5.4. Problems with the BALs
Absorption line optical depths and column densities are derived from the observed intensities, I, by the relation, I = Io exp(-), where Io is the unabsorbed (continuum) intensity and is the optical depth at frequency (Junkkarinen et al. 1983, Grillmair & Turnshek 1987, Korista et al. 1992, Savage & Sembach 1991, Jenkins 1996, Arav et al. 1999). Partial coverage of the background light source(s) can fill in the troughs and thus lead to underestimated optical depths and column densities. A critical difference between BALs and NALs is in our ability to resolve adjacent multiplet lines and thereby directly measure any partial coverage effects (see Hamann et al. 1997b and HF99). In general, this type of analysis is possible for only the NALs, where resolved doublets like CIV 1548,1550 or SiIV 1394,1403 are easily measurable. The blending of these features in BALs has generally made the partial-coverage analysis impossible.
Hamann (1998) used explicit calculations of the line optical depths to show that BAL spectra are consistent with much lower (solar) metallicities and solar relative abundances if at least the stronger transitions (such as CIV 1549) are more optically thick than they appear. In particular, those calculations suggest that the anomalously large ratio of PV 1121 / CIV 1549 absorption lines results from severe saturation in CIV rather than extreme P/C abundances. The line saturation is disguised in the observed (modest) BAL troughs by partial line-of-sight coverage of the background light source(s).
Arav et al. (1999) provided direct evidence for partial coverage in one BAL system based on some far-UV lines (whose true optical depth ratios are fixed by atomic physics). Similar evidence has come from a few BALs or BAL components where the profiles are narrow enough to resolve multiplet transitions that are close in wavelength (Telfer et al. 1998, Wampler et al. 1995). More indirect evidence for partial coverage in BALs has come from spectropolarimetry (Cohen et al. 1995, Goodrich & Miller 1995, Hines & Wills 1995). It is also interesting that the only known NAL system with PV absorption (Barlow et al. 1999) has resolved doublet ratios in CIV, SiIV and NV that require severe line saturation and partial line-of-sight coverage.
It is therefore likely that the column densities derived so far from BALs are generally underestimated and the abundances are generally incorrect. Future studies that involve relatively narrow BALs (Telfer et al. 1998) or the plethora of hard-to-measure far-UV lines (Arav et al. 1999) might yet allow the needed partial-coverage analysis and thus provide reliable abundance estimates.