A&A 401, 895-901 (2003)
DOI: 10.1051/0004-6361:20030124
G. Risaliti1,2 - L. Woltjer1,3 - M. Salvati1
1 -
INAF - Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, 50125 Firenze, Italy
2 -
Harvard-Smithsonian Center for Astrophysics, 60 Garden Street,
Cambridge, MA 02138, USA
3 -
Observatoire de Haute-Provence, CNRS, 04870 Saint-Michel l'Observatoire, France
Received 7 October 2002 / Accepted 21 January 2003
Abstract
We present BeppoSAX observations of the two radio galaxies PKS 1934-63 and S5 1946+708.
Strong Iron K
lines are detected in both objects indicating that the
two sources are absorbed by column densities higher than
1024 cm-2.
Combining radio continuum, HI absorption and X-ray data we can constrain the physical state and the
dimensions of the absorber. We find that the putative obscuring torus is molecular and located
at a distance higher than 20 pc from the center in S5 1946+70. In PKS 1934-63
no radio nucleus has been observed. If this is due to free-free absorption a
radius of only a few pc is inferred. Since the K
lines have been
detected only at 90% confidence, we also discuss the implications if they
would be not as strong as found from our data. From our results it appears
that the combination of hard X-ray data and spatially resolved HI absorption
measurements provides a powerful diagnostic for conditions in the absorbing
"torus''.
Key words: galaxies: active - X-rays: galaxies
In unified models of Seyfert galaxies Sy1/Sy2 and of quasars/radio galaxies Q/RG the Broad Line
Region BLR and the optical nucleus are hidden at certain orientations by an "absorbing torus''.
Direct evidence for such a torus comes from absorption at the lower energies X-rays from
the nucleus. While the "torus'' needs not have a truly toroidal shape, it has to be geometrically thick. For
the Seyferts the observed ratio of Sy1/Sy2 (1/3-1/6) suggests that the torus as seen from the
nucleus subtends an angle of order ,
and corresponding estimates have been made for RG.
Polarized optical broad emission lines in some Sy2 and RG have been ascribed to material above
the torus, which scatters nuclear light towards us. Also X-rays may be scattered by relatively
cool material with the iron K
line at 6.4 keV appearing in emission. When column densities in
the torus approach
cm-2, the direct X-rays are much attenuated and the equivalent
width (EW) of the iron line may become 1 keV or more. If the scattering material is warm
iron lines at 6.7-6.9 keV may appear.
While the X-ray data yield values or limits for ,
they do not contain information on the
physical and kinematical conditions in the torus. Absorption at 21 cm by atomic hydrogen
(HI) may be informative in this respect. HI absorption has been detected mainly in powerful
compact radio galaxies. Line widths of up to nearly 1000 km s-1 have been observed,
indicating that the HI is located rather close to the nucleus, presumably in the torus.
In other cases the velocities are smaller and the location of the HI more uncertain.
The absorbing matter generally covers only a small part of the radio source. It is therefore essential to have high resolution, preferably VLBI, data to determine the optical depth in HI in front of the nucleus. Conway & Blanco (1995) obtained VLA data for CygA and found that the X-ray and HI data may be quantitatively fitted to simple models of the torus. They also concluded that the radius of the torus should exceed 15 pc if strong free-free absorption is to be avoided, which would make the radio nucleus unobservable at 1.4 GHz. This radius is larger than inferred for typical Seyferts.
To obtain further information on the tori in RG we have observed the X-ray spectra of two
RG with Beppo-SAX. PKS 1934-63 is a powerful compact double with a separation of 158 pc.
VLBI observation failed to locate a nucleus between these components even at 8 GHz
(Tzioumis et al. 1999). It has been suggested that very strong free-free absorption might
be responsible (Woltjer 2000). HI measurements (Véron-Cetty et al. 2000) also failed to find
HI absorption in front of the radio source, except for a narrow weak feature, presumably due
to an isolated cloud. The second source observed with BeppoSAX is S5 1946+708, a compact symmetric
object (CSO
with a nucleus and two jets and an overall dimension of 100 pc.
The HI absorption profile in front of the nucleus has a total width of 600-800 km s-1 and
the absorption in front of the jets shows that the projected thickness of the "torus'' is about 30 pc with some narrower absorption further out (Peck et al. 1999). Further data
on the two sources are given in Table 1. Everywhere we have taken H0= 67 km s-1 Mpc-1.
