6 Comparison with literature data

Although most published spectra of high redshift galaxies lack the S/N required to carry out a study of the type presented here we tried to compare our results with the limited information available on this subject in the literature. For this purpose we measured the C  IV equivalent width for all high-z galaxy spectra published in the papers listed in the caption of Table 4. As in our samples, spectra with strong Ly$\alpha$emission were disregarded, and only equivalent width values with estimated (rest frame) mean errors $\leq$1.0 Å were used for the comparison. The number of the (for our purpose) usable spectra of each publication is listed in Col. 4 of Table 4. (Tests showed that including less accurate data lead to similar results, but with much larger statistical errors). For the objects observed by Yee et al. (1996) and by Pettini et al. (1998, 2000) the author kindly made their spectra available to us in electronic form. Hence, to measure $W_{0}(\mbox{C~{\sc iv}})$ in these literature objects we were able to apply exactly the same procedure as used for the galaxies presented in this work. For the objects investigated by Steidel et al. (1996a, 1998), Lowenthal et al. (1997) and Trager et al. (1997) we measured the $W_{0}(\mbox{C~{\sc iv}})$ from enlarged tracings. We tested the reliability of measuring $W_{0}(\mbox{C~{\sc iv}})$ from tracings using some of our own high-z galaxies. The difference between the two measurement methods turned out to be $\leq$$10 \%$. Although the spectral resolutions of the different investigations are not exactly the same, the individual resolutions are sufficiently close to allow a direct comparison within the accuracy needed here. Table 4 and Fig. 10 show that the average C  IV values derived from the literature high-z spectra are in reasonably good agreement with the mean values derived for the galaxies investigated in this work. However, the scatter of the literature data at high redshift is larger. Although this larger scatter is presumably dominated by the on average low S/N of the literature spectra, we cannot exclude that environmental effects may influence the evolution of the C  IV strength at high redshifts.

Figure 9: Measured C  IV $\lambda$ 1550 rest-frame equivalent widths of the high-z FDF galaxies as a function of the UV luminosity $L_{{\rm UV}}$, derived from the flux between 1432 Å and 1532 Å (as defined by Kinney et al. 1993).

The best investigated individual high-z galaxy is, so far, the gravitationally magnified object MS1512-cb58 (z= 2.727) (cf. e.g. Yee et al. 1996; Seitz et al. 1998; Pettini et al. 2000; Teplitz et al. 2001; Savaglio et al 2002). By measuring the C  IV equivalent width on low resolution spectra and using Eq. (3) we obtain for this galaxy a metallicity of $0.4~Z_{\odot}$ and $0.2~Z_{\odot}$ from P00's and Y96's data, respectively. Within our error limits these values are in good agreement with the result of Pettini et al. (2000) (who derive $0.25~Z_{\odot}$ by comparing the galaxy spectrum with synthetic starburst galaxy spectra from Leitherer et al. 2001) and Teplitz et al. 2001 (who found $0.32~Z_{\odot}$ from measuring its oxygen abundance on NIR spectra using the strong line index R₂₃ which relates (O/H) to the relative abundance of [OII], [OIII] and H$\beta$). This comparison seems to support our assumption that our calibration of the C  IV strength in terms of metallicity is applicable to high redshift objects at z=2.7, although the redshift of MS1512-cb58 is too small to estimate the accuracy of the method for the interesting z>3 objects.

Figure 10: Comparison of the observed C  IV $\lambda$ 1550 rest frame equivalent widths as a function of redshift for our galaxy sample (asterisks) with the mean values of $W_{0}(\mbox{C~{\sc iv}})$ derived from high-zgalaxy spectra from Pettini et al. 1998 (P98; open circle), Steidel et al. (1996a) (S96; open square), Steidel et al. (1998) (S98; open hexagons) and Lowenthal et al. (L98; open diamond). The bars denote mean errors. Single measurements from galaxies observed by Pettini et al. (2000) (P00; filled circle), Yee et al. (1996) (Y96; filled square) and Trager et al. (1997) (T97; filled triangle) are also shown. Note that the Y96 and P00 observed the same object namely the well known lensed galaxy MS1512-cb58.

 

 

Table 4: Lines 1-4: Average of the measured C  IV $\lambda$ 1550 rest-frame equivalent widths for high-z galaxies observed by Pettini et al. (1998; P98), Lowenthal et al. (1997; L97) and Steidel et al. (1996a & 1998; S96 & S98). Lines 5-7: Measured C  IV $\lambda$ 1550 rest frame equivalent width for MS1512-cb58 observed by Yee et al. (1996; Y96) and Pettini et al. (2000; P00) and for object DG-433 observed by Trager et al. (1997; T97). A * indicates that the equivalent widths were measured using the same method applied to the the high-z galaxies presented in this work. A indicates that the equivalent widths were measured from enlarged tracings.

<z>	$<\mbox{C~{\sc iv}}>$	m.e.(C  IV)	N	Reference
	[Å]	[Å]
2.92	3.55	1.23	5	P98*
3.05	1.80	0.39	5	L97$^{\dag }$
3.09	2.28	0.59	4	S98$^{\dag }$
3.22	2.20	0.50	2	S96$^{\dag }$
<z>	$W_{0}(\mbox{C~{\sc iv}})$	d $W_{0}(\mbox{C~{\sc iv}})$ )	N	Reference
	[Å]	[Å]
2.73	5.14	1.20	1	P00*
2.73	2.91	1.11	1	Y96*
3.34	2.30	0.70	1	T97$^{\dag }$

Figure 11: Mean of the estimated metallicities as a function of redshift for all galaxies shown in Fig. 3. The oxygen abundances and the statistical uncertainties derived from R23 ratio for 5 Lyman-break galaxies investigated by Pettini et al. (2001) and Teplitz et al. (2000) are indicated by the vertical lines: solid line: upper branch results; dashed line: lower branch result. Solar and LMC metallicities are indicated by the horizontal dotted lines.

For four further high-z Lyman-break galaxies Pettini et al. (2001) determined the oxygen abundance from NIR spectra using the strong line index R₂₃. Since the relation between R₂₃ and O/H has an upper and a lower branch these abundances show the well-known two-value ambiguity. Hence these results do not provide a reliable test of our conclusions. Nevertheless, in Fig. 11 we plot the allowed ranges of oxygen abundance for these 4 Lyman-break galaxies as well as for MS 1512-cb58 together with the metallicities of our starburst galaxies (as derived from the C  IV strength via the calibration described above) as a function of redshift. This comparison shows that all data are at least mutually compatible, although for the two highest-z galaxies from Pettini et al. (2001) only the lower-branch results are in reasonable agreement with a strong increase of metallicity from redshift $\approx$3.2 to $\approx$2.3suggested by our results.

Our results are also in line to those obtained by de Breuck et al. (2000), who find a qualitative increase of metallicity from higher to lower redshift for a sample of high-z radio galaxies. Furthermore Pettini et al. (1997) and Savaglio et al. (2000) report on evidence for a gradual chemical enrichment of the gas producing the damped Ly$\alpha$ lines in QSO spectra, although their trends are only weakly significant. Compared to Savaglio et al. (2000) we find a zero point offset of the metallicity-redshift relation of about 0.7 in $\Delta \log Z$ at z = 2.5. Such a difference is not unexpected since the metal absorbers in damped Ly$\alpha$ systems most likely sample the outermost regions of galaxies and therefore a different environment than the dense interstellar matter of which the massive stars seen in starbursts have been formed.