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

4. The X-ray properties

From the Einstein IPC data it is known that radio-loud and radio-quiet quasars show different power law slopes ( ; Wilkes & Elvis 1987, Canizares & White 1989, Brunner et al. 1989, Shastri et al. 1993). Further, various correlations of the photon index with other source properties have been claimed for subclasses of objects, for example with core dominance (Wilkes 1994, and references therein), with redshift (Schartel et al. 1992), or with radio spectral index (Brunner 1992). As the current sample is considerably larger than all samples used before we are able to draw statistically more reliable conclusions on the claimed correlations - however, in the somewhat softer (0.1 - 2.4 keV)  ROSAT   PSPC energy band.

4.1. The average photon index

The results of a maximum-likelihood analysis for the distribution of power law slopes for flat spectrum and for steep spectrum quasars is given in Fig. 5 (for details of the analysis see Maccacaro et al. 1988, Worrall & Wilkes 1990). To minimize evolutionary effects, we restricted the redshift range to which resulted in a sample of 102 flat and 104 steep spectrum sources for which a spectral index is available for fits with free and for fits assuming Galactic . A discussion of the redshift dependence of the power law slopes will be given in the next section.

Fig. 5. Best-fit mean spectral index and Gaussian standard deviation for power law fits to flat (FS) and steep radio spectrum (SS) sources assuming Galactic absorption (dashed curves) and free-fit absorption (full curves). Contours correspond to 90% confidence levels.

In Fig. 5 the contours correspond to 90% confidence levels. The fits are done assuming Galactic absorption only (dashed lines) or the absorbing column density is left free in the fit (full lines).

Evidently, flat spectrum quasars show a flatter X-ray power law spectrum than steep radio spectrum objects. Remarkable is the fact that in flat spectrum sources, leaving the amount of absorption as a free parameter, the dispersion of the spectral indices is compatible with zero. I.e., all flat spectrum quasars have a very similar soft X-ray power law slope. Forcing the absorption to the Galactic value increases the dispersion considerably demonstrating that a noticeable number of objects show absorption deviating from the Galactic value or have spectra not following a simple power law. Interestingly, in the case of the steep spectrum sources the two distributions are very similar. This implies that either steep spectrum objects form an intrinsically inhomogeneous group where the physical conditions determining the X-ray spectra are differing from object to object, or the X-ray emitting region is intrinsically absorbed in many of the objects. This might be caused by orientation dependent absorption in a molecular torus or by conditions similar to those found in CSS, where it has been proposed that the sources are being inhibited from growing to larger dimensions by unusual conditions of the interstellar medium (Fanti & Fanti 1994).

We have analyzed separately the small group of objects classified as `Compact Steep Spectrum' (CSS, Fanti et al. 1990) and as `Gigahertz Peaked Spectrum' sources (GPS, O'Dea et al. 1991). Interestingly, the GPS sources have a flatter photon index = 1.86 (for free) and = 1.50 for Galactic , and a dispersion which is consistent with zero for the fits with free . For the CSS sources we find = 1.88 ( free) and =1.99 (Galactic ), both with large dispersion .

It is not clear whether this dichotomy of the X-ray spectral slopes is a 'fixed' property of the flat spectrum and steep spectrum classes or whether there is a continuous mutual dependence of the two spectral indices. Taking the power law indices obtained with the assumption of fixed Galactic absorption and using only objects with errors in the indices a linear regression analysis (Draper & Smith 1966) gives for a fit of the X-ray photon index versus radio spectral index with a non-parametric Spearman rank correlation coefficient = -0.23 for 264 d.o.f., at a probability level of . is the probability that the observed correlation occurs by chance for uncorrelated data sets (Press et al. 1986). All errors given are 68% confidence limits (1  ). It should be noted that we obtain here and in the following identical results (inside the mutual errors) for fits with free absorption. The regression curve is given as dashed line on the data in Fig. 6. Throughout the paper the following symbols will be used for the different object classes: steep spectrum objects, flat spectrum objects, CSS objects, GPS objects.

Fig. 6. X-ray photon index (assuming fixed Galactic absorption) as a function of the radio spectral index . steep spectrum objects, flat spectrum objects, CSS objects, GPS objects.

