We develop and implement numerical methods for including stellar feedback in galaxy-scale numerical simulations. Our models include simplified treatments of heating by Type I and Type II supernovae, gas recycling from young stars and asymptotic giant branch winds, heating from the shocked stellar winds, H ii photoionization heating and radiation pressure from stellar photons. The energetics and time dependence associated with the feedback are taken directly from stellar evolution models. We implement these stellar feedback models in smoothed particle hydrodynamic simulations with pc-scale resolution, modelling galaxies from Small Magellanic Cloud (SMC) like dwarfs and Milky Way (MW) analogues to massive z∼ 2 star-forming discs. In the absence of stellar feedback, gas cools rapidly and collapses without limit into dense sub-units, inconsistent with observations. By contrast, in all cases with feedback, the interstellar medium (ISM) quickly approaches a statistical steady state in which giant molecular clouds (GMCs) continuously form, disperse and re-form, leading to a multiphase ISM. In this paper, we quantify the properties of the ISM and GMCs in this self-regulated state. In a companion paper we study the galactic winds driven by stellar feedback.
Our primary results on the structure of the ISM in star-forming galaxies include the following.
- 1Star-forming galaxies generically self-regulate so that the cool, dense gas maintains Toomre’s Q∼ 1. Most of the volume is occupied by relatively diffuse hot gas, while most of the mass is in dense GMC complexes created by self-gravity. The phase structure of the gas and the gas mass fraction at high densities are much more sensitive probes of the physics of stellar feedback than integrated quantities such as the Toomre Q or gas velocity dispersion.
- 2Different stellar feedback mechanisms act on different spatial (and density) scales. Radiation pressure and H ii gas pressure are critical for preventing runaway collapse of dense gas in GMCs. Shocked supernova ejecta and stellar winds dominate the dynamics of the volume-filling hot gas. However, this gas primarily vents out of the star-forming disc and contributes only modestly to the mid-plane ISM pressure.
- 3The galaxy-averaged star formation rate is determined by feedback, with different mechanisms dominating in different galaxy types. For a given feedback efficiency, restricting star formation to molecular gas or modifying the cooling function has little effect on the star formation rate in the galaxies we model (including an SMC-mass dwarf). By contrast, changing the feedback mechanisms or assumed feedback efficiencies directly translates to shifts off of the observed Kennicutt–Schmidt relation.
- 4Self-gravity leads to GMCs with an approximately self-similar mass function ∝M−2, with a high-mass cut-off determined by the characteristic Jeans/Toomre mass of the system. In all of our galaxy models, GMCs live for a few dynamical times before they are disrupted by stellar feedback. The net star formation efficiency in GMCs ranges from ∼1 per cent in dwarfs and MW-like spirals to nearly ∼10 per cent in gas-rich rapidly star-forming galaxies. GMCs are approximately virialized, but there is a large dispersion in the virial parameter for a given GMC mass, and lower mass GMCs tend to be preferentially unbound.