4 Atomic line data

4.1 Spectral lines

All unblended lines with symmetric profiles having equivalent widths larger than 20 mÅ in the solar atlas (Moore et al. [1966]) were cautiously selected. The equivalent width limit ensures that lines are not disappearing in the most metal-poor disk stars at $\mbox{\rm [Fe/H]}= -1.0$. Given that very weak lines would lead to an increase of random errors in the abundance determination and that too strong lines are very sensitive to damping constants, only weak and intermediate-strong lines with 3 < EW < 100 mÅ in the stellar spectra were adopted in our abundance analysis except for potassium, for which only one line ( $\textup{K\, {\mdseries\textsc{i}}}$ $\lambda$7699) with an equivalent width range of 50 - 190 mÅ, is available.

   
4.2 Oscillator strengths

Due to the large number of measurable lines in the spectra, Fe I lines were used for microturbulence determination and temperature consistency check. Hence, careful selection of oscillator strengths for them is of particular importance. Of many experimental or theoretical calculations of oscillator strengths for Fe I lines, only three sources with precise gfvalues were chosen. They are: Blackwell et al. ([1982b], [1982c]), O'Brian et al. ([1991]) and Bard & Kock ([1991]) or Bard et al. ([1994]). The agreements between these sources are very satisfactory, and thus mean loggf values were adopted if oscillator strengths are available in more than one of the three sources. A few oscillator strengths with large differences between these sources were excluded.

References for other elements are: O I (Lambert [1978]), Na I (Lambert & Warner [1968]), Mg I (Chang [1990]), Al I (Lambert & Warner [1968]), Si I (Garz [1973]), K I (Lambert & Warner [1968]), Ca I (Smith & Raggett [1981]), Ti I (Blackwell et al. [1982a], [1986a]), V I (Whaling et al. [1985]), Cr I (Blackwell et al. [1986b]), Fe II (Biémont et al. [1991]; Hannaford et al. [1992]), Ni I (Kostyk [1982]; Wickliffe & Lawler [1997]) and Ba II (Wiese & Martin [1980]).

These experimental or theoretical gf values were inspected to see if they give reliable abundances by evaluating the deviation between the abundance derived from a given line and the mean abundance from all lines of the same species. A significant mean deviation in the same direction for all stars (excluding the suspected binaries) were used to correct the gf value. Lines with large deviations in different directions for different stars were discarded from the experimental or theoretical loggf list. The two sets of values are presented in columns "abs.'' and "cor.'' in Table 4. Based on these gf values, we derived abundances for 10 "standard'' stars: HD60319, HD49732, HD215257, HD58551, HD101676, HD76349, HD58855, HD22484, the Sun and HD34411A (in metallicity order). Oscillator strengths for lines with unknown gf values were then determined from an inverse abundance analysis of the above 10 stars, which are distributed in the metallicity range $-0.9 \leq \mbox{\rm [Fe/H]}\leq 0.0$ dex and were observed at high $S/N~\sim~250 - 400$. Generally, the gf values for a given line from different "standard'' stars agree well, and a mean value (given in the column "dif'' in Table 4) was thus adopted.

4.3 Empirical enhancement factor

It has been recognized for a long time that line broading derived from Unsöld's ([1955]) approximation to Van der Waals interaction is too weak, and an enhancement factor, $E_{\gamma}$, should be applied to the damping parameter, $\gamma$. Usually, enhancement factors of Fe I lines with excitation potential of the lower energy level ( $\chi_{{\rm low}}$) less than 2.6 eV are taken from the empirical calibration by Simmons & Blackwell ([1982]). For Fe I lines with $\chi_{{\rm low}}>2.6$ eV, $E_{\gamma}=1.4$ is generally used in abundance analysis. Recently, Anstee & O'Mara's ([1995]) computed the broadening cross sections for s-p and p-s transitions and found that $E_{\gamma}$ should be $2.5\sim3.2$ for lines with $\chi_{{\rm low}}>3.0$ eV, whereas lines with $\chi_{{\rm low}}<2.6$ eV have broadening cross sections more consistent with Simmons & Blackwell's ([1982]) work.

Following EAGLNT and other works, we adopted Simmons & Blackwell's ([1982]) $E_{\gamma}$ for Fe I lines with $\chi_{{\rm low}}<2.6$ eV and $E_{\gamma}=1.4$ for the remaining Fe I lines, while a value of 2.5 was applied for Fe II lines as suggested by Holweger et al. ([1990]). Enhancement factors for Na I, Si I, Ca I and Ba II were taken from EAGLNT (see references therein). Finally, a value of 1.5 was adopted for the K I, Ti I and V I lines considering their low excitation potentials, and a factor of 2.5 was applied to the remaining elements following Mäckle et al. ([1975]). The effects of changing these values by 50% on the derived abundances are discussed in Sect. 5.2.

The atomic line data are given in Table 4.