibration by Alonso et al. (1999) and a reddening of E(B−V)=0.15, in the mid- thespectrographintheblueregionandfringingor dle of literature determinations. We prefer not to severetelluriccontaminationredwardoftheupper wavelengthlimit. Aroundeachlines(within±2˚A) use temperatures from the B−V colors because of their strong dependence on metal abundance. we extracted lines of neutral and singly ionized We have found a posteriori that the values of the atomicspeciesfromthe Kuruczdatabase(Kurucz effectivetemperaturescorrespondingtotheB−V 1995b). We also included molecular lines, in par- colorsagreewellwiththosefromtheV −K colors ticular of contaminant CN and hydrides. Lines forthe metalabundanceofNGC 6791determined unaccountedforintheKuruczdatabaseand/orin in the present analysis. Also, we prefer not to the solar tables (Moore et al. 1966)were assumed use temperatures derived from line excitation be- to be Fe i lines with an excitation potential of cause for the lines consideredhere there is a quite 3.5eV. The transitionprobabilitiesofthe Fe lines strong degeneracy between effective temperatures we wanted to synthetize were left untouched with and microturbulent velocities (see below). respecttothelinelistweusedintheEW analysis; thegf valuesofthenearbylineswereadjustedone The calibrationof the colour-temperaturerela- by one by matching the high resolution spectrum tions of Alonso et al. (1999) was used for consis- of HR 3627. For more details about HR 3627 and tency with the analyses we are performing of sev- these line lists, see Carretta et al. (2004). 2but other photometries arealmostidentical, seee.g. Stet- The synthetic spectra were obtained using sonetal.(2003) 3 eralotheropenclusters. Itshouldbe noticedthat whatGratton&Sneden(1990)andSmith&Ruck their calibrationonly extends up to [Fe/H]=+0.2, (2000) adopted. Uncertainties in these microtur- andapplicationto the starsofNGC 6791requires bulentvelocitiesareapproximatelyof0.08kms−1 anextrapolation. However,theV −K colorindex for a given effective temperature; this is obtained usedinthispaperisonlyverymarginallysensitive bymodifying v fromitsbestvalue untiltheslope t tometalabundance,sothaterrorsinourtempera- becomes equal to its statistical error. Note how- tures cannotbe large. Errorsin our effective tem- everthatthereisastrongcorrelationbetweenmi- peratures are mainly due to uncertainties in the croturbulentvelocitiesandeffectivetemperatures: assumedreddening: anerrorof∆E(B−V)=0.04 adopting temperatures 100 K higher, we would magnitudes, a reasonable upper limit in the case have derived microturbulent velocities ∼ 0.13 km of NGC 6791, implies an error of ∼ 100 K in the s−1 larger. assumed temperatures. For consistency, a similar Finalmetallicitiesareobtainedbyinterpolating approach was adopted for µ Leo, with the V −K in the Kurucz (1995a) grid of model atmospheres color from Johnson (1966) photometry. Notice (with the overshootingoption on) the model with thatthis Teff for µLeois slightlycooler(by 50K) the proper atmospheric parameters whose abun- thanthatderivedbyGratton&Sneden(1990)us- dance matches that derived from Fe i lines. ing the Infrared Flux Method. Surface gravities were obtained from the loca- 3.2. Abundances from individual lines tion of the stars in the color-magnitude diagram, Fig. 2showsanexampleofthequalityofthefit assumingadistancemodulusof(m−M) =13.45 V of three of the observed iron lines with synthetic -inthemiddleofliteraturedeterminations-bolo- spectra. The synthetic spectra have been com- metriccorrectionsfromAlonsoetal.(1999),anda puted with atmospheric parameters appropriate mass of 0.9 M⊙. Most of the errors in the surface for the star, and Fe abundances of logn(Fe)=7.6, gravities stem from uncertainties in the distance 7.8, 8.0, 8.2, 8.4, 8.6, and 8.8. The lines consid- modulus: an error of (m − M) = 0.5 mag, a V ered in the analysis are marked with a dash in reasonableupper limit, implies anerrorof0.2dex Fig. 2. From these three comparisons we con- in the surface gravities. For µ Leo, we adopted cluded for a best value of the Fe abundance of the value of the gravity given by Gratton & Sne- logn(Fe)=8.00. Note that other Fe i lines falling den(1990)usingavarietyofmethods(equilibrium in