Red AGN in XMM-Newton/SDSS fields

Abstract

In this paper we combine archival and proprietary XMM-Newton observations (about 5deg^2) that overlap with the Sloan Digital Sky Survey to explore the nature of the moderate-z X-ray population. We focus on X-ray sources with optically red colours (g-r>0.4), which we argue are important for understanding the origin of the X-ray background. Firstly, these systems constitute a significant fraction, about 2/3, of the z<1 X-ray population to the limit f(2-8keV)~2e-14cgs. Secondly, their luminosity function under evolution of the form ~(1+z)^3 suggests that they could be responsible for about 17 per cent of the diffuse X-ray background to z=1. Thirdly, their stacked X-ray spectrum in the range 1-8keV is consistent with a power-law distribution with Gamma~1.4 (without fitting intrinsic absorption), i.e. similar to the diffuse X-ray background. We find that the optically red X-ray population comprises a mixed bag of objects, both obscured (N_H>1e22 cm^{-2}) and unobscured (N_H<1e22 cm^{-2}), with a wide range of X-ray luminosities up L_X~1e44cgs. We argue that dilution of the AGN light by the host galaxy may play a role in shaping the continuum optical emission of this population. Finally, we explore a possible association of these sources and the moderate-z red (J-Ks>2mag) AGNs identified in the Two Micron All Sky Survey (2MASS). The median N_H of the red X-ray sources studied here is ~1e21cm^{-2}, lower than that found for the 2MASS AGNs, suggesting different populations. Comment: Accepted by MNRAS

Full text

Page 1

arXiv:astro-ph/0511317v1 10 Nov 2005
Mon. Not. R. Astron. Soc. 000, 000–000 (0000) Printed 5 February 2008 (MN LATEX style file v1.4)
Red Active Galactic Nuclei in XMM-Newton/SDSS fields
A. E. Georgakakis1,2⋆, I. Georgantopoulos1& A. Akylas1
1Institute of Astronomy & Astrophysics, National Observatory of Athens, I. Metaxa & V. Pavlou, Athens, 15236, Greece
2Imperial College of Science Technology and Medicine, Blackett Laboratory, Prince Consort Rd, 2BZ SW7, London, UK
5 February 2008
ABSTRACT
In this paper we combine archival and proprietary XMM-Newton observations (about
5deg2) that overlap with the Sloan Digital Sky Survey to explore the nature of the
moderate-z X-ray population. We focus on X-ray sources with optically red colours
(g − r > 0.4), which we argue are important for understanding the origin of the X-
ray background. Firstly, these systems constitute a significant fraction, about 2/3,
of the z<∼1 X-ray population to the limit f(2 − 8keV) ≈ 2 × 10−14ergs−1cm−2.
Secondly, their luminosity function under evolution of the form ∝ (1 + z)3suggests
that they could be responsible for about 17 per cent of the diffuse X-ray background
to z = 1. Thirdly, their stacked X-ray spectrum in the range 1-8keV is consistent
with a power-law distribution with Γ ≈ 1.4 (without fitting intrinsic absorption),
