Only lines clean from blends on very high resolution solar spectra were considered in the analysis; for Fe, the line list was extracted from Rutten & van der Zalm (1984), and Blackwell et al. (1980). The adopted values for the oscillator strengths gfs were determined following the same precepts of Clementini et al. (1995), who performed a high-dispersion study of the metal abundances of field RR Lyrae variables. Briefly, whenever possible laboratory gfs were considered: for Fe I lines they were taken from papers of the Oxford group (for references, see Simmons & Blackwell 1982) and Bard et al. (1991) and Bard & Koch 1994), gfs of the Oxford group being corrected upward to account for the systematic difference with those of Bard et al. (0.03 dex; see Clementini et al. 1995). gfs for Fe II lines were taken from Heise & Kock (1990), Biémont et al. (1991) and Hannaford et al. (1992). For lines lacking accurate laboratory determinations, gfs were derived from an inverse solar analysis using the Holweger & Müller (1974: HM) model atmosphere and the Fe abundance derived from the other lines. The adopted gf values are also listed in Table 4 (click here); they can be integrated for other lines using gfs in the larger line list of Clementini et al. (1995, Tables 3a, b).
In any differential analysis (i.e. a comparative analysis in which the
zero point of the [Fe/H] scale is set by the Sun) the assumed solar abundance
is of paramount importance. The solar Fe abundance obtained with the set of
gfs described above and the K92 solar model
atmosphere (
(Fe)=7.52), taken as the
reference value for the Sun throughout the present
work, agrees with the meteoritic value of Anders
& Grevesse (1989: 
(Fe)=7.51).
It is also very close to the value obtained using the
HM model atmosphere (
(Fe)=7.56:
Castelli et al. 1996). A very similar
value is obtained using Fe II lines when adopting the K92 model. The trend with excitation potential is also very small in the Sun.
We wish to stress here the importance of the consistency between these various
determinations of the solar Fe abundances: in fact, the whole scale of metal
abundance previously determined for globular clusters
was uncertain, due to the rather large difference (
dex) between the
solar Fe abundances obtained using the HM and the Bell et al. (1976: hereinafter
BEGN) model atmospheres generally adopted in the analysis of cool cluster
giants (see e.g., Leep et al. 1987). In fact, it was not
clear what solar Fe abundance to use: either the one determined using a
model extracted from the same grid used for cluster giants, or that obtained
using the best solar model. Use of K92 atmospheres solves this problem, since we
may now use models for giants extracted from the same grid which gives a solar
Fe abundance in agreement with the best photospheric and meteoritic
determinations.
Finally, we note that this choice of lines and of gf values allows a direct
comparison with the metallicity scale for the low-dispersion
index 
in
RR Lyrae stars derived by Clementini et al. (1995).
Figure 4: Comparison of new [Fe/H] values with those from the original
analyses. Different symbols represent stars of different samples studied