4 Uncertainties in the elemental abundances

The uncertainties in the derived abundances are caused by errors in the determination of the atmospheric parameters, in the equivalent width measurements, and also in the quality of oscillator strengths. For spectra with S/N ratios larger than 50 the errors in the equivalent widths are between 5 and 8%. The errors in gf values vary from element to element. For Fe I lines, experimental values of good accuracies (better than 10%) do exist, for other Fe-peak elements the range in errors could be within 10 to 25%. For heavier elements, particularly for s-process elements, the errors could be larger than 25%. For the stars HD 725, HD 158616, HD 172481, HD 173638, HD 218753 and HD 331319 we could measure a very large number of unblended lines, and the estimated errors in $T_{\rm eff}$, $\log g$ and $\xi_{\rm t}$ are $\pm200$ K, $\pm0.25$ and $\pm0.2$ km s^-1, respectively. The sensitivity of the derived abundances to changes in the model atmospheric parameters are described in Table 4 of Gonzalez et al. (1997) for two RV Tau stars. We have used the same grid of atmospheric models and the same database for line oscillator strengths, hence the procedure will not be repeated here. For the Fe-peak elements we could measure a sufficiently large number of lines and the gf values used being of good quality, we expect these abundances to be accurate within 0.2 to 0.25 dex. For heavier elements, particularly the s-process elements, the uncertainty could be above 0.3 dex. Similarly for light elements like oxygen where few lines are available the uncertainty could be above 0.3 dex. For HD 172324 and HDE 341617, we will discuss the uncertainties in their respective sections.