Description
The Milky Way’s formation history is encoded in its chemical diversity, shaped by both in-situ star formation and accretion. Using GALAH DR3, we construct a high-dimensional chemical space from 17 element abundances for 9,923 metal-poor stars. This space is then transformed into a lower-dimensional latent representation using Principal Component Analysis, capturing dominant patterns in chemical variation. We apply Extreme Deconvolution to cluster this latent space, identifying 10 chemically distinct stellar groups, including five within the dynamically defined thick disc, challenging the view of the thick disc as a single, homogeneous component.
We are currently deriving precise stellar ages for these found populations by mapping GALAH DR3 spectra to APOGEE DR17 using machine learning and applying the VAE-based age-determination method from Leung et al. (2023). Additionally, we are testing our chemical tagging framework on Milky Way analogues from the L-Galaxies semi-analytic cosmological simulation (Yates et al. 2024) to assess its ability to recover known merger histories. These efforts will refine our understanding of the thick disc’s formation and its role in the Milky Way’s evolution while validating this new chemical tagging approach.
