SUPER-AGB–AGB EVOLUTION AND THE CHEMICAL INVENTORY IN NGC 2419

Photometric and spectroscopic analysis of globular cluster (hereafter GC) stars in the last several decades have changed the traditional framework describing the stellar content of these systems; it is now clear that most GCs harbor multiple stellar populations, differing in their original chemistry and characterized by a spread in light elements such as aluminum, sodium, oxygen, and magnesium (Gratton et al. 2012 and references therein). These results were reinforced by high-quality photometric data, showing that some GCs harbor multiple main sequences that can be interpreted by invoking the presence of a helium-enriched population (Norris 2004; D'Antona et al. 2005; Piotto et al. 2007; Milone et al. 2012). This finding is in agreement with the suggestion that helium could be the second parameter traditionally invoked to account for the anomalous shape of the horizontal branch (HB) of some GCs (D'Antona et al. 2002; Caloi & D'Antona 2005).

The above results indicate that the stars with anomalous chemistry were born from matter contaminated by advanced p-capture nucleosynthesis, exposed to temperatures sufficiently large (exceeding ∼100 MK) to activate the CNO, Ne–Na, and Mg–Al nuclear channels (Prantzos et al. 2007). Pollution by ejecta of rapidly rotating massive stars (Decressin et al. 2007a, 2007b), massive binary stars (De Mink et al. 2009), or intermediate mass asymptotic giant branch (AGB) and super asymptotic giant branch (SAGB)⁵ stars (Cottrell & Da Costa 1981; Ventura et al. 2001) are among the scenarios suggested in the literature. In this work, we follow the hydro-dynamical model for the formation of multiple populations by D'Ercole et al. (2008) and focus on the hypothesis that AGB and SAGB stars via the strong mass loss at low velocity experienced during their post core helium-burning evolution, produced the polluted material, from which the new stars formed.

Note that Mg depletion, which is an important signature of the extreme anomalies we will be dealing with, is a severe constraint on the scenarios, as it cannot be found in the ejecta of massive stars (Decressin et al. 2007b), or of massive binaries, subject to similar nucleosynthesis limitations. Only SAGB models with non-extreme mass loss rates can predict it (Siess 2010) or massive AGB models with efficient convection (see Ventura et al. 2012). We point out that modeling of AGB and SAGB evolution is subject to considerable uncertainty concerning the efficiency of the hot bottom burning (HBB) and of the third dredge up. Consequently, the AGB–SAGB scenario can explain the chemical patterns in GCs only for those model providing compatible yields. Adoption of an efficient convection (see the discussion in Ventura & D'Antona 2005) provides both reasonable sodium yields and oxygen depletion (plus some Mg depletion at low metallicity), while models with low efficiency of convection and strong dredge up (Karakas & Lattanzio 2003; Karakas 2010; Herwig 2004; Stancliffe et al. 2004) are at variance with the AGB–SAGB scenario (Fenner et al. 2004).⁶

D'Ercole et al. (2010, 2012) showed that the abundance patterns observed can be reproduced provided that the star formation of the second generation (SG) starts immediately after the epoch of the SNII explosions, and is limited to the first ∼100 Myr, thus involving only stars of mass M ⩾ 5 M_☉. The spread observed in the O–Na and Mg–Al planes can be explained by dilution of gas ejected by AGBs with pristine gas present in the cluster. Of extreme interest is the case of massive GCs, where conditions occur for the early formation of a stellar population directly from the winds of SAGBs, with no dilution; these stars are characterized by a strong helium enhancement (Y ≳ 0.35), and populate the blue main sequence (MS) and the blue tail of the HB of their host clusters. D'Ercole et al. (2012) showed that the SAGB yields are a very important ingredient of the models for clusters of intermediate metallicity ([Fe/H] ∼−1.5, i.e., Z = 10⁻³). Here we extend our study of the yields from massive AGBs and SAGB stars to the much lower metallicity of Z = 3 × 10⁻⁴. This metallicity roughly corresponds to [Fe/H] ∼−2, the iron content of NGC 2419. This cluster is characterized by an HB hosting a blue, faint population, clearly detached from the main component (Ripepi et al. 2007), which can be explained only by invoking a stellar component greatly enriched in helium (Di Criscienzo et al. 2011a). Thus, according to the model by D'Ercole et al. (2008), these stars should be born directly from the ejecta of SAGB and from the most massive AGB stars. Consequently, the yields computed here should be directly relevant to the chemical composition of the anomalous stars in this cluster. The formation of the SG can be different in other very low metallicity clusters such as M 15 (Sneden et al. 1997), showing non-extreme HB morphology, so here we concentrate on the comparison with NGC 2419 only, and compare our yields to the abundances of O, Na, Mg, Al, Si, and K recently made available by Cohen & Kirby (2012) and Mucciarelli et al. (2012).

