A High-Resolution Spectrum of the Highly Magnified Bulge G-Dwarf MOA-2006-BLG-099S
Jennifer A. Johnson, B. Scott Gaudi, Takahiro Sumi, Ian A. Bond, Andrew Gould
aa r X i v : . [ a s t r o - ph ] J a n A High-Resolution Spectrum of the Highly Magnified BulgeG-Dwarf MOA-2006-BLG-099S Jennifer A. Johnson , B. Scott Gaudi , Takahiro Sumi , Ian A. Bond and Andrew Gould ABSTRACT
We analyze a high-resolution spectrum of a microlensed G-dwarf in the Galac-tic bulge, acquired when the star was magnified by a factor of 110. We measurea spectroscopic temperature, derived from the wings of the Balmer lines, that isthe same as the photometric temperature, derived using the color determined bystandard microlensing techniques. We measure [Fe/H]=0 . ± .
18, which placesthis star at the upper end of the Bulge giant metallicity distribution. In partic-ular, this star is more metal-rich than any bulge M giant with high-resolutionabundances. We find that the abundance ratios of alpha and iron-peak elementsare similar to those of Bulge giants with the same metallicity. For the first time,we measure the abundances of K and Zn for a star in the Bulge. The [K/Mg]ratio is similar to the value measured in the halo and the disk, suggesting that Kproduction closely tracks α production. The [Cu/Fe] and [Zn/Fe] ratios supportthe theory that those elements are produced in Type II SNe, rather than Type IaSNe. We also measured the first C and N abundances in the Bulge that have notbeen affected by first dredge-up. The [C/Fe] and [N/Fe] ratios are close to solar,in agreement with the hypothesis that giants experience only canonical mixing. Subject headings: gravitational lensing – stars: abundances – Galaxy: abun-dances – Galaxy: bulge – Galaxy: evolution This paper includes data gathered with the 6.5 meter Magellan Telescopes located at Las CampanasObservatory, Chile. Department of Astronomy, Ohio State University, 140 W. 18th Ave., Columbus, OH 43210, USA;jaj,gaudi,[email protected] Solar-Terrestrial Environment Laboratory Nagoya University, Nagoya, Japan;[email protected] Institute for Information and Mathematical Sciences, Massey University, Auckland, NewZealand;[email protected]
1. Introduction
The Galactic Bulge underwent an intense burst of star formation early in the formationof the Galaxy, leading to a very different stellar population and chemical evolution historythan found in the Milky Way disk or halo (e.g., Ortolani et al. 1995; Zoccali et al. 2003;McWilliam & Rich 1994). In particular, massive stars may dominate the pollution at almostall metallicities, leading to unique abundance patterns (e.g., Lecureur et al. 2007). Similarevents are thought to mark the formation of other galactic spheroids, making the Bulgestellar population a template for interpreting extragalactic observations. As a result of itsunique formation history in the Galaxy, the Bulge has been the subject of intensive study.The detection of RR Lyrae stars (Baade 1946) first indicated that the Bulge containedold stars. With deeper photometry, the main sequence turnoff (MSTO) of the Bulge wasdetected. Terndrup (1988) found a mean age of 11-14 Gyr for stars in Baade’s Window, witha negligible fraction of stars with ages < Hubble Space Telescope confirmed the generally old nature of the Bulge(Ortolani et al. 1995; Holtzman et al. 1998), although Feltzing & Gilmore (2000) included areminder that a young metal-rich population has similar MSTO colors and luminosities toan older, more metal-poor population. Therefore, deriving ages reliably from photometryof the MSTO requires adequate knowledge of the Bulge metallicity distribution function(MDF), specifically for the MSTO stars. However, because of the faintness of those stars,the measurement of the Bulge MDF has historically relied on giants.Since the discovery of both M giants and RR Lyr stars, it has been known that theBulge giants span a wide range in metallicity. Low-dispersion spectra provided the firstquantitative measure of the MDF (Whitford & Rich 1983; Rich 1988). Sadler et al. (1996)measured indices from low-dispersion spectra of 268 K giants (both red clump stars and firstascent giants) to derive a mean metallicity ) h [Fe / H] i = − .
11, with a dispersion of 0.46dex. Recalibration by Fulbright et al. (2006) based on high-resolution spectra of 15 starsin common with the Sadler et al. (1996) sample reduced the mean metallicity to − . h [Fe / H] i = − .
21 from low-dispersion near-infrared spectrafor 72 M giants in the inner Bulge. The good agreement between Ram´ırez et al. (2000) resultand the recalibrated Sadler et al. (1996) result is somewhat surprising. At the bright endof the giant branch, only metal-rich first ascent giants become M giants. However, both Kgiants and M giants become red clump stars, and lower luminosity metal-rich giants are Kstars as well. Therefore, the inclusion of red clump stars and fainter giants in the Sadler et al.(1996) sample make the biases in their sample more similar to those of the Ram´ırez et al. We adopt the usual spectroscopic notation that [A/B] ≡ log (N A /N B ) ⋆ – log (N A /N B ) ⊙ K and (V-K) color magnitude diagram for a low-reddening window at( l, b )=(0.277, − K < − .
5) RGB stars with globularcluster fiducials of known metallicity. After correcting for small biases in their MDF causedby their magnitude cutoff, they find an MDF with a peak at [M/H]= − .
