Astron. Astrophys. 362, 69-74 (2000)

2. BeppoSAX observation

2.1. Data reduction

The Italian-Dutch satellite BeppoSAX (Boella et al. 1997a) carries four co-aligned Narrow-Field Instruments (hereafter NFI), two of which are gas scintillation proportional counters with imaging capabilities: the Low Energy Concentrator Spectrometer (LECS, 0.1-10 keV, Parmar et al. 1997) and the Medium Energy Concentrator Spectrometer (MECS, 1.5-10 keV, Boella et al. 1997b). The remaining two instruments are the High Pressure Gas Scintillation Proportional Counter (HPGSPC, 4-120 keV, Manzo et al. 1997) and the Phoswich Detector System (PDS, 13-200 keV, Frontera et al. 1997). The quasar was observed by BeppoSAX on 1998 August 19-20. Data presented in this paper were obtained only with the imaging spectrometers, as PDS 456 has not been detected in the other two instruments. Standard reduction techniques and screening criteria were applied in order to produce useful linearized and equalized event files. A total of 62.7 ks and 29.7 ks have been accumulated for MECS and LECS instruments, respectively. Spectra have been extracted from circular regions of radius 4´ and 6´ around the source centroid for the MECS and LECS, respectively. Background spectra have been extracted from both blank-sky event files and from source-free regions in the target field-of-view. No apparent difference between the two spectra have been revealed. The background contributes to about 10% and 25% to the MECS and LECS count rates, the total net counts being 6.35 10^-2 in the 1.5-10 keV band and 2.12 10^-2 in the 0.4-4 keV energy range, respectively.

The XSPEC (version 10, Arnaud 1996) program has been extensively used to perform spectral analysis. In this paper errors are quoted at 90% confidence level for one interesting parameter ( = 2.71, Avni 1976), and energies are reported in the source rest frame. In all our models, for both neutral and ionized absorbers, we assume solar abundances as tabulated in Anders & Grevesse (1989).

2.2. BeppoSAX spectral results

PDS 456 lies fairly close to the Galactic plane (b 11^o). The Galactic H I column density derived by radio measurements is about 2-2.4 10²¹ cm^-2 (Dickey & Lockman 1990; Stark et al. 1992). We have checked whether molecular hydrogen, traced by CO emission, is present towards the PDS 456 direction, with negative result (Lebrun & Huang 1984; Dame et al. 1987). Therefore a value of 3 10²¹ cm^-2 (which is similar to the value found from the near-infrared observation of PDS 456 by Simpson et al. 1999) for the Galactic column density has been employed in all the spectral fits.

The combined BeppoSAX MECS and LECS spectra were first fitted with a single power law plus Galactic absorption. The complexity of the spectrum is evident from the quality of the fit (/dof = 253/105) and from the data-to-model ratio (Fig. 1). This shows systematic deviations over the entire LECS+MECS band. In particular, an excess of counts below 1 keV, as well as a strong deficit above 7-8 keV are clearly present, suggesting the presence of additional components at those energies. Furthermore, the monotonic rising of the model/counts ratio between and 5 keV suggests that additional absorption, exceeding the Galactic value, is required.

Fig. 1. The data/model residuals for a single power law fit to the LECS and MECS data.

To cure the strong residuals at energies above 7 keV (see Fig. 1) we added an absorption edge to the power law model (model A in Table 1). The improvement in the fit is highly significant (/ dof = 110/2). The best-fit energy of the edge is consistent at 90% with that of the photoelectric K-edge of FeXXIV -FeXXVI . If this identification is correct, then the measured optical depth (see Table 1) implies that a high column density ( cm^-2) of highly ionized matter is reprocessing the primary X-ray continuum of PDS 456 along our line of sight.

Table 1. Results of BeppoSAX LECS (0.4-4 keV) and MECS (1.5-10 keV) spectral fits.
Notes:
a) In units of erg cm s^-1
b) U = / is the ionization parameter, defined as the ratio of the ionizing photons density at the surface of the cloud ( = Q/(4R²c)) and the electron density of the gas.

Though the fit is largely improved by the addition of the absorption edge, the is still unacceptably high (143/103) and the residuals continue to show deviations at E 1 keV. Moreover, the spectral index is very flat ( = 0.94). An acceptable fit and a more reasonable value for the power law slope ( 1.3) is obtained adding to model A two additional spectral components (model B ): (a) an intrinsic neutral absorber obscuring the hard nuclear power law continuum, and (b) a soft power law, absorbed by the Galactic column density only. The best-fitting soft power law is very steep (though only poorly constrained: Table 1), indicating the presence of a luminous source of soft photons ( 5.2 10⁴⁴ erg s^-1) in the nuclear environment of PDS 456. The column density of neutral gas absorbing the hard X-ray power law largely exceedes the Galactic value along the line of sight to PDS 456, suggesting a "type-2-like" orientation for this high luminosity radio-quiet quasar.

