Fe and Al Abundances for 180 Red Giants in the Globular Cluster Omega Centauri (NGC 5139)
We present radial velocities, Fe, and Al abundances for 180 red giant branch (RGB) stars in the Galactic globular cluster Omega Centauri ( Cen). The majority of our data lie in the range 11.0V13.5, which covers the RGB from about 1 mag. above the horizontal branch to the RGB tip. The selection procedures are biased towards preferentially observing the more metal–poor and luminous stars of Cen. Abundances were determined using equivalent width measurements and spectrum synthesis analyses of moderate resolution spectra (R13,000) obtained with the Blanco 4m telescope and Hydra multifiber spectrograph. Our results are in agreement with previous studies as we find at least four different metallicity populations with [Fe/H]=–1.75, –1.45, –1.05, and –0.75, with a full range of –2.20[Fe/H]–0.70. [Al/Fe] ratios exhibit large star–to–star scatter for all populations, with the more than 1.0 dex range of [Al/Fe] decreasing for stars more metal–rich than [Fe/H]–1.4. The minimum [Al/Fe] abundance observed for all metallicity populations is [Al/Fe]+0.15. The maximum abundance of log (Al) is reached for stars with [Fe/H]–1.4 and does not increase further with stellar metallicity. We interpret these results as evidence for type II SNe providing the minimum [Al/Fe] ratio and a mass spectrum of intermediate mass asymptotic giant branch stars causing the majority of the [Al/Fe] scatter. These results seem to fit in the adopted scheme that star formation occurred in Cen over 1 Gyr.
The Galactic globular cluster Omega Centauri ( Cen) presents a unique opportunity to study the chemical evolution of both a small stellar system and stars with common formation histories covering a metallicity range of more than a factor of 10, a defining characteristic of Cen that has been known since the initial discovery of its unusually broad red giant branch (RGB) by Woolley (1966). Although Cen is the most massive Galactic globular cluster, with an estimated mass of 2–710 M (Richer et al. 1991; Meylan et al. 1995; van de Ven et al. 2006), it does not appear to have an exceptionally deep gravitational potential well (Gnedin et al. 2002). This seems to negate a simple explanation that Cen evolved as a typical globular cluster that was more easily able to retain supernova (SN) and asymptotic giant branch (AGB) ejecta for self–enrichment. This fact coupled with the cluster’s retrograde orbit and disk crossing time of 1–210 years (e.g., Dinescu et al. 1999), which could severely inhibit star formation, are some of the strongest arguments against Cen having a Galactic origin. Instead, it has been proposed (e.g., Dinescu et al. 1999; Smith et al. 2000; Gnedin et al. 2002; Bekki & Norris 2006) that Cen may be the remaining nucleus of a dwarf spheroidal galaxy that evolved in isolation and was later accreted by the Milky Way, suggesting the progenitor system was perhaps a factor of 100–1000 times more massive than what is presently observed.
Recent spectroscopic and photometric studies (Norris & Da Costa 1995; Norris et al. 1996; Suntzeff & Kraft 1996; Lee et al. 1999; Hilker & Richtler 2000; Hughes & Wallerstein 2000; Pancino et al. 2000; Smith et al. 2000; van Leeuwen et al. 2000; Rey et al. 2004; Stanford et al. 2004; Piotto et al. 2005; Sollima et al. 2005a; Sollima et al. 2005b; Kayser et al. 2006; Sollima et al. 2006; Stanford et al. 2006; Stanford et al. 2007; van Loon et al. 2007; Villanova et al. 2007) have confirmed the existence of up to five separate stellar populations ranging in metallicity from [Fe/H]–2.2 to –0.5, with a peak in the metallicity distribution near [Fe/H]–1.7 and a long tail extending to higher metallicities. In addition to the metal–poor and intermediate metallicity populations initially seen in the Woolley (1966) photometric study, Lee et al. (1999) and Pancino et al. (2000) discovered the existence of the most metal–rich RGB at [Fe/H]–0.5, commonly referred to as the anomalous RGB (RGB–a). The RGB–a is primarily observed in the central region of the cluster and contains approximately 5 of the total stellar population (Pancino et al. 2000), in contrast to the dominant metal–poor population that contains roughly 75 of cluster stars. Additionally, there is some evidence (Norris et al. 1997) that the metal–rich population exhibits smaller radial velocity dispersion and rotation than the metal–poor population. Sollima et al. (2005b) confirmed the Norris et al. (1997) results but also showed that the most metal–rich stars ([Fe/H]–1) exhibit an increasing velocity dispersion as a function of increasing metallicity, which could be evidence for accretion events occurring within Cen’s progenitor system (Ferraro et al. 2002; Pancino et al. 2003); however, this result is not yet confirmed (Platais et al. 2003, but see also Hughes et al. 2004). It should be noted that Pancino et al. (2007), using radial velocity measurements of 650 members with measurement uncertainties of order 0.5 km s, have found no evidence for rotational differences among the different metallicity groups.
The distribution of main–sequence turnoff (MSTO) and subgiant branch (SGB) stars matches that observed on the RGB, such that one can trace the evolutionary sequence of each population from at least the MSTO to the RGB using high precision photometry (e.g., Villanova et al. 2007). The main–sequence (MS) has proved equally as complex as the SGB and RGB, with the discovery by Anderson (1997) of a red and blue MS (BMS). Interestingly, Piotto et al. (2005) discovered that the BMS was more metal–rich than the red MS, suggesting the BMS could be explained assuming a higher He content, perhaps as high as Y0.38 (Bedin et al. 2004; Norris 2004; Lee et al. 2005; Piotto et al. 2005).
While it is clear that multiple populations are present in this cluster, there has been some debate regarding the age of each population. There is general agreement that the age range is between about 0 and 6 Gyrs (Norris & Da Costa 1995; Hilker & Richtler 2000; Hughes & Wallerstein 2000; Pancino et al. 2002; Origlia et al. 2003; Ferraro et al. 2004; Hilker et al. 2004; Rey et al. 2004; Sollima et al. 2005a; Sollima et al. 2005b; Villanova et al. 2007), though the recent work by Stanford et al. (2006) suggests the most likely age range is 2–4 Gyrs, with the metal–rich stars being younger. For the case of monotonic chemical enrichment in a single system, one would expect the more metal–rich stars to be younger than the more metal–poor; however, this assumption has been questioned by Villanova et al. (2007) who suggested the metal–rich stars and 33 of the metal–poor stars are the oldest with the remaining 2/3 of the metal–poor population being 3–4 Gyrs younger. The picture of Cen’s formation is further compounded by observations of RR Lyrae horizontal branch (HB) stars that reveal a bimodal metallicity distribution without a trend in He enhancement as a function of [Fe/H] (Sollima et al. 2006). The important point here is that a group of RR Lyrae stars exists with the same metallicity as the BMS but without the presumed He enhancement. A He–rich secondary population would not produce a significant RR Lyrae population unless a 4 Gyr age difference was present with respect to the dominant metal–poor population (Sollima et al. 2006). The required age difference is therefore inconsistent with most age spread estimates that put 4 Gyrs.
Cen’s chemical evolution history has so far proved difficult to interpret from measured abundances of light (Z27), , Fe–peak, s–process, and r–process elements. In “normal” Galactic globular clusters, C, N, O, F, Na, Mg (sometimes), and Al often exhibit large star–to–star variations, in some cases exceeding more than a factor of 10 (e.g., see recent review by Gratton et al. 2004). In contrast, the heavier –elements (e.g., Ca and Ti) show little to no variation and are enhanced relative to Fe at [/Fe]+0.30, with a decreasing ratio for clusters with [Fe/H]–1. Likewise, Fe and all other Fe–peak, s–process, and r–process elements show star–to–star variations of 0.10–0.30 dex. Additionally, nearly all globular clusters are enriched in r–process relative to s–process elements by about 0.20 dex. In Cen, [Fe/H] covers a range of more than 1.5 dex and, as previously stated, it has a potential well comparable to that of other globular clusters, suggesting it had to be different in the past to undergo self–enrichment. The scenario of two or more globular clusters merging seems unlikely now given the results of Pancino et al. (2007) and the typically large orbital velocities coupled with the small velocity dispersions of clusters (Ikuta & Arimoto 2000). While Cen exhibits large abundance variations for several of the light elements at various metallicities (e.g., Norris & Da Costa 1995; Smith et al. 2000), the mean heavy –element enhancement is surprisingly uniform at [/Fe]+0.30 to +0.50 (Norris & Da Costa 1995; Smith et al. 2000; Villanova et al. 2007), with perhaps a trend of decreasing [/Fe] at [Fe/H]–1 (Pancino et al. 2002). The s–process elements show a clear increase in abundance relative to Fe with a plateau occurring at [Fe/H]–1.40 to –1.20 (Norris & Da Costa 1995; Smith et al. 2000). However, unlike in globular clusters, s–process elements are overabundant with respect to r–process elements, where [Ba/Eu] typically reaches between 0.5 and 1.0 (Smith et al. 2000), indicating a strong presence of AGB ejecta.
Many globular cluster giants show clear C–N, O–Na, O–Al, Mg–Al, and in the case of M4 (Smith et al. 2005), F–Na anticorrelations alongside a Na–Al correlation (e.g., Gratton et al. 2004). In addition to these anomalies being present in the atmospheres of RGB stars, similar relations have been observed in some globular cluster MS and MSTO stars (e.g., Cannon et al. 1998; Gratton et al. 2001; Cohen et al. 2002; Briley et al. 2004a; 2004b; Boesgaard et al. 2005). According to standard evolutionary theory, first dredgeup brings the products of MS core hydrogen burning to the surface and homogenizes approximately 70–80 of the star, resulting in C depletion, N enhancement, and a lowering of the C/C ratio from about 90 to 25 (e.g., Salaris et al. 2002). The decline in [C/Fe] and C/C has been verified via observations in both globular cluster (Bell et al. 1979; Carbon et al. 1982; Langer et al. 1986; Bellman et al. 2001) and field stars (Charbonnel & do Nascimento 1998; Gratton et al. 2000; Keller et al. 2001) as strong evidence for in situ mixing occurring along the RGB. However, as the advancing hydrogen–burning shell (HBS) crosses the molecular weight discontinuity left by the convective envelope’s deepest point of penetration, extra mixing not predicted by canonical theory occurs in both field and cluster stars, driving down [C/Fe] further and allowing C/C to reach the CN–cycle equilibrium value of 4. The mechanism responsible for this extra mixing is not known, though thermohaline mixing (Charbonnel & Zahn 2007) may ameliorate the problem. While halo field and cluster giants share these same trends, differences arise when considering O, Na, and Al abundances. Field stars do not exhibit most of the familiar correlations/anticorrelations and large star–to–star variations seen in globular cluster stars and instead remain mostly constant from the MS to the RGB tip (e.g., Ryan et al. 1996; Fulbright 2000; Gratton et al. 2000).