the same spectral ranges of the program lines of ionization for Fe, dissociation equilibrium for generally give Fe abundances in good agreement MgH, and pressure broadened lines): this value is with those indicated by the lines selected in our in fact almost identical to that of the NGC 6791 analysis, although they were not actually used in stars. the estimate of the best value for each star. Microturbulentvelocities(vt)wereobtainedby Table 2 gives abundances for Fe lines3 as ob- eliminating trends of abundances with respect to tained from the comparison of observations with the expected line strength (Magain 1984) X = spectral synthesis. Lines with expected line loggf − EP ×Θ , where loggf is the oscilla- exc strength X > −5.6 were not used in the anal- torstrengthofthelines,EP theexcitationpoten- ysisbecause thereis atrendfor these linesto give tial (in eV), and Θexc = 5040/(0.86×Teff) rep- too low abundances. This is likely a reflection of resents the approximate temperature of the lay- the weightwe gavein our abundance estimates to ers where most of the lines form. This implies the cores of these lines. These form at very tiny that abundances are roughly independent of line optical depths, where the adopted model atmo- strength. We considered here the average values spheres are probably not adequate and deviations of the abundances derived for the individual lines from LTE become important. Anyhow, such lines of all the stars of NGC 6791, in order to reduce are strongly saturated, so that they would not be the scatter and better evidentiate possible trends. Thismeansthatthesamevalueofthemicroturbu- 3Theseabundancesarebynumber;weusetheusualspectro- lent velocity was adopted for all stars. The same scopic notations: logn(A) is the abundance (by number) value was also adopted in the analysis of µ Leo, oftheelementAinthescalewherelogn(H)=12;[A/H]is and it is slightly smaller (by 0.15 km s−1) than thelogarithmicratiooftheabundances ofelementsAand Hinthestar,minusthesamequantityintheSun. 4 good abundance indicators. Since the four target small difference wouldhave canceled, hadwe cho- starsarevery similarto eachother,it is meaning- sentemperatures about 50 K lowerthanadopted, ful to average the results for the individual stars, or gravities about 0.15 dex larger. and to derive the average abundance provided by each line. This is given in the last column of 3.3. Errors on the derived abundances the table; the associated error bar is simply the Table 4 gives the sensitivity of the derived dispersion of the mean. abundances on the assumptions on the atmo- Fig. 3 shows graphically the lack of trends in spheric parameters of the program stars (listed this average Fe abundances with wavelength, EP, in Column 1). Column 2 gives the considered and line intensity parameter X, thus reinforcing parametervariation,andColumns 3 and4 the re- the reliability of the derived abundances. sulting variation in the abundances derived from Table 3 summarizes the abundances obtained Fe i and Fe ii lines respectively. Column 5 lists for each star. The second, third and fourth theestimatedvalueforthesystematicerrorinour columns give the values adopted for the atmo- analysis for each of the parameters, and Columns spheric parameters (Teff, logg, [A/H]), while the 6 and 7 the corresponding uncertainties in the microturbulent velocity is the same for all stars abundances from Fe i and ii lines, respectively. (vt =1.05 km s−1). In Columns 5 to 10 we The total errors listed on the bottom line of give the average values of the abundances of Fe Table 4 are obtained summing quadratically the from neutral and singly ionized lines, along with contribution of each source of error,including the the number of lines used in the analysis and the fitting error. This estimate of the total error is r.m.s. scatter of individual abundances. Finally, computedassumingazerocovariancebetweenthe in the lasttwo columns we give the [Fe/H]values. effects of errors in the atmospheric parameters. The reference solar abundances adopted here are In principle, this assumption is not strictly valid, logn(Fe)=7.54 and 7.49 from neutral and singly since there are correlations between different pa- ionized lines, respectively; those