The complexity of the soft X-ray emission is confirmed by the fact that, in a few sources, no model gives an acceptable fit. This is the case of NGC 3783 and Fairall 9 (even if for the latter the fit with the black body model is just above the acceptability threshold), while for another source (NGC 5548) the only acceptable model (the absorption edge), is just below the acceptability threshold. These three objects may thus represent the case when the spectral complexity of the soft X-ray excess shows clearly due to a combination of statistics and specific emission characteristics.
We have therefore carried out a more detailed analysis,
being guided in the choice of more complex models
by realistic scenarios.
A first case is that of a warm absorber covering the central
source where an accretion disk produces a soft excess
either by thermal emission or by reflection. A second case is that
of an emission line added to the two models in which it
can be expected, i.e. the absorption edge and the reflection
model. In fact, warm absorbers can manifest themselves not only through
absorption features, but also with emission lines (e.g. Netzer 1993),
provided that the covering factor is a significant fraction of 4. Also
in reflection models, strong line emission is predicted (Ross & Fabian 1993).
Due to the limited energy resolution of the PSPC, the number of free parameters in each fit cannot exceed 4-5. When adding models, therefore, we fixed some fit parameters in order to never exceed 5 free parameters. In particular we fixed the absorption column density to the Galactic value.
A simultaneous Ginga-ROSAT observation of Fairall 9
suggested (despite
the rather short exposure time, 578 s, of the ROSAT observation,
obtained during the All Sky Survey) the presence of a
soft excess (parametrized with a black body) and an emission line at 0.89
(0.15) keV whose identification is not obvious (Pounds et al. 1994).
The presence of a "true" soft X-ray emission is also suggested by
our analysis, as the black body model gives the best fit, and indeed
a which is very close to our acceptability threshold.
We first tried to add an absorption edge to the
black body and the reflection models. The results are summarized in Table
7 (click here). Both fits are acceptable (
of 0.96 and 1.16,
respectively) and the parameters are reasonable
(the edge energies correspond to C V in the fit with the black body,
and C VI in the fit with the reflecton model. In the last fit,
E0 is close to the O VIII edge). It must be noted, however, that
the normalization of the reflection component is about three times
that of the primary power law. The unfolded spectrum, as well as the
residuals, for the absorption edge plus black body fit is shown in panel (a) of
Fig. 4.1 (click here).
Figure 5: Unfolded spectra and residuals for: a) absorption edge plus
black body model for Fairall 9; b) edge plus reflection model for NGC 5548;
c) edge plus line for NGC 3783
We then tried to add an emission line (as suggested by Pounds et al. 1994)
to the absorption edge or the reflection models. The fit with the absorption
edge is unacceptable (), while that with the reflection
model is acceptable (
), with E0 corresponding to the
C VI edge (see Table 7 (click here)),
with a line energy of 0.76 keV (the two closest lines are
O VIII
recombinatin line at
0.65 keV and the O VII
recombination directly to the ground state at 0.74 keV).
Again, however, the
normalization of the reflection component is about three times
that of the primary power law.
We therefore conclude that in Fairall 9 a "true" emission is likely to be present, and that a further feature (probably an edge) is strongly suggested by the data.
Absorption edge + black body (![]() | ||
![]() ![]() | 2.82 (fixed) | |
![]() ![]() | ![]() | |
![]() | 1.94 (fixed) | |
E (keV) | 0.40+0.04-0.12 | |
![]() | 1.63+2.97-0.34 | |
kT (keV) | 0.107+0.004-0.004 | |
![]() ![]() | ![]() | |
Absorption edge + Reflection (![]() | ||
![]() ![]() | 2.82 (fixed) | |
![]() ![]() | ![]() | |
![]() | 1.94 (fixed) | |
E (keV) | 0.50+0.02-0.01 | |
![]() | 1.12+0.11-0.10 | |
E0 (keV) | 0.88+0.04-0.02 | |
![]() ![]() | ![]() | |
Emission line + Reflection (![]() | ||
![]() ![]() | 2.82 (fixed) | |
![]() ![]() | ![]() | |
![]() | 1.94 (fixed) | |
![]() | 0.76+0.02-0.03 | |
![]() | 0.01 (fixed) | |
![]() ![]() | 2.13+0.37-0.28 | |
E0 (keV) | ![]() | |
![]() ![]() | ![]() | |
For NGC 5548 an acceptable fit is obtained with
the absorption edge (and in fact the presence of an highly ionized
oxygen edge is well established since Nandra et al. 1993, who analysed
the same observation discussed here). However, the
of 1.23 is close to the "acceptability threshold",
suggesting that this model alone is possibly not sufficient to fully
describe the PSPC spectrum. Indeed, both Nandra et al. (1993) and
Done
et al. (1995) suggest the presence of a "true" soft excess.
In fact,
the inclusion of a black body or a reflection model in the absorption
edge fits reduces significantly the , which becomes about 1
in both cases. In the reflection model fit, E0 is
consistent with the C V edge, but the normalization of the reflection
component is twice that of the primary radiation.
The best fit parameters are summarized in Table 8 (click here) while
the unfolded spectrum and the residuals for
the edge plus reflection fit is shown in panel (b) of
Fig. 4.1 (click here).
We tried also to add an emission line to either the absorption edge
or reflection model. In the former case, the fit is completely
unacceptable (), while in the latter case
the fit is just above our acceptability threshold (
)
with a line energy consistent with the O VII recombination line, and
E0 consistent with C V (again, the normalization of the reflection
component is twice that of the primary radiation).