1, a sharp cutoffat [M/H]= − . < −
1. Adopting the metallicities derived fromthe giants for the MSTO dwarfs, they estimated that the Bulge is coeval with the halo andargued that the lack of stars above the prominent MSTO of the Bulge ruled out a significantlyyounger population.These measurements of the Bulge giant MDF can be improved by metallicity measure-ments from high-dispersion measurements of many stars in several fields throughout theBulge. Recently, Lecureur et al. (2008) and Zoccali et al. (2008) have obtained a total of ∼ ∼ α /Fe] ratios in giants, in particular, high[Mg/Fe] for [Fe/H] values up to solar, arguing for little Type Ia SN contribution of Fe com-pared to the thick or thin disks. Fulbright et al. (2007) confirmed the overall enhancementin [Mg/Fe] and strengthened the conclusion of McWilliam & Rich (1994) that the other α abundance ratios do not track [Mg/Fe] exactly. [O/Fe], [Si/Fe], [Ca/Fe] and [Ti/Fe] beginto decrease around [Fe/H]=0, while [Mg/Fe] does so at supersolar [Fe/H]. Fulbright et al.(2007) suggested that metallicity-dependent Type II SN yields could explain the differentbehaviors of the α elements. McWilliam et al. (2007) explained the low [O/Mg] ratios inmetal-rich Bulge giants through a different metallicity-dependent mechanism: Wolf-Rayetwinds leading to less effective O production in metal-rich massive stars. By [Fe/H] ∼ α /Fe] ratios have begun to decline, probably indicating the introduction of largeamounts of Fe from Type Ia SNe (e.g., Cunha & Smith 2006).Several studies have looked at the abundances of the light elements Na and Al in the 4 –Bulge, two elements whose production should depend on the metallicity of the massive starsthat exploded as Type II SNe. Cunha & Smith (2006) measured Na in 7 K and M gi-ants and found the predicted increase in [Na/Fe] and [Na/O] in the most metal-rich stars.Lecureur et al. (2007) found supersolar [Al/Fe] at all metallicities and, for [Fe/H] >
0, en-hanced [Na/Fe] compared to the ratios in disk stars. Interestingly, at higher metallicities,the scatter in [Al/Fe] and [Na/Fe] increased and became larger than could be explained byobservational errors. Interpreting the Na and Al abundances as the result of Type II SNproduction may be problematic. The surface abundances of Al and Na have been shownto be increased by large amounts of internal mixing in metal-poor globular cluster stars(e.g. Shetrone 1996), where the products of proton-capture reactions deep inside the starare mixed up to the surface, leading to enhancements in these two elements. However,Lecureur et al. (2007) argued that the C and N abundances in the giants they studied wereconsistent with only mild mixing and, therefore, that the high Na and Al had to be due tothe overall chemical evolution of the Bulge. Cunha & Smith (2006) also found evidence formild mixing in giants, affecting C and N, but not O, Na, or Al.Finally, there is little information on the neutron-capture elements in the Bulge. Theabsorption lines for these elements are concentrated in the blue part of the optical spectrum,where the crowding from Fe, CN, and other lines is severe. Near-IR spectra have essentiallyno lines of these elements. While there are a few lines of Ba in the red, these lines inmetal-rich giants are so saturated that reliable measurements are very difficult. As a result,only the neutron-capture element Eu has published results so far. McWilliam & Rich (1994)found [Eu/Fe] > , the Optical GravitationalLens Experiment (OGLE), the Microlensing Follow Up Network ( µ FUN) and the Prob-ing Lensing Anomalies Network (PLANET), provides an opportunity to study otherwiseunobservable Bulge dwarfs. During high-magnification microlensing events, it is possible toobtain high-resolution, high signal-to-noise ratio spectra of faint stars with a huge savingsin observing time: a factor A × . m where A is the magnification and ∆ m is the numberof magnitudes below sky of the unmagnified star. In Johnson et al. (2007) , we reported thedetailed abundances for a highly-magnified Bulge dwarf, OGLE-2006-BLG-265S, which froma 15 minute exposure at magnification A ∼ A = 110.Finally, we respond the challenge: “Ask not what microlensing can do for stellar spec-troscopy – ask what stellar spectroscopy can do for microlensing.” There is one importantway that the spectroscopic study of bulge dwarfs can benefit microlensing. Whenever a sourceapproaches or transits a “caustic” (line of infinite magnification) caused by the lens, one canmeasure ρ , the ratio of the angular source radius to angular Einstein radius ρ = θ ∗ /θ E , fromthe microlens lightcurve. Then θ ∗ is inferred from the dereddened color and magnitude ofthe source to yield θ E = θ ∗ /ρ , which in turn provides important constraints on the lensproperties. Because spectroscopy is not normally available for these microlensed sources,the dereddened color and magnitude are estimated by comparing the source position onan instrumental color-magnitude diagram with that of the clump and then assuming thatthe Bulge clump is similar to the local clump as measured by Hipparcos (Yoo et al. 2004a).This procedure undoubtedly suffers some statistical errors and could suffer systematic er-rors as well. For example, the Bulge clump may have a different color from the local one.High-resolution spectra of an ensemble of microlensed bulge sources will test this proce-dure for both statistical and systematic errors. For OGLE-2006-BLG-265S, the standardmicrolensing procedure yielded ( V − I ) = 0 . ± .
05, whereas Johnson et al. (2007) ob-tained ( V − I ) = 0 . ± .
04 from high-resolution spectroscopy. This difference hints at apossible discrepancy, but only by repeating this procedure on a number of dwarfs can thisbe confirmed. ∼ iabond/alert/alert.html ∼ ogle/ogle3/ews/ews.html ∼ microfun/ http://planet.iap.fr/
2. Observations
MOA-2006-BLG-99 was alerted as a probable microlensing event toward the GalacticBulge (J2000 RA = 17:54:10.99, Dec = − l = − . b = − .
