A&A 376, 413-421 (2001)
DOI: 10.1051/0004-6361:20011025
A. Orr1,2 - P. Barr1 - M. Guainazzi3 - A. N. Parmar1 - A. J. Young4
1 - Astrophysics Division, Space Science Department of ESA, ESTEC,
Postbus 299, 2200 AG Noordwijk, The Netherlands
2 -
Institut für Astronomie und Astrophysik Tübingen,
Waldhäuser Str. 64, 72076 Tübingen, Germany
3 -
XMM SOC, VILSPA, Villafranca del Castillo, Spain
4 -
Department of Astronomy, University of Maryland,
College Park, MD 20742, USA
Received 20 March 2000 / Accepted 9 July 2001
Abstract
We present the 0.1-200 keV BeppoSAX spectrum of MR2251-178 observed at two epochs in 1998 separated by 5 months.
Both epochs show identical spectral shape and X-ray flux.
Analysis of the combined spectra allow us to
confirm the presence of the ionized Fe K
line
detected by ASCA and to test the presence of
reflection from ionized material.
The best self-consistent spectral fit is obtained when including a
contribution from a mildly ionized reflector
(
erg cm s-1, i.e.
)
with a reflection fraction
.
An exponential cut-off to the direct power-law continuum is
then required at
keV.
The low energy spectrum is absorbed by ionized matter
with a column density
cm-2 and an X-ray ionization parameter
.
The warm absorber column is slightly lower
than that measured with ASCA. This change could be caused
by bulk motion. The BeppoSAX absorber ionization parameter globally
agrees with the "U versus Flux'' correlation found for MR2251-178 with
EXOSAT, Ginga and ASCA. This suggests that the
absorber is in a state of instantaneous ionization equilibrium.
Key words: galaxies: active - galaxies: quasars: individual - X-rays: galaxies
The large variety of observed spectral and temporal variability properties
of AGN is at the origin of the complex AGN classification scheme.
Attempts to find a simple explanation for the different classes have
resulted in the "unified model'' of AGN (Antonucci 1993), in which the
viewing angle of the observer is the key parameter.
AGN luminosity is an additional parameter which can directly influence
the physical state of matter located in the innermost regions
of the nucleus.
However the role of AGN luminosity is at present still not well
understood. For instance, recent X-ray studies are still inconclusive about
the existence of any fundamental differences between Radio Quiet (RQ) QSOs
and RQ Seyfert 1 galaxies (Sy 1). These two classes have often been regarded
as high- and low-luminosity manifestations of a same phenomenon
(e.g. Staubert & Maisack 1996; Lawson & Turner 1997).
For example, 2-10 keV power-law spectral slopes of low-z RQ QSOs have been found
to be similar to those of low luminosity Sy 1s
(
-2,
e.g. George et al. 2000; Reeves & Turner 2000).
Variability and warm absorber properties appear to be comparable
in RQ QSOs and Sy 1s (George et al. ibid.).
However, the strong Compton reflection "hump'' which is
common in Sy 1s has not been clearly detected in luminous RQ QSOs
(Williams et al. 1992;
Lawson & Turner 1997; Reeves et al. 1997; George et al. 2000; Reeves & Turner 2000).
An analysis of ASCA data for high z RQ QSOs (Vignali et al. 1999)
shows that the 2-30 keV (source frame) spectra are well described
by a single power-law.
Nandra et al. (1997b) showed that Fe K
line emission in RQ QSOs is
weaker than in Sy 1s and that the line profile and centroid energy
both seem to
depend on the AGN luminosity. This trend has recently been confirmed
by George et al. (2000).
Several mechanisms have been proposed to explain the differences between RQ QSOs and Sy 1s.
For instance relativistic beaming could enhance the underlying X-ray continuum emission
but not the iron line,
which is thought to arise by fluorescence from cold or hot reflecting
material (Williams et al. 1992).
Another interpretation is that the X-ray luminosity of the central
source
regulates the ionization state of the Compton scattering medium which
is probably the accretion disk itself. As the luminosity increases the
atoms in the disk become increasingly ionized. At high ionization states the iron line
is emitted at higher energies and its flux may decrease (Nandra et al.
1997b and references therein; Matt 1998; Ross et al. 1999, hereafter RFY99).
