Astron. Astrophys. 363, 863-868 (2000)

4. Data analysis and results. Spectral analysis

The spectral analysis was performed on the LECS and MECS instruments only, because of the possible confusion problems in the PDS outlined above. The PDS data have been used only to check a posteriori that a LECS+MECS best fit spectrum, when extrapolated to higher energies, would not exceed the observed data.

All fits described below have been performed with the XSPEC software package. Quoted errors refer to 90% confident level for two interesting parameters (i.e. =4.61).

Because of the significant spectral variability, it is not wise to use the spectrum averaged over the whole observation. On the other hand, it is important to collect as many photons as possible to search for spectral details. As a trade-off we have divided the observation in two parts, and extracted spectra from: a) the beginning of the observation till 8 sec (see Fig. 1), because within this interval variations in the hardness ratio, especially in the MECS, are not dramatic; b) from this time till the end of the observation.

As it will be seen in the following (Table 1 and Table 2), no models will give a fully acceptable fit, the null hypothesis probability being 0.12 at most. This is likely due to the spectral variability discussed in the previous section, as our time selection makes this problem alleviated but not completely cured.

4.1. First half of the observation

Let us start discussing the first half of the observation. A simple power law absorbed by the Galactic column (i.e. 1.7710²⁰ cm^-2; Elvis et al. 1989) is a very poor fit to the data (see Fig. 3 and Table 1, model 1), with an unacceptable reduced of 1.70. From the figure, it is clear that most of the contribution to the comes from the softest part of the spectrum, i.e. below 1 keV, while no evidence for typical Seyfert 1 components like the iron line or the reflection continuum is present. A deficit of counts above 7 keV is also apparent. Leaving the column density free to vary (model 2), the quality of the fit significantly improves, but the fit is still unacceptable (=1.50).

Fig. 3. The spectrum along with the best fit model (upper panel) and data/model ratio (lower panel) for the first half of the observation, when fitted with a simple absorbed power law.

4.1.1. The warm absorber

Inspection of the residuals suggests the presence of warm absorption features. We therefore added the ABSORI model (model 3 in Table 1), obtaining a significant improvement in the quality of the fit. The warm absorber material has been assumed to be ionized by the observed power law. The best fit value of the ionization parameter, , is about 240 erg cm s^-1; with this value, the most prominent edge is the O VIII one. The column density of the warm absorber is 810²¹ cm^-2.

4.1.2. The Oxygen line

A further inspection of the residuals reveals an excess around 0.6 keV, which can be well fitted (model 4) by a narrow gaussian line at 0.59(0.05) keV (source rest frame), with an equivalent width of 68() eV. The statistical significance of this line is 97.3% (F-test). The line energy is consistent with He-like oxygen (0.57 keV); a fit with the line fixed at this energy, plus a line at 0.65 keV (H-like oxygen) does not provide a statistical improvement; the best fit EWs are 60 and 30 eV, respectively. The value of is now lower. The best fit ionization structures of the absorber and the emitter are somewhat different (the most relevant ion being O VII in the emitter, and O VIII in the absorber), but still consistent each other within the (fairly large) errors. Finally, a relativistic line (DISKLINE model in XSPEC ) fits the line equally well as a narrow gaussian line. Fixing the inner radius at 6 , i.e. the last stable orbit for a Schwarzschild black hole, we obtain an outer radius of thousands of and an inclination angle consistent with zero (i.e. face-on disc).

4.1.3. The iron edge. Ionized reflection of ionized absorption?

While the above fit is satisfactory from a statistical point of view, a deficit of counts above 7 keV is still apparent. Adding an absorption edge to model 4 the fit actually improves, the best fitting parameters being an edge energy of 7.55 keV, and =0.26. This edge may be related to reflection from circumnuclear matter, and therefore we added a Compton reflection component. We allowed the matter to be ionized (PEXRIV model in XSPEC ), both because the edge is at energies larger than 7.1 keV, the value of neutral iron, and because the lack of any observed iron K line (see next paragraph) is only possible, in presence of a Compton reflection component, if the iron is mildly ionized and therefore resonant destruction possible (e.g. Matt et al. 1993, 1996). The results are reported in Table 1 as model 5. Interestingly, the value of the ionization parameter of the reflecting matter, even if poorly determined, is consistent with what is expected in order to have the iron line significantly destroyed. It is worth noticing that such an ionized disc could also account for the observed oxygen emission line, even if the equivalent width is larger than expected by a factor of a few, pointing to a possible oxygen overabundance. Moreover, the best fit parameters of the warm absorber material are now such that the most important oxygen ion is the Helium-like. It is therefore possible that both the accretion disc and the warm absorber contribute to the line flux.

