Gaia Data Release 3: Analysis of the Gaia BP/RP spectra using the General Stellar Parameterizer from Photometry

Le résumé fourni par la source

We present the General Stellar Parameterizer from Photometry (GSP-Phot), which is part of the astrophysical parameters inference system (Apsis). GSP-Phot is designed to produce a homogeneous catalogue of parameters for hundreds of millions of single non-variable stars based on their astrometry, photometry, and low-resolution BP/RP spectra. These parameters are effective temperature, surface gravity, metallicity, absolute $M_G$ magnitude, radius, distance, and extinction for each star. GSP-Phot uses a Bayesian forward-modelling approach to simultaneously fit the BP/RP spectrum, parallax, and apparent $G$ magnitude. A major design feature of GSP-Phot is the use of the apparent flux levels of BP/RP spectra to derive, in combination with isochrone models, tight observational constraints on radii and distances. We carefully validate the uncertainty estimates by exploiting repeat Gaia observations of the same source. The data release includes GSP-Phot results for 471 million sources with $G<19$. Typical differences to literature values are 110 K for $T_{\rm eff}$ and 0.2-0.25 for $\log g$, but these depend strongly on data quality. In particular, GSP-Phot results are significantly better for stars with good parallax measurements ($\varpi/σ_varpi>20$), mostly within 2kpc. Metallicity estimates exhibit substantial biases compared to literature values and are only useful at a qualitative level. However, we provide an empirical calibration of our metallicity estimates that largely removes these biases. Extinctions $A_0$ and $A_{\rm BP}$ show typical differences from reference values of 0.07-0.09 mag. MCMC samples of the parameters are also available for 95% of the sources. GSP-Phot provides a homogeneous catalogue of stellar parameters, distances, and extinctions that can be used for various purposes, such as sample selections (OB stars, red giants, solar analogues etc.).

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.