Light, Alpha, and Fe–Peak Element Abundances in the Galactic Bulge
We present radial velocities and chemical abundances of O, Na, Mg, Al, Si, Ca, Cr, Fe, Co, Ni, and Cu for a sample of 156 red giant branch stars in two Galactic bulge fields centered near (l,b)=(5.25,–3.02) and (0,–12). The (5.25,–3.02) field also includes observations of the bulge globular cluster NGC 6553. The results are based on high resolution (R20,000), high signal–to–noise (S/N70) FLAMES–GIRAFFE spectra obtained through the ESO archive. However, we only selected a subset of the original observations that included spectra with both high S/N and that did not show strong TiO absorption bands. The present work extends previous analyses of this data set beyond Fe and the –elements Mg, Si, Ca, and Ti. While we find reasonable agreement with past work, the data presented here indicate that the bulge may exhibit a different chemical composition than the local thick disk, especially at [Fe/H]–0.5. In particular, the bulge [/Fe] ratios may remain enhanced to a slightly higher [Fe/H] than the thick disk and the Fe–peak elements Co, Ni, and Cu appear enhanced compared to the disk. There is also some evidence that the [Na/Fe] (but not [Al/Fe]) trends between the bulge and local disk may be different at low and high metallicity. We also find that the velocity dispersion decreases as a function of increasing [Fe/H] for both fields, and do not detect any significant cold, high velocity population. A comparison with chemical enrichment models indicates that a significant fraction of hypernovae are required to explain the bulge abundance trends, and that initial mass functions that are steep, top–heavy (and do not include strong outflow), or truncated to avoid including contributions from stars 40 M are ruled out, in particular because of disagreement with the Fe–peak abundance data. For most elements, the NGC 6553 stars exhibit nearly identical abundance trends to comparable metallicity bulge field stars. However, the star–to–star scatter and mean [Na/Fe] ratios appear higher in the cluster, perhaps indicating additional self–enrichment.
Subject headings:stars: abundances, Galactic bulge: general, bulge: Galaxy: bulge, stars: Population II
Understanding the formation and subsequent evolution of the Galactic bulge is important both for interpreting observations of extragalactic populations and for constraining Galaxy chemodynamical formation models. Recent large sample spectroscopic surveys, such as the Bulge Radial Velocity Assay (BRAVA; Rich et al. 2007a; Howard et al. 2008; Howard et al. 2009; Kunder et al. 2012), the Abundances and Radial velocity Galactic Origins Survey (ARGOS; Freeman et al. 2013; Ness et al. 2012; 2013b), the Apache Point Observatory Galactic Evolution Experiment (APOGEE; Majewski et al. 2010; Nidever et al. 2012), and the GIRAFFE Inner Bulge Survey (GIBS; Zoccali et al. 2014) provide a coherent view of the bulge as a barred, triaxial system exhibiting cylindrical rotation. Photometric and star count studies have also discovered a double red clump along some bulge sight lines that traces out an X–shaped structure (McWilliam & Zoccali 2010; Nataf et al. 2010; Saito et al. 2011). This structure appears to be dominated by stars with [Fe/H]–0.5 on bar–supporting orbits (Soto et al. 2007; Babusiaux et al. 2010; Ness et al. 2012; Uttenthaler et al. 2012; but see also Nataf et al. 2014).
Inclusive with these data are detailed composition analyses of field stars from
moderate and high resolution spectroscopy (McWilliam & Rich 1994;
Ramírez et al. 2000; Rich & Origlia 2005; Cunha & Smith 2006;
Fulbright et al. 2006; Zoccali et al. 2006; Fulbright et al. 2007; Lecureur et
al. 2007; Rich et al. 2007b; Cunha et al. 2008; Meléndez et al. 2008;
Zoccali et al. 2008; Alves–Brito et al. 2010; Bensby et al. 2010a; Ryde et al.
2010; Bensby et al. 2011; Gonzalez et al. 2011; Hill et al. 2011; Johnson et
al. 2011; Johnson et al. 2012; Rich et al. 2012; Uttenthaler et al. 2012;
Barbuy et al. 2013; Bensby et al. 2013; García Pérez et al. 2013;
Johnson et al. 2013a; Ness et al. 2013a; Jönsson et al. 2014) finding, at
least in a general sense, that the bulge is composed of stars spanning more
than a factor of 100 in [Fe/H]
While the observational data continue to grow, the difficult task of assembling the pieces into a fully self–consistent model of the bulge’s formation remains open. The chemodynamical bulge data are challenging to interpret. The bulge’s predominantly old age, enhanced [/Fe] ratios, vertical metallicity gradient, and the existence of possible “primordial building blocks” such as Terzan 5 (e.g., Ferraro et al 2009; Origlia et al. 2011; 2013) are more consistent with the classical, merger built formation scenario. However, the bulge’s boxy X–shape, similar composition characteristics to at least the thick disk, and cylindrical rotation profile suggest that the bulge formed via secular evolution from a buckling disk instability and may be a “pseudobulge” (e.g., Kormendy & Kennicutt 2004; but see also Zoccali et al. 2014). While Shen et al. (2010) rule out a classical bulge component that exceeds 8 of the disk mass, it may still be possible for a bar to form within a pre–existing classical bulge (e.g., Saha et al. 2012). Additionally, evidence such as the metallicity gradient may not be unique to the classical bulge scenario, and may be consistent with a secular evolution model in which a radial metallicity gradient in the buckling disk is transformed into a vertical gradient in the resultant bar (Martinez–Valpuesta & Gerhard 2013). The bulge may also be composed of at least two stellar populations with different composition and kinematics (Babusiaux et al. 2010; Hill et al. 2011; Bensby et al. 2011; Bensby et al. 2013; Ness et al. 2013a). However, at the moment the exact nature of these potentially distinct stellar populations is far from certain.
Although most of the chemical abundance work mentioned previously has focused on the [Fe/H] and [/Fe] ratios in comparison with the thin and thick disks, the light odd–Z and Fe–peak (and also neutron–capture) elements also provide discriminatory power between models and other stellar populations (e.g., see Kobayashi et al. 2011; their Figure 14). The Fe–peak elements in particular are useful as they may be sensitive to formation environment and metallicity. For example, the metallicity dependent yields and increased contributions of massive stars are predicted to produce enhanced [Cu/Fe] and [Zn/Fe] ratios in the bulge compared to the local disk. Similarly, if the bulge formed with a significantly flatter initial mass function (IMF) than the disk then bulge stars should exhibit very large [Co/Fe] and [Zn/Fe] ratios (Nomoto et al. 2013). Therefore, here we measure abundances of the Fe–peak elements Cr, Fe, Co, Ni, and Cu, in addition to the light odd–Z and –elements O, Na, Mg, Al, Si, and Ca, in 156 red giant branch (RGB) stars in two Galactic bulge fields at (l,b)=(5.25,–3.02) and (0,–12), and compare the abundance ratios with other bulge fields, the Galactic disk, and chemical enrichment models.
