It has been established that a direct link exists between the formation history of star cluster populations and their host galaxies. However, our lack of understanding of the assembly of star cluster populations impedes our ability to use them as tracers of galaxy evolution. In this work we introduce a new variation of the L-Galaxies 2020 semi-analytic galaxy formation model that includes the formation of star clusters above 104 M⊙ and probes different physical assumptions that affect their evolution over cosmic time. We used properties of different galaxy components and localised star formation to determine the bound fraction of star formation in disks. After randomly sampling masses from an environmentally dependent star cluster initial mass function, we assigned to each object a half-mass radius, metallicity, and distance from the galaxy centre. We considered up to 2000 individual star clusters per galaxy and evolved their properties over time while taking into account stellar evolution, two-body relaxation, tidal shocks, dynamical friction, and re-positioning during galaxy mergers. Our simulation successfully reproduces several observational quantities, such as the empirical relationship between the absolute V-band magnitude of the brightest young star clusters and the host galaxy star formation rate, the mass function of young star clusters, and the mean metallicities of the star cluster distributions versus galaxy masses. The simulation reveals great complexity in the z = 0 star cluster population resulting from differential destruction channels and origins, including in situ populations in the disk, a major merger-induced heated component in the halo, and accreted star clusters. Model variations point out the importance of the shape of the star cluster initial mass function, the initial distribution of half-mass radii, and the relationship between the sound speed of cold gas and the star formation rate. Our new model provides new avenues to trace individual star clusters and test cluster-related physics within a cosmological set-up in a computationally efficient manner.
Hoyer et al. (2026) studied this question.