Synthesized Chemical Abundances from Gaia RVS Spectra

The chemical compositions of stars trace the enrichment history of the Milky Way, yet extracting precise abundances from Gaia Radial Velocity Spectrometer (RVS) spectra remains challenging due to the instrument's narrow wavelength range (~25 nm) and systematic mismatches between observations and synthetic models. Here we present a hybrid approach that cross-matches RVS stars with APOGEE DR17 to build a training set, uses the data-driven method Lux*to determine stellar labels (Teff, log g, [Fe/H], [Si/Fe]), and fixes these labels in the spectral synthesis code Korg to derive chemical abundances. We recover multiple abundances at the 0.2 dex level, and demonstrate clear spatial chemical gradients across the Galaxy as well as the well-known [α/Fe] bimodality between the thin and thick disk populations. This framework is scalable to the order-of-magnitude increase in RVS spectra anticipated with Gaia DR4.