We further explore the possibility that the bimodal potassium abundances recently discovered by Mucciarelli et al. (2012) in NGC 2419 are a signature of the production of potassium by proton capture on the argon nuclei in the same HBB environment that produces the other abundance anomalies. This can be a powerful indication that the massive AGB/SAGB scenario is indeed operating for the formation of multiple populations.

The yields from massive AGB stars of interest in this work are determined by HBB, i.e., an advanced, p-capture nucleosynthesis active at the bottom of the convective envelope (Blöcker & Schöenberner 1991). Ventura & D'Antona (2005) showed that the use of the Full Spectrum of Turbulence (Canuto & Mazzitelli 1991) model for turbulent convection, coupled with the treatment of mass loss by Blöcker (1995), easily leads to HBB conditions for all stars with initial mass M > 4 M_☉.

Here we present the yields of AGB and SAGB models experiencing HBB for Z = 3 × 10⁻⁴, [α/Fe] = +0.4, and put into evidence the difference with the Z = 10⁻³ models discussed in Ventura & D'Antona (2009, 2011). A detailed discussion of these models will be presented in a forthcoming paper. To allow a more direct comparison with the observations, we show for the ith species the quantity [i/Fe] = log (X_i/X_Fe) − log (X_i/X_Fe)_☉.

2.1. Oxygen and Sodium

The degree of the HBB experienced can be deduced by the extent of the oxygen depletion; among all the elements considered, oxygen is the only one whose nuclear activity is a pure destruction process, different from sodium and aluminum, for which both creation and destruction channels are active, and magnesium, whose nucleosynthesis complicated by the distribution of the overall magnesium content among the three isotopes (Ventura et al. 2012). In the bottom-right panel of Figure 1 we show the typical O–Na pattern for the models from 4 to 7.5 M_☉. Similar to the Z = 10⁻³ case, the degree of the p-capture nucleosynthesis does not increase monotonically with mass and reaches a maximum around the threshold of ∼6 M_☉⁷ separating the AGB from the SAGB regime. The reason (see discussion in Ventura & D'Antona 2011) is in the high mass loss rate of SAGBs, which consumes the envelope before a very advanced nucleosynthesis is experienced. In the panel we note the straight correlation between the oxygen and sodium content of the ejecta. The sodium yields are positive due to the initial phase of sodium synthesis via proton capture onto ²²Ne, whereas the correlation is due to simultaneous destruction of oxygen and sodium once the temperature at which the nucleosynthesis occurs exceeds ∼80 MK. The maximum depletion in the surface oxygen reaches [O/Fe] ∼ −1.2 dex at Z = 3 × 10⁻⁴, while it was limited to ∼ − 0.8 dex at Z = 10⁻³ (Ventura & D'Antona 2009, 2011). This is a direct consequence of the increasing temperatures at which the bottom of the surface convective zone is exposed for models of decreasing metallicity. Note that in the models with maximum oxygen depletion there cannot be any sodium enhancement.