PKS 1934-63 was observed on Nov. 11/12, 2000 and again on May 03-04 2001, each time for about
50 ksec. While in May the K
line appeared to have a lower EW than before, the low S/N ratio
was insufficient to conclusively demonstrate variability and so we combined the two data sets. The total
exposure times were 38 ksec for the LECS instrument (0.1-10 keV), 95 ksec for
the MECS (1.65-10 keV) and 48 ksec for the PDS (20-200 keV).
The source was not detected with the PDS.
The LECS and MECS images show the presence of an unknown serendipitous source 5 arcmin NW
of the target, with comparable flux.
To avoid contamination, the spectrum of PKS 1934-63 was extracted in a circular region with
radius 2 arcmin using the XSELECT reduction package. This small radius is the one for which
we obtain the highest S/N. The spectral analysis has been performed
with the XSPEC code.
The LECS and MECS spectra have been convolved with the response matrices provided by the
BeppoSAX Science Data Center and background subtracted on the basis of a long exposure of a
blank field. To check for systematic errors in this procedure we also extracted the background
spectrum from a source free region in our field. After rescaling for the vignetting factor the
two spectra agreed very well.
PKS 1934-63 | S5 1946+708 | |
z | 0.182 | 0.101 |
![]() |
15 | 1.0 |
![]() |
27.1 | 25.3 |
AV(mag) | 1.5 | 2.2 |
![]() |
43.2 | 42.0 |
Given the low S/N we can only fit simple models. A power law fit to the continuum yields a photon
index
,
but leaves a clear excess at 5.3 keV corresponding to 6.3 keV in the rest frame.
This is interpreted as a Fe K
line with very large EW=2.0 keV.
If we interpret this spectrum as reflection-dominated, we can assume a
reflection efficiency, R, to estimate the intrinsic continuum and a lower limit
for the absorbing column density,
.
Assuming R=5% (the maximum allowed
according to reflection models as in Ghisellini et al. 1994, and therefore
the one giving the lowest intrinsic continuum) we find
cm-2.
The results
are given in Table 2.
This interpretation is not unique, due to the poor signal-to-noise.
An alternative possibility is suggested by the separate analysis of the two observations.
The 2-10 keV continuum is higher in the second observation, while the line flux remains
constant. We fitted the two spectra with the photon indexes and the line flux fixed at the
values obtained in the combined fit, leaving the power law normalization free.
We found that the continuum variation is significant at a confidence level of 90%.
In the high-continuum observation the line is
detected only at a 1.2
level.
Therefore, we cannot exclude that the observed spectrum is the intrinsic emission from
the active nucleus, but that a variation by a factor of 2 in the continuum (not followed
by a line variation, indicating that the line is produced by
material far from the center) gives rise to a measured equivalent width in the combined
spectrum higher than the usual values for type 1 AGNs (100-300 eV).
The spectra from the two observations are plotted in Fig. 1. The results of the separate fits are listed in Table 2.
![]() |
Figure 1: Spectrum and best fit model for the BeppoSAX observation of PKS 1934-63. Triangles: LECS data of the first observation; circles: MECS data of the first observation; squares: MECS data of the second observation. |
Open with DEXTER |
PKS 1934-63 | |||
Parameter | Merged obs. | First obs. | Second obs. |
Power law: | |||
Photon index |
![]() |
![]() |
![]() |
Normalization (10-5 cts s-1 keV-1) |
![]() |
![]() |
![]() |
Iron line: | |||
Peak energy (keV, rest frame) |
![]() |
![]() |
6.4a |
Equivalent width (keV) |
![]() |
![]() |
![]() |
Flux 2-10 keV (10-13 erg cm-2 s-1)b |
![]() |
![]() |
![]() |
![]() |
14.2/17 | 4.3/7 | 4.4/7 |
![]() |
Figure 2: Spectrum and best fit model for the BeppoSAX observation of S5 1946+708. |
Open with DEXTER |
In conclusion both objects have rather steep spectra.
Combined with the high EW of the iron lines this suggests that the sources are Compton thick
(log
), with the observed X-rays resulting from scattering.
Observations with higher S/N ratio are needed to confirm these results.
