Identificación de candidatas a estrellas Be utilizando redes neuronales

Abstract

Astronomical databases currently provide large volumes of spectroscopic and photometric information. In particular, as photometric data is relatively easier to obtain due to the shorter use time of the telescope, there is an increasing need to use those data in order to automatically identify specific objects and study them in detail afterwards. In this work, we focus on the photometric identification of Be stars, early-type stars with Hα line in emission. These kind of objects are very interest for understanding the evolution of fast rotating stars, and also for the study of the physics of circumstellar disks. For their identification, we use photometric (VPHAS+, 2MASS, AlWISE) and spectroscopic (LAMOST) databases, together with machine learning techniques, such as neural networks. Our results show that using the reddening-free Q indices as features provides a significant improvement in the photometric identification of Be stars.