are the values we rameters. Inpractice,the Fe abundanceis mainly obtainedfromasolaranalysiscompatiblewiththe a function of the adopted effective temperatures. present one (see Gratton et al. 2003). In fact, adopting Teff values e.g. 200 K higher,we Thelastline ofTable 3givesthe averageabun- wouldhave obtained higher microturbulentveloc- dance for the cluster. The average Fe abundance ities (by0.26kms−1),andFeabundancessmaller fromneutrallines is[Fe/H]=+0.47±0.04(r.m.s. by about 0.15 dex. Notice that since the effect on ofindividualstarsequalto0.07dex). Theline-to- microturbulent velocity is larger than the direct line scatter foreachindividualstaris in the range effect of temperature variations, the abundances 0.11-0.18dex. This is dominated by uncertainties getsmallerwithincreasingmodeltemperatures,at still presentin the locationofthe continuum level variance with the usual dependence for late type within the small spectral windows considered in stars. Unluckily, giventhe correlationbetween ef- this analysis. fective temperature and microturbulent velocity, If we consider the averageabundances for each wecouldnotderivereliableeffectivetemperatures line derived from the four spectra, the line-to-line from the excitation equilibrium. scatter is 0.08 dex, as expected by assuming that There are quite strong arguments favoring the thefourresultsareindependentofeachother. For solution adopted throughout this paper. In fact, theseaverageabundances,thereisasmall,notsig- had we adopted the hypothetical temperatures nificant,trendwithlineexcitation: ∆[Fe/H]/∆EP warmer by 200 K mentioned above (that would =0.009±0.012dex/eV.Thisimpliesthattemper- be obtained assuming a reddening of E(B−V)= atures from line excitation are 38±50 K higher 0.23, a value near the upper limit in the analy- thanthoseadoptedhere. Wedeemthisagreement ses of NGC 6791), abundances from Fe ii lines as fully satisfying. wouldhaveresultedmuchlowerthanthosederived There is alsoa small- but againnot significant from Fe i lines, with an offset of 0.24 dex, rather - difference between the abundances of Fe i and than 0.04 dex. In principle, it should be possi- Fe ii: [Fe/H]i − [Fe/H]ii = 0.04±0.07 dex. This ble to compensate for such an offset by increasing the surface gravities by 0.5 dex (up to values of 5 logg ∼ 2.8). However, this would require a dis- of the abundance agrees with that of Gratton & tance modulus of (m−M) ∼12.15, incompati- Sneden(1990). The[Fe/H]valuegivenbyGratton V ble with the distances derived from the literature & Sneden (1990) is slightly lower than that esti- color-magnitude diagrams. The absolute magni- mated in this paper ([Fe/H]=+0.34±0.03), due tudeofcoreHe-burningstars(i.e.,theclumpstars to the use of a different solar abundance, derived we are observing) is expected to be M ∼ 1.2 using the Holweger & Muller (1974) model atmo- V according to the models by Girardi & Salaris sphere, rather than that obtained from the same (2001). This corresponds to a distance modulus grid considered for the red giants. More recently, of (m − M) ∼ 13.4−13.5, in agreement with Smith&Ruck(2000)usedaverysimilartechnique V the value used in our analysis. Hence, an upper and set of atmospheric parameters, and obtained limitfortheeffectivetemperaturesoftheprogram a lower value of logn(Fe) = 7.79±0.03 (internal starsinNGC6791isabout100Khigherthanthe error). This value is about 0.13 dex lower than adopted values (corresponding to a reddening of the presentoneandthe differencecanbe ascribed E(B −V) < 0.18, and a microturbulent velocity to the larger value of the microturbulent velocity of v < 1.18 km s−1). This implies that the lower adopted by Smith & Ruck (2000) (1.22 km s−1 t limit of the Fe abundance is [Fe/H]> +0.39 dex. rather than 1.05 km s−1: see Table 4). These Similar arguments can be used to define a robust comparisons show that systematic errors in our upper limit of [Fe/H]<+0.55. analysis are within the adopted error bar. Summarizing, the Fe abundance we derive for By comparing the abundances in NGC 6791 NGC 6791is [Fe/H] = +0.47±0.04±0.08, where with those of µ Leo, we conclude that the for- the first error value represents the random term, mer has an Fe abundance larger than the latter