The best fit parameters for the latter case
are also summarized in Table 8 (click here).
Absorption edge + black body (![]() | |
![]() ![]() | 1.93 (fixed) |
![]() ![]() | ![]() |
![]() | 1.81 (fixed) |
E (keV) | ![]() |
![]() | 0.50+0.12-0.11 |
kT (keV) | 0.046+0.008-0.006 |
![]() ![]() | 0.069+0.015-0.008 |
Absorption edge + Reflection (![]() | |
![]() ![]() | 1.93 (fixed) |
![]() ![]() | ![]() |
![]() | 1.81 (fixed) |
E (keV) | 0.79+0.02-0.03 |
![]() | 0.53+0.10-0.11 |
E0 (keV) | 0.28+0.12-0.01 |
![]() ![]() | ![]() |
Emission line + Reflection (![]() | |
![]() ![]() | 1.93 (fixed) |
![]() ![]() | ![]() |
![]() | 1.81 (fixed) |
![]() | 0.55+0.05-0.06 |
![]() | 0.01 (fixed) |
![]() ![]() | ![]() |
E0 (keV) | 0.28+0.14-0.01 |
![]() ![]() | 10.6+1.2-0.6 |
NGC 3783 was by far the most intractable case in our previous analysis. All
the three models are completely unacceptable, the best fit being given
by the absorption edge with . This cannot be simply
due to the relative high statistic of the ROSAT observation, as there are
sources with similar or even greater total number of counts which are well
fitted by one or more models. Even excluding all channels above 2 keV
(see below) the fits remain unacceptable.
Therefore, there must be something intrinsic
in this source which prevents any simple model to fit satisfactorily.
The clue is given by an ASCA observation (George et al. 1995)
which discovered in the spectrum of this source a line emission identified
with the O VII recombination line at 0.57 keV. There is
evidence, even if less compelling, also for the O VIII line at 0.65 keV.
We then added a narrow emission line to both the absorption edge and reflection
models.
At a first glance, the inclusion
of the line does not make the fit acceptable if
the whole energy band is used. However, the inspection of the residuals
reveals a systematic excess above 2 keV, maybe due to calibration
problems (this is the only source in which we have found such a problem).
Excluding all channels above 2 keV,
the reflection + line fit remains fully unacceptable (
),
but the absorption edge plus line
fit turns out to be marginally acceptable (
; see
panel (c) of Fig. 4.1 (click here)).
The best fit parameters are summarized in Table 9 (click here).
For the sake of completeness
we tried also to add an absorption edge to the black body or the
reflection models. In the latter case the fit is unacceptable
(); on the contrary, in the former case it is
acceptable
(), with an edge energy close to that of
O VII. The best fit parameters are summarized in Table 9 (click here). Even if the
edge + black body model gives a better fit, the edge plus line fit seems
to be preferred after the ASCA result mentioned above.
Absorption edge + black body (![]() | |
![]() ![]() | 9.41 (fixed) |
![]() ![]() | ![]() |
![]() | 2.11 (fixed) |
E (keV) | 0.77+0.011-0.02 |
![]() | 1.87+0.20-0.19 |
kT (keV) | 0.137+0.014-0.011 |
![]() ![]() | ![]() |
Absorption edge + emission line (![]() | |
![]() ![]() | 9.41 (fixed) |
![]() ![]() | 13.0+0.6-0.5 |
![]() | 2.11 (fixed) |
E (keV) | 0.85+0.02-0.01 |
![]() | ![]() |
![]() | 0.53+0.01-0.02 |
![]() | 0.01 (fixed) |
![]() ![]() | 5.5+0.5-0.6 |
In their analysis of the ASCA observations of this source,
Guainazzi et al. (1997) attribute the observed spectral variations
to changes in the slope of the intrinsic power law correlated
with the luminosity of the object.
The slope varies from to
when
L(2-10 keV) goes from
to
.
Assuming this relationship, the intrinsic slope corresponding
to the ROSAT luminosity would be
.
We then repeated the fits adopting for the intrinsic
slope this very conservative value.
The results are basically unchanged: the only acceptable model
remains the blackbody (
), with a relative normalization
greater by about 30%. The other parameters are similar
to those obtained previously within 20%.
From the previous analysis on Fairall 9, NGC 5548 and NGC 3783 we have learned that complex scenarios may be acceptable when simple ones are not.
Since for all the remaining sources at least one model provides a good fit, we could not draw any significant information from a more detailed analysis. Of course any mixture of models will provide a good representation.
However, we had a further feedback from the analysis of the three sources that is important to verify.
Namely, in the case of warm absorber or reflection models, the presence of emission lines is possible. We have therefore added an emission line to all the sources where those two models were not able to fit the data. In most cases, the fit turns out to be formally acceptable (which is not surprising, as the number of free parameter is now equal to six, and the data are therefore somewhat overfitted), but the derived value of the line and/or edge energies are physically meaningless. In particular, this is the case for the edge model for Mrk 335, Ark 120, Mkn 509, MCG-2-58-22 and NGC 7213, while the same model for NGC 4051 remains statistically unacceptable. For the reflection model, the fit is unacceptable for MCG-6-30-15, while unphysical for Mkn 509. For Mrk 335, any line is set to zero. Finally, for NGC 4051 and NGC 3516 the fit is acceptable; the line energy is consistent with C V, while E0 is consistent with O VII, a situation which seems rather unplausible.
We therefore conclude that, in general, the inability of the warm absorber or reflection models to fit the spectrum of many sources is not due to the inadequacy of the parameterization adopted.