WIYN Open Cluster Study LXXXV. Li in NGC 2243 -- Implications for Stellar and Galactic Evolution
DDraft version February 11, 2021
Typeset using L A TEX twocolumn style in AASTeX63
WIYN Open Cluster Study LXXXV. Li in NGC 2243 -Implications for Stellar and Galactic Evolution
Barbara J. Anthony-Twarog, Constantine P. Deliyannis, and Bruce A. Twarog Department of Physics and Astronomy, University of Kansas, Lawrence, KS 66045-7582, USA Department of Astronomy, Indiana University, Bloomington, IN 47405-7105, USA
ABSTRACTHigh-dispersion spectra in the Li 6708 ˚A region have been obtained and analyzed in the old, metal-deficient cluster, NGC 2243. From Hydra spectra for 29 astrometric and radial-velocity members, wederive rotational velocities, as well as [Fe/H], [Ca/H], [Si/H], and [Ni/H] based on 17, 1, 1, and 3lines, respectively. Using ROBOSPECT, an automatic equivalent width measurement program, wederive [Fe/H] = − . ± − . ± − . ± − . ± B − V ) = 0.055, appropriate isochrones imply( m − M ) = 13.2 ± ± (cid:12) /dex for the high mass edge ofthe Li-dip. The A(Li) distribution among giants reflects the degree of Li variation among the turnoffstars above the Li-dip, itself a function of stellar mass and metallicity and strongly anticorrelated witha v rot distribution that dramatically narrows with age. Potential implications of these patterns for theinterpretation of Li among dwarf and giant field populations, especially selection biases tied to age andmetallicity, are discussed. INTRODUCTIONWith improved precision and rapidly expandingdatabases due to the application of larger telescopes insurvey mode (see, e.g.
Casali et al. (2019); Gao etal. (2019)), the abundance of atmospheric Li (A(Li) )has become an increasingly valuable signature of bothstellar structure and galactic chemical evolution. Sincelower mass main sequence and red giant stars represent acrucial interface between both research areas, an exten-sive spectroscopic program has been underway to surveymembers of a key set of open clusters from the tip of thegiant branch to as far down the main sequence as thetechnology allows. New spectroscopy and/or reanaly-ses of published data have been discussed for NGC 752, [email protected]@[email protected] A(Li)=log N Li - log N H + 12.00 NGC 3680, and IC 4651 (Anthony-Twarog et al. 2009),NGC 6253 (Anthony-Twarog et al. 2010; Cummingset al. 2012), NGC 2506 (Anthony-Twarog, Deliyan-nis, & Twarog 2016; Anthony-Twarog et al. 2018),Hyades and Praesepe (Cummings et al. 2017), NGC6819 (Anthony-Twarog, Deliyannis, & Twarog 2014;Lee-Brown et al. 2015; Deliyannis et al. 2019), and,most recently, NGC 188 (Sun et al. 2020). The regu-larly repeated rationale for investigation of these clustersand their role within the bigger picture of Li evolutioncan be found in the respective papers and will not be de-tailed here. An overview of the goals and some currentinsight into the benefits of a more comprehensive ap-proach to open cluster analyses can be found in Twaroget al. (2020).The focus of this investigation is the metal-deficient,older open cluster, NGC 2243. Just how metal-deficientthe cluster is will be a point we return to in Section3. There is, however, longstanding agreement that, ir-respective of the technique used to derive the metallic-ity and/or the exact zero point of the metallicity scale,when ranked by metallicity among the well-studied open a r X i v : . [ a s t r o - ph . GA ] F e b Anthony-Twarog, Deliyannis, Twarog clusters within 5 kpc of the Sun, NGC 2243 sits consis-tently at or near the bottom of the list (Twarog et al.1997; Friel et al. 2002; Netopil et al. 2016). Thiscombination of distance, metallicity, and an age com-petitive with that of M67 (see Section 4) makes NGC2243 an important testbed for the effects of metallicityon the rate and degree of Li evolution among lower massstars while on the main sequence and, to an even greaterextent, in subgiant and giant branch phases. Isolatingthe metallicity-dependent effects of atmospheric evolu-tion also allows exposure of the underlying patterns ofGalactic Li evolution at both the metal-poor and metal-rich ends of the scale (Randich et al. 2020). Moreover,for now, the main sequence stars provide one of the clos-est links to their analogs within the halo population anda constraint on the apparent contradiction between stel-lar and cosmological predictions for the primordial A(Li)(Spite & Spite 1982a,b).In an attempt to expand the spectrosopic sample overthe full range of luminosity from the tip of the red giantbranch to the main sequence below the Li-dip (Boes-gaard & Tripicco 1986), our own Hydra data, detailedin Section 2, have been supplemented when possible byhigh resolution spectra obtained at the VLT, and by re-considered A(Li) from published sources when no spec-tra were accessible. The layout of the discussion is asfollows: Section 2 details the new spectra, their pro-cessing, reduction, and velocity derivation, as well asthe additional spectra from other sources, and a com-pilation of precision broad-band photometry for color-magnitude diagram (CMD) analysis; Section 3 lays outthe cluster metallicity determination from the Hydraspectra and comparison with previous results; Section4 combines the metallicity, reddening, and BV CMDto define the cluster age and distance, and indirectlythe masses of the stars populating the turnoff and giantbranch, through comparison to appropriate isochrones;Section 5 describes the Li abundance estimation for newmembers within the Hydra and VLT samples and com-bines these with previously published determinations todelineate the evolutionary trend within the cluster andplace the cluster within the context of Li changes withboth age and metallicity. Section 6 summarizes our re-sults for NGC 2243 while illuminating the impact of thevarying boundaries of the Li-dip due to metallicity ef-fects on our understanding of Galactic Li evolution andselection bias. OBSERVATIONAL DATA: NEW, OLD, ANDREVISITED2.1.
Spectroscopy: WIYN Hydra Data - Acquisitionand Reduction
NGC 2243 is a compact cluster and, while rather farsouth for northern hemisphere observations, access tothe Hydra multi-object spectrograph on the WIYN 3.5-meter telescope motivated spectroscopic observationsin this critical cluster.Our candidate list of cluster members was con-structed long before Gaia
DR2 (Gaia Collaboration etal. 2018) and was based primarily upon analyses ofprecision extended Str¨omgren photometry (Anthony-Twarog, Atwell, & Twarog 2005) (ATAT) using thetraditional intermediate and narrow-band indices of thesystem. Under the assumption that cluster membersshould exhibit similar chemical composition and red-dening, we additionally compiled a new parameter H ,defined as hk − . b − y ), and compared this photo-metric construct to then current membership informa-tion, mainly radial velocities. H values for the mostlikely member candidates centered on − .
40; any starthat deviated significantly from this target figure waseliminated as a spectroscopic target. Table 1 lists basicinformation about our Hydra sample, including coor-dinates, WEBDA identification numbers, and V and B − V photometry compiled as described below. All ofthe stars above the dividing horizontal line in Table 1are judged members by Cantat-Gaudin et al. (2018)based on Gaia
DR2 astrometry. Additional data abouteach star’s rotational and radial velocities are presented,with further descriptions to follow.Two Hydra fiber configurations were constructed, onewith nine bright stars and another with an eventual to-tal of 33 fainter targets designed for longer, multipleexposures. The brighter configuration was observed forthree 30-minute exposures on 21 Jan., 2015, where thedate indicates the UT date. Stars in the fainter config-uration were observed with exposures ranging from 53to 90 minutes, for 180 minutes total on 21 Jan., 2015,268 minutes on 22 Jan., 2015 and 246 minutes on 12Jan., 2016, amounting to 11.6 hours of total exposure.Between the 2015 and 2016 runs, one fiber fell out of useso that one star, observed only in 2015, was replaced byanother star observed only in 2016; the other 31 starswere observed in both years. Notes to Table 1 indicatewhich stars were observed in only one epoch.Our spectrograph setup is designed to cover a wave-length range ∼
400 ˚A centered on 6650 ˚A, with dispersionof 0.2 ˚A per pixel. Examination of Thorium-Argon lampspectra indicates a line resolution comprising 2.5 pixels, The WIYN Observatory was a joint facility of the University ofWisconsin-Madison, Indiana University, Yale University, and theNational Optical Astronomy Observatory. https://webda.physics.muni.cz i Evolution in NGC 2243 Table 1.
