Seminar: Gaia's low-resolution BP/RP spectra and its use for stellar characterisation

220 million low-resolution BP/RP spectra have been published in Gaia DR3. These are nearly complete down to Gaia magnitudes brighter than 17.6mag, cover the whole sky, and are homogeneously derived. I will give an introduction into BP/RP spectra, explaining their basic characteristics and different formats.

I will explain how I have used these in a forward-modelling approach to infer the stellar parameters for 471 million sources (down to G=19) that have been published in the release. In particular, I will emphasise the importance of not normalising BP/RP spectra and instead exploiting their apparent flux levels: This, together with the parallax and photometry from Gaia, not only provides powerful constraints on stellar parameters, but also imposes specific requirements on the design of the algorithms and the models that are used. However, I will also underline that the forward modelling of BP/RP spectra is currently limited by systematics that prevent, e.g., a reliable estimation of metallicities. This limitation can be circumvented with an empirical approach and I will show how I have used machine learning in order to empirically estimate metallicity. This has lead to the discovery of about ten thousand red-giant stars with [M/H]<-2 in the inner Galaxy. Using Gaia DR3 radial velocities and APOGEE DR17 abundances, I will argue that these reveal the old metal-poor heart of the Milky Way. This application in particular highlights the quality and usefulness of the new Gaia BP/RP spectra.

---

We will go to lunch at 12:20. There will be fika in the astronomy corridor (Hus 9 Plan 5) after the seminar.