2. Observations, Target Selection, and Data Reduction
The FLAMES–GIRAFFE spectra for this project are based on data obtained from the ESO Science Archive Facility under request number 51251, which are based on observations collected at the European Southern Observatory, Paranal, Chile (ESO Program 073.B–0074). Details regarding the selection of targets and input parameters (e.g., photometry and astrometry) are given in Zoccali et al. (2008). To briefly summarize, fibers were placed on K giants approximately 1–2 magnitudes brighter than the bulge red clump, and the spectra were obtained in high resolution mode (R/20,000). The original program by Zoccali et al. (2008) included four fields centered at (l,b)=(1.14,–4.18), (0.21,–6.02), (0,–12), and (5.25,–3.02). While the (1.14,–4.18) and (0.21,–6.02) fields were observed in the HR 13, HR 14, and HR 15 setups (spanning 6100–6950 Å), the (5.25,–3.02) and (0,–12) fields were observed in the HR 11, HR 13, and HR 15 setups (5590–5835 Å; 6100–6400 Å; 6600–6950 Å). Since the HR 11 setup is the only one containing measurable copper lines, we have only analyzed GIRAFFE spectra from the (5.25,–3.02) and (0,–12) fields. We note that the (5.25,–3.02) field also includes the bulge globular cluster NGC 6553.
Figure 1 shows a 2MASS (Skrutskie et al. 2006) color–magnitude diagram of our final target selection from the archival data. The raw data set obtained from the ESO archive included observations of 205 RGB stars in the (5.25,–3.02) field and 109 RGB stars in the (0,–12) field. However, we only analyzed spectra for which the co–added signal–to–noise (S/N) ratio exceeded 70. We also discarded spectra that exhibited strong TiO absorption bands, for which a “standard” equivalent width (EW) analysis would be inappropriate. The final sample utilized here includes 75/205 stars (37) in the (5.25,–3.02) field and 81/109 stars (74) in the (0,–12) field. In Figure 1 we also identify stars that are likely members of NGC 6553 (see Section 3.5). In particular, note the broad dispersion in the color–magnitude diagram of cluster members, as well as with stars within 5 of the cluster center. This highlights the combined effects of differential reddening and population mixing along the line–of–sight toward the (5.25,–3.02) field. The star names and coordinates from the raw image headers and Zoccali et al. (2008), as well as available 2MASS photometry and star identifiers, are provided in Table 1.
The raw science and calibration data were downloaded and re–reduced using
the GIRAFFE Base–Line Data Reduction Software (girBLDRS)
3. Data Analysis
3.1. Model Stellar Atmospheres
The four primary model atmosphere input parameters of effective temperature
(T), surface gravity (log(g)), metallicity ([Fe/H]), and
microturbulence (vt) were determined via spectroscopic analyses. For stars in
the (0,–12) field, we used the model parameters given in Zoccali et al.
(2008) as a starting point before converging to a solution. However, the
adopted model atmosphere parameters for stars in the (5.25,–3.02) field
are not provided in Zoccali et al. (2008) nor Gonzalez et al. (2011).
Therefore, we adopted the generic values T=4500 K, log(g)=2.0
cgs, [Fe/H]=–0.20 dex, and vt=1.5 km s before converging to a
solution. The final parameters given in Table 1 were derived by enforcing Fe I
excitation equilibrium for T, ionization equilibrium between
Figure 2 shows our spectroscopically determined temperature and surface gravity values in comparison with the the spectroscopic T and photometric log(g) values given in Zoccali et al. (2008). As is evident in Figure 2, the spectroscopic determination of both parameters leads to a more extended distribution of surface gravities. This is likely due to the unavoidable problem that one must assume a distance (and mass) when deriving a photometric surface gravity. However, this is only a major issue when determining abundances of elements from transitions that are strongly sensitive to log(g). The model atmosphere parameters determined here are well–bounded by and follow the expected trends of the 10 Gyr isochrones with [Fe/H]=–1.5 (–enhanced) and [Fe/H]=0.5 ([/Fe]=0) shown in Figure 2.
We do note that 25/156 (16) stars in our sample converged to a solution
in which log(g)3–3.5. The derived higher gravity values suggest some of
these stars may be foreground lower RGB and subgiants rather than more
evolved bulge RGB stars. A better measurement of surface gravity, either from
the addition of more than 2–3 Fe II lines or the inclusion of more sensitive
atmospheric pressure indicators, would better constrain the true nature of
these stars. We do not find any strong systematic differences in
the derived [X/Fe] ratios between stars of “low” and “high”
In Figure 3 we compare the derived model atmosphere parameters between this study, Zoccali et al. (2008), and Gonzalez et al. (2011). We find good agreement in the derived T values with an average difference of only 2 K (=98 K). The dispersion of 100 K is reasonable given the different line lists and model atmospheres (but similar technique of excitation equilibrium) used. As mentioned previously, there is some discrepancy in log(g), especially for the highest gravity stars, between the present work and Zoccali et al. (2008). For stars with log(g)2.5, the average difference in log(g) is 0.01 dex (=0.29 dex), but for stars with log(g)2.5 the magnitude of the average gravity difference is 0.64 dex (=0.37 dex). Comparing the microturbulence values, which may be particularly sensitive to line choice and can vary as a function of gravity, we find an average difference of 0.18 km s (=0.27 km s).
When comparing derived [Fe/H] values, we find good agreement for [Fe/H]0.2 with an average difference of 0.03 dex (=0.13 dex). However, as is evident in Figure 3 our derived [Fe/H] values are systematically higher on average by 0.18 dex (=0.13 dex), for stars with [Fe/H]0.2. The source of this discrepancy may be related to the large 1 [Fe/H] uncertainties given in Zoccali et al. (2008) for stars with [Fe/H]0. This is illustrated in Figure 4 where we plot the 1 [Fe/H] uncertainties between our work and Zoccali et al. (2008) as a function of [Fe/H]. Ideally one expects to have measurement errors that are not correlated with metallicity, as is the case here. For the Zoccali et al. (2008) subsample in common with the present analysis, the line–to–line dispersions are comparable only for stars with [Fe/H]0.2.
3.2. Equivalent Width Abundance Determinations
The abundances of Fe I, Fe II, Si I, Ca I, Cr I, and Ni I were determined by measuring EWs via an interactive, semi–automatic code developed for this project. The measurement process followed the “standard” procedure of fitting single or multiple Gaussian profiles to the spectra for isolated and weakly blended lines, respectively. However, the measurement time frame was significantly reduced by implementing a simple machine learning algorithm that kept track of user input on a per–line basis to make an educated first guess for subsequent measurements in other stars of: the number of profiles to fit, profile fitting edges, and the central wavelength, width, and central depth of all associated nearby features. While all EW measurements were manually inspected, as was mentioned in Section 2 we selected stars from the archival data based primarily on S/N considerations in an effort to reduce measurement uncertainties. Sample spectra for stars of similar temperature but different metallicity are shown in Figure 5 to illustrate typical data quality in the three spectrograph setups.
The line lists for this project were created by visually examining the high S/N spectra of cool metal–poor and metal–rich giants in the sample, finding all isolated and/or weakly blended features for elements of interest, and merging the two line list sets. This was done to ensure that a roughly equivalent number of lines could be used in metal–rich and metal–poor stars, and the manual inspection of each fit enabled us to discard prohibitively strong and weak lines. On average, the Fe I, Fe II, Si I, Ca I, Cr I, and Ni I abundances were based on the measurement of 70, 2, 8, 6, 6, and 16 lines, respectively. The log(gf) values were set via an inverse abundance analysis relative to Arcturus. We adopted the Arcturus model atmosphere parameters from Fulbright et al. (2006). Similarly, for Fe, Si, and Ca we adopted the Arcturus abundances from Fulbright et al. (2006), and for Cr and Ni we adopted the Arcturus abundances from Ramírez & Allende Prieto (2011). The final line list, including the adopted Arcturus and derived solar abundances (based on measurements of the Hinkle et al. 2000 Arcturus and solar atlases), are provided in Table 2. The derived solar abundances for Fe, Si, Ca, and Cr agree within 0.05 dex of the values given in Asplund et al. (2009).
