CHEMICAL ANALYSIS OF ASYMPTOTIC GIANT BRANCH STARS IN M62^*

The radial velocities of our targets have been obtained by using the code DAOSPEC (Stetson & Pancino 2008) and by measuring the position of over 300 metallic lines distributed along the whole spectral range covered by the 580 Red Arm of UVES-FLAMES. The uncertainties have been computed as the dispersion of the velocities measured from each line divided by the square root of the number of lines used, and they turned out to be smaller than 0.05 km s⁻¹. Finally, we applied the heliocentric corrections computed with the IRAF task RVCORRECT. For each spectrum, the zero point of the wavelength calibration has been accurately checked by means of a few emission lines of the sky. The final velocities are listed in Table 1. They range from −109.8 to −53.4 km s⁻¹, with a mean value of −76.7 ± 3.6 km s⁻¹ and a dispersion σ = 15.6 km s⁻¹. These values are in good agreement with the derivations of Dubath et al. (1997, v_r = −71.8 ± 1.6 km s⁻¹, σ = 16.0 km s⁻¹) and Yong et al. (2014, v_r = −70.1 ± 1.4 km s⁻¹, σ = 14.3 km s⁻¹), the small differences being likely due to the small statistics.

The most discrepant target (id = 79), with a radial velocity of −109.85 km s⁻¹, is still within 2σ from the systemic velocity of the cluster. By using the Besançon Galactic model (Robin et al. 2003), we extracted a sample of about 5,300 field stars in the direction of M62, finding a quite broad and asymmetric radial velocity distribution, with mean v_r ≃ −60 km s⁻¹ and dispersion σ = 80 km s⁻¹, which partially overlaps that of the cluster. On the other hand, only a few percent of the stars studied in that region close to the Galactic bulge have a [Fe/H] < −1.0 dex (see, e.g., Zoccali et al. 2008; Hill et al. 2011; Johnson et al. 2013; Ness et al. 2013). Thus, taking into account the metallicity of star 79 (see below), its position in the CMD, and its distance from the cluster center (d ∼ 385), we conclude that it is likely a genuine cluster member, and we therefore keep it in the following analysis.

First-guess effective temperature (T_eff) and surface gravity (log g) values for each target have been derived by using the photometric information. Temperatures have been estimated by using the ${(U-V)}_{0}-{T}_{{\rm{eff}}}$ calibration of Alonso et al. (1999). Gravities have been computed with the Stefan–Boltzmann equation by adopting the color excess quoted above, a distance modulus ${(m-M)}_{0}=14.16$ mag (Harris 1996), and the bolometric correction from Alonso et al. (1999). For the RGB stars, we adopted a mass of 0.82 M_⊙, according to the best-fit isochrone retrieved from the PARSEC data set (Bressan et al. 2012), and computed for an age of 12 Gyr and a metallicity Z = 0.0013. For the AGB stars, we adopted a mass of 0.61 M_⊙, according to the median value of the HB mass range estimated by Gratton et al. (2010).

Then we performed a spectroscopic analysis as done in Lapenna et al. (2014) and Mucciarelli et al. (2015a), constraining the atmospheric parameters as follows: (1) spectroscopic temperatures have been obtained by requiring that no trend exists between iron abundance and excitation potential, (2) the gravity was derived by using the Stefan–Boltzmann equation with the value of T_eff thus obtained, and (3) the microturbulent velocity was determined by requiring that no trend exists between iron abundance and line strength. In order to evaluate the effects of a different procedure in the derivation of the atmospheric parameters and abundances, we have also performed a spectroscopic determination of the surface gravities by modifying condition (2) and imposing that the same abundance is obtained from neutral and single-ionized iron lines (ionization balance).

