5. Traditional approach of abundance analysis and its demerits
As usual, lines are practically always used for microturbulence velocity determination, sometimes for effective temperature specification, and together with lines for gravity determination. The number of unblended lines of different excitation potentials of the lower level and with equivalent widths spanning the interval from 0 to 200 mÅ is rather large in the spectra of yellow supergiants, making them a very attractive tool for determination of the stellar atmospheric parameters.
Up to now, the majority of the investigators used lines for this purpose. Nevertheless, there is accumulating evidence, that the lines of neutral elements (iron, in particular) should be used with caution.
As was pointed out by Lyubimkov & Boyarchuk (1983, 1985) and Rentzsch-Holm (1996), the NLTE effects for lines are very pronounced and depend upon the equivalent width (although, in the latter work only the stars with =7000 K-12000 K and =3.5-4.5 were examined, there is no doubt that NLTE corrections should increase along with the gravity decreasing). According to Lyubimkov & Boyarchuk (1983), the NLTE corrections to the iron abundance in supergiants achieve about 0.6 dex, when lines with 200 mÅ are used, and 0.1-0.2 dex in the case of lines having 50 mÅ.
At the same time, lines are practically not sensitive to the departure from LTE. Severe overionization of is unlikely in supergiants, while for atoms it can be expected. Recently this conclusion was confirmed by the calculations of statistical equilibrium for / in the atmosphere of late-type stars (Thevenin & Idiart 1999). Accurate atomic models for neutral and ionized iron were applied to investigate the NLTE effects in iron abundance for various stellar parameters. The authors found that particularly in the case of metal deficient stars, surface gravities derived by LTE analysis are in significant error.
Generally this problem exists not only for intermediate mass supergiants. Steenbock (1985) showed that deviations from the LTE for neutral iron affect the stellar parameters found from the spectroscopic analysis, particularly in the evolved giants such as Pollux. The NLTE effects in metal-poor stars and their spectral manifestations were explored by Gehren et al. (1991).
Fuhrmann et al. (1997) proposed the use of pressure-broadened Mg Ib lines to derive the gravity parameter for F and G stars. These lines appear to be more reliable tracers compared to the ionization equilibrium condition for /, which is susceptible to overionization effects. They demonstrated that the strong line method circumvents the long-standing problem of gravity discrepancy. The same conclusion about the systematic errors in the standard spectroscopic LTE analysis based on the keeping of iron ionization equilibrium was also drawn in the work by Fuhrmann (1998). It was found that for neutral iron lines formed in the atmospheres of some stars, much stronger deviations are expected than for the corresponding lines in the solar photosphere. While no such influence was found for the lines, one can suspect that the different behaviour of and lines would cause a noticeable change in the value if it is determined from the / balance.
If the NLTE corrections progressively depend upon the equivalent widths, then we should expect a monotonous decreasing of the iron abundance derived from the lines along with an increase of their equivalent widths. Standard spectroscopic analysis implies that the microturbulence parameter is determined by avoiding any dependence of the resulting abundances from individual lines upon their equivalent widths. If lines are used, then taking into account their anomalous behaviour, one can suspect that one will inevitably obtain an artificially decreased value in order to equalize the abundances resulting from the stronger and weaker lines of the neutral iron. Such an incorrectly determined microturbulence parameter can then affect the gravity determination and finally introduce an error in the elemental abundances.
Let us consider the Cep example. Starting the analysis with the preliminary estimated (photometry) and (e.g., physical gravity) values, we have to specify the model parameters (in the main it refers to the gravity value). With value derived by using only lines ( 3 km s-1, see Table 1), lines produce an apparent iron overabundance (Fig. 4).
To equalize the mean abundance results from and lines (i.e., to preserve the / ionization balance), we are forced to decrease the model gravity by approximately 0.5 dex compared to the physical one. Thus, the averaged abundances from and lines become more or less similar (see Fig. 3), but at the same time the discrepancy between spectroscopic and physical gravity values appears.
As was already mentioned, such a problem was discussed by Luck & Lambert (1985) and, for example, by Luck (1994), Luck et al. (1998). In the latter work, where the sample of LMC and SMC supergiant stars was investigated, the authors found high disaccord between physical and spectroscopic values for the supergiants having physical gravities less than 1 dex (see Fig. 4 from that work). The difference between the two values in some cases achieved 1-1.5 dex. This fact clearly signals that something is wrong in the traditional spectroscopic analysis of the intermediate mass supergiants. We believe that this problem will be more severe in the case of very luminous supergiants and those having metal deficient atmospheres, where departure from LTE and additional overionization of the neutral species are very likely.
Hill et al. (1995) analysed several F supergiants from LMC and also concluded that atmospheric parameters derived with a law of overionization gave the gravity increased by 0.6 dex, and microturbulent velocity increased by 0.5 km s-1 compared with the parameters determined in pure LTE approach. Moreover, the gravity determined in this way appeared to be close to the value which can be derived from photometry.
© European Southern Observatory (ESO) 1999
Online publication: November 3, 1999