Lower metallicity limit of the galactic globular cluster system: Calcium triplet spectroscopy of metal-poor globular cluster giants

Medium-resolution spectra at the Ca II infrared triplet are presented for 164 giants in 12 metal-poor Galactic globular clusters. The data for 4 of these clusters are combined with published data for 7 other clusters with well-known metal abundances and used to calibrate the derivation of metallicity values for the remaining clusters from the summed strength of the 2 strongest Ca lines at a given V - V_HB. The mean [Fe/H] values we find are: -1.82 for NGC 2298, -2.10 for NGC 4372, -1.84 for NGC 4833, -2.10 for NGC 5053, -1.87 for NGC 5694, -1.94 for NGC 5897, -1.86 for NGC 6101 and -1.86 for NGC 6144, with a typical mean error of 0.05 dex from an average of 10 member giants per cluster. Upper limits to any intrinsic intracluster metallicity dispersion are ≤0.05 dex in all of the clusters except ∼0.15 dex in NGC 5053 and NGC 5694, where the higher limits are due to the lower quality spectra for these more distant clusters. The mean metallicity values for most of the program clusters are in excellent agreement with the Zinn [ApJ, 293, 424 (1985)] scale, while those for NGC 5897 and NGC 5053 are 0.26 dex lower and 0.48 dex higher than given by Zinn, respectively. The value for NGC 5897 is intermediate to that given by the Washington photometry of Geisler et al. [AJ, 104, 627 (1992b)] and Zinn. The high value for NGC 5053 increases the already very significant difference between the metallicity distribution functions of halo clusters and field stars. Our results, combined with previous determinations for other clusters, indicate that the lower metallicity limit for globular clusters in the Galaxy is -2.25±0.10. A similar value also appears to hold for other nearby globular cluster systems. We also derive mean cluster radial velocities to a typical precision of 1.2 km/s (mean error). The values for several of the program clusters are significantly different from previously published values. An updated analysis slightly strengthens van den Bergh's [AJ, 105, 971 (1993)] finding that halo clusters on retrograde orbits have different horizontal branch morphologies than clusters on prograde orbits.