as derived from the star-to-star scatter, and the by 0.09 ± 0.05, where the error here is simply second error value the systematic errors, due to the quadratic sum of the internalerrors,since the the assumptions on reddening, distance modulus, other uncertainties should cancel out in the dif- mass, temperature scales, etc. ferential abundance analysis. This small positive differencewellagreeswiththeeyeimpressionfrom 3.4. A comparison with µ Leo Figure 1. To further assess the soundness of the results 4. Abundances of oxygen and carbon of our spectrum synthesis analysis, we also ana- lyzed the well known, metal-rich field star µ Leo, We estimated the abundance of oxygen from strictly using the same criteria. The average Fe the [O i] line at 6300.31 ˚A. The second, weaker, abundance we derived is logn(Fe) = 7.92±0.03 [O i]line at6363.79˚A is barelymeasurableinour (29 lines, r.m.s.=0.15dex)fromneutrallines, and spectra, and the three O lines near 7775 ˚A (that logn(Fe) = 7.79±0.04 (6 lines, r.m.s.=0.10 dex) are not the best suited for this kind of stars) fall fromsinglyionizedlines. Thesevaluescorrespond in a regionwhere strong interference fringes make to [Fe/H]=+0.38±0.03 and [Fe/H]=+0.30±0.04 it difficult to extract accurate information from respectively. The difference in the abundances spectra. providedbyneutralandsinglyionizedlinesisonly The6300.31˚Alineoftenhappenstobestrongly marginally larger than the internal error, and is affected by telluric contamination. However, in similar to the corresponding difference found for the case of NGC 6791, the cluster RV relative to the stars in NGC 6791. Earth motion makes the [O i] line non contami- Thisvalueofthe FeabundanceofµLeocanbe natedbytelluricabsorptionsinthemajorityofour compared with literature estimates. We will con- spectra. Compromising between a slightly lower sider only abundances based on very high quality S/N and a lower data manipulation, we decided spectra (R ∼ 100,000, S/N > 200) and compar- to average only the undisturbed spectra for each isons with synthetic spectra, since use of equiva- star, and measure the O abundance from them. lent widths is not reliable for such line-rich spec- InordertoderivereliableOabundances,acare- tra. Gratton&Sneden(1990)derivedlogn(Fe)= ful synthesis of the [O i] lines is necessary,includ- 7.97 ± 0.02 ± 0.15 (36 lines, r.m.s.=0.12 dex). ingnotonlythecontributionofthenearbyNiline Within the internal errors, the present estimate 6 butalsotherelevantcouplingwithCabundances, etal.2005),reachingalmostthehydrogen-burning and the contamination by CN lines. While a full limitonthemainsequenceanddefiningtheWhite discussion of CNO abundances is deferred to a Dwarfs cooling sequence. forthcomingpaper,herewepreliminarilyestimate Spectroscopic studies of NGC 6791 are fewer, the C abundances from the spectral synthesis of but present interesting results. Since the pioneer- the C2 molecularfeaturesat5086˚A.Fig. 4shows ing work by Spinrad & Taylor (1971) the cluster an example of matching of synthetic spectra for has been recognized to be more metal-rich than star 3019 in the forbidden [O i] line and in the the Sun; they found [M/H] = +0.75, but with a C2 regions. The synthetic spectra were computed methodthatattributes[M/H]≃+0.6toNGC188 assuming [N/Fe]=−0.2,and 12C/13C=8. The lat- andM67,bothpresentlyknowntohavenearlyso- ter value is adequate for low mass evolved giants, lar metallicity. while the former one provides CN features in rea- More recently, the red HB (that we call red sonable agreement with observations for the best clump) has been studied by low resolution spec- C abundance. troscopybyHufnageletal.