SP_Ace: a new code to derive stellar parameters and elemental abundances

Context. Ongoing and future massive spectroscopic surveys will collect large numbers (10(6)-10(7)) of stellar spectra that need to be analyzed. Highly automated software is needed to derive stellar parameters and chemical abundances from these spectra. Aims. We developed a new method of estimating the stellar parameters T-eff, log g, [M/H], and elemental abundances. This method was implemented in a new code, SP_Ace (Stellar Parameters And Chemical abundances Estimator). This is a highly automated code suitable for analyzing the spectra of large spectroscopic surveys with low or medium spectral resolution (R = 2000-20 000). Methods. After the astrophysical calibration of the oscillator strengths of 4643 absorption lines covering the wavelength ranges 5212-6860 angstrom and 8400-8924 angstrom, we constructed a library that contains the equivalent widths (EW) of these lines for a grid of stellar parameters. The EWs of each line are fit by a polynomial function that describes the EW of the line as a function of the stellar parameters. The coefficients of these polynomial functions are stored in a library called the "GCOG library". SP_Ace, a code written in FORTRAN95, uses the GCOG library to compute the EWs of the lines, constructs models of spectra as a function of the stellar parameters and abundances, and searches for the model that minimizes the chi(2) deviation when compared to the observed spectrum. The code has been tested on synthetic and real spectra for a wide range of signal-to-noise and spectral resolutions. Results. SP_Ace derives stellar parameters such as T-eff, log g, [M/H], and chemical abundances of up to ten elements for low to medium resolution spectra of FGK-type stars with precision comparable to the one usually obtained with spectra of higher resolution. Systematic errors in stellar parameters and chemical abundances are presented and identified with tests on synthetic and real spectra. Stochastic errors are automatically estimated by the code for all the parameters. A simple Web front end of SP_Ace can be found at http://dc.g-vo.org/SP_ACE, while the source code will be published soon.