Speaker
Description
Globular clusters are among the most well-studied objects in astronomy (Renaud, 2018). The continued study thereof will likely reveal key insights into the spatial, dynamical and chemical properties of galaxies (particularly the Milky Way), stellar formation and evolution, as well as assist in applying constraints on dark matter and initial mass function models.
My Master’s project was centred on using machine learning (ML) algorithms to detect and characterise the kinematic properties of globular clusters (GCs) within the Milky Way (MW). The GCs were searched in regions residing away from the Galactic disc, described by $|b|>20^\circ$ and $l \in (0^\circ,220^\circ)$. The techniques used in our study are inspired by the OCfinder framework (Castro-Ginard et al, 2022; Hunt & Reffert, 2024).
In our study, a robust clustering algorithm, HDBSCAN, was applied to 5-D astrometric data (l, b, parallax, proper motion in RA and DEC), to detect spatial overdensities. Subsequently, a convolutional neural network (CNN) was trained using colour-magnitude diagrams (CMDs) of synthetic clusters to identify the isochrone pattern of true clusters. The CNN was then applied to the CMDs of the detected overdensities to distinguish real clusters from statistical overdensities. All the astrometric and photometric data was sourced from the Gaia mission's most recent data release, DR3. The main aims of the research were as follows: i) derive globular cluster properties that are in agreement with those reported by Vasiliev & Baumgardt (2021); ii) if possible, detect other stellar cluster types, streams, or moving groups within our search field.
Of the 28 known GCs residing in our search region, 23 were recovered. We found 6 possible GC candidates, 2 of which are likely to be components of the Sagittarius stream. We also managed to recover, by serendipity, 2 dwarf spheroidal galaxies, Draco and Ursa Minor I.
| Stream | Science or Engineering |
|---|