A&A 444, 165-174 (2005)
DOI: 10.1051/0004-6361:20053441
A. Wolter 1 - I. M. Gioia 2 - J. P. Henry 3 - C. R. Mullis 4
1 - INAF - Osservatorio Astronomico di Brera, via Brera 28, 20121
Milano, Italy
2 -
INAF - CNR - Istituto di Radioastronomia, via Gobetti 101,
40129 Bologna, Italy
3 -
Institute for Astronomy, University of Hawai'i, 2680 Woodlawn Drive, Honolulu, HI 96822, USA
4 -
Department of Astronomy, University of Michigan, 918 Dennison Building, Ann Arbor, MI 48109, USA
Received 16 May 2005 / Accepted 18 August 2005
Abstract
We present in this paper optical and X-ray follow up observations
for three X-ray selected objects extracted from the ROSAT
North Ecliptic Pole survey which is a flux-limited, completely
identified survey. All three objects have X-ray luminosities in the
1044 erg s-1 regime and show narrow emission lines in their
optical discovery spectra, typical of QSO 2 type objects. Spectroscopic
data for the three QSO 2 candidates, obtained with the Telescopio
Nazionale Galileo, confirm the widths of the H
or H
emission lines are less than 750 km s-1. On the other hand
XMM-Newton data do not show any sign of obscuration as expected
for this class of objects. The X-ray spectra of the three objects
are all well fit by a single power law model with
with low energy absorption fixed to the Galactic value along the line
of sight to each object. Most observational evidence supports
the scenario where optical and X-ray obscurations are linked,
contrary to our findings. We discuss the unanticipated results of
these observations, and compute the space density in soft X-ray surveys
of this possibly new class of objects. Their spatial density in the
ROSAT NEP survey is 2.8
h3 Mpc-3 in a
CDM model with h =0.7. Unobscured QSO 2
candidates could go unrecognized in current X-ray surveys where the
low hydrogen column density is inferred by a hardness ratio rather than
a more precise X-ray spectrum measurement.
Key words: galaxies: active - galaxies: quasars: general - X-rays: galaxies - X-rays: individuals: RX J1715.4+6239 - X-rays: individuals: RX J1724.9+6636 - X-rays: individuals: RX J1737.0+6601
Absorbed AGN are a major fraction in the total emission budget of the diffuse X-ray Background (XRB; Schmidt et al. 1998), as was first predicted by Setti & Woltjer (1989). The model was later refined by Madau et al. (1994), Comastri et al. (1995), Gilli et al. (1999), among others. In more recent years the XRB was directly resolved with Chandra by Mushotzky et al. (2000), Brandt et al. (2001), Giacconi et al. (2001), Rosati et al. (2002), and with XMM-Newton by Hasinger et al. (2001). The original requirement of the Setti & Woltjer model, that the number of absorbed sources is approximately equal to that of unabsorbed ones, still holds. QSO 2 are deemed necessary to explain the hard source counts at fluxes of the order of 10-13 erg cm -2 s-1 in the ASCA 2-10 keV band where they should contribute at least 30% of the total emission (Gilli et al. 2001). The relative densities of absorbed and unabsorbed objects in the model can be compared to the relative densities found in the optical band. The local ratio between Sy 2 and Sy 1 in the optical band ranges from 2 (Huchra & Burg 1992) to 4 (Maiolino & Rieke 1995). Since there is noreason to expect a different ratio for high luminosity objects, we should expect a factor of 2-4 more QSO 2 than QSO 1. What is observed instead is a dearth of QSO 2 (see among others: Akiyama et al. 2002; Fiore et al. 1999; Della Ceca et al. 2003; Barcons et al. 1998).
Table 1: ROSAT NEP properties of the three QSO 2 candidates.
Since QSO 2 are a key ingredient in models of the diffuse XRB (Gilli et al. 2001; Comastri et al. 2000), they should be more easily detected in X-ray
selected samples because they have high X-ray luminosities.
No X-ray survey found a sizeable number of such objects
(e.g. there are no QSO 2 in the complete ASCA Large Sky
Survey; Akiyama et al. 2000), and only a few candidates have been found using ASCA
(Akiyama et al. 2002; Della Ceca et al. 2003; Ohta et al. 1996), ROSAT (Almaini et al. 1995; Barcons et al. 1998; Georgantopoulos et al. 1999),
and Beppo-SAX (Fiore et al. 1999).
