Red supergiants represent the most common final stage of the evolution of stars with initial masses between 8 and 30-35 times the mass of the Sun 1 . During this phase of lifetime lasting ≈ 10 5 yrs 1 , they experience substantial mass loss of unknown mechanism 2 . This mass loss can affect their evolutionary path, collapse, future supernova light curve 3 , and ultimate fate as a neutron star or a black hole 4 . From November 2019 to March 2020, the second closest red supergiant (RSG, 222 +48 −34 pc 5, 6 ) Betelgeuse experienced a historic dimming of its visible brightness, witnessed worldwide. Usually between 0.1 and 1.0 mag, it went down to 1.614 ± 0.008 mag around 7-13 February 2020 7 . Here we report high angular resolution observations showing that the southern hemisphere of the star was ten times darker than usual in the visible. Observations and modeling support the scenario of a dust clump recently formed in the vicinity of the star due to a local temperature decrease in a cool patch appearing on the photosphere. The directly imaged brightness variations of Betelgeuse evolved on a timescale of weeks. This event suggests that an inhomogeneous component of red supergiant mass loss 8 is linked to a very contrasted and rapidly changing photosphere.

Context. Modeling stellar atmospheres is a complex and intriguing task in modern astronomy. A systematic comparison of models with multi-technique observations is the only efficient way to constrain the models. Aims. We intend to perform self-consistent modeling of the atmospheres of six carbon-rich AGB stars (R Lep, R Vol, Y Pav, AQ Sgr, U Hya, and X TrA) with the aim of enlarging the knowledge of the dynamic processes occurring in their atmospheres. Methods. We used VLTI/MIDI interferometric observations, in combination with spectro-photometric data, and compared them with self-consistent, dynamic model atmospheres. Results. We found that the models can reproduce spectral energy distribution (SED) data well at wavelengths longer than 1 µm, and the interferometric observations between 8 µm and 10 µm. Discrepancies observed at wavelengths shorter than 1 µm in the SED, and longer than 10 µm in the visibilities, could be due to a combination of data-and model-related effects. The models best fitting the Miras are significantly extended, and have a prominent shell-like structure. On the contrary, the models best fitting the non-Miras are more compact, showing lower average mass loss. The mass loss is of episodic or multi-periodic nature but causes the visual amplitudes to be notably larger than the observed ones. A number of stellar parameters were derived from the model fitting: T Ross , L Ross , M, C/O, andṀ. Our findings agree well with literature values within the uncertainties. T Ross , and L Ross are also in good agreement with the temperature derived from the angular diameter T (θ (V−K) ) and the bolometric luminosity from the SED fitting L bol , except for AQ Sgr. The possible reasons are discussed in the text. Finally, θ Ross and θ (V−K) agree with one another better for the Miras than for the nonMiras targets, which is probably connected to the episodic nature of the latter models. We also located the stars in the H-R diagram, comparing them with evolutionary tracks. We found that the main derived properties (L, T eff , C/O ratios and stellar masses) from the model fitting are in good agreement with TP-AGB evolutionary calculations for carbon stars carried out with the COLIBRI code.

Context. The mass-loss process from evolved stars is a key ingredient for our understanding of many fields of astrophysics, including stellar evolution and the chemical enrichment of the interstellar medium (ISM) via stellar yields. Nevertheless, many questions are still unsolved, one of which is the geometry of the mass-loss process. Aims. Taking advantage of the results from the Herschel Mass loss of Evolved StarS (MESS) programme, we initiated a coordinated effort to characterise the geometry of mass loss from evolved red giants at various spatial scales. Methods. For this purpose we used the MID-infrared interferometric Instrument (MIDI) to resolve the inner envelope of 14 asymptotic giant branch stars (AGBs) in the MESS sample. In this contribution we present an overview of the interferometric data collected within the frame of our Large Programme, and we also add archive data for completeness. We studied the geometry of the inner atmosphere by comparing the observations with predictions from different geometric models. Results. Asymmetries are detected for the following five stars: R Leo, RT Vir, π 1 Gruis, omi Ori, and R Crt. All the objects are O-rich or S-type, suggesting that asymmetries in the N band are more common among stars with such chemistry. We speculate that this fact is related to the characteristics of the dust grains. Except for one star, no interferometric variability is detected, i.e. the changes in size of the shells of non-mira stars correspond to changes of the visibility of less than 10%. The observed spectral variability confirms previous findings from the literature. The detection of dust in our sample follows the location of the AGBs in the IRAS colour-colour diagram: more dust is detected around oxygen-rich stars in region II and in the carbon stars in region VII. The SiC dust feature does not appear in the visibility spectrum of the U Ant and S Sct, which are two carbon stars with detached shells. This finding has implications for the theory of SiC dust formation.