Supernova-driven winds in simulated dwarf galaxies

Abstract

We investigate galactic winds driven by supernova (SN) explosions in an isolated dwarf galaxy using high-resolution (particle mass |$m_{\rm gas} = 1\, {\rm M_\odot}$|⁠, number of neighbour Nngb = 100) smoothed-particle hydrodynamics simulations that include non-equilibrium cooling and chemistry, individual star formation, stellar feedback, and metal enrichment. Clustered SNe lead to the formation of superbubbles which break out of the disc and vent out hot gas, launching the winds. We find much weaker winds than what cosmological simulations typically adopt at this mass scale. At the virial radius, the time-averaged loading factors of mass, momentum, and energy are 3, 1, and 0.05, respectively, and the metal enrichment factor is 1.5. Winds that escape the halo consist of two populations that differ in their launching temperatures. Hot gas acquires enough kinetic energy to escape when launched while warm gas does not. However, warm gas can be further accelerated by the ram pressure of the subsequently launched hot gas and eventually escape. The strong interactions between different temperature phases highlight the caveat of extrapolating properties of warm gas to large distances based on its local conditions (e.g. the Bernoulli parameter). Our convergence study finds that wind properties converge when the cooling masses of individual SNe are resolved, which corresponds to |$m_{\rm gas}=5 \, {\rm M_\odot}$| with an injection mass of |$500 \, {\rm M_\odot}$|⁠. The winds weaken dramatically once the SNe become unresolved. We demonstrate that injecting the terminal momentum of SNe, a popular sub-grid model in the literature fails to capture SN winds irrespective of the inclusion of residual thermal energy.