Detection of QSO 2 is very controversial. Several QSO 2 candidates
have been proposed in the literature, but they have often been classified
later with other known types of objects. This is the case of AXJ0341.4-4453,
proposed as a type 2 AGN by Boyle et al. (1998), and later identified as a Narrow
Line Sy 1 (Halpern et al. 1999) on the basis of a broad component in the
H emission line (not observed in the discovery spectrum) plus other
spectral properties (see for the classification of Narrow Line Sy 1: Goodrich 1989; Osterbrock & Pogge 1985).
CHANDRA and XMM-Newton data have included new candidates
(Caccianiga et al. 2004; Mainieri et al. 2002; Dawson et al. 2001; Fiore et al. 2003; Norman et al. 2002; Stern et al. 2002)
but the number of confirmed QSO 2 is
still small. Their absence in X-ray selected samples could be explained if
"all'' sources are perfectly Compton-thick
(
cm-2), implying a complete absorption
of the X-ray radiation. Alternatively, the luminosity of the nucleus could
be so high that the obscuring material is effectively removed from the
circumnuclear regions, leaving a totally unblocked view of the Broad Line
region (BLR). Both interpretations seem unlikely. It has been
suggested (Risaliti et al. 2003) that some soft X-ray weak quasars with red
optical counterparts might be optically broad-lined objects with an
intrinsically weak and flat X-ray spectrum.
This model seems to be supported by the fact that there is no
one-to-one correspondence between optical and X-ray absorption. A number of
galaxies, optically classified as Sy 1, shows high intrinsic hydrogen
columns
or at least soft energies spectral curvatures that might be attributed
to ionized absorbers (Fiore et al. 2001; Nandra & Pounds 1994).
Also a number of objects,
optically classified as Sy 2, shows absorptions smaller than 1022 cm-2 even if they might still be reconciled with the unifiying scenario
(see e.g. Panessa & Bassani 2002; Pappa et al. 2001; Guainazzi et al. 2001).
The comparisons of optical and X-ray absorption has not yet been done
with a large statistical sample. This
will be possible in the near future for instance with the new samples
identified by the XMM Survey Science Center (Della Ceca et al. 2004) or from surveys
with CHANDRA (Silverman et al. 2005). First results seem to confirm
that the optical and X-ray absorption are linked (Caccianiga et al. 2004).
Therefore objects optically appearing as type 2 should present a high
degree of X-ray absorption.
We have selected from the ROSAT North Ecliptic Pole survey
(NEP; Gioia et al. 2003) three QSO 2 candidates on the basis of high X-ray
luminosity and absence of broad optical emission lines. We present in this
paper new optical and X-ray observations of the three NEP QSO 2 candidates
present in the survey catalog and discuss the results obtained. We address
the issue of the density of QSO 2 and their spectral properties in a
completely identified, X-ray-flux limited sample.
Throughout the paper quoted uncertainties are 90% confidence levels for
one interesting parameter. We assume the currently-favored cosmological
concordance model (
,
0.3, 0.7).
The main constituents of the catalog are AGN (49%), either type 1 or
type 2 according to the broadness of their permitted emission lines
with the cut-off at
km s-1 for classification purposes.
There are 198 AGN 1 and 21 AGN 2, where under the term AGN we include
both Seyfert and QSO. We further divide QSO and Seyfert galaxies
according to their X-ray luminosity and put the dividing line at
erg s-1 in the 0.5-2.0 keV energy band. Only
three of a total of 219 AGN are QSO 2 candidates. These objects have narrow
permitted emission lines (MgII 2798 Å and H
with
km s-1) and show high-ionization emission lines like [NeIII] 3869 Å
(see two of the spectra in Figs. 1 (top) and 2).
