4 Towards an interpretation

4.1 Observational constraints

First, it is important to note that the observational facts discussed in detail in Paper I and arguing against an instrumental bias and/or a contamination by interstellar polarization in our Galaxy are still valid. They are even strengthened since additional objects have been observed, several of them by different authors with different instrumentations. For the new measurements presented here, the polarization of field objects has been measured simultaneously and found to be very small (Lamy & Hutsemékers 2000). The comparison of the quasar polarization position angles with those of neighbouring stars has also been re-investigated for the new sample. Using the most recent compilation of Heiles (2000), we confirm the absence of correlation between quasar and Galactic star polarization angles, especially in region A1. And finally, the fact that the polarization vectors of quasars on the same line of sight but at lower redshifts are not accordingly aligned certainly remains one of the strongest arguments against artifacts. Let us now discuss some observational results providing us with possible constraints on the phenomenon. This discussion is mostly based on quasars in region A1. Looking at Fig. 1, we found - by chance - that the plane of the Local Supercluster (in the direction of its center) roughly passes through the structure formed by the aligned objects. We have then transformed the polarization position angles in the supergalactic coordinate system (de Vaucouleurs et al. 1991) using Eq. (16) of Paper I. A map is illustrated in Fig. 3. It shows a rough alignment of quasar polarization vectors with the supergalactic plane, the effect being more prominent for those objects close to the supergalactic equator. The most polarized objects (with $p \geq$ 5%) do follow the trend. This behavior is very reminiscent of the alignment of Galactic star polarization vectors with the Galactic plane (Mathewson & Ford 1970; Axon & Ellis 1976).

Figure 3: A map in the supergalactic coordinate system of the polarization vectors of the 29 polarized quasars belonging to region A1. The vector length is arbitrary. Thicker lines refer to objects with $p \geq 5\%$

Other constraints come from the relation between quasar intrinsic properties and polarization. In Table 2, we give all quasars in region A1 with good polarization measurements, either polarized ( $p \geq 0.6 \%$ and $\sigma _{\theta } \leq 14\hbox {$^\circ $ }$ the latter constraint being equivalent to $p / \sigma_{\rm p} \simeq 2$), or unpolarized ($p < 0.6\%$ with $\sigma_{\rm p} \leq 0.3\%$). The quasar type is also given: broad absorption line (BAL), radio-loud non-BAL (RL), radio-quiet non-BAL (RQ), or optically selected non-BAL (O). We may first notice that the quasars with aligned polarization vectors do belong to all types, i.e. radio-quiet/optically selected (3 objects), radio-loud (9), or BAL (13). Note however that significantly polarized radio-quiet non-BAL quasars are definitely less numerous, and that two of them, B1115+080 and B1429-008, are possibly gravitationally lensed. Furthermore, we can see that the known polarization difference between BAL and non-BAL radio-quiet quasars also prevails in region A1 (Fig. 4). A Kolmogorov-Smirnov test gives a probability of only 0.6% that these two samples of are drawn from the same parent population. The illustrated distributions are also in good agreement with those reported by Hutsemékers et al. (1998). In addition, the distribution of p for non-BAL radio-quiet quasars is very similar to that found by Berriman et al. (1991) for the Palomar-Green quasar sample. This clearly indicates that, also in region A1, quasar polarization is related to the intrinsic properties of the objects.

 

 

