Spectral type, temperature, and evolutionary stage in cool supergiants

Abstract

Context. In recent years, our understanding of red supergiants has been questioned by strong disagreements between stellar atmospheric parameters derived with different techniques. Temperature scales have been disputed, and the possibility that spectral types do not depend primarily on temperature has been raised. Aims. We explore the relations between different observed parameters, and we explore the ability to derive accurate intrinsic stellar parameters from these relations through the analysis of the largest spectroscopic sample of red supergiants to date. Methods. We obtained intermediate-resolution spectra of a sample of about 500 red supergiants in the Large and the Small Magellanic Cloud. From these spectra, we derive spectral types and measure a large set of photospheric atomic lines. We explore possible correlations between different observational parameters, also making use of near- and mid-infrared colours and literature on photometric variability. Results. Direct comparison between the behaviour of atomic lines (Fe i, Ti i, and Ca ii) in the observed spectra and a comprehensive set of synthetic atmospheric models provides compelling evidence that effective temperature is the prime underlying variable driving the spectral-type sequence between early G and M2 for supergiants. In spite of this, there is a clear correlation between spectral type and luminosity, with later spectral types tending to correspond to more luminous stars with heavier mass loss. This trend is much more marked in the LMC than in the SMC. The population of red supergiants in the SMC is characterised by a higher degree of spectral variability, early spectral types (centred on type K1) and low mass-loss rates (as measured by dust-sensitive mid-infrared colours). The population in the LMC displays less spectroscopic variability and later spectral types. The distribution of spectral types is not single-peaked. Instead, the brightest supergiants have a significantly different distribution from less luminous objects, presenting mostly M subtypes (centred on M2), and increasing mass-loss rates for later types. In this regard, the behaviour of red supergiants in the LMC is not very different from that of Milky Way objects. Conclusions. The observed properties of red supergiants in the SMC and the LMC cannot be described correctly by standard evolutionary models. The very strong correlation between spectral type and bolometric luminosity, supported by all data from the Milky Way, cannot be reproduced at all by current evolutionary tracks.