78) by MOA on 22July 2006. On 23 July, MOA issued a further alert that this would be a high-magnificationevent, with
A > µ FUN, primarily with the aim of searching for planets (Mao & Paczy´nski1991; Griest & Safizadeh 1998; Udalski et al. 2005; Gould et al. 2006). Results of that searchwill be presented elsewhere. The event actually peaked on 23 July (HJD 2453940.349) at A max ∼ µ FUN emails describing this event. He theninterrupted his normal program to obtain a 20-minute exposure of this event at the begin-ning of the night, just after peak, when the magnification was A = 215. Unfortunately, theatmospheric transparency was poor, and the observation had to be interrupted after 1089seconds when the cloud cover became too thick. Conditions cleared several hours later, andhe obtained two 20-minute exposures centered at UT 02:09:36 and UT 02:30:34 24 July,when the magnification was A = 113 and A = 107. We base our results on these higherquality spectra.The observations of MOA-2006-BLG-99 were made using the Magellan Inamori KyoceraEchelle (MIKE) double spectrograph (Bernstein et al. 2003) mounted on the Clay telescopeon Las Campanas. with seeing of 0.7–1.0 arcsec. We used the 1.0 arcsec slit, which producesR ∼ ∼
3. Data Reduction
The data were reduced using the MIKE Python data reduction pipeline (D. Kelson,2007, private communication), with the exception of the bluest orders containing the CHand CN lines, which were reduced using the IRAF echelle package. The bias and overscanwere subtracted. The wavelength calibration was derived from Th-Ar data. Flatfields weretaken through a diffusor slide to create “milky flats” that made the orders sufficiently wideto get good flatfields along the edges of the orders of the data frames. Parts of orderswith overlapping wavelength coverage were coadded together before analysis. These overlapregions are larger for the bluer parts of the spectrum. Over the wavelength region where IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the As-sociation of Universities for Research in Astronomy, Inc., under cooperative agreement with the NationalScience Foundation. ∼
30. The CNand CH bandhead regions had lower S/N (S/N ∼
4. Abundance Analysis
We analyzed both the spectrum of MOA-2006-BLG-99S and the spectrum of the Sun(Kurucz et al. 1984) using the same set of lines, line parameters, and model atmospheregrids. We used TurboSpectrum (Alvarez & Plez 1998), a 1-dimensional LTE code, to deriveabundances. TurboSpectrum uses the recent treatment of damping from Barklem et al.(2000). We interpolated the model atmosphere grid of ATLAS9 models updated with newopacity distribution functions (Castelli & Kurucz 2003). The abundances of most elementswere determined from analysis of the equivalent widths (EWs). The EWs for both MOA-2006-BLG-99S and the Sun are presented in Table 1. We restricted the analysis to lineswith EW ≤ (C. Sneden, 2007, privatecommunication). We compared synthetic with observed spectra to determine abundances forlines that were blended, lines that had substantial hyperfine splitting and for C and N thatwere measured from CH and CN lines, respectively. We used the solar atlas of Moore et al.(1966) to check for blending with telluric features and eliminated the few lines that wereaffected.The linelists for CH and CN are from B. Plez (2006, private communication). The effectof hyperfine splitting (HFS) was included for Sc, Mn, Cu and Ba. The HFS constants weretaken from the sources listed in Johnson et al. (2006). HFS information was not availablefor the Na or Al lines, so neither this study nor the literature studies we use for comparisoncan correct for those effects. Therefore, the comparison between [Na/Fe] and [Al/Fe] val-ues for MOA-2006-BLG-99S and the literature values is robust, but the absolute values ofthese abundance ratios for all studies have a systematic error. Table 1 lists the transitionprobabilities (listed as log gf -values) and sources for all the atomic lines. We measured a T eff =5800K using the wings of the H- α and H- β lines (Figure 4.1).Our uncertainty in the temperature is based on the uncertainty in this fit, in particular http://kurucz.harvard.edu/grids.html http://verdi.as.utexas.edu/spectre.html V − K colors (Mashonkina et al. 1999; Barklem et al. 2002). ForMOA-2006-BLG-99S, we compared the temperature derived from the Balmer lines withthat derived from excitation equilibrium of the Fe I lines. The temperature derived fromthe Fe I lines was very uncertain because the abundances derived from individual lines werepoorly determined because of the S/N of the spectrum. The excitation equilibrium favoreda higher temperature (+200K) with an uncertainty of 300K.Given the uncertainties, thistemperature is in agreement with the Balmer line temperature. Next, we measured themicroturbulent velocity, ξ , by ensuring that the abundances derived from the Fe I linesdo not depend on their reduced EW. Our best fit value was ξ =1.5 km/s. Changing ξ by ± ξ . Thegravity was measured by ionization balance for Fe I and Fe II and for Ti I and Ti II . Asa sanity check, in Figure 2, we compare the atmosphere parameters for MOA-2006-BLG-99S, OGLE-2006-BLG-265S, and the Sun to the parameters from the Yonsei-Yale (Yi et al.2001; Demarque et al. 2004) isochrones. Finally, the metallicity of the model atmosphere,[m/H], was set equal to the [Fe/H] given by the Fe I lines. Because the log g of the modelatmosphere affects the [m/H] and the [m/H] affects the abundances (and therefore the log g measurement), we iterated to find a solution for which Fe I and Fe II were equal and the[m/H] of the model was equal to [Fe I /H].Using standard microlensing techniques (e.g., Yoo et al. 2004a), µ FUN determinedthat the dereddened color and magnitude of the source were ( V − I ) = 0 . ± . I = 18 . ± .
10. The error is due to possible differential reddening between the mi-crolensed source and the red clump, which is assumed to have the same ( V − I ) = +1 . M V = 4 . ⊙ .That is, it is a solar-type star. We combined the µ FUN color and the Ram´ırez & Mel´endez(2005) color-temperature relation (an update of the Alonso et al. (1996) relation) to derivean estimate of the temperature of 5806 ±
200 K. This photometric temperature agrees withthe temperature from the Balmer lines. For OGLE-2006-BLG-265S, the spectroscopic andphotometric temperatures were different, suggesting possible systematic errors in estimatingthe dereddened color and magnitude of the source star. As outlined in the Introduction,quantifying any systematic errors is important for obtaining the most accurate informationabout microlensing events, and additional data will be crucial for quantifying any systematicerror. 9 –Fig. 1.— Fits to the H- α line for three different temperatures: 5600K, 5800K and 6000K.The atmosphere with T eff =5800K is the best fit to the hydrogen line. 10 –Fig. 2.— The position of MOA-2006-BLG-99S (square), OGLE-2006-BLG-265S (circle) andthe Sun (triangle) in the H-R diagram. We also show isochrones from Yi et al. (2001). Thesolid lines show isochrones for [Fe/H]=0.385 and for ages 2, 3, 4, 5 and 7 Gyr. The dashedline shows a 5 Gyr isochrone for a solar metallicity. 11 – Our uncertainties are 200K for T eff and 0.3 km/s for ξ . The distribution of EWs andexcitation potential for the Fe I lines were sufficiently uncorrelated that ξ did not depend onT eff . Our measurements of T eff and ξ therefore did not depend on accurate assessments of theother model atmosphere parameters, namely log g and [m/H]. Our measurement of log g and[m/H], on the other hand, depended strongly on the other model atmosphere parameters.Because we set the gravity by ionization equilibrium, log g depends on T eff , ξ , and [m/H],as well as the abundance measured from the Fe I and Fe II lines. Therefore, the uncertaintyin log g depends on the uncertainties in those quantities. The uncertainty in [m/H], in turn,depends on the uncertainty in T eff , log g , and ξ , as well as the scatter in the abundancegiven by different Fe I lines for a particular model atmosphere. The standard error of themean for the gravity derived from the 35 Fe I lines for a single model atmosphere was 0.04dex, and for the 4 Fe II lines was 0.14. We adopt 0.04 dex as the uncertainty from EWand log gf errors for Fe I for inclusion in the [m/H] uncertainty. Adding the Fe I and Fe II uncertainties in quadrature give us 0.15 dex as our uncertainty in the difference between theFe I and the Fe II log ǫ values arising from the EW and oscillator strength uncertainties.The total uncertainty in log g is 0.66 dex and in [m/H] is 0.21 dex.We ran the Fe I and Fe II EWs through a series of models: ± ± ± g , and 0.13 dex in [m/H] (smaller because of the limits of the Kurucz grid), andcalculated the difference in Fe abundance with these different model atmospheres.Finally, we calculate uncertainties using modified equations (A5) and (A20) from McWilliam et al.(1995). We considered the covariance between T eff : log g , T eff : [m/H] , log g : [m/H] , ξ : log g , and ξ : [m/H]. The covariances were calculated by a Monte Carlo technique. Forexample, to calculate the covariance between T eff and log g , we first found ∂ log g/∂ T eff andnoted the remaining scatter that was caused by uncertainty in ξ , etc. Next, we randomlypicked 1000 T eff from a Gaussian distribution with a σ of 200K. We used the derivative tocalculate log g and then added an extra random ∆log g drawn from a Gaussian distributionwith a σ equal to the uncertainty from non-T eff causes. We calculated the covariance usingthese 1000 T eff -log g pairs. A similar calculation was done for the other covariances. log ǫ (A) ≡ log (N A /N H ) + 12.0
12 –
5. Results
In Table 2, we summarize the abundances measured for 17 elements in MOA-2006-BLG-99S. We include both log ǫ and its error, as well as [X/Fe] and its error. To give an idea of theuncertainty due to scatter from the lines, rather than from atmospheric parameters, we give σ , the rms of abundances derived from individual lines as well as the number of lines. Wealso give our measurements of the solar abundances, which we will use to calculate ratios.For reference, we include the Grevesse & Sauval (1998) solar abundances in the final column. We measure [Fe/H]= 0 . ± .