i.e. similar to the diffuse X-ray background. We find that the optically red X-ray
population comprises a mixed bag of objects, both obscured (NH > 1022cm−2) and
unobscured (NH < 1022cm−2), with a wide range of X-ray luminosities up LX ≈
1044ergs−1. We argue that dilution of the AGN light by the host galaxy may play a
role in shaping the continuum optical emission of this population. Finally, we explore
a possible association of these sources and the moderate-z red (J − Ks > 2mag)
AGNs identified in the Two Micron All Sky Survey (2MASS). The median NHof the
red X-ray sources studied here is ≈ 1021cm−2, lower than that found for the 2MASS
AGNs, suggesting different populations.
Key words: Surveys – X-rays: galaxies – X-rays: general
1INTRODUCTION
The origin of the diffuse X-ray background (XRB) remains
one of the most debated issues of X-ray astronomy. Deep
X-ray surveys with the new generation X-ray missions, the
Chandra and the XMM-Newton have revolutionised this field
demonstrating that the bulk of the XRB can be resolved
into discrete point sources (e.g. Brandt et al. 2001; Gia-
conni et al. 2002). These surveys have confirmed previous
results that luminous (LX>
∼1044ergs−1) broad emission-
line QSOs peaking at redshifts z ≈ 1.5 − 2 are undoubtedly
a major component of the XRB, especially at energies be-
low ≈ 2keV (e.g. Lehmann et al. 2001). These sources alone
however, cannot produce the bulk of the XRB. For example
their steep X-ray spectra (Γ ≈ 1.9) are inconsistent with the
spectral shape of the X-ray background (Γ ≈ 1.4; e.g. Gen-
dreau et al. 1995). An additional population of X-ray sources
is clearly required to account for the XRB properties.
⋆age@imperial.ac.uk
There is in particular, accumulating evidence suggest-
ing that a large fraction of the XRB, particularly at energies
> 2keV may arise at redshifts z<
systems (LX<
∼1044ergs−1). For example, previous missions
at hard energies (> 2keV; HEAO1 A-2, Ginga), although
they lacked imaging capabilities, suggest a statistically sig-
nificant cross-correlation signal between the detected XRB
fluctuations and nearby galaxy catalogues (e.g. ESO, UGC,
IRAS). This implies that about 30 per cent of the XRB could
be produced at low-z (Jahoda et al. 1991; Lahav et al. 1993;
Miyaji et al. 1994; Carrera et al. 1995). In the Chandra and
the XMM-Newton era, a non-negligible fraction of the hard
X-ray population has been identified with optically extended
sources suggesting z<
∼1 galaxies (Koekemoer et al. 2002;
Grogin et al. 2003; Georgantopoulos et al. 2004). This is
further confirmed by spectroscopic follow-up observations,
which show a redshift distribution with a peak at z<
for hard X-ray selected samples (Barger et al. 2002; Rosati
et al. 2002; Fiore et al. 2003; Georgantopoulos et al. 2004;
Georgakakis et al. 2004a). Although the difficulty to spec-
troscopically identify optically faint sources (R > 25mag)
∼1 in moderate luminosity
∼1
c ? 0000 RAS