The final abundances of Fe I, Fe II, Si I, Ca I, Cr I, and Ni I, determined using the abfind driver of the LTE line analysis code MOOG (Sneden 1973; 2010 version), are given in Table 3. Note also that the [Fe/H] values given in Table 1 are the average of the [Fe I/H] and [Fe II/H] abundances given in Table 3. However, the average difference in the sense [Fe I/H]–[Fe II/H] is 0.00 dex with a small dispersion (=0.02 dex).
3.3. Spectrum Synthesis Abundance Determinations
For the element abundances derived from transitions involving a small number of lines that are affected by significant blends from prevalent spectral features, such as molecules and Ca I autoionization, and/or broadened due to isotopes and/or hyperfine structure, we used spectrum synthesis rather than EW analyses. For the present work this list includes [O I], Na I, Mg I, Al I, Co I, and Cu I. The abundances were determined using the parallelized version of the synth driver for MOOG (Johnson et al. 2012). For O, Na, Mg, and Al, we adopted as a reference point the Arcturus abundances given in Fulbright et al. (2006). However, as described below the reference Arcturus abundances for Co and Cu are based on measurements using the Kurucz (1994) and Cunha et al. (2002) hyperfine structure line lists.
The specific reasons for using synthesis are slightly different for each element given above. The 6300.30 Å [O I] line is blended with both a Sc II feature at 6300.69 Å and a Ni I feature at 6300.33 Å. Additionally, for most stars in this sample the oxygen abundance is sensitive to the molecular equilibrium calculations set by the carbon and nitrogen abundances as well. Using the CN line list from the Kurucz (1994) database, we iteratively solved for the O and CN abundances in each star. For sodium, the 6154.23 Na I line is relatively clean, but the 6160.75 Na I line is partially blended with two relatively strong Ca I lines. The three Mg I lines at 6319 Å are strongly affected by a broad Ca I autoionization feature, which we set by fitting the slope of the pseudo–continuum from 6316–6318 Å. The 6696.02 and 6698.67 Å Al I lines are both affected by CN, particularly in cooler and more metal–rich stars. Therefore, as with [O I] we simultaneously fit the Al I doublet and nearby CN features. The odd–Z isotope Co constitutes almost 100 of the cobalt abundance. While the 5647.23 and 6117.00 Å Co I lines are relatively weak (EW 50 mÅ), we included the hyperfine structure components from the Kurucz (1994) line list in our syntheses. For copper, which is dominated by the two odd–Z isotopes Cu and Cu, we assumed a solar system mixture of 69.17 and 30.83, respectively. We adopted the hyperfine line list of Cunha et al. (2002) and derive a similar solar abundance of log (Cu)=4.04 but a slightly lower Arcturus abundance than McWilliam et al. (2013). Although the 5782.11 Å Cu I line is strong in most of our stars, the hyperfine broadening helps desaturate the line profile to some extent.
In Figure 6 we show sample syntheses of the O, Mg, and Cu features for a typical metal–rich spectrum. We note that the 5782 Å Cu I line is also sometimes affected by a nearby diffuse interstellar band (DIB). The width and depth of the DIB feature was found to be highly variable. The level of contamination depends on the relative velocity between the interstellar cloud and the individual star and also the reddening value. Therefore, stars in the (5.25,–3.02) field, which have an average E(B–V)=0.7, were more strongly affected than those in the (0,–12) field, which have an average E(B–V)=0.2 (Zoccali et al. 2008). Most of the stars listed in Table 3 that do not have a [Cu/Fe] abundance listed were omitted because of obvious contamination by the DIB feature.
3.4. Radial Velocities
Radial velocities were measured using the XCSAO code (Kurtz & Mink 1998) for each individual exposure of every star and in all three filters. For reference templates we generated synthetic spectra ranging in temperature from 4250 to 5000 K (250 K steps), log(g) from 0.5 to 3.5 cgs (0.5 dex steps), [Fe/H] from –1.5 to 0.5 dex (0.5 dex steps), and vt from 1 to 2 km s (0.25 km s steps). Radial velocities were determined relative to the nearest template. We found the average agreement between exposures to be 0.15 km s (=0.13 km s). The heliocentric corrections were taken from the headers of the pipeline reduced files, and the heliocentric radial velocities (RV) listed in Table 1 represent the average value of all exposures and filters for each star.
The kinematic properties of the bulge have been extensively discussed in dedicated survey papers (e.g., Rich et al. 2007a; Howard et al. 2009; Rangwala et al. 2009a; Babusiaux et al. 2010; Kunder et al. 2012; Ness et al. 2013b; Nidever et al. 2012; Babusiaux et al. 2014; Zoccali et al. 2014). Therefore, here we seek only to place our results in context with those surveys. Figure 7 shows velocity histograms for both fields, the velocity distribution as a function of [Fe/H], and the velocity dispersion as a function of [Fe/H]. While a detailed comparison between our measured velocities and those in Babusiaux et al. (2010) is not possible because their individual velocities were not published, for both fields we can compare our average results with those given in Figure 13 of Zoccali et al. (2008) and Table 3 of Babusiaux et al. (2010). For the (5.25,–3.02) field, ignoring NGC 6553 stars, we find average velocity and dispersion values of 4.55 km s and 95.51 km s, respectively. This compares well with the Zoccali et al. (2008) average velocity of 11 km s and velocity dispersion of 107 km s. Similarly, in the (0,–12) field we measured an average heliocentric radial velocity of –8.61 km s (=85.56 km s) compared to the Babusiaux et al. (2010) values of –14 km s (=80 km s). Additionally, as can be seen in Figure 8 our galactocentric radial velocity (V) distributions are similar to those of nearby fields from the BRAVA, GIBS, and APOGEE surveys.
With the exception of the stars obviously related to NGC 6553, we find in agreement with previous bulge studies that, at least away from the Galactic plane, the velocity distributions are normal with no evidence for significant cold populations (but see also Rangwala et al. 2009b). This contrasts with Nidever et al. (2012) and Babusiaux et al. (2014), which find kinematically cold populations with V200 km s. However, their fields are significantly closer to the Galactic plane than those analyzed here. We do note however that these high velocity populations are also not found in the BRAVA, ARGOS, nor GIBS analyses, nor is there yet a satisfactory theoretical explanation for their origin (e.g., Li et al. 2014).
For the (0,–12) field we observe the same trend of a decrease in velocity
dispersion with increasing [Fe/H] found by Babusiaux et al. (2010). However,
while Babusiaux et al. (2010) find an increase in velocity dispersion with
increasing [Fe/H] in the (1.1,–4) field of Baade’s window, our off–axis
but similar Galactic latitude field at (5.25,–3.02) still exhibits a trend
of decreasing velocity dispersion with increasing [Fe/H]. This further
contrasts with recent fields observed close to the plane by Babusiaux et al.
(2014; see their Figure 16) that also show a possible increase in velocity
dispersion with increasing [Fe/H].
