Elemental Abundance Analyses with DAO Spectrograms. XXXV. On the Iron Abundances of B and A Stars

Whenever astronomers improve the precision and accuracy of stellar elemental abundance analyses, more stringent tests for a variety of astrophysical problems can be produced. Better analysis methods combined with more accurate and precise effective temperature and surface gravity determinations now appear possible in the coming years (see, e.g., Yüce & Adelman 2014). Major improvements in the atomic data, especially the oscillator strengths of key atomic species which are not of the highest quality, are especially desirable. Fuhr & Wiese (2006), hereafter FW, is a relatively recent major critical compilation of Fe I and Fe II values by two well respected NIST authors. Its values have been used in many abundance analyses. In the optical region many Fe I oscillator strengths had a higher quality than those of Fe II. The gf values of Melendez & Barbuy (2009), hereafter MB, which these authors claim are more accurate and precise optical region Fe II gf values, can be used both to check the FW Fe II values and possibly to supersede them. It would have been very useful to have error estimates for each individual value. An advantage of the FW gf values is that they cover a greater range of wavelengths and are available for more lines than those of MB which cover the region λλ 4087–7912. Most of the lines being compared have MB values based on laboratory data rather than solar data. This year Deb & Hibbert (2014) (hereafter DH) presented a theoretical study of all of the lines studied by MB. There is generally good agreement for most of the transitions, but there are differences. A line by line comparison is given in their Table 3.

I used the LTE physics Programs ATLAS9 and the WIDTH9 (Kurucz 1993) to perform fine analyses. The high dispersion and high signal-to-noise ratio (>200) spectra of B and A stars were obtained with Reticon and CCD detectors with the long camera of coudé spectrograph of the 1.22-m telescope of the Dominion Astrophysical Observatory for papers in the "Elemental Abundance Analyses with DAO Spectrograms" series (see Table 1 for references) which are also the sources of the effective temperatures and surface gravities. The minimal wavelength range is λλ 3820–4750. The published equivalent widths were measured with the graphical measurement program VLINE (Hill et al. 1982) using Gaussian and rotational fits to the metal line profiles. The data for each star used in this paper consist of 16 to 39 Fe II lines which had gf values from both FW and MB.

I derived the iron abundances only for nonblended lines with gf values from both FW and MB. I found the microturbulences in two ways: by (1) minimizing the scatter about the mean value and by (2) making the derived abundances independent of equivalent width. These results for a given star are usually so similar that in Table 1 I give the averages of these calculations. My published results using FW gf values are similar, but not identical to those presented here as more lines were often analyzed in the references. For this paper I selected the Fe II lines used with the FW gf values in the same manner as my original analyses. If I had started with the MB gf value analyses, some lines might have been deleted and others added.

For 50 lines, the mean difference between gf values from FW and those from MB is -0.01 ± 0.15 dex. Thus, the average values are the same, but there are line-to-line differences and in different stars the Fe II lines used for analyses can be slightly different. The worst agreements are for λ 4413.60 -0.40 dex, λ 4549.47 0.36 dex, and λ 4729.025 -0.34 dex. For about 60% of the sample the agreement is better than 0.10 dex. The mean difference between the values of FW and DH for 50 lines is -0.13 ± 0.22 dex, which indicates a modest shift in the zero point and a greater scatter than the FW gf values with those of MB. The largest difference is 0.90 dex for λ 4720.15.

For the 34 stars studied, the mean difference (FW—MB) in log Fe/N_T is 0.04 ± 0.06 dex with the standard deviation of the mean being 0.02 dex. Fe and N_T are the number of iron and all atoms, respectively, per unit volume. The mean difference in microturbulence is -0.47 ± 0.87 km s^-1. Thus, the MB gf values slightly decrease the derived iron abundance and slightly decrease the standard deviation of the mean while increasing slightly the microturbulence. These small decreases in the abundance will only slightly affect Fe II/Fe I equilibrium. The decrease in the scatter of the derived abundances is a signature of a more consistent set of gf values. The increase of the microturbulence will reduce the individual abundances of lines beyond the linear part of the curve of growth which will be seen in abundance results for elements whose strong lines were used in their determinations. However, the range of abundance changes is 0.15 to -0.07 dex, except for six stars the standard deviation of the mean changes is no more than 0.03 dex, and the microturbulence changes range between no change and an increase of 1.00 km s^-1. In the difference of results between FW and MB gf values, there is a change in the behavior of the difference which occurs near T_eff = 11000 K; namely, cooler than this value the FW values usually give the higher values, while for the hotter stars the MB gf values give the higher values. The results of this comparison suggest that the remaining systematic errors can be reduced. But the result for the microturbulences are worrisome.

I derived abundances using DH gf values for three stars as trial calculations: for o Peg, I found log Fe/N_T = -4.53 ± 0.22 and ξ = 1.9 km s^-1, for π Dra log Fe/N_T = -4.47 ± 0.22 and ξ = 3.8 km s^-1, and for υ Her log Fe/N_T = -4.97 ± 0.21 and ξ = 1.1 km s^-1. The scatter about the mean is about twice those for FW and MB gf values, the values of log Fe/N_T are smaller, and the values of ξ are closer to those for MB gf values than for FW gf values. It is primarily due to the increase in the scatter about the mean that I did not calculate results for the remaining stars. The increased scatter is mostly due to a few discrepant values for each star. In a fine analysis using DH gf values, the values for these lines would have been deleted.

As a check on which set of Fe II gf values to use, I derived the results for Fe I whose gf values are from FW. Since there has been a debate between the advocates of using the lines with the best gf values and those of using all possible gf values, I examined what would happen if I deleted the Fe I lines I used with the lower accuracy FW gf values. To have sufficient Fe I lines, I examined mostly those stars which are both sharp-lined and have effective temperatures cooler than the ensemble average (see Table 2). The reference numbers are those from Table 1. The Δ log Fe and Δξ are the differences between the Fe II MB results and those for the Fe I FW A & B quality lines whose uncertainties of less than 3% and less than 10%, respectively. C, D, and E quality lines used when all lines with FW gf values are studied have uncertainties of less than 25%, of less than 50%, and greater than 50%, respectively.

Removing the lower quality Fe I lines makes little difference to the results. The mean Δ log Fe and Δξ values are -0.10 ± 0.06 and 0.57 ± 0.34, respectively, indicating that the Fe I lines are giving larger Fe/N_T values than the Fe II lines with MB gf values and the microturbulences for the Fe II lines are greater than those of the Fe I lines. The results for the Fe I lines are in better agreement with the Fe II lines with FW gf values.

Use of the MB gf values produces slightly more consistent results than the FW gf values for the Fe II lines in common. Using Fe I lines with the two highest quality gf values often yields only a slight improvement over when all quality FW Fe I gf values are used. To some extent this may represent that in selecting lines for the final averages those lines with the poorest quality gf values are often eliminated in the analysis process.

Some of the offset between the results for Fe I lines and those for Fe II lines may be due to non-LTE effects. The Fe II lines have minor corrections while those for Fe I lines are larger for a main sequence star with T_eff = 9500 K (Gigas 1986). The sense of the Δ log Fe results given in Table 2 are consistent with this idea, although the magnitudes are often less. Other sources of difficulty could be due to measurement of line profiles and continuum placement, isotopic shifts which could move some weak components out of the profiles usually measured, and errors in the effective temperatures and surface gravities. In conclusion, the MB gf values might be better than those of FW, but further study is needed.