Ultimately,
the optical depth for photo-electric absorption decreases to such an extent
that the continuum flux <30 keV is mostly reflected from the ionized
disk and no longer absorbed (Basko et al. 1974;
RFY99).
Narrow band X-ray observatories have failed to detect any "hard tails''
in RQ QSOs
and the reflection component in these sources may so far have been confused
with the underlying continuum emission.
However, at high X-ray energies the cut-off due to Compton recoil
and Klein-Nishina effects may be detectable.
In this paper we present broad band BeppoSAX spectroscopy of a RQ QSO. We show that this source has an ionized iron line and exhibits a high energy cut-off, despite lacking a Compton reflection hump. We compare our data with the predictions of reflection models in order to test for the presence of an ionized accretion disk.
MR2251-178 is a nearby (z = 0.064) X-ray bright RQ QSO
(
erg s-1).
Early X-ray observations of this source with the Einstein Observatory
revealed both a soft excess and variable intrinsic absorption. This was
the first detection of a warm absorber in an AGN (Halpern 1984).
The source was later monitored by EXOSAT between 1983 and 1984 with a
sequence of 15 observations (Pan et al. 1990).
During this period the 2-10 keV flux
varied between 1.8 and
erg s-1 cm-2.
The changes occurred on time-scales as short as
10 d.
In the same period the column density of the warm absorber also varied.
Strong correlations between the soft excess flux,
the warm absorber column density variability and the source luminosity were found.
A subsequent reanalysis of the EXOSAT data together with Ginga observations
(Mineo & Stewart 1993) showed that the spectral behavior can be well
described using a self-consistent warm absorber model, and that the ionization
parameter of the gas is clearly correlated with the continuum flux.
ASCA observations in 1993 and 1996 (Otani et al. 1997;
Reeves et al. 1997) reveal a weak emission
line at
6.57-0.07+0.09 keV due to partially ionized iron.
In this case too, the ionization state of the warm absorber correlates well
with the measured X-ray flux. High ionization absorption lines
have been observed in the UV by Monier et al. (2001), these appear to
be related to the X-ray warm absorber.
The BeppoSAX observations of MR2251-178 extend the spectral range of previous X-ray spectroscopy of this object. At the same time they also expand the luminosity range of "warm-absorbed'' AGN observed by BeppoSAX. A preliminary analysis of the warm absorber data is given by Orr et al. (2000).
We present results obtained with the Low-Energy Concentrator Spectrometer (LECS; 0.1-4 keV; Parmar et al. 1997), the Medium-Energy Concentrator Spectrometer (MECS; 1.8-10 keV; Boella et al. 1997), the High Pressure Gas Scintillation Proportional Counter (HPGSPC; 4-120 keV; Manzo et al. 1997) and the Phoswich Detection System (PDS; 15-300 keV; Frontera et al. 1997) on-board BeppoSAX.
All these instruments are co-aligned and collectively referred to as the Narrow Field Instruments, or NFI. The MECS consists of two grazing incidence telescopes with imaging gas scintillation proportional counters in their focal planes. The LECS uses an identical concentrator system as the MECS, but utilizes an ultra-thin entrance window and a driftless configuration to extend the low-energy response to 0.1 keV. The non-imaging HPGSPC consists of a single unit with a collimator that remained on-source during the entire observation. The non-imaging PDS consists of four independent units arranged in pairs each having a separate collimator. Each collimator was alternatively rocked on- and off-source during the observation.
The region of sky containing MR2251-178 was observed by BeppoSAX at two epochs:
from 1998 June 14 17:58 UT to June 18 05:33 UT and
from 1998 November 12 21:23 UT to November 16 05:32.
Good data were selected from intervals when the elevation angle
above the Earth's limb was >
and when the instrument
configurations were nominal, using the SAXDAS 2.0.0 data analysis package.
The standard PDS collimator dwell time of 96 s for each on- and
off-source position was used together with a rocking angle
of 210
.
LECS and MECS data were extracted centered on the position of MR2251-178
using radii of 8
and 4
,
respectively.
The June exposures in the LECS, MECS, HPGSPC, and PDS instruments
are 60.3 ks, 82.7 ks, 53.3 ks, and 41.1 ks, respectively.
The November exposures
are in the same order 30.5 ks, 61.2 ks, 13.9 ks, and 35.2 ks.