18 for MOA-2006-BLG-99S. In Johnson et al. (2007), wemeasured [Fe/H]= 0 . ± .
19 for the dwarf OGLE-2006-BLG-265S. The stars that are mi-crolensed are unbiased in metallicity. The criterion for spectroscopic follow-up is that theunmagnified source be faint enough to be a Bulge dwarf, regardless of color. Therefore,especially considering the large and variable reddening toward the Bulge, we are not biasedin our high-resolution follow-up toward high metallicity sources. The high metallicities ofthese two dwarfs is surprising given that work on giants has indicated an average metallicitynear solar. In Figure 3, we compare the metallicity distribution function (MDF) of thetwo dwarfs with several MDFs based on studies of giant stars. The MDF of Rich et al.(2007), which is based on high-resolution analysis of M giants and should be biased toward the highest metallicity objects, lacks any giants as metal-rich as these dwarfs. This is partic-ularly notable, since the work of Sadler et al. (1996) on K giants with low-dispersion spectrashows some extremely metal-rich stars. A complicating factor is the possible presence of ametallicity gradient in the Bulge. In Figure 4, we compare the metallicities of the dwarfswith MDFs derived by Zoccali et al. (2008) from high-resolution spectra for giants in threeBulge fields: 4 ◦ , 6 ◦ , and 12 ◦ away from the Galactic center. The inner field is more metal-rich than the outer. However, these two dwarfs are located 6.5 ◦ (MOA-2006-BLG-99S) and4.9 ◦ (OGLE-2006-BLG-265S) away from the Galactic Center, and therefore gradients cannotexplain their anomalously high metallicities.These results hint that the MDFs of the Bulge giants and dwarfs may be different.Whether this is true can be established by more observations of Bulge dwarfs and by res-olution of the discrepancies among the MDFs derived for giants, particularily between thelow-dispersion and high-dispersion studies. Because the microlensed dwarfs are found at arange of distances from the Galactic Center, comparison of the giant and dwarf MDFs alsodepends on measuring the metallicity gradient (and the size of deviations from that gradient)in the Bulge. Ideally, the MDF for giants in the same field as the microlensed dwarf would 13 –be measured. With our measurements of T eff , log g , and [Fe/H] from spectroscopy, we can comparethe position of MOA-2006-BLG-99S on the Hertzsprung-Russell diagram with theoreticalisochrones (Fig. 2). We use the Yonsei-Yale isochrones (Yi et al. 2001; Demarque et al.2004) with [Fe/H]=0.385, [ α /Fe]=0 for comparison. The best fit age for this star is ∼ ≥ eff on the lower edge of our range (5600K), the metallicity calculatedfrom the Fe I lines drops to [Fe/H]=0.16 dex. Using isochrones of this metallicity, the newT eff gives an age of ∼ As stars move up the giant branch, they pass through first dredge-up, which brings upmaterial that has been processed in the CN cycle. The C and N abundances measured ingiants no longer represent the original C and N endowments of the stars, although C+Nwill remain constant as long as only material processed in the CN cycle, and not the ONcycle, is mixed to the surface. Cunha & Smith (2006) and Cunha et al. (2007) measured Cand N in giants in the Bulge. They found that the giants lie to the N-rich side of the linedefined by the C/N ratio of the Sun (Figure 6). Cunha et al. (2007) concluded that a smallamount of mixing had occurred in the giants. This conclusion is only valid if the originalabundances in the giants lie close to the line. Otherwise, if the Bulge dwarfs have non-solarC/N ratios, the C/N ratios measured in Bulge giants could imply either no mixing (and aN-rich original composition) or substantial mixing (and a C-rich original composition). Theabundances of C and N for MOA-2006-BLG-99S are also plotted on Figure 8 and shows that 14 –Fig. 3.— The MDF for MOA-2006-BLG-99S and OGLE-2006-BLG-265S compared to theMDF from Sadler et al. (1996), which was measured on K and M giants and with the MDFsof Ram´ırez et al. (2000), Rich & Origlia (2005), and Rich et al. (2007) for M giants. TheMDF of the dwarfs is shifted to higher metallicities compared to the M giants, which issurprising since the most metal-rich stars should end their red giant phase as M giantsrather than K giants. 15 –Fig. 4.— The MDF for MOA-2006-BLG-99S and OGLE-2006-BLG-265S compared to theMDF derived by Zoccali et al. (2008) from high-dispersion spectra of giants in three fields:Baade’s Window at 4 ◦ as well as a 6 ◦ and 12 ◦ field. The fraction of the two-bulge-dwarfsample has been scaled down from 0.5 to fit clearly on the graph. The average metallicityof the bulge giants decreases as the distance from the Galactic center. The dwarfs are morethan 4 ◦ away from the center, making their high metallicities even more surprising. 16 –Fig. 5.— The position of MOA-2006-BLG-99S (square) in the H-R diagram (M I -T eff ). M I was calculated using the I magnitude and assuming a distance of 8.5 kpc. The T eff is thetemperature from the Balmer lines. We also show isochrones from Yi et al. (2001) The solidlines show isochrones for [Fe/H]=0.385 and for ages 2, 3, 4, 5 and 7 Gyr. The dashed lineshows a 5 Gyr isochrone for a solar metallicity. 