4.3. The Oxygen Abundance in H II Regions
How typical are these results of the Lyman break galaxy population as a whole? In the nearby universe, element abundances in star forming regions have traditionally been measured from the ratios of optical emission lines from H II regions. At z = 3 these features move to near-infrared (IR) wavelengths and have only become accessible in the last two years with the commissioning of high resolution spectrographs on the VLT (ISAAC) and Keck telescopes (NIRSPEC). Using these facilities, our group has recently completed the first spectroscopic survey of Lyman break galaxies in the near-IR, bringing together data for 19 LBGs; the galaxies are drawn from the bright end of the luminosity function, from ~ L* to ~ 4 L* (Pettini et al. 2001). Figure 26 shows an example of the quality of spectra which can be secured with a 2-3 hour integration. In five cases we attempted to deduce values of the abundance of oxygen by applying the familiar R23 = ([O II]+[O III]) / H method first proposed by Pagel et al. (1979). We found that generally there remains a significant uncertainty, by up to 1dex, in the value of (O/H) because of the double-valued nature of the R23 calibrator (see Figure 27).
Figure 26. Example of a NIRSPEC K-band spectrum of a Lyman break galaxy from the survey by Pettini et al. (2001). The objects targeted typically have K = 21 (on the Vega scale) and remain undetected in the continuum. However, the nebular emission lines of [O III] 4859, 5007, [O II]3727 (not shown), and H usually show up clearly with exposure times of 2-3 hours. The dotted line is the 1 error spectrum.
Figure 27. Oxygen abundance from the R23 = ([O II] + [O III]) / H ratio. In each panel the continuous lines are the calibration by McGaugh (1991) for the ionisation index O32 = [O III] / [O II] appropriate to that object. The shaded area shows the values allowed by the measured R23 and its statistical 1 error. The broken horizontal line gives for reference the solar abundance 12 + log(O/H) = 8.83 from the compilation by Grevesse & Sauval (1998); the recent revision by Holweger (2001) would bring the line down by 0.09 dex.
Thus, in the galaxies observed, oxygen could be as abundant as in the interstellar medium near the Sun or as low as ~ 1/10 solar. When the R23 method is applied to cB58, a similar ambiguity obtains (Teplitz et al. 2000). The results from the analysis of the interstellar absorption lines described above (Section 4.2) resolve the issue by showing that the upper branch solution is favoured (we have no reason to suspect that the neutral and ionised ISM have widely different abundances). It remains to be established whether this is also the case for other LBGs.
In the near future this work will shift to lower and more easily accessible redshifts near z = 2.2 where all the lines of interest, from [O II] 3727 to H, fall in near-IR atmospheric transmission windows. Nevertheless, the determination of element abundances from nebular emission lines will remain a time consuming task until multi-object spectrographs operating at near-IR wavelengths become available on large telescopes, or until the launch of the Next Generation Space Telescope (Kennicutt 1998b).