Page 2

2
Georgakakis, Georgantopoulos & Akylas
may skew the above distribution to lower z (e.g. Treister
et al. 2004), it is nevertheless accepted that these optically
faint systems (about 25 per cent in the Chandra deep fields)
cannot drastically modify the observed distribution.
The evidence above suggests that parallel to the deep
X-ray surveys targeting the high-z Universe, it is important
that we also study the nature of the moderate-z X-ray pop-
ulation. This may indeed hold important clues for the origin
of the XRB and for interpreting deeper X-ray samples.
A large fraction of the z<
∼1 hard X-ray sources in par-
ticular, are associated with optically red galaxies suggesting
either dust reddening (e.g. Seyfert-2s) or continuum emis-
sion dominated by stars rather than the central AGN (Geor-
gantopoulos et al. 2004; Georgakakis et al. 2004a). There
is also evidence that this population has an average X-ray
spectrum similar to that of the X-ray background (Γ ≈ 1.4),
further underlying its significance for XRB studies (Geor-
gantopoulos et al. 2004).
In this paper we explore the nature of the optically red
nearby (z < 1) X-ray sources and their significance in shap-
ing the XRB by combining public XMM-Newton data with
the Sloan Digital Sky Survey (SDSS; York et al. 2000) to
exploit the uniform 5-band optical photometry and spec-
troscopy available in this area. Wide field coverage is es-
sential to probe large enough volume at moderate redshifts
to provide a representative sample of this class of sources.
The XMM-Newton with 4 times the Chandra field-of-view
provides an ideal platform for such a study. Throughout
this paper we adopt Ho = 70kms−1Mpc−1, ΩM = 0.3 and
ΩΛ = 0.7.
2THE X-RAY DATA
In this paper we use 28 XMM-Newton archival observations
selected to overlap with the second data release of the SDSS
(DR2; Stoughton et al. 2002) and with a proprietary pe-
riod that expired before September 2003. A total of 8 of
these fields are part of the XMM-Newton/2dF survey (Geor-
gakakis et al. 2003, 2004b; Georgantopoulos et al. 2004),
while the remaining 20 pointings are presented by Georgan-
topoulos & Georgakakis (2005). The exposure times are in
the range 2-67ks with a median of about 15ks. The Galactic
column density in the direction of these fields varies between
1.3−13×1020cm−2with a median of about 2×1020cm−2.
A full description of the data reduction, the event file gener-
ation and the X-ray image production is presented by Geor-
gantopoulos & Georgakakis (2005).
For this paper the source extraction is performed in
the 2-8keV merged PN+MOS images (when available) us-
ing the ewavelet task of sas with a detection threshold of
6σ. This choice of threshold is to minimise spurious detec-
tions in the final catalogue. The extracted sources for each
field were visually inspected and spurious detections clearly
associated with CCD gaps, hot pixels or lying close to the
edge of the field of view were removed. We further exclude
from the final catalogue the target of a given XMM-Newton
pointing (e.g. nearby galaxies or clusters) and a total of 7
sources that appear extended on the XMM-Newton EPIC
images and are clearly associated with diffuse cluster emis-
sion. Fluxes are estimated using an 18arcsec radius aper-
ture corresponding to an encircled energy fraction of about
Figure 1. Solid angle as a function of 2-8keV flux (6σ) for our
survey.
70 per cent at 1.5keV. For the spectral energy distribution
we adopt a power-law with Γ = 1.8 and Galactic column
appropriate for each field. Using Γ = 1.4 instead of Γ = 1.8,
typical of the mean spectrum of the XRB and the red AGN
studied here (see section 4.2), has a minimal effect on the
estimated fluxes (about 8 per cent) and does not affect the
results presented here. For the background estimation we
use the background maps generated as a by-product of the
ewavelet task of sas. The final hard X-ray selected sam-
ple comprises a total of 507 sources above the 6σ flux limit
fX(2 − 8keV) = 5 × 10−15ergs−1cm−2. The area curve
giving the cumulative area of our survey as a function of
limiting flux is shown in Figure 1.
3THE SAMPLE
3.1Selection criteria
We use the SDSS DR2 catalogue to optically identify the
hard X-ray selected sources following the method of Downes
et al. (1986), as described in Georgakakis et al. (2004b) to
calculate the probability, P, a given candidate is a spurious
identification. Here we apply an upper limit in the search
radius, r < 7arcsec and a cutoff in the probability, P < 0.05,
to limit the optical identifications to those candidates that
are least likely to be spurious alignments.
In this study we use 6σ hard X-ray selected sources with
2-8keV flux fX(2−8keV) > 2×10−14ergs−1cm−2. This is
to ensure sufficient photon-statistics to perform X-ray spec-
tral analysis for most sources and to minimise the number
of objects without optical identifications. Out of 230 sources
above the 6σ detection threshold and fX(2 − 8keV) >
2 × 10−14ergs−1cm−2a total of 189 have optical coun-
terparts to the SDSS limit (82 per cent completeness).
The choice of the X-ray flux limit above is a trade-off be-
tween maximum optical identification completeness and suf-
ficiently large sample size to avoid small number statistics.
c ? 0000 RAS, MNRAS 000, 000–000