The reason for the observed differences between cluster and field giants is not known, but obviously the higher density cluster environment is a key factor. Coupled O depletions and Na/Al enhancements are clear signs of high temperature (T4010 K) H–burning via the ON, NeNa, and MgAl proton–capture cycles, but this does not necessarily mean those cycles are operating in the RGB stars we presently observe and instead may be from the ejecta of intermediate mass (IM) AGB stars (3–8 M) that underwent hot bottom burning (HBB) and polluted the gas from which the current stars formed. One of the strongest arguments against in situ mixing is the observed abundance relations on the MS and MSTO matching those on the RGB because these stars are both too cool for the ON, NeNa, and MgAl cycles to operate and their shallow envelope convection zones do not reach deep enough to bring up even CN–cycled material. Additionally, Shetrone (1996) showed that at least in M13 giants, Mg is anticorrelated with Al instead of Mg and/or Mg, which means temperatures not achievable in low mass RGB stars (at least 7010 K) are needed to activate the full MgAl chain (Langer et al. 1997); however, these temperatures are reached in HBB conditions. Current models of low mass RGB stars (e.g., Denissenkov & Weiss 2001) indicate Al is only produced deep in the stellar interior by burning Mg and convective mixing reaching these depths would cause a second increase in the surface abundance of both Na and He. It should be noted that if it is instead Al (110 yrs) causing the abundance anomalies on the upper RGB, then the O–Na and Na–Al relations can be explained in a self–consistent manner via in situ mixing (Denissenkov & Weiss 2001). Also, there is some evidence that O depletions and Na/Al enhancements become stronger in the upper 0.7 mag before the RGB tip in M13 (e.g., Sneden et al. 2004; Johnson et al. 2005), indicating the possible operation of additional deep mixing episodes in some stars. Although it is more difficult to believe in situ mixing is responsible for the Mg–Al anticorrelation, the same may not be true for O and Na. In or just above the HBS of a metal–poor low mass RGB star, the O–Na anticorrelation can be naturally explained because the ON and NeNa cycles can operate at T4010 K (Denisenkov & Denisenkova 1990; Langer et al. 1993). Of course, this cannot be the case for any O–Na anticorrelation observed in MSTO and SGB stars and does require convective mixing in RGB stars to penetrate past the radiative zone separating the bottom of the convective envelope and the top of the HBS.
While pollution from a previous generation of more massive AGB stars seems an attractive explanation, there are a few important issues. Predicted IM–AGB stellar yields are sensitive to the adopted treatment of convection because it affects other important parameters such as luminosity, number of thermal pulses, third dredgeup efficiency, envelope temperature structure, and mass loss (Ventura & D’Antona 2005a). The two most common methods employed are mixing length theory (MLT) (e.g., Fenner et al. 2004) and the full spectrum of turbulence (FST) model (e.g., Ventura & D’Antona 2005b), with the latter providing more efficient convection. In Cen and all other globular clusters observed, the [C+N+O/Fe] sum is constant (Pilachowski et al. 1988; Dickens et al. 1991; Norris & Da Costa 1995; Smith et al. 1996; Ivans et al. 1999), but models based on MLT indicate stars forming from different generations of AGB ejecta should show a large increase in the CNO sum (e.g., Lattanzio et al. 2004). In contrast, FST models keep [C+N+O/Fe] constant to within about a factor of 2 due to enhanced mass loss and fewer third dredgeup episodes (Ventura & D’Antona 2005b). Although Na and Al production could be due to HBB, it is difficult to produce the observed O depletion of 1.0 to 1.5 dex along with the required Na enhancement (e.g., Denissenkov & Herwig 2003; but see also Ventura & D’Antona 2005b). Self–consistent models of globular cluster enrichment from AGB ejecta fail to reproduce the MgAl anticorrelation seen in several globular clusters, including Cen, where Mg increases relative to Al instead of decreases (Fenner et al. 2004). Without an evolutionary scenario, O deficient, Na/Al enhanced stars must have preferentially formed out of enriched gas relative to “O–normal” stars (i.e., [O/Fe]+0.30) and Yong et al. (2003) point out that even with no O present in the enriched gas, these stars would require a composition of 90 enriched, 10 “normal” material to obtain the observed O deficiency. Lastly, AGB stellar envelopes contain roughly 36 He by mass (Lattanzio et al. 2004), but O–poor, Na/Al–rich stars do not appear to be particularly He–rich; however, this does not rule out AGB stars as the source of the He–rich BMS observed in Cen. Given the evidence for and against evolutionary and primordial processes, a hybrid scenario probably needs to be invoked to explain all abundance anomalies.
Given the inherently large spread in metallicity of stars in Cen and that Al is the heaviest element sensitive to proton–capture nucleosynthesis at temperatures achieved in the interiors of low mass metal–poor RGB stars, we present radial velocities, Fe, and Al abundances for 180 RGB stars covering –2.20[Fe/H]–0.70. With additional data from the literature covering from the MS to the RGB tip, we address the issues of star formation and possible pollution sources driving the chemical evolution of Cen as a function of metallicity.
2 Observations and Reductions
The observations of all 180 giants in Cen were obtained with the Blanco 4m telescope using the Hydra multifiber positioner and bench spectrograph at the Cerro Tololo Inter–American Observatory. All observations were obtained using the “large” 300 (2) fibers. The full spectral coverage ranged from 6450–6750 Å, centered on 6600 Å; however, wavelengths blueward of 6500 Å lie on the shoulder of the filter response curve, making continuum placement difficult. Therefore, we truncated the spectra to include only the region from 6500–6750 Å. The 316 line mm echelle grating and Blue Air Schmidt Camera provided a resolving power of R(/)13,000 (0.5 Å FWHM) at 6600 Å. A list of our observation dates and exposure times is provided in Table 1.
Target stars, coordinates, photometry, and membership probability were taken from the proper motion study by van Leeuwen et al. (2000). Stars were given priority in the Hydra assignment program based on V magnitude, with a focus on stars in the range 11.0V14.0, which includes all giants in the cluster brighter than the HB up to the RGB tip. Only stars with membership probabilities 80 were included for possible study. All observations took place between 2003 July 17 and 2003 July 19. Three different Hydra setups were used with exposure times ranging from 1800 to 3600 seconds. Each setup allowed approximately 100 fibers to be placed on targets, yielding a total initial sample size of nearly 300 stars. At V13.5, reaching a signal–to–noise (S/N) ratio of 100 requires 3 hours of total integration time. Unfortunately, weather and time constraints led to one of the setups receiving less than 2 hours of integration time with an average S/N of less than 50. Many of these stars had to be excluded from analysis due to poor S/N; however, the final sample size still includes nearly 200 stars. These are shown in Figure 1 along with the complete sample given in van Leeuwen et al. (2000) for 11.0V14.0.
Due to Cen’s broad RGB, selection effects must be taken into account when interpreting abundance results. Figure 2 shows our observed completion fraction of RGB stars both as a function of V magnitude and B–V color compared to the deeper photometric study by Rey et al. (2004). Since our observing program is biased towards selecting brighter stars, our sample includes more metal–poor than metal–rich stars because metal–rich stars have lower V magnitudes due to H opacity increasing with increasing metallicity. While we observed 75 of all RGB tip stars available, the fraction of stars observed decreases to 15–50 in the range 11.5V13.0. Likewise, in considering completeness in B–V color, our sample includes stars of higher luminosity for a given B–V, biasing our results towards the more metal–poor regime.
Figure 3 shows the location of our observed stars in right ascension and declination relative to the cluster center, defined by van Leeuwen et al. (2000) as 132645.9, –472837.0 (J2000) and marked with a cross in the figure. Since some evidence exists for a correlation between metallicity and distance from the cluster center (Norris et al. 1996; Suntzeff & Kraft 1996; Norris et al. 1997; Hilker & Richtler 2000; Pancino et al. 2000; Rey et al. 2004), we have observed stars as uniformly as possible at radii extending out to 20. Near the cluster center, crowding and the physical size of the fibers limited the number of observations inside about 2 core radii, where the core radius is approximately 1.40 (Harris 1996; rev. 2003 February). We illustrate this effect with the ellipses in Figure 3 that correspond to 1, 5, and 10 core radii.
Basic data reductions were accomplished using the IRAF111IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. package ccdproc to trim the bias overscan region and apply bias level corrections. The IRAF task dohydra was employed to correct for scattered light, extract the one–dimensional spectra, remove cosmic rays, apply a flat–field correction, linearize the wavelength scale, and subtract the sky spectrum. Wavelength calibrations were carried out using a high S/N solar sky spectrum because the ThAr lamp was unavailable. Standard IRAF tasks were used to co–add and normalize the spectra. Typical S/N ratios for individual exposures ranged from 25–50, with co–added spectra having S/N between 75 and 150.
3 Radial Velocity Determinations
Cen’s location in the thick disk (Dinescu et al. 1999) makes field star contamination a more serious problem than for typical halo globular clusters. While we initially only chose targets with high membership probabilities from van Leeuwen et al. (2000), direct measurements of target radial velocities assist with membership confirmation. Radial velocities were determined using the IRAF tasks rvcor, to correct for heliocentric motion, and fxcor, to determine the heliocentric radial velocity. For the comparison spectrum, we used the same high S/N daylight sky spectrum that was used for wavelength calibration. A summary of our determined radial velocities along with membership probabilities from van Leeuwen et al. (2000) are given in Table 2.
