Astron. Astrophys. 319, 413-429 (1997)

7. Discussion

The large sample of radio-loud quasars available from  ROSAT  observations allowed a detailed study of the spectral X-ray properties of various subsamples, the correlations between X-ray and radio properties, as well as the broad band energy distribution of the objects. For the statistical analyses we used the restricted sample of radio selected quasars only. However, the corresponding results from the whole sample of all radio-loud quasars are only slightly different in their numerical values (all inside the mutual 1 errors) and show the same general trends. This means that the inclusion of originally optically or X-ray selected quasars does not lead to detection biases.

In Sect. 4.1 we showed that the power law indices of the soft X-ray spectra differ for the various radio classes of quasars, with steep spectrum sources having a steeper slope than the flat spectrum quasars. However, the correlation of X-ray spectral index with redshift indicates that at redshift zero both classes have similarly steep soft X-ray spectra. As the spectral slopes for steep spectrum sources flatten considerably less with redshift, their average power law index appears to be steeper for a large sample of objects. The correlations of X-ray spectral index with redshift strongly suggests that radio-loud quasars have two spectral components: a flat power law at higher energies, which is related to the radio core emission and an additional steeper component in the  ROSAT  energy band. At z = 0 the soft X-ray spectra are dominated by the steep component, and flat and steep spectrum sources have similar indices. With increasing z this steep component moves out of the PSPC's energy band and the X-ray spectrum is a combination of the two components. The generally flatter soft X-ray power law spectra can then be attributed to a more dominant flat X-ray component in flat spectrum radio sources - a scenario which also accounts for the fact that flat spectrum radio sources are X-ray louder than their steep spectrum counterparts. The flat component can be recognized in Fig. 6 for the flat spectrum quasars but, unfortunately, there are no data available for steep spectrum quasars at higher redshifts. If the flat power law component is related to beamed emission from the core and the steeper low energy component with an isotropic unbeamed component (Browne & Murphy 1987), the above results are in accordance with current unification schemes which attribute the different classes of objects to different viewing conditions (Barthel 1989).

Similar ideas have been followed by Jackson et al. (1993) to explain the different spectral slopes generally found for AGN in the  ROSAT  and in the wider Einstein IPC energy band. Whereas the steeper  ROSAT  spectra could be reproduced the simulated Einstein spectra were too steep. However, as Bühler et al. (1994) showed recently, these results strongly depend on the parameterization of the soft component in these models.

However, the above simple two component interpretation is not quite consistent with expectations of these schemes: in that framework one would expect that the X-ray emission from radio galaxies is predominantly the unbeamed, isotropic component with steep spectral slope. But the average X-ray slope found for radio galaxies (Papers I, II) is around .

Another important issue in that respect is the relation between the X-ray spectral properties and the radio properties of the objects. We found correlations between the power law photon index and the radio spectral slope , the radio loudness , and the core dominance R. These parameters themselves are, to a high degree, interrelated and indicators for orientation effects (for example, the radio spectral index can directly be related to the core dominance parameter). They further depend, at least via selection effects, on the redshift z as well and it remains unclear which physically dominating parameter is accounting for the correlations. Again, viewing conditions might be responsible - at least in the extreme cases of flat spectrum and steep spectrum objects. In most correlations the transition between these two object classes is rather abrupt.

A final question related to the X-ray spectral properties is that of a possible correlation of the amount of absorption towards a source with its distance (redshift). There seems to be no additional absorption increasing with redshift, at least up to . However, we cannot answer this question with any reliable statistical significance as the quality of the spectra is relatively poor for most of the distant sources and there is substantial scatter in the data. This scatter seems to be related to individual sources, i.e., some objects (especially at higher redshifts) show additional absorption associated with the quasar itself. Further, there is growing evidence that this extra absorption is not a 'fixed' property of a particular object but that the intrinsic absorption can vary strongly with time (Comastri et al. 1996, Schartel et al. 1996).