(1995);theywantedto Combining the C and O abundances provided investigate possible correlations between CH and by these two abundance indicators, we find aver- CN band strengths and found none, at variance age abundance ratios of [C/Fe] = −0.23 dex (a with similar studies on globular clusters. rather normal value for clump stars), and [O/Fe] Worthey & Jowett (2003) took low resolution = −0.32 dex. The comparisons with synthetic spectra of 23 K giants in NGC 6791 and derived, spectra clearly show that the O abundance can- usingLick/IDSindices,[Fe/H]=+0.320±0.023± not be much larger than this value. It cannot be σ , where σ , the systematic errorattachedto sys sys much lower either, otherwise the features due to their metallicity scale, is unknown. We have only C-bearing molecules would be much stronger(the one star in common (3019, their R25) for which spectra of the stars clearly shows that O≫C). the quoted abundance is [Fe/H] = +0.341. Asa comparison,Gratton& Sneden(1990)de- All the four stars analyzed here were observed rived for µ Leo [C/Fe] = [O/Fe] = −0.15 dex. at low resolution by Friel et al. (1989), where NGC 6791 is then slightly more deficient in both we adopted identification and membership sta- O and C with respect to µ Leo, along the stan- tus from. The same group (Friel & Janes 1993) dardtrendofmorepronounceddeficiencyofthese later published their metallicities: [Fe/H] values elements with increasing metal abundance (see are +0.32 ± 0.28, +0.23 ± 0.33, +0.39 ± 0.15, Bensbyetal.2005;Andersson&Edvardsson1994) +0.41±0.21forstars2014,3009,3019,andSE49, respectively. These values (that would give an 5. Discussion and summary average metallicity of about +0.34) do not dif- fer much from our new ones, considering that the The first detailed study of NGC 6791 of which scaleusedbyFriel&Janes(1993)givesonaverage we are aware of is by Kinman (1965), who pub- metal-poorer values than others. The final result lished a photographic color-magnitude diagram of their analysis depends on the calibration of in- and derived information on membership and in- dicesusingstandardstarswithmetallicityderived trinsic colors (i.e., reddening) from low resolution by several different sources. Due to a recent revi- spectra. sion of their abundance scale (Friel et al. 2002), There is a general agreement that this cluster these abundances were lowered to +0.12±0.13, is at about the same Galactocentric distance as +0.11±0.13,+0.15±0.11and+0.16±0.11forthe the Sun, is about twice as old, and about twice same stars; for their total sample of 39 stars they as metal-rich. Nevertheless, there are significant derived an average metallicity of +0.11, σ = 0.10 differencesbetweenpropertiesderivedbydifferent dex. authors: (m−M)0 ∼ 12.6−13.6, E(B −V) ∼ 0.09−0.23,age∼7−12Gyr,[Fe/H]∼0.1−0.4(for To date, the only published analysis based on high resolution spectroscopy remains that by Pe- arecentreviewseeStetsonetal.2003). NGC6791 terson & Green (1998). They obtained 4 hours hasalsobeenobservedwiththeACSonboardthe of integrationat the 4 m Mayalltelescope at Kitt Hubble Space Telescope (King et al. 2005; Bedin 7 Peak on a blue HB star that is a cluster member −0.32 and [C/Fe] = −0.23 dex, on average, for bothbypropermotionandRV.Thesummedspec- NGC 6791. trumhasS/Nandresolutionlowerthanours(S/N NGC 6791 represents a challenge to any model ∼ 30per pixel, R=20000),but is lessaffectedby for the formation of the Galactic disk, a possible blends because of the much higher temperature missing link between globular and open clusters, (about7300K).TheyemployedbothEWs(forFe a genuine puzzle that modern abundance analysis and atmospheric parameters determination) and basedon highresolutionspectroscopywill help to spectrumsynthesis(foralltheothermeasuredele- unveil. ments). They discussedthe use ofdifferentscales, e.g. for loggf’s and solar reference abundances. We thank the TNG staff who helped collecting On the scale they chose to adopt, their star has the spectra. We gratefully acknowledge the use [Fe/H] = +0.37 (±0.10) dex; since their solar Fe of the BDA, created and updated by J.-C. Mer- is logn = 7.67 and ours is 7.54, their value corre- milliod for many years. This project has been sponds to [Fe/H] = +0.50 on our scale. partially supported by the Italian MIUR, under Takenatfacevalue,theagreementwithourre- PRIN 2003029437. sultsappearsexcellent,butitdoesn’ttakeintoac- count many differences like e.g., loggf’s or model REFERENCES atmospheresandthe like. For instance,we have4 Alonso, A., Arribas, S., & Martinez-Roger, C. FelinesincommonwithPeterson&Green(1998), 1999,A&AS, 140, 261 1 of Fe i and 3 of Fe ii (see Table 2 and their table 1), and the average difference in loggf’s Andersson,H.,&Edvardsson,B.1994,A&A,290, is about 0.29 (our values minus theirs). They 590 also measured