3.5. Identifying NGC 6553 Members
Members of the globular cluster NGC 6553 in the (5.25,–3.02) field are best identified in the velocity–metallicity diagram in Figure 7. The likely members (12 stars total) are clustered near [Fe/H]–0.10 and RV0 km s. Literature measurements of the cluster’s average [Fe/H] value vary considerably, with estimates that include: –0.55 (Barbuy et al. 1999), –0.16 (Cohen et al. 1999), –0.7 (Coelho et al. 2001), –0.3 (Origlia et al. 2002), –0.2 (Meléndez et al. 2003), and –0.2 (Alves–Brito et al. 2006). However, we find in agreement with the most recent estimates that [Fe/H]=–0.11 (=0.07). While the cluster is slightly iron–deficient relative to the Sun, the moderate enhancements of the cluster’s [/Fe] ratio (see Section 4.1) gives it an overall metallicity that is roughly solar. NGC 6553 is therefore one of the most metal–rich globular clusters in the Galaxy.
We find similar agreement with literature values for the cluster’s radial velocity, with RV=–2.03 km s (=4.85 km s). This is compared with recent values of: –1 km s (Coelho et al. 2001), 1.6 km s (Meléndez et al. 2003), and –1.86 km s (Alves–Brito et al. 2006). Finally, we note that the stars identified in Table 1 as possible cluster members have an average, projected radial distance from the cluster center of about 6 (=5). We have adopted a more lenient radial distance discriminator than the 2 limit used by Zoccali et al. (2008) and Gonzalez et al. (2011), and instead rely more on the [Fe/H] and velocity measurements to identify possible cluster members.
3.6. Abundance Ratio Comparisons with Previous Work
As noted previously, Zoccali et al. (2008) and Gonzalez et al. (2011) presented [Fe/H], [Si/Fe], [Ca/Fe], and [Ti/Fe] abundances based on the same GIRAFFE data utilized here. Therefore, in Figures 9–10 we compare our results with theirs for stars and elements in common. While a quantitative comparison of the individual [Fe/H] values is given in Section 3.1 (see also Figure 3), in Figure 9 we compare the general shapes and bulk properties of the metallicity distribution functions. For the (5.25,–3.02) field the average and median [Fe/H] ratios are similar, but the distribution from the present work is somewhat broader and extends to higher [Fe/H]. In contrast, there are no significant differences in the [Fe/H] distribution functions between the present work and the same stars from Zoccali et al. (2008), for the (0,–12) field. We also reconfirm one of the primary conclusions of Zoccali et al. (2008) that interior bulge fields have a higher average metallicity than outer bulge fields. Finally, we note that the distribution functions shown in Figure 9 do not provide strong evidence supporting the existence of multiple, discreet populations, as has been suggested in some studies (Bensby et al. 2011; Hill et al. 2011; Bensby et al. 2013; Ness et al. 2013a). However, the number of stars per field presented here is 100.
In Figure 10 we compare our derived [Mg/Fe], [Si/Fe], and [Ca/Fe] ratios
to those given in Gonzalez et al. (2011). The average differences
between the present work and that of Gonzalez et al. (2011) are
[Mg/Fe]=0.00 (=0.14), [Si/Fe]=0.00
(=0.13), and [Ca/Fe]=0.06 (=0.14). The relatively
consistent star–to–star scatter of 0.14 dex is a reasonable estimate
of the attainable precision between the two studies, which derive
–element abundances from different techniques (synthesis in Gonzalez
et al. 2011 and EW measurements here). We note that the –elements
oxygen (measured here) and titanium (measured in Gonzalez et al. 2011) were
not both measured in each study.
3.7. Abundance Uncertainty Estimates
We investigated the sensitivity of derived abundances for each element in every star by taking the abundances given in Table 3, determining theoretical EWs using the line list in Table 2, and then varying the model atmosphere parameters T, log(g), [Fe/H], and vt individually while holding the other parameters fixed. We selected parameter changes of 100 K in T, 0.30 dex in log(g), 0.15 dex in [M/H], and 0.30 km s in vt, which are reasonable when comparing our derived parameters with those of the independent analysis by Zoccali et al. (2008; see also Section 3.1). The total uncertainty for each element ratio in each star resulting from this exercise is provided in Table 4.
In general, most elements are not affected by changes in T of 100 K at more than the 0.1 dex level. However, the two species presented here that reside in their dominant ionization states ([O I] and Fe II) are strongly affected by changes in surface gravity. For a change in log(g) of 0.3 dex, the log (O) and log (Fe II) abundances can change by 0.1–0.3 dex, but these effects are mitigated when the [O I/H] abundance is normalized with [Fe II/H]. These two species are also more strongly affected by changes in the model metallicity, and the larger [Fe II/H] measurement and sensitivity uncertainties are a contributing factor to the increased dispersion in the [O/Fe] ratios compared to other –elements (e.g., [Mg/Fe]). As expected, microturbulence sensitivity is correlated with a star’s overall metallicity (i.e., line strength). Among the transitions under consideration here, in metal–rich stars those of Na, Ca, and Cu typically have the strongest lines and are thus more strongly affected by the microturbulence uncertainty.
In Table 5 we also provide the 1 line–to–line dispersion values for all species measured here. These values should be mostly representative of the combined measurement error that includes effects such as: continuum placement, line deblending, synthesis fits via visual inspection, log(gf) uncertainties, and model atmosphere deficiencies. Typical line–to–line dispersion values are 0.08 dex. The measurement error of Cu may be underestimated because of the line’s large EW, non–negligible blending (see Figure 6), and possible contamination with a nearby DIBS feature. A more reasonable measurement uncertainty for Cu is, in most cases, 0.15–0.20 dex.
4. Results and Discussion
4.1. The Elements Oxygen, Magnesium, Silicon, and Calcium
The –elements have been the primary focus of detailed composition work in the Galactic bulge. To first order there is agreement among the various studies that: (1) the [/Fe] ratios are enhanced by 0.3 dex at [Fe/H]–0.3, (2) for stars with [Fe/H]0.3 there is a mostly monotonic decline in [/Fe] with increasing [Fe/H], (3) the bulge and thick disk may share similar chemistry over a wide range in metallicity, and (4) there are no significant variations in the [/Fe] trends between different bulge sight lines (McWilliam & Rich 1994; Cunha & Smith 2006; Zoccali et al. 2006; Fulbright et al. 2007; Lecureur et al. 2007; Meléndez et al. 2008; Alves–Brito et al. 2010; Bensby et al. 2010a; Ryde et al. 2010; Bensby et al. 2011; Gonzalez et al. 2011; Hill et al. 2011; Johnson et al. 2011; Uttenthaler et al. 2012; Bensby et al. 2013; García Pérez et al. et al. 2013; Johnson et al. 2013; Ness et al. 2013a; Jönsson et al. 2014). Additionally, there is evidence that the [O/Mg] ratio declines with increasing metallicity (Fulbright et al. 2007; Lecureur et al. 2007; McWilliam et al. 2008; Alves–Brito et al. 2010).
