Comparison with Other Determinations

The final stellar parameters are listed in Table A in the Appendix. We compared these values with those obtained by JW95, who constructed fitting functions and SPS models based on the same spectral library we adopt in this work, so that the number of stars in common is maximum. Overall, there is no major systematic difference between the two sets of stellar parameters. The average differences (Ours - JW95) are as follows: $\Delta T_{\rm eff}$ = 19 $\pm$ 260 K, $\Delta \log g$ = -0.06 $\pm$ 0.4, and $\Delta$ [Fe/H] = -0.02 $\pm$ 0.2. Moreover, the 1$-\sigma$ error bars in $T_{\rm eff}$ and $\log g$ drop to $\sim$ 130 K and $\sim$ 0.2 dex when roughly 30 stars hotter than 7000 K, for which stellar parameters are more uncertain, are excluded from the statistics.

However, further scrutiny reveals the presence of systematic differences worthy of mention, for instance when we split the comparisons between dwarfs and giants. It makes sense to look into comparisons within these sub-samples, because different procedures are followed to determine stellar parameters for dwarfs and giants both in this work and by JW95.

Dwarfs

We first focus on dwarf stars. In Figure 3 differences between the two sets of stellar parameters (this work - JW95) are compared as a function of JW95 values. The most outstanding differences revealed by the comparisons in Figure 3 are those between the two sets of [Fe/H]s. Our values are on average 0.15-0.2 dex higher than those of JW95. While JW95 adopted [Fe/H]s from Edvardsson et al. (1993), ours are based on Strömgren photometry using the calibration from Schuster & Nissen (1989), as revised by Clementini et al. (1999). The latter explains the discrepancy, as Clementini et al. added an extra 0.15 dex to Schuster & Nissen's [Fe/H] values.

There are also systematic differences, albeit more subtle, between the two sets of $T_{\rm eff}$'s. Our $T_{\rm eff}$s are hotter by up to 250 K (average $\sim$ 100 K) for stars hotter than 6200 K. JW95's $T_{\rm eff}$s for dwarf stars are based on broadband color-$T_{\rm eff}$ calibrations from the literature, while ours come from Strömgren photometry, thus being consistent with the values estimated by Edvardsson et al. (1993). In fact, JW95 note that their $T_{\rm eff}$s were cooler than those of Edvardsson et al. by a similar amount, and they decided to use stars in common with Edvardsson et al. to convert the latter set of $T_{\rm eff}$ into their own. Since our $T_{\rm eff}$-scale is already consistent with that of Edvardsson et al. (1993), the difference found here is not surprising.

No substantial systematic effect is seen for $\log g$, but the scatter is higher for this parameter. This is not surprising. Uncertainties in $\log g$ are usually large because they are affected by uncertainties in $T_{\rm eff}$, adopted mass, distance and bolometric correction.

Giants

Figure 4 repeats Figure 3 restricting the plot to giant stars. While no systematic effect is found for $\log g$, our [Fe/H]s tend to be lower than those of JW95, especially in the high-[Fe/H] end, where the average residual reaches $\sim$ -0.25 dex. At [Fe/H] $\sim$ -0.5, the two scales are essentially the same. There is also a small systematic effect in the $T_{\rm eff}$ values in that ours are slightly lower (on average $\mathrel {\copy \simlessbox }$ 100 K) than those of JW95. It is natural to suppose that the two effects might be correlated, given the degenerate effects of $T_{\rm eff}$ and [Fe/H] on colors and absorption line features. However, there is no correlation between $\Delta T_{\rm eff}$ and $\Delta$ [Fe/H]. Our atmospheric parameters for giant stars are rooted in the Soubiran et al. (1998) scale (see Paper I for details), while the JW95 scale is based on that of Dickow et al. (1970), so we believe that our parameters, being based on updated stellar parameter determinations, are more reliable.

In summary, the differences found here are not unexpected, and we stress that they not only are not substantial but in fact are commensurate with the uncertainties associated with $T_{\rm eff}$ and [Fe/H] determinations from broadband colors.