Page 3

Red AGN in XMM-Newton/SDSS fields
3
From the sample above we select sources with red op-
tical colours, g − r > 0.4mag, to exclude bluer broad-line
QSOs. Optically red sources have been identified in non-
negligible numbers in recent surveys (e.g. Koekemoer et al.
2002) and may play a key role in shaping the diffuse X-ray
background. Indeed, this population may be associated with
obscured AGNs and/or the population of X-ray Bright Opti-
cally Normal Galaxies (XBONGs; Comastri et al. 2002) that
are also suggested to harbor deeply buried AGNs (but see
Georgantopoulos & Georgakakis 2005). Figure 2 plots the
optical colours of the hard X-ray selected sample with the
expected tracks for different galaxy types (E/S0, Sbc, Scd,
Im; Coleman, Wu & Weedman 1980) and optically selected
QSOs (Cristiani et al. 2004). In this figure it is clear that
the colour cut g − r > 0.4mag selects against typical broad
line QSOs providing a sample that is dominated by sources
with galaxy-like colours. Nevertheless, high-z QSO (z > 2;
e.g. Richards et al. 2002; Kitsionas et al. 2005) or reddened
QSOs (e.g. Wilkes et al. 2002; White et al. 2003) with SEDs
different to those of optically selected ones may still exist
within our sample. Such sources are expected to be identi-
fied by their unresolved (e.g. point-like) optical light profile.
We also caution the reader that the colour cut g − r = 0.4
also eliminates from the sample low- (z<
(z>
∼0.8) systems with SEDs similar to irregular (Im) type
galaxies.
A total of 102 hard X-ray selected sources fulfil the cri-
teria fX(2−8keV) > 2×10−14ergs−1cm−2and g−r > 0.4.
Figure 2 shows that most of them (total of 84) have extended
optical light profile suggesting moderate-z (<
where the central AGN does not dominate the optical light
profile. The 18 sources with unresolved optical light profile
are associated with either high-z QSOs or Galactic stars. We
note however, that the SDSS star-galaxy separation is reli-
able at the 95% confidence limit to r = 21mag and becomes
less robust at fainter magnitudes. A total of 25 out of 102
sources in the sample are fainter that this magnitude limit.
Also there are only 29 [≈ 25 per cent, 29/(29+84)] optically
extended sources with colours bluer than g −r = 0.4 within
the fX(2−8keV) > 2×10−14ergs−1cm−2subsample. The
red subpopulation, comprising 75 per cent of the optically
extended X-ray sources, clearly dominates at moderate-z.
∼0.2) or high-z
∼1) systems
3.2 Redshift estimation
Spectroscopic redshifts are available for 24 hard X-ray se-
lected sources with g −r > 0.4 from either the SDSS or our
own follow-up campaign of the XMM-Newton/2dF survey.
A total of 22 of the 24 sources are optically extended and are
assigned moderate redshifts (z < 1). The remaining two X-
ray sources are optically unresolved and are identified with
a Galactic star and a z = 1.335 broad-line QSO (Georgan-
topoulos et al. 2004). For the spectroscopically unidentified
sources in the sample we use photometric methods to esti-
mate redshifts.
In particular for systems with extended optical light
profile we use the SDSS-DR2 photometric redshifts that are
based on galaxy templates (Csabai et al. 2002). As a byprod-
uct of the photometric redshift method each source is as-
signed a best-fit SED which is a continuous parameter be-
tween 0 and 1. The two extreme values 0 and 1 correspond
to ellipticals and actively starforming (irregulars) systems
Figure 2. r − i against g − r colour for optically identified 6σ
sources with fX(2−8keV) > 2×10−14ergs−1cm−2. Filled circles
and crosses are for optically extended and point-like X-ray sources
respectively. The expected colours of different galaxy types (E/S0,
Sbc, Scd, Im) using the template SEDs from Coleman, Wu &
Weedman (1980) are also shown for redshifts z = 0 − 1. Also
plotted is a QSO template SED for z = 0 − 3 (Cristiani et al.
2004). Our colour selection g −r > 0.4 is shown with the vertical
dotted line.
respectively. Galaxies with intermediate best fit SEDs are
assigned photometric types between 0 and 1.
Previous studies on photometric redshifts of X-ray
sources suggest that galaxy templates work well for sys-
tems with red optical colours, similar to those studied here
(Barger et al. 2002, 2003; Mobasher et al. 2004; Georgakakis
et al. 2004a; Kitsionas et al. 2005). The success rate using
galaxy templates for these sources is demonstrated in Fig-
ure 3 which compares the photometric and spectroscopic
redshift estimates for the optically extended objects with
available spectroscopic observations. The agreement is good
with |(zphot−zspec)|/(1+zspec) ≈ 0.06. A number of sources
significantly deviate (δz > 0.1) from the zspec = zzphot re-
lation in Figure 3. All of them show broad emission-lines
and it is likely that AGN light is contributing to their op-
tical continuum. Figure 3 also shows there is a fair agree-
ment between the observed spectral classification (emission,
absorption line) and the galaxy type of the best fit SED
(early, late types). For absorption-line objects 4 out of 5 are
best-fit by early-type SEDs (< 0.2). For narrow emission-
line galaxies and broad-line systems 5/7 and 9/9 are best-fit
by late type SEDs respectively. One spectroscopically iden-
tified source has no spectral classification in Figure 3 (see
below for details).
A small number of 4 optically faint (r ≈ 22mag; see Ta-
ble 1) galaxies in our sample are assigned unrealistically low
photometric redshifts, z ≈ 0.001, by the SDSS algorithm.
For these systems we use the relation between X-ray lumi-
c ? 0000 RAS, MNRAS 000, 000–000

