X-SHOOTER OBSERVATIONS OF MAIN-SEQUENCE STARS IN THE GLOBULAR CLUSTER NGC 2808: FIRST CHEMICAL TAGGING OF A He-NORMAL AND A He-RICH DWARF^*

The main sequence (MS) of the globular cluster (GC) NGC 2808 shows a wide color distribution (D'Antona et al. 2005). Very accurate HST ACS photometry (Piotto et al. 2007) reveals that the MS splits into three sequences that are likely formed in discrete episodes of star formation, slightly separated in age, and with different initial chemical composition. The MS of the massive, multi-metallicity GC ω Cen is also split into two (Bedin et al. 2004) or even more (Bellini et al. 2010) separate branches. NGC 2808 and ω Cen are presently the only GCs where the MS is clearly separated into discrete sequences; see the review by Piotto (2009).

The main role in producing these broad MSs is probably played by helium (e.g., Norris 2004). From the analysis of stacked spectra of 17 stars in each sequence of ω Cen, Piotto et al. (2005) clearly showed that the blue main sequence (bMS) is more metal rich than the red MS (rMS). The most reasonable way to reconcile these observations with stellar evolutionary theory is to suppose that the bMS is populated by stars born with larger helium abundance than the rMS stars. By analogy, He variations provide the simplest explanation for the three distinct NGC 2808 MSs, since the separation cannot be explained easily by age and/or metallicity variations. The three MSs can be fit by theoretical models with He content ranging from a normal Y = 0.24 to an extreme Y = 0.38 value (see Figure 2 in Piotto et al. 2007).

Helium abundances cannot be directly measured by spectroscopy, except for high temperature, highly evolved stars (e.g., Moehler & Sweigart 2006; Villanova et al. 2009). Fortunately, spectroscopic investigations of GCs over several decades have found large star-to-star and cluster-to-cluster variations in other light-element abundances, which strongly indicate the existence of multi-populations. In all clusters studied so far⁹ large star-to-star variations in light elements O, Na, Mg, Al, and Si are present (e.g., Carretta et al. 2009b, 2009c; and see Gratton et al. 2004 for a recent review). The variations have distinctive patterns: O and Mg abundances are positively correlated and are anticorrelated with Na, Al, and Si abundances. Such patterns leave little doubt about the chief nucleosynthesis culprit: high-temperature hydrogen fusion that includes CNO, NeNa, and MgAl cycles. Moreover, Na–O and Mg–Al anticorrelations have been found among unevolved MS and subgiant stars (Gratton et al. 2001), arguing that this pattern is produced by the ejecta of a first generation of now extinct more massive stars (Denisenkov & Denisenkova 1989; Langer et al. 1993). The site of the H burning is still unclear. It might have occurred either in intermediate-mass asymptotic giant branch (AGB) stars (D'Antona et al. 2002) or in fast-rotating massive stars on the MS (Decressin et al. 2007).

The main outcome of H burning, helium, is expected to be directly related to the observed chemical pattern of light elements in GCs. Stars on the MS that are highly enriched in He should have large depletions of O and Mg and large enhancements of Na and Al (possibly Si as well). They should also populate the extreme blue horizontal branch (HB; see Bragaglia 2010 and Gratton et al. 2010) and the bluer MS.

NGC 2808 is the ideal target to derive a clear-cut confirmation of these effects. The close link between He enhancement and the simultaneous depletion/increase in light elements may be investigated in this GC because (1) the three distinct MSs allow unambiguous selection of stars with likely different He contents; (2) this cluster has an unusual HB, strongly bimodal (or even trimodal), and with a long blue (hot) tail (e.g., Bedin et al. 2000) that has been connected to different He enrichments (D'Antona & Caloi 2004); and (3) it shows a very extended Na–O anticorrelation (Carretta et al. 2006), also implying very large He enhancements (see also Bragaglia et al. 2010). As discussed by Piotto et al. (2007), there seems to be a clear connection between the various groups of MS, red giant branch (RGB), and HB stars, according to star counts in the studies of Piotto et al. (2007), Carretta et al. (2006), and Bedin et al. (2000), respectively.

