Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 344, 857-867 (1999)

Previous Section Next Section Title Page Table of Contents

Appendix A: comparison with results and "ambiguities" from the literature for the data below 10 keV

To allow a direct comparison of the BeppoSAX results with the results reported in the literature (Sect. 1), we restrict here the analysis to the data below 10 keV. In agreement with previous findings (G98 and ref. therein), the spectrum of Mkn 3 below 10 keV is very complex, with prominent soft excess emission below 3 keV and strong iron line emission above a flat underlying 3-10 keV continuum. This is clearly illustrated by the residuals shown in Fig. 8 obtained by simultaneously fitting the LECS and MECS data with a single power-law model, with an extremely flat best-fit photon index [FORMULA] [FORMULA].

[FIGURE] Fig. 8. LECS and MECS spectra and residuals with a continuum model consisting in a single power-law model ([FORMULA] [FORMULA] -0.5) absorbed by the Galactic absorption [FORMULA]=8.46 [FORMULA] 1020cm-2.

The three basic descriptions for the continuum shown here (Table A1) comprise a) a double power-law model; b) a soft power-law plus a hard, absorbed, power-law model; and c) model b) plus an unabsorbed pure reflection continuum model. A Gaussian emission line is added in all fits. The resulting best-fit parameters are given in Table A1.


Table A1. "Literature" models fitted to the data below 10 keV (LECS + MECS data).
a) See text for a description of the models and parameters used; b) Ratio of the power-law normalizations calculated at 1 keV; c) In units of 1022 cm-2. d) Line energy in the source rest-frame, in units of keV; e) Line width, in units of keV; f) Observed line intensity in units of 10-5 photons cm-2 s-1.
Note: Intervals are at 90% confidence for 2 interesting parameters.

We find that model c gives the best description of the data at energies lower than 10 keV, with a [FORMULA] compared with model b for one additional free parameter, in agreement with Turner et al. (1997b). However, it is difficult to clearly prefer one model to the other. Inclusion in model b of extra emission/absorption features in the soft and/or hard component (rather plausible if the scattering gas and/or absorbing material is ionized) could account for most of the remaining residuals of the fit (i.e, the 3-5 keV bump, iron edge structure and excess emission above [FORMULA] 7 keV, as shown in Fig. 9b). This ambiguity is similar to the one resulted from the ASCA data that led to different parameterizations and interpretation of the same data by different authors (I94, Turner et al. 1997b, G98). However, we have shown in Sect. 3.2 how such an ambiguity can be solved thanks primarily to the BeppoSAX detection of the high-energy ([FORMULA] keV) emission. This is also illustrated in Fig. 10 which shows that an extrapolation of model b) to higher energies falls short of the PDS data, thus requiring a more complex model with larger absorption.

[FIGURE] Fig. 9a-c. Unfolded spectra with best-fit models as obtained from the fits of the LECS+MECS data. Pannels a , b and c correspond to lines 1, 2 and 3 of Table A1, respectively. For clarity, only the MECS data have been plotted here.

[FIGURE] Fig. 10. Unfolded spectrum with the best-fit model b) as obtained from the fit of the LECS+MECS data and with the PDS data added subsequently.

Because of the spectral complexity, we also find that any residual feature is a strong function of the adopted model for the underlying continuum. For example, from the ratios shown in Figs. 9a, 9b and 9c there is some evidence of a broad absorption structure between [FORMULA] 7-8 keV that, at first glance, could be ascribed to absorption by ionized material (G98). However, a comparison of the different figures shows that the feature is significant within model a and model b but is only marginal in model c. Indeed, in our baseline model, such an edge is entirely accounted for ([FORMULA] 0.5 if an extra absorption edge is added at energies between 7-8 keV) by the combination of the deep Fe K edge of the large absorption column density plus the broad Fe K edge produced by the reflection component.

Moreover, as shown in Table A1, the same data give differences of about a factor of two in the line width (and thus line intensity) depending on the adopted continuum model. Even for equal width (compare lines in pannels b and c of Fig. 9), a different continuum modeling can result in a [FORMULA] 20% difference in the observed line intensity.

