We use moderate-resolution optical spectrophotometry and the new MARCS stellar atmosphere models to determine the effective temperatures of 74 Galactic red supergiants (RSGs). The stars are mostly members of OB associations or clusters with known distances, allowing a critical comparison with modern stellar evolutionary tracks. We find we can achieve excellent matches between the observations and the reddened model fluxes and molecular transitions, although the atomic lines Ca i k4226 and Ca ii H and K are found to be unrealistically strong in the models. Our new effective temperature scale is significantly warmer than those in the literature, with the differences amounting to 400 K for the latest type M supergiants (i.e., M5 I). We show that the newly derived temperatures and bolometric corrections give much better agreement with stellar evolutionary tracks. This agreement provides a completely independent verification of our new temperature scale. The combination of effective temperature and bolometric luminosities allows us to calculate stellar radii; the coolest and most luminous stars (KW Sgr, Case 75, KY Cyg, HD 206936 = Cep) have radii of roughly 1500 R (7 AU), in excellent accordance with the largest stellar radii predicted from current evolutionary theory, although smaller than that found by others for the binary VV Cep and for the peculiar star VY CMa. We find that similar results are obtained for the effective temperatures and bolometric luminosities using only the dereddened V À K colors, providing a powerful demonstration of the self-consistency of the MARCS models.

Mass loss of red supergiants (RSG) is important for the evolution of massive stars, but is not fully explained. Several empirical prescriptions have been proposed, trying to express the mass-loss rate (Ṁ) as a function of fundamental stellar parameters (mass, luminosity, effective temperature). Our goal is to test whether the de Jager et al. (1988, A&AS, 72, 259) prescription, which is used in some stellar evolution models, is still valid in view of more recent mass-loss determinations. By considering 40 Galactic RSGs with an infrared excess and an IRAS 60-μm flux larger than 2 Jy, and assuming a gas-to-dust mass ratio of 200, we find that the de Jager rate agrees within a factor 4 with mostṀ estimates based on the 60-μm signal. It also agrees with six of the only eight Galactic RSGs for whichṀ can be measured more directly through observations of the circumstellar gas. The two objects that do not follow the de Jager prescription (by an order of magnitude) are μ Cep and NML Cyg. We also considered the RSGs of the Magellanic Clouds. Thanks to the results of previous research, we find that the RSGs of the Small Magellanic Cloud have mass-loss rates consistent with the de Jager rate scaled by (Z/Z ) α , where Z is the metallicity and α is 0.7. The situation is less clear for the RSGs of the Large Magellanic Cloud. In particular, for L > 1.6 × 10 5 L , one finds numerous RSGs (except WOH-G64) withṀ significantly smaller than the de Jager rate and indicating thatṀ would no longer increase with L. Before this odd situation is confirmed through further analyses of LMC RSGs, we suggest to keep the de Jager prescription unchanged at solar metallicity in the stellar evolutionary models and to apply a (Z/Z ) 0.7 dependence.

Context. It has been proposed that convection in red supergiant (RSG) stars produces large-scale granules causing observable surface inhomogeneities. This convection is also extremely vigorous and is suspected to be one of the main causes of mass-loss in RSGs. It should thus be understood in detail. Evidence has accumulated of asymmetries in the photospheres of RSGs, but detailed studies of granulation are still lacking. Interferometric observations provide an innovative way of addressing this question, but they are still often interpreted using smooth symmetrical limb-darkened intensity distributions, or simple, spotted, ad hoc models. Aims. We explore the impact of the granulation on visibility curves and closure phases using the radiative transfer code OPTIM3D. We simultaneously assess how 3D simulations of convection in RSG with CO 5 BOLD can be tested by comparing with these observations. Methods. We use 3D radiative hydrodynamical (RHD) simulations of convection to compute intensity maps at various wavelengths and time, from which we derive interferometric visibility amplitudes and phases. We study their behaviour with time, position angle, and wavelength, and compare them to observations of the RSG α Ori. Results. We provide average limb-darkening coefficients for RSGs. We describe the prospects for the detection and characterization of granulation (i.e., contrast, size) on RSGs. We demonstrate that our RHD simulations provide an excellent fit to existing interferometric observations of α Ori, in contrast to limb darkened disks. This confirms the existence of large convective cells on the surface of Betelgeuse.