Parametrization of Binary Stars with Gaia Observations

Abstract

A large fraction of the 10⁹ sources that will be observed during the 5-year Gaia survey will be binary stars. In order to estimate the main astrophysical parameters for both stars of those systems, we have developed the multiple star classifier (MSC). The code, which will be part of the Gaia Astrophysical parameters processing chain, is based on the support vector machine algorithm for regression and will use the low-resolution spectra that will be obtained by the Gaia satellite. First tests on the performance of MSC on simulated Gaia spectra show that even though the parameter estimation for the primary stars is quite accurate, the results are very poor for the secondary star of the system. In order to improve the performance of MSC, we test how the results change by the use of additional prior information. The method makes explicit use of domain knowledge by employing a Hertzsprung–Russell diagram to constrain solutions and to ensure that they respect stellar physics. In addition, we use the parameters extracted by MSC for the primary star and the line of sight extinction in order to further constrain the parameter values of the secondary star.