The largest radial velocity study of Cen stars to date is by Reijns et al. (2006), who determined radial velocities for 2,000 RGB stars. Their study finds a strongly peaked distribution near 232 km s, with a median uncertainty of less than 2 km s and a velocity dispersion of 6 km s for the inner 25 of the cluster. Similarly, Mayor et al. (1997) find V=232.8 0.7 km s (17.5 km s) for 471 stars and Suntzeff & Kraft (1996) find V=234.7 1.3 km s (=11.3 km s) for their “bright” sample of 199 stars. Recently, Pancino et al. (2007) determined radial velocities for 650 RGB stars and found V=233.4 0.5 km s (=13.2 km s). We find in agreement with these studies: V=231.8 km s 1.6 km s (=11.6 km s). Our observations do not provide an absolute velocity calibration, but comparison with the other observations of the average velocity of cluster stars suggests that the systematic error of our radial velocities is less than about 2 km s. Since all of our stars listed in Table 2 are less than 3 away from the cluster averaged velocity and Cen’s velocity is high relative to the general field population, it is unlikely any of our targets are field stars.
We have derived Fe and Al abundances using lines available in the spectral range 6500–6750 Å with either equivalent width or synthetic spectrum analyses. Spectrum synthesis was used to determine Al abundances in metal–rich and/or CN–strong stars. When multiple lines were available, the stated abundances represent the average of the individual lines. Effective temperatures (T) and gravities (log g) were estimated using published (V–K) photometry. T and microturbulence (V) were further refined via spectroscopic analyses. The analysis follows the methods described in Johnson et al. (2005) and Johnson & Pilachowski (2006).
4.1 Model Stellar Atmospheres
Using V photometry from van Leeuwen et al. (2000) and K photometry from 2MASS, we estimated T with the color–temperature relation described in Alonso et al. (1999; 2001), which is based on the infrared flux method (Blackwell & Shallis 1977). However, the Alonso et al. (1999) method requires the photometry to be on the Carlos Sánchez Telescope (TCS) photometric system. We transformed the V and K magnitudes onto the TCS system using the transformations provided in Alonso et al. (1994; 1998) and Carpenter (2001), as summarized in Johnson et al. (2005). To correct for interstellar reddening and extinction, we applied the correction recommended by Harris (1996; rev. 2003 February) of E(B–V)=0.12 and used E(V–K)/E(B–V)=2.7 (Johnson 1965). While Calamida et al. (2005) claim differential reddening, perhaps differing by as much as a factor of two near the core, could be a problem, the well defined evolutionary sequences seen in Villanova et al. (2007) seem to indicate it is not too severe. Van Loon et al. (2007) find some evidence for interstellar absorption by gas in the cluster, but this is concentrated near the core where our observations are sparse. Therefore, we have only applied a uniform reddening correction. Bolometric corrections were applied using the empirical relations given in Alonso et al. (1999) assuming a distance modulus of (m–M)=13.7 (van de Ven et al. 2006).
Applying the proper color–temperature relation requires knowledge of a star’s metallicity. Therefore, we took the empirical relation given in van Leeuwen et al. (2000; their eq. 15), which gives [Ca/H] as a function of V and B–V, and assumed [Ca/Fe]+0.30 for [Fe/H]–1.0 (e.g., Norris & Da Costa 1995), with a linear decrease towards [Ca/Fe]=0.0 at [Fe/H]=0.0. This gave a rough estimate of [Fe/H] for each star and allowed us to choose the proper equation in Alonso et al. (1999).
Since only one Fe II line was available for analysis (6516 Å), we determined surface gravity using the standard relation,
instead of the ionization equilibrium of Fe. We assumed M=0.80 M for all stars, regardless of metallicity. Though there may be an intrinsic age spread of a few Gyr on the RGB (see 5 for further discussion on this issue), this will lead to a mass difference only of order 0.05 M, which is negligible for surface gravity determinations.
In addition to T, log g, and [Fe/H] estimates, we also needed a starting point with V. Initial estimates were based on the empirical relation derived in Pilachowski et al. (1996), which gives V as a function of T for metal–poor field giants and subgiants. Typical V values ranged from about 1.3–2.3 km s in the temperature range 5000–3800 K, respectively.
We generated the model stellar atmospheres by interpolating in the ATLAS9222The model atmosphere grids can be downloaded from http://cfaku5.cfa.harvard.edu/grids.html. (Castelli et al. 1997) grid of models without convective overshoot. Initial models were created using the T, log g, [Fe/H], and V estimates as described above. T was further refined by removing trends in Fe abundance as a function of excitation potential. Likewise, V was improved by removing trends in Fe abundance as a function of reduced width (log(EW/)). A comparison between photometric and spectroscopically determined temperatures is given in the top panel of Figure 4. Typical photometric and spectroscopic temperature estimates agree to within approximately 100 K. The bottom panel of Figure 4 shows our spectroscopically determined V as a function of T for different metallicity bins with a linear least squares fit given by,
which is independent of metallicity. This fit agrees to within 0.10–0.15 km s to that given in Pilachowski et al. (1996). Figure 5 shows our derived [Fe II/H] given as a function of [Fe/H]. As stated above, we only had one Fe II line available for analysis, but the fact that both Fe estimates agree to within 0.16 dex on average (=0.12 dex) leads us to believe our surface gravity estimates are not in serious error. A complete list of our adopted model atmosphere parameters is provided in Table 3.
4.2 Derivation of Abundances
Abundances were determined using equivalent width analyses for all Fe lines and most Al lines, with the exception of cases where evidence for considerable CN contamination near the 6696, 6698 Å Al doublet (i.e., metal–rich and/or CN–strong stars) existed and spectrum synthesis was used instead. We measured equivalent widths using a FORTRAN program developed for this project that interactively fits a Gaussian curve to each absorption line by implementing a Levenberg-Marquardt algorithm (Press et al. 1992) to find the least–squares fit given a continuum level and limits of integration. A high resolution, high S/N Arcturus spectrum333The Arcturus Atlas can be downloaded from the NOAO Digital Library at http://www.noao.edu/dpp/library.html. was simultaneously overplotted for each spectrum to aide in continuum placement and line identification. The program also has the ability to fit up to five Gaussians simultaneously for deblending purposes; however, all equivalent widths were verified independently using IRAF’s splot package.
4.2.1 Equivalent Width Analysis
Final abundances were calculated using the abfind driver in the 2002 version of the local thermodynamic equilibrium line analysis code MOOG (Sneden 1973). Adopted log gf values are the same as those employed in Johnson et al. (2006), which were adapted from line lists provided in Thévenin (1990), Sneden et al. (2004; modified from Ivans et al. 2001), and Cohen & Meléndez (2005). A summary of our line list is given in Table 4 and the measured equivalent widths are provided in Table 5.
While we had identified 20 Fe I lines for analysis, in most cases only 10–15 lines could be used due to severe line blending, bad ccd pixels, or line strength. In this sense, only lines lying on the linear part of the curve of growth were used, which meant neglecting almost all lines with a reduced width larger than about –4.5 (roughly 200 mÅ at 6600 Å). This unfortunately meant that many lines in metal–rich stars are too strong to give accurate abundances using our techniques. For the cases where Al abundances were determined using equivalent width measurements, weak line blends were taken into account using deblending methods. As stated above, stars with strong line blending or molecular line blanketing in the region near the Al doublet were analyzed with spectrum synthesis.
Typical uncertainties are quite small for [Fe/H] determinations with internal line–to–line spreads of 0.10–0.15 dex and / 0.05 dex on average. Sample spectra for stars of approximately the same T but different metallicities are shown in Figure 6. Here we illustrate that our [Fe/H] determinations are at the very least consistent in a relative sense as one notices the increasing Fe line strengths and CN–band strengths with increasing metallicity. The uncertainty in Al abundances is larger given that only two lines are available, but the two lines give a remarkably consistent abundance, with an average /=0.08 dex. It should be noted that in several of our spectra only one Al line could be confidently measured due mostly to bad pixels. In Figure 6, the reader can see the stark contrast in line strength between a star such as 51021, which has [Al/Fe]=+0.15 at [Fe/H]=–1.44, and star 61085, which has [Al/Fe]=+0.97 at [Fe/H]=–1.15. A summary of all derived abundances and associated / values is given in Table 6.
4.2.2 Spectrum Synthesis Analysis
As mentioned above, we determined Al abundances for metal–rich and/or CN–strong stars using the synth driver in MOOG. Candidates for spectrum synthesis were chosen based on visual inspection of the 6680–6700 Å region, where the majority of lines surrounding the Al doublet are CN lines. Stars where CN contamination was seen between the Al lines were designated for synthetic spectrum analysis (e.g., see Figure 6, lower two spectra).
The atomic line list (with the exception of the two Al lines) was taken from the Kurucz atomic line database444The Kurucz line list database can be accessed via http://kurucz.harvard.edu/linelists.html.. We adjusted the oscillator strengths from this line list so the line strengths matched those in the solar spectrum. For the CN molecular line list, we used a combination of one available from Kurucz and one provided by Bertrand Plez (2007, private communication; for a description on how the line list was prepared, see Hill et al. 2002).
Since most of the program stars do not have known C, N, or C/C abundances, we started with [C/Fe]=–0.5, [N/Fe]=+1.5, and C/C=5, values roughly consistent with previous work (e.g., Norris & Da Costa 1995; Smith et al. 2002). We then treated the nitrogen abundance as a free parameter and adjusted it until a satisfactory fit was achieved. Typical best fit [N/Fe] values were +1.0 to +1.5. To test the effect of different C/C ratios, we generated two sets of spectra with C/C=5 and C/C=1000. The fits to the CN lines were indistinguishable between the two cases, meaning C is the dominant isotope in this spectral region and thus synthesized CN lines are insensitive to the C abundance.
