2. New analysis of solar Fe I lines
We selected a sample of 65 good solar Fe I lines. The solar data (, equivalent widths) as well as the atomic data (excitation potentials, log gf-values) are given in Table 1. Among these 65 lines, 33 lines of rather low excitation energy (0-2.6 eV) have gf-values measured at Oxford (see Blackwell et al. 1995a, for the references) whereas for 32 lines of higher excitation energy (1.6-4.6 eV), the gf-values have been measured at Hannover (Bard et al. 1991; Bard & Kock 1994). Broadening of these lines is essentially due to collisions with atomic hydrogen atoms. Damping constants have been computed from the recent data cited above (Anstee & O'Mara 1995; Barklem & O'Mara 1997; Barklem et al. 1998a).
Table 1. Fe I lines in the solar spectrum
The problem of the accuracy of the two sets of gf-values has been discussed at large in the original papers and also in Blackwell et al. (1995a, b) and Holweger et al. (1995). For the Oxford data, we shall adopt a conservative estimate for the mean uncertainty of the oscillator strengths of log gf (Oxford) = 0.02 dex (= 5%); in reality the uncertainty is probably better than this value (see Blackwell et al. 1995b). The gf-values measured at Hannover have generally uncertainties of the order of 10%. Actually, the mean uncertainty of the log gf-values for our 32 Hannover Fe I lines is log gf (Hannover) = 0.045 ( 0.015) dex. Oxford and Hannover data sets have lines in common: a comparison shows that log gf (Hannover - Oxford) = +0.026 0.044 dex (Blackwell et al. 1995b). This excellent agreement between the two sets, obtained by entirely different techniques, is a pretty good test of the quality of the data in both sets (see however Sects. 3 and 4).
We have to comment on two other recent sets of transition probabilities for Fe I . O'Brian et al. (1991) have measured a large number of gf-values for Fe I lines of different excitation energies. We do not select these lines because the uncertainty of individual results is too large. The comparison of the Hannover and O'Brian et al. results shows that log gf (Hannover - O'Brian) = +0.003 0.15 (Bard & Kock 1994). If the mean absolute scales agree, the spread is uncomfortably large! We also showed that if a large number of good solar lines with gf-values from O'Brian et al. are used to derive the solar abundance of iron, the spread of the results is extremely large (see Fig. 4 of Grevesse & Noels 1993 and Grevesse et al. 1995) and essentially due to uncertainties in the gf-values. Milford et al. (1989) also measured intensities of a sample of Fe I lines relative to reference lines for which the absolute gf-values have been measured by others. Some of the reference lines have uncertain accuracy and moreover some of the reference lines are too far away from the line of interest. Although some of their new results are very accurate, essentially those referred to lines measured at Oxford, other results are of lower accuracy. The use of some of their lines of rather high excitation energy by Milford et al. (1994) to derive the solar iron abundance leads to = 7.54 0.05, a result in agreement with the meteoritic value. No line from this list is used in the present study.
A comment has also to be made on the equivalent widths because differences in for lines in common between Oxford and Kiel-Hannover can explain part of the difference between the results of the two groups. The problem is already discussed at large by Blackwell et al. (1995a, b), Holweger et al. (1995) and Kostik et al. (1996). Our equivalent widths, measured on the Jungfraujoch atlas of the solar photospheric spectrum of Delbouille et al. (1973), agree pretty well with the 's obtained at Oxford [(this work)/(Oxford) = 0.990 0.015] as well as with the 's measured at Kiev by Kostik et al. [(this work)/(Kiev) = 1.002 0.026]. When the same comparison is made with the equivalent widths from Kiel, we find (this work)/(Kiel) = 1.06 0.07 (for faint lines, the value is 1.08 0.08). Such a large disagreement has much larger effects on the abundance: for medium-strong lines, the difference might amount to 0.10 dex (25%). We feel confident in our measurements because of the very good agreement with Oxford and Kiev and the low dispersion of the results and because in a few cases, it is evident that the Kiel equivalent widths are too small. But, as the Kiel measurements are not made on the same solar photospheric spectrum as the others, maybe the sun is really a slightly variable star.
We have redone the abundance determination using the Fe Ilines of Table 1 and the widely used, since many decades, photospheric model of Holweger & Müller (1974). The results are shown in Fig. 2 where a strong dependence against the excitation energy is clearly visible: lines with low excitations lead to results substantially higher than the high excitation lines which lead to an abundance in agreement with that of the meteorites ( = 7.50).
The last but very important parameter in abundance studies, namely the microturbulent velocity, has been varied between 0.85 and 1 km s-1. In going from = 0.85 km s-1, the value preferred by the Oxford group, to 1 km s-1, the value favoured by the Kiel-Hannover group, the abundance might decrease by 0.08 dex for medium-strong lines. In the plot of Fig. 2, the detailed results are given for = 0.85 km s-1, but the excitation dependence is not drastically changed if = 1 km s-1 is used. It is true however that the lines of the Oxford group are more sensitive to the microturbulent velocity than the lines from the Kiel-Hannover group: (Oxford) = -0.040 (7.617-7.577) whereas (Kiel-Hannover) = -0.012 (7.513-7.501) if is increased from 0.85 to 1 km s-1.
© European Southern Observatory (ESO) 1999
Online publication: June 18, 1999