The chemical abundances of Fe, Ti, Na, Mg, and Al have been derived with the package GALA⁵ (Mucciarelli et al. 2013b), which adopts the classical method to derive the abundances from the measured EWs of metallic unblended lines. The EW and the error of each line were obtained using DAOSPEC, iteratively launched by means of the 4DAO⁶ code (Mucciarelli 2013a). The lines considered in the analysis have been selected from suitable synthetic spectra at the UVES-FLAMES resolution and computed with the SYNTHE package (Sbordone 2005) by using the guess atmospheric parameters and the metallicity derived by Yong et al. (2014). The model atmospheres have been computed with the ATLAS9⁷ code. We adopted the atomic and molecular data from the last release of the Kurucz/Castelli compilation⁸ and selected only the lines predicted to be unblended. The selected lines and the atomic data adopted in the analysis are listed in Table 2.

As detailed in Table 3, we used 100–150 Fe i lines and 7–12 Fe ii lines to derive the iron abundances, and 25–60 lines of Ti i and 6–15 lines of Ti ii to derive the abundances of titanium. For Na i, Mg i, and Al i, only few lines are available, namely those at 5682–5688 and 6154–6160 Å for Na i, the line at 5711 Å and the doublet at 6318–6319 Å for Mg i, and the doublet at 6696–6698 Å for Al i. The O abundances have been derived from spectral synthesis in order to take into account the blending between the forbidden [O i] line at 6300.3 Å and a Ni transition. For Ni we adopted the average abundance obtained by Yong et al. (2014), while for stars located in the upper RGB we assumed average C and N abundances according to Gratton et al. (2000), all rescaled to the assumed solar reference values (Grevesse & Sauval 1998). Because in some spectra the [O i] line was also partially blended with a telluric line, the spectra have been cleaned by using suitable synthetic spectra obtained with the TAPAS tool (Bertaux et al. 2014). For some stars, the [O i] line is not detectable, so only upper limits are obtained. As solar reference abundances we adopted the Caffau et al. (2011) value for O, and those of Grevesse & Sauval (1998) for all of the other elements.

For the computation of the global uncertainties on the final abundances, we took into account two main sources of errors, which have been added in quadrature.

• 1.  
  The error arising from the EW measurements. For each star we computed this term by dividing the line-to-line dispersion by the square root of the number of lines used. Thanks to the high quality of the spectra and to the number of lines that can be exploited, this term turned out to be very small, especially for Fe i and Ti i (providing up to 150 lines). For these species, the line-to-line scatter is smaller than 0.1 dex, leading to internal uncertainties lower than 0.01–0.02 dex. For Fe ii and Ti ii, the number of lines ranges from seven up to 15, leading to an uncertainty of about 0.02–0.03 dex. For the other chemical species, the number of measured lines is much smaller (1–4). Hence, the average uncertainties are of the order of 0.06–0.08 dex for O i, Na i, Mg i, and Al i.
• 2.  
  The error arising from atmospheric parameters. For the computation of this term, we varied each parameter by the 1σ error obtained from the previous analysis. We have found that representative errors for T_eff, log g, and v_turb are ∼50 K, 0.1 dex, and 0.1 km s⁻¹, respectively, for both the RGB and the AGB samples. Thus we decided to adopt these values as the 1σ error for all stars. We also checked the effect of a ±0.1 dex change in the metallicity of the model atmosphere, finding variations smaller than ±0.01 dex on the final abundances.

By using spectroscopic gravities (thus imposing that the same iron abundance is obtained from neutral and from single-ionized lines), the average values measured for the RGB and the AGB subsamples are [Fe/H]_RGB = −1.10 ± 0.01 (σ = 0.04 dex) and [Fe/H]_AGB = −1.18 ± 0.01 (σ = 0.03 dex). These values are consistent (within 1–2 σ) with previous abundance determinations of M62 giants, whether they are on the RGB or on the AGB: [Fe/H] =−1.12 dex (Kraft & Ivans 2003),⁹ [Fe/H] = −1.18 ± 0.07 dex (Carretta et al. 2009a), and [Fe/H] = −1.15 ± 0.02 dex (σ = 0.05 dex, Yong et al. 2014).

