Stellar models with self-consistent Rosseland opacities

Context. The building of a stellar structure requires knowing the Rosseland mean opacity at each layer of the model. This mean opacity is very often interpolated in pre-computed tables due to the overwhelming time to compute it from monochromatic cross sections. The main drawback to using tables is that the opacities can be inconsistent with the actual local chemical composition, for instance in the regions of the star where nucleosynthesis occurs. Aims. We study the effects of self-consistent Rosseland mean opacity calculations on the stellar structure and evolution, in comparison with models where the metal mixture remains equal to the initial one. Methods. We developed a strategy that allows very fast calculations of Rosseland opacities from monochromatic cross sections. We are then able to compute evolutionary tracks with models whose Rosseland opacities are fully consistent with the chemical mix everywhere in the star. This method has been implemented in the Toulouse-Geneva evolution code. Results. Our self-consistent models show very small structural differences compared to models where the Rosseland opacity is computed with a fixed metal mixture. As a consequence, the main-sequence evolutionary tracks are almost the same for models of mass ranging from 2 to 8 M⊙. At a given surface gravity the relative difference in age is lower than 2% and generally below 1% between the two kinds of calculations, the self-consistent model being younger most of the time. Unless such a precision in age is sought out, the use of tabulated Rosseland opacities with a metal content defined globally is still acceptable, at least in main-sequence stars where the chemical mix changes only through nucleosynthesis.