Annu. Rev. Astron. Astrophys. 1999. 37: 487-531 Copyright © 1999 by Annual Reviews. All rights reserved

3.3. Broad Absorption Line Results

One common characteristic of BAL spectra is that the metallic resonance lines like CIV 1548,1951, SiIV 1394,1403, NV 1239,1243, and OVI 1032,1038 are typically strong (deep) compared with Ly (e.g. Figure 8). This result and the fact that low-ionization lines like MgII 2796,2804 and FeII (UV) are usually absent indicates that the BALR ionization is generally high (Turnshek 1984;, Weymann et al. 1981, 1985). However, quantitative studies of the ionization have repeatedly failed to explain the measured line strengths with solar abundances. These difficulties were first noted by Junkkarinen (1980), Turnshek (1981; see also Weymann & Foltz 1983), who showed that photoionization models with power-law ionizing spectra and solar abundances underpredict the metal ions, especially SiIV, by large factors relative to HI. A straightforward conclusion is that the metallicities are well above solar. Turnshek (1986, 1988), Turnshek et al. (1987) estimated metal abundances (C/H) of 10 to 100 times solar and provided tentative evidence for some extreme metal-to-metal abundance ratios such as P/C 100 times solar.

Better data in the past 10 years have done nothing to change these startling results (e.g. Turnshek et al. 1996). The early concerns about unresolved line components (Junkkarinen et al. 1987, Kwan 1990) have gone away, thanks to spectroscopy with the Keck 10-m telescope at resolutions (~ 7 km s^-1) close to the thermal speeds (Barlow & Junkkarinen 1994, VT Junkkarinen, personal communication). The previously tentative detections of PV 1118,1128 absorption, which led to the large P/C abundance estimates, have now been confirmed in two objects by excellent wavelength coincidences, by the predicted weakness of nearby lines like FeIII 1122, and in one case by the probable presence of PIV 951 absorption (Junkkarinen et al. 1997, Hamann 1998; Figure 8). The commonality of PV absorption is not yet known (see also Korista et al. 1992, Turnshek et al. 1996), but its relative strength in just the two cases is surprising because the solar P/C ratio is only ~ 0.001.

More complex theoretical analyses, considering a range of ionizing spectral shapes or multiple ionization zones, also do not change the main result for metallicities and P/C ratios well above solar (Weymann et al. 1985;, Turnshek et al. 1987, 1996, 1997;, Korista et al. 1996). Hamann (1997) used the analysis in Section 3.2 to determine how high the abundances must be, given the measured column densities and a photoionized BALR. He showed that average BALR column densities require [C/H] and [N/H] > 0 and [Si/H] > 1.0 for any range of ionizations and reasonable spectral shapes. The conservatively low values of IC [corresponding to the f(M_i) at their peaks] indicate [C/H] and [N/H] 1.0 and [Si/H] 1.7. The results for individual BAL systems can be much higher. In PG1254+047 (Figure 8; Hamann 1998) the inferred minimum abundances are [C/H] and [N/H] 1.0, [Si/H] 1.8 and [P/C] 2.2.

However, we now argue that all of these BAL abundance results are incorrect, because partial coverage effects have led to generally underestimated column densities.

3.3.1. Uncertainties and Conclusions

There is now direct evidence for partial coverage in some BALQSOs based on widely separated lines of the same ion (Arav et al. 1999) and resolved doublets in several narrow BALs and BAL components [Telfer et al. 1999, Barlow & Junkkarinen 1994, Wampler et al. 1995, Korista et al. 1992 - confirmed by VT Junkkarinen (personal communication)]. Although most of this evidence applies to narrow features, it is noteworthy that there are no counter-examples to our knowledge - in which narrow line components associated with BALs indicate complete coverage (also VT Junkkarinen, personal communication).

There is also circumstantial evidence for partial coverage in BAL systems; namely, (a) spectropolarimetry indicates that BAL troughs can be filled in by polarized flux (probably from an extended scattering region) that is not covered by the BALR (Figure 12; Goodrich & Miller 1995, Cohen et al. 1995, Hines & Wills 1995, Schmidt & Hines 1999); (b) some BAL systems have a wide range of lines with suspiciously similar strengths or flat-bottom troughs that do not reach zero intensity (Arav 1997); (c) Voit et al. (1993) made a strong case for low-ionization BALRs being optically thick at the Lyman limit, which implies large optical depths in Ly; yet the Ly troughs are not generally black in these systems; (d) the larger column densities that follow assuming partial coverage and saturated BALs [N_H 10²² cm^-2 , (Hamann 1998)] are consistent with the large absorbing columns inferred from X-ray observations of BALQSOs (Green & Mathur 1996, Green et al. 1997, Gallagher et al. 1999).

More indirect evidence comes from the abundance results themselves. Voit (1997) noted that the derived overabundances tend to be greater for rare elements like P than for common elements like C. This is precisely what would occur if line saturation is not taken into account. The surprising detections of PV might actually be a signature of line saturation (and partial coverage) in strong lines like CIV, rather than extreme abundances (Hamann 1998). This assertion is supported by the one known NAL system with PV 1118,1128 absorption, where the doublet ratios in CIV, NV, and SiIV clearly indicate >> 1 (Barlow et al. 1997; TA Barlow, personal communication).

We conclude that BAL column densities have been generally underestimated and that the true BALR abundances are not known. Observed differences between BAL profiles that resemble simple optical-depth effects are probably caused by a mixture of ionization, coverage fraction, and optical depth differences in complex, multizone BALRs. This conclusion paints a grim picture for BAL abundance work, but it might still be possible to derive accurate column densities and therefore abundances for some BALQSOs or some portions of BAL profiles (Wampler et al. 1995, Turnshek 1997, Arav et al. 1999). Most needed are spectra at shorter rest frame wavelengths to measure widely separated lines of the same ion and thereby diagnose the coverage fractions and true optical depths (Section 3.2.2, "Column Densities and Partial Coverage"; Arav 1997, Arav et al. 1999).