## 4. Soft X-ray spectraIn order to establish soft X-ray spectra in the energy range (0.1 -
2.4)keV for all data sets, the vignetting corrected photon events
accumulated from the source extraction area were sorted into 256 pulse
height channels to yield the source plus background count rate
spectrum. The same procedure is performed for the background counts
gathered from an off-source sky field and renormalized to the source
extraction area. After subtracting the background spectrum and
employing the dead time correction, a binned up pulse height source
count rate spectrum, corrected for instrumental effects, is obtained.
It is used to compare it with
## 4.1. The mean spectrumTo get an idea of the shape of the quasar's soft X-ray spectrum we decided to merge those data sets of the pointed observations, for which the hardness ratio in Fig. 4 does not indicate a spectral change, i.e. we merged all data sets of the June observations in 1992 and 1993. The extracted source and background counts of the merged orbits were accumulated and processed as described above. The resulting mean count rate spectrum of Ton S180 contained 63.000 cts registered in 15130s. Firstly, we fit a simple power law modified by low energy absorption to the source spectrum. Thereby we restrict the fit to data between channel 10 and channel 240 as beyond these boundaries the detector response matrix is not well defined. The channel spectrum was then compressed to 22 spectral bins with an equal significance of their count contents. In the case of the mean spectrum the significance per spectral bin was 47 standard deviations. The simple power law fit with the absorption as a free parameter proved to be unacceptable. The best fit parameters were: photon index , normalization at 1 keV , and cold matter absorption . The minimum reduced of the fit was 38.5/19 d.o.f. = 2.03. The errors given correspond to one standard deviation (68.3% confidence level) for three interesting parameters. The residuals between the actual count rate spectrum and the model spectrum are depicted in Fig. 5a showing systematic trends troughout the spectral range. The -value as resulting from the fit was inconsistently lower than the Galactic value towards Ton S180, (Stark et al., 1992), even at a 90% confidence level. This might indicate the existence of an additional spectral component at very low photon energies. Fixing to the Galactic value worsens the fit even more.
An appreciable improvement is achieved by fitting a two-component model spectrum consisting of a power law, a black body, and cold matter absorption. The fit of this model to the merged data yielded the following parameters: , , , , and . The goodness of fit is given by a reduced -value of 14.3/17 d.o.f. = 0.84. The improved residuals are reproduced in Fig. 5b. The fit reveals that the black body component dominates the quasar's count rate spectrum below 0.3 keV. The remaining range between 0.1 keV, the low energy limit of the spectrum due to absorption essentially in our Galaxy, and 0.3 keV is too narrow to determine the exact spectral shape of the additional low energy component. Therefore, the actual choice of its spectral shape is rather arbitrary. The fit is effectively not changed by substituting the black body by thermal bremsstrahlung or by a second steep power law. The -value resulting from the fit is at a 90% confidence level inconsistently larger than the Galactic value towards Ton S180. This might be due to a small amount of intrinsic absorption in the host galaxy of the QSO or in its nuclear region. It is further remarkable that the fit temperature of the black body component is unusually low compared to the spectra of Seyfert 1 galaxies showing soft X-ray excesses (Walter & Fink, 1993; Walter et al., 1994).
For the assessment of the statistical interdependence and significance of the spectral parameters resulting from the fit we studied the significance diagrams of the mutually dependent parameter pairs and . The corresponding significance contour diagrams are given in Fig. 6. As expected, the parameters , which determine the shape of the continuum and which are dominant in different energy bands, are less mutually dependent. They show narrow error ranges and their values are well defined. This is not the case for , the significance contours of which are shown in Fig. 6a. The interplay of the exponential low energy cut-off due to absorption by interstellar matter and the strength of the steep energy component of the intrinsic quasar spectrum determines the rise of the observed count rate spectrum below 0.2 keV. Since the low energy component is dominant only below 0.3 keV, the interdependence of both parameters is very strong. Consequently, the -valley is extended and shallow causing particularly the strength of the steep component to be poorly defined. This effect is further increased for the fits of the orbit spectra discussed in the next section, which contain many fewer counts than the merged spectrum. Fig. 7 summarizes the statistical characteristics of the low energy spectral component and illustrates how poorly the normalization of the black body can be determined, while its temperature is well defined. ## 4.2. Orbit spectraAs the hardness ratio given in Fig. 4 indicates the possibility of a slight change of the quasar's spectrum between Dec 1992 and Jan 1993, we wished to investigate the spectral parameters as a function of time in more detail. All subsets of data, RASS and the individual orbits of the pointed observations, comprise enough counts to allow the construction of statistically significant X-ray spectra. Again, we firstly fit the orbit spectra with a simple absorbed power law. Due to the poorer statistics the representation of the orbit spectra by simple power laws cannot be rejected in all fits as in the case of the