The origin of the galaxy mass-metallicity relation and implications for galactic outflows

Using cosmological hydrodynamic simulations that dynamically incorporate enriched galactic outflows together with analytical modelling, we study the origin of the stellar mass-gas-phase metallicity relation (MZR). We find that metallicities are driven by an equilibrium between the rate of enrichment owing to star formation and the rate of dilution owing to infall of unenriched gas. This equilibrium is in turn governed by the outflow strength. As such, the MZR provides valuable insights and strong constraints on galactic outflow properties across cosmic time. We compare three outflow models: no outflows, a 'constant-wind model that emulates the popular Dekel & Silk scenario, and a 'momentum-driven wind' model that best reproduces z greater than or similar to 2 intergalactic medium metallicity data. Only the momentum-driven wind scaling simulation is able to reproduce the observed z similar to 2 MZR's slope, amplitude, and scatter. In order to understand why, we construct a one-zone chemical evolution model guided by simulations. This model shows that the MZR in our outflow simulations can be understood in terms of three parameters: (i) the equilibrium metallicity Z(g,eq) = y circle dot(SFR)/circle dot(ACC) (where y = net yield), reflecting the enrichment balance between star formation rate circle dot(SFR) and gas accretion rate circle dot(ACC); (ii) the dilution time t(d) = M-g/M-SFR, representing the time-scale for a galaxy to return to Z(g,eq) after a metallicity-perturbing interaction; and (iii) the blowout mass M-blowout, which is the galaxy stellar mass above which winds can escape its halo. Without outflows, galaxy metallicities exceed observations by approximately two to three times, although the slope of the MZR is roughly correct owing to greater star formation efficiencies in larger galaxies. When outflows with mass-loading factor eta(W) are present, galaxies below M-blowout obey Z(g,eq) approximate to y/(1 + eta(W)), while above M-blowout, Z(g,eq) -> y. Our constant-wind model has M-blowout similar to 10(10) M-circle dot, which yields a sharp upturn in the MZR above this scale and a flat MZR with large scatter below it, in strong disagreement with observations. Our momentum-driven wind model naturally reproduces the observed Z(g) proportional to M-*(0.3) because Z(g,eq) proportional to eta(-1)(W) proportional to M-*(1/3) when eta(W) > 1 (i.e. at low masses). The flattening of the MZR at M-* greater than or similar to 10(10.5) M-circle dot observed by Tremonti et al. is reflective of the mass-scale where eta(W) similar to 1 rather than a characteristic outflow speed; in fact, the outflow speed plays little role in the MZR except through M-blowout. The tight observed MZR scatter is ensured when t(d) less than or similar to dynamical time, which is only satisfied at all masses in our momentum-driven wind model. We also discuss secondary effects on the MZR, such as baryonic stripping from neighbouring galaxies' outflows.