C and N Abundances in Stars at the Base of the Red Giant Branch in M15

We present an analysis of a large sample of moderate-resolution Keck Low Resolution Imaging Spectrometer spectra of subgiants and stars at the base of the red giant branch (RGB) in the Galactic globular cluster (GC) M15 (NGC 7078), most within the range 16.5 < V < 19.5 (1.2 < MV < 4.2), with the goal of deriving C abundances (from the G band of CH) and N abundances (from the NH band at 3360 Å). Star-to-star stochastic variations with significant range in both [C/Fe] and [N/Fe] are found at all luminosities extending to the subgiants at MV ∼ +3. The C and N abundances appear anticorrelated, as would be expected from the CN-cycle processing of stellar material. Yet these M15 stars are considerably fainter than the RGB bump, the point at which deep mixing is believed to set in. On this basis, while the observed abundance pattern is consistent with proton-capture nucleosynthesis, we infer that the site of the reactions is likely not within the present sample. The range of variation of the N abundances is very large, and the sum of C + N increases as C decreases. To reproduce this requires the incorporation not only of CN but also of ON-processed material. Combining our work with that of Trefzger and coworkers for the brighter giants in M15, we find strong evidence for additional depletion of C among the most luminous giants. This presumably represents the first dredge-up (with enhanced deep mixing) expected for such luminous cluster RGB stars in the course of normal stellar evolution as they cross the RGB bump. We compare the behavior of these patterns for C and N in GCs covering a wide range of metallicity and current mass. While all clusters studied show strong anticorrelated variations of C and N at all luminosities probed, the metal-rich clusters (M71, 47 Tuc, and M5) do not show evidence for the first dredge-up among their most luminous giants, while the metal-poor ones (M13, M92, and M15, plus M5) do. Conversely, the metal-poor clusters do not show evidence for the bimodality in CH and CN line strengths seen in the metal-rich clusters. The collected data on C and N abundances in low-luminosity GC stars cannot be explained by the commonly invoked models for the chemical evolution of GC stars; in particular, "pollution" of existing low-mass stars by ejecta from intermediate-mass asymptotic giant branch (AGB) stars can be ruled out. Pollution of cluster gas by such stars prior to the formation of the lower mass stars we observe today can also be ruled out unless current models of nucleosynthesis and dredge-up into the surface layers of AGB stars are flawed; such models agree qualitatively but disagree quantitatively with our data. We are forced to assume that there was an extended period of star formation in GCs, and that a previous generation of more massive stars evolved, ejected mass, and polluted the GC gas with light elements; the low-mass stars we see today formed afterward. A tentative scenario is developed involving an initial phase of star formation heavily biased toward high-mass stars, with subsequent formation of intermediate-mass, then low-mass stars.