Numerical simulations of standard cosmological scenarios have now reached the degree of sophistication required to provide tentative answers to the fundamental question: Where and when were the heavy elements formed? Averaging globally, these simulations give a metallicity that increases from 1% of the solar value at $z=3$ to 20% at present. This conclusion is, in fact, misleading, as it masks the very strong dependency of metallicity on local density. At every epoch higher density regions have much higher metallicity than lower density regions. Moreover, the highest density regions quickly approach near solar metallicity and then saturate, while more typical regions slowly catch up. These results are much more consistent with observational data than the simpler picture (adopted by many) of gradual, quasi-uniform increase of metallicity with time.

ApJ(Letters) in press, 15 latex pages and 4 figures