The new data presented here, and summarized in Figure 11, reinforce many
observations from the previous studies mentioned above. In particular, we find
that for [Fe/H]–0.3 all of the [/Fe] ratios are enhanced and
exhibit minimal star–to–star scatter with [O/Fe]=0.54
(=0.10), [Mg/Fe]=0.33 (=0.08),
[Si/Fe]=0.28 (=0.07), and
[Ca/Fe]=0.34 (=0.09). For stars with
[Fe/H]–0.3, we find that the [/Fe] ratios decrease with increasing
[Fe/H]. However, Figure 11 illustrates the disparate trends for
individual elements and highlights the information loss that can occur when
averaging abundance ratios for multiple –elements. The [O/Fe] ratios
are higher by 0.2 dex in metal–poor stars than those of other
–elements, but this trend reverses for stars with [Fe/H]0 where
[O/Fe] is, on average, lower by 0.2 dex. While both O and Mg are
significant products of hydrostatic burning in massive stars (e.g., Woosley &
Weaver 1995), the [Mg/Fe] trend exhibits a more shallow decline with
increasing [Fe/H] than [O/Fe]. This is most clearly seen in Figure 12,
which shows a sharply declining [O/Mg] ratio at [Fe/H]–0.1. Although
massive star production of Si and Ca involves both hydrostatic and explosive
burning (e.g., Woosley & Weaver 1995), the [Si/Fe] and [Ca/Fe] trends are more
similar to [Mg/Fe] than [O/Fe]. Given the disparate trend of [O/Fe] compared
to other –elements, and the low production of most –elements
relative to Fe in Type Ia supernovae (SNe; e.g., Nomoto et al. 1997), we
conclude in agreement with past work (e.g., McWilliam et al. 2008) that the
strong decline in [O/Fe] at [Fe/H]–0.3 is likely a result of metallicity
dependent yields in massive stars
When comparing the individual [/Fe] trends between the two fields analyzed here, Figure 11 shows no significant variations. Similarly, in Figure 13 we combine our two fields and compare with literature giant and dwarf [/Fe] data. A comparison between the present work and literature giant trends, which span a variety of bulge sight lines, leads us to find in agreement with Johnson et al. (2011; 2013) and Gonzalez et al. (2011) that no significant field–to–field [/Fe] variations exist over a broad region of the bulge. The microlensed dwarf data exhibit the same qualitative and quantitative distributions for [O/Fe] and [Mg/Fe], at least for [Fe/H]0, as the giant data, but there may be small systematic offsets with [Si/Fe] and [Ca/Fe]. In particular, the dwarf abundances are 0.1 dex lower for a given [Fe/H], when considering [Fe/H]0. At super–solar [Fe/H] values, the dwarf and giant data are in excellent agreement for [O/Fe], but the leveling–off or slight increase in [Mg/Fe], [Si/Fe], and [Ca/Fe] seems to be unique to the dwarf measurements. Unfortunately, the source of this discrepancy is not clear and may be related to analysis differences between dwarfs and giants.
In Figure 14 we compare the [O/Fe], [Mg/Fe], [Si/Fe], and [Ca/Fe] abundances between the bulge, thick disk, and thin disk. For stars with [Fe/H]–0.5, the bulge and thick disk stars exhibit similar abundance patterns for all four element ratios. However, we note that on average the bulge stars have [O/Fe] and [Mg/Fe] ratios that are slightly enhanced by 0.03 dex and [Si/Fe] and [Ca/Fe] ratios that are enhanced by 0.05 dex compared to similar metallicity thick disk stars. In contrast, the most metal–poor thin disk stars exhibit significantly lower [X/Fe] ratios for all of the –elements measured here. The bulge and thin disk stars with [Fe/H]0 are not strikingly different, but the star–to–star scatter, especially for [O/Fe], is significantly larger for the bulge giants. For the intermediate range of [Fe/H]–0.5 to 0, the bulge stars still exhibit significantly larger [/Fe] ratios than the thin disk and may remain enhanced to a higher [Fe/H] value than the thick disk.
The chemical similarities between especially the metal–poor bulge and thick
disk found here have been documented in previous work (Meléndez et
al. 2008; Alves–Brito et al. 2010; Bensby et al. 2010a; Ryde et al. 2010;
Bensby et al. 2011; Gonzalez et al. 2011; Hill et al. 2011; Johnson et al.
2011, 2013). The apparent homogeneity between the most metal–poor bulge and
thick disk stars lends credibility to the idea that the Galactic bulge formed
in situ with the disk. However, there is not universal agreement in the
literature that the metal–poor bulge and disk trends are identical. In
particular, earlier work by Zoccali et al. (2006), Fulbright et al. (2007),
and Lecureur et al. (2007) found that the bulge stars exhibited both larger
[/Fe] ratios and remained enhanced to higher [Fe/H] values than the
local thick disk
Finally, we note that combining the present data set with those available in the literature (e.g., see Figure 13) totals 10 bulge stars that have had [/Fe] measurements made from high resolution spectroscopy. Despite the large sample size, there is a paucity of stars with [/Fe] ratios that deviate significantly from the bulk trend. In agreement with work suggesting the Galactic bulge did not form predominantly from a build–up of merger events (e.g., Shen et al. 2010), we can effectively rule out significant contributions from the infall of objects with chemistry similar to those of many present–day dwarf galaxies (i.e., low [/Fe]; e.g., see Venn et al. 2004 and references therein). Additionally, as can be seen in Figure 11 (see also Gonzalez et al. 2011) the [X/Fe] abundance ratios of individual –elements for NGC 6553 stars are nearly identical to those of bulge field stars with similar [Fe/H]. Specifically, the average [X/Fe] values for NGC 6553 are: [O/Fe]=0.24 (one star), [Mg/Fe]=0.16 (=0.08), [Si/Fe]=0.17 (=0.10), and [Ca/Fe]=0.22 (=0.12), which compare well with the average abundances for nearby bulge field stars in the range [Fe/H]=–0.20 to 0.00: [O/Fe]=0.24 (=0.29), [Mg/Fe]=0.25 (=0.09), [Si/Fe]=0.15 (=0.08), and [Ca/Fe]=0.19 (=0.13). These values are in good agreement with past work that finds the cluster to be moderately –enhanced (Barbuy et al. 1999; Cohen et al. 1999; Coelho et al. 2001; Origlia et al. 2002; Meléndez et al. 2003; Alves–Brito et al. 2006). The similar [/Fe] abundances between the cluster and field stars suggests that NGC 6553 likely formed in situ with the bulge field population and is not a captured cluster.
4.2. The Light, Odd–Z Elements Sodium and Aluminum
In a similar fashion to the –elements, the light, odd–Z elements Na and Al provide clues of the processes that dominated the chemical enrichment of a stellar population. Furthermore, these elements are useful for “chemical tagging” analyses, and both the [Na/Fe] and [Al/Fe] ratios can vary significantly between stellar populations that have otherwise identical [/Fe] and [Fe/H] values. The large (0.5 dex) star–to–star [Na/Fe] and [Al/Fe] abundance variations present in metal–poor globular cluster but not halo/disk stars of the same metallicity are perhaps the most well–known example of this phenomenon (e.g., see reviews by Gratton et al. 2004; 2012 and references therein). While the production of Na and Al is dominated by hydrostatic helium, carbon, and neon burning in massive stars, the final yields are expected to grow significantly with increasing progenitor mass and metallicity (e.g., Woosley & Weaver 1995; Kobayashi et al. 2006; 2011). Intermediate mass (4–8 M) asymptotic giant branch (AGB) stars and the hydrogen–rich envelopes of massive stars can also produce significant amounts of Na and Al via the NeNa and MgAl proton–capture cycles (e.g., Decressin et al. 2007; de Mink et al. 2009; Ventura & D’Antona 2009; Karakas 2010). Since Na and Al are thought to result from similar production mechanisms, we expect their abundance patterns to reflect a comparable morphology.
