演講資訊

摘要：

The growing diversity of observed exoplanet systems suggests that multiple formation and migration pathways contribute to shaping present-day planetary architectures. In this study, we apply machine-learning techniques to investigate whether dynamically distinct exoplanet populations retain statistical signatures of their underlying formation histories. Using a two-stage Gaussian Mixture Model (GMM), we classify close-in exoplanets within a physically motivated parameter space including the Hill radius, Safronov number, and planet–star mass ratio. The resulting clusters are subsequently mapped onto pebble-accretion-based synthetic populations to explore trends in formation timing, gas accretion efficiency, and volatile-rich solid growth. Our results indicate that different exoplanet sub-populations are associated with systematically different formation environments and migration pathways. In particular, very massive gas giants preferentially correspond to earlier formation in gas-rich disks, while warm-Jupiter-dominated populations appear linked to distinct migration histories and accretion conditions. This work demonstrates the potential of combining machine learning with physically motivated population synthesis models to connect observed exoplanet demographics with planet formation theory at the population level.