Astron. Astrophys. 331, 1051-1058 (1998)

6. A note on differences in coding

All computations presented in this work have been based on a unix implementation of WIDTH6. Computations for a particular test line (Fe I 5247) were also made using two additional codes. One was a unix implementation of the spectrum synthesis program of Cottrell (Cottrell & Norris 1978), and the second was the PC code of Ross (1997) ⁵. As this line has a very low excitation potential, it is only weakly sensitive to damping, but nevertheless ought to give the same sensitivity to line broadening irrespective of which code is used. Regrettably, this is not the case. Calculations were performed for the solar line strength in a Holweger-Müller model using =0.85 km s^-1. When pure Unsöld damping was used in both programs, Cottrell's gave an iron abundance lower than WIDTH6 by only 0.007 dex; i.e. in good agreement. The Ross code gave an abundance 0.019 dex lower than WIDTH6, but as it's damping implementation ignores the inner level, this difference is of minor concern. More disturbing, however, was the result when was enhanced by a factor of 2.0. The resulting abundance change was 0.062 dex for WIDTH6, but only 0.043 dex for the Cottrell code and 0.038 dex for the Ross code.

The origin of the difference between the WIDTH6 and Cottrell implementations is unclear. As the sensitivity given by WIDTH6 differs from that found with the Cottrell code by a factor of 1.4 for the particular Fe I line tested, it raises the possibility that the differences shown earlier in this paper might be upper limits on the extent of the errors. However, even scaling down the size of the errors will not undermine the main results of this work, which is that weak lines are not immune to the effects of uncertain damping, that the errors can be strongly dependent upon equivalent width and excitation potential, and that efforts to exploit high S/N spectra utilising accurate gf values may ultimately be limited by errors in the adopted damping values and (relatedly) in our characterisation of microturbulence. Errors in the model atmospheres must also be considered.

© European Southern Observatory (ESO) 1998

Online publication: March 3, 1998
helpdesk.link@springer.de