While the bulge abundance patterns of [Na/Fe] and [Al/Fe] have not been investigated to the extent of the –elements, the combined literature sample now totals of order a few hundred stars. Interestingly, the agreement between studies regarding the [Na/Fe] and [Al/Fe] trends is worse than for the –elements. While all high–resolution analyses (McWilliam & Rich 1994; Fulbright et al. 2007; Lecureur et al. 2007; Alves–Brito et al. 2010; Bensby et al. 2010a, 2011; Johnson et al. 2012; Bensby et al. 2013) tend to agree that the average [Na/Fe] ratio rises with increasing metallicity, significant scatter is present at [Fe/H]–1 and [Fe/H]0. Similarly, there is general agreement that [Al/Fe] is enhanced in bulge stars at [Fe/H]–0.3. However, some studies find that [Al/Fe] remains enhanced at super–solar metallicities (McWilliam & Rich 1994; Fulbright et al. 2007; Lecureur et al. 2007; Alves–Brito et al. 2010) while others find a decline in [Al/Fe], similar to [/Fe] (Bensby et al. 2011, 2011; Johnson et al. 2012; Bensby et al. 2013). Additionally, there is general agreement that the [Na/Fe] and [Al/Fe] trends as a function of [Fe/H] are similar between the bulge and disk over a broad metallicity range, but differences could be present at the metal–poor and metal–rich ends of the bulge distribution. It is also not yet clear if any significant [Na/Fe] and [Al/Fe] abundance differences exist between different bulge sight lines.
Figure 11 shows our derived [Na/Fe] and [Al/Fe] abundances as a function of [Fe/H] for both fields and the possible NGC 6553 stars, and in Figure 15 we compare our results with those from previous work. For Na we find general agreement with literature values such that the average [Na/Fe] ratio rises with increasing [Fe/H]. However, we find only a small number of metal–rich stars with [Na/Fe]0.4 and do not reproduce the very large [Na/Fe] ratios of Lecureur et al. (2007). Additionally, we do not find significant evidence supporting large [Na/Fe] variations between the two bulge sight lines probed here. At [Fe/H]–0.5, the mean [Na/Fe] trend and star–to–star dispersion for our measured RGB stars is in good agreement with those of the microlensed bulge dwarfs (e.g., Bensby et al. 2013).
The primary discrepancy between our work and some of the literature values
occurs for stars with [Fe/H]–0.7, with the present work and Johnson et
al. (2012) finding that the average Na trend decreases from [Na/Fe]0 at
[Fe/H]=–0.5 to [Na/Fe]=–0.3 at [Fe/H]=–1.7. It is not immediately clear
if the discrepancy, especially between the bulge RGB and dwarf data, is real
or caused by analysis differences (e.g., NLTE, 3D, or spherical/plane–parallel
effects between dwarfs and giants). The inclusion of NLTE corrections would
minimize the differences at low metallicity between bulge RGB and dwarf stars,
and also between bulge RGB and metal–poor thick disk dwarfs (see
Figure 16), if the largely positive Na corrections for RGB stars from
Gratton et al. (1999) were applied. However, more recent NLTE calculations
(e.g., Lind et al. 2011) instead find that the sign of the Na correction is
negative for the lines and atmospheric parameters used here. Similarly, the
NLTE corrections for log (Fe I) appear to be positive (e.g.,
Lind et al. 2012; Bergemann et al. 2012) for most stars in our
When comparing the [Na/Fe] and [Al/Fe] trends in Figure 11, it is immediately clear that the two elements exhibit discrepant trends. While [Na/Fe] gradually rises with increasing [Fe/H], the [Al/Fe] trend is nearly indistinguishable from that of most –elements. In particular, we find in agreement with Bensby et al. (2010a; 2011; 2013) and Johnson et al. (2012) that [Al/Fe]0.3 in bulge stars until [Fe/H]–0.3 and then steadily declines at higher [Fe/H]. As mentioned previously, the decline in [Al/Fe] with increasing metallicity contrasts with other literature results that find [Al/Fe] remains enhanced even at [Fe/H]=0.5 (Fulbright et al. 2007; Lecureur et al. 2007; Alves–Brito et al. 2010). The data quality among the various studies is comparable, and it is not clear why the derived [Al/Fe] trends are in disagreement at high metallicity. We do note however that for cool, high metallicity stars the 6696 and especially 6698 Å Al I lines, as well as the continuum placement, can be affected by CN blending.
The discrepant [Na/Fe] and [Al/Fe] trends as a function of [Fe/H] are not limited to the bulge and may also be present in the disk, as can be seen in Figure 16. Despite nucleosynthesis models predicting similar production of Na and Al in massive stars (e.g., Woosley & Weaver 1995), Figure 16 shows that, at least in the metallicity range probed here, Al is over–produced relative to Na in both bulge and disk stars for [Fe/H]–0.3. The increased production of Na relative to Al in metal–rich stars, and especially in the bulge, suggests that metallicity dependent yields from massive stars vary more strongly for Na than Al. Contributions from intermediate mass AGB stars may also help explain the Na and Al trends since the AGB [Na/Fe] yields tend to increase at higher [Fe/H] while those of [Al/Fe] decline (e.g., Ventra & D’Antona 2009). Interestingly, we find that, unlike the case for [Na/Fe], the [Al/Fe] ratios are nearly indistinguishable between the bulge and thick disk at [Fe/H]0. Similarly, the [Al/Fe] ratios for bulge stars are identical to those in the thin disk at [Fe/H]0.
Given the similar behavior of [Al/Fe] to many of the –elements, in Figure 17 we provide a detailed comparison between [Al/Fe], [O/Fe], [Mg/Fe], [Si/Fe], and [Ca/Fe] for the bulge stars analyzed here. While the [O/Fe] trend is clearly different than that of [Al/Fe], there are no similarly strong discrepancies between [Al/Fe] and the other –elements. At [Fe/H]–0.8 both [Mg/Fe] and [Ca/Fe] are 0.10–0.15 dex enhanced compared to [Al/Fe], but those differences disappear at higher [Fe/H]. On the other hand, the [Si/Fe] and [Al/Fe] trends are essentially identical at all [Fe/H] with an average difference of 0.01 dex (=0.13 dex).
Examining the NGC 6553 stars in Figure 11 shows that Na, and to a lesser extent Al, exhibit larger star–to–star [Na/Fe] and [Al/Fe] variations than similar metallicity field stars. In particular, the average Na and Al abundances for the cluster stars are [Na/Fe]=0.16 (=0.20) and [Al/Fe]=0.17 (=0.13), which can be compared to the similar metallicity fields stars having [Na/Fe]=0.03 (=0.11) and [Al/Fe]=0.16 (=0.10), respectively. The larger [Na/Fe] abundance and dispersion values for the cluster stars suggests NGC 6553 experienced some degree of self–enrichment. However, unlike low metallicity globular clusters, NGC 6553 does not exhibit a strong Na–Al correlation. This is in agreement with the observed trend that the Na–Al correlation is more mild and [Al/Fe] dispersions smaller in metal–rich as opposed to metal–poor globular clusters (e.g., Carretta et al. 2009; O’Connell et al. 2011; Cordero et al. 2014). Unfortunately, the 6300 Å telluric oxygen emission feature combined with NGC 6553’s relatively low radial velocity prohibited us from obtaining an [O/Fe] abundance for more than one star in NGC 6553. Therefore, we cannot comment further on the existence or extension of the likely O–Na correlation. Finally, we note that our mean [Na/Fe] and [Al/Fe] values and abundance dispersions are in excellent agreement with those found by Alves–Brito et al. (2006), but are considerably lower than the values (based on two stars) of Barbuy et al. (1999).