mean spectrum. More than half of the fits was revealed to be statistically acceptable. Considering one standard deviation as error, the mean index of the observations in June 1992, , is inconsistently larger than that determined for the observations in June 1993, . This inconsistency vanishes at a confidence level of only 90%. The mean photon index of the observations in Dec 1992 and Jan 1993, is consistent with that of the previous as well as of the following measurements. The derived values are for all three observational epochs less than the Galactic value: , , and , respectively. This strengthens also for the orbit spectra the conjecture that a realistic spectral model must be more complex than a simple power law. Making use of the experience with the mean spectrum of the merged data set we choose as model spectrum a power law and a black body. remains as a free fit parameter. The spectral parameters resulting from the fits of the two-component model to the orbit spectra are given in Table 2. In general, the adding of a very soft component to the power law continuum improves the fit of the orbit spectra in almost all cases with the one exception of the spectrum of orbit No.12, the residuals of which show unexplained systematic wiggles between 0.1 keV and 0.5 keV (see Fig. 10). From Table 2 it is obvious that the actual strength of the low energy component is not well constrained. However, the necessity of adding such a spectral component to the power law is certain: considering here only the June observations in 1992 and 1993, for which spectral variability is not obvious, and taking one standard deviation as uncertainty, the normalization of the black body is never consistent with omitting such an additional component. Only in the case of the less significant RASS observation is no additional component needed. The different behaviour of the quasar spectrum in late 1992 and in early 1993 will be discussed just below. The temperature (or steepness) of the low energy component is, however, well defined. Its mean value is and the corresponding average temperatures for 1992 and for 1993 are consistent with each other. The photon index has a mean value of for the observational epochs in June 1992 and 1993: ; . These photon indices as well as the black body temperature of the orbit spectra are consistent with those values determined from the fit of the merged mean spectrum. Neither the temperature nor the index varies significantly for observations as can be seen from Fig. 8. For the pointed observations in June 1992 an averaged -value of was determined, whereas in June 1993 the average value amounts to . At a significance level of 68.3% these mean values are consistent with each other, but only the 1992 mean value is consistent with the derived for the mean spectrum. Anyway, both averaged -values are significantly larger than the Galactic value. ## 4.2.1. Spectral variabilityIn this section we consider only those observations which took place on Dec 18th, 1992, and on Jan 10th, 1993, respectively. During the two adjacent orbits in December the spectrum of the QSO was as steep as during previous epochs: . 23 days later the spectrum has become unusually flat, , - flatter than ever before and after. Together with a slight increase of the total count rate the flattening of the spectrum is accompanied by an increase of the normalization of the power law component at 1 keV. Fitting the spectra for normalizations at several photon energies in ROSAT's energy range shows that the change of the power law component between Dec 18 1992 and Jan 10 1993 is a rotation around a crossover point situated between 0.5 keV and 0.7 keV. Thereby the flux due to the power law component increased slightly by 16 per cent. The simultaneous behaviour of the low energy component is remarkable: while the characteristic parameters of the black body, temperature and strength, are positive definite even at a 90% confidence level during the December observations, the fit of the spectrum measured in January revealed that the same parameters are consistent with zero already at a confidence level of less than 68.3%. This situation is shown in Fig. 9. Using the best fit parameters the flux of the black body component decreased by about a factor of four between Dec 18, 1992, and Jan 10, 1993. Therefore we conclude that there are epochs at which the quasar shows spectral variability without a noticeable change of the total soft X-ray emission. Simultaneously with a flattening of the power law component, the component, usually dominant at very soft photon energies, vanishes. To date, such a behaviour of a soft X-ray excess has only been observed in the X-ray spectra of Seyfert 1 galaxies (e.g. NGC4051: Pounds et al. (1994), Komossa et al. (1996)). ## 4.2.2. Warm absorber?Although the entirety of the residuals of the fit of a two-component model to the mean spectrum is in accordance with the expected statistical distribution, it is noticeable that there are low bins just around 0.7 keV and beyond 1 keV (Fig. 5b). This could lead to the suspicion that these negative deviations from the model are caused by the effects of a warm absorber on the line of sight as has been found for several Seyfert galaxies, but not yet for QSOs (Nandra & Pounds, 1992; Ptak et al., 1994; Mihara et al., 1995). We checked the residuals of the fits of all orbit spectra (Fig. 10). In the individual sequences we could not detect any systematic trend of low bins around 0.7 keV, where possible absorption edges of highly ionized oxygen should be located and should be visible for all physical conditions of a warm absorber with cosmic abundances. Therefore, we conclude that the ROSAT spectra do not indicate the existence of absorbing ionized material close to the nucleus of Ton S180. © European Southern Observatory (ESO) 1997 |