Stellar Characteristics, Hydra SampleWEBDA ID α (2000) δ (2000) V B − V S/N V rad σ V rad v rot σ vrot Noteper pix. km-s − km-s − km-s − km-s −
519 97.307140 -31.350243 15.724 0.524 157 55.54 1.89 16.54 1.03611 97.317848 -31.279542 15.384 0.861 185 55.67 1.64 11.92 0.69 a
716 97.327288 -31.225097 15.731 0.504 154 44.64 2.18 19.56 1.32873 97.342210 -31.281069 15.712 0.557 164 55.20 1.35 8.86 0.431044 97.357060 -31.273251 15.702 0.493 135 53.16 2.13 13.56 0.931133 97.362952 -31.307905 16.074 0.497 60 39.72 3.60 31.96 2.82 a a a a,b a,b a,c a,b b · · · a,d a,b a,b a,b a,b
259 97.272604 -31.284006 14.714 0.957 175 58.33 1.20 9.70 0.41507 97.305208 -31.280300 12.112 1.048 130 80.21 0.88 9.01 0.28 b
552 97.310604 -31.234522 14.820 0.508 249 46.93 1.21 12.58 0.55718 97.327242 -31.188708 15.862 0.519 149 74.54 3.00 22.14 2.172451 97.444504 -31.384656 13.985 0.594 248 -13.58 1.94 13.93 1.072528 97.450467 -31.324308 15.736 0.539 121 67.83 2.44 24.58 1.672704 97.467308 -31.281325 15.070 0.967 203 3.74 1.84 16.44 0.982719 97.468896 -31.359297 13.881 0.531 282 3.90 1.62 17.18 0.923063 97.510746 -31.266553 15.284 0.560 147 45.06 1.41 15.04 0.783726 97.304417 -31.309617 11.576 1.166 36 117.27 1.10 10.39 0.39 b a Li analysis from additional VLT spectra b Radial velocity obtained from only 2015 spectra. c Radial velocity obtained from only 2016 spectra. d Radial velocity for W2135 was 39.7 km-s − in 2015, 74.4 km-s − in 2016. Anthony-Twarog, Deliyannis, Twarog yielding a spectral resolution over 13,000. The data weresubsequently processed using standard reduction rou-tines in IRAF , including, in order of application, biassubtraction, division by the averaged flat field, disper-sion correction through interpolation of the comparisonlamp spectra, throughput correction for individual fibersusing daytime sky exposures in the same configuration,and continuum normalization. After flat field divisionand before the dispersion correction, the long-exposureprogram images were cleaned of cosmic rays using “L.A. Cosmic” (van Dokkum 2001). Real-time sky sub-traction was accomplished by using the dozens of fibersnot assigned to stars and exposed to the sky during eachintegration.Cumulative spectra for each configuration were con-structed by additive combination by night and by year;the differences between 2015 and 2016 values were alsoexamined for signs of potential radial-velocity variationbefore construction of cumulative combination spectra.Other than WEBDA 2135 (W2135), discussed in detailin Anthony-Twarog, Deliyannis, & Twarog (2020)(Pa-per I), no other stars show significant variation betweenthe two epochs. The signal-to-noise ratio per pixel (S/N)has been estimated by direct inspection of the compos-ite spectra within IRAF’s SPLOT utility, using meanvalues and r.m.s. scatter from a relatively line-free re-gion between 6680 and 6694 ˚A. S/N estimates of Table1 characterize the summed composite spectra, with amedian value of 138 achieved for the sample.2.2. Spectroscopy: VLT Data
Largely because early analyses indicated a fascinat-ing range of anomalies for member giant W2135 (PaperI), we sought to expand our sample of spectroscopicallyobserved stars by searching for archival spectra withinthe field of NGC 2243 and with spectral coverage in theregion of the Li line. We gratefully acknowledge use ofthe ESO Science Archive Facility from which we re-trieved fully processed spectra obtained for the public Gaia
ESO Survey at the Very Large Telescope (VLT)with the Fibre Large Array Multi Element Spectograph(FLAMES) fiber-feed assembly to both the high reso-lution UVES spectrograph and the GIRAFFE spectro-graph, with pixel-resolutions of 16.9 m˚A and 50 m˚A re- IRAF is distributed by the National Optical Astronomy Obser-vatory, which is operated by the Association of Universities forResearch in Astronomy, Inc., under cooperative agreement withthe National Science Foundation. http://archive.eso.org/scienceportal spectively. As our intention was to use these spectrafor Li abundance analysis only, we performed no fur-ther processing other than to import them to a formatacceptable to standard IRAF routines. The higher res-olution of the VLT spectra, particularly those from theUVES instrument, provides a considerable advantage forLi analysis although S/N above 100 is still desirable evenfor higher resolution spectra. Our initial archive searchwas limited to spectra described as having S/N above70; where possible, multiple exposures were co-added toproduce spectra with higher S/N.As a result, additional spectra for some of the starslisted in Table 1 were found and examined to take ad-vantage of the higher resolution of the VLT spectra.Notes to Table 1 indicate the Hydra sample stars forwhich VLT spectra were also examined. VLT spectrafor eleven probable member stars not included in theHydra sample were also identified. Table 2 presents thebasic information for these stars, comparable to thatpresented for the Hydra stars in Table 1, although ra-dial and rotational velocity values have not been esti-mated for the VLT spectra; in the place of these data,atmospheric parameters used in later analysis steps, areincluded in this table. S/N values are reproduced fromthe processed spectra’s headers or report quadratic sumsfor co-added spectra. Direct measurements of the S/Nbetween 6680 and 6694 ˚A largely confirm these listedvalues as adequate representation of the S/N per pixel.2.3. WIYN Hydra Data: Radial and RotationalVelocities
We utilized the fxcor
Fourier-transform, cross-correlation utility in IRAF to assess velocites for eachstar in each year’s composite spectra. In fxcor , starsare compared to a stellar template of similar T eff for thewavelength range of our spectra redward of H α . Out-put of the fxcor utility characterizes the cross correlationfunction (CCF), from which estimates of each star’s ra-dial velocity are easily inferred.From the procedure for radial-velocity estimation, ro-tational velocities can also be estimated from the crosscorrelation function full-width half maximum (CCFFWHM) using a procedure developed by Steinhauer(2003). This procedure exploits the relationship be-tween the CCF FWHM, line widths and v rot , using a setof numerically “spun up” standard spectra with compa-rable spectral types to constrain the relationship. Ouranalyses of both red giant and turnoff spectra demon-strate that f xcor cross-correlation profiles have signifi-cantly reduced accuracy when attempting to reproducerotational velocities above 30 km-s − . Our spectral reso- i Evolution in NGC 2243 Table 2.
Stellar Characteristics, VLT SampleWEBDA ID α (2000) δ (2000) V B − V S/N T eff log g ξ
239 97.269875 -31.357306 13.644 0.891 94 4990 2.5 1.50365 97.288917 -31.175694 14.009 0.975 151 4824 2.5 1.50910 97.345833 -31.291639 13.717 0.916 196 4939 2.5 1.501261 97.372208 -31.359833 15.177 0.873 130 5027 3.1 1.391679 97.395958 -31.322167 15.113 0.874 99 5025 3.1 1.391686 97.396333 -31.338639 15.675 0.496 130 6324 3.8 1.811923 97.408167 -31.276917 15.838 0.449 114 6538 3.9 1.972195 97.424500 -31.263194 15.321 0.766 133 5267 3.2 1.412536 97.451542 -31.377111 15.234 0.892 135 4987 3.1 1.382613 97.458250 -31.278306 15.146 0.828 94 5124 3.1 1.412648 97.462536 -31.245210 14.449 0.942 88 4888 2.75 1.50 lution implies that rotational velocities below 10 km-s − are not meaningful.Radial and rotational velocities for each of our pro-gram stars are included in Table 1, divided into likelymembers and nonmembers based on Gaia astrometry.Listed errors reflect the precision of the template fit by fxcor . As noted earlier, the lines of Table 1 are separatedby astrometric membership categories applied after sam-ple selection and data acquisition. The top 32 starsare considered members in the compilation by Cantat-Gaudin et al. (2018) based on
Gaia
DR2 astrometry,while the bottom ten stars, initially selected by photo-metric criteria to be potential members, are now desig-nated as unlikely astrometric members. In nearly everycase, our radial-velocity information for the ten likelynonmembers is consistent with the astrometric classifi-cation. The exception is star W259 for which DR2 as-trometry indicates a parallax ∼ / ∼ / . ± . − , where theerror statistic is the standard deviation among the 29stars.Star W2135 has been discussed in detail in Paper I.In summary, in addition to a measurable (and thereforedistinctive for the cluster’s age and the star’s evolution-ary state) Li line, this giant had been flagged as a pho-tometric variable by ATAT and as V14 by Kaluzny etal. (2006a). Limited information regarding photometricamplitude or periodicity emerged from either photomet-ric study, beyond a period estimated in days rather than hours. From Hydra spectra, we were only able to notea significant change in radial velocity between our twoobservational epochs, from 39 . − in 2015 to 74 . − one year later, for an average fortuitously nearthe cluster mean. Our velocity analyses indicated abnor-mally large v rot sin i for a giant star, at least 40 km-s − ,although we note again that analysis of the f xcor cross-correlation profiles probably can’t accurately reproducerotational velocities above ∼
30 km-s − .Reiterating a note added in proof to Paper I, theAll-Sky Automated Survey for SuperNovae (ASAS-SN)variable stars data base (Shappee et al. 2014) in-cludes star W2135. The cataloged star, ASASSN-VJ062941.11-311906.9/V0412 CMa, has a variability classof ROT, implying that the variability is likely due torotationally-induced brightness modulation, with a pe-riod of 6.755421 days (Jayasinghe et al. 2019). Thus,the estimated true v rot , i.e. with the sin i effect re-moved, using stellar parameter estimates derived fromthe CMD location of the star becomes 86 km-s − (PaperI).The spectra for W1133 are essentially featureless otherthan H α , indicating either a higher temperature thansuggested by the star’s color, rapid rotation, or both.This star is also a variable candidate from Kaluzny etal. (2006a)(V13), with a tentative classification as a γ Doradus variable and a period of ∼ .
77 days. We wereunable to obtain f xcor results for one of the two epochsand consider its one radial-velocity determination tenta-tive. Star W716 had a consistent radial velocity indica-tive of non-membership for both epochs, in spite of aclassification as a highly probably member based on as- https://asas-sn.osu.edu/variables Anthony-Twarog, Deliyannis, Twarog trometry. If a cluster member, it must be a single-linedspectroscopic binary (SB1) and we were unlucky in ourphase sampling, a result confirmed below.2.4.
Comparison With Other Radial-velocityDeterminations
A few radial-velocity standards were observed onsome, but not all, of the 2015 and 2016 nights on whichNGC 2243 was observed. Our radial velocities derivedfrom f xcor analysis yield V rad that are, on average,0 . ± . − too large in comparison to the veloci-ties quoted for our adopted standards on the SIMBADdatabase . Although the absolute scale of our radialvelocities was considered less critical than the internalor relative values for purposes of confirming commonmembership or detecting SB candidates, there are sev-eral radial-velocity surveys of NGC 2243 with which wecan compare our mean results, even though the directoverlap in stellar samples is significant for only two sur-veys. Comparisons of the radial velocities of overlappingstars to other past surveys have been constructed andare summarized in Table 3.Fran¸cois et al. (2013)(FR13) derive a cluster meanradial velocity of 61.9 km-s − , some 6.1 km-s − higherthan our Hydra-based result of +55 . − from the29 member stars. Only one star is common to bothstudies, the likely nonmember W552; our derived ra-dial velocity is 5.2 km-s − smaller than that quoted byFR13. The most recent, and largest, analysis of motionsin the cluster by Jackson et al. (2020)(G3D) indicates acluster velocity of 59 . ± .
58 km-s − , again larger thanour average value by 4.2 km-s − , where the quoted errorreflects only the internal dispersion within the cluster.We have referred to the G3D analysis to re-evaluatethe membership credentials of stars described in Tables1 and 2 (Hydra and VLT samples), as well as samplesreported on by FR13 and Hill & Pasquini (2000)(HP).Referring once again to stars excluded from our Hydraradial-velocity analysis, radial-velocity discrepancies forstars W2135, W1133 and W716 are confirmed. In Table1, V rad for W716 is noted as 44.6 km-s − , quite a bitlower than the internal average for the cluster of 55.8km-s − . Results from G3D indicate a radial velocityof 68.3 km-s − , but likely membership, suggesting evenmore strongly an SB1 nature for this star. The situationis much the same for W1133, known to be a variable andpotential binary; the G3D V rad is even lower than the39.7 km-s − listed in Table 1. Finally, W2135 exhibitsa G3D radial velocity of 80.4 km-s − , similar to our operated at CDS, Strasbourg, France (Wenger et al. 2000),http://simbad.u-strasbg.fr/simbad Figure 1.
CMD of stars with Li spectroscopy. Filled blackcircles and crosses are the members and nonmembers, re-spectively, from Tables 1 and 2. Filled blue circles are themembers from FR13, and the filled red circles are membersfrom HP. W2135 is noted with an asterisk symbol. larger 2016 measurement and further confirmation of itsSB character.2.5.
Photometry and Reddening
Reliable photometry plays a dual role in the discus-sion of NGC 2243. When adjusted for metallicity andreddening, it (a) is the basis for deriving the T eff usedin spectroscopic abundance estimation and, (b) permitsage and distance determination through comparison toappropriate isochrones, leading to mass estimates for in-dividual stars at a variety of CMD positions. A compre-hensive discussion of the broad-band BV photometryavailable at the time can be found in ATAT. For NGC2243, only two significant CCD BV surveys of the clus-ter were then available (Bonifazi et al. 1990; Bergbusch,VandenBerg, & Infante 1991), focusing on the centralregion of the cluster due to smaller format CCDs. For-tunately, comparison of the two data sets showed excel-lent agreement, indicating that both were well tied tothe standard BV system at the time as defined by Lan-dolt (1983) and Graham (1982). The merger of thesetwo photometric sets is described in detail in ATAT andwill not be repeated. Somewhat surprisingly, with theexception of the Gaia
DR2 dataset, no wide area broad-band studies, BV or otherwise, have been added to theliterature since then.For our purposes the photometric compilation hasbeen broken down into two distinct regions, the red gi-ants (( B − V ) ≥ .
