The INT search for metal-poor stars: Spectroscopic observations and classification via artificial neural networks

With the dual aims of enlarging the list of extremely metal-poor stars iden tified in the Galaxy and boosting the numbers of moderately metal-deficient stars in directions that sample the rotational properties of the thick dis k, we have used the 2.5 m Isaac Newton Telescope and the Intermediate Dispe rsion Spectrograph to carry out a survey of brighter (primarily northern he misphere) metal-poor candidates selected from the HK objective-prism-interf erence-filter survey of Beers and collaborators. Over the course of only th ree observing runs (15 nights) we have obtained medium-resolution (lambda/d elta lambda similar or equal to 2000) spectra for 1203 objects (V similar o r equal to 11-15). Spectral absorption-line indices and radial velocities h ave been measured for all the candidates. Metallicities, quantified by [Fe/ H], and intrinsic (B-V)(0) colors have been estimated for 731 stars with ef fective temperatures cooler than roughly 6500 K by using artificial neural networks (ANNs) trained with spectral indices. We show that this method per forms as well as a previously explored Ca II K calibration technique, yet i t presents some practical advantages. Among the candidates in our sample we identify 195 stars with [Fe/H] less than or equal to -1.0, 67 stars with [ Fe/H] less than or equal to -2.0, and 12 new stars with [Fe/H] less than or equal to -3.0. Although the effective yield of metal-poor stars in our sam ple is not as large as that in previous HK survey follow-up programs, the r ate of discovery per unit of telescope time is quite high. Further developm ent of the ANN technique, with the networks being fed the entire spectrum, rather than just the spectral indices, holds the promise to produce fast, a ccurate, multidimensional spectral classifications (with the associated phy sical parameter estimates), as is required to process the large data flow p rovided by present and future instrumentation.