Alternatively, the iron edge may be due to absorbing material. We therefore added a second ionized absorber (model 6), instead of the reflector. The fit is as good as the one with the reflector. This second absorber results to be more thick and ionized than the other one (but the parameters are loosely constrained). It may be interesting to note that a similar double-absorber solution has been found in the (broad line) Seyfert 1 NGC 3516 (Costantini et al. 2000).

4.1.4. The iron line

Leighly et al. (1996) detected a narrow 6.4 keV iron line with an equivalent width of about 100 eV in the ASCA spectrum of Mrk 766, but only when the source was in a high state. We searched for an iron line in the BeppoSAX data, but could find only upper limits. The upper limits to a narrow iron line are 110, 43 and 36 eV if the line energy is fixed to 6.4 keV (neutral iron), 6.7 keV (H-like iron) or 6.97 keV (H-like iron), respectively. Therefore, our result is marginally consistent with the line found by Leighly et al. (1996). Of course, the upper limits would increase if the line is broad. As explained in the previous section, the lack of an observable line is still consistent with the reflection scenario, because the best fit ionization state is such to have the line destroyed by resonant re-absorption.

4.1.5. PDS data

As shown in Fig. 4, the extrapolation of model 5 to the PDS data is rather good, which may be an indication (but unfortunately not a proof) that the contribution to the PDS count rate from 2A 1219+305 is low. Also the extrapolation of model 6 does not exceed the observed PDS data.

Fig. 4. The spectrum and best fit model (upper panel) and residuals (lower panel) for model 5 in Table 1. The fit has been performed on the LECS and MECS data only, and then the best fit model has been extrapolated to the PDS data.

4.1.6. The soft excess

We then searched for a soft excess (modeled as a black body), a component which has been already observed in this source (see Sect. 1), and whose presence may be expected by analogy with other Narrow Line Seyfert 1 Galaxies (e.g. Boller et al. 1996). The fit with the soft excess instead of the warm absorber (other components as in model 5) gives a worse (1.23/119; best fit temperature of about 30 eV), while the addition of the black body to model 5 does not provide a significant improvement (=1.13/117; best fit temperature of 200 eV). We therefore conclude that such a component is not required by the data, and confirm that the warm absorber is real, and not an artifact of not having included the soft excess. However, it should be noted that while the fits presented in Table 1 have been obtained with a column density of the cold absorber in excess of the Galactic value, the column density turns out to be close to the Galactic one when the black body component is included. Therefore, the presence of the soft excess remains rather ambiguous. It is worth noting that the BeppoSAX power law index is significantly steeper than the one observed by ASCA, while the average 2-10 keV flux is about the same in the two observations, making any soft excess in the BeppoSAX data difficult to observe.

4.2. Second half of the observation

We then analysed the spectra extracted from the second half of the observation. For the sake of conciseness, only models 5 and 6 are reported in Table 2. The oxygen line is now not required by the data, and we fixed its energy to 0.6 keV (see Table 1; this energy would correspond to a blend of He- and H-like atoms) to get an upper limit. The fit with two absorbers is now preferable to that with one absorber and one reflector, from a statistical point of view. Again, a blackbody component instead of the warm absorber gives a much worse fit. This time, however, adding the blackbody component a significantly better fit is obtained, but at the expense of a very steep power law component (=3.4) and an unplausibly large reflection component (r=21). As there is significant spectral variations during the second half of the observation, it is possible that this improvement occurs because a further component may help fitting the fictitious spectral complexity arising from time-averaging over different spectral states. Not surprisingly, the is significantly higher than in the first half of the observation.

Let us now compare the results obtained in the two halves of the observation. Two main differences are evident: first, in the second half of the observation the oxygen line is not required by the data, and only an upper limit on its EW can be obtained. This limit, however, is consistent with the value for the first half. Second, the WA (both of them in model 6) appears to be thicker and more ionized in the second half than in the first half (even if, within the errors, the values of the ionization parameters are consistent with each other). The two best fit absorption models are shown in Fig. 5, where it can be seen that the differences between the two models are large in the 0.9-2 keV interval. Intriguingly, these are the energies where the NEV has a maximum (Fig. 2). Unfortunately, the limited statistics prevent from a more detailed analysis.

Fig. 5. Best fit warm absorber model for the two halves of observations, obtained with model 5 in the tables. Solid line: first half. Dash-dotted line: second half.

© European Southern Observatory (ESO) 2000

Online publication: December 5, 2000
helpdesk.link@springer.de