Hopkins et al., available on arXiv.

Abstract: The Feedback In Realistic Environments (FIRE) project explores the role of feedback in cosmological simulations of galaxy formation. Previous FIRE simulations used an identical source code (FIRE-1) for consistency. Now, motivated by the development of more accurate numerics (hydrodynamic solvers, gravitational softening, supernova coupling) and the exploration of new physics (e.g. magnetic fields), we introduce FIRE-2, an updated numerical implementation of FIRE physics for the GIZMO code. We run a suite of simulations and show FIRE-2 improvements do not qualitatively change galaxy-scale properties relative to FIRE-1. We then pursue an extensive study of numerics versus physics in galaxy simulations. Details of the star-formation (SF) algorithm, cooling physics, and chemistry have weak effects, provided that we include metal-line cooling and SF occurs at higher-than-mean densities. We present several new resolution criteria for high-resolution galaxy simulations. Most galaxy-scale properties are remarkably robust to the numerics that we test, provided that: (1) Toomre masses (cold disk scale heights) are resolved; (2) feedback coupling ensures conservation and isotropy, and (3) individual supernovae are time-resolved. As resolution increases, stellar masses and profiles converge first, followed by metal abundances and visual morphologies, then properties of winds and the circumgalactic medium. The central (~kpc) mass concentration of massive (L*) galaxies is sensitive to numerics, particularly how winds ejected into hot halos are trapped, mixed, and recycled into the galaxy. Multiple feedback mechanisms are required to reproduce observations: SNe regulate stellar masses; OB/AGB mass loss fuels late-time SF; radiative feedback suppresses instantaneous SFRs and accretion onto dwarfs. We provide tables, initial conditions, and the numerical algorithms required to reproduce our simulations.