17 –the assumption of Cunha et al. (2007) is justified and that the expected amount of mixinghas occurred in the giants. Li has been created since the Big Bang by stellar nucleosynthesis and by cosmic rayspallation. A Li abundance for a star in the Bulge, measuring how fast Li was made in theearly Galaxy, would be very interesting. However, most stars no longer have the same amountof Li on their surfaces as was present in their natal gas clouds. Li is easily burned duringpre-main sequence and main-sequence phases of stars and is either destroyed throughout theconvective envelope during the RGB phase or (for a brief time) created in the star itself anddredged up. We have no detection of Li in this star, only a 3- σ upper limit of log ǫ (Li)=1.84dex based on a χ fit to the data (see Johnson et al. 2007 for more details). In Figure 7 weshow this upper limit compared with Li measurements in open cluster stars having a rangeof ages as well as field stars from Lambert & Reddy (2004). We also include the upper limitfrom OGLE-2006-BLG-265S. Lower values can be expected in field stars because of astrationon the main-sequence and because they are often older than the clusters featured in Figure 7and were formed out of gas that had not been polluted by as much Li. Figure 7 shows thatthe Li upper limits in the bulge dwarfs are consistent with the upper limits in field dwarfs.A dwarf with T eff > The Bulge has a different star formation history than the halo/disk. The ratios ofType II/Type Ia pollution or Type II/AGB star pollution at a given [Fe/H] are thereforedifferent as well, and the abundance ratios reflect this. We compare the abundances for bothMOA-2006-BLG-99S and OGLE-2006-BLG-265S with Bulge giants and field stars from thethick/thin disk and halo from literature sources. In Table 3, we summarize the literaturesources we use for each element.
For many elements, observing red giant stars in the Bulge is an effective method ofmeasuring their abundances. However, as shown in § eff vs. log ǫ (Li) for the Hyades and M67 from Balachandran (1995) and for fieldstars from Lambert & Reddy (2004) compared with the upper limit for MOA-2006-BLG-99Sand OGLE-2006-BLG-265S. The bulge dwarf Li limits are consistent with disk stars, whichis not surprising given the age and temperature of these stars. 20 –alters the abundances of C and N. The abundances of C and N that we measure in MOA-2006-BLG-99S therefore represent the first observations of the primordial C and N producedby the chemical evolution of the Bulge.MOA-2006-BLG-99S has [C/Fe]=0.04 ± .
22 and [N/Fe]= − ± .
43 (Fig. 8). The solarvalues of [C/Fe] and N/Fe] show that C and N production kept pace with the Fe productionin the Bulge. There are many sources of C in the Universe (Gustafsson et al. 1999, e.g.).Type II SNe and AGB stars certainly contribute substantial amounts of C and N; the roles ofnovae and Wolf-Rayet stars are less clear. These contributions, whatever they are, track theproduction of Fe in the chemical evolution of the Bulge. Finally, we attempted to measurethe C/ C ratio, which is sensitive to the source of C, being low for low-mass AGB starsand high for Type II SNe. We could only set an uninteresting limit ( C/ C ¿ 1) on thisvery interesting number.
The abundances of Na and Al are elevated in Bulge giants (McWilliam & Rich 1994;Lecureur et al. 2007) compared to disk stars. Metal-rich Type II SNe are predicted toproduce more of the odd-Z elements such as Na and Al than metal-poor Type II SNe andcould potentially be the explanation of the difference between the Bulge and the disk. InFigure 9, we show the [Na/Fe] and [Al/Fe] value for MOA-2006-BLG-99S and OGLE-2006-BLG-265S. These two unmixed stars are on the lower end of the scatter seen in the giants.This could be a hint that the larger [Na/Fe] and [Al/Fe] values seen in giants are due tointernal mixing, but because the dwarf values fall within the scatter outlined in the scatterindicates that more measurements in dwarf stars are needed before any differences in thedistribution of [Na/Fe] and [Al/Fe] in dwarfs and giants can be seen. α elements Figure 10 shows the [O/Fe], [Mg/Fe], [Si/Fe], and [Ca/Fe] for MOA-2006-BLG-99Scompared with halo stars, thin and thick disk stars, and Bulge giants. The metallicityof MOA-2006-BLG-99S is in the range of metallicities measured for Bulge giants, allowingdirect comparison of [ α /Fe] ratios between dwarfs and giants. The agreement is good, asboth the giants and the dwarf have [ α /Fe] below the high [ α /Fe] values of the more metal-poor ([Fe/H] ≤
0) stars. This decline in all [ α /Fe] ratios suggests that Fe from Type Ia SNe isbeing added and that the more metal-rich stars formed sufficiently later to have this ejectum 21 –Fig. 8.— [C/Fe] and [N/Fe] for MOA-2006-BLG-99S (filled black square) compared to Bulgegiants (filled blue circles) and disk dwarfs (open red triangles). The low [C/Fe] values forthe Bulge giants are the result of internal mixing. The [C/Fe] and [N/Fe] values in MOA-2006-BLG-99S, on the other hand, are the result of the pollution of the gas of the Bulge byprevious generations of stars. The solar ratios for these elements impose constraints on theinefficiency of C and N production in the Bulge. 22 –Fig. 9.— [Na/Fe] and [Al/Fe] for MOA-2006-BLG-99S and OGLE-2006-BLG-265S (filledblack squares) compared to Bulge giants (filled blue circles) and field stars (open red tri-angles). The dwarfs fall within the distribution of [Na/Fe] and [Al/Fe] values seen in thegiants, but at the lower edge of that distribution. 23 –in their gas. K is an odd-Z element and is predicted in nucleosynthesis models to be underproducedrelative to the α elements in metal-poor SNe. There are few measurements in the literature,and those that exist show the opposite trend of increasing [K/Fe] with decreasing [Fe/H](Gratton & Sneden 1987; Chen et al. 2000; Cayrel et al. 2004). However, the only K lineavailable for study in most stars, the resonance line at 7698˚A, is affected by non-LTE effects,and these corrections have not yet been applied to large samples. Zhang et al. (2006) derivedNLTE corrections for each star in their sample. The [K/Fe] values were still supersolar atlow metallicities, with the thin disk stars showing a drop in [K/Fe] for[Fe/H] ≥ −
1. The[K/Fe] ratios in the thick disk stars remain high. However, the [K/Mg] ratios showed muchsmaller variations among the different Galactic populations. They argued that the constant[K/Mg] ratio ([K/Mg]= − . ± .