70) and the turnoff region (( B − V ) < .
70) through the unevolved main sequence stars to V i Evolution in NGC 2243 Table 3.
Radial-Velocity Comparisons: Table 1 -LIT Survey ∆( V rad ) σ Numberkm-s − km-s − of stars Gaia
DR2 -1.7 1.3 6
Gaia
3D -4.2 1.8 24FR13 -5.2 · · · · · · · · ·
Note —DR2: Gaia Collaboration et al. (2018);G3D: Jackson et al. (2020); FR13: Fran¸coiset al. (2013); JFP: Jacobson, Friel, & Pila-chowski (2011); Gratton (1982); Mermilliod,Mayor & Udry (2008); FJ93: Friel & Janes(1993); FR02: Friel et al. (2002); Kaluzny:Kaluzny, Krzemi´nksi, & Mazur (1996); Minitti(1995); Cameron & Reid (1987). ∼ uvbyCa H β CCD photometryof ATAT. ATAT used the multicolor indices to iden-tify highly probable cluster members and to convert thehigh precision ( V , b − y ) for likely members to the tradi-tional ( V , B − V ) system defined by the earlier, smallerCCD samples (Bergbusch, VandenBerg, & Infante 1991;Bonifazi et al. 1990). As detailed in Paper I, the Gaia
DR2 ( G , B p − R p ) photometry for the red giants wasreadily transferred to the ( V , B − V ) system using 25member giants (Cantat-Gaudin et al. 2018) in com-mon, with the scatter in residuals between the two sys-tems measured at ± ± V and ( B − V ), respectively. This allowed a merger of thetwo photometric datasets and reliable expansion of the BV system for giants to Gaia
DR2 members outside thefield studied by ATAT. As detailed in Paper I, this leadsto 39 red giant members in the color range of interest.For the turnoff and main sequence stars, only themerged ( V , B − V ) data of Bergbusch, VandenBerg,& Infante (1991); Bonifazi et al. (1990) were adoptedfor stars within the survey areas. For a small subsam-ple of stars far enough beyond the cluster core to lieoutside the coverage of both BV surveys, simple ( G , B p − R p ) transformations to ( V , B − V ) with linear terms in ( B p − R p ) were defined using only astromet-ric members of the cluster (Cantat-Gaudin et al. 2018)within the same magnitude range as the stars of interestat the turnoff. Unlike the multiple merger process forbroad-band and intermediate-band photometry appliedto the giants, the transformed Gaia data were adoptedonly for the stars not found within the BV CCD surveys.A comprehensive discussion of the cluster redden-ing as of 2005 is given in ATAT. An early photomet-ric study by Kaluzny, Krzemi´nksi, & Mazur (1996)used
V I photometry to obtain a reddening estimateE( V − I ) = 0 .
10, implying E( B − V ) = 0 .
077 usingE( V − I ) = 1.35*E( B − V ). This estimate depends inpart on analysis of a nearby field RR Lyr which permitssetting 0.08 as an upper limit for E( B − V ). We note,however, that Kaluzny, Pych, & Rucinski (2006b), intheir analysis of the eclipsing binary (EB), NV CMa,adopt the ATAT reddening value and an [Fe/H] similarto the value obtained therein to derive their EB mass,cluster distance, and age estimates. From the extendedStr¨omgren photometry of 100 stars at the turnoff, ATATderived E( B − V ) = 0.055 ± E ( B − V ) = 0.074, the maximum value along the line Anthony-Twarog, Deliyannis, Twarog of sight in the direction of the cluster. Using the onlinetool , more recent corrections to the earlier reddeningmaps (Schlafly & Finkbeiner 2011) have reduced thisvalue to a maximum reddening of 0 . +0 . − . in the di-rection of NGC 2243. Unfortunately, NGC 2243 sits atthe edge of the map generated by Green et al. (2019)and attempts to derive the trend along the line of sightfail to converge.Finally, in Figure 1 we illustrate the location of thestars under discussion within the CMD of NGC 2243.Filled black circles and crosses are the members andnonmembers, respectively, from the combined samplesof Tables 1 and 2. An asterisk notes the location ofW2135. Filled blue circles are the members from FR13,and the filled red circles are members from HP. Of the10 identified nonmembers, only two near the top of theturnoff fall within the CMD trend defined by the clustermembers. Four nonmembers populate the blue stragglerregion well above the turnoff and the other four occupythe red giant zone, again well off the track defined bythe members.A few stars considered members by FR13 and HP arenot included in this plot nor in further analysis, due toambiguous or missing astrometric information. Theseinclude W3628 (referred to by 1035 by Fran¸cois et al.(2013), using the second of two numbering schemes in-troduced by Kaluzny, Krzemi´nksi, & Mazur (1996)),and W2512 (1106). An additional star, W1910, fromthe HP sample has also been excluded from further con-sideration.FR13 provide no additional information about thestars they identify as probable nonmembers based onradial velocities, but note that one star, Kaluzny 1410(W2363), has an essentially featureless spectrum, somuch so that no V rad estimate is possible. This de-scription bore a striking resemblance to the appearanceof the Hydra spectrum for W1133; we were promptedto verify that like W1133, W2363 has been identified asvariable star V3 by Kaluzny et al. (2006a), an eclipsingbinary NW CMa of W UMa class with a period of 0.356days. Although confirmed as an astrometric member byCantat-Gaudin et al. (2018), the star is not includedin the G3D sample, presumably because of a lack of areliable V rad measurement. ABUNDANCE ANALYSIS: HYDRA SPECTRA3.1.
Spectroscopic Processing With ROBOSPECT
As described in Lee-Brown et al. (2015), we have em-ployed the automated equivalent width (EW) measuring https://irsa.ipac.caltech.edu/applications/DUST/ software ROBOSPECT (Waters & Hollek 2013). De-tails (Lee-Brown et al. 2015) are provided of our testsof the software and refinement of a linelist to exploit theiron and other heavy element lines included in the ∼ gf values for any line.This approach ensures that our abundances are relativeto the solar value and solar spectrum rather than anyadopted solar abundance value. From all ROBOSPECToutput, we discarded negative EW (ROBOSPECT’sdesignation of emission lines) due to non-convergent fit-ting solutions, artifacts of cosmic ray removal, largenoise spikes, or non-existent lines in the measured spec-trum. We also excluded measured EW below 5 m˚A orabove 150 m˚A.3.2. T eff , Surface Gravity, and Microturbulent Velocity From measured EW, abundances are derived in thecontext of a stellar atmosphere model constructed withappropriate values of T eff , log g and microturbulent ve-locity, ξ . The T eff determination was approached ina manner to ensure consistency with previous spectro-scopic studies by this group. For dwarfs, we continue touse the calibration defined in Deliyannis, Steinhauer, &Jeffries (2002); for giants, the T eff -color-[Fe/H] calibra-tion of Ram´ırez & Mel´endez (2005) was used. In bothregimes, reddening values of E( B − V ) = 0 .
055 were ap-plied to B − V colors, with a value of [Fe/H] = − . V , B − V values to a Victoria-Regina isochrone(VandenBerg et al. 2006)(VR), constructed with theappropriate age, metallicity, and corrections for distanceand reddening, as discussed in Section 4. Our approachto microturbulent velocity estimation follows the oneused in Lee-Brown et al. (2015). For stars with sur-face gravity below 3.0, ξ was estimated using a surface-gravity-based algorithm, ξ = 2 . − . g . For lessevolved stars, we employ the T eff , log g formulation de-veloped by Bruntt et al. (2012), based on comparison ofspectroscopic and asteroseismic parameters for Kepler stars. i Evolution in NGC 2243 T eff , log g , and microturbulent velocity values. Each star’s mea-sured equivalent widths and model serve as input to the abfind routine of MOOG (Sneden 1973) to produce in-dividual [A/H] estimates for each successfully measuredline in each star. With high confidence in the member-ship credentials for our spectroscopic sample, it makessense to follow the statistical approach outlined in Lee-Brown et al. (2015), including the consistent use of me-dians to estimate cluster values, as well as constructedMedian Absolute Deviations (MAD) statistics to esti-mate the range for each estimated quantity.Figure 2 illustrates information presented in Table 4,including the [A/H] values obtained from each of the 22lines in our linelist; error bars indicate the MAD statis-tic for abundance values derived from metal lines in asmany as 29 stars (or as few as 14). Clearly some linesproduce noisier results than others, but we stress thatuse of median estimators lessens the impact of theselines. The dashed horizontal line indicates an overall[Fe/H] value of − .
55, based on the following considera-tions. Perhaps the most robust statistical estimation ofthe cluster’s [Fe/H] is derived from the median of all 354line measures, with a resulting median value of [Fe/H]= − .
55 and a MAD of 0.13 dex for all 354 separateestimators of [Fe/H]. An approximation of a standarddeviation for our sample is given by 1.48*MAD, or 0.19.Consideration of the [Fe/H] values for each of the 17 Felines leads to a median value of [Fe/H] = − .
53 amongthe 17 estimations, MAD = 0 .
10 dex, estimated stan-dard deviation 0.15. Taking the 29 separate [Fe/H] esti-mations for the sample stars, a median value of [Fe/H]= − .
56 is obtained, MAD = 0 .
12 dex, σ = 0 .
18. As aprecaution, we recomputed this median using only starswith at least 8 successfully measured Fe lines, producingthe same result.From a consistency standpoint, this result is encourag-ing. ATAT derived a weighted average of [Fe/H] = − . ± m and hk for 100 stars at the NGC2243 turnoff, slightly lower than the typical result above.For NGC 2506, analyzed in the same manner as NGC2243, Anthony-Twarog et al. (2018) derived a spectro-scopic abundance of [Fe/H] = − . ± − . ± Figure 2.
Median abundances and median absolute devia-tion (MAD) statistics for each analyzed line in Hydra spec-tra. line measurements. From 24 stars, the median [Ca/H]= − . ± .
19, where the quoted error indicates theMAD statistic, from 29 stars, [Ni/H] = − . ± . − . ± .
11. Rela-tive to the adopted cluster [Fe/H] = − .
55, these valuesrepresent [A/Fe] values of 0 . , − .