The 99.6% identification rate implies that QSO 2 are a relatively small
fraction of the total AGN population, at least in the soft X-ray
0.5-2.0 keV energy band. The properties of the three candidates are given
in Table 1. The explanation of each column is given below:
![]() |
Figure 1: NEP 1239: ( top) longslit spectrum of RXJ 1715.4+6239, QSO 2 candidate at z=0.85, obtained with the University of Hawai'i 2.2 m telescope; ( bottom) longslit spectrum of the same object at redder wavelengths taken at the TNG 3.6 m telescope. The dashed lines indicate the positions of the emission lines at the object redshift. Wavelengths of atmospheric absorption (or bad sky subtracion) are also indicated. |
Open with DEXTER |
![]() |
Figure 2: NEP 1640: ( top) longslit spectrum of RXJ 1724.9+6636, QSO 2 candidate at z=0.6758, obtained with the University of Hawai'i 2.2 m telescope; ( center) longslit spectrum of the same object taken at the TNG 3.6 m telescope with the High Resolution grism and ( bottom) the Medium Resolution grism. The dashed lines indicate the positions of the emission lines at the object redshift. Wavelengths of atmospheric absorption (or bad sky subtracion) are also indicated. |
Open with DEXTER |
In this section we present the discovery optical spectra for two of the three QSO 2 candidates obtained at the University of Hawai'i (UH) 2.2 m telescope and the new optical spectroscopy performed at the Italian 3.6 m Telescopio Nazionale Galileo (TNG), plus the X-ray follow-up observations acquired with XMM-Newton. We first describe the optical spectroscopy, and then present the X-ray data.
The optical spectra of the NEP QSO 2 candidates, taken at the UH 2.2 m
telescope of the Mauna Kea Observatory with the Wide Field Grism Spectrograph,
are shown in Figs. 1 and 2 for two of the three QSO 2
candidates. These spectra are the original discovery spectra and have
relatively low signal to noise ratio.
However, the presence of [NeIII] is indicative of high-ionization line
objects. New spectroscopic observations were carried out at the TNG 3.6 m
telescope located at Roque de Los Muchachos Observatory in the Canary islands
on 2004, May 19, and June 19, 22 and 23.
The new spectra were acquired to obtain higher resolution and larger
wavelength coverage in the optical-IR regime in order to confirm that
the permitted lines are narrow, including H
when present in
the observed wavelength range.
We used the low resolution spectrograph DOLORES (Device Optimized
for the LOw RESolution). The Dolores camera was operated at the f/3.2 focus.
The camera is equipped with a Loral thinned and back-illuminated CCD
with
,
15
pixels which provided a derived scale of 0.275 arcsec pixels-1, and a field of view of about
arcmin.
A combination of medium and high resolution spectra were obtained centered in
the blue or red/infrared part of the spectrum, according to the redshift
of each QSO 2 candidate. We used the MR-I grism for NEP 1239, the
highest redshift object at z=0.85 for a total exposure of 2.5 h
divided into three exposures of 3000 s each.
We requested also an infrared spectrum to check for the presence of H
.
This spectrum however was never acquired due to the faintness
of the object which was not visible when the observation was performed.
We may speculate that the object probably varied (as is common for AGN)
at the time of the TNG observations. In fact NEP 1239 is the only object
with a ROSAT PSPC flux (see Tables 1 and 4)
which is about twice the value of the XMM-Newton flux
measured almost simultaneously to
the TNG observations. For NEP 1640 (z=0.6758) the HR-I and
MR-I grisms were used for a total exposure of about 2h divided into 4500 s for the HR-I and 3000 s for the MR-I, while for NEP 2131 (the
lowest redshift object at z=0.358) the MR-B (700 s) and MR-I
(1100 s) grisms were used. The HR-I grism provides a
pixel scale in spectroscopic mode of 0.8 Å pixels-1.
The slitlet width of 1'' provided a spectral resolution of
3 Å FWHM. The wavelength coverage was approximately
7360-8900 Å. The MR-I grism gives a pixel scale of 1.8 Å pixels-1, a spectral resolution of
7 Å FWHM, and a
wavelength coverage of approximately 8400-12 000 Å which in reality
turned out to be only up to 10 000 Å. The MR-B grism provides
a pixel scale of 1.7 Å pixels-1, a spectral resolution
of
6 Å FWHM, and a wavelength coverage of approximately 3500-7000 Å. All the values are for 1
slit used.
