Astron. Astrophys. 325, L13-L16 (1997)

3. Spectral analysis and results

Spectral fits have been performed with the XSPEC 9.0 package, using the response matrices released on Jan 1997. To cure the mismatch between MECS and PDS absolute normalizations in the current matrices, PDS data have been divided by a factor 0.7, constant over energy (Cusumano et al., in preparation). Note that even allowing for a 10% remaining uncertainty in the cross-calibration, the continuum best fit parameters would change only slightly, and the basic picture would by no means be altered.

In the following, all quoted errors correspond to 90% confidence level for one interesting parameter ( =2.7).

As we are interested here only in the high energy part of the spectrum (i.e. that believed to be due to the reflection of the obscured nucleus), we restricted our analysis to energies greater than 4 keV to avoid contamination from other spectral components. The overall spectrum between 4 and 100 keV is shown in Fig.1. The spectrum confirms that NGC 1068 is substantially Compton-thick, otherwise it would have been detected at a much higher level in hard X-rays, similarly to NGC 4945 (Iwasawa et al. 1993; Done et al. 1996). A prominent, broad iron line is clearly seen. According to ASCA-SIS results (Ueno et al. 1994; Iwasawa et al. 1997), we fitted this broad feature with a blend of three narrow lines, with energies fixed at 6.4 keV (corresponding to neutral iron), 6.7 keV (He-like iron) and 6.97 keV (H-like iron), respectively. The results are presented in Table 1, and have been obtained adopting a power law for the continuum (fitted to MECS data only). No significant differences in the parameters of the lines have been found with a more complex description of the continuum (see below). If the ionized lines energies are instead fixed at 6.61 and 6.86 keV, as suggested by Iwasawa et al. (1997), the results are somewhat different, the H-like line being now more intense than the He-like one. Leaving these energies free, the best fit values are 6.64 and 7.01 respectively (but consistent within the errors with the nominal atomic values). From a statistical point of view, the first and third fits seem to be preferred, but all three results are acceptable at the 90% confidence level. The power law index of the underlying continuum is insensitive to the details of the line modeling. What is important to remark is that the lines' equivalent widths are of the expected order if these lines were produced by reflection of an invisible primary radiation (Matt et al. 1996).

Fig. 1. The observed BeppoSAX spectrum of NGC 1068 (folded with the instruments response) above 4 keV. Best fit parameters (two-reflectors model) are given in Table 2.

Table 1. Iron lines parameters. The lines have been fitted with -functions, the continuum with a simple power law over the 4 to 10.5 keV range (i.e. MECS data only). Line energies refer to the source rest frame (z =0.0038). Line fluxes are in 10^-5 ph cm^-2 s^-1.

If the MECS+PDS continuum is fitted with a simple power law, a good fit is obtained (see Table 2). (The iron lines have been modeled for simplicity with a single, broad gaussian feature). However, fitting the PDS data alone gives a photon index of 1.83 , inconsistent at the 90% level with the index of the total spectrum. This is due to the fact that the index is driven basically by the MECS, owing to its better statistics, while the PDS data, even if lying, on average, on the extrapolation of the lower energy spectrum, have a different spectral shape. This high energy steepening, together with the flatness of the 4-10 keV continuum and the presence of a strong 6.4 keV iron line suggests that part of the continuum is due to reflection of the nuclear radiation by circumnuclear neutral matter, possibly the inner surface of the torus. The presence of intense He- and H-like iron lines, and the fact that the 4-10 keV continuum, even if flatter than usual in Seyfert galaxies, is not as flat as a pure cold reflection spectrum would be, indicates contribution from an ionized reflector too, to be identified with the same medium responsible for the scattering and polarization of the optical broad lines. We then fitted the MECS+PDS continuum with a cold reflection component (XSPEC 's model PLREFL) plus reflection from free electrons, i.e. a power law with Compton downscattering in the assumption of a parallel incident beam and temperature of the electrons negligible with respect to the photon energy; in formulae (e.g. Matt 1996 and Poutanen et al. 1996):

where is the scattering angle, the solid angle subtended by the illuminated matter, and

Table 2. MECS+PDS joint fits with either a simple power law or the two-reflectors model (see text for details). Fluxes are in 10^-12 erg cm^-2 s^-1 and refer to the continuum only. C and W stand for the cold and warm reflector respectively.

The power law index of the illuminating radiation has been assumed to be the same for the two reflection components (so assuming no angular dependence of the nuclear emission spectral shape). To agree with the current wisdom on Seyfert 1 X-ray spectra, a reflection component from the accretion disc should actually have been included in modeling the nuclear radiation; however, it appears to be an unnecessary sophistication here. The fit results are shown in Table 2, and the best fit model in Fig.2. The power law index of the primary emission (1.74 ) is consistent with typical Seyfert 1 values (Nandra & Pounds 1994). The scattering angle (which, with our assumptions, is the same as the system inclination angle) is unfortunately not constrained, affecting the spectrum only at the highest energies, where the statistics is poor. The warm reflection component is the most important in the 4-10 keV range, while the cold reflection component dominates above  10 keV (see Fig.2).

Fig. 2. The two-reflectors best fit model. Warm and cold reflection continua and the iron line are also plotted separately. See text for details and Table 2 for best fit parameters.

If the nuclear X-ray luminosity is of the order of 10⁴⁴ erg s^-1, as indicated by several and independent pieces of evidence (Iwasawa et al. 1997 and references therein), then the 20-100 keV observed flux is about 2 orders of magnitude lower than the nuclear one. Adopting the Ghisellini et al. (1994) geometry, both a very high equatorial column density of the torus () (to avoid a significant transmitted intensity) and a high inclination angle (to reduce the observable fraction of the inner surface of the torus) are implied (see Fig.6 of Ghisellini et al. 1994). A high inclination disagrees with estimates based on both polarization properties (Miller et al. 1991), assuming that the X-ray and optical/near IR scatterers are one and the same, and infrared mapping of the torus (Young et al. 1996); such a disagreement would suggest that either the geometry is more complex than commonly assumed or the luminosity is smaller than estimated. Interestingly, however, both the large column

density and the high inclination are consistent with recent water maser measurements (Gallimore et al. 1996; Greenhill et al. 1996).

© European Southern Observatory (ESO) 1997

Online publication: April 28, 1998

helpdesk.link@springer.de