06 and 0.10 for Ca,Ni and Si respectively, consistent within the uncertain-ties with little to no differential elemental enhancementin the cluster.Table 5 summarizes the input atmospheric parametersand spectroscopic results for each likely member starwith a Hydra spectrum, with the exceptions of W2135,W1133 and W716 for reasons summarized above. Fig-ure 3 illustrates the run of [Fe/H] values as a function of T eff for which significant sensitivity might be a concern,and as a function of S/N for our sample. No signifi-cant trends are discernible here, as was also the case for[Fe/H] with respect to log g or ξ .3.3. Hydra Spectroscopic Uncertainties andComparisons With Other Spectroscopic Work
Probing the effects of incremented/decremented at-mospheric parameters was particularly rewarding in theanalysis of NGC 6819 Hydra spectroscopy by Lee-Brownet al. (2015) due to the large number of stars and thethoroughly populated CMD distribution, with an ex-tensive error propagation map presented in Figure 5 ofthat paper. We probed a subset of our smaller sam-ple to verify that similar trends are found in NGC 2243.Four stars in different parts of the CMD were reanalyzedwith incremented/decremented atmospheric parametersto probe the sensitivity of [Fe/H] estimates to incor-0
Anthony-Twarog, Deliyannis, Twarog rect parameter choices, resulting in similar effects. Al-though the results are not necessarily symmetric for in-crements/decrements in relevant quantities, we will con-form to the convention presented in Lee-Brown et al.(2015) and summarize the consequences of adopted T eff values too high by 100 K, surface gravities too smallby 0.25, and microturbulent velocities too high by 0.25km-s − .For cooler stars, assignment of a T eff too high by 100K results in a derived [Fe/H] too high by 0 .
08 dex, witha marginally smaller effect (+0.07) for stars near theturnoff. It should be noted that a ∆ T eff this large wouldimply a B − V color or reddening error of 0.06 mag forcooler giants or 0.02 mag for stars nearer the turnoff.The abundance effect of log g values assumed to betoo small by 0.25 is only discernible for cooler giants,amounting to ∆[Fe/H] = − .
02, ( i.e. , lower log g leadsto lower abundance), again similar to results from Lee-Brown et al. (2015). Using microturbulent velocitiestoo small by 0.25 km-s − will produce a higher [Fe/H]although the effects are more T eff -dependent, smallestfor turnoff stars (+0.02) to ∆[Fe/H]= +0 .
07 dex for thecooler giants.As noted earlier, NGC 2243’s place as one of themost metal-poor open clusters is fairly clear despite ahistoric range of [Fe/H] estimates that is distressinglylarge. This history is summarized well by recent stud-ies, including Jacobson, Friel, & Pilachowski (2011) andKovalev et al. (2019). It appears that differences be-tween one study and another cannot always be simplyexplained as due to different T eff scales although tem-perature is likely to have the dominant effect on [Fe/H]determination. One of the lower estimates, [Fe/H] = − .
63, was derived by Houdashelt, Frogel & Cohen(1992), who employed
JHK photometry to derive T eff .From four stars in common between our analysis setand theirs, we estimate that our T eff values are hotterby 109 ±
69 K, which alone would account for our 0.08dex higher estimate of [Fe/H].With respect to the parametric precepts of Jacobson,Friel, & Pilachowski (2011), we adopt T eff only 35 Kwarmer and log g values 0.3 dex lower. For giant stars,we would therefore expect these differences to generateapproximately offsetting increments to the metallicity,yet the cluster metallicity from Jacobson, Friel, & Pila-chowski (2011) using Hydra spectra of higher resolutionbut lower S/N is [Fe/H] = − .
42, higher than our esti-mate by more than 0.10 dex.Magrini et al. (2017) provide detailed T eff , log g and ξ estimates for 13 cool stars in common with thepresent study, with a resulting cluster average [Fe/H] of − . ∼ BV photometry, we would assign T eff
60 K cooler, log g values 0.1 dex smaller and ξ values 0.13 km/sec larger,which would work in the same sense to lower the Magriniet al. (2017) [Fe/H] estimate by no more than 0.07 dex,placing it in the [Fe/H] = − . − . BV photometry exists for stars studied byFR13 to estimate atmospheric parameters for their sam-ple stars in a manner consistent with precepts followedin this study. As noted by FR13, their temperature scaleis relatively hot. Differences between their adopted T eff and values assigned by color-temperature relations from BV photometry range from 150 K for stars below theclump to 250 K or higher for stars near the turnoff. De-spite a prediction that lowering their temperature scalewould result in lower [Fe/H] and [Li/H] values by ≥ . − .
54, virtually identical tothe median value derived from EW measures in the cur-rent study.Accounting for differences between spectroscopic anal-yses may require looking beyond the most traceable ef-fects ( e.g. differing temperature scales) to causes moredifficult to probe. These may include atmospheric mod-els, linelists and adopted log gf values in addition tooperational differences, e.g. , continuum level estimationand line measurement techniques. Alternatively, it maybe useful to consider methods that train and utilize neu-ral networks to analyze spectra in a holistic manner.We previously developed and used such a neural net-work classification approach called ANNA (Lee-Brownet al. 2015; Lee-Brown 2017, 2018) in similar contexts,especially NGC 2506 and NGC 6819 (Anthony-Twaroget al. 2018; Deliyannis et al. 2019). Data from NGC2506 successfully analyzed using ANNA were obtainedin the course of the same runs as our Hydra spectra inNGC 2243, so the same training parameters developedcould be applied to NGC 2243 Hydra spectra. This ap-plication is not obviously straightforward because theANNA training set was developed using Kurucz atmo-spheric models more metal-rich than [Fe/H] = − . − . ± .
03, keeping in mind thatthis value is a likely upper limit to the true cluster metal-licity.Kovalev et al. (2019) have also developed a neural-network approach to parameter estimation, using bothLTE and non-LTE atmospheric models. As with ANNA,“the Payne” code first trains on a large set of synthe-sized spectra then performs fits. With a training setbuilt upon atmospheric models that include consider-ably lower [Fe/H] values than ANNA, “the Payne” is i Evolution in NGC 2243 Figure 3. [Fe/H] as a function of T eff and S/N. well situated to provide parameter estimates for globularclusters as well as open clusters, as described in Kovalevet al. (2019) where NGC 2243 and a more metal-richopen cluster, NGC 3532, are discussed. 82 likely mem-bers of NGC 2243 were selected using radial-velocity and Gaia proper-motion criteria. An intra-cluster spread of0.07 dex in [Fe/H] is estimated for NGC 2243. Fromthe limited specific data provided for NGC 2243 stars,it appears that their derived T eff for a turnoff star isapproximately 150 K hotter than our BV photometry-based values. The LTE result from Kovalev et al.(2019), probably the most appropriate comparison toour LTE-based results, is [Fe/H] = − . ± .
11, wherethe quoted error refers to a mean systematic error; theNLTE estimate is 0.05 dex higher.Possible α -element enhancement in NGC 2243 hasbeen a topic of interest due to its location in the outerdisk of the Milky Way. Gratton & Contarini (1994) an-alyzed two giants to arrive at [Fe/H] of = − . ± . α -elements, concluding thata scaled solar abundance pattern was neither excludednor strongly favored, with typical enhancements [ α /Fe]of 0.07. More recently, Jacobson, Friel, & Pilachowski(2011) find elemental enhancements [X/Fe] ≤ .
15, con-sistent with scaled solar abundances. In particular, theycite strong evidence of solar [O/Fe] for one of the giantsin NGC 2243.Similarly, Kovalev et al. (2019) cite only modestevidence for α -element enhancement, +0.20 ± ± Table 4.
Characteristics and Results for In-dividual Metal LinesElem. λ [A/H] Number MAD˚A of StarsFe 6597.56 -0.39 25 0.10Fe 6609.11 -0.74 23 0.32Fe 6609.68 -0.76 20 0.17Fe 6627.54 -0.61 22 0.09Fe 6646.93 -0.52 14 0.06Fe 6653.91 -0.33 14 0.14Fe 6677.99 -0.73 20 0.14Fe 6703.57 -0.47 24 0.06Fe 6710.32 -0.48 18 0.04Fe 6725.36 -0.47 17 0.19Fe 6726.67 -0.53 25 0.06Fe 6733.15 -0.52 20 0.07Fe 6750.15 -0.61 28 0.10Fe 6806.86 -0.62 21 0.04Fe 6810.27 -0.68 24 0.08Fe 6820.37 -0.53 23 0.11Fe 6837.01 -0.44 16 0.19Ca 6717.68 -0.48 25 0.18Si 6721.85 -0.44 28 0.11Ni 6643.63 -0.68 28 0.14Ni 6767.77 -0.57 29 0.09Ni 6772.31 -0.56 27 0.044. THE CMD: CLUSTER AGE AND DISTANCETo place the stellar population within NGC 2243 andthe cluster itself in the appropriate context for evalu-ation and interpretation of the Li patterns describedbelow, we show in Figure 4 the CMD of well-definedcluster members relative to an appropriate set of VRisochrones. The filled circles include all member clustergiants with B − V > .
70; the open circles are core clus-ter members with
Gaia probabilities ≥ BV photometry while the asterisk again notes W2135. Thecompilation of the BV photometry is described in Sec-tion 2.4. The adopted isochrones have [Fe/H] = − . − .
53 to − .
56 (Section 3.2), the minor offset of the isochronescan account, in part, for the possibility of a minor α -element enhancement. For consistency with the stan-dard adopted in our previous cluster analyses using VRisochrones, we first zero the isochrones by requiring thata star of solar mass and metallicity at an age of 4.6 Gyr2 Anthony-Twarog, Deliyannis, Twarog
Table 5.
Individual Abundance Information for NGC 2243 Stars, Hydra SampleW ID T eff log g ξ [Fe/H] MAD Num. [Ca/H] [Ni/H] [Si/H]519 6199 3.75 1.72 -0.62 0.17 10 -0.65 -0.65 · · ·
611 5052 3.25 1.34 -0.56 0.14 17 -0.40 -0.55 -0.35873 6055 3.75 1.61 -0.62 0.19 11 -0.46 -0.55 -0.471044 6337 3.80 1.82 -0.26 0.37 10 -0.74 -0.62 -0.551230 5321 3.60 1.28 -0.67 0.24 11 -0.42 -0.74 -0.391263 5172 3.20 1.38 -0.60 0.07 15 -0.43 -0.63 -0.711266 6089 3.75 1.64 -0.37 0.10 10 -0.57 -0.41 -0.211271 4935 2.50 1.50 -0.55 0.12 15 -0.31 -0.60 -0.401294 5052 3.25 1.34 -0.55 0.26 17 -0.48 -0.66 -0.531313 4610 2.00 1.60 -0.50 0.05 15 · · · -0.53 -0.341421 5129 3.40 1.30 -0.66 0.10 14 -0.66 -0.72 -0.121436 6396 3.85 1.85 -0.57 0.12 8 -0.92 -0.77 -0.661467 4911 2.50 1.50 -0.52 0.11 15 -0.21 -0.49 -0.671696 4843 2.65 1.47 -0.53 0.04 16 -0.28 -0.56 -0.351738 4921 2.50 1.50 -0.66 0.15 15 -0.12 -0.51 -0.401847 4849 2.65 1.47 -0.47 0.08 16 -0.25 -0.61 -0.251871 6387 3.85 1.84 -0.89 0.15 5 -0.70 -0.83 -0.551995 5133 3.40 1.30 -0.58 0.07 14 -0.37 -0.61 -0.292003 6446 3.85 1.90 -0.82 0.23 4 -0.83 -0.78 -0.962098 5414 3.60 1.31 -0.74 0.12 13 -0.59 -0.76 -0.442394 6128 3.75 1.67 -0.51 0.06 6 -0.65 -0.56 -0.492410 4899 2.50 1.50 -0.43 0.12 16 -0.09 -0.47 -0.452434 5140 2.60 1.48 -0.54 0.09 17 -0.24 -0.55 -0.332676 6506 3.95 1.92 -0.33 0.28 8 -0.73 -0.71 -0.182696 6529 3.95 1.94 -0.68 NA 1 -0.77 -0.62 -0.502908 6510 3.95 1.92 -0.36 0.19 8 -0.66 -0.60 -0.873618 4696 2.35 1.53 -0.60 0.10 16 · · · -0.74 -0.563633 4127 1.20 1.76 -0.64 0.07 15 · · · -0.71 -0.803728 4793 2.25 1.55 -0.57 0.15 16 · · · -0.55 -0.31
Note —No equivalent width analysis for W2135 was carried out; correspondingatmospheric model parameters would be 4652 K, 2.35 and 1.53 km/sec. have M V = 4.84 and B − V = 0.65, leading to minoradjustments, ∆ V = 0.02 and ∆( B − V ) = +0.013 mag.The assumed cluster reddening is E( B − V ) = 0.055(Section 2.5). The isochrones have been shifted by anapparent distance modulus of 13.20 and have ages of 3.4,3.6, and 3.8 Gyr.The average parallaxes for the stars plotted in Figure 4yield an apparent modulus of ( m − M ) = 13.55 with anappropriate contribution from the adopted reddening.However, this does not account for the often discussedlikelihood of a zero-point offset of -0.05 ± Gaia parallaxes, especially for stars at greater distance(see, e.g.