![]() |
Figure 3: NEP 2131: longslit spectrum of RX J1737.0+6601, QSO 2 candidate at z=0.358, obtained with the TNG 3.6 m telescope and the Medium Resolution grism in the bluer ( top) and redder ( bottom) wavelength region. The dashed lines indicate the positions of the emission lines at the object redshift. Wavelengths of bad sky subtraction are also indicated. |
Open with DEXTER |
The optical data were analyzed using standard IRAF reduction packages. The redshifts measured
at the TNG 3.6 m for the three objects are in perfect agreement with the
previously determined redshifts from the discovery spectra as can be
seen in Table 1.
The TNG spectra for the three objects are shown in Fig. 1 (bottom),
Fig. 2 (center and bottom) and Fig. 3. The following
properties, listed also in Table 2, can be derived from the
optical spectra of the three objects:
Table 2: Rest-frame optical emission line widths.
Table 3: Journal of XMM- Newton observations.
Table 4: Spectral analysis results.
XMM-Newton (Jansen et al. 2001) observed NEP 1239, NEP 1640 and NEP 2131
as part of the GO program in four epochs, March through June 2004,
with the European Photon Imaging Camera (EPIC) pn
(Struder et al. 2001) and with the EPIC MOS CCD arrays (Turner et al. 2001).
The journal of the observations is reported in Table 3.
The pn and MOS instruments were operating in full-frame mode with the
thin filter (thick for NEP 1640) applied. Event files produced from
the standard pipeline processing have been examined to recognize
high background time intervals (using the version 6.1.0 of the Science
Analysis Software, SAS, and the latest calibration files released by the
EPIC team). Only events corresponding to pattern 0-12 for MOS and pattern 0-4 for pn have been used.
Good time intervals were selected by applying thresholds of 0.35 counts s-1 in the MOS and 1 counts s-1 in the pn to the photons at
energies greater than 10 keV. At these higher energies counts come mostly
from background.
The net exposure times, before data cleaning, are of the order of 24-32 ks
as listed in Table 3. These exposure times reduce to 5-21 ks after
data cleaning thus reducing dramatically the count statistics.
The available statistics after cleaning becomes so poor that we decided
to retain time intervals of background flares. Given the point-like
morphology of the sources the total background count rate under their
extraction regions is still not very high even in the presence of
possible flares. On the other hand the improvement in source statistics
is much higher than the drawback of having a high noise.
In any case, we have performed analysis on both the total and flare-free
datasets and we find results consistent within the errorbars for
all parameters. Net counts and exposure times in seconds, before and
after cleaning (Total and Flare-free), are reported in Table 3.
X-ray source counts were extracted in circular regions centered on the
source with radius 30
(except for
NEP 1640 where a smaller 20
radius was used due to
the presence of other sources close to the target) and background
counts were extracted in a nearby source-free circular region
using a radius of 100
.
For only one source, NEP 1640,
the X-ray observations were performed in two diferent epochs.
The datasets were independently analyzed. Both source and background
spectra have been summed, while response matrices and auxiliary
files are the average of the two exposures, weighted by the different
exposure times. Response matrices at the position of the target source
have been generated using the SAS tasks arfgen and rmfgen.
The two MOS spectra are summed to improve the statistics.
Since the resulting relative normalization between the MOS and pn instruments
is always within 10% we give only fluxes for the MOS.
We used XSPEC (version 11.3.1) for spectral modelling. Spectral bins below 0.5 keV were excluded because of uncertain calibration, and those above 6 (10) keV where excluded for MOS (pn) because of the low statistical
significance. Spectra were binned
to have at least 20 counts and at least a significance of 4
in each
bin. MOS and pn spectra were jointly fitted with independent
relative normalizations to account for possible intercalibration differences
or mismatches. The spectra are well fitted by a single power law model with
low energy absorption fixed at the Galactic value along the line of sight
to each object (Stark et al. 1992; Elvis et al. 1994, see Table 1).
Results are summarized in Table 4. The second column gives the
values of the power low slope
when the
is fixed to the
Galactic values of Table 1, while the third column indicates
the range of additional intrinsic
when it is left free to vary. The following columns
list the
values and degrees of freedom, null hypothesis
probabilities, XMM-Newton fluxes and luminosities in different
energy bands. The X-ray spectra are shown in Figs. 4 through 6.