Background subtraction for the LECS and MECS were performed using the standard background files. For the HPGSPC it was carried out using data obtained when the instrument was observing the dark Earth. Finally the PDS background was estimated from the offset field according to the standard procedure.
![]() |
Figure 1: 0.1-2 keV and 2-10 keV light curves and hardness ratios of the June and November 1998 BeppoSAX observations of MR2251-178. |
Open with DEXTER |
We have examined the LECS 0.1-2 keV and the MECS 2-10 keV light curves
for the 1998 June and November observations as well as the corresponding
hardness ratios. The data were binned into intervals of 10 ks.
The count rates and the hardness ratios are both relatively similar
at the two epochs.
The 1998 June light curves are well fit with a
constant count rate. This gives, respectively, for the soft and hard
X-ray bands
and
counts s-1
with corresponding fit statistics
and 0.9 (degrees of freedom,
d.o.f. = 29).
Fits to the November data give
and
counts s-1, respectively, with
(d.o.f. = 24) and
(d.o.f. = 21).
The poor statistics of the constant fits to the November data are an
indication of variability. In fact, the November light curves are
better fitted with slowly decreasing linear functions (minus
15%
in 70 hours).
Finally, an interval of 30 ks beginning 10 ks after
the observation start was ignored in the MECS November light curve. This
time interval covers a drop in MECS count rate due to a change in pointing
which could not be corrected by using the standard housekeeping data.
The cleaned and linearized MECS November image shows what appears to
be an extended, oblong source at the position of MR2251-178. A comparision of
light curves for different event extraction radii shows that this is
an artefact caused by the displacement of the source in the MECS field
of view. A similar displacement occurred in the LECS, however the
available housekeeping data allow this effect to be corrected.
The 1998 June 2-10 keV flux is
erg s-1 cm-2 (MECS) and the November
flux is
erg s-1 cm-2.
model | ![]() |
![]() ![]() |
EW |
![]() |
![]() |
![]() |
keV | eV | keV | ||||
PL |
![]() |
4.23 (162) | ||||
PL+FeK![]() |
![]() |
6.54-0.10+0.11 | ![]() |
1.22 (156) | ||
COPL+FeK![]() |
![]() |
6.54 -0.11+0.12 | ![]() |
0.74 | 0.59-0.14+0.15 | 0.94 (155) |
133-35+64 | 0.87 | 0.33 -0.13+0.12 | ||||
1.32 -0.07+0.08 | 0.20-0.05+0.06 | |||||
model | ![]() |
![]() ![]() |
EW | R | ![]() |
![]() |
pexrav+FeK![]() |
![]() |
6.54-0.11+0.13 | 73-23+24 | <0.38 | 0.93 (154) | |
pexriv+FeK![]() |
1.62-0.02+0.01 | 6.53-0.12+0.14 | 62-25+12 | 0.16-0.09+0.15 | 465.1-240+2003 | 0.89 (153) |
115-30+38 | ||||||
RFY99 |
![]() |
102-26+39 | 0.11-0.05+0.06 | 1625 -930+1422 | 0.94 (155) |
![]() |
Figure 2: The summed June and November 1998 BeppoSAX spectra of MR2251-178. The MECS data are represented by small filled squares, the HPGSPC data by open circles. The fit model is RFY99 with a warm absorber. |
Open with DEXTER |
![]() |
Figure 3:
Contour plot of the parameters
![]() ![]() |
Open with DEXTER |
![]() |
Figure 4: Data-to-model ratio for a simple power-law with galactic neutral absorption. The MECS data are shown with solid squares and the HPGSPC with open circles. The warm absorption and the high energy cut-off are evident. |
Open with DEXTER |
The June and November spectra of MR2251-178 were investigated both individually and as a
single, summed spectrum by simultaneously
fitting data from all the BeppoSAX NFI with the help of the XSPEC 11.01
spectral fitting package.
The LECS and MECS spectra were rebinned to oversample the full
width half maximum (FWHM) of the energy resolution by
a factor 3 and to have, additionally, a minimum of 20 counts
per bin to allow use of the
statistic.
The HPGSPC spectrum was rebinned to have a minimum of 40 counts per bin
and the PDS data were rebinned using the standard logarithmic binning recommended for this instrument.