Cantat-Gaudin et al. (2018); Riess et al.(2018); Stassun & Torres (2018); Zinn et al. (2019)). Applying a typical correction of +0.05 mas to the cluster
Gaia
DR2 parallax generates ( m − M ) = 13.1, clearlyin much better agreement with the isochrone fit.While the isochronal giant branches are an excellentmatch to the observational trends, the colors of theturnoff stars extend across the full color range of theisochrones, showing more structure than implied by themodels. The luminosities of the subgiant stars also ap-pear to lie above the isochrones, which might reflectan age on the lower side of our proposed range. As acompromise, we will adopt 3.6 ± i Evolution in NGC 2243 Figure 4.
CMD comparison to the isochrones of Vanden-Berg et al. (2006). Isochrones have ages of 3.4, 3.6, and3.8 Gyr. Filled circles include all cluster members with B − V > .
7. Open circles are highly probable membersfrom the cluster core with CCD B − V photometry while anasterisk notes the location of W2135.5. LITHIUM ABUNDANCES AND TRENDS5.1.
Spectrum Synthesis and Data Merger
We have utilized spectrum synthesis techniques to es-timate A(Li) for all high S/N spectra in our Hydra sam-ple, augmented by available archive spectra from theVLT; these stars have summarized in Tables 1 and 2 re-spectively. For this purpose, we used the synth mode ofMOOG (Sneden 1973), referencing the spectra to modelatmospheres appropriate to each star. As described inSection 3.2, T eff are derived in an internally consistentmanner from homogeneous B − V colors using two color-temperature relations adopted in previous investigationsby this group. Interactive estimation of A(Li) in thesynthesis process also provides an opportunity to verifythat the adopted T eff for each star is appropriate, basedon simultaneous agreement of the several temperature-dependent lines near 6710 ˚A.Table 6 summarizes the results of synthesis on all spec-tra, with consensus A(Li) estimates that summarize re-sults from Hydra, Giraffe and UVES spectra. S/N val-ues for the Hydra and VLT samples may be found inTables 1 and 2. While a standard for S/N values gener-ally ≥
100 is desirable, a few exceptions were made forVLT spectra to deliberately expand the very small over-lap with the previous Li abundance samples of FR13and HP. Adopted atmospheric model parameters arefound in Tables 5 and 2 for the two spectroscopic sam-ples. The consensus or overall A(Li) value was arrived at with the following general considerations: spectra withhigher S/N were favored, with higher resolution dom-inating only if S/N values are comparable. All otherfactors being comparable, results from our own Hydraspectra were favored. Of 14 Hydra stars for which VLTspectra were also available, 13 fell within the same cate-gory for both sources, i.e. classed as detections or upperlimits in both cases. Only W3618 was switched from anupper limit of A(Li) = 0.65 using Hydra spectra to adetection of A(Li) = 0.65 from VLT.Comparison of these Li abundances with prior work iscomplicated by differences of methodology and temper-ature scales, with the expectation that the latter wouldprovide the dominant influence. FR13 make primary useof
V I photometry from Kaluzny, Krzemi´nksi, & Mazur(1996), coupled with their fairly high reddening cor-rection, equivalent to almost E( B − V ) = 0 .
08. FR13note in their paper that a prior study by HP employsa lower reddening value with consequently lower tem-peratures. Figure 6 of FR13 illustrates both samples ina magnitude, T eff plane with a substantial temperatureoffset between the two samples amounting to 200 K andresulting in Li abundances lower by 0.15 dex in HP.Figure 1 illustrated our sample (filled black points),as well as those analyzed by FR13 (blue points) andHP (red points); the photometry is the BV photometrydescribed in Section 2.5. Clearly no intrinsic T eff offset,as defined by B − V , between the samples is evident.It remains important, therefore, to investigate the T eff differentials before discussing all available Li abundancesin a homogenous manner.Having found or synthesized colors for all of HP’s andFR13’s sample, we estimated T eff values for all starsbased on color-temperature relations discussed above.FR13’s use of a higher reddening value than used in thepresent study implies T eff for near turnoff stars that are ≥
250 K higher in their analysis, while cooler giantsare taken to be ≥
150 K hotter than our assessmentwould be. FR13 themselves estimate that temperatureshigher by 200 K imply A(Li) values higher by 0.15 dex.We sought to validate this estimate in two ways. First,there are a handful of stars with abundances based onour synthesis analysis that overlap with the FR13 sam-ple, none of them from our original Hydra sample butfrom the augmented archival VLT spectra. Three starsfrom Table 2 with detections overlap with FR13 detec-tions, W1261, W1686 and W2648. In the case of W1686,our analysis provides A(Li) that is 0.24 dex higher thanFR13; conversely, our analysis yields a Li abundancefor W2648 that is 0.36 lower. Our estimated A(Li) forW1261 is only 0.04 dex smaller than that of FR13. No4
Anthony-Twarog, Deliyannis, Twarog direct overlap of samples between the results presentedin Tables 1 and 2 with that of HP exists at all.Another way to estimate the effect of increments in T eff is to repeat the synthesis analysis on our avail-able spectra using models with higher temperatures.Test cases suggest that use of models with temperatureshigher by 200 K for cooler stars and 300 K for turnoffstars would produce Li abundances values higher by 0.2dex. It is unclear whether other differences in method-ology (continuum placement, use of equivalent widths, ahigher assumed overall metallicity) between our resultsand those from other surveys mask or negate some ofthis offset. Since the direct comparison of our A(Li)values for two stars in common with FR13 is indetermi-nate, we elect to present Li abundances from FR13 andHP with no adjustment in the following discussion.5.2. Li Trends: The Turnoff Region
Figure 5 illustrates the run of A(Li) values with re-spect to V magnitude and B − V color for all samples,with inverted triangles noting the location of upper limitvalues. A number of features are immediately appar-ent. First, in the CMD region where the FR13 data,the stars of Tables 1 and 2, and the A(Li) results of HPoverlap, i.e. at the top of the turnoff with V < .
0, theFR13 sample (blue points) does not separate toward ahigher A(Li) from the current sample (black points) orHP (red points), as one might expect given the highertemperature scale used by FR13. The largest detectedLi abundance in the FR13 analysis is A(Li) = 2.9, 0.2dex lower than the highest abundance derived for oursample. The probability of attaining a higher limitingA(Li) is undoubtedly enhanced by a sample of 12 starsfrom the current study relative to 4 from FR13, one ofthe reasons for boosting the population of stars observedwithin this critical evolutionary phase. Second, in con-trast, the generally fainter stars explored in great depthby FR13 map out the Li-dip extraordinarily well, par-ticularly at the faint end, but the addition of 5 membersnear V = 16.0 from the current study and HP combineto illustrate the strikingly steep high-mass edge of theLi-dip, hereinafter referred to as the wall, a feature thatemerges in all well-populated cluster samples older thanthe Hyades (Twarog et al. 2020) where stars enter-ing the subgiant branch have masses outside the rangeof the Li-dip. The addition of these stars shifts thestart of the Li-dip approximately 0.2 mag fainter thanallowed from the FR13 data alone. Whether the limitcould be pushed another 0.1 mag fainter still requires aneven more expanded sample and cannot be answered atpresent. Third, the cool edge of the Li-dip where mea-surable detections of Li, rather than upper limits, recur Figure 5.
A(Li) estimates for stars at the CMD turnoff.Symbol colors have the same meaning as in Figure 1. Filledcircles are Li detections while triangles denote upper limitsto A(Li). is starkly apparent near V = 16.95. While one couldargue that the spread in A(Li) at this magnitude levelruns from the upper limit above the Li-dip to the lowestmeasured value within the Li-dip, FR13 has emphasizedthat the stars below V = 16.95 with only detection lim-its are the likely product of spectra with low S/N andunlikely to be indicative of the true Li abundance amongthe fainter stars. Unfortunately, FR13 don’t supply S/Ndata for individual stars, so the role of this parameterin setting the scatter among the fainter dwarfs remainsambiguous. Fourth, among the stars above the Li-dip( V < . ∼ V at the turnoffinto Li versus mass, as shown in Figure 6. Stars fromthe subgiant branch beyond B − V = 0.7 and all red gi-ants have been excluded from the plot because the massrange among these stars is too small to allow differen-tiation on this scale. Taking the discrete nature of themass estimates into account and reiterating the greateruncertainty in the Li estimates at the lower mass end ofthe distribution, a plausible mass range encompassingthe Li-dip for NGC 2243 runs from 1.09 ± (cid:12) to i Evolution in NGC 2243 Table 6.
A(Li) Derived from SynthesisID Adopted Hydra Giraffe UVESA(Li) A(Li) EW A(Li) EW A(Li) EW a ≤ . · · · · · · ≤ . ≤ . · · · · · · · · · · · · · · · · · · · · · · · ·
611 1.1 1.1 31.5 1.0 26 · · · · · ·
873 2.45 2.45 44.2 · · · · · · · · · · · · ≤ . · · · · · · ≤ . ≤ . · · · · · · · · · · · · ≤ . ≤ . · · · · · · · · · · · · · · · · · · · · · · · · ≤ . ≤ . ≤ . · · · · · · · · · · · · ≤ . ≤ . ≤ . ≤ . ≤ . ≤ . ≤ . · · · · · · ≤ . ≤ · · · · · · ≤ − . ≤ . ≤ . · · · · · · · · · · · · · · · · · · · · · · · · ≤ . ≤ . ≤ . ≤ . ≤ . · · · · · · ≤ . · · · · · · · · · · · · · · · · · · ≤ . ≤ . · · · · · · ≤ . ≤ . ≤ . · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ≤ . · · · · · · ≤ . · · · · · · ≤ . ≤ . · · · · · · · · · · · · · · · · · · · · · · · · ≤ . ≤ . · · · · · · · · · · · · ≤ . · · · · · · ≤ . · · · · · · · · · · · · · · · · · · ≤ . ≤ . · · · · · · ≤ . ≤ . ≤ . · · · · · · · · · · · · ≤ . · · · · · · ≤ . · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ≤ .
