CN-Strong Models for Globular Clusters

We have seen in the previous section that the base models do an excellent job of reproducing the Balmer, Fe, and Mg indices for one metal-poor and one mildly metal-rich Galactic globular cluster. The same level of agreement was not reached, however, for the carbon, nitrogen and calcium-sensitive indices. In order to address this problem we computed a new set of models adopting the known abundance pattern of 47 Tuc, following the procedure outlined in Section 4.3.2, so as to generate CN-strong models for this cluster. Such models may be very useful given the fact that the CN-strong phenomenon in integrated spectra of globular clusters seems to be ubiquitous, being probably caused by very high nitrogen abundances (e.g., Burstein et al. 1984, Brodie & Huchra 1991, Papers I and II, Li & Burstein 2003, Beasley et al. 2004, Burstein et al. 2004).

The elemental abundances adopted for 47 Tuc are those listed in Table 25. Because carbon and nitrogen have a bimodal distribution, we computed two sets of models, corresponding to the CN-strong and CN-weak abundance patterns. These models were then combined, with weights determined by the relative numbers of CN-strong and CN-weak stars in the core of 47 Tuc, from Briley (1997), who found CN-strong/CN-weak $\sim$ 2. In view of the mass-segregation effects discussed in the previous section, we adopted a dwarf-poor IMF in these computations, with x=-4 (equation 3).

In Figure 24 these models are compared with our base models (1-5 in Table 24, in a few interesting index-index plots. Gray lines represent the base models, while dark lines represent the CN-strong models described above. Same-[Fe/H] models are connected by dotted lines. As expected, CN-strong models are characterized by stronger CN$_2$, and weaker carbon indices (G4300 and C$_2$4668). The behavior of Ca4227 is more complex. It is weaker in metal-poor models, where [Ca/Fe] and [O/Fe] in the CN-strong and base models are similar, so that the differences arise purely due to the stronger CN strengths in the former. At higher metallicities, Ca4227 becomes stronger in the CN-strong models because those have higher [Ca/Fe] and [O/Fe] than the base models, by 0.1 and 0.6 dex, respectively.

Also shown in Figure 24 are line indices in the spectrum of 47 Tuc. It can be seen that in all plots 47 Tuc is well matched by the 14 Gyr-old, CN-strong, models with [Fe/H]=-0.7 (the second dotted line from left to right), except for the case of G4300. For Ca4227, C$_2$4668, and CN$_2$ the CN-strong models are a vast improvement over CN-normal models (the same is true of CN$_1$, not shown). In the case of Ca4227, this result highlights an important fact, seldom appreciated in the literature: the Ca4227 index is very sensitive to the abundances of carbon and nitrogen, due to a contamination of its blue pseudo-continuum window by a CN band-head. In order to circumvent this problem, Prochaska et al. (2005) propose the definition of a new index, Ca4227r, which is far less sensitive to the CN contamination, thus being a more reliable indicator of the calcium abundances. See Prochaska et al. (2005) for details. Finally, we note that adoption of these CN-strong models only affects predictions of these carbon/nitrogen-sensitive indices, so that the quality of the match to all other indices is the same as in the previous plots. Models computed for both the CN-strong and CN-normal abundance patterns of 47 Tuc stars (Table 25) and a dwarf-depleted IMF ($x=-4$) are provided in Tables A and A in the Appendix.