01) in the stars indicated that the nucleosynthesis of K isclosely coupled to that of the α -elements, which is somewhat surprising given the theoreticalpredictions.We measured the 7698˚A line in MOA-2006-BLG-99 and the Sun. We also measured theK abundance in OGLE-2006-BLG-265S using TurboSpectrum and the model atmospheredescribed in Johnson et al. (2007). The [K/Fe] we measure for OGLE-2006-BLG-265S is − . ± .
20 Figure 11 compares the results for the Bulge to the Zhang et al. (2006) results.We applied no NLTE correction, but assumed that the LTE abundances in the Sun andMOA-2006-BLG-99S would be affected by the same amount.The [K/Mg] values are 0.07 ± .
23 for MOA-2006-BLG-99S and 0.12 ± .
27 for OGLE-2006-BLG-265S; the [K/Mg] abundance is still within a narrow range, even in this verydifferent chemical evolution history.
The abundances for Ti, Sc, Mn and Ni are shown in Figure 12. McWilliam & Rich(1994) found that [Ti/Fe] behaves like [O/Fe] in the Bulge, with supersolar [Ti/Fe] ratiosfor many stars, followed by [Ti/Fe] decreasing to solar for [Fe/H >
0. The abundance of Tiis also enhanced in halo stars, leading it to be classified as an “ α -element” for observationalpurposes. In MOA-2006-BLG-99S, [Ti/Fe] is close to solar, in line with the α -elementsdiscussed above. The ratios of [Sc/Fe] and [Ni/Fe] are observed to be close to solar for a 24 –Fig. 10.— [O/Fe], [Mg/Fe], [Si/Fe] and [Ca/Fe] for MOA-2006-BLG-99S and OGLE-2006-BLG-265S (filled black squares) compared to Bulge giants (filled blue circles) and field stars(open red triangles). The [ α /Fe] ratios in MOA-2006-BLG-99S agree well with the valuesmeasured in giants of similar metallicity. 25 –Fig. 11.— [K/Fe] for MOA-2006-BLG-99S and OGLE-2006-BLG-265S (filled black squares)compared to field stars (open red triangles). The [K/Fe] values in the dwarfs fall in the rangeseen in the disk stars. 26 –wide range of populations: halo, thick and thin disk and Bulge. The data for MOA-2006-BLG-99S show that abundances in this Bulge dwarf agree with this picture.For [Fe/H] < − . ∼ − . The percent of Cu and Zn production to be ascribed to different nucleosynthesis sites(e.g., Type II SNe, AGB stars, Type Ia SNe) is uncertain. The observations show that[Cu/Fe] ∼ − ∼ − . ∼ α -rich freeze out in Type II SNe, which is supplemented at higher metallicities by 27 –Fig. 12.— [Sc/Fe], [Ti/Fe], [Mn/Fe], and [Ni/Fe] for MOA-2006-BLG-99S and OGLE-2006-BLG-265S (filled black squares) compared to Bulge giants (filled blue circles) and field stars(open red triangles).The [Ti/Fe] value in MOA-2006-BLG-99S agrees well with [Ti/Fe] ratios measured in bulgegiants of similar metallicity. Within the error bars, the other [iron-peak/Fe] ratios follow thetrends seen the disk stars. 28 –Fig. 13.— [Cu/Fe] and [Zn/Fe] for MOA-2006-BLG-99S and OGLE-2006-BLG-265S (filledblack squares) compared to field stars (open red triangles). The Matteucci et al. (1999)models predict [Zn/Fe] ≈ . The [Ba/Fe] for MOA-2006-BLG-99S falls considerably below the solar value ([Ba/Fe]= − . − ǫ (Ba)=2.39) reportedin Johnson et al. (2007), OGLE-2006-BLG-265S has a [Ba/Fe]= − . α /Fe] values in MOA-2006-BLG-99S are the result of Type Ia pollution. If we assume that Type Ia SNe havecontributed significantly to the abundances of MOA-2006-BLG-99S, which is reasonable,then AGB pollution must trail Type Ia SN production. The evidence on this point is mixed.Simmerer et al. (2004) saw, in addition to a wide range at any given metallicity, a rise inthe [La/Fe] ratios at [Fe/H] > −
2. Because La, like Ba, is mostly due to the s-process,this would indicates the s-process from AGB stars is added before Fe from Type Ia SNecauses the [ α /Fe] ratios to turn over. On the other hand, Mel´endez & Cohen (2007) argued 30 –Fig. 14.— [Ba/Fe] for MOA-2006-BLG-99S and OGLE-2006-BLG-265S (filled black squares)compared to field stars (open red triangles). [Ba/Fe] shows the largest deviation from thetrends seen in the disk stars of any element studied in this paper. This low Ba value isconsistent with the idea that the r-process is the dominate producer of heavy elements inthe Bulge. It also puts constraints on s-process contributions to Ba (and the accompanyingC and N contributions) from AGB stars. 31 –based on the isotope ratios of Mg that 3-6 M ⊙ stars did not start contributing to the halountil [Fe/H] ≥ .
5, at the same time or later than the Type Ia SNe. The [Ba/Fe] measured inOGLE-2006-BLG-265S is closer to the solar value, and suggests that by [Fe/H]=0.5 the Bulgehad reached the same point in chemical evolution as the solar neighborhood, with substantialamounts of Ba supplied by the s-process in AGB stars balancing the iron supplied by TypeIa SNe. Ba abundances for Bulge stars with a wide range of metallicity would help clarifythe origin of the Ba abundance in the Bulge.