With the advent of X-shooter (Vernet et al. 2009) at the Very Large Telescope (VLT), medium-resolution spectra useful for abundance analysis are currently within reach for faint MS stars of NGC 2808. In this Letter, we present the first results of the chemical tagging of two stars observed in situ on the bluest and the red MS of this cluster.

We selected our targets among the brightest stars for which the separation of the MSs is still possible, since the three MSs merge near the turnoff of the cluster (Figure 1). As confirmed on theoretical grounds (Salaris & Cassisi 2005), the combined effects of larger brightness and shorter lifetime during the MS stage when the initial He content increases largely cancel out. As a consequence, two isochrones of same age and different He contents almost overlap in the turnoff and subgiant regions. From the ACS photometric catalog (Piotto et al. 2007), we selected two MS stars in NGC 2808 that (1) have proper motions typical of cluster stars, (2) have a probability >85% of being on the bMS and >95% of being on the rMS, based on their magnitude and colors, respectively, and (3) do not have neighbors closer than 15.

Zoom In Zoom Out Reset image size

As part of the Italian X-shooter guaranteed time observations (GTO), we observed these two MS stars with the X-shooter spectrograph at VLT-UT2 on 2010 January 22–23. The wavelength coverage of X-shooter ranges from the atmospheric UV limit to the near-infrared. For our observations, the slit width was set at 1'' or 08, but the resolution (R∼ 10,000) was mostly dictated by the sub-arcsecond seeing, especially for the second night. We took four and five 1 hr exposures for star 5 and 6, respectively. Information on the targets is given in Table 1 and their position in the color–magnitude diagram (CMD) is shown in Figure 1. The observations were optimized for the UV and visual part of the spectrum; i.e., we did not use the nod option, essential to subtract the sky in IR. Therefore, we use here mostly the UV/blue spectrum (∼3300–5500 Å) and part of the visible (∼5500–10000 Å). The spectra were reduced using the preliminary X-shooter pipeline (v0.9.4; see Goldoni et al. 2006) and standard IRAF¹⁰ routines. Each stellar exposure was bias-corrected and flat-fielded, calibrated in wavelength, then extracted and corrected for sky background. Our targets are near the limit of the instrument response, and the sky level is comparable to the star signal (or higher than it, in the presence of even a small Moon contribution). The individual spectra have signal-to-noise ratio (S/N) from about 10 to about 20. They were combined, weighting them with their S/N, and shifted to zero radial velocity. The radial velocities were measured using about 35 lines; the heliocentric values are 75 and 80 km s⁻¹ (rms 10 km s⁻¹), for the rMS star and the bMS star, respectively. Given the resolution and the uncertainty due to the use of a (not filled) slit, they well compare to the value of 93.6 km s⁻¹ reported by Harris (1996).

To derive temperatures and gravities, we used a combination of photometric information and isochrone fitting. Isochrones (Pietrinferni et al. 2006), distance modulus, and reddening are those used in Piotto et al. (2007) to fit the discrete MSs observed in NGC 2808. The isochrone consistent with the rMS position of the first star has a canonical He abundance (Y = 0.248) and that for the bMS position of the second star has enhanced He (Y ≃ 0.40). In the current interpretation, rMS star is of first generation and bMS star of second generation. The individual values for T_eff and log g are indicated in Table 1. For both stars, we adopted the NGC 2808 metallicity derived from high-resolution spectra by Carretta et al. (2006), [Fe/H] = −1.1, and assumed a microturbulent velocity v_t = 0.8 km s⁻¹. These are reasonable assumptions, given the identical evolutionary states and the internal homogeneity in heavy elements and metallicity.¹¹ The exact value of v_t does not strongly influence the results presented here.