Zoom In Zoom Out Reset image size

To illustrate the dependence of the results on the mass loss treatment, and to compare our results with the investigation by Siess (2010), based on smaller rates of mass loss in the high mass domain, for the 6 M_☉ and 7 M_☉ models we show the results obtained by reducing the mass loss rate to 1/4: thanks to the longer time spent in the HBB phase, oxygen depletion reaches −1.7 dex for the 6 M_☉, while sodium is not touched in a significant way. A larger O depletion (and further Na depletion) is also found in the 7 M_☉ evolution. Sodium remains quite large, [Na/Fe] > +0.3, only in the mass range 7.2–7.5 M_☉.

2.2. Magnesium, Aluminum, and Silicon

2.3. Helium

The large distance of NGC 2419 (87.5 ± 3.38 Kpc from the Sun; Di Criscienzo et al. 2011b) has precluded an extensive abundance analysis of its stars. However, some fundamental properties have emerged from the work of Cohen et al. (2010, 2011) and Cohen & Kirby (2012), showing the absence of a spread in iron, indicating that the system evolved like a typical GC. Cohen & Kirby (2012) provide abundances for a total of 13 giants, showing the presence of stars with very low Mg abundances, and confirm recent additional data by Mucciarelli et al. (2012) showing an interesting anticorrelation Mg–K. Figure 1 adds the Cohen & Kirby (2012) data to the theoretical patterns. Cohen & Kirby (2012) noticed that the extreme Mg depletion of some stars did not correspond to a large increase in the Al abundance. Our models, however, show that the stronger is Mg burning, the more moderate the Al abundance tends to be, as nucleosynthesis proceeds toward nuclei of larger atomic number, with Si production. This is shown clearly by comparing the standard models to those with reduced mass loss, where the time to proceed to more advanced nucleosynthesis is longer, and the Al yields are in the range shown by the Mg-poor giants.

In the panels of Figure 1, we subdivide Cohen & Kirby data into two main groups: Mg-poor (black dots) and Mg-rich (open dots) giants. We also point out three possibly "peculiar" giants with moderately large sodium: two giants are shown as asterisks, while a third (S1673 in Cohen & Kirby), very Mg-poor, is shown as a triangle.

First of all, we note that the oxygen abundances are determined only in a few stars having large abundances. We draw the upper limits in the panel O–Na at arbitrary locations [O/Fe] = 0.25–0.5 to avoid superposition of the points. In this way, we show that the sodium variation encompasses almost 1 dex, but most points cluster at low [Na/Fe] ≲ 0.3. The Mg-depleted giants also show low sodium (top-right panel): these data are compatible with our models showing the strongest HBB processing (5–6 M_☉). S1673 (shown as a triangle) is an exception.

The open points have normal O, Na, and Mg, and thus should be plain "first-generation" (FG) stars; of the two giants denoted as asterisks, one is Mg-normal, Si-normal, so it is probably an FG star. The other giant (S1305) is very Na-rich ([Na/Fe] = 0.66 ± 0.15) and also Si-rich, but not Mg-poor; we tentatively suggest that it is an SG star formed by SAGB ejecta in the 7.2–7.5 M_☉ range, stars that did not have enough time to deplete Mg by much, but were able to increase Si.

Although there are very limited data available, the observations currently available suggest that the SG of NGC 2419, identified by Di Criscienzo et al. (2011a), with the blue hook stars can be directly formed from the SAGB–massive AGB ejecta. Only the Na-rich, Mg-poor giant S1673 is out of any qualitative interpretation, as the sodium survival is at variance with the strong Mg burning and Si production. We predict that the oxygen abundance in the stars with low Mg of NGC 2419 must be extremely low.

3.1. Understanding K Production in the HBB Phase

Mucciarelli et al. (2012) discovered the presence of two distinct, well-separated populations in NGC 2419, differing in their magnesium and potassium contents, with the Mg-poor population significantly enriched in potassium by a factor ∼10. In Figure 2 we plot their data together with the data by Cohen & Kirby (2012) which confirm the K-dichotomy. Note also that in this plot the giant shown as a (black) triangle is very rich in K, confirming the advanced nucleosynthesis (and the difficulty in explaining the high sodium).

