Towards simulating star formation in turbulent high-z galaxies with mechanical supernova feedback

Abstract

Feedback from supernovae is essential to understanding the self-regulation of star formation in galaxies. However, the efficacy of the process in a cosmological context remains unclear due to excessive radiative losses during the shock propagation. To better understand the impact of SN explosions on the evolution of galaxies, we perform a suite of high-resolution (12 pc), zoom-in cosmological simulations of a Milky Way-like galaxy at z=3 with adaptive mesh refinement. We find that SN explosions can efficiently regulate star formation, leading to the stellar mass and metallicity consistent with the observed mass-metallicity relation and stellar mass-halo mass relation at z~3. This is achieved by making three important changes to the classical feedback scheme: i) the different phases of SN blast waves are modelled directly by injecting radial momentum expected at each stage, ii) the realistic time delay of SNe, commencing at as early as 3 Myr, is required to disperse very dense gas before a runaway collapse sets in at the galaxy centre via mergers of gas clumps, and iii) a non-uniform density distribution of the ISM is taken into account below the computational grid scale for the cell in which SN explodes. The last condition is motivated by the fact that our simulations still do not resolve the detailed structure of a turbulent ISM in which the fast outflows can propagate along low-density channels. The simulated galaxy with the SN feedback model shows strong outflows, which carry approximately ten times larger mass than star formation rate, as well as smoothly rising circular velocity. Other feedback models that do not meet the three conditions form too many stars, producing a peaked rotation curve. Our results suggest that understanding the structure of the turbulent ISM may be crucial to assess the role of SN and other feedback processes in galaxy formation theory.