By using photometric gravities (and not imposing ionization balance), we determined the iron abundances separately from neutral and from single-ionized lines. For the 13 RGB stars we obtained [Fe i/H]_RGB = −1.07 ± 0.01 dex (σ = 0.04 dex) and [Fe ii/H]_RGB = −1.04 ± 0.02 dex (σ = 0.06 dex). For the five AGB stars we measured [Fe i/H]_AGB = −1.19 ± 0.01 dex (σ = 0.04 dex) and [Fe ii/H]_AGB = −1.06 ± 0.02 dex (σ = 0.06 dex). The average difference between the values of log g derived spectroscopically and those derived photometrically are 0.09 dex (σ = 0.10 dex) and 0.30 dex (σ = 0.20 dex) for the RGB and the AGB samples, respectively. Figure 2 shows the generalized histograms of the iron abundances for the RGB and the AGB samples separately, obtained by using spectroscopic (left panels) and photometric gravities (right panels). By construction, the distributions of [Fe i/H] and [Fe ii/H] essentially coincide if spectroscopic gravities are assumed. Instead, the two distributions significantly differ in the case of AGB stars if photometric gravities are adopted. In particular, the average iron abundances of RGB stars measured from neutral and single-ionized lines are consistent within the uncertainties, whereas a difference of −0.13 dex, exceeding 5σ, is found for the AGB sample. Moreover, RGB and AGB stars show very similar (well within 1σ) average values of [Fe ii/H], whereas the neutral abundances of AGB stars are significantly lower (by 0.12 dex) than those of the RGB targets.

Zoom In Zoom Out Reset image size

When using photometric gravities, similar results are obtained for titanium, the only other chemical species presenting a large number of neutral and single-ionized lines. For the RGB sample we find [Ti i/H]_RGB = −0.88 ± 0.01 dex (σ = 0.06 dex) and [Ti ii/H]_RGB = −0.92 ± 0.01 dex (σ = 0.05 dex). For the AGB stars we measure [Ti i/H]_AGB = −1.10 ± 0.01 dex (σ = 0.02 dex) and [Ti ii/H]_AGB = −0.95 ± 0.02 dex (σ = 0.06 dex). In this case, the average abundance of AGB stars from neutral lines is lower than that of the RGB sample by 0.21 dex (whereas such a difference amounts to only 0.04 dex for the RGB sample). In Figure 3 we report the differences between the iron (top left panel) and the titanium (top right panel) abundances derived from neutral and from single-ionized lines, as a function of the abundances from the neutral species, obtained for each observed star assuming photometric gravities. Clearly, with the only exception of star 96 (plotted as an empty circle in the figure), the AGB and the RGB samples occupy distinct regions in these planes because of systematically lower values of the AGB abundances derived from the neutral species.

Zoom In Zoom Out Reset image size

Such a difference can also be directly appreciated by visually inspecting the line strengths in the observed spectra and their synthetic best fits. As an example, in Figure 4 we show the observed spectra of an RGB and an AGB star around some Fe i and Fe ii lines, together with synthetic spectra calculated with the appropriate atmospheric parameters and the metallicity derived from Fe ii and from Fe i lines. As is apparent, the synthetic spectrum computed adopting the Fe ii abundance well reproduces all of the observed lines in the case of the RGB star, whereas it fails to fit the neutral features observed in the AGB target, independent of the excitation potential (thus guaranteeing that the effect cannot be due to inadequacies in the adopted temperature). On the other hand, the abundance measured from Fe i lines is too low to properly reproduce the depth of the ionized features of the AGB star. This clearly demonstrates a different behavior of iron lines in AGB and RGB stars.

Zoom In Zoom Out Reset image size

To investigate the origin of the discrepancy between Fe i and Fe ii abundances obtained for the AGB sample, we checked the impact of the adopted stellar mass on the estimate of the photometric gravity. As described in Section 2.2, for the AGB stars we assumed a mass of 0.61 M_⊙, corresponding to the median value of the distribution obtained for HB stars by Gratton et al. (2010), ranging from 0.51 to 0.67 M_⊙. By adopting the lowest mass (0.51 M_⊙), the average value of log g decreases by ∼0.08 dex, but when assuming the largest value, log g increases by 0.04 dex. Such small gravity variations¹⁰ have a negligible impact on the abundances derived from the neutral iron lines, and the impact is still modest (at a level of a few hundredths of a dex) on the abundances derived from single-ionized lines. The only way to obtain (by construction) the same abundance from Fe i and Fe ii lines is to use the spectroscopic values of log g derived from the ionization balance (Section 2.2). However, these gravities correspond to stellar masses in the range 0.25–0.3 M_⊙, which are totally unphysical for evolved stars in GCs.