Taken at face value the X-ray spectra
are not consistent with an obscuration scenario, that is the expected
significant absorption at low energies, typical of QSO 2 objects,
is not seen in any of the three XMM spectra. This is very surprising since
the three objects all have the characteristics of QSO 2, that is narrow
optical emission lines and X-ray luminosities at the 1044 erg s-1
level or higher.
![]() |
Figure 4: X-ray spectra and model ( upper panel) and ratio between the two ( lower panel) for NEP 1239. |
Open with DEXTER |
![]() |
Figure 5: X-ray spectra and model ( upper panel) and ratio between the two ( lower panel) for NEP 1640. |
Open with DEXTER |
The X-ray spectrum for the first object, NEP 1239, is given in
Fig. 4. The spectrum is well fitted by a power law model with
and shows no sign of absorption at low energies.
When left free the total
is <
cm-2 at the quasar redshift.
Its 0.5-2 keV flux is about half the ROSAT NEP flux.
Such variations are not unusual for AGN in the X-ray band (Mushotzky et al. 1993).
The source probably was in a bright state at the time of the ROSAT
observation thus favouring its detection. Even at the lower flux
observed by XMM-Newton NEP 1239 would have made the cut-off
threshold for inclusion in the sample of QSO 2 candidates. The faintness
of the source was probably the reason why the NICS spectrum could not be
taken (see Sect. 3.1). The fainter X-ray continuum should give more contrast
to the presence of a Fe-line, which is however not detected (see below).
The spectrum of NEP 1640 (Fig. 5) is again well fit by a
power law model with
,
with Galactic absorption.
When left free to vary the total
ranges between
cm-2 at the quasar redshift.
![]() |
Figure 6: X-ray spectra and model ( upper panel) and ratio between the two ( lower panel) for NEP 2131. |
Open with DEXTER |
![]() |
Figure 7:
Contour plot of the two free parameters in absori + power
law, ![]() ![]() ![]() |
Open with DEXTER |
For the three sources, if the observed spectrum is due to reflection over a
neutral medium, we would expect a prominent iron K line at around 6.4 keV rest-frame (see e.g. Matt et al. 1996).
No line is evident in any of the three objects
spectrum. However, given the relatively poor statistics, we fitted a
narrow 6.4 keV line to each spectrum and derived the following equivalent
width (EW) upper limits. For NEP 1239 we derive EW < 250 eV,
for NEP 1640 EW < 190 eV, for NEP 2131 EW <240 eV.
To better investigate the
possibile presence of a Fe line we also used the software FLEX
(described in Braito et al. 2005; Maccacaro et al. 2004) which searches for high S/N excesses
in unbinned spectra. Even with this ad-hoc procedure no
excess due to a possible Fe line at the source redshift was found in
any of the three objects.
From the optical point of view the three NEP objects all have
the properties of Seyfert 2 objects both in the UH 2.2 m and in
the higher resolution TNG 3.6 m spectra.
The absence of a broad component in the H
or H
emission
lines implies an extinction of at least a magnitude which for standard
galactic relationship,
/
cm-2 (Predehl & Schmitt 1995),
corresponds to
cm-2.
We cannot set a more stringent limit from optical data
since the broad components of H
and H
are not
detected. However this lower limit is larger than the upper limit derived
from simple modeling of the X-ray data. Within the Unified Model
(Antonucci 1993) we assume that the BLRs in these objects are obscured by
some kind of medium, usually referred to as the "torus''.
The optical light is mainly absorbed by dust, and the X-ray radiation
by gas. The torus contains both media and therefore it absorbs both the
optical and the X-ray emission. This relation between optical and X-ray
absorption is supported by observational evidence (see e.g. Caccianiga et al. 2004),
where optical and X-ray classification for obscuration agree, that is the
optical and X-ray measured absorptions seem to be linked, or at least
present in the same objects.
There are a few known counter-examples (e.g. Page et al. 2003; Brusa et al. 2003; Nandra & Pounds 1994)
in which the X-ray derived absorption is higher than what is seen
in the optical. This situation can have different explanations.
It could be due to a different dust-to-gas ratio, to the presence
of "large'' dust grains (Maiolino et al. 2001), or to the dust being swept
out by the ionizing radiation.
In the three objects under study here
we have a situation where an absorber masks the radiation coming from
an external zone, the BLR, but not the one from the inner zone, the nucleus.