Data were selected in the energy ranges
0.1-4.0 keV (LECS), 1.8-10 keV (MECS), 8.0-12 keV (HPGSPC),
and 15-200 keV (PDS)
where the instrument responses are well determined and sufficient
counts obtained.
The photo-electric absorption
cross sections of Morrisson & McCammon (1983) and the
solar abundances of Anders & Grevesse (1989) are used throughout.
Factors were included in the spectral fitting to allow for flux normalization uncertainties between the instruments. For all instruments factors were constrained to be within their usual ranges during the fitting (see Fiore et al. 1999). In this paper all uncertainties are 90% confidence intervals for one parameter of interest, unless stated otherwise.
Because of the lower exposure times during the November 1998 observation the
uncertainties on the fitted parameters are slightly larger than for those of the June spectrum.
However, our fit results indicate that absolutely no significant spectral
change occurred between the two epochs which are separated by 5 months.
All the fit parameters for the individual spectra are compatible within the
statistical uncertainties for any of
the models tested. For this reason and because the observed fluxes in June and November are
nearly identical with no large structure change in the light curves we have summed the spectra
for the two epochs in order to obtain a higher signal to noise ratio.
We tried fitting the summed continuum with several models (see Table 1), starting with a power-law
which gives a totally unacceptable fit
(
,
d.o.f. = 162). Strong deviations appear in the ranges
-3 keV and at
keV as illustrated in Fig. 4. The spectral complexity in the
soft X-rays is due to the warm absorber and is discussed further on.
At high energies the data points clearly fall below a simple power-law.
Highly significant improvement is brought with an exponential
high energy cut-off to the power-law at
keV,
(the confidence level of the improvement is >99.9%;
F-statistic F > 40 for
d.o.f. = 1, with respect to a
simple power-law, with or without an Fe K
line or a warm absorber;
the value of the cut-off energy is discussed below).
However, as described in the following sections,
the best fit to the summed 0.1-200 keV
BeppoSAX spectrum of MR2251-178 is obtained with an ionized reflection model.
Such a model predicts a drop in high X-ray flux but, unlike the case of
reflection onto cold matter, a weaker or absent "Compton hump''.
The knowledge of the spectral shape above 10 keV is an important key
to the interpretation of the present BeppoSAX data. Therefore we have taken
special
care in verifying the quality of the PDS data (for details about the PDS
instrument see Frontera et al. 1997).
It has been shown (by Guainazzi & Matteuzzi 1997)
that the total systematics in the PDS 20-200 keV range
due to background subtraction are lower than 0.026 s-1(corresponding to approximately 5.8% of the PDS 20-200 keV
source count rate for MR2251-178, which is 0.45 cts s-1). The observed flux of
a one mCrab source with photon index 1.5-2,
is well above the systematics ()
up to 60 keV and remains
above the systematics (
)
up to 100 keV.
(MR2251-178 has a flux of
3 mCrab in the 20-200 keV range).
We have also checked the consistency of the spectra measured by the
two PDS half-arrays
(A & B) which see the sky (source and background field) through two
different collimators. This was done by fitting either one of the PDS
half-arrays (A & B) spectra together with the LECS, MECS and HPGSPC
data. Because the data were obtained during two epochs (June and
November 1998), we first tested the half array data for June only
and then for the summed
June and November. This is justified because the PDS rocking
collimator positions happen to coincide within 0.6 degrees.
Using the "cut-off power-law + 3 edges + Fe K
line'' model
in Table 1 one finds that the cut-off energies
are all compatible with those
found using the combined PDS data (A + B half-arrays). Also, in all
cases the presence of the cut-off is if significant at greater than the
99.5% confidence level using the F-statistic.
We have also tested whether the PDS so-called "spike'' events can contribute to the PDS spectral shape. The answer is no: in our two observations of MR2251-178 the contribution of the "spike'' events to the PDS spectral shape is negligible.
However, PDS event filtering using the so-called "Variable Rise Time'' (VRT) selection does bring a significant improvement to the quality of fit, by increasing the signal to noise ratio of the PDS data. Therefore this filtering has been applied to all our PDS data.
Our results with a pexriv model including a warm absorber and an Fe
K
line (see also Fig. 3) are the following:
a power-law slope
,
a high energy exponential cut-off of the
incident power-law at
keV,
the temperature of the
ionized reflector
being fixed at 106 K,
an ionization parameter of the reflector
erg cm s-1 and
finally the scaling factor for reflection
,
which is in fact the solid angle
subtended by the reflecting disk.