Zoom In Zoom Out Reset image size

Potassium is produced by proton capture on the argon nuclei. The reactions involved start from ³⁶Ar, an α-nucleus that will be the most abundant form of argon at so low metallicities, as shown by models of galactic chemical evolution (Timmes et al. 1995; Kobayashi et al. 2011). The chain is³⁶Ar(p, γ) ³⁷ K(e⁺, ν) ³⁷Cl(p, γ)³⁸Ar(p, γ)³⁹ K. We checked the relevant cross sections by means of the NETGEN tool (Aikawa et al. 2005); all of them increase by orders of magnitude at temperatures above 10⁸ K, so that the details of the envelope structure of our AGB and SAGB models can be relevant whether or not they achieve the required nucleosynthesis. In order to reach the K production, we had to increase the cross section ³⁸Ar(p, γ)³⁹ K by a factor of 100; a similar result can be achieved with a stronger HBB, leading to temperatures at the base of the envelope exceeding ∼150 MK. As the initial argon largely exceeds the initial potassium content, even a soft activation of the Ar-burning reactions may favor a great increase in the potassium abundance. The result of exploratory evolution of the 6 M_☉ can be seen in the top panel of Figure 2, where the increase in the surface ³⁹K and the concomitant ³⁸Ar burning is shown as a function of [Mg/Fe] for the standard track with an increased cross section, and for the speculative case in which the mass loss rate is reduced to 1/4 of the standard rate. In this latter case, Mg reduction and K production are maximized. The main panel of Figure 2 shows the Mg and K abundances in the sample by Mucciarelli et al. (2012) and in the Cohen & Kirby (2012) sample. The double pentagons show the yields of the 6 M_☉ evolution, (1) in the standard case (unchanged cross sections), (2) in the case of increased cross section, and (3) in the case of increased cross section and reduced mass loss. The models show that this path is worth a more complete exploration.

Thus far we have only discussed the elemental abundances in Cohen & Kirby (2012) for which HBB provides a possible nucleosynthesis path. Nevertheless, it is necessary to touch on the problem of Calcium. Cohen et al. (2010, 2011) found larger Ca abundances in their Mg-poor sample. Mucciarelli et al. (2012) questioned whether the difference in abundance could be attributed to the different atmospheric structure in the presence of such Mg differences, but Cohen & Kirby (2012) discuss in depth that this cannot be the case. So we must accept the Ca variations. Where do these variations come from? At low metallicity, supernova (SN) ejecta contain more α-elements than iron (Kobayashi et al. 2006), a result dependent on the assumptions made on the mass cut (e.g., Limongi & Chieffi 2003). A small contamination from SN ejecta could then be a viable solution to the calcium problem, although a detailed model should be produced to determine whether this is feasible, maintaining the lack of iron variations and the magnesium depletion. Alternatively, we stretch our results and leave this problem open, by asking whether the same p-capture chains could reach ⁴⁰Ca from ³⁹K.

The abundances of light elements observed in NGC 2419 have been compared to the chemical patterns expected for SAGB and massive AGB evolving through HBB and mass loss, and having the low metallicity of this cluster. We directly compare the yields with observations, following the idea that the SG stars in this cluster are born directly from SAGB plus massive AGB ejecta (Di Criscienzo et al. 2011a). The low metallicity of the models allows a very strong HBB, especially for masses around 6 M_☉, at the edge between the AGB and the SAGB regime. These ejecta are thus predicted to produce the most extreme contamination, with a strong depletion of oxygen, a significant reduction of the initial magnesium, and only a modest increase in sodium. The models are compatible with the abundance trends in NGC 2419. The presence of a K-rich population of Mg-poor stars in this cluster is a signature of extreme nucleosynthesis, and we have shown that potassium can be produced by proton capture on Argon nuclei if the relevant cross section is higher (by a factor of 100) than the standard rate.

The results should be tested by a more detailed spectroscopic investigation. Confirmation that the Mg-poor stars are extremely depleted in O and have normal Na would support our model.

We thank Marco Limongi for useful discussion. This work has been supported by PRIN–INAF 2011: Multiple populations in globular clusters: their role in the Galaxy assembly (PI: E. Carretta). E.V. was supported in part by grant NASA-NNX10AD86G.