Using the statistics of pixel optical depths , we compare H i , C iv and C iii absorption in a set of six high quality z \sim 3 - 4 quasar absorption spectra to that in spectra drawn from two different state-of-the-art cosmological simulations that include galactic outflows .
We find that the simulations predict far too little C iv absorption unless the UVB is extremely soft , and always predict far too small C iii/C iv ratios .
We note , however , that much of the enriched gas is in a phase ( T \sim 10 ^ { 5 } -10 ^ { 7 } K , \rho / \langle \rho \rangle \sim 0.1 - 10 , Z \gtrsim 0.1 Z _ { \odot } ) that should cool by metal line emission – which was not included in our simulations .
When the effect of cooling is modeled , the predicted C iv absorption increases substantially , but the C iii/C iv ratios are still far too small because the density of the enriched gas is too low .
Finally , we find that the predicted metal distribution is much too inhomogeneous to reproduce the observed probability distribution of C iv absorption .
These findings suggest that strong z \lesssim 6 winds can not fully explain the observed enrichment , and that an additional ( perhaps higher- z ) contribution is required .