With the CN lines fit, we were then able to adjust the Al abundance until the synthetic spectrum matched the observed. Sample synthesis fits are given in Figure 7 for a metal–poor and metal–rich case. Aside from the CN lines, the Fe I line near the 6696 Å feature is the only other contaminating line in the region, but this line has an excitation potential of nearly 5 eV, making its contribution mostly negligible in these cool stars. Generally, the abundances given by the 6696 and 6698 Å lines agreed to within about 0.10 dex. Since a significant percentage of our Al abundances were determined using synthesis analyses, we tested for systematic offsets between synthesis and equivalent width methods. For sample stars that were both metal–poor and did not show signs of CN contamination, the difference in [Al/Fe] determined via both methods was less than 0.05 dex. However, for higher metallicity stars and those with possible CN contamination, the difference was 0.10–0.20 dex, with equivalent width analyses always overestimating the abundance. The quoted values for Al abundances derived via spectrum synthesis are given as the average from those two lines. A summary of our derived abundances is given in Table 6. Stars with Al determinations via synthesis are designated by “Syn” in the 6696 and 6698 Å columns of Table 5.
4.2.3 Abundance Sensitivity to Model Atmosphere Parameters
We tested the effects on derived abundances from changes in model atmosphere parameters by altering T 100 K, log g 0.25 cm s, and V 0.25 km s for models of [Fe/H]=–2.0, –1.5, and –1.0. As can be seen in Table 7, T uncertainties are the primary source of error for Fe I and Al I, and surface gravity is the primary source for Fe II abundances. This seems logical given that Fe I and Al I reside in a subordinate ionization state, and Fe II exists in the primary ionization state.
Following Table 7, an uncertainty of order 100 K in T leads to an error of 0.10–0.20 dex in Fe I, though the effect is somewhat reduced at higher metallicity. The opposite is true for Fe estimates based solely on the Fe II line, where the error range is 0.05–0.10 dex and the uncertainty becomes larger with increasing metallicity. Though the variation in Al I abundance as a function of T uncertainty is smaller than for Fe I, it is still of order 0.10 dex with a weak dependence on metallicity.
The effects of surface gravity uncertainty are of order 0.10 dex for the Fe II line, but are negligible for the neutral Fe and Al lines. For this reason, enforcing ionization equilibrium between different species is often used for constraining surface gravity estimates. As mentioned in 4.2.1, having only one Fe II line means the Fe abundance derived from Fe II is probably no more accurate than the typical line–to–line scatter present in Fe I (0.10–0.15 dex). Combined with the sensitivity of Fe II to surface gravity estimates of order 0.25 cm s, the fact that agreement between Fe I and Fe II is better than about 0.10 dex (see Figure 5) suggests estimates based on evolutionary arguments provide a decent approximation to the surface gravity; however, Table 7 shows this has little effect on our derived Fe I and Al I abundances. From this, we can safely assume that contamination from AGB stars, which have M0.60 M and thus a lower surface gravity, will not significantly alter our results.
The ad hoc microturbulence parameter, adjusted to remove abundance trends as a function of reduced width, has the strongest effect for lines lying on the flat part of the curve of growth. As is seen in Table 7, the effect on the Fe I abundance due to uncertainty in V increases with increasing metallicity because the lines become progressively stronger. However, Fe II and Al I are mostly unaffected due to their relatively small equivalent widths and the effect on Fe I is still 0.10 dex even at [Fe/H]=–1.0.
In addition to variations in model stellar atmosphere parameters we tested the sensitivity of Al abundance to CN strength via spectrum synthesis by varying [N/Fe]0.30 dex. Changing the nitrogen abundance by this amount worsens the fit to the CN lines in the spectrum, but alters the derived [Al/Fe] abundance less than 0.10 dex at all metallicities. Note that since [O/Fe] is unknown for most of our program stars and [O/Fe] can have values ranging from about +0.30 to less than –0.50, it is not possible to constrain the molecular equilibrium equations to derive true [C/Fe] and [N/Fe]. We present the [Al/Fe] results for each metallicity bin in Table 7.
4.3 Comparison with the Literature
While Cen has been the subject of multiple abundance studies (see 1 for a brief review), most of these are low resolution studies that do not involve elements other than Fe and/or Ca. Therefore, we are only comparing results in the literature for which moderate to high resolution Al data are available and with which we have three or more stars in common. This limits the comparison to Brown & Wallerstein (1993; 3 stars), Norris & Da Costa (1995; 24 stars), Zucker et al. (1996; 4 stars), and Smith et al. (2000; 3 stars).
In Figure 8, we present the values of T, log g, [Fe/H], and V given in the literature versus those obtained in this study. As can be seen from the figure, agreement is quite good for the temperature and surface gravity estimates, with the scatter increasing slightly for the metallicity and microturbulence estimates. For T, the average offset between our study and the literature is –7 K (50 K), and the average difference for log g is –0.02 cm s (0.10 cm s). This indicates that any disagreement between literature Fe and Al abundances and ours is not due to choices of T and log g. Similarly, [Fe/H] measurements agree to within 0.02 dex on average (0.20 dex). The reason for the larger dispersion in microturbulence estimates is not entirely clear, but it could be due to factors such as the number of lines available, data quality, continuum placement, and type of lines used (i.e., high and/or low excitation potential). However, on average the agreement is within 0.10 km s (0.25 km s).
Comparison between our derived [Al/Fe] abundances versus those in the literature are provided in Figure 9. Given the various data qualities, choices of model atmospheres and parameters, and adopted atomic line data, agreement is again quite good. The average offset between our derived abundances and those available in the literature is 0.06 dex (0.30 dex). Given that typical uncertainties in [Al/Fe] are of order 0.10–0.20 dex, agreement is comparable to that range.
5 Results and Discussion
5.1 Fe Abundances
As discussed in 1, it has been known for many years and shown by several authors that Cen has a considerable spread in metallicity that ranges from slightly less than [Fe/H]=–2.0 to more than [Fe/H]=–0.7. While several lower resolution spectroscopic (Norris et al. 1996; Suntzeff & Kraft 1996; Sollima et al. 2005b; Kayser et al. 2006; Stanford et al. 2006; Stanford et al. 2007; van Loon et al. 2007555The referee noted discrepancies between the [Fe/H] values derived by Norris & Da Costa (1995) and van Loon et al. (2007). We note that our results agree with Norris & Da Costa and a detailed resolution of this problem is beyond the scope of this paper.; Villanova et al. 2007) and photometric (Lee et al. 1999; Hilker & Richtler 2000; Hughes & Wallerstein 2000; Pancino et al. 2000; van Leeuwen et al. 2000; Rey et al. 2004; Stanford et al. 2004; Sollima et al. 2005a; Stanford et al. 2006) studies have obtained metallicity estimates for a large number of stars (N500 in some cases), there have only been a few high resolution spectroscopic studies with a significant number (N10) of stars observed (Norris & Da Costa 1995; Smith et al. 2000; Piotto et al. 2005; Sollima et al. 2006). However, aside from the present study, Norris & Da Costa (1995) still represents the largest (N=40) single high resolution analysis of Cen RGB stars. The general results from the metallicity studies can be summarized as: (1) few stars exist at [Fe/H]–2.0, (2) a primary peak in the metallicity distribution is observed at [Fe/H]–1.8 to –1.6, (3) there is a long tail of increasing metallicity up to [Fe/H]–0.5, and (4) there appear to be multiple peaks in the distribution at various [Fe/H] values.
In Figure 10, we present a histogram of our derived metallicity distribution function for all 180 stars. We find in agreement with previous studies that there are at least four distinct populations with the most metal–poor having [Fe/H]–1.75, the two intermediate metallicity populations have [Fe/H]–1.45 and –1.05, and the most metal–rich population has [Fe/H]–0.75. While our observations are skewed towards observing more metal–poor stars (see Figure 2), there are intrinsically more metal–poor than metal–rich stars, as can be seen in Figure 1. This means our derived metallicity distribution is affected by both the actual distribution and observational selection effects. Given that we only observed one star on the most metal–rich branch, it is possible that stars with metallicities higher than [Fe/H]=–0.75 exist. However, since our observed completion fraction is significantly higher for the most metal–poor stars, it is likely that our observed distribution function is accurate in a relative sense such that the cluster was rapidly enriched from the primordial metallicity of [Fe/H]–2.15 to the first major epoch of star formation at [Fe/H]–1.75. The absence of stars more metal–poor than [Fe/H]–2.2 means the proto– Cen environment was already pre–enriched, perhaps from processes such as cloud–cloud collisions (Tsujimoto et al. 2003), when the primary metal–poor population formed. In contrast, field stars in the Galactic halo exhibit a wide range of metallicities from [Fe/H]0.0 to [Fe/H]–4.0 (e.g., Gratton et al. 2004), indicating that the two do not share a common chemical enrichment history.
The distribution shown in Figure 10 suggests that if Cen evolved as a single entity (i.e., without significant contributions from mergers), then there were four to five significant star formation episodes that occurred. This seems to fit the high resolution photometric data from Sollima et al. (2005a) and Villanova et al. (2007) that show the multiple giant branches appear in discrete groups instead of as a continuous distribution. This trend is similarly reproduced in Figure 11, where our derived metallicities are superimposed on the photometric data from van Leeuwen et al. (2000). Here, even when binning by the approximate 3 value of each peak in the distribution from Figure 10 (0.3 dex), the different metallicity groups can be separated. The metallicity distribution from Figure 10 is very well produced in the hydrodynamical chemical enrichment simulations of Marcolini et al. (2007), where they assumed Cen is the core remnant of a dwarf spheroidal galaxy that was captured and tidally stripped 10 Gyr ago with star formation occurring over roughly 1.5 Gyr. The simulated metallicity peaks from Marcolini et al. (2007) lie at [Fe/H]–1.6, –1.35, –1.0, and –0.70, which are very similar to ours at [Fe/H]=–1.75, –1.45, –1.05, and –0.75.