It should be noted that, for these reflection models, we use
as it is
defined in RFY99, i.e. with the ionizing flux integrated between 0.01-100 keV.
Because in the pexriv model
is defined
between 5-20 keV we have made the necessary
conversions, assuming a photon index
(i.e.,
).
The fit statistics are good, with
for 153 d.o.f.
The low value of the scaling factor R tells us that most of the flux
in this model is still coming from the direct, non-reflected component.
Furthermore, one should bear in mind that a moderate ionization
parameter such as obtained with this model, still produces a significant
reflection "hump''.
Since the present data obviously lack such a hump the scaling factor
cannot be large.
In all our reflection fits we fixed the inclination angle of the disk with
respect to the observer so that
.
This was necessary because
in our fits the viewing angle and the reflection scaling factor are not
independent, nor are they well constrained with respect to one another
at low values (
or
).
In fact, in the present pexriv fit the entire range of numerically
permitted
values, i.e.
-0.95 lies within the
99% confidence interval for this parameter. At
R is poorly
constrained with a lower limit value of R > 0.4. At
we have
0.05< R < 0.22.
R and
are linked in the fitting process because smaller values
of i produce stronger reflection humps
and vice-versa.
Reflection on cold matter (e.g. neutral Compton reflection model
pexrav in XSPEC) gives, compared to pexriv, a slightly worse
fit with
for 154. The difference is significant at
the 99% confidence level.
The neutral reflection scaling factor is weakly constrained with an upper limit
.
The pexriv model in XSPEC does not include Fe Kemission and one therefore needs to add a separate Gaussian line component
to the fit model.
Because of this drawback we have also tested the self-consistent
ionized reflection models published by Ross and collaborators (RFY99)
which include Fe K
line emission and take into account the destruction
of K
photons by Auger effects.
The models were specially computed for the values of the incident
power-law slope observed in MR2251-178.
As with pexriv we set the iron abundance at the solar
value and included a warm absorber. The RFY models do not include the
"incident'' continuum so it is necessary to add
it separately, in this case an exponentially
cut-off power-law. We hereafter refer to this combination of components
as the "RFY99 model''.
The power-law slope of the incident continuum and the slope specifying the
reflection component were fixed to have the same value.
It should be noted that, unlike pexriv, the RFY99 reflection
assumes a sharp, step-like cut-off
of the incident spectrum at exactly 100 keV. This causes the RFY99 reflection
component to decline steeply above
50 keV.
The RFY99 model gives a good a fit with
for 155 d.o.f., see Fig. 2.
If the incident spectrum is taken to be a power-law without an exponential
cut-off the fit becomes significantly worse (
).
The "RFY99'' reflected fraction is
0.11 and is compatible with the pexriv value.
Both models also give comparable high energy cut-offs
at
100 keV. We have checked that this value is not an artefact caused by the sharp cut-off
assumed in the RFY99 reflection component (at
keV and
keV). Indeed,
if a "hard tail'' is added to the RFY99 reflection component, with an exponential
cut-off at 200 keV, the fitted continuum cut-off energy remains at
100 keV with a 90%
confidence upper limit
keV. The reason for this is most likely the relatively low reflection scaling factor,
.
A noteworthy analogy can be seen between the pexriv and RFY99 models concerning the
ionization parameter of the reflector. Due to non-linear
space
two distinct fits are possible with the RFY99 models. In the
first case the ionization parameter is
erg cm s-1
and in the second
erg cm s-1. This second value corresponds
to a nearly neutral reflector and gives a worse fit than with the first value
(
). Such a behavior is
analogous to the pexriv versus pexrav solutions and therefore
gives support to the validity of the ionized reflection models and their fitted parameter values.
These reasons lead us to consider hereafter only the fit solution yielding the
higher value of .
The ionization parameter measured with pexriv is
poorly constrained making it difficult to compare with the
RFY99 value.
The value of
(with the RFY99 model + 3 edges)
is driven mainly by the two following factors:
first, the intensity of the Fe K
line
and second,
the low energy (<6.4 keV) spectral shape.