Page 4

4
Georgakakis, Georgantopoulos & Akylas
Figure 3. Photometric against spectroscopic redshift estimates
for the hard X-ray selected sources with available spectroscopic
observations. Open circles are systems with absorption line spec-
tra, open squares correspond to sources with narrow-emission line
spectra and stars are broad line AGNs. The source in our sample
(#84 in Table 1) without spectral classification is shown with the
filled triangle. A cross on top of a symbol indicates a late-type best
fit SED (photometric type >0.2) to the photometric data, while
small dots are for early type SEDs (photometric type <0.2).The
dashed lines are the δz = ±1 envelope around the zspec = zphot
continuous line.
nosity and X-ray–to–optical flux ratio (see next section for
the definition of logfX/fopt) suggested by Fiore et al. (2003)
to get a rough redshift estimate. Figure 4 plots this relation
for the optically red sources in our sample with spectroscopic
redshift available. Using this relation we estimate z ≈ 1 − 2
for the 4 optically faint sources. These systems are marked
in Table 1. The redshift uncertainty using the 1σ rms scatter
around the best-fit relation in Figure 4 is estimated to be
δz/(1 + z) ≈ 0.2.
Figure 5 plots the redshift and spectral type distribu-
tion of the optically extended subsample. About half of the
sources lie at z<
∼0.4 and have best fit spectral types < 0.4.
A total of 18 X-ray sources with red colours have point-
like optical light profile and are most likely associated with
either Galactic stars, or high-z QSOs. Using the photometric
methods described by Kitsionas et al. (2005) we find that
5 of these sources are best fit by stellar templates with the
remaining assigned photometric redshifts z > 1.3. These 18
sources will not be considered in the rest of the paper.
We also exclude from the analysis source #47 in Table
1 at z = 0.005 with LX ≈ 2 × 1039ergs−1. This system
has X-ray–to–optical flux ratio < −2 (see Figure 6 below)
and has been classified ‘normal’ galaxy by Georgantopou-
los et al. (2005) with X-ray emission associated with stellar
processes. For completeness this is included in Table 1 (see
below). The final sample used in the analysis comprises a
total of 83 systems with g − r > 0.4 and extended optical
light profile. For luminosity estimates photometric redshifts
are used unless spectroscopic redshifts are available
Figure 4. 2–8keV X-ray luminosity against X-ray–to–optical flux
ratio for red X-ray sources with available optical spectroscopy.
Open circles are systems with absorption line spectra, open
squares correspond to sources with narrow-emission line spectra
and stars are broad line AGNs. The source in our sample (#84
in Table 1) without spectral classification is shown with the filled
triangle. The continuous line is the best-fit relation to the data.
The dashed line is the best-fit relation of Fiore et al. (2003) for
their sample of non type-I AGNs.
Figure 5. Redshift (photometric or spectroscopic) and photomet-
ric type distribution for the red optically extended X-ray sources.
3.3X-ray spectral analysis
We explore the X-ray spectral properties of our sample us-
ing the xspec v11.2 package. For sources with small num-
ber of net counts we use the C-statistic technique (Cash
1979) specifically developed to extract information from low
signal-to-noise ratio spectra. The data are grouped to have
at least one count per bin. We adopt an absorbed power-
c ? 0000 RAS, MNRAS 000, 000–000