The phenomenological model B provides a good fit to the observed 0.4-10 keV spectrum (Fig. 2) and good constraints to the various spectral parameters (Table 1, Fig. 3 and Fig. 4). To test the uniqueness of this model and to investigate whether the observed soft excess of photons, compared to the absorbed nuclear power law, is indeed due to genuine emission and not instead to an artifact due to additional complexity in the geometrical or physical status of the matter covering our line of sight to PDS 456, a few different spectral models were tested. More specifically, the soft X-ray power law was eliminated and the full-covering neutral absorber was either replaced by a partial covering absorber, allowing a fraction of the nuclear radiation to escape unabsorbed, or by a mildly ionized absorber transparent at energies lower than  keV. None of these models was able to provide good quality fits to our data. We also replaced the soft power law with thermal emission from a collisionally ionized plasma, and refitted the data. This did not modify the best-fit parameters of the remaining spectral components, nor improved further the quality of the fit, compared to model B. Moreover, a significant contribution to the soft X-rays from starburst emission is unlikely. Indeed, given a 60-100 µm luminosity of 6-10 10⁴⁵ erg cm^-2 s^{- 1} (depending on the assumed IR spectral slope) and relation [2] in David et al. (1992), only about 5% of the soft X-ray emission may be due to star-forming emission. We then conclude that PDS 456 shows genuine soft X-ray emission below  keV. Such a component is likely to be related to the strong UV bump present in the multiwavelength energy distribution (Reeves et al. 2000).

Fig. 2. The BeppoSAX spectrum (MECS+LECS, model B ) and residuals

Fig. 3. 68, 90 and 99% edge energy - optical depth confidence contours (model B )

Fig. 4. Intrinsic cold - confidence contours for the model B in Table 1. Absorption in excess to the Galactic value is clearly present

The energy and optical depth of the absorption edge in model B strongly suggest the presence of highly ionized gas absorbing and/or reflecting the primary radiation. To verify this hypothesys we then replaced the photoelectric edge in model B, with either an ionized reflector (parameterized by the model PEXRIV in XSPEC , Magdziarz & Zdziarski 1995: model C ) or an ionized absorber (using the CLOUDY - Ferland 1996 - based model described in Nicastro et al. 2000a, and adopting the Mathews & Ferland 1987 parameterization for the AGN continuum: model D ), and refitted the data. In both cases we left the soft and the hard, reflected/absorbed, power law free to vary independently. Both parameterizations gave acceptable . The best-fit values of and were consistent with each other between the two parameterizations. The hard reflected/absorbed continuum ( 1.4-1.6) is still much flatter than the soft X-ray power law. Finally, the best-fit ionization parameters are consistent with each others, and their values correspond to a very high ionization state of the matter reflecting/absorbing the nuclear radiation, with iron mainly distributed among species XXI and XXVII. Both models provide then an acceptable description of our data. However, according to Matt et al. (1993), such a high ionized reflector should produce strong (EW 300-500 eV) fluorescence iron K emission lines at 6.7-6.97 keV. These lines are not observed in our data, the 90% upper limits on their equivalent width being of 120, 115 and 80 eV for neutral, He-like and H-like ionic species respectively. The corresponding 3 limits lie in the range 150-180 eV. We then conclude that, while statistically acceptable, the model including an ionized reflector is not fully consistent with the BeppoSAX data of PDS 456. K emission from highly ionized iron is instead not expected to be strong in the case of an ionized absorber. The predicted strength of these lines depends on the fraction of solid angle covered by the absorber (as seen by the central source), and on the dynamics of the gas (see Sect. 4), but does not exceed few tens of eV for the strongest line, fully compatible with our data. A spectrum transmitted by a high column ( 4.5 10²⁴ cm^-2) of highly ionized matter would then self-consistently account for both the observed deep absorption FeXXIV-XXVI K edge and the absence of K line emission from the same iron species (model D, Table 1).

The best-fit spectrum gives a 2-10 keV flux of about 5.7 10^{- 12} erg s^-1 cm^-2, which corresponds to an intrinsic 2-10 keV luminosity of about 1.2 10⁴⁵ erg s^-1. At this flux level and given the relatively short exposure time, the source has not been detected at high energies by the PDS instrument (which spends about half of the time to monitor the background). In fact, a 3 detection would have been possible only for a power law harder that 1.3 (Guainazzi & Matteuzzi 1997). The present upper limit provides therefore only a loose contraint on the high energy spectrum.

© European Southern Observatory (ESO) 2000

Online publication: October 30, 19100
helpdesk.link@springer.de