There is some evidence that different metallicity populations may be spatially and kinematically unique (Norris et al. 1996; 1997; Suntzeff & Kraft 1996; Hilker & Richtler 2000; Pancino et al. 2000; 2003). In Figure 12, we present Fe and Al abundances as a function of distance from the cluster center. Keeping in mind our observational bias, we find a marginal tendency for the more metal–rich stars to be located in the inner regions of the cluster while the more metal–poor stars are rather evenly distributed at all radii sampled here. However, given our small sample size in the metal–rich regime, we are unable to make any definitive arguments for or against a metallicity–radius relationship. It should be noted though that Ikuta & Arimoto (2000) and Rey et al. (2004) do not find any strong evidence for the metal–poor and metal–rich populations having a spatially different structure. Even though the relaxation time for Cen is thought to exceed 5 Gyr (Djorgovski 1993; Merritt et al. 1997), any correlation between projected spatial position and metallicity is apparently subtle. However, it has been pointed out in deep photometric surveys (e.g., Rey et al. 2004) that the most metal–rich RGB–a is predominately seen in CMDs of the inner region of the cluster.
The main result indicating that at least the most metal–rich population may have a different formation history is that those stars appear to have a lower velocity dispersion (i.e. are kinematically cooler) than the other populations and do not show signs of rotation (Norris et al. 1997). In Figure 13 we show our derived radial velocities plotted both as a function of log (Fe)666log (X)=log(N/N)+12 and log (Al), where the error bars indicate the velocity dispersion in the data. To within one standard deviation, we do not find significant evidence for any of the stellar populations having a different bulk radial velocity or velocity dispersion. It seems unlikely that a larger sample size would provide significantly different results because Reijns et al. (2006) determined radial velocities for nearly 2000 Cen members and concluded the RGB–a stars had radial velocity and dispersion values consistent with the entire cluster. Pancino et al. (2007) have shown the rotational velocities for all populations are comparable to one another, but interestingly they find an underlying sinusoidal pattern in their measured velocities as a function of position angle. However, the metal–poor, intermediate metallicity, and anomalous giant branches all show the same sinusoidal pattern. Whether any true kinematic anomaly exists for this cluster or not remains to be seen.
5.2 Al Abundances
The bulk of aluminum production in galaxies and globular clusters is thought to arise from quiescent carbon and neon burning in massive stars (M8 M) and HBB occurring in the envelopes of IM–AGB stars via the MgAl cycle (e.g., Arnett & Truran 1969; Arnett 1971). In most Galactic globular clusters, there is a very small (0.10 dex) spread in the abundance of heavy and Fe–peak elements, with a somewhat larger spread (0.3–0.6 dex) in s– and r–process elements (e.g., Sneden et al. 2000). However, the lighter elements carbon through aluminum are typically not uniform and in some cases show star–to–star variations of more than a factor of 10. While Cen does not share all of the same chemical characteristics as globular clusters, the primary production locations of each element should be similar to globular clusters and/or the Galactic halo. The lesson learned from the monometallicity of “normal” globular clusters is that however Al manifests itself onto the surface of stars, the process must not alter Fe–peak, s–process, or r–process abundance ratios. This means that the often large star–to–star variation of [Al/Fe] seen in globular clusters (but not in halo field stars) are not due to supernova yields or the s–process, leaving either in situ deep mixing or HBB as the possible sites for [Al/Fe] variation. With these two scenarios in mind, we explore Al abundances with the goal of helping to constrain the source of Al variation and chemical evolution in Cen.
While the literature on Fe abundances for both evolved and main sequence stars is quite extensive, the spectroscopic surveys by Norris & Da Costa (1995) and Smith et al. (2000) represent the only studies to consider light element abundances that include Al for a large (N10) number of RGB stars in Cen. The results of those two studies indicate that the full range of [Al/Fe] is larger than 1.0 dex, Al and Na are correlated, Al and O are anticorrelated, and there is a hint of a decrease in [Al/Fe] with increasing [Fe/H]. We present the results of our larger sample plotting [Al/Fe] as a function of [Fe/H] in Figure 14. Even for the lowest metallicity stars, a large range in [Al/Fe] of 0.70 dex is already present. Near the first metallicity peak at [Fe/H]=–1.75, where it is assumed the first episode of star formation after the initial enrichment period occurred, the full range in [Al/Fe] reaches a maximum value of 1.3 dex. This star–to–star variation remains mostly constant until about [Fe/H]=–1.4, where the variation begins to decrease smoothly with increasing [Fe/H]. Interestingly, the “floor” Al abundance remains mostly constant at [Al/Fe]+0.15, regardless of the star’s metallicity; a characteristic shared with many globular clusters of various metallicity and in agreement with [Al/Fe] values typical of Galactic halo stars in Cen’s metallicity regime.
In Figure 15, we overlay a boxplot on top of the underlying distribution from Figure 14. The median [Al/Fe] ratio typically resides between about 0.45 and 0.80 dex for all well–sampled metallicities, with a relatively constant interquartile range. This implies that the average amount of Al in the cluster must increase with increasing Fe abundance, at least up to [Fe/H]–1.4. This result is confirmed in Figure 16, where log (Al) is plotted against log (Fe). It appears that for metallicities higher than about log (Fe)=6.0 ([Fe/H]–1.50), log (Al) no longer increases beyond log (Al)6.40 and the star–to–star scatter decreases. This result is likely robust against our observational bias because all stars observed in the metal–rich regime are located at or near the RGB tip (see Figure 1), where it is believed any Al enhancements due to deep mixing should be the most apparent. However, no obvious trend is seen between Al abundance and evolutionary state.
As discussed previously, there is some evidence for a correlation between Fe abundance and distance from the cluster center and we show the results from this study in the bottom panel of Figure 12. In the top panel of Figure 12, we present the same data but for Al instead of Fe. While there may be a tendency for the most metal–rich stars to be located inwards of about 10–15, there is no evidence of a trend for Al. Instead, stars of varying Al abundance are uniformly spread throughout the entire region sampled, at least out to 20. Likewise, the top panel of Figure 13 shows average radial velocities for Al abundances in 0.10 dex bins. To within uncertainties, there appears to be no trend in either radial velocity or velocity dispersion with log (Al). The fact that we do not find any preference of Al abundance or star–to–star dispersion with distance from the cluster center or radial velocity suggests star formation occurred on timescales shorter than those required to uniformly mix the gas.
5.3 Possible Implications on Chemical Evolution
From our available spectroscopic data for 180 RGB stars, we have confirmed the existence of at least four stellar populations ranging in metallicity from –2.2[Fe/H]–0.70, in agreement with previous photometric, low resolution spectroscopic, and smaller sample high resolution spectroscopic studies. Additionally, we have determined [Al/Fe] abundances for about 165 giants, most of which for the first time, with a sample larger by more than a factor of four than what was previously available in the literature. We find a constant Al abundance floor of [Al/Fe]+0.15 present at all metallicities, but with a largely varying and metallicity dependent spread above the floor. The star–to–star variation reaches a maximum extent in the intermediate metallicity regime, which is consistent with the second peak in the metallicity distribution, and begins to decline at higher metallicities. The floor itself is consistent with observations of field stars and is predicted by Galactic chemical evolution models, but the large [Al/Fe] variations are not predicted. Observations of some Galactic globular cluster stars, especially more metal–poor than [Fe/H]–1.5, show similar large star–to–star variations in [Al/Fe]. Combining our determined Fe and Al abundances with those available in the literature for these and other elements now allows us to examine each metallicity regime in turn.
5.3.1 The Metal–Poor Population
A prominent feature of the metal–poor stars ([Fe/H]–1.6) in Cen is the rapidly increasing abundances of Na, Al, and light and heavy s–process elements relative to Fe as the metallicity increases from [Fe/H]=–2.2 to the first metallicity peak at [Fe/H]=–1.75 (e.g., Norris & Da Costa 1995; Smith et al. 2000). These increases are accompanied by nearly constant heavy [/Fe]+0.30, low Cu abundances ([Cu/Fe]–0.60), and low r–process abundances ([Eu/Fe]–0.50). These results seem to indicate that massive stars exploding as type II SNe are the primary contributors for Fe–peak and heavy –element enhancement in the cluster, but the low Eu abundances, which should be synthesized in the same stars, are puzzling. Additionally, the growing s–process component appears to be best fit by models of 1.5–3 M AGB ejecta (Smith et al. 2000). The lack of clear evidence for type Ia SNe having contributed to the chemical composition of metal–poor stars in Cen (e.g., Smith et al. 2000; Cunha et al. 2002; Pancino et al. 2002; Platais et al. 2003) is consistent with the 1 Gyr timescales needed for type Ia SNe to evolve and the fact that they might not efficiently form in metal–poor environments (Kobayashi et al. 1998).
As mentioned above, the majority of Al present in the atmospheres of these RGB stars was likely produced in type II SNe explosions that polluted the pristine gas from which these stars formed. While the heavy element data do not support high mass (8M) stars being the source for the more than 1.0 dex [Al/Fe] variations, that may be explained from HBB occurring in IM–AGB stars, in situ deep mixing, or a hybrid scenario. In Figures 14–16, we have shown that [Al/Fe]0 for all metal–poor stars sampled, but a constant Al abundance floor is setup at [Al/Fe]+0.15 with a rapidly increasing star–to–star dispersion that reaches about 1.3 dex in extent by [Fe/H]=–1.75. For the neutron capture elements, which are the only other group exhibiting a variations with metallicity, Smith et al. (2000) showed stars with [Fe/H]–2 are dominated by an r–process component with a shift to a primarily s–process component by [Fe/H]–1.8.
