SIRIUS project. I. Star formation models for star-by-star simulations of star clusters and galaxy formation (2005.12906v3)

Abstract: Most stars are formed as star clusters in galaxies, which then disperse into galactic disks. Upcoming exascale supercomputational facilities will enable performing simulations of galaxies and their formation by resolving individual stars (star-by-star simulations). This will substantially advance our understanding of star formation in galaxies, star cluster formation, and assembly histories of galaxies. In previous galaxy simulations, a simple stellar population approximation was used. It is, however, difficult to improve the mass resolution with this approximation. Therefore, a model for forming individual stars that can be used in simulations of galaxies must be established. In this first paper of a series of the SIRIUS (SImulations Resolving IndividUal Stars) project, we demonstrate a stochastic star formation model for star-by-star simulations. An assumed stellar initial mass function (IMF) is randomly assigned to newly formed stars. We introduce a maximum search radius to assemble the mass from surrounding gas particles to form star particles. In this study, we perform a series of N-body/smoothed particle hydrodynamics simulations of star cluster formations from turbulent molecular clouds and ultra-faint dwarf galaxies as test cases. The IMF can be correctly sampled if a maximum search radius that is larger than the value estimated from the threshold density for star formation is adopted. In small clouds, the formation of massive stars is highly stochastic because of the small number of stars. We confirm that the star formation efficiency and threshold density do not strongly affect the results. We find that our model can naturally reproduce the relationship between the most massive stars and the total stellar mass of star clusters. Herein, we demonstrate that our models can be applied to simulations varying from star clusters to galaxies for a wide range of resolutions.