We adopt a simple model in which a torus with radius r
parsec is illuminated by a point source
nucleus emitting X-rays with an energy spectral index
= 0.7 between 1 and 100 keV with a
luminosity of 1044 L44 erg s-1. The torus is taken to be isobaric with
temperature 103 T3 K and
particle density 106 n6 cm-3. We neglect the curvature of the torus.
Our aim is to use radio and X-ray data to constrain the density and size of
the circumnuclear absorber.
We shall make use of the
calculations of Maloney (1996) and Neufeld et al. (1994) for single phase isobaric models at
high pressure (
K cm-3), and of Maloney et al. (1996) at low pressure
(
K cm-3).
Conditions in the torus depend on the effective ionization parameter
which from Maloney et al.
we write, for
,
as
![]() |
(1) |
S5 1946+708 | |
Parameter | Best fit value |
Power law: | |
Photon index |
![]() |
Normalization |
![]() |
high ![]() |
![]() |
low ![]() |
![]() |
Iron line: | |
Peak energy (rest frame) |
![]() |
Equivalent width |
![]() |
Flux 2-10 keV (10-13 erg cm-2 s-1)b |
![]() |
![]() |
16.5/21 |
All tori have an atomic zone towards the nucleus, but if the total column density through the
torus
is large enough this is followed by a mainly molecular zone. From the calculations of
Neufeld et al. and of Maloney it is seen that the transition from atomic to molecular occurs in a
very narrow region from log
when the gas is fully atomic to
when it is fully molecular, for pressure
in the range
10
10- 1012 K cm-3. In the high pressure range the temperature averages about 7000 K
in the atomic zone and some 600 K in the molecular zone.
From Maloney et al. (1996) we see that at lower pressures
in the range
107-109 K cm-3 the transition is more gradual and temperatures as low as 100 K may be reached
in the atomic zone.
In a typical atomic zone the average
ionized fraction is around log(
and may be lower at pressures near 107 K cm-3.
In the molecular zone
is negligible.
The condition for a fully atomic torus, log
,
can be rewritten as
a condition in the torus radius versus density plane:
![]() |
(2) |
The ratio
in the molecular zone depends on the destruction of H2 molecules by X-rays and
on their formation on the surface of grains. Maloney et al. obtain an expression that in the
molecular zone may be written as
![]() |
(3) |
An additional condition may be that in a steady state the pressure in the torus should be higher than
or equal to the radiation pressure (Maloney 1996).
This can be written as follows:
![]() |
(4) |
![]() |
(5) |
For a molecular torus
and Eq. (5) is trivially
satisfied.
We can now complement the information on the total column density coming from
the X-ray data with the radio data, which provide constraints on the
free-free absorption,
,
and the column density of atomic hydrogen
.
If the torus is mainly molecular
there is still an atomic zone on the inside, which is responsible for the free-free opacity.
Neufeld et al. (1994) give an expression for
which at 1.4 GHz corresponds to
![]() |
(6) |
Since the nucleus of S5 1946+708 has been observed at 1.4 GHz, the free-free
opacity must satisfy
.
From Eq. (6) we then obtain:
![]() |
(7) |
Peck et al. (1999) have measured the HI absorption in front of the
nucleus of the source, obtaining for the HI column density log
,
with
the spin temperature.
At high densities radiative
excitation is negligible and in the molecular zone we should have
K, or log
which for
requires log(
,
or from Eq. (3) log
.
Treating the torus as a uniform layer we should evaluate
at the mid point
N24=0.5.
From Eq. (1), we then obtain:
![]() |
(8) |
With n6=200 and
the path length through the dense gas
is only 0.002 pc.
It is therefore likely that such a torus would be composed of small high density clouds or filaments.
As stated above, the models being used here are single phase at any given
radius. If the absorber is made by dense clouds, there is probably a confining
medium which could be relevant in the absorption processes we are studying.
We note however that a medium with temperature T> 108 K and
density
would have the right pressure to
confine our clouds, without significantly contributing either to the radio
or to the X-ray absorption. A denser, colder (T< 107 K,
confining gas could instead be relevant.
However, this possibility can be ruled out, for it would imply a
very high thermal emission in the soft X-rays, which is not observed.
