Cool carbon stars in the halo: A new survey based on 2MASS

We present the first results of a new survey for finding cool N-type carbon (C) stars in the halo of the Galaxy. Candidates were first selected in the 2MASS Second Incremental Release database with JHKs colours typical of red AGB C stars and $K_{\rm s} < 13$, and subsequently checked through medium resolution slit spectroscopy. We discovered 27 new C stars plus one known previously and two similar objects in the Fornax and Sculptor dwarf galaxies. We determine and discuss the properties of our sample, including optical and near-infrared colours, radial velocities, as well as $\rm H\alpha$ emission and variability that are frequent, all these characteristics being compatible with an AGB C-type classification. Surprisingly, of the 30 studied objects, 8 were found to have small but measurable proper motions (μ) in the USNO-B1.0 catalogue, ranging over $8 < \mu < 21$ mas yr-1 and opening the possibility that some objects could perhaps be dwarf carbon stars. Yet, a detailed analysis based on comparison with the sample of known carbon dwarfs leads us to consider these μ as incompatible with the broader picture suggested by the other data taken as a whole. So, we adopt the view that all objects are of AGB type, i.e. luminous and distant. Because the stream of Sagittarius dwarf galaxy is known to be the dominant source of luminous C stars in the halo, we chose to determine distances for our sample by scaling them on the 26 known AGB C stars of the Sgr galaxy itself, which are found to be, in the Ks-band, ~0.5 mag less luminous than the average LMC C stars for a given $J-K_{\rm s}$ colour. The obtained distances of our halo stars range from 8 to 80 kpc from the Sun. Then, examination of position and radial velocities show that about half belong to the Sgr stream. Our findings suggest that numerous AGB C stars remain to be discovered in the halo. Long term Ks-band monitoring would be of great value to ascertain distance estimates through the period-luminosity relation, because a large fraction of our sample is probably made of Mira variables