6. Conclusions
The results for MOA-2006-BLG-99S demonstrate the unique information that can be ob-tained from high-resolution spectra of microlensed dwarfs in the Bulge. Because microlensingis not biased in metallicity, we can measure the MDF of the Bulge main-sequence stars whensufficient number of highly magnified Bulge dwarfs have been observed with high-dispersionspectrographs. However, the two dwarfs we have studied so far both have [Fe/H] > . α -elements (O, Mg, Si, Ca, and Ti) fall on the trend observedin the Bulge giants. The solar values for the [ α /Fe] ratios suggest that Type Ia SN ejecta areresponsible for some of the Fe, which would be expected if these metal-rich stars were partof a younger population in the Bulge. The [C/Fe] and [N/Fe] ratios are also ∼
0, suggestingthat AGB stars have also begun to contribute these light elements to offset the contributionof Fe from Type Ia SNe. It is therefore unexpected to find [Ba/Fe] ∼ -0 .
5. Such low [Ba/Fe]values, when found in the halo, imply that only Type II SNe and the r-process have pollutedthe gas, and not the s-process from AGB stars.The C and N abundances we measure in MOA-2006-BLG-99S are unaffected by anymixing or dilution and therefore are the result of the chemical evolution of the Bulge. Com-paring the total amount of C+N to that measured in Bulge giants, we conclude that onlymild mixing, as expected in theoretical models, has occurred in the Bulge. This agrees withthe assertions by Lecureur et al. (2007), Zoccali et al. (2006), and McWilliam et al. (2007)that the O, Na and Al abundances in Bulge giants should not have been affected by mixing.The [Na/Fe] and [Al/Fe] ratios for MOA-2006-BLG-99S fall on the lower end of the valuesseen in Bulge giants of this metallicity. The [Na/Fe] and [Al/Fe] ratios in giants are markedabove all else by large scatter and more abundance ratios in dwarfs are needed before a 32 –comparison of scatter can be made.We can use our measurements of the abundances of K, Cu and Zn, rare for Bulge stars,to exploit the different star formation history of the Bulge to study the nucleosynthesis sitesof these elements. The supersolar [Cu/Fe] and subsolar [Zn/Fe] values agree with the modelof Bisterzo et al. (2005) and the production of these two elements predominantly in Type IISNe. The chemical evolution of K is more difficult to understand, because the [K/Mg] valuein the Bulge is similar to that in the metal-poor halo, although the production of K shouldbe enhanced in metal-rich SNe.Finally, we note that the standard microlensing technique of estimating the intrinsiccolor and magnitude of the source star by using the offset from the position of the redclump can be tested by deriving a temperature and gravity from the spectrum. For MOA-2006-BLG-99S, the photometric and spectroscopic temperature agree very well, although theagreement was considerably less good for OGLE-2006-BLG-265S.In summary, the abundances in MOA-2006-BLG-99S provide unique information aboutthe formation and evolution of the Bulge. Larger samples of dwarfs observed in this waywould allow the derivation of the dwarf MDF and the trends in abundance ratios over awide range of metallicity. By taking advantage of microlensing events in the Bulge, we canachieve this goal with a modest amount of telescope time.Our thanks to Ian Thompson for help with the echelle observations at Las Campanas.Thomas Masseron and Bertrand Plez provided invaluable support in installing and runningTurboSpectrum. We would also like to thank Daniel Kelson for codes and support for reduc-ing the MIKE data. Jon Fulbright, Manuela Zoccali, and Solange Rami´rez kindly providedmetallicity data for the bulge giant metallicity distribution functions. We acknowledge sup-port from: NSF AST-042758 and NASA NNG04GL51G (AG). Work by B.S.G. was partiallysupported by a Menzel Fellowship from the Harvard College Observatory. 33 –Table 1. Line Parameters and Equivalent WidthsIon Wavelength E.P. log gf EW star EW sun SourceO I I I I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − gf EW star EW sun SourceCa I − II II − II − I − I − I − I − II − I − I − I I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − gf EW star EW sun SourceFe I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − I − II − II − II − II − I − I − I − I − I − I − I − I − I − I − I − gf EW star EW sun SourceNi I − I − I − I − I − I − I − I − I − I − I − II − ǫ σ ǫ [X/FeI] a σ [ X/F eI ] σ N lines Solarmeas. GS98C (CH) 8.8 0.26 0.04 0.22 0.20 . . . 8.4 8.52N (CN) 8.2 0.56 − I − I I I I I I − II − I − II − I − I II − I I − I − II − a [X/Fe I ] given for all species except Fe I , where [Fe I /H] is given. Table 3. Literature Sources
Source C N O Na Mg Al Si K Ca Sc Ti Mn Ni Cu Zn BaBulge StarsFulbright et al. (2007) x x x x x xRich & Origlia (2005)Rich et al. (2007) x x x x x xLecureur et al. (2007)Cunha & Smith (2006) x x x x xHalo/Disk DataReddy et al. (2006) x x x x x x x x x x x x x xReddy et al. (2003) x x x x x x x x x x x x x x xFeltzing et al. (2007) xBensby & Feltzing (2006) xBensby et al. (2005) x x x x x xBensby et al. (2004) xBensby et al. (2003) x xChen et al. (2004) x x x x x x x x x xCarretta et al. (2000) x xZhang et al. (2006) x
