Abstract
We contrast the latest observations of the cosmic metal density in neutral gas (.met,neu) with three cosmological galaxy evolution simulations: L-GALAXIES 2020, TNG100, and EAGLE. We find that the fraction of total metals that are in neutral gas is <40 per cent at 3 less than or similar to z less than or similar to 5 in these simulations, whereas observations of damped Lyman-alpha (DLA) systems suggest greater than or similar to 85 per cent. In all three simulations, hot, low-density gas is also a major contributor to the cosmic metal budget, even at high redshift. By considering the evolution in cosmic SFR density (rho(SFR)), neutral gas density (rho(HI)), and mean gas-phase metallicity ([< M/H >](neu), we determine two possible ways in which the absolute rho(met,neu) observed in DLAs at high redshift can be matched by simulations: (i) the rho(SFR) at z greater than or similar to 3 is greater than inferred from current FUV observations, or (ii) current high-redshift DLA metallicity samples have a higher mean host mass than the overall galaxy population. If the first is correct, TNG100 would match the ensemble data best, however there would be an outstanding tension between the currently observed rho(SFR) and rho(met,neu). If the second is correct, L-GALAXIES 2020 would match the ensemble data best, but would require an increase in neutral gas mass inside subhaloes above z similar to 2.5. If neither is correct, EAGLE would match the ensemble data best, although at the expense of overestimating [< M/H >](neu). Modulo details related to numerical resolution and HI mass modelling in simulations, these incompatibilities highlight current tensions between key observed cosmic properties at high redshift.