7 Conclusions

Our study shows that the FDF starburst galaxies at $z \approx 3.2$ have on average significantly lower C  IV equivalent widths than starburst galaxies at lower redshifts. In view of the known close relation of the C  IV strength to the metallicity in local starburst galaxies, it appears likely that this effect is due to a significant evolution of the average metallicity in such objects at high redshifts. Using data from well studied local starburst galaxies we calibrated the C  IV strength in terms of the heavy element content of these objects. Assuming that this calibration is applicable to high redshift starburst galaxies we find that the mean cosmic metallicity as observed in starburst galaxies has increased significantly between the cosmic epochs corresponding to $z \approx 3.2$ and $\approx$2.3 (when the universe was between 2 and 3 Gyrs old). If this interpretation of the increase of the C  IV absorption strength with decreasing redshift is correct, an intense phase of star formation and evolution of massive stars must have occurred during this period. At lower redshifts (z <2.5) our data indicate little further increase of the average metallicity of starburst galaxies. Hence, the further cosmic chemical enrichment seems to have been insignificant during the last 11 Gyrs. The metallicity evolution indicated by our data is in reasonable agreement with published theoretical chemical enrichment models due to star formation at early cosmological epochs. Our results are also in agreement with a metallicity evolution found in high-z radio galaxies (de Breuck et al. 2000) and the tentative evidence for a gradual chemical enrichment of the gas producing the damped Ly$\alpha$ lines (Pettini et al. 1997; Savaglio et al. 2000).

Since the FDF observations provide a fairly complete sample of the bright starburst galaxies in the observed direction and redshift ranges, the observed chemical evolution should be characteristic of the cosmic volumes with the most intense star formation at the corresponding epochs. However, it is also clear, that our results do not apply to all objects and volumes at a certain redshift or epoch. It is, e.g., well known that the BLR gas of high-redshift QSOs is characterized by high metallicities and that no significant chemical evolution is observable in these objects up to at least z=5 (see e.g. Hamann & Ferland 1999; Dietrich et al. 1999; Dietrich & Wilhelm-Erkens 2000). Obviously, in the environment of these early QSOs much star formation and stellar evolution must have taken place at epochs corresponding to even higher redshifts. It is also known, however, (and confirmed from the FDF survey) that high-z starburst galaxies are much more frequent than bright high-z QSOs. Hence, the starburst galaxies are expected to be a more representative tracer of the history of the overall cosmic chemical evolution than the QSOs.

From our data we cannot determine whether at high redshift a metallicity-luminosity relation does exist, since we do not have any faint objects in our high-z sample. But it is evident that the high-redshift galaxies are on average overluminous for their metallicities compared with local starburst galaxies. This trend is also found by Pettini et al. (2001) and Kobulnicky & Koo (2000) for Lyman break galaxies. From tests on various subsamples we find that the observed dependence of $W_{0}(\mbox{C~{\sc iv}})$ on redshift is not caused by a luminosity effect. Furthermore we showed that this dependence is also not cause by the possible selection effect of preferentially observing galaxies with low dust content and hence low metallicity at high redshifts.

Differences in the Si  IV to C  IV ratios between local and high-zgalaxies in our sample suggest differences in the population and star formation history in the galaxies with z>3. Short bursts of star formation may have been more important (relative to periods of "continuous star formation'') at these early epochs.

Acknowledgements

We are greatly indebted to Drs. M. Pettini and H. K. C. Yee for providing the ASCII files of their high redshift spectra published in Yee et al. (1996) and Pettini et al. (1998, 2000). We thank the referee C. Leitherer for valuable comments. We also want to thank C. Tapken for helpful comments and the Paranal staff for their support. This research was supported by the German Science Foundation (DFG) (Sonderforschungsbereiche 375 and 439).