Correlations between the emission at different wavelengths are indicators for the underlying emission processes. A vast number of papers addressed this question - with differing results. The quoted reasons for these discrepancies range from the definitions of the samples studied to the mathematical methods applied in the analysis. We have shown how the latter influence the luminosity correlations in our sample and came to the conclusion that methods taking into account errors in both variables were the most appropriate ones.

Even with these techniques the slopes of the l _x - l _r - correlations were less than unity, that for the l _x - l _o - correlations could be consistent with a slope of unity, expected from the study of the luminosity functions. However, the correlations involving the radio luminosities clearly show that there are higher order effects in the data and that a single straight line very likely does not represent the true physical connection between the variables.

Our analysis in Sect. 5.4 indicates that the observed X-ray luminosity depends on the total intrinsic radio power of the quasar and, additionally, on the amount of beaming in the source. Assuming that the observed total X-ray luminosity is composed of two contributions, one related to the radio core luminosity, one to the extended emission, i.e., log(l _x) = log () +   log () + const, and using the fitted slopes for high R and low R objects from Sect. 5.4 the representation of the X-ray data with this model results in a largely improved value compared to the fits of l _x versus or versus . This strongly indicates that the two-component model is a better description of the physical situation in the objects.

It must further be noted, that the set of parameters for the correlations found between the various luminosities and their ratios is not self consistent. The reasons are either, that some (or all) of the correlations are not linear, that there is a strong redshift or selection dependency in the data, or that the considered subclasses of objects are exhibiting different luminosity dependencies.

Some correlations, like that of the X-ray luminosity versus core dominance or X-ray loudness versus core dominance can be traced back directly to the different emission properties of the flat and steep spectrum sources. We do not find correlations of broad band spectral properties , or with redshift z, however, these quantities seem to correlate with the luminosities.

One of the most discussed correlations is that of the X-ray loudness versus optical luminosity l _o  (Kriss & Canizares 1985, Pickering et al. 1994, Green et al. 1995, Avni et al. 1995). The slope of found in chapter 6 (similar values are obtained in the other papers as well) is inconsistent with a slope of unity in the log(l _x) - log(l _o) correlation and various physical explanations can be found for the apparent inconsistencies (Pickering et al. 1994, Della Ceca et al. 1994).

Taking the slope determined for the - l _o   dependence, we fail to find the expected - l _x   correlation in our data. Therefore, we believe that at least a major fraction of the claimed correlation is caused by selection effects. Assuming a constant with a certain dispersion as well as upper and lower luminosity limits for the X-ray data the sources should occupy a phase space region as depicted in Fig. 19.

Fig. 19. Phase space of the X-ray loudness versus optical luminosity of a sample of objects with limits on the X-ray luminosity as found for the current quasar sample. Shaded areas denote phase space regions devoid of sources (see Fig. 15).

With a Monte-Carlo simulation we `filled' the rhomboid with the same number of test particles as data points in Fig. 17 with various constant values for and dispersion .We then performed regression analyses on these data as above. With and a dispersion we obtained the same slope of found in chapter 7.

In addition to luminosity boundaries for l _x  we see optical luminosity boundaries in the data: there are hardly any data below log(l _o) = 29.6 and above log(l _o) = 31.8. These regions correspond to the shaded regions in Fig. 19. Taking into account these limits in the Monte-Carlo simulations we obtain the above fitted slope with a dispersion of and the same constant value for .

Thus we conclude that the claimed correlation between and l _o   is mainly caused by luminosity selection effects in conjunction with the intrinsic dispersion of the distribution (similar arguments hold for the - l _r  and the - l _r  correlation) and that a linear correlation between log(l _x) and log(l _o) might quite well hold. There are, however, indications for a lack of very X-ray luminous quasars at very high optical luminosities in the data and we can certainly not rule out a non-linear correlation or correlations of higher order.

© European Southern Observatory (ESO) 1997

Online publication: July 3, 1998
helpdesk@link.springer.de