65 33.3 · · · · · · ≤ − . ≤ − . · · · · · · ≤ − . ≤ . ≤ . ≤ . ≤ . a Measured equivalent widths for UVES spectra refer only to Li line at6707.78 ˚A. EWs listed for HYDRA and Giraffe spectra include contribu-tion from nearby Fe line. Anthony-Twarog, Deliyannis, Twarog
Figure 6.
Mass map of the stars populating the region of theLi-dip as shown in Figure 5. Symbols have the same meaningas in Figure 5. Masses are derived from a VR isochrone of3.6 Gyr age as delineated in Figure 4. Crosses illustrate thetypical error bars for A(Li) and mass at the extreme ends ofthe mass scale. ± (cid:12) , for a center at 1.15 ± (cid:12) . Asdiscussed in D19, defining the optimal position for inter-comparison of the Li-dips among multiple clusters canbe a challenge. Clusters such as NGC 3680 and NGC752 are inadequately populated and/or spectroscopicallysampled to provide definitive boundaries for the edgesof the Li-dip, though attempts have been made to ap-proximate the location of the center of the compositecluster sample to estimate the trend with central masswith varying [Fe/H]. Anthony-Twarog et al. (2009),assuming a symmetric Li-dip and comparing the com-posite cluster sample with the well-defined Hyades dis-tribution, derived a linear mass relation of M center /M (cid:12) = 0.4*[Fe/H] + 1.38. With [Fe/H] = − .
04 and a muchricher sample to outline the boundaries of the Li-dip,D19 found excellent agreement between this relation andNGC 6819. If we extrapolate to [Fe/H] = − .
55, thecenter of the Li-dip in NGC 2243 should sit at 1.16 M (cid:12) ,clearly consistent with the observations.It should be emphasized that attempts have beenmade to use alternate markers to test the impact ofmetallicity on the evolution of the Li-dip, most notablythe cool boundary of the feature (see, e.g.
Cummings etal. (2012), FR13). The clear benefit in this approachis the survival of the lower mass boundary to a muchgreater age, allowing extension of the possible trend with[Fe/H] to a wider array of clusters with a greater rangein metallicity, if precision spectroscopy can be obtainedat the fainter luminosities of these stars. The greater age of a cluster also has an important drawback: A(Li)for stars lower in mass than the Li-dip does not remainfixed with time. A(Li) for stars at lower mass rises out ofthe Li-dip before plateauing and declining with decreas-ing mass. The rise to, and the level of, the Li plateauchanges with age, making the exact definition of theboundary of the Li-dip increasingly indeterminate.While it does have applicability to a limited range ofcluster ages, with a range that depends upon [Fe/H], theA(Li) wall at the high mass boundary is usually read-ily identifiable if a sufficient sampling of the mass rangecan be accomplished. With the addition of NGC 2243,we can extend the mass boundary from the rich, well-sampled Hyades cluster at [Fe/H] = +0.15 (Anthony-Twarog et al. 2009; Cummings et al. 2017) and 1.494M (cid:12) to NGC 6819 at [Fe/H] = − .
04 and 1.43 M (cid:12) toNGC 2243 at [Fe/H] = − .
55 and 1.21 M (cid:12) . A sim-ple linear fit through these data generates M wall /M (cid:12) = 0.41*[Fe/H] + 1.44. We will return to this well-documented decline in mass with decreasing [Fe/H] inSection 6.5.3. Li Trends: Post-Main-Sequence Evolution
Figures 7 and 8 illustrate the trends of A(Li) for starsleaving the main sequence and evolving across the sub-giant branch, up the giant branch, and eventually tothe red giant clump. Figure 7 separates the data by V magnitude while Figure 8 details the trends with color.The turnoff region for each plot includes only stars atmasses that place them outside the Li-dip. Keeping inmind the typically high S/N for the spectra at the topof the turnoff, it must be concluded that the order-of-magnitude spread in A(Li) among stars with Li detec-tions at the top of the turnoff is real. Equally important,the range in A(Li) is not correlated with position in theCMD, either in color or luminosity.By the time these stars evolve to the base of the giantbranch, A(Li) is further reduced to a degree that placesall the stars at or below the canonical Li-rich boundaryof A(Li) = 1.5 (Kumar et al. 2020). Continuing up thegiant branch, there is a general decline in A(Li) for starswith detectable Li; above the level of the red giant clumpand within the clump itself, no star has detectable Li,with the upper limits set at A(Li) = 0.7 or lower.5.4. Li Patterns: The Role of NGC 2243
In the same way that NGC 2243 extends and strength-ens the pattern of declining mass boundaries for theLi-dip with decreasing [Fe/H], its combination of low[Fe/H] and significant age place it in a unique positionfor testing the evolutionary trends for stars more mas-sive than the Li-dip, while on the main sequence and i Evolution in NGC 2243 Figure 7.
A(Li) for stars from the top of the turnoff throughthe giant branch as a function of V . Symbols have the samemeaning as in Figure 5 Figure 8.
A(Li) for stars from the top of the turnoff throughthe giant branch as a function of B − V . Symbols have thesame meaning as in Figure 5, with an asterisk noting theposition of W2135. beyond. Using the Hyades-Praesepe ([Fe/H] = +0.15,age = 0.65 Gyr) as a zero-point and a mix of clustersranging in age from 1.5 Gyr (NGC 7789) to 2.25 Gyr(NGC 6819) and [Fe/H] as low as − . − . − .
3) should approximate the Li appearance of NGC2243 at a younger age (1.85 Gyr vs. 3.6 Gyr), keeping inmind that the differential “Li age” is smaller due to thehigher mass boundaries of the Li-dip for a more metal-rich cluster. Note also that the T eff scale of the turnoffstars in NGC 2506 is hotter than that of NGC 2243 dueto the younger isochronal age of the cluster.Two striking features emerge from a comparison of thetwo distributions. First, among the turnoff stars in NGC2506, 85% have A(Li) within 0.6 dex of the maximumcluster value at the turnoff; the remaining 15% outsidethis range include both detections and upper limits. Bycontrast, keeping in mind the sparser sample from theturnoff of NGC 2243, there is an almost uniform dis-tribution of stars extending from the upper bound inA(Li) to 1.1 dex lower. For the older cluster, more than2/3 of the turnoff stars lie below the 0.6 dex boundaryin A(Li). Second, the turnoff distinction translates di-rectly to the giant branch distribution. As discussed inAnthony-Twarog et al. (2018), despite its moderatelyyoung age, the subgiant branch in NGC 2506 is suffi-ciently populated to allow delineation of the depletionof Li as stars evolve from a typical turnoff value nearA(Li) ∼ ∼ ∼ ∼ Anthony-Twarog, Deliyannis, Twarog
Figure 9.
Contrast of the A(Li) distribution with T eff forstars more massive than the Li-dip in NGC 2506 (red) andNGC 2243 (black). Filled symbols are detections, open tri-angles and three-point stars are upper limits with the lat-ter tagging clump stars. An asterisk notes the location ofW2135. quence A(Li) distribution are depleted to a comparableextent, placing them as only upper limits near the giantbase, accounting for the handful of non-detections near5300 K. For NGC 2243, clump stars aside, the ratio ofnon-detections to detections is 3/2 among the red gi-ants, consistent within the uncertainties with the wideand almost uniform A(Li) distribution range among theturnoff stars. We conclude that the pattern observedamong the giants merely reflects a range in extant Liin stars on the point of leaving the main sequence, con-volved with one or more depletion mechanisms ( e.g. , di-lution and/or rotationally induced mixing) as the starsevolve along the subgiant branch and the first-ascentgiant branch.To get some clarity on the potential source of the dif-ferences between NGC 2506 and NGC 2243, we turn toan older and more metal-rich cluster, NGC 6819. Whileonly 0.4 Gyr older than NGC 2506, from the standpointof “Li age”, the cluster is significantly more advanceddue to its higher metallicity ([Fe/H] = − . (cid:12) higherthan in NGC 2506; its turnoff stars (above the Li-dip)are in a closer evolutionary state relative to the Li-dipto those in NGC 2243 rather than NGC 2506. As asimplistic means of quantifying this evolutionary phase,we note that the range in magnitude for stars above theLi-dip in the three clusters of interest is, binaries aside,0.6 mag, 1.0 mag, and > . Figure 10.
Same as Figure 9 for NGC 6819 (blue) andNGC 2243 (black). Clump stars in NGC 6819 are, like themajority of giants, all upper limits. No distinction is madebetween first-ascent and clump giants in NGC 6819. Aster-isks note the presence of Li-rich giants W2135 in NGC 2243and W7017 in NGC 6819. reveals the NGC 6819 analog to Figure 9, with the no-table change that no distinction is made between upperlimits in A(Li) for clump giants relative to first-ascentgiants since so few giants of any category in NGC 6819exhibit Li detections.Comparison of Figure 10 with Figure 9 demonstratesthat the NGC 6819 turnoff distribution bears a muchstronger similarity to NGC 2243 than NGC 2506. Keep-ing in mind the statistically richer sample of NGC 6819,the range in A(Li) among detections is 1.4 dex; includ-ing upper limits the range extends to 2.0 dex. Morerelevant, the distribution is even more heavily weightedtoward lower A(Li) than in NGC 2243; 30% of the sam-ple lies within 0.6 dex of the cluster limit but 63% aremore than 1 dex below the cluster limit and are domi-nated by upper limits rather than detections (D19). Be-cause the range in A(Li) among the stars hotter than5800 K is uncorrelated with either luminosity or T eff above the Li-dip, stars reaching the base of the giantbranch should exhibit a comparable spread in A(Li), as-suming the depletion mechanism on the subgiant branchapplies equally to all stars independent of their initialA(Li) upon entering the subgiant branch. The ubiquityof upper limits among the giants in NGC 6819 is consis-tent with this prediction given the turnoff distributionweighted toward depleted Li. Clearly some stars retainenough Li to fall within the detection limits, supposedlythe descendents of the stars leaving the main sequencewith Li relatively unchanged from the primordial clus-ter value. Only 5 stars cooler than 5600 K in NGC 6819have detectable Li. Of the two stars with A(Li) greaterthan 2.0, one lies on the subgiant branch between the i Evolution in NGC 2243 v rot as a function of the star’s position above the wall of theLi-dip, a more restricted version of their Figure 11, in-cluding only the three clusters under discussion, NGC2243, NGC 6819, and NGC 2506 and plotting only starsoutside the Li-dip. ∆ V is the distance in magnitudesa star lies positioned above the wall of the Li-dip, inthe sense ( V star − V wall ). The velocity range within theyoungest cluster (from a “Li age” standpoint) is just un-der 100 km-sec − ; for both NGC 6819 and NGC 2243,the range is virtually identical, with only one star ineach cluster above 25 km-sec − .When placed in the context of all the clusters detailedin D19, it is concluded that the primary control of theLi distribution with age for stars above the Li wall isthe rotational spindown of stars from an initial rangeclose to or greater than 100 km-sec − (NGC 7789 andNGC 2506) to less than 25 km-sec − (NGC 6819 andNGC 2243). The spindown mechanism, while still on themain sequence, is either correlated with and/or triggersthe variable decline in Li among stars with a significantrange in primordial v rot , generating a smaller range in v rot with increasing age, but an increasing range down-ward in A(Li) from the primordial cluster value.It should be emphasized that the A(Li) distributionfor stars above the wall prior to rotational spindownon the main sequence need not be a single-valued peakset at the initial cluster value. Diffusion of Li, withthe degree of depletion defined by the level of rotationalmixing, could produce a range of A(Li) among turnoffstars prior to rotational spindown, as seen, for exam-ple, among the turnoff stars in NGC 2506 in Figure 9.The spindown would then initially raise the atmosphericLi content as Li from stable layers below the surface ismixed to the surface, supposedly returning the star toits primordial value before enhanced rotational mixingdrives the retrieved Li content to even greater depth anddestruction. This process could be indicated by the Litrend among the subgiants in Figure 8 between B − V =0.45 and 0.6, but the sample remains too small and thedata sources too heterogeneous to reach any definitiveconclusion. Figure 11.