4.3. The Fe–Peak Elements: Chromium, Cobalt, Nickel, and Copper
Unlike the lighter elements, the abundance patterns of Fe–peak elements in the Galactic bulge are not well–explored. The production of Fe–peak elements occurs through a variety of processes in the late stages of massive star evolution, the resulting core collapse SNe, and also in Type Ia SNe. The Fe–peak abundance patterns can also be useful indicators of a stellar population’s IMF, with odd–Z elements in particular providing some diagnostic power (e.g., Nomoto et al. 2013). Some initial work on the bulge Fe–peak abundance distribution was included in McWilliam & Rich (1994), which found [V/Fe], [Cr/Fe], and [Ni/Fe] ratios near solar and a possible enhancement in [Co/Fe] and [Sc/Fe]. More recent work analyzing the Fe–peak abundance trends in the bulge has come from microlensed dwarf studies (Cohen et al. 2008; Johnson et al. 2008; Cohen et al. 2009; Bensby et al. 2010a; Epstein et al. 2010; Bensby et al. 2011; Bensby et al. 2013). The bulge [Mn/Fe] trend in RGB stars has also been investigated recently by Barbuy et al. (2013). The results of the these analyses indicate that the bulge Fe–peak trends are similar to that of the local disk, except that the bulge may have different [Mn/O] ratios than the thick disk for a given [O/H] value.
The general [X/Fe] versus [Fe/H] abundance trends derived here are shown in Figure 11. From these data we find that: (1) Cr is the element that most closely tracks Fe with [Cr/Fe]=0.00 (=0.11), (2) [Co/Fe] exhibits low level variations as a function of [Fe/H] but is generally enhanced with [Co/Fe]=0.14 (=0.11), (3) [Ni/Fe] shows similar variations to [Co/Fe] but at a much smaller amplitude and is slightly enhanced with [Ni/Fe]=0.09 (=0.06), (4) the Cu abundance increases monotonically from [Cu/Fe]=–0.84 in the most metal–poor star to [Cu/Fe]0.40 in the most metal–rich stars, and (5) there are no significant Fe–peak abundance variations between NGC 6553 stars and the field stars.
Although the exact nature of Cu nucleosynthesis is complex (e.g., see Mishenina et al. 2002 and references therein), the significant secondary (i.e., metallicity–dependent) production of Cu (and also Na) is evident in Figure 11. Additionally, Figure 12 shows that despite the larger measurement errors in both O and Cu abundances, the [Cu/O] ratio is strongly correlated with [Fe/H]. This trend has been noted previously and is prevalent in stellar populations with different star formation histories, such as the local disk and Sagittarius Dwarf Galaxy (e.g., McWilliam et al. 2013). The [Cu/O] trend is taken as evidence that a significant portion of Cu is synthesized in massive stars, perhaps via the weak s–process (e.g., Sneden et al. 1991). However, some component of Cu may also be produced by Type Ia SNe (Matteucci et al. 1993).
In Figure 18 we compare our derived Fe–peak abundance trends with those in the literature. For Cr there is general agreement between the bulge RGB stars analyzed here and the literature microlensed dwarf data. However, the small number of bulge literature data points for Co and Cu makes a direct comparison difficult. The [Ni/Fe] comparison also shows excellent agreement overall, but the RGB stars appear systematically enhanced by 0.1 dex in the range [Fe/H]=–0.3 to 0.1. Note also the similarly small star–to–star dispersion in especially [Ni/Fe] between the RGB and dwarf data.
A comparison between the bulge Fe–peak abundance trends and those of the thin/thick disk is shown in Figure 19. Interestingly, at least for [Fe/H]–1.5, the [Cr/Fe] distribution is seemingly independent of formation environment with the bulge, thick disk, and thin disk stars all having [Cr/Fe]0. For [Co/Fe], [Ni/Fe], and [Cu/Fe], there is significant overlap between the bulge and thick disk trends at [Fe/H]–0.5. At higher [Fe/H], the bulge may be enhanced in all three elements relative to both the thick and thin disks. This is especially evident in the Figure 19 panel showing [Ni/Fe] versus [Fe/H]; the low star–to–star scatter in [Ni/Fe] for all three populations highlights the possible composition difference between the local disk and bulge from [Fe/H]–0.4 to 0.2. While the strong rise in [Cu/Fe] with metallicity is, as mentioned previously, a common feature in many different stellar populations, the bulge stars at [Fe/H]–0.3 appear to extend to higher abundances than the local disk. However, the increased measurement uncertainty of Cu and paucity of disk [Cu/Fe] ratios at [Fe/H]0 prevents us from undertaking a more comprehensive analysis.
4.4. Comparing Composition Data to Bulge Chemical Enrichment Models
Accurately modeling the chemical enrichment history of a stellar system
requires solving for a variety of free parameters that may include the IMF,
star formation rate, star formation efficiency, supernova/hypernova
The baseline Galactic bulge model shown in Figures 20–21 is
from Kobayashi et al. (2006; 2011) and is designed to reproduce the metallicity
distribution in Baade’s Window from Zoccali et al. (2008), assumes a Kroupa
(2008) IMF, and assumes a star formation time scale of 3 Gyr (see Kobayashi et
al. 2011; their Table 1 and Section 2.4 for more detail regarding model input
parameters). In general, the baseline model does a reasonable job of
reproducing the observed abundance trends of all abundance
ratios, except [Na/Fe] and [Al/Fe]. All of the models shown in
Figures 20–21 predict large over–abundances of both
[Na/Fe] and [Al/Fe] that are not observed, suggesting the massive star
yields of both elements may be too high
In Figure 21 we examine how changes in the IMF could affect the expected abundance trends. Compared to the Kroupa (2008) IMF adopted in our baseline bulge model, a steep IMF (x=1.6) is completely ruled out by the data. Additionally, adopting a Kroupa (2008) IMF that truncates at 40 M, and thus ignores contributions from the most massive stars, is inconsistent with the [Cu/Fe] abundances, and to a lesser extent those of [Co/Fe]. While a flatter, top–heavy IMF (x=0.3) alone leads to [X/Fe] ratios that are too high for nearly every element, a reduction in the yields from outflow and/or slow star–formation combined with a high Type Ia SN rate, artificially enhanced with a ten times larger binary fraction, could bring such a model into agreement with the data. However, bulge formation models with slow star–formation are likely unrealistic, and the observed [Co/Fe] and [Cu/Fe] data appear to rule out these models. Based on the present data it does not appear that the bulge required a uniquely “non–standard” IMF to reach its present–day composition (but see also Ballero et al. 2007, for example).
Finally, in Figure 20 we also compare the measured bulge abundance trends with our adopted baseline model and similar models representing the composition distributions of the local thick disk and thin disk. Comparing the three predicted trends indicates that in the range –0.8[Fe/H]–0.3 the bulge [/Fe] ratios should be similar or modestly enhanced and remain enhanced to higher [Fe/H] than the thick disk. Similarly, at [Fe/H]0 the bulge and thin disk should exhibit similar, if not identical, [/Fe] ratios. Both of these predictions match our observations (see Section 4.1). The predicted enhancements in the bulge for [Na/Fe] and [Al/Fe] compared to the local disk are not supported by observations, but this could be related to the previously mentioned possible over–production issues of the adopted stellar yields. However, in addition to Na and Al, Figure 20 shows that Co and Cu may also exhibit some discriminating power between the bulge and local disk populations. In particular, the data support bulge stars with [Fe/H]–0.5 having [Co/Fe] and [Cu/Fe] ratios that are higher than the local disk. Therefore, the data and models presented here provide some supporting evidence that the bulge experienced a different chemical enrichment path than the thick disk.
