P. Ghafour, S. Ansarifard, M. H. Jalali Kanafi, S. M. S. Movahed

In this study, a parameter-free group-finding method named GraphShed is introduced and evaluated using the IllustrisTNG100-1 simulation. The method utilizes top-down watershed segmentation applied to the set of separated Voronoi-induced graphs, facilitating the recognition of aggregations directly from the density field without tunable parameters or density thresholds. A galaxy group catalog constructed with GraphShed is compared with a Friends-of-Friends catalog generated from the same dataset. The $M_{200}$ distributions of the two catalogs are statistically consistent; nevertheless, other structural properties, including $R_{200}$, sphericity, compactness, spin, and centroid shift show significant differences, suggesting that GraphShed could improve several internal characteristics of the identified systems. Conversely, the two-point correlation function and the mass function of the identified galaxy systems, derived from the aforementioned methods, show consistency. A velocity-based classification of interacting pairs indicates that GraphShed provides improved separation of nearby over-densities which might otherwise be considered as components of a single larger system in position-only methods due to their positional proximity. These results demonstrate that GraphShed effectively preserves cosmological statistics while offering a more refined detection of galaxy systems and their dynamical interactions.