Inelastic e+Mg collision data and its impact on modelling stellar and supernova spectra

Results of calculations for inelastic e+Mg effective collision strengths for the lowest 25 physical states of Mg I (up to 3s6p 1P), and thus 300 transitions, from the convergent close-coupling (CCC) and the B-spline R-matrix (BSR) methods are presented. At temperatures of interest, ~5000 K, the results of the two calculations differ on average by only 4%, with a scatter of 27%. As the methods are independent, this suggests that the calculations provide datasets for e+Mg collisions accurate to this level. Comparison with the commonly used dataset compiled by Mauas et al. (1988), covering 25 transitions among 12 states, suggests the Mauas et al. data are on average ~57% too low, and with a very large scatter of a factor of ~6.5. In particular the collision strength for the transition corresponding to the Mg I intercombination line at 457 nm is significantly underestimated by Mauas et al., which has consequences for models that employ this dataset. In giant stars the new data leads to a stronger line compared to previous non-LTE calculations, and thus a reduction in the non-LTE abundance correction by ~0.1 dex (~25%). A non-LTE calculation in a supernova ejecta model shows this line becomes significantly stronger, by a factor of around two, alleviating the discrepancy where the 457 nm line in typical models with Mg/O ratios close to solar tended to be too weak compared to observations.