The Relationship between Globular Cluster Mass, Metallicity, and Light-element Abundance Variations

We investigate aluminum abundance variations in the stellar populations of globular clusters using both literature measurements of sodium and aluminum and APOGEE measurements of nitrogen and aluminum abundances. For the latter, we show that the Payne is the most suitable of the five available abundance pipelines for our purposes. Our combined sample of 42 globular clusters spans approximately 2 dex in [Fe/H] and 1.5 dex in $\log{M_{GC}/M_{\odot}}$. We find no fewer than five globular clusters with significant internal variations in nitrogen and/or sodium with little-to-no corresponding variation in aluminum, and that the minimum present-day cluster mass for aluminum enrichment in metal-rich systems is $\log{M_{GC}/M_{\odot}} \approx 4.50 + 2.17(\rm{[Fe/H]}+1.30)$. We demonstrate that the slopes of the [Al/Fe] vs [Na/Fe] and [Al/Fe] vs [N/Fe] relations for stars without field-like abundances are approximately log-linearly dependent on both the metallicity and the stellar mass of the globular clusters. In contrast, the relationship between [Na/Fe] and [N/Fe] shows no evidence of such dependencies. This suggests that there were (at least) two classes of non-supernovae chemical polluters that were common in the early universe, and that their relative contributions within globular clusters somehow scaled with the metallicity and mass of globular clusters. The first of these classes is predominantly responsible for the CNO and NeNa abundance variations, and likewise the second for the MgAl abundance variations. 47 Tuc and M4 are particularly striking examples of this dichotomy. As an auxiliary finding, we argue that abundance variations among Terzan 5 stars are consistent with it being a normal globular cluster.