This is verified by performing the fits over limited energy
ranges. For instance, excluding the 5.5-8 or 5-12 keV range leads to
significantly higher ionization parameters, and weaker Fe lines.
Likewise, if the range below 5 keV is excluded, then
decreases significantly.
implies a hot Fe K
line in the reflection component,
with
keV and
keV.
This is "hotter'' than the line energies found with the other,
non-self-consistent, fit models in Table 1.
However, from Fig. 5 we see that the "hot'' line is
still consistent with the MECS data.
Indeed, no large fit residuals appear at the Fe K
line
with the RFY99 model.
Nevertheless, an improvement in
can be brought to
the global RFY99 fit if an additional Fe K
line
(
keV,
eV)
is included (
,
for
(d.o.f.)=2). But the
model is then no longer self-consistent and, in order to "compensate''
for the additional line, the ionization parameter of the reflector tends to
increase and the reflected fraction to decrease.
It may be a hint that a slightly different Fe abundance is required.
For instance, increasing the Fe abundance increases both the line flux and
the depth of the Fe K-edge. However, any iron over-abundance must be
relatively low.
This is because the ionization parameter derived from the RFY99 model
indicates that the reflecting matter is in the regime where
ions are already too highly ionized to permit Auger destruction of iron
line photons. Therefore these photons escape from the reflector to
produce the observed hot iron line (Fabian et al. 2000).
Another possibility is that the line is due to neutral reflection from matter
further out, e.g. the torus or a cold phase of the accretion disk.
![]() |
Figure 5:
Fe K![]() ![]() ![]() |
Open with DEXTER |
Fits of the Fe K
line (see Fig. 5)
show that it is weak but very significant at >99.9% confidence level
(the F-test value for the presence of a line in the cut-off
power-law model of Table 1 is F=15.9 for
a difference of 2 d.o.f.).
A narrow line fit gives
keV, an equivalent width
eV
(the continuum is here the best fit pexriv model).
The line energy is consistent with ionization stages Fe XIII-XXIII.
A significantly worse fit is obtained if one forces the line energy to be
6.4 keV (>95% confidence level).
In the fit models of Table 1
the width of the narrow line has been frozen at
keV because no there is no significant
improvement in the fit statistics when leaving
free.
Nevertheless, we checked whether broad wings are present beneath
the narrow core.
We fitted the MECS continuum spectrum with a power-law continuum and
galactic absorption to which we added, in turn, a narrow line
or a "narrow plus broad Gaussian line''. In the second case
both lines were required to have the same centroid energy.
There is no strong evidence for the presence of a broad Gaussian component of
the Fe K line. Such a broad line is
not required by the data, the confidence level for its inclusion
being only 80%, using the F-statistic. The resulting line
width is
keV.
The equivalent widths of the core and the broad components are then
53.0 and 116.0 eV, respectively. So, we cannot completely
rule out some broadening of the line by e.g. Compton scattering.
Likewise, we have performed fits with the XSPEC diskline model using
the MECS data alone and compared them with the broad Gaussian line fits.
The fit quality of either model is statistically
equivalent.
Therefore it is not possible to formally exclude
the presence of a relativistically "blurred'' line.
The width of the peak line emission remains lower than the instrument's
energy resolution at 6 keV (8% in the MECS).
When testing the diskline model we tried to minimize the amount of
free parameters by estimating the outer radius of the disk.
We considered the simple case of an annulus illuminated from one
side and a non-rotating central black hole
with a mass of
.
This mass was derived for MR2251-178 by
Brunner et al. (1997) by fitting geometrically
thin
accretion disc models. They also found an inclination of the
disc of 70 (39-82) degrees.
From the value of the ionizing X-ray flux and the ionisation
parameter obtained with the RFY99 fit it is possible to estimate the
surface area of the ionized disk, assuming a hydrogen number density of
cm-3. From the value of this surface area one
can derive the outer radius of the disc (
6.46
), the inner
radius being fixed at 6
.
The resulting diskline feature is double peaked, several
keV broad and skewed. The blue peak is, however,
about 4 times stronger in intensity than the red
peak. And the FWHM of the blue peak is <0.1 keV (not resolved by
BeppoSAX).
In conclusion, our results with a narrow emission line
confirm the previous measurements of the line made at higher energy
resolution with ASCA (
keV,
eV,
keV, Reeves et al. 1997)
and they also are consistent with the EW-luminosity relation suggested by
Iwasawa & Taniguchi (1993) and confirmed by Nandra et al.
