Galactic Archaeology using Machine learning and Novel Data

Galactic Archaeology seeks to reconstruct the formation and evolution of the Milky Way using stars as fossil records of past processes. Their ages, motions, and chemical compositions provide unique insights into how spiral galaxies assemble. This thesis addresses open questions in Galactic Archaeology by combining survey data with modern methods, focusing on three pillars: stellar ages, kinematics, and chemical abundances. The first study derives precise stellar ages by combining high-resolution spectroscopy with asteroseismology. Spectroscopic parameters and abundances were used with asteroseismic scaling relations, alongside theoretical ages from stellar evolutionary models and chemical ages from [Y/Mg] and [C/N]. Asteroseismic and theoretical ages showed excellent agreement, while chemical clocks proved unreliable, underscoring the need for refinement. The second study explores the kinematics of metallicity populations in Omega Centauri, the Galaxy’s most massive and complex globular cluster. Despite its wide metallicity spread (Δ[Fe/H] ≈ 2 dex) and distinct chemical behaviours between metal-poor and metal-rich stars, Gaia and Hubble astrometry revealed no significant kinematic differences among populations. All exhibit uniform rotation, suggesting either an accreted population that has fully mixed or in-situ formation through self-enrichment. The third study develops LRPayne, a neural-network–based spectroscopic pipeline optimized for low-resolution optical spectra. Trained on synthetic grids and validated on benchmark and metal-poor stars, it recovers stellar parameters and elemental abundances with good internal accuracy and robustness to moderate signal-to-noise. Limitations remain, particularly systematic biases and mismatches between models and observations. Together, these studies integrate ages, kinematics, and abundances to refine our picture of the Milky Way. They highlight where methods converge or diverge, constrain the dynamical state of a complex stellar system, and establish machine learning as a practical tool for survey-scale spectroscopy.