39 –
REFERENCES
Alvarez, R., & Plez, B. 1998, A&A, 330, 1109Balachandran, S. 1995, ApJ, 446, 203Alonso, A., Arribas, S., & Martinez-Roger, C. 1996, A&A, 313, 873Baade, W. 1946, PASP, 58, 249Barklem, P. S., Piskunov, N., & O’Mara, B. J. 2000, A&AS, 142, 467Barklem, P. S., Stempels, H. C., Allende Prieto, C., Kochukhov, O. P., Piskunov, N., &O’Mara, B. J. 2002, A&A, 385, 951Bensby, T., Feltzing, S., & Lundstr¨om, I. 2003, A&A, 410, 527Bensby, T., Feltzing, S., & Lundstr¨om, I. 2004, A&A, 415, 155Bensby, T., & Feltzing, S. 2006, MNRAS, 367, 1181Bensby, T., Feltzing, S., Lundstr¨om, I., & Ilyin, I. 2005, A&A, 433, 185Bernstein, R., Shectman, S. A., Gunnels, S. M., Mochnacki, S., & Athey, A. E. 2003, Pro-ceedings of the SPIE, 4841, 1694Bisterzo, S., Pompeia, L., Gallino, R., Pignatari, M., Cunha, K., Heger, A., & Smith, V.2005, Nuclear Physics A, 758, 284Carretta, E., Gratton, R. G., & Sneden, C. 2000, A&A, 356, 238Castelli, F., & Kurucz, R. L. 2003, Modelling of Stellar Atmospheres, 210, 20Cayrel, R., et al. 2004, A&A, 416, 1117Chen, Y. Q., Nissen, P. E., Zhao, G., Zhang, H. W., & Benoni, T. 2000, A&AS, 141, 491Chen, Y. Q., Zhao, G., Nissen, P. E., Bai, G. S., & Qiu, H. M. 2003, ApJ, 591, 925Chen, Y. Q., Nissen, P. E., & Zhao, G. 2004, A&A, 425, 697Cunha, K., & Smith, V. V. 2006, ApJ, 651, 491Cunha, K., Sellgren, K., Smith, V. V., Ramirez, S. V., Blum, R. D., & Terndrup, D. M.2007, ArXiv e-prints, 707, arXiv:0707.2610 40 –Demarque, P., Woo, J.-H., Kim, Y.-C., & Yi, S. K. 2004, ApJS, 155, 667Feltzing, S., Fohlman, M., & Bensby, T. 2007, A&A, 467, 665Feltzing, S., & Gilmore, G. 2000, A&A, 355, 949Fulbright, J. P., McWilliam, A., & Rich, R. M. 2006, ApJ, 636, 821Fulbright, J. P., McWilliam, A., & Rich, R. M. 2007, ApJ, 661, 1152Gould, A., et al. 2006, ApJ, 644, L37Gratton, R. G., & Sneden, C. 1987, A&A, 178, 179Grevesse, N., & Sauval, A. J. 1998, Space Science Reviews, 85, 161Griest, K. & Safizadeh, N. 1998, ApJ, 500, 37Gustafsson, B., Karlsson, T., Olsson, E., Edvardsson, B., & Ryde, N. 1999, A&A, 342, 426Holtzman, J.A., Watson, A.M., Baum, W.A., Grillmair, C.J., Groth, E.J., Light, R.M.,Lynds, R., & O’Neil, E.J. 1998, AJ, 115, 1946Johnson, J. A., Gal-Yam, A., Leonard, D. C., Simon, J. D., Udalski, A., & Gould, A. 2007,ApJ, 655, L33Johnson, J. A., Ivans, I. I., & Stetson, P. B. 2006, ApJ, 640, 801Kurucz, R. L., Furenlid, I., Brault, J., & Testerman, L. 1984, National Solar ObservatoryAtlas, Sunspot, New Mexico: National Solar Observatory, 1984,Lambert, D. L., & Reddy, B. E. 2004, MNRAS, 349, 757Lecureur, A., Hill, V., Zoccali, M., Barbuy, B., G´omez, A., Minniti, D., Ortolani, S., &Renzini, A. 2007, A&A, 465, 799Lecureur, A. et al. 2008, in prepMcWilliam, A., & Rich, R. M. 1994, ApJS, 91, 749McWilliam, A., Preston, G. W., Sneden, C., & Searle, L. 1995, AJ, 109, 2757McWilliam, A., Rich, R. M., & Smecker-Hane, T. A. 2003, ApJ, 592, L21McWilliam, A., Matteucci, F., Ballero, S., Rich, R. M., Fulbright, J. P., & Cescutti, G. 2007,ArXiv e-prints, 708, arXiv:0708.4026 41 –Mao, S. & Paczy´nski, B. 1991, ApJ, 374, L37Mashonkina, L., Gehren, T., & Bikmaev, I. 1999, A&A, 343, 519Matteucci, F., Raiteri, C. M., Busson, M., Gallino, R., & Gratton, R. 1993, A&A, 272, 421Matteucci, F., Romano, D., & Molaro, P. 1999, A&A, 341, 458Mel´endez, J., & Cohen, J. G. 2007, ApJ, 659, L25Mishenina, T. V., Kovtyukh, V. V., Soubiran, C., Travaglio, C., & Busso, M. 2002, A&A,396, 189Moore, C. E., Minnaert, M. G. J. & Houtgast, J. 1966. The Solar Spectrum 2935˚A- 8770 ˚A,US Govt. Printing Office, Washington, D. C.Ortolani, S., Renzini, A., Gilmozzi, R., Marconi, G., Barbuy, B., Bica, E., & Rich, R. M.1995, Nature, 377, 701Pickering, J. C., Thorne, A. P., & Perez, R. 2001, ApJS, 132, 403Piskunov, N. E., Kupka, F., Ryabchikova, T. A., Weiss, W. W., & Jeffery, C. S. 1995, A&AS,112, 525Ram´ırez, S. V., Stephens, A. W., Frogel, J. A., & DePoy, D. L. 2000, AJ, 120, 833Rich, R. M., & Origlia, L. 2005, ApJ, 634, 1293Rich, R. M., Origlia, L., & Valenti, E. 2007, ApJ, 665, L119Ram´ırez, I., & Mel´endez, J. 2005, ApJ, 626, 465Reddy, B. E., Lambert, D. L., & Allende Prieto, C. 2006, MNRAS, 367, 1329Reddy, B. E., Tomkin, J., Lambert, D. L., & Allende Prieto, C. 2003, MNRAS, 340, 304Rich, R. M. 1988, AJ, 95, 828Sadler, E. M., Rich, R. M., & Terndrup, D. M. 1996, AJ, 112, 171Shetrone, M. D. 1996, AJ, 112, 1517Simmerer, J., Sneden, C., Cowan, J. J., Collier, J., Woolf, V. M., & Lawler, J. E. 2004, ApJ,617, 1091 42 –Smith, G., & Raggett, D. S. J. 1981, Journal of Physics B Atomic Molecular Physics, 14,4015Terndrup, D. M. 1988, AJ, 96, 884Udalski, A., et al. 2005, ApJ, 628, L109Whitford, A. E., & Rich, R. M. 1983, ApJ, 274, 723Yi, S., Demarque, P., Kim, Y.-C., Lee, Y.-W., Ree, C. H., Lejeune, T., & Barnes, S. 2001,ApJS, 136, 417Yoo, J., et al. 2004a, ApJ, 603, 139Zhang, H. W., Gehren, T., Butler, K., Shi, J. R., & Zhao, G. 2006, A&A, 457, 645Zoccali, M., et al. 2003, A&A, 399, 931Zoccali, M., et al. 2006, A&A, 457, L1Zoccali, M., et al. 2008, in prep