Astron. Astrophys. 340, 371-380 (1998)

5. Discussion and conclusions

New broad-band linear polarization measurements have been obtained for a sample of 42 optically selected QSOs including 29 BAL QSOs (14 LIBAL and 13 HIBAL). The polarization properties of the different sub-classes have been compared, and possible correlations with various spectral indices searched for. The main results of our study are:

• Nearly all highly polarized QSOs of our sample belong to the LIBAL class (provided that BAL QSOs with weaker low-ionization features are included in the class).

• The range of polarization is significantly larger for LIBAL QSOs than for HIBAL and non-BAL QSOs. It extends from 0% to 4.4%, with a peak near 2%.

• There is some indication that HIBAL QSOs as a class may be more polarized than non-BAL QSOs and therefore intermediate between LIBAL and non-BAL QSOs, but the statistics are not compelling from the sample surveyed thus far.

• We confirm the fact that LIBAL QSOs (including weaker ones) have larger balnicity indices and more reddened continua than HIBAL QSOs.

• The continuum polarization appears correlated with the line profile detachment index, especially in the LIBAL QSO sub-sample.

• No correlation is found between polarization and the strength of the low- or the high-ionization absorption features, nor with the strength or the width of the emission lines. The apparent correlation between polarization and the slope of the continuum is probably due to the different distribution of these quantities within the HIBAL and LIBAL sub-samples.

The fact that LIBAL QSOs have different polarization properties is an additional piece of evidence that these objects could constitute a different class of radio-quiet QSOs, as suggested by several authors (WMFH, Sprayberry & Foltz 1992, and Boroson & Meyers 1992), HIBAL QSOs being much more similar to non-BAL QSOs. The higher maximum polarization observed in LIBAL QSOs is probably related to the larger amount of absorbing material and/or dust, either via the presence of additional scatterers (dust or electrons), or via an increased attenuation of the direct continuum.

The correlation between the continuum polarization and the detachment index was unexpected, especially since the latter index is a rather subtle characteristic of the line profiles which involves both absorption and emission components. The correlation is in the sense that LIBAL QSOs with detached profiles are less polarized in the continuum, while those with P Cygni-type profiles are more polarized. The most obvious explanation for such a correlation is that the high-ionization line profiles and the continuum polarization both depend on the geometry and/or the orientation of the LIBAL QSOs. This would explain that a range of polarization degrees is in fact observed, the maximum value being characteristic of the class. It is not excluded that HIBAL QSOs behave similarly within a smaller polarization range.

Murray et al. (1995) proposed a BAL flow model which accounts for many of the observed BAL profiles including the detached ones. Instead of being accelerated radially from a central source, the flow emerges from the accretion disk at some distance from the central source. It is then exposed to the continuum radiation and accelerated, rapidly reaching radial trajectories. The wind has naturally a maximum opening angle, and may produce polarization in the continuum via electron scattering. Other recent models are also based on such a "wind-from-disk" paradigm, and may result in roughly similar geometry and kinematics although acceleration mechanisms, photoionization, cloud size and filling factor could significantly differ (de Kool & Begelman 1995, Königl & Kartje 1994, Emmering et al. 1992).

Murray et al. (1995) show that for a flow seen nearly along the disk, P Cygni-type profiles with black troughs at low velocities are naturally produced. For the flow seen at grazing angle along the upper edge of the wind, high-velocity detached absorptions are obtained. Since the direct continuum is expected to be more attenuated for lines of sight near the disk, the continuum polarization is expected to be higher for orientations which produce P Cygni-type profiles than for orientations which produce detached profiles. This is in good qualitative agreement with the observed correlation. This mechanism has already been proposed by Goodrich (1997) to explain the higher polarization of some PHL5200-like (i.e. P Cygni-type) BAL QSOs. The polarization being uncorrelated with the slope of the continuum in the LIBAL QSO sub-sample, this differential attenuation should be dominated by electron scattering in the wind. In fact, the electron scattering models of Brown & McLean (1977) also account for the observed behavior. For the cylindrical sector geometry which roughly characterizes the "wind-from-disk" models, Brown & McLean (1977) found that the observed polarization is given by , where i is the inclination of the system ( for the disk in the plane of the sky), the opening half-angle of the wind, R its maximum extension, and a uniform electron density. With this geometry, polarization is higher along the equatorial line of sight (i = ) than along any other line of sight, again in good agreement with the observed correlation. In addition to this orientation effect within the LIBAL QSO sub-class, we see that the higher wind opacities (from either density or size) or the larger covering factors (up to ) which possibly distinguish LIBAL QSOs from HIBAL QSOs lead to higher maximum polarizations, as observed. While these models are certainly too simple to reproduce quantitatively the observations, the good overall agreement is encouraging.

A problem with the "wind-from-disk" model is that low-ionization features are assumed to be formed near the disk and therefore only observable for nearly equatorial lines of sight (Murray et al. 1995); low-ionization absorption troughs and high-ionization detached profiles are apparently mutually exclusive. Since this is not the case observationally, we have to admit that low-ionization features could form at large distance from the core also along inclined views. In this case, low-ionization features could be observed not only at the low-velocity end of the high-ionization troughs, but also at higher velocities. And indeed, more complex velocity structures are observed in the low-ionization troughs of two LIBAL QSOs with detached profiles, 0335-3339 and 1231+1320 (Voit et al. 1993), giving some support to this hypothesis. Assuming more extended LIBAL regions would also imply that LIBAL and HIBAL QSOs are different objects, in agreement with other studies (e.g. Boroson & Meyers 1992). Possibly, the efficiency of the X-ray shielding could make the difference.

While unexpected a priori, the correlation found between LIBAL QSO line profiles and continuum polarization fits reasonably well the "wind-from-disk" models, without the need of ad-hoc explanations. Clearly, the possibility of more extended LIBAL regions should be investigated theoretically. More detailed polarization differences between objects with detached and with P Cygni-type profiles should be carefully investigated, namely using spectropolarimetry. Also, possible differences between the X-ray properties of LIBAL and HIBAL QSOs would be worthwhile to detect.

© European Southern Observatory (ESO) 1998

Online publication: November 9, 1998
helpdesk.link@springer.de