We have measured radial velocities and chemical abundances of O, Na, Mg, Al, Si, Ca, Cr, Fe, Co, Ni, and Cu in a sample of 156 RGB stars located in Galactic bulge fields centered near (l,b)=(5.25,–3.02) and (0,–12). The (5.25,–3.02) also includes 12 stars identified as likely members of the bulge globular cluster NGC 6553, based on their radial velocity and [Fe/H] values. The results are based on high resolution archival spectra obtained with the FLAMES–GIRAFFE instrument, and originally used to derive [Fe/H] and [/Fe] abundances in Zoccali et al. (2008) and Gonzalez et al. (2011). We culled the original target list and selected only those stars with co–added S/N70 that also lack strong TiO bands. The abundance analysis was carried–out using standard EW and spectrum synthesis techniques.
Our derived heliocentric radial velocity distributions for both fields are in good agreement with past surveys (BRAVA, GIBS, and APOGEE) covering nearby fields. We do not confirm the existence of a significant population of high velocity stars noted by Nidever et al. (2012) and Babusiaux et al. (2014). However, our targeted fields are farther away from the plane than most of those in which Nidever et al. (2012) and Babusiaux et al. (2014) observe the cold, high velocity stars. For both fields analyzed here we also find that the velocity dispersion monotonically decreases with increasing [Fe/H]. This is not unexpected for the outer bulge field at (0,–12), but the similar trend in the (5.25,–3.02) field appears to contradict the findings of Babusiaux et al. (2010; 2014) that the velocity dispersion of bulge stars with [Fe/H]0 increases at lower Galactic latitude. The reason for this discrepancy is not clear, but we note that previous analyses finding increased velocity dispersion at low Galactic latitude for metal–rich stars have all focused on minor–axis fields. The inner bulge field included here is several degrees off–axis.
The composition data reconfirm the already well–documented metallicity gradient in the bulge. Similarly, we find good agreement between our derived [Mg/Fe], [Si/Fe], and [Ca/Fe] abundances and those of Gonzalez et al. (2011). Additionally, we confirm that there are no significant field–to–field [/Fe] abundance variations among various bulge sight lines. Our new –element measurements also reinforce the previously held notion (e.g., McWilliam et al. 2008) that the decline in [O/Mg] with increasing metallicity is likely the result of metallicity dependent yields from massive stars. While we find that the bulge and thick disk exhibit nearly identical [/Fe] ratios at [Fe/H]–0.5, the bulge stars appear to remain enhanced in [/Fe] by up to 0.1–0.2 dex higher in [Fe/H] than the local thick disk. The bulge [/Fe] ratios at [Fe/H]0 are well–matched to the local thin disk trends. These results are in agreement with recent differential abundance analyses of microlensed bulge dwarfs (Bensby et al. 2013), and suggest the bulge experienced faster enrichment than the local thick disk. However, similar differential analyses comparing bulge and thick disk giants find no significant differences between the two populations (Meléndez et al. 2008; Alves–Brito et al. 2010; Gonzalez et al. 2011).
Combining the new data set of [/Fe] abundances with those available in the literature now totals several hundred stars. However, the combined data set does not reveal any significant population with “anomalous” chemistry, such as the low [/Fe] ratios reminiscent of many present–day dwarf galaxy stars. Therefore, we can effectively rule out these types of objects as major contributors to any portion of the present–day Galactic bulge field population. This further supports the idea that the Galactic bulge is not a merger–built system. Similarly, the [/Fe] ratios of the NGC 6553 stars are identical to those of similar metallicity field stars. This suggests NGC 6553 formed in situ with the bulge and is not a captured system.
With regard to the light, odd–Z elements, we find that Na and Al exhibit discrepant trends as a function of metallicity. In particular, bulge stars exhibit a steady increase in [Na/Fe] with increasing [Fe/H], but the [Al/Fe] trend almost exactly matches that of the –elements (except oxygen). While we do not find any significant field–to–field variations in either [Na/Fe] or [Al/Fe], our results indicate that the bulge and thick disk have different [Na/Fe] abundances at [Fe/H]–0.5 but similar [Al/Fe]. Interestingly, the “–like” behavior of [Al/Fe] contrasts with several previous bulge studies that found [Al/Fe] was enhanced up to [Fe/H]=0.5. Instead, our results are in agreement with the abundance patterns of microlensed bulge dwarfs (e.g., Bensby et al. 2013). The discrepant behavior of Na and Al suggests metallicity dependent yields from massive stars, and perhaps intermediate mass stars, leads to significantly more production of Na than Al at high metallicity. We also find that the NGC 6553 stars have nearly identical [Al/Fe] ratios as similar metallicity field stars, but both the average [Na/Fe] abundance and star–to–star dispersion of cluster stars are higher. This suggests NGC 6553 experienced some light element self–enrichment, which is typical for globular clusters.
The abundance trends of the Fe–peak elements are distinctly different: (1) the average [Cr/Fe] ratio is essentially solar over the full range in [Fe/H] and shows no variations over the metallicity range probed here, (2) both [Co/Fe] and [Ni/Fe] are enhanced by 0.1 dex at nearly all [Fe/H] and exhibit some low amplitude, metallicity–dependent variations, and (3) [Cu/Fe] exhibits a large increase from the metal–poor to metal–rich end of the distribution. In a similar fashion to [Na/Fe], the strong secondary (metallicity–dependent) production of Cu is evident in bulge stars, and the correlation between [Cu/O] and [Fe/H] suggests massive stars produce significant portions of Cu. However, Cu production from another source (e.g., Type Ia SNe) seems required to explain the high [Cu/Fe] abundances at super–solar metallicities. Interestingly, at [Fe/H]–2 the [Cr/Fe] trend is identical between the bulge, thick disk, and thin disk, but the heavier Fe–peak [X/Fe] ratios appear to all be enhanced in the bulge relative to the local disk. Additionally, the NGC 6553 Fe–peak abundance trends are in agreement with similar metallicity field stars.
Despite predicting [Na/Fe], [Al/Fe], and [Ni/Fe] ratios that are too high, our adopted baseline bulge chemical enrichment model from Kobayashi et al. (2006; 2011) does a reasonable job fitting the abundance trends of the and other Fe–peak elements. However, better agreement between the data and model is found when a significant fraction of HNe, which produce more Fe, are included. Unfortunately, setting the HN fraction is best constrained using abundance patterns at [Fe/H]–1, where the bulge data are sparse. While a Kroupa (2008) IMF provides a reasonable fit to the observed abundance trends, a top–heavy IMF including strong outflow cannot be ruled out. In contrast, the Fe–peak abundance data strongly rule out IMFs that are truncated to exclude the contributions of stars 40 M, steep IMFs (e.g., x=1.6), and top–heavy IMFs that do not include outflow. We conclude that the bulge likely does not require a particularly unusual IMF to explain its present–day abundance patterns, and that its enhanced abundances for several and Fe–peak elements match model predictions in which the bulge experienced a different enrichment history than the local disk.
||2MASS||RA (J2000)||DEC (J2000)||V||V–I||J||H||K||T||log(g)||[Fe/H]||vt||RV||RV
|(degrees)||(degrees)||(mag.)||(mag.)||(mag.)||(mag.)||(mag.)||(K)||(cgs)||(km s)||(km s)||(km s)|
||log (X)||log (X)||[X/Fe] or [Fe/H]|
||[Na/Fe]||[Mg/Fe]||[Al/Fe]||[Si/Fe]||[Ca/Fe]||[Cr/Fe]||[Fe I/H]||[Fe II/H]||[Co/Fe]||[Ni/Fe]||[Cu/Fe]|