The X-ray spectrum is in fact well fitted by a canonical power law
slope (
)
with no requirement for additional intrinsic
absorption at the source. Only if one assumes a more complex model,
which is not statistically required by the data (as in NEP 2131 where
the model has two additions: the absorption at low energies and an extra
thermal component in the same energy range to account for the residuals),
then a larger value for the intrinsic absorption is required.
These three objects also differ from other objects, like NGC 4698,
which is a peculiar Seyfert 2 with no intrinsic absorption,
possibly the end product of a merger event (see e.g. Bertola et al. 1999),
because they have high X-ray luminosities. Pappa et al. (2001) describe
NGC 4698 as either a dusty warm absorber, or as an object with intrinsic
lack of Broad Line clouds due to its low X-ray luminosity (i.e. lack of
ionizing radiation to illuminate the clouds).
Georgantopoulos & Zezas (2003) recompute the source luminosity to avoid contamination
of other sources and reach the same conclusion of a lack of Broad
Line Regions possibly due to low accretion rate. The three NEP objects
all have luminosities at
erg cm-2 s-1 level
thus making this explanation even less viable.
By using a sample of optically selected Seyfert 2, Panessa & Bassani (2002)
estimate that the fraction of such galaxies which is Compton thin (
cm-2) is of the order of 10-20%. They propose the following
explanations: either the BLRs are covered by a dusty obscuring material
(dust patches, dust lanes, HII regions) or the BLRs are absent, weak or
faded away. This interpretation is supported by the low (
erg s-1) luminosity of the nuclei. In addition the percentage of Compton
thin objects seems to decrease with increasing luminosity, so this
interpretation it's not useful for our case. At higher luminosity,
erg s-1, Caccianiga et al. (2004) report a Sy 1.9 with
no absorption representing, albeit large uncertainties, a fraction in
their sample of 12%, similar to that found by Panessa & Bassani (2002).
The presence of BLR clouds has been linked to the accretion rate
(Nicastro 2000) in such a way that there is an upper limit
to the width of the lines corresponding to a minimum accretion rate dictated
by the conditions of a Shakura & Sunyaev (1973) accretion disk.
Thus low luminosity, low accretion rate nuclei cannot have Broad Emission
Lines. The case for the present QSO2 is however more complex: they might
have a very low accretion rate or, conversely, a very high one that
would give origin to a BLR very far away from the nucleus and
therefore in a region of low velocities that do not widen the lines.
There is always the possibility that the torus in the three NEP objects
is Compton thick. The X-ray component would then be the reflection off the
far side of the torus, that mantains the same slope as the direct component
which is however completely blocked in the observed energy range.
However, the /
ratio
for the three NEP objects has values
consistent with "normal'' type 1 objects, albeit within the calibration
uncertainties of the optical spectra (for the use of the
/
as a diagnostic to find Compton thick objects,
(see Maiolino et al. 1998; Panessa & Bassani 2002).
Furthermore, we would expect in the reflection spectra a prominent line
(
keV) from neutral Fe (Matt et al. 2001). This line is not observed
in any of the three objects. A further obstacle to the Compton
thick interpretation is the high
observed. If this is just a fraction
(0.1-10%) of the total power, the intrinsic luminosity would be very
high (10
45-47 erg s-1). The spatial/volume density
would need to be compared with the density of known AGN, that
decrease accordingly by 2-5 orders of magnitude for increasing luminosity
(see e.g. the XLF in Miyaji et al. 2000).
The NEP QSO2 would represent a higher and higher fraction, even exceeding
by orders of magnitude the density of known AGN.
The most similar object we have found in literature is H1320+551
(Barcons et al. 2003), which is a Seyfert 1.8/1.9. We do not have the same
spectral resolution and S/N as the above authors, so we cannot
exclude that also our candidates have a broad base to the Balmer lines,
but the overall shape is the same. Barcons et al. (2003) suggest that the narrow
line region is internally reddened in H1320+551, but with a small covering
factor over the nuclear emission, and that the Balmer decrement of the BLR is an instrisic property rather than caused by reddening or absorption.
H1320+551 has a weak Fe K detection (EW = 400 eV) too faint
for the source to be Compton-thick. A similar example is
XBSJ031146.1-550702 classified as Sy 1.9 by Caccianiga et al. (2004). Its X-ray
emission does not show significant absorption and is probably Compton-thin.
