Keck Hires Spectroscopy of M92 Subgiants: Surprising Abundances near the Turnoff

Using high-resolution, moderate signal-to-noise ratio spectroscopy obtained with the 10 m Keck I Telescope and efficient HIRES echelle spectrograph, we derive abundances of several elements in subgiants near the M92 turno†. As a consistency check, we also analyze the metal-poor Ðeld star HD 140283 and Ðnd an Fe abundance in Ðne agreement with many previous determinations. However, our M92 value ([Fe/H]\ [2.52) is a factor of 2 lower than the abundance derived from its red giant members. Di†erences in model atmospheres, gf-values, and instrumental e†ects might account for this di†erence, but whether they in fact do so is unclear. We note possible evidence for [Fe/H] di†erences within M92. Our spectroscopic analysis suggests that the M92 reddening, E(B[V ), may be 0.04È0.05 mag greater than canonical values, but various uncertainties mean that this conclusion is not deÐnitive ; the signiÐcant di†erence in interstellar Na I line strengths in the M92 and HD 140283 spectra may be consistent with an increased reddening. Regardless, the conclusion that either the [Fe/H] of M92 has been signiÐcantly overestimated from red giants or current reddening/photometry estimates are too small/red is not easily escaped. If the reddening/photometry were in error by this amount, turno† colorÈ based ages for M92 could be reduced by D4 Gyr. The adjustment to the M92 distance modulus required for a similarly reduced turno† age that is luminosity-based can be accommodated by increases in extinction and alterations to the metal-poor Ðeld star distance scale recently inferred from Hipparcos Cepheid and subdwarf data. Our M92 subgiants demonstrate [Cr/Fe], [Ca/Fe], and [Ti/Fe] ratios that are unremarkable and essentially identical to the values for HD 140283. [Ba/Fe] is 0.45 dex larger for the M92 subgiants than for HD 140283. Surprisingly, we Ðnd [Mg/Fe] to be 0.55 dex lower in our M92 subgiants than in HD 140283, and [Na/Fe] to be 0.76 dex larger in our M92 subgiants than in HD 140283. These di†erences (and indeed nearly all our abundance ratios) seem immune to various data, analysis, and parameter errors. If real, this striking abundance pattern is suggestive of material in our M92 starsÏ photospheres that has undergone Ne ] Na and Mg ] Al cycling like that inferred for red giants in M92 and other clusters. While this is generally believed to be an in situ process in cluster giants, the presence of abundant Li in our M92 objects suggests a polluting source acting either primordially or via accretion after cluster star formation. This may be consistent with CN and Na variations on the 47 Tucanae main sequence, recently reported Ba and Eu variations in M15 red giants, possible cluster-to-cluster n-capture abundance di†erences, and very low [O/Fe] ratios observed near the base of the M13 giant branch. We thus suggest that a polluting source of light-element alteration, in addition to the in situ source for more evolved stars, may be required for M92. Comparison of our M92 subgiant abundance ratios with those of M92 red giants may indicate that pollution occurred after the present generation of cluster stars formed, but until the cause or causes of the subgiant versus giant Fe abundance discrepancy are deÐnitively identiÐed, this conclusion is uncertain. A polluting source of our Na and Mg anomalies produced via processing in a previous stellar generation also has complications ; namely, how the Mg and Na anomalies arise without apparently any net inÑuence on our subgiantsÏ Li abundances and on the C abundances of other M92 subgiants. A similar quandary may exist in some 47 Tuc turno† stars. An understanding of cluster abundance variations (by whatever mechanisms) and their behavior with evolutionary state may be needed for a complete understanding of absolute and relative globular clusters ages, and for derivation of the primordial Li abundance.