Astron. Astrophys. 343, 33-40 (1999)

3. Spectral analysis

The 3C 390.3 data were initially divided by the 3C 273 data collected during the BeppoSAX Science Verification Phase (SVP). This method allows a quick qualitative look at the spectrum without the need for complex models. The SVP spectrum of 3C 273 was chosen because it is well represented by a simple power law in the 1.5-200 keV. Only a very small deviation occurs at keV (observer frame) due to the presence of a weak iron line (EW eV). The LECS data were not considered because of the spectral complexity of 3C 273 below 1 keV (Grandi et al. 1997b). The spectral ratio immediately evidenced the presence of a line at keV (as expected from neutral iron, taking into account the redshift of the source), and a clear excess above 10 keV.

Because of the complexity of the 3C 390.3 spectrum, we decided to perform the spectral analysis in two steps, studying the continuum and the emission line separately. Special care was also taken in determining the absorbing column density along the line of sight.

3.1. Continuum: primary emission and reflection component

We fitted the BeppoSAX data in XSPEC between 0.12 and 100 keV excluding the points in the energy interval 5.5-7.0 keV, where the emission line contributes significantly, aiming for the best possible determination of the continuum. The relative flux normalization between the LECS and MECS was left free to vary, whereas a miscalibration of 15 between the MECS and the PDS was assumed as indicated by NFI intercalibration analysis based on observations of 3C 273 (Grandi et al. in preparation).

We first tested a single power law model (PL), with low energy absorption. The best fit parameters are reported in Table 1, with 90 confidence limits. This model is unacceptable as it gives a large value of , corresponding to a chance probability of about 0.004. A broken power law model (BKP) gives a better fit (see Table 1), but still not good enough in terms of the statistic (chance probability of 0.054).

Table 1. Fits to BeppoSAX continuum^a of 3C 390.3
Notes:
^a Data in the 5.5-7 keV energy band are not included in the fits.
^b Normalization factor at 1 keV ( 10^-3 photons cm^-2 sec^-1 keV^-1).
^c High-energy power law for model BKP.
^d Power-law break in model BKP, e-folding cut-off energy in model CPL+REF.

We therefore tested a more complex spectral model including, in addition to a power law, a reflection component (PEXRAV in XSPEC; see Magdziarz & Zdziarski 1995). The PEXRAV model assumes that a cold reflector is irradiated by a primary isotropic X-ray source. The input X-ray spectrum can be modeled by a simple power law or by a power-law with a high energy exponential cut-off. The general functional form for this model is:

where A is a global normalization factor at 1 keV (photons cm^-2 sec^-1 keV^-1), the photon index, and the cutoff e-folding energy. Setting =0, the exponential term is excluded from the fitting formula, and a simple power law is taken as primary continuum. The reflection component is computed assuming a plane parallel semi-infinite medium irradiated by a point-like or optically thin X-ray source, and is a function of the angle i between the line of sight and the normal to the slab. We fixed i to 26^o as deduced from the UV and optical emission line measures (Eracleous & Halpern 1994 and Wamsteker et al. 1997), and from the radio jet superluminal motion (Eracleous et al. 1996). Finally, R is an additional scaling factor introduced to take into account roughly the solid angle subtended by the cold reflecting material to the X-ray source located above it, when different from 2. R=1 corresponds to the scale free geometry assumed in the computation of . In principle, a change of the subtended solid angle would imply a change in the angular distribution of impinging photons, which in turn would change not just the normalization of the reflected spectrum, but also its shape . However, such an effect is small in comparison to the quality of available data, and the geometry is usually assumed to be simply described by the parameter R = .

The MECS and PDS data are better reproduced by the model including reflection than by the single (PL) or broken power law (BKP) models. We initially assumed a simple power law input spectrum (PL+REF). The reduced decreases from 1.17 to the well acceptable value of 1.09 (chance probability of 0.17). The reflection component is unambiguously detected at 99 confidence level, as evident from the contour plots shown in Fig. 1. We then included a high energy cutoff (CPL+REF), allowing to vary. As expected, when the input X-ray spectrum is a power law with cut-off (E_c 400 keV), the deduced value of R (R=1.2) is somewhat larger, but the increase is not statistically significant. In fact, in order to reproduce the observed Compton hump, the reflection component has to be a bit boosted to compensate for the reduction of primary photons at very high energies, which preferentially emerge, once down-scattered in the cold layers, around 20 keV. In any case, the improvement in is not significant (), and therefore the inclusion of a cut-off is not statistically required by the data. We note, however, that non-simultaneous ASCA , Ginga , and OSSE observations (Wozniak et al. 1988) suggest the presence of a break in the high energy spectrum of 3C 390.3. Also, the spectral index variations measured by ASCA (Leighly et al. 1997) indicate a pivot point at keV.