If one assumes that what is seen in these sources at
X-ray wavelengths is actually the direct nuclear component then one could
still keep the Unified Model paradigm by assuming that the intervening gas
is completely ionized, and therefore transparent to the X-ray radiation,
while the dust in the torus absorbs the optical BLR. This in turn implies
that the accretion is inefficient: moderate accretion rate with low
radiative efficiency in the form of either an Advection Dominated Accretion
Flow (ADAF; see e.g. Narayan & Yi 1995) or a Radiatively Inefficient Accretion
Flow (RIAF; see e.g. Quataert 2003), and therefore at sub-Eddington
regimes. To be sub-Eddingtion
(i.e.
/
)
implies
erg s-1 for these objects and therefore
.
At the moment we cannot estimate the black hole
mass, but this would be a crucial information to acquire to
push the limits of the Unified Models.
The virtually complete identification rate of the ROSAT NEP
survey (99.6%, Gioia et al. 2003) allows us to derive for the first time
an estimate of the spatial density of such sources in soft X-ray
sky. The spatial density of the NEP QSO 2 candidates is
with a density-weighted
average luminosity of
erg s-1.
We can compare it with the most recent estimates in literature
(see discussion in Ueda et al. 2003).
The NEP point is an order of magnitude below any of the different
redshift bin curves in Fig. 15 of Ueda et al. (2003). Therefore this
indicates that this class of soft X-ray selected unobscured QSO2,
if it is indeed a new class, and if the are really unobscured
provides a very small contribution to the total luminosity density.
Since these are most probably objects with no absorption, they
do not make up the missing population of the XRB. This result is not
completely unexpected since the sources have been selected to be bright
in the soft X-ray band of ROSAT, while obscured objects are better
detected in hard X-rays where the reflection component becomes dominant,
and/or the direct component is not absorbed.
Similar objects might be present in current X-ray surveys. For instance,
unobscured QSO 2 candidates might be lurking in hard X-ray surveys like
the ASCA Medium Sensitivity Survey (Akiyama et al. 2003), where the low hydrogen
column density is inferred by a hardness ratio rather than from a more
precise X-ray spectrum. See the large crosses at the bottom of Fig. 10a
in Akiyama et al. (2003) which represent AGN with no broad H
,
low
(log
)
but large X-ray luminosity
(log
). Also the XMM-Newton Bright Serendipitous
Survey (Della Ceca et al. 2004) contains possibile unobscured QSO2 candidates.
A large fraction (more than 60%) of the Narrow Line AGN in this survey
(see Fig. 4a in Della Ceca et al. 2004, and relative discussion) has
an hardness ratio
in the same range as the Broad Line AGN. Only a couple of these sources
might have a luminosity in the QSO range (Caccianiga private communication).
A similar interesting object with narrow optical emission lines, high
X-ray luminosity and no evidence for obscuration in the X-ray spectrum,
namely XLEO J0332-2744, was recently published in Braito et al. (2005).
However, differently from the three NEP objects, XLEO J0332-2744 shows
a strong (
EW = 0.5-1.9 keV) Fe-line which is not present in our X-ray
spectra.
Further observations are needed to strengthen our
findings. We need a better coverage of the H region, we need to
measure the optical extinction directly from the optical continuum shape,
and possibly the central Black Hole mass that is expected
to be in the 10
range. Ultimately higher energy
X-ray observations will say the final word on the presence of a
Compton thick absorber in these objects.
Acknowledgements
The work presented here is based on observations made with the Italian Telescopio Nazionale Galileo (TNG) operated on the island of La Palma by the Centro Galileo Galilei of the INAF (Istituto Nazionale di Astrofisica) at the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias and on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA member states and NASA. A special thank to Valentina Braito for performing the FLEX check. We would like to thank our colleagues Luigi Foschini, Laura Maraschi, Nick Scoville and Paola Severgnini for many useful discussions. We thank an anonymous referee for a number of useful suggestions and comments. I.M.G. thanks the hospitality of the Institute for Astronomy of the University of Hawai'i where part of this paper was written. She also notes that this work was done in spite of the continued efforts by the Italian government to dismantle publicly-funded fundamental research. This work was partially supported by NASA grant NNG05GA70G.