We tested this general trend against regression fits of the two classes of objects separately. The regression line for the subsample of 114 steep spectrum objects is given by , that for the 150 flat spectrum objects . While for flat spectrum sources the Spearman rank correlation analysis gave a probability for no correlation of , this hypothesis can be ruled out with 98.7% confidence for the steep spectrum objects. For flat spectrum sources the assumption that the data set can be described by the 'general' slope of can be ruled out with almost 95 % confidence, for steep spectrum objects with 79% only.

This result strongly suggests that flat spectrum and steep spectrum sources are from an X-ray point of view intrinsically different types of objects, but so far the physically relevant parameter responsible for this difference has not been found. If these differences can be related to orientation effects the changes of the emission characteristics must occur rather abruptly as there seem to be no smooth transitions of the X-ray properties between the different sub-classes or the radio spectral index is not an appropriate measure of the quasar's intrinsic properties.

4.2. Core dominance

The X-ray slopes of core and lobe-dominated radio-loud quasars were found to be different (Boroson 1989, Wilkes & Elvis 1987, Canizares & White 1989, Brunner et al. 1992, Shastri et al. 1993) which was interpreted as a slope continuously flattening with core dominance R = . Here and are the core and extended flux densities at 5GHz, respectively, K-corrected assuming = 0 and . The core fluxes were obtained from the literature and from recent VLA observations of radio-loud ROSAT sources (Laurent-Muehleisen et al. 1996).

A plot of the photon index versus core dominance actually shows a correlation between these two quantities with a Spearman rank coefficient and a probability level .

The common explanation for the flattening of the X-ray spectrum with core dominance would be that the beaming angle influences the slope of the spectrum. However, Fig. 7 indicates that this result can also be interpreted as a correlation between radio slope and X-ray photon index : the flatter X-ray spectrum sources are predominantly found at higher core dominance, a quantity which can be connected directly to flatter radio spectra. Similar correlations were noticed previously, for example by Kembhavi et al. (1986) and at higher X-ray energies by Williams et al. (1992) using Ginga data. However, a partial Spearman rank correlation analysis did not reveal which of the underlying parameters is fundamentally correlated with .

Fig. 7. X-ray photon index as a function of the core dominance . Symbols as in Fig. 6.

4.3. Radio loudness

The use of the extended flux density for normalizing the boosted flux of the core introduces considerable scatter in the determination of the beaming angle as the emission from the radio lobes depends strongly on the interaction of the jet with the environment. Therefore, the ratio of radio core to optical continuum luminosity, i.e., the radio loudness, has been proposed as a more suitable core dominance parameter (Wills & Brotherton 1995). Indeed, the plot of the core dominance R versus radio loudness for our sample shows a relatively well defined correlation over a large parameter range - with some scatter. From Ginga observations Williams et al. (1992) find a possible correlation between the X-ray medium energy (2 - 10 keV) spectral index and radio loudness at the 90% level, however for a rather small sample of radio-loud quasars.

In Fig. 8 we show the photon index as function of radio loudness, again for flat spectrum and steep spectrum sources. There is certainly a trend of decreasing with increasing radio loudness. However, this is partly masked by the effect that steep spectrum objects are predominantly found at lower radio loudness, flat spectrum sources at higher radio loudness. The combined two classes of objects show a trend of the form , with with a Spearman rank coefficient of and a probability level .

Fig. 8. X-ray power law photon index as a function of the radio loudness . Symbols have the usual meaning.

Interestingly, comparing Fig. 7 and Fig. 8 we find that the trends of versus core dominance R or versus radio loudness are similar, except at high values of R and where seem to decrease with but to increase with R. Similar deviations from a linear relation at higher values of these parameters are seen as well in the plot R versus . This is not unexpected as the discussion (chapter 5) of the luminosity correlations clearly shows that the optical luminosity is correlated with the (beamed) X-ray luminosity indicating that the optical emission is, at least partly, beamed as well (Baker et al. 1994). Thus its use as an orientation indicator based on an assumed isotropy of the optical emission seems to be questionable.

