A supernova feedback implementation for the astrophysical simulation software Arepo

TL;DR

This paper presents a new supernova feedback implementation for the Arepo simulation software, enabling the study of supernova-driven turbulence and its effects on the interstellar medium in galaxy simulations.

Contribution

The paper introduces a stochastic supernova feedback model for Arepo, specifically targeting high-density gas in the ISM, and validates it through comparison with theoretical models and fixed supernova rates.

Findings

Supernova remnant evolution matches theoretical predictions.

Feedback implementation induces turbulence in simulated ISM.

Model effectively simulates supernova effects over 20 Myrs.

Abstract

Supernova (SN) explosions play an important role in the development of galactic structures. The energy and momentum imparted on the interstellar medium (ISM) in so-called "supernova feedback" drives turbulence, heats the gas, enriches it with heavy elements, can lead to the formation of new stars or even suppress star formation by disrupting stellar nurseries. In the numerical simulation at the sub-galactic level, not including the energy and momentum of supernovas in the physical description of the problem can also lead to several problems that might partially be resolved by including a description of supernovas. In this thesis such an implementation is attempted for the combined numerical hydrodynamics and N-body simulation software Arepo (Springel, 2010) for the high density gas in the ISM only. This allows supernova driven turbulence in boxes of 400pc cubed to be studied. In a stochastic process a large amount of thermal energy is imparted on a number of neighbouring cells, mimicking the effect of a supernova explosions. We test this approach by modelling the explosion of a single supernova in a uniform density medium and comparing the evolution of the resulting supernova remnant to the theoretically-predicted behaviour. We also run a simulation with our feedback code and a fixed supernova rate derived from the Kennicutt-Schmidt relation (Kennicutt, 1998) for a duration of about 20 Myrs. We describe our method in detail in this text and discuss the properties of our implementation. vii