In the pure pollution scenario, which does not invoke deep mixing affecting elements heavier than N, type II SNe, low and IM–AGB stars, and perhaps winds from less evolved very massive stars (e.g., Maeder & Meynet 2006) are responsible for all abundance anomalies. Adding our large Al data set to the sample of stars previously observed may help constrain enrichment timescales and polluting AGB masses. Conventional theory suggests light and s–process elements do not share the same origin and Cen’s s–process component is best fit with lower mass AGB stars, but masses lower than 3–4 M undergo third dredgeup without significant HBB (e.g., Karakas & Lattanzio 2007) and thus should not appreciably alter their envelope Al abundances. Additionally, Ventura & D’Antona (2007) suggest globular cluster light element anomalies can only be explained with ejecta from AGB stars in the mass range of 5–6.5 M. While our sample only includes two stars with [Fe/H]–2 (36036 & 51091), the elevated [Al/Fe] ratios of +0.40 and +1.13 suggest IM–AGB stars, with lifetimes of about 50–15010 yrs (Schaller et al. 1992), have already polluted the Cen system. In this case, the low metallicity environment would favor high [Al/Fe] yields from HBB processes occurring in IM–AGB stars. The rapidly rising average value of log (Al) shown in Figure 16 in the metallicity regime –2.0[Fe/H]–1.6 implies a continued contribution from IM–AGB stars, presumably forming from the same star formation event that creates the first peak in the metallicity distribution. The top two panels of Figure 17 show binned [Al/Fe] for this metallicity regime and we note approximately four sub–populations with [Al/Fe]+0.15, +0.45, +0.85, and +1.05. Predicted yields from type II SNe (e.g., Woosley & Weaver 1995) and measurements of field stars (e.g., Fulbright 2000) suggest type II SNe should enrich the ISM with [Al/Fe]+0.10 to +0.30 while 5–6.5 M AGB stars should produce [Al/Fe]+0.50 to +1.10 (e.g., D’Antona & Ventura 2007), which could explain our observed distribution. Given the rather short lifetimes of stars believed to produce Al and the fact that evidence for 1.5–3.0 M pollution does not appear until [Fe/H]–1.8, it would seem that Cen was probably enriched from [Fe/H]=–2.2 to –1.75 in 0.5–1.0 Gyr.
5.3.2 The Intermediate Metallicity Populations
For the two intermediate metallicity populations ([Fe/H]=–1.45 and [Fe/H]=–1.05), the heavy [/Fe] ratio remains constant and the s–process abundances level off with very little star–to–star dispersion (Norris & Da Costa 1995; Smith et al. 2000). As in the most metal–poor stars, r–process and Cu ratios relative to Fe remain low and mostly unchanged. However, the star–to–star scatter in O, Na, and Al is still quite large. It is interesting to point out that log (Al) reaches its maximum value at about the same metallicity at which the s–process elements reach a constant ratio relative to Fe. The [Al/Fe] abundance floor is constant throughout this metallicity regime at [Al/Fe]+0.15, which means the scatter, still considerably larger than for [Ba/Fe], decreases as a function of increasing metallicity. This trend should presumably be present for Na and in the opposite sense for O assuming the Na–Al correlation and O–Al anticorrelation exist at all metallicities.
Had the scatter in Al abundances been comparable to that of other heavier elements in this metallicity range (0.10–0.30 dex) with a nearly constant [Al/Fe] ratio, as is seen in field stars, we might be inclined to believe Al enhancement in the cluster was due solely to production in massive stars and that typical type II SNe ejecta have [Al/Fe]+0.15. It is interesting to note that the [Al/Fe] floor tracks closely (with a slight offset of 0.2-0.3 dex) to the Galactic chemical evolution model presented in Timmes et al. (1995; their Figure 19), assuming the amount of Fe ejected is decreased by a factor of two, and Samland (1998; their Figure 10), with an increase in secondary (i.e., metal–dependent) Al production by a factor of five. If the well–known light element correlations/anticorrelations seen in previously observed Cen stars (e.g., Norris & Da Costa 1995) holds at all metallicities and for all stars, those with [Al/Fe]+0.15 should also have [O/Fe]+0.30, heavy [/Fe]+0.30, and [Na/Fe]–0.20, which are consistent with predicted yields from type II SNe (e.g., Woosley & Weaver 1995). It could be that these stars formed preferentially out of SNe ejecta without significant IM–AGB contamination.
While the maximum observed log (Al) increases with metallicity for the most metal–poor Cen giants, this trend halts at [Fe/H]–1.4, which coincides with the second peak in the metallicity distribution (i.e., the next round of star formation). We know the heavy [/Fe], [Ba/Fe], and floor [Al/Fe] ratios remain constant at higher metallicities, indicating an increase in log (Ba), log (), and the minimum log (Al) that track with Fe. The question now posed by the Al data is why does the process producing the high Al values shut off or become less efficient at [Fe/H]–1.45? Increases in metallicity lead to lower temperatures at the bottom of the convective envelope and require higher masses for HBB to occur. It may be that we are observing the result of lower convective efficiency at higher metallicity and/or that fewer IM stars form in higher metallicity environment. IM–AGB models in the metallicity range of –1.5[Fe/H]–0.7 (e.g., Fenner et al. 2004; Ventura & D’Antona 2007; 2008) predict [Al/Fe] yields of +0.5 to +1.0, with lower [Al/Fe] yields at higher [Fe/H]. This may explain the bimodal distribution in the bottom panels of Figure 17, with the abundances in between possibly being due to varying degrees of ejecta dilution. The fact that the metallicity at which the heavy elements cease to increase in abundance more quickly than Fe and the metallicity where the maximum [Al/Fe] begins to decrease coincide suggests an important parameter changed in Cen at this point in its evolution. It may even be the case that this is when the progenitor dwarf galaxy began to change structurally via encounters with the Galactic disk. It appears that at metallicities higher than [Fe/H]=–1.45, the cluster slowly approaches a constant [Al/Fe], which is consistent with values observed in the halo.
While type Ia ejecta have been mostly ruled out by previous studies as contributors to the most metal–poor population, the metallicity at which they become important contributors is unclear. Marcolini et al. (2007) claim that their intermediate metallicity peak at [Fe/H]–1.4 is due primarily to inhomogeneous pollution by type Ia SNe. It is interesting to note that in this same metallicity bin we find a median [Al/Fe] value about 0.40 dex lower than the two surrounding bins as well as the only star with [Al/Fe]+0.15. It is uncertain whether this is a real effect or simply due to an anomalous selection of stars. Inhomogeneous pollution by type Ia SNe may also explain the bimodal distribution seen in the bottom panels of Figure 17 where stars polluted by both type Ia ejecta and IM–AGB stars exhibit lower [Al/Fe] ratios and “normal” stars polluted by type II SNe and IM–AGB stars have higher [Al/Fe] values. While the same trend is not particularly apparent for s–process elements (e.g., Smith et al. 2000), this may be due to a smaller sample size, especially if inhomogeneous pollution only affected a small percentage of intermediate metallicity stars; however, this could explain the few observations in the literature of stars with [Fe/H]–1.4 and [Ba/Fe]0 (e.g., Smith et al. 1995).
5.3.3 The Metal–Rich Population
For stars more metal–rich than [Fe/H]–1, there is some evidence of a decrease in [/Fe] and an increase in [Cu/Fe] (Pancino et al. 2002; but see also Cunha et al. 2002), which, if true, likely indicates an increased contribution from type Ia SNe. Similarly, there appears to be a decrease in [Eu/Fe] with perhaps a similar decrease in the abundance of s–process elements relative to Fe (Norris & Da Costa; Smith et al. 2000). Although the Al data are rather incomplete in this metallicity regime, the general trends seen in slightly more metal–poor stars appear to continue.
While the scope of an age spread amongst the various metallicity populations is still unknown, the Al data presented here seem to indicate that the age difference between the intermediate and metal–rich populations is not especially large. In particular, stars with the largest values of log (Al) appear with [Fe/H] ranging from –1.5 to –0.7, perhaps indicating that they formed from gas polluted by the same generation of IM–AGB ejecta. In this scenario, the lower [Al/Fe] ratios at high metallicity might be due to those stars forming in regions where [Fe/H] increased due to inhomogeneous pollution by type Ia SNe, as mentioned in Marcolini et al. (2007). In their scenario, this effect should be more important for the inner regions of the cluster. This may be corroborated by our finding that there is no apparent relationship between log (Al) and distance from the cluster center, but a trend might be present for Fe such that stars with [Fe/H]–1 are preferentially located closer to the cluster center. In any case, additional data are required in this metallicity regime to determine whether the decreasing [Al/Fe] ratios are a real effect or the result of incomplete statistics. It will be interesting to see if O and Na display similar behavior to Al as a function of [Fe/H].
We have determined radial velocities, Fe, and Al abundances for 180 RGB stars in the Galactic globular cluster Cen using moderate resolution (R13,000) spectroscopy. The bulk of our sample includes stars with V14.0, but an observational bias is present such that we preferentially observed more luminous and more metal–poor stars. The spectra ranged from 6500–6750 Å and Fe abundances were based on an average of approximately 10–20 Fe I lines. Al abundances were determined using either spectrum synthesis or equivalent width analyses of the 6696, 6698 Å Al I doublet, with synthesis being reserved for CN–strong and/or metal–rich stars.
With respect to our determined Fe abundances, we find in agreement with previous studies that at least four or more different metallicity populations are present in the cluster. Peaks in the metallicity distribution function appear at [Fe/H]=–1.75, –1.45, –1.05, and –0.75, indicating the presence of multiple star formation episodes. We do not find evidence suggesting any of the different metallicity populations are kinematically or spatially unique, but it should be noted that our observed completion fraction is low for stars more metal–rich than [Fe/H]–1.0 and we only observed stars between about 2 and 20 from the cluster center.
Our Al data corroborate the Fe results such that there does not appear to be any correlation between Al abundance and distance from the cluster center or radial velocity. This suggests that the cluster gas was not significantly mixed while star formation was still occurring. In a plot of [Al/Fe] versus [Fe/H], the data reveal a star–to–star variation of nearly 1.3 dex that stays mostly constant until [Fe/H]–1.45, in which case the spread in [Al/Fe] declines monotonically with increasing [Fe/H]. Additionally, the [Al/Fe] floor remains nearly constant across all metallicities sampled here at [Al/Fe]+0.15. This result is similar to what is predicted based on type II SNe yields and closely mimics the trend seen in Galactic halo field stars. The anomalously low median [Al/Fe] ratio at [Fe/H]=–1.45 may be evidence for inhomogeneous pollution from type Ia SNe and could explain the bimodal [Al/Fe] distribution seen in intermediate metallicity stars, but more observations are required to confirm whether this is real or the result of an incomplete sample.