Fig. 1. Confidence contours for the photon index () and the amount of reflection (R) when the continuum is fitted with a power law reflected by cold matter (model PL+REF in Table 1).

In any case, the amount of reflection required by the BeppoSAX data is larger than that previously measured by Ginga (R, (Nandra & Pounds 1994, Wozniak et al. 1998), possibly suggesting variations of the relative strength of the reflection hump with respect to the primary continuum.

3.2. Column density

The low energy absorption requires a column density larger than the Galactic value, cm^-2 (Stark et al. 1992), for every model of the continuum we tested. Comparing our results with ROSAT data (Wamsteker et al. 1997), we noted that our determination of the column density is slightly larger. We then collected all the available information from the literature in order to investigate possible temporal variations of . We re-analyzed two old EXOSAT observations performed in 1985 Feb. 3 and 1996 Mar. 17-18, respectively, as they provide a quasi-simultaneous coverage of the 0.01-10 keV band, and have a good signal-to-noise in the ME spectra (quality flag ). The ME (1.8-10.0 keV) and LE (0.01-2 keV) data were simultaneously fitted with a simple power law with cold photoelectric absorption, which turned out to provide an adequate fit to the data in both the observations. The best fit values are reported in Table 2. No soft X-ray excess is required by the EXOSAT data. The LE data points lie on the extrapolation of the power-law continuum, which is basically determined by the ME statistics. This result is in disagreement with Ghosh and Soundararajaperumal (1991), who claimed that a soft excess is present in this source.

Table 2. Best-fit parameters when a simple absorber power-law model is applied to the Feb. 3, 1985, and Mar 17-18, 1986, EXOSAT observations of 3C 390.3

When the historical column density is plotted as a function of the date of observation, several interesting variations are evident, clearly indicating that temporal modifications of the cold material along the line of sight occurred, with an estimated time scale of a few years (see Fig. 2). In particular, the value of  seems to be steadily increasing since 1992, but was always found much lower than the first Einstein detection (Kruper et al. 1990).

Fig. 2. (upper panel ) Historical X-ray light curve of 3C 390.3. (lower panel ) The column density is plotted as a function of time. The  trend was decreasing before 1987 and increasing after 1991. The dashed line corresponds to the Galactic column density

We did not find any correlation between  and the intensity at 1 keV (see Fig. 2 and Table 3). This fact suggests that changes in the geometry of the absorber, rather than variations of its ionization state (as expected in the case of a warm absorber responding to an ionizing continuum), are probably responsible for the observed long term variability of . Similar long-term absorption variability has been also observed in NGC4151 (Yaqoob et al. 1993).

Table 3. Absorbing column density history

3.3. Emission line

An excess in emission with respect to the power law plus reflection continuum is clearly present in the 5.5-7.0 keV energy interval (see Fig. 3). The statistical significance of this excess, estimated by the quadratic sum of the deviations in each bin, is at the level of . The excess is most likely produced by an Fe fluorescent line, for which we could estimate, using a gaussian profile, a flux of 3.63 (-1.45,+0.78) ph cm^-2 sec^-1 (at the 90 confidence level for one interesting parameter), corresponding to an equivalent width of 136 (-36,+40) eV. The line centroid lies at 6.39 (-0.09,+0.10) keV (rest frame), compatible with K line emission from neutral iron located at the source redshift. The line is not resolved: the intrinsic width can be estimated from the fits as 73 (-73,+207) eV, indicative of a narrow feature. The line flux and the intrinsic width confidence contours are shown in Fig. 4. These line best fit values are consistent with those of previous ASCA observations (Eracleous et al. 1996, Leighly et al. 1997): in fact the line intensity derived from either Ginga , ASCA or BeppoSAX is consistent with being constant within the rather large errors (cf. Wozniak et al. 1988).

Fig. 3. MECS residuals () to the data, when the continuum of emission is fitted with a power law reflected by cold material. The error bar corresponds to one sigma

Fig. 4. Confidence contours for the line flux and the intrinsic width when the feature is fitted with a gaussian profile.

In Fig. 5 the total 0.1-100 keV photon spectrum is shown with the residuals to a power-law-plus-reflection model when the gaussian line is added to the fit.

Fig. 5. Photon spectrum (upper panel ) and residuals (lower panel ) when a power law plus reflection and a gaussian line are fitted to 0.12-100 keV BeppoSAX data

© European Southern Observatory (ESO) 1999

Online publication: March 1, 1999
helpdesk.link@springer.de