The main interest of our analysis lies in the differential analysis of the two stars. However, the temperatures adopted are supported by the Balmer lines, and we confirmed the assumed metallicity by computing synthetic spectra near a few Fe i lines with good gf values. We did not explicitly took into account an enhancement in He in the stellar atmospheres; however, the expected effect on iron abundance determination is negligible (see discussions in Carretta et al. 2006; Bragaglia et al. 2010) and this is most probably true for the other elements. Although we did not do a detailed analysis of other heavy-element species, we noted that lines of, e.g., Ti and Ca have approximately the same strength in both stars. None of the suggested pollution sources that contribute hydrogen-burning products to newly forming stars should contribute elements beyond Si. Heavier α or Fe-peak abundances should be the same in all NGC 2808 stars, and our spectra do not contradict this expectation.

Adopting the stellar parameters defined above, we used standard routines to compute synthetic spectra for some particularly interesting elements: N (from the NH feature at 3360 Å), Al (from the 3961 Å resonant line), Mg (from the Mg b lines near 5180 Å), C (from the CH features in the G band near 4300 Å), and Na (from the 8183–94 Å doublet). The spectral syntheses shown in the figures were computed with the LTE spectroscopic analysis code ROSA (Gratton 1988) for the atomic lines (Al, Mg, and Na) and MOOG (Sneden 1973) for the molecules (NH, CH). However, all the synthesis work was checked independently using both codes; results are in very good agreement.

Abundances for all elements are presented in Table 1. The derived values have conservative error estimates (mostly due to the uncertain continuum placement) of 0.1 dex for Na (and Fe), and 0.2 dex for N, C, Mg, and Al. We stress however that the main result of our analysis lies in the difference between the abundance patterns of the two stars, more than in the absolute values for the chemical abundances. We will show that the light-element differences between rMS star and bMS star exceed their uncertainties.

Comparison of observed and synthetic spectra are shown in Figure 2 for NH, Al i, CH, and Mg i features. From the closest observed/synthetic matches, we estimated the abundances that are given in Table 1. The expectations are that N and Al should be increased and C and Mg should be decreased in bMS star with respect to the values for rMS star (the one of supposedly normal, primordial composition), following what has been found for evolved RGB stars (e.g., Ivans et al. 2001; Cohen et al. 2002; Ramírez & Cohen 2003; Carretta et al. 2009b). Figure 2 demonstrates that the two stars have different light-element spectra. This is most obvious for N: the NH absorption is much stronger in bMS star than in rMS star. Neither details in the spectrum normalization nor (small) differences in the atmospheric parameters can account for this difference.

Zoom In Zoom Out Reset image size

The Al abundance was derived only from the 3961 Å resonance line, since its doublet partner at 3944 Å is a blend (Arpigny & Magain 1983). The synthesis was computed adopting the Ca abundance appropriate for NGC 2808 ([Ca/Fe] = +0.34; Carretta et al. 2009b) to reproduce the Ca ii H and K lines. The values for Al and Mg given in Table 1 are corrected for NLTE effects according to Gehren et al. (2004); the corrections are about +0.5 dex for Al and +0.06 dex for Mg, respectively, for both stars.

The huge abundance of N found for bMS star (which also had decreased C) can be explained only with the transformation of (virtually) all oxygen into nitrogen. Our findings seem to indicate that we are seeing, in the gas from which this star formed, the outcome of the complete CNO cycle. Actually, if we combine the [C/Fe] and [N/Fe] values of Table 1 with the solar C, N, and O abundances by Asplund et al. (2009) and with the maximum [O/Fe] ratio for RGB stars in NGC 2808 (Carretta et al. 2006), even large depletions of O ([O/Fe] <−1) cannot reproduce a constant sum of the CNO elements. This would be reproduced by assuming [N/Fe] ∼1.4 for bMS star. We note that a systematic offset of ∼0.6 dex in N abundances from our analysis would produce a roughly solar scaled [N/Fe] ratio for rMS star, which would agree fairly well with the values usually assumed for field halo stars (Gratton et al. 2000). Once again, what is most important is the difference between the derived abundances and this is a sound result.

