mufasa: galaxy formation simulations with meshless hydrodynamics
We present the <scp>mufasa</scp> suite of cosmological hydrodynamic simulations, which employs the <scp>gizmo</scp> meshless finite mass (MFM) code including <it>H</it><inf>2</inf>-based star formation, nine-element chemical evolution, two-phase kinetic outflows following scalings from the Feedback in Realistic Environments zoom simulations, and evolving halo mass-based quenching. Our fiducial (50 <it>h</it>−1 Mpc)3 volume is evolved to <it>z</it> = 0 with a quarter billion elements. The predicted galaxy stellar mass functions (GSMFs) reproduces observations from <it>z</it> = 4 → 0 to ≲ 1.2σ in cosmic variance, providing an unprecedented match to this key diagnostic. The cosmic star formation history and stellar mass growth show general agreement with data, with a strong archaeological downsizing trend such that dwarf galaxies form the majority of their stars after <it>z</it> ∼ 1. We run 25 and 12.5 <it>h</it>−1 Mpc volumes to <it>z</it> = 2 with identical feedback prescriptions, the latter resolving all hydrogen-cooling haloes, and the three runs display fair resolution convergence. The specific star formation rates broadly agree with data at <it>z</it> = 0, but are underpredicted at <it>z</it> ∼ 2 by a factor of 3, re-emphasizing a longstanding puzzle in galaxy evolution models. We compare runs using MFM and two flavours of smoothed particle hydrodynamics, and show that the GSMF is sensitive to hydrodynamics methodology at the ∼×2 level, which is sub-dominant to choices for parametrizing feedback.