Distribution in v rot for stars above Li-dip forNGC 2506, NGC 6819, and NGC 2243. From the standpointof Li evolution, NGC 6819 and NGC 2243 are of comparableage and much older than NGC 2506, in contrast with thelisted isochronal ages. ∆ V = 0 is defined by the wall of theLi-dip. We have shown that both age and metallicity are im-portant parameters for Li depletion. Two clusters ofroughly similar age (NGC 2506 and NGC 6819) but dif-ferent metallicity show substantially different subgiantand giant Li patterns, with NGC 2506 being the moreorderly of the two. But it’s also true that among thetwo metal-poor clusters (NGC 2506 and NGC 2243), theolder one (NGC 2243) is less orderly, and is more similarto NGC 6819. The older, metal-poor cluster bears moreresemblance to the younger, metal-rich cluster than itdoes to the younger, metal-poor cluster. One straight-forward implication of this trend is that, assuming thatthe degree of post-main-sequence Li depletion is inde-pendent of the initial Li abundance upon entering thesubgiant branch, the range in A(Li) among giants ata given phase of evolution will depend strongly uponthe mass and metallicity of the stars leaving the mainsequence, i.e. classification of a giant as Li-rich or Li-poor requires boundaries that vary with mass, metallic-ity, and age (Twarog et al. 2020). NGC 2243 SUMMARY AND GALACTICIMPLICATIONSHigh dispersion spectra have been obtained of 42 starswithin the field of the metal-deficient open cluster, NGC2243. From both astrometry and radial velocities, 32cluster members have been identified and their spectraanalyzed. From 29 likely single stars, the overall clus-ter [Fe/H] is well-defined at − . ± Anthony-Twarog, Deliyannis, Twarog ror describes the median absolute deviation about thesample median value, leading to internal precision forthe cluster [Fe/H] below 0.03 dex. A much more re-stricted sample of lines for α -elements produces effec-tively scaled-solar abundances, though the uncertain-ties are larger. With the metallicity and previously de-termined reddening in hand, the cluster age and dis-tance are obtained by comparison to an appropriate setof isochrones, generating an age and apparent distancemodulus of 3.6 ± m − M ) = 13.2 ± − .
38 fromGaia ESO pipeline analysis; references within Randichet al. (2020). Randich et al. (2020) also correctly noteupper turnoff stars “may have already undergone somepost-MS dilution” to their surface Li abundances mak-ing a detection of the peak A(Li) value stochasticallyvulnerable to sample size and selection.Moreover, the spread in A(Li) among stars on theverge of entering the subgiant branch is at least 1.1 dexand could be greater than 1.3 dex. This spread is cru-cial because it is not dependent upon specific positionwithin the turnoff region, i.e. stars occupying virtuallyidentical locations within the CMD differ in their A(Li)by 0.9 dex. This spread translates directly into a largespread among giants ascending the giant branch for the first time. Excluding the post-He-flash members of thered clump, all of which have only upper limits to A(Li),the majority of the red giants exhibit only upper lim-its below A(Li) = 1.5, the nominal value predicted forstars of solar Li upon arrival at the base of the giantbranch. (Note, the limits become more restrictive withincreasing luminosity due to the declining T eff amongbrighter giants and should not be interpreted as evidencefor declining A(Li) with evolution up the giant branch;only the Li detections can illuminate this pattern.) Thistrend is consistent with the pattern found in the moremetal-rich and numerically richer but younger cluster,NGC 6819. The implication is that, due to the shift inthe Li-dip to lower mass at lower [Fe/H], the evolution-ary phase of a cluster with respect to Li evolutionarypattern, the “Li age”, requires a more metal-deficientcluster to attain a greater age to approach the same “Liage” as a metal-rich cluster.Within the discussion of Li evolution among stars, therole of the main sequence/turnoff stars above the wallhas been twofold. First, by evaluating the sensitivity ofthe wall to changes in mass, metallicity, and age, someinsight could be gained into the physical process(es) trig-gering the depletion of Li among main sequence starsboth within the Li-dip and beyond. As discussed in de-tail in D19 and confirmed with the addition of NGC2243, a controlling factor in the evolutionary history ofLi for any star, whether above the wall, within, or be-low the Li-dip appears to be the spindown mechanismof the star. The key, but not necessarily the only, differ-ence among the stars at varying masses on the main se-quence is whether the decline in rotation occurs rapidly, e.g. during the pre-main-sequence phase, or more grad-ually (and/or nonlinearly) during main-sequence evolu-tion. The challenge in identifying such trends comesfrom the fact that the degree of Li depletion is likely tobe correlated with the degree of rotational decline, butwhat one measures is an instantaneous snapshot of thecurrent rotational speed with no a priori knowledge ofthe star’s initial rate upon arrival at the main sequence,though statistical distributions of the kind shown in Fig-ure 11 (and in Figure 11 of D19) can help constrain thelikely degree of change. If correct, the spindown patternamong the cluster samples and the long-standing de-cline in the mass location of the Li-dip with decreasing[Fe/H] points toward a metallicity dependence in the ro-tational evolution of the main sequence stars, through aninitial v rot distribution with mass which depends upon[Fe/H] and/or a spindown mechanism whose effective-ness changes with [Fe/H], a question whose resolution,while intriguing, is well beyond the scope of the currentsample. i Evolution in NGC 2243 e.g. Cum-mings et al. (2012); Bouvier et al. (2018); Anthony-Twarog et al. (2018) among many others), the starsabove the wall have represented a potential pristine in-dicator of the primordial cluster Li abundance, a clearreference marker against which changes to lower massstars could be calibrated, again assuming that the highermass stars were minimally affected by processes such asdiffusion. It is now apparent that as a cluster ages the Liabundance of the stars above the wall does evolve down-ward, with the degree of depletion growing with age dueto the declining range in v rot , usually well beyond therange defined by the statistical uncertainty in the deter-mination of individual A(Li). Not only does this weakenthe value of these stars in defining the original clusterabundance, among the stars above the wall it essentiallyrecreates the Li-dip, though at a slower pace and lesseffectively (D19). The impact on the giant branch iscrucial. The masses of the stars reaching the base of thegiant branch with severely reduced A(Li) are no longerdefined by the mass of the Li wall, but now extend sig-nificantly higher. For the numerically rich cluster NGC6819 at [Fe/H] = − .
04, Li-depletion is well underwayamong stars leaving the main sequence with minimalmasses of 1.57 M (cid:12) , even though the Li wall doesn’t oc-cur until one reaches 1.43 M (cid:12) . Clearly these boundarieswill shift downward will decreasing metallicity. Becausethe level of the Li plateau among lower mass stars be-yond the cool edge of the Li-dip on the main sequencewill decrease with time (Cummings et al. 2012), the ma-jority of the red giants older than ∼ i.e. on the warm side of the Li-dip, and the metallicity-dependent change in the mass boundaries of the Li-diphave subtle but relevant impacts on the interpretation ofthe Li distribution among field stars. In a recent analysisof the exquisite GALAH database for F and G dwarfs,Gao et al. (2020) identify and analyze over 100000stars between [Fe/H] = − T eff between5900 K and 7000 K. After sorting the stars into three Licategories, “warm” (above the wall), within the Li-dip,and “cool” (below the Li-dip), the trends with [Fe/H]for the “warm” and “cool” stars were compared. Asexpected, for stars with [Fe/H] below -1.0, only “cool”stars were present due to the evolution of “warm” starsoff the main sequence by the age typical of these metal-poor dwarfs. The approximately uniform A(Li) for thesestars is consistent with Li abundances for stars occupy-ing the Spite plateau (Spite & Spite 1982a,b). Between[Fe/H] = − . − .
5, a “warm” sample of 117 starsreappeared and scattered about a mean value of A(Li)= 2.69 ± − .
5, A(Li) among both the “warm” andthe “cool” stars exhibits a steady increase, implying acontribution from Galactic nucleosynthesis superposedupon the initial undiluted primordial Li abundance de-fined by the stars at [Fe/H] = − . − . − . − . − .
5, themean A(Li) would be 2.52 ± − . − . i.e. the defining lines for “warm” and “cool”stars. Fig. 12 illustrates the competing effects of ba-sic stellar evolution with the potential development ofthe boundaries of the Li-dip. Using the VR isochronesdefined with [ α /Fe] set at +0.4, the mass range of the2 Anthony-Twarog, Deliyannis, Twarog stars populating the CMD turnoff at an age of 12.5 Gyrwith decreasing [Fe/H] is shown as a pair of solid lines.The mass range of the turnoff is defined from the bluestpoint at the of the main sequence CMD to a position 0.10mag redder in B − V , the latter boundary selected toavoid stars normally classified as subgiants. The massesare set using isochrones with an age of 12.5 Gyr acrossthe board. Clearly, shifting the models to younger ageswould slide the relations to a systematically higher mass.By contrast, the dashed lines show the mass boundariesof the Li-dip, created by simple extrapolation of thelinear trends defined by multiple clusters from [Fe/H]= − . (cid:12) (Fran¸cois et al.2013) to 1.09 M (cid:12) (Figure 6).As expected, the simplistic extrapolation of the Li-dipmass range crosses the globular cluster mass trend be-tween [Fe/H] = − .
25 and − .
50, in contradiction withthe absence of the Li-dip within globular cluster CMDsand the existence of the Spite plateau. An obvious so-lution to the contradiction is that the mass/metallicityslope flattens below [Fe/H] = − .
5, preventing the Li-dip from ever crossing the evolutionary trend with lowermetallicities except among younger isochrones. Becausethe mass/metallicity relation is based upon clusters withscaled-solar abundances, it is possible that it is inappli-cable to stars/clusters with enhanced [ α /Fe], i.e. metal-licity should refer to [m/H] rather than [Fe/H]. If thetypical star below [Fe/H] = − . α /Fe] = +0.4, asassumed for the isochrones, the mass boundaries shouldbe boosted by as much as 0.14 to 0.16 M (cid:12) at the “cool’and “warm” boundaries, pushing the crossover of thetwo trends to [Fe/H] below − . − .
