Using Machine-Learning to obtain Be star candidates

YAEL J. AIDELMAN; CARLOS ESCUDERO

La Plata

Congreso; IX LA PLATA INTERNATIONAL SCHOOL SPLUS: The Universe in True Colors; 2020

Facultad de Ciencias Astronómicas y Geofisicas de la Universidad Nacional de La Plata

Classical B-type emission line (Be) stars are non-supergiant stars whose spectrum exhibit or have exhibited emission in Hα line, interpreted as arising from an optically-thin gaseous circumstellar disk. One technique commonly used to identify classical Be stars candidates is the use of photometric 2-colour diagrams (2-CDs) that utilize a narrow-band filter centered on Hα and an associated filter that samples the nearby continuum region. However, it is necessary to consider that these 2-CDs can identify other astrophysical sources that emit at Hα besides classical Be stars, such as supergiants, luminous blue variables, Wolf-Rayet stars, B[e] stars, Herbig Be stars, etc. In this work, we use supervised machine learning algorithms to disentangle OB-type stars with emission lines, sub- and over-luminous, and normal OB stars. The algorithm learning was trained using VPHAS data in filters u, g, r, Hα, and i. Subsequently, this classification algorithm was applied to a new IPHAS + SDSS dataset in order to separate the OB candidates in the different classes mentioned above.