(1997b.)
This relation is a strong anticorrelation between the luminosity and the
EW, an "X-ray Baldwin effect'' which suggests Fe K
line core EWs lower than
80 eV at
erg s-1,
which is the luminosity of MR2251-178.
A soft excess component is not required by the data.
The ionized gas of the warm absorber
can account for the spectral complexity below 2 keV.
Likewise, we see that the best fitting models (which include a warm absorber)
do not require excess cold absorption. The upper limit on excess neutral
absorption is
cm-2, using the RFY99 model.
The soft X-ray spectrum of MR2251-178 is complex and can be well fit by considering the
effects of strong absorption
by partly ionized matter. Below 1 keV this absorption is essentially due to
highly ionized oxygen.
BeppoSAX cannot spectrally resolve the two oxygen edges O VII and O VIII
at 0.74 and 0.87 keV in the source rest frame.
Therefore fits have been performed by multiplying the continuum model by
either a single "blended'' absorption edge or two oxygen edges with
energies fixed at the values above.
The "blended'' edge parameters are:
keV,
.
In the second case,
and
.
A third edge brings further significant improvement to the fit statistics
(F-statistic F=21.41 for
d.o.f. = 2).
This third edge (
keV,
)
can be
due to Ne X K, Mg I-IV K or Ni XIV-XV L.
Alternatively, fits with warm absorber models were tested. The models
were calculated using ION98, the 1998 version of the photo-ionization code ION
(Netzer 1996). They were combined with a cut-off power-law continuum
and a Gaussian Fe K
line.
The best fit gives a column density of the warm absorber
cm-2 and an oxygen ionization parameter
.
The oxygen ionization parameter
is the unitless
ratio of ionizing photon flux to the
electron density and is defined by integrating the rate of photon emission over the oxygen
K-shell continuum, i.e. between 0.538-10 keV. In the present case
translates to an X-ray ionization parameter (defined by integrating
between 0.1-10 keV)
U0.1-10= 0.060-0.009+0.012.
The fit statistics with the warm absorber model are significantly better (at more than
90% confidence) than with the triple absorption edge fits:
for
d.o.f. = 2, leading to an F-statistic F = 2.9.
We note that the ionization parameters of the reflection component
and the warm absorber are similar: using the apropriate conversion
and flux integration between
and
U0.1-10 (e.g. Netzer 1996;
George et al. 1998) the reflector has
U0.1-10 = 0.04
(0.02-0.22), with the pexriv fit and
U0.1-10 = 0.14
(0.06-0.27), with the RFY99 fit, respectively.
But the warm absorber is optically thin and the RFY99 reflector
optically thick. So, despite their similar ionization parameters
they most likely are distinct.
The comparison of our warm absorber fit results with previous warm absorber
results for MR2251-178 is complicated by the different formalisms used in the
literature
to define the ionization parameter, in particular the integration range
of the ionizing photon flux.
For a start, we note that our measured warm absorber column,
cm-2
is consistent with values derived from Ginga and EXOSAT data
(Mineo et al. 1993), even when MR2251-178 was in weaker or brighter
states than seen by BeppoSAX. An earlier study of the same EXOSAT data
by Pan et al. (1990) seemed to suggest a column versus flux trend
whereby the warm absorber column decreases with increasing X-ray flux.
However, results from Mineo et al. (1993) as well as from Otani et al.
(1997), for ASCA data, excluded such a trend. Unlike the EXOSAT
data with relatively low statistics, the high signal to noise
ASCA spectra indicate slightly lower values of
than
the BeppoSAX spectra do (
cm-2).
Finally, the Ginga, EXOSAT and ASCA data all show a well defined correlation
between the ionization parameter and the X-ray flux. Using the
ionization parameter "translation'' curves of George et al.
(1998, Fig. 1) which give the relations between various
definitions of the ionization parameter,
,
U and
U0.1-10,
we estimate that the ionization parameter of the warm absorber in
MR2251-178 globally agrees with the EXOSAT, Ginga and ASCA "U versus flux''
correlation curves.
