Removing Reddening to Enhance Gaia’s XP Spectra Clustering

Abstract

Reddening, or extinction, distorts stellar spectra due to dust between the source and the observer, in this case, the Gaia satellite. It shifts energy from the blue to the red part of a spectrum, potentially confusing hot, reddened stars with cooler ones, since the cooler the star, the higher the SED at the red part. Gaia currently provides 2D extinction values, which can lead to errors for nearby stars, where a 3D model would be ideal. Our research group clusters Gaia sources using SOM maps, taking into account only their XP spectra. To improve our results, we are developing AI methods to remove extinction using denoising autoencoders and disentangling techniques trained on spectra with similar astrophysical parameters but different extinction levels.