We develop a new particle spray algorithm to model the formation of globular cluster streams. The algorithm ejects stream tracer particles based on 6D phase space distributions calibrated to N-body simulations. It successfully matches the morphology and kinematics of simulated streams with higher accuracy than most existing particle spray methods. The new algorithm is implemented in galactic dynamics codes agama, gala, galax, and galpy for public use. See the tutorial notebook and Chen et al. (2024) for more details.

Figure: positions and velocity vectors of escaped stream particles near the Lagrange points.

We apply an updated version of the Chen & Gnedin (2023) model to a sample of simulation galaxies matching various observational characteristics of the Milky Way. The model generates a catalogue of model globular clusters reproducing the total mass, mass function, metallicity distribution, radial profile, and velocity dispersion of the Galactic globular cluster systems, see Chen & Gnedin (2024a). The catalogue is publicly available here. Based on this catalogue, I examine widely-used clustering methods to categorize the progenitors of globular clusters, using their age, chemical, and kinematic properties, and subsequently apply the same methods and properties to Galactic clusters. The categorization is detailed in Chen & Gnedin (2024b).

Figure: classification of the Milky Way globular clusters shown in the integral of motion plane.

We adjust Chen & Gnedin (2022) to satellite galaxies in the local group from a suite of collisionless simulations. We validate the near-linear globular cluster mass-halo mass correlation down to M_h ~ 10⁸ M_sun, where the majority of dwarf galaxies do not host any cluster. By studying two Fornax-like satellites in the simulations, we reproduce the radial profile of globular clusters in Fornax and show that observational samples can be notably biased by incompleteness below detection limit and at large radii. See Chen & Gnedin (2023) for detais.

Figure: near-linear globular cluster mass-halo mass correlation.

The advent of the Gaia mission has enabled detailed chemical and kinematic studies of the Galactic globular clusters and revolutionized our understanding of the connections between globular cluster properties and galaxy assembly. I update the globular cluster formation model developed by Choksi & Gnedin (2019) by assigning globular clusters to particles in the IllustrisTNG simulation based on age and location. This adds spatial and kinematic information to the modeled globular clusters. The model successfully reproduces the radial distribution and various kinematic properties of the Galactic globular clusters. See Chen & Gnedin (2022) for details.

Figure: effective radii of globular cluster systems vs total mass of host halos.

We perform a suite of hydrodynamical simulations to investigate the effects of initial density profiles on the evolution of star clusters in giant molecular clouds using the moving-mesh code arepo. We find that the uniform profile follows a "hierarchical" cluster formation mode, while the steep power-law profiles show an "accretion" dominated mode. These two cluster formation modes lead to different proprieties of the most massive clusters in giant molecular clouds. See our first paper (Chen, Li, & Vogelsberger 2021) for details.

Figure: "hierarchical" mode of cluster formation.

Based on the cosmic light speed variation, we find a novel pre-burst stage for gamma-ray bursts. We also employ a primary clustering method of machine learning to classify this stage with the data from the Fermi telescope. The work was completed in 2019 but was accepted for publication in 2021 due to the COVID-19 pandemic (see, Chen & Ma 2021).

Figure: a classification of gamma-ray burst photons that distinguishes pre-burst from main-burst.