The source of the [Al/Fe] spread that has also been observed in other light elements remains an open problem, but the results obtained here pose some interesting questions. A plot of log (Al) versus log (Fe) shows that log (Al) no longer increases beyond about 6.40 at metallicities higher than [Fe/H]–1.45, which is coincident with the second peak in the metallicity distribution function. Apparently, whatever process is responsible for manifesting very high Al abundances shuts down or becomes less efficient at intermediate and high metallicities. In “normal” metal–poor globular clusters, the large star–to–star variations seen in the light elements are not shared by Fe–peak and neutron capture elements, and it has been suggested that HBB occurring in IM–AGB stars or in situ deep mixing are responsible for the light element abundance anomalies. Without a comparable sample of O and Na data to supplement the Al abundances here, it is difficult to determine the role either source plays. However, AGB yields of stars undergoing HBB indicate stars forming from material polluted by AGB ejecta can only reach [Al/Fe] ratios between about +0.5 and +1.0, with perhaps slightly lower and higher values being reached in higher and lower metallicity environments, respectively.
It may be possible to explain the Al data such that core–collapse SNe drive the [Al/Fe] floor and an AGB mass spectrum with varying HBB efficiencies and mixing depths are responsible for much of the additional scatter present. The decrease in the maximum [Al/Fe] with increasing [Fe/H] might then be attributed to requiring higher mass stars for HBB to occur at temperatures adequate to activate the full Mg to Al cycle, which means the burning material is exposed for a shorter amount of time and thus leads to less [Al/Fe] enhancement. Whether this can be made to work quantitatively in light of the problems associated with AGB pollution scenarios (see 1) remains to be seen.
- Alonso et al. (1994) Alonso, A., Arribas, S., & Martinez-Roger, C. 1994, A&A, 282, 684
- Alonso et al. (1998) Alonso, A., Arribas, S., & Martinez-Roger, C. 1998, A&AS, 131, 209
- Alonso et al. (1999) Alonso, A., Arribas, S., & Martínez-Roger, C. 1999, A&AS, 140, 261
- Alonso et al. (2001) Alonso, A., Arribas, S., & Martínez-Roger, C. 2001, A&A, 376, 1039
- Anders & Grevesse (1989) Anders, E., & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53, 197
- Anderson (1997) Anderson, J. 1997, Ph.D. thesis, Univ. California at Berkeley
- Arnett (1971) Arnett, W. D. 1971, ApJ, 166, 153
- Arnett & Truran (1969) Arnett, W. D., & Truran, J. W. 1969, ApJ, 157, 339
- Bedin et al. (2004) Bedin, L. R., Piotto, G., Anderson, J., Cassisi, S., King, I. R., Momany, Y., & Carraro, G. 2004, ApJ, 605, L125
- Bekki & Norris (2006) Bekki, K., & Norris, J. E. 2006, ApJ, 637, L109
- Bell et al. (1979) Bell, R. A., Dickens, R. J., & Gustafsson, B. 1979, ApJ, 229, 604
- Bellman et al. (2001) Bellman, S., Briley, M. M., Smith, G. H., & Claver, C. F. 2001, PASP, 113, 326
- Blackwell & Shallis (1977) Blackwell, D. E., & Shallis, M. J. 1977, MNRAS, 180, 177
- Boesgaard et al. (2005) Boesgaard, A. M., King, J. R., Cody, A. M., Stephens, A., & Deliyannis, C. P. 2005, ApJ, 629, 832
- Briley et al. (2004a) Briley, M. M., Cohen, J. G., & Stetson, P. B. 2004a, AJ, 127, 1579
- Briley et al. (2004b) Briley, M. M., Harbeck, D., Smith, G. H., & Grebel, E. K. 2004b, AJ, 127, 1588
- Brown & Wallerstein (1993) Brown, J. A., & Wallerstein, G. 1993, AJ, 106, 133
- Calamida et al. (2005) Calamida, A., et al. 2005, ApJ, 634, L69
- Cannon et al. (1998) Cannon, R. D., Croke, B. F. W., Bell, R. A., Hesser, J. E., & Stathakis, R. A. 1998, MNRAS, 298, 601
- Carbon et al. (1982) Carbon, D. F., Romanishin, W., Langer, G. E., Butler, D., Kemper, E., Trefzger, C. F., Kraft, R. P., & Suntzeff, N. B. 1982, ApJS, 49, 207
- Carpenter (2001) Carpenter, J. M. 2001, AJ, 121, 2851
- Castelli et al. (1997) Castelli, F., Gratton, R. G., & Kurucz, R. L. 1997, A&A, 318, 841
- Charbonnel & Do Nascimento (1998) Charbonnel, C., & Do Nascimento, J. D., Jr. 1998, A&A, 336, 915
- Charbonnel & Zahn (2007) Charbonnel, C., & Zahn, J.-P. 2007, A&A, 467, L15
- Cohen et al. (2002) Cohen, J. G., Briley, M. M., & Stetson, P. B. 2002, AJ, 123, 2525
- Cohen & Meléndez (2005) Cohen, J. G., & Meléndez, J. 2005, AJ, 129, 303
- Cunha et al. (2002) Cunha, K., Smith, V. V., Suntzeff, N. B., Norris, J. E., Da Costa, G. S., & Plez, B. 2002, AJ, 124, 379
- D’Antona & Ventura (2007) D’Antona, F., & Ventura, P. 2007, MNRAS, 379, 1431
- Denisenkov & Denisenkova (1990) Denisenkov, P. A., & Denisenkova, S. N. 1990, Soviet Astronomy Letters, 16, 275
- Denissenkov & Weiss (2001) Denissenkov, P. A., & Weiss, A. 2001, ApJ, 559, L115
- Denissenkov & Herwig (2003) Denissenkov, P. A., & Herwig, F. 2003, ApJ, 590, L99
- Dickens et al. (1991) Dickens, R. J., Croke, B. F. W., Cannon, R. D., & Bell, R. A. 1991, Nature, 351, 212
- Dinescu et al. (1999) Dinescu, D. I., Girard, T. M., & van Altena, W. F. 1999, AJ, 117, 1792
- Djorgovski et al. (1997) Djorgovski S., 1993, ASP Conf. Ser. Vol. 50, Structure and Dynamics of Globular Clusters. Astron. Soc. Pac., San Francisco, p. 373
- Fenner et al. (2004) Fenner, Y., Campbell, S., Karakas, A. I., Lattanzio, J. C., & Gibson, B. K. 2004, MNRAS, 353, 789
- Ferraro et al. (2004) Ferraro, F. R., Sollima, A., Pancino, E., Bellazzini, M., Straniero, O., Origlia, L., & Cool, A. M. 2004, ApJ, 603, L81
- Fulbright (2002) Fulbright, J. P. 2002, AJ, 123, 404
- Gnedin et al. (2002) Gnedin, O. Y., Zhao, H., Pringle, J. E., Fall, S. M., Livio, M., & Meylan, G. 2002, ApJ, 568, L23
- Gratton et al. (2000) Gratton, R. G., Sneden, C., Carretta, E., & Bragaglia, A. 2000, A&A, 354, 169
- Gratton et al. (2001) Gratton, R. G., et al. 2001, A&A, 369, 87
- Gratton et al. (2004) Gratton, R., Sneden, C., & Carretta, E. 2004, ARA&A, 42, 385
- Harris (1996) Harris, W. E. 1996, AJ, 112, 1487
- Hilker & Richtler (2000) Hilker, M., & Richtler, T. 2000, A&A, 362, 895
- Hilker et al. (2004) Hilker, M., Kayser, A., Richtler, T., & Willemsen, P. 2004, A&A, 422, L9
- Hill et al. (2002) Hill, V., et al. 2002, A&A, 387, 560
- Hughes & Wallerstein (2000) Hughes, J., & Wallerstein, G. 2000, AJ, 119, 1225
- Hughes et al. (2004) Hughes, J., Wallerstein, G., van Leeuwen, F., & Hilker, M. 2004, AJ, 127, 980
- Ikuta & Arimoto (2000) Ikuta, C., & Arimoto, N. 2000, A&A, 358, 535
- Ivans et al. (1999) Ivans, I. I., Sneden, C., Kraft, R. P., Suntzeff, N. B., Smith, V. V., Langer, G. E., & Fulbright, J. P. 1999, AJ, 118, 1273
- Ivans et al. (2001) Ivans, I. I., Kraft, R. P., Sneden, C., Smith, G. H., Rich, R. M., & Shetrone, M. 2001, AJ, 122, 1438
- Johnson (1965) Johnson, H. L. 1965, ApJ, 141, 923
- Johnson et al. (2005) Johnson, C. I., Kraft, R. P., Pilachowski, C. A., Sneden, C., Ivans, I. I., & Benman, G. 2005, PASP, 117, 1308
- Johnson & Pilachowski (2006) Johnson, C. I., & Pilachowski, C. A. 2006, AJ, 132, 2346
- Karakas & Lattanzio (2007) Karakas, A., & Lattanzio, J. C. 2007, Publications of the Astronomical Society of Australia, 24, 103
- Kayser et al. (2006) Kayser, A., Hilker, M., Richtler, T., & Willemsen, P. G. 2006, A&A, 458, 777
- Keller et al. (2001) Keller, L. D., Pilachowski, C. A., & Sneden, C. 2001, AJ, 122, 2554
- Kobayashi et al. (1998) Kobayashi, C., Tsujimoto, T., Nomoto, K., Hachisu, I., & Kato, M. 1998, ApJ, 503, L155