4.4. Redshift dependence

Quasars are seen over a large range of cosmological distances and, correspondingly, the observed spectral energy band transforms into different intrinsic energy bands in the source frames. Secondly, quasar luminosity functions show evolution (cf. Ciliegi et al. 1995, Boyle et al. 1993) and, therefore, a cosmological evolution of the quasars spectral properties cannot be ruled out either.

Canizares & White (1989) find no evidence for a dependence of the power law slope on z for quasars observed with the IPC. In the softer  ROSAT   energy band Schartel et al. (1992) report a flattening of the power law spectra with redshift while Bechtold et al. (1994) claim similar slopes for high and low redshift radio-loud quasars, however, with substantial intrinsic absorption for objects at high z.

In Fig. 9 we show the photon indices (assuming Galactic absorption, again) as a function of redshift for flat spectrum (upper panel) and for steep spectrum quasars. Again, we have excluded the objects with large errors in the photon index () from the analysis and sources found in regions of exceptionally high Galactic absorption ( cm^-2) as it cannot be ruled out that the fitted values of the spectral slopes are affected by the correspondingly narrow remaining energy window. We further excluded 3 objects with very high photon statistics which show definite deviations from a simple power law. Finally, for 8 objects with clear indications of additional intrinsic absorption we used the free - fit values for the power law slopes.

Fig. 9. X-ray photon index as a function of redshift z for flat spectrum and GPS (top) and steep spectrum and CSS (bottom) quasars. For clarity only data points with are plotted and the symbols have the usual meaning. The fitted linear regression curves are given.

For flat spectrum quasars a single regression curve gives an acceptable fit (Spearman rank probability level ). However, the fit of this total slope for the subset of quasars at alone is unacceptable and the data are indicative for a redshift dependent broken line. Fitting two straight lines results in a z-dependent correlation of the form for redshifts and a redshift independent component with for . The regression analysis for the subsample of steep spectrum objects (excluding 12 CSS objects) gives , using objects with for the fit. For flat spectrum sources the hypothesis of `no correlation' can be ruled out with almost 100% confidence, for steep spectrum objects with 91%.

The break in the fitted line of the power law index around z 2 for the flat spectrum quasars is probably not related to evolutionary effects seen in the luminosity function of X-ray selected quasars (Boyle et al. 1993). The changes are more likely caused by the fact that with increasing redshift the soft X-ray excess `moves out' of the PSPC's energy window - modified by the different amounts of Galactic absorption towards the sources. The photon index thus approaches the average redshift independent value found in the higher energy band (E keV) by EXOSAT (Lawson et al. 1992), Ginga (Williams et al. 1992), and ASCA (Siebert et al. 1996, Cappi et al. 1996) for high redshift quasars.

Finally, we investigated the possible dependence of excess absorption in the fitted spectra with redshift. Absorption intrinsic to the quasars themselves could yield information about their evolution and their local environments; excess absorption along the line of sight places limits on a hot diffuse intergalactic medium and on physical conditions in damped Lyman- absorbers (Elvis 1994).

In Fig. 10 we plot, for all objects, the differences between the fitted absorption and the Galactic absorption towards the sources in units of cm^-2 as a function of redshift. A few objects which evidently show excess absorption cm^-2 and some objects with errors larger than cm^-2 are outside of the plot boundaries.

Fig. 10. The difference between the fitted absorption and the value of Galactic absorption to the source (in units of cm-2) as a function of redshift z.

For most of the objects the values are compatible with the Galactic values within their 1  errors. We do not see a statistically significant systematic trend with z and the results of a regression analysis (applying different - cuts to avoid 'outliers') are inconclusive. The differences found for individual sources can be related to intrinsic absorption in these objects or to the fact that a single power law is an inappropriate representation for the soft X-ray spectrum. Interestingly, there seems to be a higher fraction of quasars with intrinsic absorption at high redshifts an effect only found for radio-loud quasars (Elvis 1996). The GPS objects seem to have a tendency to show (on average) absorption in excess of the Galactic values, for CSS objects we find the opposite behavior. However, these two samples are too small for statistically reliable conclusions.

© European Southern Observatory (ESO) 1997

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