the abundances of severalelements by means of extensive spectrum synthesis, finding Bedin, L.R., Salaris, M., Piotto, G., King, I.R., strong overabundances with respect to the solar Anderson, J., Cassisi, S., & Momany, Y. 2005, ratio for N and Na (+0.5 or +0.4) and slightly ApJ, 624, L45 loweronesforMgandSi(+0.2)orEu(+0.1). All Bensby, T., Feltzing, S., Lundstr¨om, I., & Ilyin, the other elements, including C and O, display I., 2005, A&A, 433, 185 solar-scaled abundances. In our analysis we have measuredsofaronlyCandOabundances,finding Bragaglia, A., & Tosi, M. 2006, AJ, in press, for them sub-solarratios. A complete comparison March 2006 issue (astro-ph/0511020) between the two analyses is beyond the purpose of the present paper, also because the chosen tar- Bragaglia,A., et al. 2001, AJ, 121, 327 getsarequitedifferentinproperties,butitwillbe Carney, B.W., Lee, J.-W., & Dodson, B., 2005, done once we have measured all other elements. AJ, 129, 656 Summarizing, we have derived the metallicity of NGC 6791 analyzing the high-resolution, high Carretta,E.,Bragaglia,A.,Gratton,R.G.,&Tosi, S/N spectra of four red clump stars, finding an M. 2004, A&A, 422, 951 average [Fe/H] = +0.47±0.04±0.08, where the Carretta,E.,Bragaglia,A.,Gratton,R.G.,&Tosi, two error values represent the random term and M. 2005, A&A, 441, 131 the systematic one, respectively. This result has been obtained using spectrum Chaboyer, B., Green, E.M., & Liebert, J. 1999, synthesis of Fe i and Fe ii lines, and adopting at- AJ, 117, 1360 mospheric parameters from the photometry. The Cutri,R.M.,etal.2003,VizieROn-lineDataCat- soundness of our analysis has also been checked alog: II/246, Originally published in: Univer- relatively to a well studied, metal-rich field giant sity of Massachusetts and Infrared Processing star, µ Leo. and Analysis Center, (IPAC/California Insti- We have measured O and C abundances using tute of Technology) synthesis of the [O i] line at 6300.31 ˚A and of the C2 molecular feature at 5086 ˚A, finding [O/Fe] = Friel, E. D. 1995, ARA&A, 33, 381 8 Friel, E.D., Liu, T., & Janes, K.A 1989, PASP, Peterson, R.C., & Green, E.M. 1998, ApJ, 502, 101, 110 L39 Friel, E.D., & Janes, K.A 1993, A&A, 267, 75 Smith, G., & Ruck, M.J. 2000,A&A, 356, 570 Friel, E. D., Janes, K. A., Tavarez, M., Scott, J., Spinrad, H., & Taylor, B.J. 1971, ApJ, 163, 303 Katsanis, R., Lotz, J., Hong, L., & Miller, N. Stetson, P.B., Bruntt, H, & Grundahl, F. 2003, 2002,AJ, 124, 2693 PASP, 115, 413 Girardi,L.,&Salaris,M.2001,MNRAS,323,109 Tosi, M., Greggio, L., Marconi, G., & Focardi, P. Gratton,R.G.,Carretta,E.,Claudi,R.,Lucatello, 1991,AJ, 102, 951 S., & Barbieri, M. 2003, A&A, 404, 187 Worthey, G., & Jowett, K.J. 2003, PASP, 115, 96 Gratton, R. G., & Sneden, C. 1990, A&A, 234, 366 Holweger,H.,&Mueller,E.A.1974,SolarPhysics, vol. 39, p. 19 Hufnagel, B., Smith, G.H., & Janes, K.A. 1995, AJ, 110, 693 Johnson, H.L. 1966, ARA&A, 4, 193 KaluznyJ., &Udalski, A.1992,Acta Astron.,42, 29 King, I.R., Bedin, L.R., Piotto, G., Cassisi, S., & Anderson, J. 2005, AJ, 130, 626 Kinman, T.D. 1965, ApJ, 142, 655 Kurucz, R.L. 1995a, CD-ROM 13, Smithsonian Astrophysical Observatory,Cambridge Kurucz, R.L. 1995b, CD-ROM 23, Smithsonian Astrophysical Observatory,Cambridge Liebert, J.,Saffer,R.A., & Green,E.M.1994,AJ, 107, 1408 Magain, P. 1984, A&A, 134, 189 Mermilliod, J.-C. 1995, in Information and On- Line Data in Astronomy, D. Egret, M.A. Albrecht eds, Kluwer Academic Press (Dor- drecht), p. 127 Montgomery K.A., Janes K.A., Phelps R.L. 1994, AJ, 108, 585 Moore, C.E., Minnaert, M.G.J., & Houtgast, J. 1966, National Bureau of Standards Mono- graph, Washington: US Government Printing Office This2-columnpreprintwaspreparedwiththeAASLATEX macrosv5.2. 9 Table 1 Information on the four observed stars. ID ID RA Dec V B−V K RV S/N (FLJ) (BDA) (2000) (2000) 2MASS km s−1 2014 2562 19 21 01.1 37 46 39.6 14.563 1.403 11.468 −48.60 60 3009 3363 19 20 56.5 37 44 33.7 14.638 1.387 11.556 −48.26 50 3019 3328 19 20 56.2 37 43 07.7 14.675 1.334 11.586 −46.10 85 SE49 3926 19 20 53.1 37 45 33.4 14.540 1.346 11.513 −45.63 40 References. — FLJ: Friel et al. (1989); BDA: Mermilliod (1995); B, B−V: Montgomery et al. (1994); K: Cutri et al. (2003); RV: this paper 10