Table 2: Polarized and unpolarized quasars in region A1

Object	z	p	$\sigma_{\rm p}$	$\theta$	$\sigma_{\theta}$	Ref	Type
(B1950)		(%)	(%)	( $\hbox{$^\circ$ }$)	( $\hbox{$^\circ$ }$)
B1115+080	1.722	0.68	0.27	46	12	0	RQ
B1120+019	1.465	1.95	0.27	9	4	0	BAL
B1127-145	1.187	1.26	0.44	23	10	2	RL
B1138-014	1.270	0.38	0.23	53	21	9	BAL
B1138+040	1.876	0.10	0.24	36	-	1	RQ
B1157-239	2.100	1.33	0.17	95	4	9	BAL
B1157+014	1.990	0.76	0.18	39	7	9	RL
B1158+007	1.380	0.44	0.20	74	14	9	RL
B1203+155	1.630	1.54	0.20	30	4	9	BAL
B1205+146	1.640	0.83	0.18	161	6	9	BAL
B1206+459	1.158	0.24	0.17	132	20	1	RQ
B1210+197	1.240	0.33	0.19	76	19	9	RL
B1212+147	1.621	1.45	0.30	24	6	0	BAL
B1215+127	2.080	0.62	0.24	17	12	9	BAL
B1216+110	1.620	0.58	0.19	63	10	9	BAL
B1219+127	1.310	0.68	0.20	151	9	9	BAL
B1222-016	2.040	0.80	0.22	119	8	9	BAL
B1222+146	1.550	0.23	0.18	65	30	9	O
B1222+228	2.046	0.84	0.24	150	8	2	RL
B1225+317	2.200	0.16	0.24	150	-	2	RL
B1228+122	1.410	0.12	0.18	142	-	9	BAL
B1230-237	1.840	0.05	0.19	72	-	9	O
B1230+170	1.420	0.30	0.19	101	22	9	BAL
B1234-021	1.620	0.54	0.18	72	10	9	O
B1235-182	2.190	1.02	0.18	171	5	9	RL
B1238-097	2.090	0.18	0.18	80	-	9	O
B1239+099	2.010	0.82	0.18	161	6	9	BAL
B1239+145	1.950	0.18	0.20	21	-	9	RQ
B1241+176	1.273	0.12	0.19	120	-	1	RL
B1242+001	2.080	0.22	0.19	56	39	9	RQ
B1246-057	2.236	1.96	0.18	149	3	8	BAL
B1246+377	1.241	1.71	0.58	152	10	2	RL
B1247+267	2.038	0.41	0.18	97	12	1	RQ
B1248+401	1.030	0.20	0.19	3	-	1	RQ
B1250+012	1.690	0.21	0.18	132	37	9	BAL
B1254+047	1.024	1.22	0.15	165	3	1	BAL
B1255-316	1.924	2.20	1.00	153	12	4	RL
B1256-220	1.306	5.20	0.80	160	4	7	RL
B1256-175	2.060	0.91	0.19	71	6	9	RL
B1258-164	1.710	0.53	0.18	132	10	9	O
B1303+308	1.770	1.12	0.56	170	14	3	BAL
B1305+001	2.110	0.70	0.22	151	9	9	O
B1309-216	1.491	12.30	0.90	160	2	4	RL
B1309-056	2.212	0.78	0.28	179	11	0	BAL
B1317+277	1.022	0.15	0.20	94	-	2	RQ
B1329+412	1.930	0.36	0.21	83	16	1	RQ
B1331-011	1.867	1.88	0.31	29	5	0	BAL
B1333+286	1.910	5.88	0.20	161	1	9	BAL
B1334+262	1.880	0.23	0.19	116	34	9	BAL
B1338+416	1.219	0.37	0.19	67	15	1	RQ
B1354-152	1.890	1.40	0.50	46	10	4	RL
B1416+067	1.439	0.77	0.39	123	14	2	RL
B1429-008	2.084	1.00	0.29	9	9	0	RQ
B1429-006	1.180	0.07	0.20	107	-	9	BAL
References: (0) Hutsemékers et al. 1998, (1) Berriman et al.
1990, (2) Stockman et al. 1984, (3) Moore & Stockman 1984,
(4) Impey & Tapia 1990, (7) Visvanathan & Wills 1998,
(8) Schmidt & Hines 1999, (9) Lamy & Hutsemékers 2000.

4.2 Discussion

Figure 4: The distribution of the polarization degree p (in %) for the quasars located in the region of alignment A1 (Table 2). Upper panel: radio-quiet (RQ+O) quasars. Lower panel: BAL quasars

The apparent alignment of quasar polarization vectors with the supergalactic plane is very appealing as the starting point of an explanation, namely since this could decrease by more than one order of magnitude the scale at which a mechanism must act coherently. By analogy with the alignment of stellar polarization vectors with the plane of our Galaxy (Mathewson & Ford 1970; Axon & Ellis 1976), some dichroism could be achieved due to extinction by dust grains aligned in a magnetic field. Another possibility could be the conversion of photons into pseudo-scalars also within a magnetic field (Harari & Sikivie 1992; Gnedin & Krasnikov 1992; Gnedin 1994). In both cases the hypothetical magnetic field should be coherent on a $\sim$50 Mpc scale, which is only slightly larger than the large-scale magnetic field possibly detected by Vallée (1990) in the direction of the Virgo cluster. But this interpretation suffers some drawbacks: namely, it cannot explain why the polarization vectors of quasars at lower redshifts are not accordingly aligned (cf. Fig. 5 in Paper I).