Unfortunately, it was not possible to measure O abundances for these two stars, since the O triplet at 7771–7774 Å is weak and falls in a wavelength region where the sky subtraction is difficult for these very faint objects. However, we were able to estimate the Na abundances. Figure 3 shows observed and synthetic spectra surrounding the 8183–8194 Å Na i lines. We see in the figure that the Na lines are stronger in bMS star than in rMS star and this is reflected in the abundance ratios indicated in Table 1. These abundances include NLTE corrections (about −0.1 dex) as recommended by Gratton et al. (1999). The chief limitation is the strong telluric-line contamination in this spectral region. The tellurics were eliminated by division of the program star spectra with that of a hot, rapidly rotating star, using an IRAF routine for this task. The cleaning quality is much better for the bluest of the two lines, so that our Na abundances rest on that single feature. They are qualitatively confirmed by the other line and by the relative strengths of the Na D lines. Unfortunately, strong interstellar absorption and sky emission made the derivation of Na abundance from the D lines less secure, given the radial velocity at the time of observation and the moderate resolution of the X-shooter spectra.

Zoom In Zoom Out Reset image size

The present Al and Mg results are in good agreement with those found from high resolution UVES spectra of 12 red giants in NGC 2808 by Carretta et al. (2009b). In Figure 4, we plot the Mg–Al anticorrelation for the RGB stars and the two MS stars analyzed here. Apart from a possible small zero-point effect due to the use of different lines, corrections for NLTE, etc., the two MS stars do nicely participate in the same trend defined by the giants. The rMS and the bMS star fall in the Mg-rich/Al-poor and Mg-poor/Al-rich groups, respectively. This result indicates that the extreme abundance pattern of the Al-rich, Mg-poor stars in NGC 2808 must have been produced by pollution from a previous stellar generation in the clusters. Deep mixing has been recently revisited as an explanation for the extreme chemical abundances of RGB stars in GCs (see, e.g., D'Antona & Ventura 2007; Lee 2010). However, since deep mixing cannot have been responsible for the light-element abundances in the bMS star, it is unlikely to have caused the identical pattern in evolved RGB stars of NGC 2808.

Zoom In Zoom Out Reset image size

Finally, it is interesting to note that the Ba abundance, as indicated by the resonance Ba line at 4554 Å, seems the same for the two MS stars. If confirmed by more quantitative analysis, this would probably exclude a significant contribution from low-mass AGB stars (e.g., Yong et al. 2009) to the pool of gas from which the second-generation stars originated. The consequence is that the handful of stars observed to have strong enhancement in Ba and other s-process elements in some GCs must have another origin, likely from mass transfer from a former AGB companion in a binary system (see V. D'Orazi et al. 2010, in preparation).

In conclusion, the chemical pattern we found from the first abundance analysis of a star on the He-rich MS of NGC 2808 is exactly what is expected if stars on the bMS formed from ejecta produced by an early stellar generation via proton-capture reactions in H burning at high temperature, accompanied the main outcome of this nuclear burning, i.e., helium. The extreme Al enhancement and Mg depletion observed in the bMS star argues against a deep mixing hypothesis for the extreme chemical abundances of RGB stars in GCs.

Observations of a larger sample of unevolved stars in this cluster and others with suspected He variations would be welcome.

We thank Paolo Molaro for the nice observations made on behalf of the Italian X-shooter GTO team, and Valentina D'Odorico for help and suggestions with the data reduction. The organizing work of Paul Groot and Sofia Randich is acknowledged. Funding come from PRIN-MIUR 2007, the Italian GTO X-shooter Consortium, and US National Science Foundation grant AST-0908978. This Letter is dedicated to Roberto Pallavicini, late Italian PI of the X-shooter Consortium, who spent a lot of time and work for the realization of this fine instrument.