2016MNRAS.458..912V

Simulating dwarf galaxy haloes in a reionizing Universe puts severe constraints on the subgrid model employed in the simulations. Using the same subgrid model that works for simulations without a UV-background (UVB) results in gas-poor galaxies that stop forming stars very early on, except for haloes with high masses. This is in strong disagreement with observed galaxies, which are gas rich and star forming down to a much lower mass range. To resolve this discrepancy, we ran a large suite of isolated dwarf galaxy simulations to explore a wide variety of subgrid models and parameters, including timing and strength of the UVB, strength of the stellar feedback and metallicity-dependent Pop III feedback. We compared these simulations to observed dwarf galaxies by means of the baryonic Tully-Fisher relation (BTFR), which links the baryonic content of a galaxy to the observationally determined strength of its gravitational potential. We found that the results are robust to changes in the UVB. The strength of the stellar feedback shifts the results on the BTFR, but does not help to form gas-rich galaxies at late redshifts. Only by including Pop III feedback are we able to produce galaxies that lie on the observational BTFR and that have neutral gas and ongoing star formation at redshift zero.