A possible explanation of the observed discrepancy could be a departure from LTE conditions in the atmosphere of AGB stars. In fact, lines originated by atoms in the minority ionization state usually suffer from NLTE effects, while those originated by atoms in the dominant ionization state are unaffected (see, e.g., Mashonkina et al. 2011). Thus, if this is the case, the most reliable determination of the iron abundance is that provided by [Fe ii/H] because the majority of iron atoms are in the first ionized state in giant stars. Moreover, following Ivans et al. (2001), the degree of overionization of the neutral species should be (at least at a first order) the same as the one affecting Fe i lines. Hence, the correct way to obtain a [X/Fe] abundance ratio is to compute it with respect to the Fe i abundance if [X/H] is derived from minority species, and with respect to Fe ii if [X/H] is obtained from majority species. In the lower panels of Figure 3 we present the [Ti ii/Fe ii] and the [Ti i/Fe i] abundance ratios as a function of the iron abundance derived from single-ionized lines.

As expected, the abundances of AGB stars agree with those of the RGB sample when single-ionized (dominant state) titanium lines are used. For [Ti i/Fe i] a systematic offset of the AGB sample toward lower values is still observable (although reduced), thus indicating the possible presence of residual NLTE effects. We also note a systematic offset of +0.08 dex between [Ti i/Fe i] and [Ti ii/Fe ii], especially for RGB stars. However, taking into account that the oscillator strength values of the Ti ii lines are highly uncertain and that the offset is still reasonably small, we can conclude that the [X/Fe] abundance ratio can be safely constrained either by neutral or by single-ionized lines. It is also interesting to note that the average [Ti i/Fe] and [Ti ii/Fe] abundance ratios (+0.16 dex and +0.25 dex, respectively) of Yong et al. (2014) show a relative offset of −0.09 dex, which is similar to ours but in the opposite direction. This suggests that there is an intrinsic (although small) uncertainty in the zero-point scale of the titanium abundance.

AGB star 96: As apparent from Figure 3, AGB star 96 shows a difference between neutral and ionized abundances, both for iron and titanium, which is incompatible with those found for the other AGB targets, and which is much more similar to the values measured for RGB stars. Interestingly, star 96 presents atmospheric parameters compatible with those spanned by the RGB targets (but with a surface gravity that is 0.15–0.2 lower than that of RGB stars at the same temperature). This case is similar to that encountered in 47Tuc, where the Fe i abundance of a small subsample of AGB stars (four out of 24) has been found to agree with the value obtained from ionized lines, thus suggesting that one of the possible explanations could be the lack of LTE departures for these objects (Lapenna et al. 2014). Also in M22, one (out of five) AGB star shows a perfect agreement between [Fe i/H] and [Fe ii/H], whereas the other AGB stars show systematically low [Fe i/H] values (Mucciarelli et al. 2015b).

In most Galactic GCs, the abundances of oxygen, sodium, magnesium, and aluminum are known to show large star-to-star variations, organized in clear correlations (see Gratton et al. 2012 for a review). These are usually interpreted as the signature of self-enrichment processes occurring in the early stages of GC evolution and giving rise to at least two stellar generations with a very small (if any) age difference, commonly labeled as first and second generations (FG and SG, respectively). In particular, the variations observed in O, Na, Mg, and Al are thought to be due to the ejecta from still unclear polluters, like massive AGB stars, fast-rotating massive stars, or massive binaries (Fenner et al. 2004; Ventura & D'Antona 2005; Decressin et al. 2007; de Mink et al. 2009; Marcolini et al. 2009; Bastian et al. 2013, 2015).