All the above discussion is focused to the high pressure case. It is worth
considering the possibility of a low pressure, low temperature atomic torus.
Assuming
K cm3 and
we still have a
relatively high density. This implies that the spin temperature is
equal to the thermal temperature. Therefore, the relation of Peck et al. (1999) rules out the possibility of a Compton thick torus. We will further discuss
this scenario in the case of a Compton-thin torus (Sect. 3.2).
PKS 1934-63: Less information is available for PKS 1934-63.
Since no nucleus has been observed at 8 GHz we infer
at that frequency, corresponding to
at 1.4 GHz. However since
the dynamic range in the 8 GHz observations is not very high this condition
may be a bit too strict.
From Eq. (7) we then obtain
![]() |
(9) |
The case for an atomic torus in this source can be easily ruled out, using the
same equations as above to show that the condition
and Eq. (8) are not compatible.
Some comments are in order. The adopted models are all based on the
assumption that the tori in the two sources are Compton thick. We have taken
,
though the rather uncertain PDS results suggest perhaps even higher
values. However as far as we are aware no detailed calculations have been made
for tori with N24>1. While a thicker torus would not change
,
the inferred
might be affected. A second point is that the values of
L44 are very uncertain if the sources are indeed Compton thick and might
well be larger. And thirdly the basic assumption has been made that the same
medium is responsible for the X-ray, free-free and HI absorption.
![]() |
Figure 3:
Radius-density diagram for the torus around our radio
galaxies. a) S5 1946+708: the permitted parameters are given by the relation between
total absorbing column density, HI column density and spin temperature
(Peck et al. 1999). The two almost horizontal lines represent the constraints
on the free-free opacity. The two points on the left represent a solution with
an atomic, Compton thin torus, in the two cases of n6=0.1 and n6=0.001.
The dotted lines connecting these
points represent a possible solution in the low-density, low pressure regime.
b) PKS 1934-63: in the molecular torus case, the allowed region is at
the bottom-right of the diagram, and is delimited by the constraints on the
ionization parameter
![]() ![]() ![]() ![]() |
Open with DEXTER |
At lower pressures a cool atomic torus becomes a possibility. Since from the
data of Peck et al.
and since now
we should have
.
According to the results
of Maloney et al. (1996) for n6=0.1 this requires
and
for n6=0.001,
.
Again taking a one zone model where
these values refer to the midpoint
we obtain from Eq. (1) that
for n6=0.1 r=67 pc and for n6=0.001, r=282 pc. From the
plot we see that connecting the two points an extrapolation to higher
densities should fit to the line for the molecular torus, which, in fact, is
only marginally molecular. The pressures now are very modest: for n6=0.1 we
have
p11=0.001, a value more representative for the NLR. It may also be
verified that
is of the order of a few times 0.1.
We now analyze the case for Compton-thin absorption in PKS 1934-63. A low energy cut off is
not required by our BeppoSAX observations. From our data we estimate
cm-2.
Assuming
cm-2 we have log
L44=-0.09 and
.
With a temperature T=7000 K, the pressure condition for an atomic torus gives:
![]() |
(10) |
![]() |
(11) |
Therefore in this case an atomic torus is acceptable.
Note however that the pressure condition (Eq. (5)) is not satisfied. Therefore, this scenario is possible only if the absorbing gas is not in a steady state, or other mechanisms contribute to the internal pressure, in addition to the thermal component (for example, magnetic fields).
![]() |
Figure 4: Diagnostic diagram for Seyfert 2s of Bassani et al. (1999). We have superimposed the data for a sample of powerful radio galaxies for which a hard X-ray observation and [OIII] and Balmer lines measurements are available. Our two sources lie significantly above the region occupied by both Seyfert 2s and Compton-thin radio loud objects. |
Open with DEXTER |
The permitted region in the radius-density plane is indicated in Fig. 3b. The
pressure is high,
,
but should be possible at such a small
radius.
Some other powerful radio galaxies have data in the literature. Cyg A was
discussed by Conway & Blanco (1995) who fit their HI data with an atomic
torus with n6<0.2 and
.
For a molecular torus Maloney (1996)
finds
.