5, not including the possible role of[ α /Fe] = +0.4 within the metallicity. The red pointsdefine the same age marker, but include the enhanced α elements in the mass boundary determination, i.e. M/M (cid:12) = − . Figure 12.
The mass range of the Li-dip (dashed lines) com-pared to the mass range for the turnoff stars in a cluster ofage 12.5 Gyr (solid lines) as a function of [Fe/H]. Isochronesare assumed to be [ α /Fe] = 0.4. Age (Gyr) at which theLi wall evolves to the subgiant branch is given in blue formetallicity defined as [Fe/H], red for [m/H], α enhancementincluded. able stars. The more metal-rich the star, the youngerit has to be to be included in a “warm” star sam-ple. Second, mechanisms used to alter the slope of themass/metallicity relation by making it shallower or byshifting it to higher mass limits with metallicity, as donevia alteration of the definition of metallicity in Figure12, invariably move the age boundary for the “warm”stars to a younger age, creating an even more extremeselection bias within any “warm” star sample from thefield.The selection bias due to age limitations for popu-lating the “warm” side of the Li-dip, coupled with Lidepletion among “warm” stars prior to entering the sub-giant branch, has already been seen among stars at themetal-rich end of the scale. A number of studies haveclaimed that the limit to A(Li) among stars of super-solar metallicity declines as [Fe/H] increases (see, e.g.,Fu, Romano, Bragaglia et al. (2018); Guiglion, Chiap-pini, Romano et al. (2019); Bensby, Feltzing & Yee(2020); Stonkut˙e, Chorniy, Tautvaiˇsien˙e et al. (2020)).This apparent trend is readily understood if the Li-dipboundaries plotted in Figure 12 are extended to [Fe/H]= +0.5, using isochrones with scaled-solar abundances.From our mass/[Fe/H] relation, at [Fe/H] = − .
04 and+0.43 the mass boundaries for the wall are located at1.42 and 1.61 M (cid:12) , respectively. Using the same defini-tion as earlier, these mass boundaries enter the subgiantbranch at an age of 3.1 and 2.4 Gyr, respectively. The i Evolution in NGC 2243 ± ± ∼ >
0, by Randich et al. (2020).With one cluster at 1.4 Gyr and all others at 1.0 Gyror less, Randich et al. (2020) find no decrease in A(Li)with increasing [Fe/H]. In fact, the clusters with highermetallicity have the highest A(Li).ACKNOWLEDGMENTSNSF support for this project was provided to BJATand BAT through NSF grant AST-1211621, and to CPDthrough NSF grants AST-1211699 and AST-1909456.The assistance of KU undergraduate Steven Smith wasinstrumental in the early stages of this project; Dr. Don-ald Lee-Brown assisted with observations and develop-ment of the ANNA code.This research has made use of the WEBDA database,operated at the Department of Theoretical Physics andAstrophysics of the Masaryk University. This researchhas made use of the services of the ESO Science ArchiveFacility and is based on observations collected at theEuropean Southern Observatory under the public
Gaia
ESO Survey Program.
Facility:
WIYN: 3.5m
Software:
IRAF Tody (1986), ANNA Lee-Brown(2017, 2018), MOOG Sneden (1973), LACOSMIC vanDokkum (2001)REFERENCES
Anthony-Twarog, B. J., Atwell, J., & Twarog, B. A. 2005,AJ, 129, 872 (ATAT)Anthony-Twarog, B. J., Deliyannis, C. P., Rich, E., &Twarog, B. A. 2013, ApJ, 767, L19Anthony-Twarog, B. J., Deliyannis, C. P., & Twarog, B. A.2014, AJ, 148, 51Anthony-Twarog, B. J., Deliyannis, C. P., & Twarog, B. A.2016, AJ, 152, 192Anthony-Twarog, B. J., Deliyannis, C. P., & Twarog, B. A.2020, AJ, 160, 75 (Paper I)Anthony-Twarog, B. J., Deliyannis, C. P., Twarog, B. A.,Croxall, K. V., & Cummings, J. 2009, AJ, 138, 1171Anthony-Twarog, B. J., Deliyannis, C. P., Twarog, B. A.,Cummings, J. D., & Maderak, R. M. 2010, AJ, 139, 2034Anthony-Twarog, B. J., Lee-Brown, D. B., Deliyannis,C. P., & Twarog, B. A. 2018, AJ, 155, 138Bensby, T., Feltzing, S., & Yee, J. 2020, A&A, 634,A130Bergbusch, P. A., VandenBerg, D. A., & Infante, L. 1991,AJ, 101, 2102Boesgaard, A. M., & Tripicco, M. J. 1986, ApJL, 302, 49Bonifazi, F., Fusi Pecci, F., Romeo, G., & Tosi, M. 1990,MNRAS, 245, 15 Bouvier, J., Barrado, D., Moraux, E., et al. 2018, A&A,613, A63Bragaglia, A., & Tosi, M. 2006, AJ, 131, 1544Bruntt, H., Basu, S., Smalley, B., et al. 2012, MNRAS, 423,122Cameron, A. C. & Reid, N. 1987, MNRAS, 224, 821Cantat-Gaudin, T., Jordi, C., Vallenari, A., et al. 2018,A&A, 618, 93Carlberg, J. K., Smith, V. V., Cunha, K., et al. 2015, ApJ,802, 7Casali, G., Magrini, L., Tognelli, E., et al. 2019, A&A, 629,A62Cummings, J. D., Deliyannis, C. P., Anthony-Twarog, B. J.,Twarog, B. A., & Maderak, R. M. 2012, AJ, 144, 137Cummings, J. D., Deliyannis, C. P., Maderak, R. M., &Steinhauer, A. 2017, AJ, 153, 128Deepak, & Reddy, B. E. 2019, MNRAS, 484, 2000Deliyannis, C. P., Anthony-Twarog, B. J., Lee-Brown,D. B., Twarog, B. A. 2019, AJ, 158, 163 (D19)Deliyannis, C. P., Steinhauer, A., & Jeffries, R. D. 2002,ApJ, 577, L39 Anthony-Twarog, Deliyannis, Twarog
Fran¸cois, P., Pasquini, L., Biazzo, K., Bonifacio, P., &Palsa, R. 2013, A&A, 552, A136 (FR13)Friel, E. D., Janes, K. A., Tavarez, M., et al. 2002, AJ, 124,2693Friel, E. D. & Janes, K. A. 1993, A&A, 267,75Fu, X., Romano, D., Bragaglia, A. et al. 2018, A&A, 610,A38Gaia Collaboration, Brown. A. G. A., Vallenari, A., et al.2018a, A&A, 616, A1 (DR2)Gao, Q., Shi, J.-R., Yan, H.-L., et al. 2019, ApJS, 245, 33Gao, Q., Lind, L., Amarsi, A. M. et al 2020, MNRAS, 497,L30Graham, J. A. 1982, PASP, 94, 244Gratton, R. G. 1982, AJ, 257, 640Gratton, R. G. & Contarini, G. 1994, A&A, 283, 911Green, G. M., Schlafly, E. F., Zucker, C., Speagle, J. S., &Finkbeiner, D. P. 2019, ApJ, 887, 93Guiglion, G., Chiappini, C., Romano, D., et al. 2019, A&A,623, A99Handberg, R., Brogaard, K., Miglio, A., et al. 2017,MNRAS, 472, 979Harris, W. E. 1996, AJ, 112, 1487Hill, V., & Pasquini, L. 2000, I.A.U. Symp. 198,
The LightElements and Their Evolution , ed. L. Da Silva, J. R. deMedeiros,, & M. Spite, 293 (HP)Houdashelt, M. L., Frogel, J. A. & Cohen, J. .G. 1992, AJ,103, 163Jacobson, H. R., Friel, E. D., & Pilachowski, C. A. 2011,AJ, 141, 58Jackson, R. J., Jeffries, R. D., Wright, N. J. et al. 2020,MNRAS, 496, 4701Jayasinghe, T., Stanek, K. Z., Kochanek, C. S., et al. 2019,MNRAS, 486, 1907Kovalev, M., Bergemann, M., Ting, Y.-S. & Rix, H.-W.2019, A&A, 628, 54Kaluzny, J., Krzemi´nksi, W., & Mazur, B. 1996, A&A, 118,303Kaluzny, J., Krzemi´nksi, W., Thompson, I. B., &Stachowski, G. 2006a, AcA, 56, 51Kaluzny, J., Pych, W., & Rucinski, S. M. 2006b, AcA, 56,237Kumar, Y. B., Reddy, B. E., Campbell, S. W., et al. 2020,Nature Astronomy, doi:10.1038/s41550-020-1139-7Kurucz, R. L. 1995, in IAU Symp. 149,
The StellarPopulations of Galaxies , ed. B. Barbuy & A. Renzini(Dordrecht: Kluwer), 225Landolt, A. U. 1983, AJ, 88, 439 Lee-Brown D. B. 2017, Artificial Neural NetworkAbundances (ANNA), v0.1.0, Zenodo,doi:10.5281/zenodo.1043585Lee-Brown, D. 2018, Ph.D. Thesis, University of KansasLee-Brown, D., Anthony-Twarog, B. J., Deliyannis, C. P.,Rich, E., & Twarog, B. A. 2015, AJ, 149, 121Magrini, L., Randich, S., Kordopatis, G., et al. 2017, A&A,603, A2Mermilliod., J.-C., Mayor, M., Udry, S. 2008, A&A, 485,303Minitti, D. 1995, A&AS, 113, 299Netopil, M., Paunzen, E., Heiter, U., & Soubiran, C. 2016,A&A, 585, 150Pinsonneault, M. 1997, ARA&A, 35, 557Ram´ırez, I., & Mel´endez, J. 2005 ApJ, 626, 465Randich, S., Pasquini, L., Franciosini, L., et al. 2020,MNRAS, 640, L1Riess, A. G., Casertano, S., Yuan, W., et al. 2018, ApJ,861, 126Ryan, S. G., & Deliyannis, C. P. 1995, ApJ, 453, 819Schlafly, E. F., & Finkbeiner, D. P. 2011 ApJ, 737 103Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ,500, 525Shappee, B. J., Prieto, J. L., Grupe, D., et al. 2014, ApJ,788, 48Sneden, C. 1973, ApJ, 184, 839Spite, M., & Spite, F. 1982a, Nature, 297, 483Spite, F., & Spite, M. 1982b, A&A, 115, 357Spite, M., Spite, F. & Bonifacio, P. 2012, in
Li in theCosmos , Mem. Soc. Astron, Italiana, Suppl., 22, 9Stassun, K. G., & Torres, G. 2018, ApJ, 862, 61Steinhauer, A. 2003, PhD thesis, Indiana UniversitySteinhauer, A., & Deliyannis, C. P. 2004, ApJ, 614, L65Stonkut˙e, E., Chorniy,Y., Tautvaiˇsien˙e, G., et al. 2020, AJ,159,90Sun, Q., Deliyannis, C. P., Twarog, B. A., et al. 2020, AJ,submittedTody, D. 1986, Proc. SPIE, 627, 733Twarog, B. A., Anthony-Twarog, B. J., Deliyannis, C. P.,& Steinhauer, A. 2020, in