In conclusion, the above results highlight the need of broad-band spectroscopy for such complex sources in order to reach confidence that the continuum emission and spectral features are properly modeled.

Appendix B: alternative models to our baseline model

More complex models have also been fitted to the BeppoSAX broad-band spectrum, in particular: a neutral dual absorber model and a dual absorber model with a mixture of neutral and ionized matter. The basic idea is to explain the flat [FORMULA] 3-10 keV underlying continuum without using an unabsorbed reflection component (as proposed in our baseline model, Sect. 3.2) but with the addition of a second absorbed power-law component. The former model has been extensively used in the literature to explain complex absorption and/or flat 2-10 keV spectra of several Seyfert galaxies (EXO 055620-3820.2, Turner et al. 1996; NGC 4151, Weaver et al. 1994; NGC 5252, Cappi et al. 1996; NGC 2110, Hayashi et al. 1996, Malaguti et al., 1999; NGC 7172, Guainazzi et al. 1997; IRAS 04575-7537, Vignali et al. 1998). The latter model has been first proposed by G98 for Mkn 3 based primarily on the Ginga detection of an ionized Fe K edge at [FORMULA] 8 keV. The results obtained with these models are given in Table B1. The ionized absorber model used here is the absori model in XSPEC (Done et al. 1992) with a temperature fixed to 106 K, as in G98 (none of the following conclusions changed, though, for a temperature ranging between 10[FORMULA] K). We find that both models give acceptable fits of the broad band spectrum and are both statistically indistinguishable given our data alone. None of them, however, gives a statistically better fit than the baseline model (Sect. 3.2, Table 1) considering the increase by 1 and 2 in the number of free parameters for the former and latter model, respectively. Given the poor statistics of the data at low energies, we did not attempt to fit a more physical (and more complex) model for the ionized absorber that would take into account emission/reflection from the absorbing gas or the possible presence of dust.


Table B1. Alternative models fitted to the broad band spectrum (LECS + MECS + PDS data).
a) See text for descriptions;
b) In units of 1022 cm-2;
c) Covering factor, equivalent to Ah(1)/(Ah(1)+Ah(2));
d) Line energy in the source rest-frame, in units of keV;
e) Line width, in units of keV;
f) Observed line intensity in units of 10-5 photons cm-2 s-1.

Moreover, the main problem with both these models is that, in order to interpret physically different absorbing columns along our line of sight, one is forced to assume either that i) the source of X-rays is not a point source as seen from the two absorbers and the two different columns cover different areas of the source or that ii) the lowest of the two column densities covers the whole source while the larger one only partially covers it. The former geometry appears rather unlikely given the small dimensions generally inferred to the X-ray emitting regions in AGNs from variability arguments (e.g. Mushotzky et al. 1993). The latter geometry could be more physical and fits well into the framework of Unified Models since one could identify the highest column density ([FORMULA] in Table B1) with the BLR clouds, partially covering the source, and the lowest column ([FORMULA]) with absorbing matter, eventually ionized, associated with the torus (or its rim) or with some matter in the outer zone of the galaxy (e.g. Hayashi et al. 1996, Vignali et al. 1998). However, our best-fit results give [FORMULA] [FORMULA] 1.2 [FORMULA] 1024 cm-2 and a covering fraction of 90% (Table B1, column 7) which are at least an order of magnitude larger than commonly believed for the BLR (respectively 10[FORMULA] cm-2 and 5-30%, Netzer & Laor 1993, Kwan & Krolik 1981). Such complex models may, possibly, also find alternative interpretations in terms of different geometries involving complex scattering by matter at non-uniform density and/or with non-uniform coverage.

It should also be noted that the main reason G98 proposed an ionized absorption model was the presence of an ionized Fe K absorption edge in the Ginga data at [FORMULA] 7-8 keV. However our MECS 7-10 keV data, with unprecedent statistics, show that such edge could be entirely accounted for by the combination of the reflection component and the larger column density (required by the PDS data) as given in our baseline model (Sect. 3.2). In conclusion, we find no need from the present data to invoke more complex models and/or to invoke ionized absorption in our data.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1999

Online publication: March 29, 1999