Figure 5: The distribution of the polarization degree for the radio-quiet (RQ+O) quasars in region A1. The upper panel is an enlargement of that of Fig. 4 and refers to the observed polarization degree p (in %). The so-called polarization bias (due to the fact that p is always a positive quantity) affects the first bin (see also Berriman et al. 1991). The lower panel shows the distribution of the polarization degree after subtraction of a small but systematic polarization ( $p_{\rm s} =0.25\%$, $\theta _{\rm s} = 170\hbox {$^\circ $ }$)

It is therefore difficult to escape the conclusion that if a mechanism is able to produce the alignment of polarization vectors of high-redshift quasars by modifying the polarization state of photons during their travel towards us, this must happen at redshifts (for region A1), or all along the line of sight assuming a cumulative or oscillatory effect (see also discussion in Paper I). Interestingly, an oscillation of quasar polarization with cosmological distance has been predicted as the consequence of the conversion of photons into pseudo-scalars within a large-scale magnetic field permeating the intergalactic medium (Harari & Sikivie 1992; Gnedin & Krasnikov 1992; Gnedin 1994). However, this interpretation, like other mechanisms which affect light as it propagates towards us, cannot easily explain why correlations observed between quasar polarization and intrinsic properties are not washed away. In this view we may ask ourselves how small could be the systematic polarization which, added to randomly oriented polarization vectors, can be at the origin of an orientation effect which involves nearly all polarized quasars of region A1. To simulate this, we have vectorially subtracted a systematic polarization $p_{\rm s}$ oriented at $\theta _{\rm s} = 170\hbox {$^\circ $ }$ (the dominant direction, cf. Fig. 1) from all polarized and unpolarized quasars of region A1 (Table 2). Then we have re-selected a sample of significantly polarized quasars with the conditions $p \geq 0.6 \%$ and $\sigma _{\theta } \leq 14\hbox {$^\circ $ }$. For $p_{\rm s} =0.25\%$, 29 quasars fulfil the conditions and only 15 objects out of this new sample have their polarization position angles still in the range $146\hbox{$^\circ$ }$- $46\hbox{$^\circ$ }$. This small systematic polarization seems therefore sufficient to produce the orientation effect, and small enough to preserve the difference between BAL and non-BAL quasars. However, as seen in Fig. 5, even such a small systematic polarization significantly modifies the polarization distribution of non-BAL radio-quiet quasars which appears depleted at polarization degrees , and then quite different from the distribution found by Berriman et al. (1991) and Hutsemékers et al. (1998). If we further note that values of $p_{\rm s}$ higher than 0.25% are obviously needed to explain the alignment of the 16 objects with $p \geq 1.0\%$, we may conclude that it is quite difficult to invoke a systematic additional polarization along the line of sight without modifying the quasar intrinsic polarization properties. This simple test also provides additional evidence that a systematic instrumental effect is unlikely. Although more subtle and speculative effects modifying the polarization of light along the line of sight can probably be imagined, we may admit on the other hand that the quasars themselves i.e. their structural axes are coherently oriented on Gpc scales. For radio-loud quasars, it is well known that the optical polarization is often parallel to the structural axis of the radio core (Rusk 1990; Impey et al. 1991). Unfortunately, only one polarized quasar in region A1 (B1127-145) is spatially resolved. It is however very interesting to note that this object has a core structure parallel to its polarization vector (Impey & Tapia 1990). In this view it is worth noting that possible coherence of morphological structures from the central engines of active galactic nuclei to superclusters has been suggested by West (1991, 1994), although this observation is apparently not confirmed at the supercluster scale (Jaaniste et al. 1998). From the theoretical point of view, some studies (Reinhardt 1971; Wasserman 1978) have pointed out the possible effects of magnetic fields on galaxy formation and orientation. Extrapolating, a correlation between quasar structural axes could be settled at the epoch of formation and related to very large-scale primordial magnetic fields possibly formed during inflation (Battaner & Florido 2000). Large-scale vorticity is another possibility, at least qualitatively. In this view, the apparent alignment found with the supergalactic plane is puzzling. However, coincidence cannot be ruled out, especially if we note that the polarization vectors of quasars belonging to the other regions of alignments (regions A2 and A3 in Paper I) do not show the same alignment with the plane of the Local Supercluster.