- Langer et al. (1986) Langer, G. E., Kraft, R. P., Carbon, D. F., Friel, E., & Oke, J. B. 1986, PASP, 98, 473
- Langer et al. (1997) Langer, G. E., Hoffman, R. E., & Zaidins, C. S. 1997, PASP, 109, 244
- Lattanzio et al. (2004) Lattanzio, J., Karakas, A., Campbell, S., Elliott, L., & Chieffi, A. 2004, Memorie della Societa Astronomica Italiana, 75, 322
- Lee et al. (1999) Lee, Y.-W., Joo, J.-M., Sohn, Y.-J., Rey, S.-C., Lee, H.-C., & Walker, A. R. 1999, Nature, 402, 55
- Lee et al. (2005) Lee, Y.-W., et al. 2005, ApJ, 621, L57
- Maeder & Meynet (2006) Maeder, A., & Meynet, G. 2006, A&A, 448, L37
- Marcolini et al. (2007) Marcolini, A., Sollima, A., D’Ercole, A., Gibson, B. K., & Ferraro, F. R. 2007, MNRAS, 382, 443
- Mayor et al. (1997) Mayor, M., et al. 1997, AJ, 114, 1087
- Merritt et al. (1997) Merritt, D., Meylan, G., & Mayor, M. 1997, AJ, 114, 1074
- Meylan et al. (1995) Meylan, G., Mayor, M., Duquennoy, A., & Dubath, P. 1995, A&A, 303, 761
- Norris & Da Costa (1995) Norris, J. E., & Da Costa, G. S. 1995, ApJ, 447, 680
- Norris et al. (1996) Norris, J. E., Freeman, K. C., & Mighell, K. J. 1996, ApJ, 462, 241
- Norris et al. (1997) Norris, J. E., Freeman, K. C., Mayor, M., & Seitzer, P. 1997, ApJ, 487, L187
- Norris (2004) Norris, J. E. 2004, ApJ, 612, L25
- Origlia et al. (2003) Origlia, L., Ferraro, F. R., Bellazzini, M., & Pancino, E. 2003, ApJ, 591, 916
- Pancino et al. (2000) Pancino, E., Ferraro, F. R., Bellazzini, M., Piotto, G., & Zoccali, M. 2000, ApJ, 534, L83
- Pancino et al. (2002) Pancino, E., Pasquini, L., Hill, V., Ferraro, F. R., & Bellazzini, M. 2002, ApJ, 568, L101
- Pancino et al. (2003) Pancino, E., Seleznev, A., Ferraro, F. R., Bellazzini, M., & Piotto, G. 2003, MNRAS, 345, 683
- Pancino et al. (2007) Pancino, E., Galfo, A., Ferraro, F. R., & Bellazzini, M. 2007, ApJ, 661, L155
- Pilachowski (1988) Pilachowski, C. A. 1988, ApJ, 326, L57
- Pilachowski et al. (1996) Pilachowski, C. A., Sneden, C., & Kraft, R. P. 1996, AJ, 111, 1689
- Piotto et al. (2005) Piotto, G., et al. 2005, ApJ, 621, 777
- Platais et al. (2003) Platais, I., Wyse, R. F. G., Hebb, L., Lee, Y.-W., & Rey, S.-C. 2003, ApJ, 591, L127
- Press et al. (1992) Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. R. 1992, Numerical Recipes in FORTRAN: The Art of Scientific Computing (2nd ed.; Cambridge: Cambridge Univ. Press)
- Reijns et al. (2006) Reijns, R. A., Seitzer, P., Arnold, R., Freeman, K. C., Ingerson, T., van den Bosch, R. C. E., van de Ven, G., & de Zeeuw, P. T. 2006, A&A, 445, 503
- Rey et al. (2004) Rey, S.-C., Lee, Y.-W., Ree, C. H., Joo, J.-M., Sohn, Y.-J., & Walker, A. R. 2004, AJ, 127, 958
- Richer et al. (1991) Richer, H. B., Fahlman, G. G., Buonanno, R., Fusi Pecci, F., Searle, L., & Thompson, I. B. 1991, ApJ, 381, 147
- Ryan et al. (1996) Ryan, S. G., Norris, J. E., & Beers, T. C. 1996, ApJ, 471, 254
- Salaris et al. (2002) Salaris, M., Cassisi, S., & Weiss, A. 2002, PASP, 114, 375
- Samland (1998) Samland, M. 1998, ApJ, 496, 155
- Schaller et al. (1992) Schaller, G., Schaerer, D., Meynet, G., & Maeder, A. 1992, A&AS, 96, 269
- Shetrone (1996) Shetrone, M. D. 1996, AJ, 112, 2639
- Smith et al. (1995) Smith, V. V., Cunha, K., & Lambert, D. L. 1995, AJ, 110, 2827
- Smith et al. (1996) Smith, G. H., Shetrone, M. D., Bell, R. A., Churchill, C. W., & Briley, M. M. 1996, AJ, 112, 1511
- Smith et al. (2000) Smith, V. V., Suntzeff, N. B., Cunha, K., Gallino, R., Busso, M., Lambert, D. L., & Straniero, O. 2000, AJ, 119, 1239
- Smith et al. (2002) Smith, V. V., Terndrup, D. M., & Suntzeff, N. B. 2002, ApJ, 579, 832
- Smith et al. (2005) Smith, V. V., Cunha, K., Ivans, I. I., Lattanzio, J. C., Campbell, S., & Hinkle, K. H. 2005, ApJ, 633, 392
- Smith (2006) Smith, G. H. 2006, PASP, 118, 1225
- Sneden (1973) Sneden, C. 1973, ApJ, 184, 839
- Sneden et al. (1991) Sneden, C., Kraft, R. P., Prosser, C. F., & Langer, G. E. 1991, AJ, 102, 2001
- Sneden et al. (2000) Sneden, C., Pilachowski, C. A., & Kraft, R. P. 2000, AJ, 120, 1351
- Sneden et al. (2004) Sneden, C., Kraft, R. P., Guhathakurta, P., Peterson, R. C., & Fulbright, J. P. 2004, AJ, 127, 2162
- Sollima et al. (2005a) Sollima, A., Ferraro, F. R., Pancino, E., & Bellazzini, M. 2005a, MNRAS, 357, 265
- Sollima et al. (2005b) Sollima, A., Pancino, E., Ferraro, F. R., Bellazzini, M., Straniero, O., & Pasquini, L. 2005b, ApJ, 634, 332
- Sollima et al. (2006) Sollima, A., Borissova, J., Catelan, M., Smith, H. A., Minniti, D., Cacciari, C., & Ferraro, F. R. 2006, ApJ, 640, L43
- Stanford et al. (2006) Stanford, L. M., Da Costa, G. S., Norris, J. E., & Cannon, R. D. 2006, ApJ, 647, 1075
- Stanford et al. (2007) Stanford, L. M., Da Costa, G. S., Norris, J. E., & Cannon, R. D. 2007, ApJ, 667, 911
- Suntzeff & Kraft (1996) Suntzeff, N. B., & Kraft, R. P. 1996, AJ, 111, 1913
- Thevenin (1990) Thévenin, F. 1990, A&AS, 82, 179
- Timmes et al. (1995) Timmes, F. X., Woosley, S. E., & Weaver, T. A. 1995, ApJS, 98, 617
- Tsujimoto & Shigeyama (2003) Tsujimoto, T., & Shigeyama, T. 2003, ApJ, 590, 803
- van de Ven et al. (2006) van de Ven, G., van den Bosch, R. C. E., Verolme, E. K., & de Zeeuw, P. T. 2006, A&A, 445, 513
- van Leeuwen et al. (2000) van Leeuwen, F., Le Poole, R. S., Reijns, R. A., Freeman, K. C., & de Zeeuw, P. T. 2000, A&A, 360, 472
- van Loon et al. (2007) van Loon, J. T., van Leeuwen, F., Smalley, B., Smith, A. W., Lyons, N. A., McDonald, I., & Boyer, M. L. 2007, MNRAS, 382, 1353
- Ventura & D’Antona (2005a) Ventura, P., & D’Antona, F. 2005a, A&A, 431, 279
- Ventura & D’Antona (2005b) Ventura, P., & D’Antona, F. 2005b, ApJ, 635, L149
- Ventura & D’Antona (2008) Ventura, P., & D’Antona, F. 2008, A&A, 479, 805
- Ventura & D’Antona (2008) Ventura, P., & D’Antona, F. 2008, MNRAS, 255
- Villanova et al. (2007) Villanova, S., et al. 2007, ApJ, 663, 296
- Woolley (1966) Woolley, R. R. 1966, Royal Observatory Annals, 2, 1
- Woosley & Weaver (1995) Woosley, S. E., & Weaver, T. A. 1995, ApJS, 101, 181
- Yong et al. (2003) Yong, D., Grundahl, F., Lambert, D. L., Nissen, P. E., & Shetrone, M. D. 2003, A&A, 402, 985
- Zucker et al. (1996) Zucker, D., Wallerstein, G., & Brown, J. A. 1996, PASP, 108, 911
|1||6600||2003 July 17||1 x 1800|
|2||6600||2003 July 18||1 x 1800|
|2||6600||2003 July 18||4 x 2700|
|3||6600||2003 July 19||2 x 2700|
|3||6600||2003 July 19||1 x 3600|
|StaraaIdentifier from van Leeuwen et al. (2000).||Alt. IDbbIdentifier from Woolley (1966).||V||Error||from Mean||Mem. Prob.ccMembership probability from van Leeuwen et al. (2000).|
|[km s]||[km s]|
|Cluster Mean Values|
|StaraaIdentifier from van Leeuwen et al. (2000).||Alt. IDbbIdentifier from Woolley (1966).||V||B–V||V–K||M||T||log g||[Fe/H]||V|
|TCS||[K]||[cm s]||Spec.||[km s]|
|Element||Exc. Pot.||log gf|
|ccWavelengths are given in units of Å.||6516||6533||6546||6551||6574||6592||6593||6597||6608||6609||6625||6627||6633||6646||6648||6677||6696||6698||6703||6710||6726||6733||6739|
|9ddDesignation is from van Leeuwen et al. (2000).||63||20||151||159||130||34||40||100||51||14||15||36||170||23||28||49|
|StaraaThe designation “Syn” indicates a synthetic spectrum comparison method was used.||log (Fe)||[Fe/H]bbEquivalent widths are given in units of mÅ.||Num. Lines||/||log (Al)||[Al/Fe]ccAssumed the solar log (Al)=6.47 (Anders & Grevesse 1989).||Num. Lines||/|
|[K]||[cm s]||[km s]||[dex]|