Exploring the connection between atmosphere models and evolution models of very massive stars

Very massive stars (VMS) dominate the light of young stellar populations and are sources of intense stellar feedback. Their evolution is mainly driven by strong wind mass loss, yet current evolution models make simplistic assumptions on their atmospheric physics which are incompatible with the nature of VMS. In this work, we aim to understand VMS atmospheres throughout their evolution by supplementing structure models (computed with GENEC) with detailed atmosphere models (computed with PoWR) capable of capturing the physics of a radially-expanding medium in non-LTE. An important aspect is the computation of atmosphere models reaching into deeper layers of the star, notably including the iron-opacity peak as an important source of radiative driving. In this study, we compute atmosphere models at 16 snapshots along the main sequence of a 150 $M_\odot$ star. For each snapshot, we compute two atmosphere models connected to the underlying structure model at different depths (below and above the hot iron bump). We perform a detailed spectroscopic and structural comparison of the two sequences of model atmospheres, and present a generalized method for the correction of the effective temperature in evolution models with strong winds. The choice of connection point between structure and atmosphere models has a severe influence on the predicted spectral appearance, which constitutes a previously unexplored source of uncertainty in quantitative spectroscopy. The simplified atmosphere treatment of current stellar structure codes likely leads to an overestimation of the spatial extension of very massive stars, caused by opacity-induced sub-surface inflation. This inflation does not occur in our deep atmosphere models, resulting in a discrepancy in predicted effective temperatures of up to 20 kK. Future improvements with turbulence and dynamically-consistent models may resolve these discrepancies.