Comparing Properties of Convection in the Surface Layers of Stars With Different Masses and Evolutionary States

By comparing convective properties computed from simulations derived from a series of G and K type stellar models, we find marked differences between simulations and mixing length theory. These differences increase with the strength of the turbulence (as measured by the size of the turbulent pressure). Two of the most obvious differences are in the run of the superadiabaticity versus depth and in the size of the convective eddies. Provided the turbulence is weak (i.e. the ratio of turbulent pressure to gas pressure is a few percent) the estimate of eddy size used in a stellar model eddy (i.e. αH_p) is quite close to the mean size of the simulated eddies. However, as the strength of the turbulence is increased (as happens in more evolved or hotter or more massive stars) the mixing length estimate becomes a very poor approximation to the actual eddy size. We also show results of a comparison study between different simulation codes conducted by an independent collaborator. Despite significant differences in input physics, numerical methods and treatments of radiation, the resolved convective quantities differ by less than 10%.