2003PASP..115..143K

Assuming that in the atmospheres of low-mass, metal-poor red giant stars, one-dimensional models based on local thermodynamic equilibrium accurately predict the abundance of iron from Fe II, we derive a globular cluster metallicity scale based on the equivalent widths of Fe II lines measured from high-resolution spectra of giants in 16 key clusters lying in the abundance range -2.4<[Fe/H]_II←0.7. We base the scale largely on the analysis of spectra of 149 giant stars in 11 clusters by the Lick-Texas group supplemented by high-resolution studies of giants in five other clusters. We also derive ab initio the true distance moduli for certain key clusters (M5, M3, M13, M92, and M15) as a means of setting stellar surface gravities. Allowances are made for changes in the abundance scale if one employs (1) Kurucz models with and without convective overshooting to represent giant star atmospheres in place of MARCS models and (2) the Houdashelt et al. color-temperature scale in place of the Alonso et al. scale.

We find that [Fe/H]_IIis correlated linearly with W', the reduced strength of the near-infrared Ca II triplet defined by Rutledge et al., although the actual correlation coefficients depend on the atmospheric model employed. The correlations, limited to the range -2.4<[Fe/H]_II←0.7, are as follows:
(1.)[Fe/H]_II=0.531W'-3.279 (MARCS),
(2.)[Fe/H]_II=0.537W'-3.225 (Kurucz with convective overshooting),
(3.)[Fe/H]_II=0.562W'-3.329 (Kurucz without convective overshooting).
We also discuss how to estimate [X/Fe] ratios. We suggest that C, N, and O, as well as elements appearing in the spectrum in the singly ionized state, e.g., Ti, Sc, Ba, La, and Eu, should be normalized to the abundance of Fe II. Other elements, which appear mostly in the neutral state, but for which the dominant species is nevertheless the ionized state, are probably best normalized to Fe I, but uncertainties remain.