Carbon stars from LAMOST using machine learning

Using an efficient machine-learning algorithm, we find these stars from more than 7 million spectra. As a by-product, 17 carbon-enhanced metal- poor turnoff star candidates are also reported in this paper, and they are preliminarily identified by their atmospheric parameters.

Except for 176 stars that could not be given spectral types, we classify the other 2475 carbon stars into five subtypes: 864 C-H, 226 C-R, 400 C-J, 266 C-N, and 719 barium stars based on a series of spectral features. Furthermore, we divide the C-J stars into three subtypes, C-J(H), C-J(R), and C-J(N), and about 90% of them are cool N-type stars as expected from previous literature. Besides spectroscopic classification, we also match these carbon stars to multiple broadband photometries.