To also verify the presence of these key features in our sample of giants, we derived the abundances of O, Na, Al, and Mg from the observed spectra. The results are shown in Figure 5, where all abundance ratios are plotted as a function of the iron content as measured from the ionized lines. Since the oxygen abundance derived from the forbidden [O i] line at 6300.3 Å is not affected by NLTE, its abundance ratio is expressed with respect to the "true" iron content (measured from Fe ii lines). Instead, the other species are known to suffer from NLTE effects, and their abundances are therefore plotted with respect to Fe i (see Section 3.1). This is true also for sodium, although we have applied the NLTE corrections of Gratton et al. (1999), which take into account departures from LTE conditions driven by overrecombination (Bergemann & Nordlander 2014).¹¹ In any case, we have verified that the same results are obtained if the Na, Al, and Mg abundances are computed with respect to Fe ii or H. In agreement with what is commonly observed in Galactic GCs, we find that the Mg abundance is constant within the uncertainties, whereas O, Na, and Al show significant (several tenths of a dex) star-to-star variations in the RGB sample (see also Yong et al. 2014). As shown in Figures 6 and 7, the observed star-to-star variations are organized in the same correlations observed for GCs. In particular, oxygen and sodium are anticorrelated, independently of using Fe i or Fe ii for the computation of the sodium abundance ratio (Figure 6), whereas aluminum and sodium are positively correlated, and [Al i/Fe i] shows a ∼1 dex spread for fixed magnesium (Figure 7). Very interestingly, instead, all abundance ratios are constant for the AGB sample, with values mainly consistent with those commonly associated with the FG.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The Na–O anticorrelation derived from our RGB sample is qualitatively compatible with that measured by Yong et al. (2014), who found two groups of stars well separated both in [Na/Fe] and [O/Fe]. We note that the oxygen abundances quoted by Yong et al. (2014) are larger than ours, with an average offset of +0.5 dex for the O-rich stars. The origin of this discrepancy can be ascribable to different factors (like atomic data, telluric correction, and so on), but it is beyond the aims of this paper. A good agreement with the results of Yong et al. (2014) is found also for the [Al/Fe] and [Mg/Fe] distributions.

The derived Na–O anticorrelation of M62 is more extended than those observed in most Galactic GCs. Two discrete groups of stars can be recognized, a first one with [O/Fe] ∼ +0.2/+0.3 dex and [Na/Fe] ∼ +0.1 dex, and a second group with [O/Fe] < 0.0 dex (only upper limits) and [Na/Fe] at ∼+0.5 dex. In particular, the subsolar O-poor component (the so-called "extreme population"; see Carretta et al. 2010a) is quite prominent in M62, although these stars are usually rare, observed only in some massive systems, such as NGC 2808 (Carretta et al. 2009c), M54 (Carretta et al. 2010b), and ω Centauri (Johnson & Pilachowski 2010). We also find a significant lack of "intermediate population" stars (with enhanced Na abundances and mild oxygen depletion (Carretta et al. 2010a), which are instead the dominant component in most GCs.

We finally note that the RGB star 89 exhibits a Na abundance of [Na/Fe] = +1.08 dex, which is ∼0.5 dex larger than that measured for all the other O-poor stars. In Figure 8 we compare the spectrum of star 89 with that of another RGB target (id = 95) having very similar atmospheric parameters and iron abundances (see Table 3). As is apparent, all lines have the same strengths, with the notable exception of the two Na doublets, which are significantly stronger in star 89.

Zoom In Zoom Out Reset image size

To our knowledge, this is one of the most Na-rich giants ever detected in a genuine GC (see also the comparison with literature data in Figure 6), with a [Na/Fe] abundance even higher than the most Na-rich stars observed in NGC 2808 (Carretta et al. 2006) and NGC 4833 (Carretta et al. 2014), and comparable to a few extremely Na-rich objects observed in the multi-iron system ω Centauri (Marino et al. 2011).