3.1. The PDS spectrum ( keV)
The 60.0 ks exposure time on NGC 1052 allowed a PDS detection, with a total 13-200 keV count rate of s-1. If we take into account the typical systematic uncertainties associated with the PDS background subtraction algorithm (Guainazzi & Matteuzzi 1997), the detection is at a level higher then 5 in the 13-90 keV band (see Fig. 1). No known hard X-ray bright source is present in the (1.3o)2 PDS field of view. The number of expected sources with flux equal or higher than NGC 1052 is according to the Cagnoni et al. (1998) logN-logS relation. A simple power-law fit of the PDS spectrum yields a very flat index: .
3.2. The broadband BeppoSAX spectrum (0.1-100 keV)
The source did not show any significant X-ray variability during the BeppoSAX observation. The reduced , when a constant line is fit to the s light curve are and in the 0.1-2 keV (LECS) and 2-10 keV (MECS) energy bands, respectively. We will therefore focus in this Section on the time-averaged spectra only.
A photoelectrically-absorbed single component model provides an inadequate fit of the broadband (0.1-100 keV) BeppoSAX spectrum (e.g.: degrees of freedom, dof, if a power-law model is employed). On the other hand, a very good fit (-1.03) is obtained with a two-component model, constituted by an absorbed ( a few cm-2) power-law plus a "soft excess" below 2 keV. The limited statistics prevents us from unambiguously characterizing the latter component. In the following we will discuss, as illustrative examples, models where this soft excess is described: either with a power-law ("P" model hereinafter), whose index is held fixed to that of the high-energy absorbed power-law (thus modeling reflection of the nuclear continuum, scattered by an electron plasma - "warm mirror" - along our line of sight; Antonucci & Miller 1985); or with thermal emission from a collisionally ionized, optically thin plasma (mekal model in XSPEC ; "M" model hereinafter). Model "P" also describes a geometry, in which the absorber only covers a fraction of the line of sight. The best-fit parameters are reported in the upper panel of Table 1. As already suggested by the analysis of the PDS spectrum alone, the absorbed power-law component is rather flat (). If the power-law in model "P" is substituted by a thermal bremsstrahlung ( dof), its temperature is keV.
Table 1. Best-fit parameters and results when the models "M" and "P" (details in text) are applied to the NGC 1052 broadband spectrum of BeppoSAX (upper panel ), ASCA-ROSAT (after G99; central panel ), and ASCA-BeppoSAX-ROSAT (lower panel ). is the scattering fraction (defined as the 2-10 keV flux ratio between the transmitted and the scattered power-law components). and EW are the centroid energy and the equivalent width of the emission line, respectively.
The addition of a narrow Gaussian emission line is required at the 98.9% confidence level, according to the F-test, in the "P" model ( for a decrease of the degrees of freedom by two), whereas only at the 90.9% level in the "M" model (). The Gaussian line centroid energy is consistent, within the statistical uncertainties, with fluorescent emission from neutral iron. However, the EW of the iron line system is too large to be produced in transmission by the same cold matter, which is responsible for the attenuation of the X-ray continuum (which would imply 130 eV for a spherical distribution of matter; Leahy & Creighton 1993). No iron emission line is expected from an ADAF. The slight difference in EW between models "P" and "M" (in the latter the iron line profile is partly accounted by the emission of the thermal plasma) may suggest a multi-component structure of the iron line, which is unresolved by the MECS. Ionized iron lines could be also produced by the "warm mirror" (Netzer & Turner 1997). We have therefore repeated the fit in the "P" scenario, assuming that the iron emission actually consists of two components: one neutral ( keV) and one He-like ( keV). The fit is of comparable quality ( dof), with: EW(6.4 keV) eV; EW(6.7 keV) eV. The soft excess continuum flux at 6 keV is about 1/3 of that of the transmitted component. Therefore, the EW of the ionized iron line against its proper continuum would be of the correct order of magnitude if produced in a "warm mirror" (Matt et al. 1996). The neutral component EW is now consistent with being produced in transmission by the same matter covering the active nucleus, if its covering factor is large.
A hard X-ray continuum could be in principle due to Compton reprocessing of the nuclear continuum, by either the accretion disc (George & Fabian 1991; Matt et al. 1992) or the molecular torus encompassing the active nucleus (Ghisellini et al. 1994; Krolik et al. 1994). This scenario does not, however, match our data. A fit, where the absorbed high-energy component is a bare face-on Compton-reflection (model pexriv in XSPEC ; Magdziarz & Zdziarski 1995) is statistically unacceptable ( dof). The addition of a Compton-reflection component to the "P" model (where only the relative normalization between the reflected and the direct component, R, and the intrinsic power-law cut-off energy are left free parameters in the fit; an inclination angle of and solar abundances are assumed) does not significantly improve the fit ( dof). The 90% upper limit for two interesting parameter on R is 0.6 (for ). These results allow us to rule out one class of models, which adequately fit the ASCA-ROSAT spectra, hence favoring the G99 transmission scenario.
The observed flux in the 0.5-2 keV (2-10 keV) energy band is 0.4 (4.0) erg cm-2 s-1. This corresponds to a luminosity of 0.4 (4.2) erg s-1.
3.3. Comparison with ROSAT and ASCA data
In the central panel of Table 1, the best-fit parameters are reported, when the "M" and "P" models are applied to the ASCA and ROSAT spectra of NGC1052 (see the Table 1 in G99). NGC 1052 was comparatively bright during the August 1996 ASCA (2-10 keV flux erg cm-2 s-1) and the January 2000 BeppoSAX observations. The only spectral parameter showing a significant difference is the absorbing column density, which was about a factor of 2 higher in the later BeppoSAX observation. The spectral indices measured by BeppoSAX tend also to be slightly softer, but still consistent with the ASCA-ROSAT measurements within the statistical uncertainties.
We have performed a simultaneous fit of the ROSAT, ASCA and BeppoSAX spectra, to check whether the improved statistics allows us to distinguish between the "M" and "P" models. In these fits only the column density absorbing the primary nuclear continuum has been allowed to vary independently in the BeppoSAX and ASCA-ROSAT models (ROSAT spectra are basically insensitive to column densities of the order of cm-2). Normalization constants have been included as free parameters in the models, to account for the different fluxes measured in the three observations. The results are reported in the lower panel of Table 1. The two models yield comparably good fits ( dof; dof). Better data quality is needed to resolve this issue. The spectral index is indeed much better constrained than by BeppoSAX data alone, and still very flat (see Fig. 2).
© European Southern Observatory (ESO) 2000
Online publication: January 29, 2001