For Hya A Taylor (1996) found HI absorption corresponding to
,
while Sambruna et al. (2000) detected an absorbed
X-ray nucleus with
and
(for
H0=67 km s-1 Mpc-1). To reconcile the values of
and
we have to take
K which excludes a high pressure atomic
torus with T=8000 K. However from the results of Maloney et al. (1996) at
lower pressures we find a fit for n6=0.1 and T=250 K with
at the mid point and
.
It should be noted, however,
that iron K
line has not been measured in this source.
Bassani et al. (1999) have presented an interesting diagram for Seyfert galaxies with as coordinates
the equivalent width of the iron lines near 6.5 keV and the ratio of the observed 2-10 keV X-ray
flux to the absorption corrected [0III]
flux.
In this diagram (Fig. 4) the Sy2 follow a broad sequence from high
and low iron EWto low
and high iron EW. The interpretation of this diagram is that as the X-ray
absorption increases, the much weaker scattered component with its fluorescent iron lines becomes
more conspicuous. Plotting in this diagram the powerful narrow line radio galaxies with adequate data there
is perhaps a tendency for the average to be displaced towards higher
values,
with S5 1946+708 the most extreme case. An interesting possibility is that, since our results
point to a far absorber, the narrow line clouds emitting the [OIII] line are also
partially covered by the X-ray absorber. In this case, using the true [OIII] flux would shift
the source to the left in the plot, in the standard region for Compton-thick AGNs.
If on the other hand the K
line is not as strong as we found, then S5 1946+708 would come down vertically and fit in with other low absorption
objects.
Since in typical Seyferts the ionizing spectrum appears to have a strong "blue bump'' which
is probably absent in Liners (like S5 1946+708), thereby causing an unusually low ionization,
it may well be that its absence also causes
to be unusually high.
Alternatively since these very compact radio galaxies have been shown in some cases to be very
young (thousands of years, Conway 2002) it may be that the NLR has not yet had the time to fully develop.
More high quality X-ray spectra are needed to further investigate these possibilities.
Observations with BeppoSAX seem to show that both PKS 1934-63 and S5 1946+708
are Compton thick sources. This is suggested by the high equivalent width of
the iron K
lines (EW>1 keV) and, in the case of PKS 1934-63, by
the low X-ray to [OIII] flux ratio.
Powerful radio galaxies with X-ray absorbing column densities higher than 1024 cm-2 are expected on the basis of unified models, but are still rather elusive: to our knowledge, only one such object, 3C 321 (Sambruna et al. 1999) has been observed prior to the present work.
We have used a simple modelization of the absorbing torus (under the
hypothesis that the same medium is responsible for both the radio and X-ray
absorption) to estimate the density and distance of the nuclear absorber.
We find that in PKS 1964-93 the absorber is compact (R< 4 pc) while in S5 1946+708 the distance from the center is higher than 20 pc.
There are however some problems with the interpretation of the X-ray data as due to Compton-thick sources.
For PKS 1934-63 this does not pose particular
problems and, in fact, could serve as an explanation of the absence of a radio
core in this compact double. However there remains the difficulty that at 90%
confidence the source is variable, which seems in contradiction to it being
Compton thick on the basis of the high Fe K
line equivalent width,
also with 90% confidence.
For S5 1946+708 there are two problems: the inferred pressure at several tens of pc from the nucleus seems to be anomalously high, while its position in the Bassani et al. diagram suggests a rather low X-ray absorption. Both problems disappear when the source is not Compton thick.
As a consequence of these ambiguities, in our Discussion we have also estimated the physical parameters of the absorbers in the scenario of Compton-thin X-ray absorption.
For most of the radio galaxies the torus appears to have a radius of several tens of pc. Another case where this was suggested on the basis of the radio data alone is PKS 1353-341 (Véron-Cetty et al. 2000). Smaller tori as in the Seyfert galaxies tend to be optically thick in free-free absorption (O'Dea et al. 2000), while those in radio galaxies must in most cases be optically thin since the radio nucleus is observable.
To make further progress higher quality X-ray spectra are needed, while also further VLBI data on HI absorption are required.
Acknowledgements
This work was partially supported by the Italian Ministry for University and Research (MURST) under grant Cofin-00-02-36.
In the field of PKS 1934-63 there is another source at 5' distance at
38 56.7 and
38 2.7. This source is not identified in the
literature. With a power law fit we obtain
and F(2-10 keV)
erg cm-2 s-1.