Astron. Astrophys. 326, 885-896 (1997)

3. Spectral power law fits to the ROSAT data

As a first approach to quantify the soft X-ray excess emission in the ROSAT energy window we have fitted power law spectra to the ROSAT count rates for all sample members. The distribution of the resulting spectral indices for a column density fixed at the galactic value and a free absorbing column density are given in Fig. 2 (upper panels). For comparison, the canonical value, , and the sample mean, = 0.86, of the hard X-ray power law are indicated as vertical lines. Except for a somewhat larger width of the distribution no large difference between the fixed, galactic and free power law spectral indices is observed, thus indicating that, on average, the spectra are not much affected by intrinsic low energy absorption in excess of the galactic value in the ROSAT spectral range. The spectral indices, together with the corresponding column densities , are summarized in Table 2. The mean ROSAT spectral power law indices for free and fixed are 1.40 and 1.55, respectively, signifying a marked steepening of the spectrum as compared to the spectral slopes observed at higher energies. Looking at each object individually, a steepening of the spectral slope between the hard and soft X-ray range is found in almost all sample members. The mean change in spectral index is found to be 0.62 (free ) and 0.78 (fixed ), respectively. In Fig. 2 (lower panels) the distribution of this change in spectral index is shown.

Fig. 1. Histograms of ROSAT best-fit spectral power law indices. The canonical hard X-ray spectral power law index, , as well as the mean hard X-ray spectral power law index, , of the sample members where measurements were available are marked by vertical lines. Lower panels: Histograms of change in spectral slope between ROSAT and hard X-ray band.

Fig. 2. ROSAT spectral power law index plotted over redshift (left: linear scale; right: logarithmic scale). Each object is shown twice. Diamonds: Fits with free . Triangles: Fits with fixed . Open symbols refer to spectral fits where the statistical errors were larger than 0.3.

Table 2. in units of . The first value ( - fix) is taken from Elvis et al. 1989 if marked with a , else from Stark et al. 1992 . is the ROSAT spectral power law index, the corresponding error. ist the error of the fittet value.

When going to objects at higher redshifts, the ROSAT sensitivity window is shifted to higher source frame energies, thus in effect turning ROSAT into a higher energy X-ray instrument. We find that at these higher source frame X-ray energies ROSAT does indeed measure harder X-ray spectral indices, similar to those of low redshift objects measured by higher energy X-ray instruments (see Fig. 2), in agreement with results by Schartel et al. (1996) and others. Note, however, that Puchnarewicz et al. (1996 ) do not find a dependence of the mean X-ray power law index on redshift in their sample of AGN from the RIXOS survey which may in part be due to the selection of their objects in the hard 0.4 - 2.0 keV ROSAT energy band.

Contrary to previous results, here, the dependence of the X-ray spectral index on redshift is visible in individual, relatively bright objects as opposed to averaged properties derived from stacked spectra or mean spectral indices of many weak objects. Based on radio flux measurements and upper limits all objects in the sample are known to be radio-quiet ( ; see Sect. 2). We can therefore exclude any contamination of our sample from high-redshift, radio-loud objects which are known to be more luminous than radio-quite quasars and at the same time display harder X-ray spectra. Such a contamination has previously been suggested as a possible cause of the observed dependence of the ROSAT power law indices on redshift. For the 7 objects in the redshift range a Spearman rank correlation coefficient of is found, corresponding to a likelihood of 0.02 (0.05) for randomness of the correlation. Results for fixed and free (in brackets) spectral power law fits are given. A quantitativ analysis of this behaviour, using the accretion disk model described in the following section is presented in a forthcoming paper (Brunner et al., 1997). Note that the low redshift objects () in our sample do not follow this trend, suggesting a turnover of the relation in the redshift range with decreasing spectral indices on either side of this range. One possible interpretation for the turnover of the relation is that, as one goes to lower and lower redshifts, lower luminosity objects are detected where the accretion disk component is increasingly absorbed and the spectrum is increasingly dominated by the hard power law component. We would finally like to point out that the observed hardening of the ROSAT spectral indices with redshift rules out the possiblity that the steep AGN spectra observed by ROSAT may, as has been suggested, in part be due to errors in the cross-calibration of the ROSAT PSPC detector and previous higher energy X-ray instruments.

For the steep X-ray spectrum () subsample (17 objects) a strong correlation of the ROSAT spectral index (fixed ) and the optical to X-ray broad-band spectral index is found (see Fig. 3). The correlation index is 0.78 (Spearman rank correlation coefficient), corresponding to a probability of of randomness. This suggests that in objects with strong soft X-ray excess emission, i.e., objects with steep ROSAT spectra, the dominant contributions to the X-ray and UV/optical emission are due to the same physical emission component (i.e., the big blue bump emission). While the correlation can also be traced to objects with lower ROSAT spectral indices, a number of these objects show broad-band spectral indices which are considerably steeper than predicted by the correlation, suggesting that in these objects, the onset of the big blue bump emission is at or below the lower cutoff of the ROSAT sensitivity window. Note that a fraction of these objects (marked by filled triangles in Fig. 3) are at higher redshifts () where any blue bump emission component is expected to be shifted out of the ROSAT sensitivity window. Objects marked as open diamonds in Fig. 3 which also do not seem to follow the correlation are seen through absorbing column densities cm^-2 and may thus be affected by uncertainties in the ROSAT power law index (fixed ) and possibly the de-reddening of the optical fluxes. Similar correlations of and have also been reported by Puchnarewicz et al. 1996.

Fig. 3. Broad band power law index plotted against ROSAT spectral power law index (fixed ). Diamonds: redshift, ; triangles: ; open symbols: absorbed spectra (). errors of the slope of the linear regression line calculated for the 17 steep X-ray spectrum objects () are marked as solid lines.

While the change in spectral slope between the ROSAT and harder X-ray energy bands as well as the correlation are useful indicators for the soft X-ray excess and big blue bump emission, a quantitative analysis is best performed in the framework of a physical emission model, the most widely advocated candidate being emission from the hot inner region of an accretion disk around a super-massive central object.

© European Southern Observatory (ESO) 1997

Online publication: April 8, 1998
helpdesk.link@springer.de