CONSTRUCTING THE NEW OPEN CLUSTER CATALOGUE USING GAIA EDR3 DATA

• Main Author: Firdausya Nur C, Amatul
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/65214
• Institution: Institut Teknologi Bandung

Open clusters are a homogeneous group of stars of the same age and initial composition born from the same gas cloud; therefore, they are the ideal laboratory for studying the formation and evolution of the stellar clusters. The evolution of open clusters is driven by several prominent factors of their stellar evolution and dynamical processes, together with the external influences in form of a local galactic tidal field. This research aimed to investigate the Galactic environment effect on the spatial and velocity dispersion in a large sample of open clusters, subjected to a large sample of local open clusters in a sphere of 2.5 kpc centered on the Sun. In addition, to that, this thesis also aimed to build a very well-defined new catalogue of open clusters catalogue using the latest Gaia EDR3 data. We adopt two techniques to make the catalogue, namely the optimized DBSCAN which is applied simultaneously on 8-parameters to select the cluster’s membership, and a machine learning model to filter the cluster’s members by eliminating the fictitious members. A machine learning algorithm is implemented in Dataiku, to do the elimination task. The best model chosen, which best fit the data training set, was the Random Forest. This model was then applied to all open cluster candidates to achieve a final classification with a certain probability. The result shows that from data classification performed by ML, almost 48.6% of data are classified as non-cluster (unknown object and not a cluster); therefore, only 51.4% of the data is retained. Some possible reasons behind the false detection of the open clusters are: (i) false-negative detection, real open clusters but classified as non-cluster objects, (ii) the inability of the DBSCAN to detect accurately the stellar members from a field occupied by more than one cluster, (iii) the nature of objects classified as non-cluster by the ML selection, are probably the star-forming region (SFR), very dense region without real clusters, poor concentrations region (low number of stars), or very young clusters which not relaxed. By utilizing the catalogue result, we performed some calculations to analyze the effect of the Galactic environment on a large population of open clusters. From analysis, we obtained that there is a weak linear correlation between the clusters’ compression and their distance in the Galactic Z direction, as well as for the velocity’s dispersion ratio (Vbldisp). Nevertheless, a stronger correlation emerges between clusters’ comiii pression and their velocity dispersion ratio, implying that the Vbldisp increases as the clusters’ compression is bigger.