To summarize, it seems that the warm absorber in MR2251-178 is responding
in a consistent way to the slow variations of the ionizing continuum
flux and therefore is at most times in a state of ionization equilibrium. Any change in absorber column density between the 1996
ASCA and the 1999 BeppoSAX observations could be explained by bulk
motion of the absorber material across our line of sight.
Both the XSPEC pexriv and the RFY99 models give very good fits to the BeppoSAX spectrum of MR2251-178.
Of all the models we tested only the RFY99 model can by itself and in a
self-consistent way account for the ionized Fe line, part of the
high energy cut-off and the lack of reflection hump,
however, this model assumes a reflector with
constant density. The upper layers of the reflector, which are subject
to strong external illumination probably have lower densities and higher
effective ionization parameters than the inner layers.
This would tend to suppress Fe K
emission in the outer layers
(RFY99).
Nayakshin et al. (2000) and Ballantyne et al. (2001)
have recently calculated models of reflection on an ionized atmosphere in
hydrostatic equilibrium. Nayakshin et al. (2000)
find that in certain conditions
Fe K
photons can nevertheless be emitted:
if the incident spectrum is steep (
)
or if the
intrinsic disk flux exceeds the X-ray illuminating flux, in these cases
the reflecting layer is expected to yield ionized iron emission lines.
Ballantyne et al. (2001) show that the constant density models
of Ross & Fabian (1993) actually often give
accurate fits to their "hydrostatic'' reflection spectra,
both statistically and parametrically, assuming an ASCA-like spectral
resolution. On the other hand, the XSPEC pexriv and pexrav
models give only poor fits to their spectra.
Whereas the value of the Fe K
line energy, as measured by
ASCA and BeppoSAX in MR2251-178, clearly excludes neutral Fe, the case for
ionized reflection is less firm. The data strongly suggest
that little reflection is
present and that we are in fact mainly observing the direct ionizing
continuum since very good fits are obtained with an exponentially
cut-off power-law
(
,
d.o.f. = 155) or a pexriv or RFY99 model
allowing the direct component to "shine'' through at more than 80%
(
,
d.o.f. = 153 with pexriv, see also Table 1). Because of the weak contribution of the reflection
component the shape of the cut-off at
100 keV must be intrinsic to the primary X-ray emission.
However, despite its weak contribution, an ionized reflector can
produce the observed ionized Fe emission, as shown with the RFY99 fit.
Broad band observations of Sy1 galaxies with BeppoSAX have revealed exponential cut-off energies
100 keV
in only a couple Sy1s:
e.g. NGC 5548 (
keV, Nicastro
et al. 2000) and NGC 4151 (70 keV, Piro et al. 1998).
Whereas in other Sy1s the cut-off occurs at much higher values,
with lower limits around 200-300 keV (cf. Gondek et al. 1996;
Matt 1998).
More recently, Zdziarski et al. (2000) have measured the average OSSE spectra
of Sy 1 and Sy 2. Using a cut-off power-law they derived expontenial cut-off energies
of 120
-60+220 and 130
-50+220 keV
for Sy 1 and Sy 2s, respectively.
It appears therefore that the cut-off energy measured by BeppoSAX for the RQ QSO
MR2251-178 is compatible with the average OSSE Sy 1 value as well as the
BeppoSAX values obtained so far for Sy 1s.
We have for the first time observed MR2251-178 in the 0.1-200 keV
energy range using BeppoSAX.
The spectral fitting described above shows that this source possibly
combines two distinct signatures of ionized reflection which, to our knowledge,
have not been detected together before in an individual radio quiet AGN:
a strongly ionized
Fe K
emission line and a contribution, in this case relatively weak,
from an ionized reflection continuum.
A spectral break is apparent at higher energies which can be parametrized with the
help of an e-folded power-law. The cut-off energy is measured to be in the range
80-190 keV (90% confidence interval) and is compatible with values found for
Sy 1 galaxies.
Acknowledgements
The BeppoSAX satellite is a joint Italian-Dutch programme. AO acknowledges a Fellowship of the Swiss National Science Foundation during part of this work. She expresses gratitude to Prof. R. Staubert for having hosted her at the IAA in Tübingen during this period and wishes to thank J. Wilms and other colleagues form the IAAT for fruitful discussions about MR2251-178. AO is very thankful to Prof. Netzer for allowing to use his ION98 model and giving valuable advice about the warm absorber fitting.