The Time-averaged Mass-loss Rates of Red Supergiants as Revealed by Their Luminosity Functions in M31 and M33

Mass loss in red supergiants (RSGs) is generally recognized to be episodic, but mass-loss prescriptions fail to reflect this. Evolutionary models show that the total amount of mass lost in this phase determines if these stars evolve to warmer temperatures before undergoing core collapse. The current Geneva evolutionary models mimic episodic mass loss by enhancing the quiescent prescription rates whenever the star's outer layers exceed the Eddington luminosity by a large factor. This results in a 20 M-circle dot model undergoing 10x more mass loss than it would otherwise, but has little effect on models of lower mass. We can test the validity of this approach observationally by measuring the proportion of high-luminosity RSGs to that predicted by the models. To do this, we use our recent luminosity-limited census of RSGs in M31 and M33, making modest improvements to membership, and adopting extinctions based on the recent panchromatic M31 and M33 Hubble surveys. We then compare the proportions of the highest luminosity RSGs found to that predicted by published Geneva models, as well as to a special set of models computed without the enhanced rates. We find good agreement with the models which include the supra-Eddington enhanced mass loss. The models with lower mass-loss rates predict a larger fraction of high-luminosity RSGs than observed, and thus can be ruled out. We also use these improved data to confirm that the upper luminosity limit of RSGs is log L L-circle dot similar to 5.4, regardless of metallicity, using our improved data on M31 and M33 plus previous results on the Magellanic Clouds.