Images from VLT observations of nearby AGB stars

These data were then used to train a random forest machine learning model to determine the most discriminating variables for a resolved envelope around a given star. We constructed polarization maps for all stars in the sample, revealing a wide diversity of circumstellar morphologies. We detected 16 dusty circumstellar envelopes, including three never observed before.

They display a wide range of morphologies, all of them showing a clear departure from spherical symmetry, indicating interaction with a companion or asymmetric mass ejections. The random forest model identified optimal thresholds for several physical parameters, thus providing robust criteria to anticipate SPHERE's ability to resolve dust envelopes around AGB stars. These results facilitate the selection of targets for future observations and contribute to a better understanding of the evolution mechanisms of circumstellar envelopes.