The determination of the abundances of the light elements in stars of differing metallicities is important for understanding the chemical evolution of the Milky Way. The first stellar generations are supposed to produce mostly -elements during massive supernova events of type II. This is observed as a supersolar Mg/Fe abundance ratio in very metal-poor stars, [Mg/Fe] and has been reported by a number of researchers (Wallerstein 1961; Gratton & Sneden 1987; Magain 1987; Hartmann & Gehren 1988; Fuhrmann et al. 1995; McWilliam et al. 1995). Studies of -element synthesis in SNe II have recently been undertaken by Thielemann et al. (1996), Arnett (1991), and Woosley & Weaver (1995). Due to small differences in the way stellar winds and semi-convection are treated and in the specification of the mass cut and explosion mechanism their predictions differ, mainly because of differences in the respective Fe yields but also due to non-negligible differences in the pre-explosion yields of Mg. There is no question that Mg, in principle, is less affected by the fine-tuning of SN II explosions. Therefore, inasmuch as the products of SN II nucleosynthesis are mixed into interstellar space 24 Mg should constitute a reliable reference of the early evolutionary time scale of the Galaxy.
Due to the number of strong absorption lines found in the visible spectra of even the more metal-poor stars, neutral magnesium is easier to observe than e.g. O I. However, it shares the disadvantage of most neutral metals in the atmospheres of moderately cool stars, with Mg II being the dominant ionization stage above 5000 K. Consequently, neutral Mg I is sensitive to deviations from local thermodynamic equilibrium, particularly as its ionization balance is dominated by photoionization from the 3p state. In metal-poor cool stars most of the free electrons have vanished and the collision rates are correspondingly smaller. Together with an increased UV radiation field that leads to large photoionization rates this makes Mg I even more sensitive to non-LTE and affects any careful abundance analysis of Mg in these objects. The question then is: when do such deviations from LTE become important? An analysis of the solar spectrum will enable us to answer this question.
Several studies have been carried out analyzing the Mg I spectrum in the solar atmosphere starting with the non-LTE analysis of Athay & Canfield (1969) and the LTE analysis of intercombination line formation by Altrock & Cannon (1972). Lambert & Luck (1978) determined the Mg abundance of the solar photosphere assuming LTE and using the Holweger & Müller (1974) model with a constant, isotropic microturbulent velocity = 1.0 km/s.
A non-LTE study of the solar Mg I emission lines near 12 µm was published by Lemke & Holweger (1987) who analyzed the statistical equilibrium of Mg I and the influence of various input data on the line profile. The Mg I model atom they used includes 38 bound levels and 62 line transitions. Their standard non-LTE calculations did not reproduce the emission in the infrared lines. Mauas et al. (1988) used a twelve level atomic model for Mg I line synthesis. They investigated how the computed profiles at 4571Å and 5173Å are influenced by the model atom and the choice of its parameters. Carlsson et al. (1992) have explained the formation of the emission lines of Mg I in the solar spectrum near 12 µm employing standard plane-parallel non-LTE line formation with a radiative-equilibrium model atmosphere. They obtained excellent agreement with the observational constraints from a comprehensive atomic model. Although they neglected inelastic collisions with neutral hydrogen particles they were able to reproduce the IR emission line profiles. Recent work on the Mg I non-LTE problem comes from Mashonkina et al. (1996) who explored the corresponding abundance variations in cool stars.
In our present study, we carefully check the various unblended Mg I lines visible in the solar spectrum over a wavelength range extending from the blue to the far infrared. All Mg I lines considered here are reproduced using standard non-LTE line formation techniques with the radiative transfer solved in the Auer-Heasley scheme (DETAIL; Giddings, 1981, Butler & Giddings, 1985), taking the population processes between all levels of the Mg model atom into account. The principal aim is not merely to reproduce the observed solar spectrum but also to gather empirical information about the interplay between electron or heavy particle collisions and photoionization.
It is appropriate here to point out that we cannot hope to provide the reader with new results about the physics inherent to atomic parameters as inferred e.g. from theoretical considerations or laboratory experiments. It is obvious that the solar atmosphere is more complex in structure than any laboratory plasma; therefore a bad representation of an atmospheric model such as the plane-parallel hydrostatic approach inevitably produces may well mimic details of the interaction processes used in either LTE or NLTE spectrum synthesis. There is, however, reasonable evidence from comparing synthesized and time-integrated observed solar spectra that semi-empirical fits to atomic data such as collision cross-sections indicate some trends that have not yet been predicted by either theoretical atomic physics or terrestrial laboratories. With this in mind we have to employ our model of the solar atmosphere as the single access to atomic data that have not been determined otherwise. Note also that our aim to use both atomic and atmospheric models to analyze magnesium in other stars requires that we model the solar magnesium spectrum with the same set of approximations.
© European Southern Observatory (ESO) 1998
Online publication: April 15, 1998