The results of the non-LTE calculations from various Mg I lines shown in Table 2 confirm that deviations from LTE of the level populations of neutral magnesium are quite small in the Sun. The resulting effects on line formation are almost negligible in the visible solar spectrum. We find that only lines in the infrared show a significant non-LTE effect. The singlet lines at 8806, 8923 and 11828 Å begin to show significant deviations from LTE at the line cores. A Mg I atomic model without neutral hydrogen collisions probably cannot reproduce the observed profiles of these lines. The lines at m allow the best discrimination between the models, they cannot be fitted when the hydrogen collisions are too strong. In fact, the additional exponential dependence on excitation energy of the hydrogen collisions obtained by applying a cross-section significantly reduced with respect to Drawin's (1969) formula produces the most realistic profile centres, and the m line emission lines at the solar limb are also well reproduced. The interacting levels of the Mg I atom in the solar photosphere are well represented by our choice of lines analyzed, with the principal levels below 6 eV included. All lines in the blue, notably multiplets 12 to 16 are heavily blended, and there will be no loss of information due to excluding these lines. Lines in the infrared are combining even higher excited levels of which we have only analyzed the lines to put strong limits on the hydrogen collison rates. Thus we do not expect to improve the calculations significantly by comparison with additional IR lines that have been observed in the solar spectrum. The question of how reliable our results are must be answered with respect to our choice of free parameters. Since we have used standard formulae, each with only 1 or 2 parameters to fit, we are confident that our atomic model is well determined.
Table 2. Comparison of calculations using different atomic models with hydrogen collisions following Drawin's formula (B), exponentially scaled hydrogen collisions (C), and no hydrogen collisions (D)
The present investigation will serve as a basis for further analyses of the statistical equilibrium in metal-poor stars, and it is interesting to estimate the changes expected with reduced metallicity. With the density of free electrons decreasing in proportion to the star's metal abundance we expect considerably reduced collisional interaction. This could lead to substantially stronger deviations from LTE at optical depths between -3 and 0, if the reduced electronic interaction is not compensated by hydrogen collisions (Baumüller & Gehren, 1996, 1997). Corresponding calculations will be presented in a forthcoming paper.
© European Southern Observatory (ESO) 1998
Online publication: April 15, 1998