Page 5

Red AGN in XMM-Newton/SDSS fields
5
law model (wabs*pow) to constrain the absorbing column
density NH by fixing the power law index to Γ = 1.8, i.e.
in-between radio loud and radio quiet AGNs. For sources
with sufficient net counts (about 150) we perform standard
χ2spectral fitting. The data are grouped to have a mini-
mum of 15 counts per bin to ensure that Gaussian statistics
apply and we require that the spectrum has at least 10 spec-
tral bins. In the case of χ2analysis an absorbed power-law
(wabs*pow) is fit to the data yielding both the NH and
the power-law photon index Γ. For both the χ2and the
C-statistic analysis the fit was performed in the 0.2-8keV
energy range where the sensitivity of the XMM-Newton is
the highest. The estimated errors correspond to the 90 per
cent confidence level.
The column densities estimated above are in the ob-
server’s frame and therefore lower than the rest-frame NH
because the k-effect shifts the absorption turnover to lower
energies. The relation between intrinsic and observed col-
umn density scales approximately as (1+z)2.65(e.g. Barger
et al. 2002). This redshift correction is applied to all sources,
after subtracting from the observed NH the appropriate
Galactic column in the direction of the source. The rest-
frame NH is then used to estimate the unabsorbed X-ray
luminosities.
The optical and X-ray properties of the hard X-ray
selected sample with g − r > 0.4 and fX(2 − 8keV) >
2 × 10−14ergs−1cm−2used in this paper is presented in
Table 1 which has the following format:
1. Identification number.
2-3. Right ascension and declination of the X-ray cen-
troid position in J2000.
4. SDSS r-band magnitude of the optical counterpart.
5. Probability, P, the optical counterpart is a chance
coincidence.
6. 2-8keV X-ray flux in ergs−1cm−2.
7. X-ray–to–optical flux ratio, logfX/fopt, estimated
from the relation
logfX
fopt
= logfX(2 − 8keV) + 0.4r + 5.60.(1)
The equation above is derived from the X-ray–to–optical
flux ratio definition of Stocke et al. (1991) that involved 0.3-
3.5keV flux and V -band magnitude. These quantities are
converted to 2-8keV flux and r-band magnitude adopting a
mean colour V − R = 0.7, the transformations of Fukugita
et al. (1996) and a power-law X-ray spectral energy distri-
bution with index Γ = 1.8.
8. Spectroscopic redshift measurement if available.
9. Spectroscopic classification. AB is for absorption line
systems, NL is for narrow emission-line sources and BL sig-
nifies broad emission-line optical spectra. Source #84 has a
spectroscopic redshift estimate from the NED but no spec-
tral classification is available.
10. Photometric redshift estimate.
11. 2-8keV X-ray luminosity in units of ergs−1, cor-
rected for X-ray absorption.
12. Rest-frame column density, NH, estimated from the
spectral fitting analysis in units of 1022cm−2.
13. Power-law spectral index, Γ, for those sources with
sufficient net counts (about 150) to perform χ2analysis. The
median Γ for these sources is ≈ 1.8.
Figure 6. r-band magnitude against 2-8keV flux for the opti-
cally extended g − r > 0.4 sources. The lines logfX/fopt = ±1
delineate the region of the parameter space occupied by powerful
unobscured AGNs.
Figure 7. NH versus photometric redshift spectral type. Ellip-
tical SEDs correspond to spectral type=0 while irregulars are
assigned type=1. Upper limits in the hydrogen column density
NH are shown with an arrow. A cross on top of a symbol is for
sources with broad-line optical spectra. Similarly, open squares
and open circles on top of a dot correspond to sources with nar-
row emission-line and absorption optical spectra respectively.
c ? 0000 RAS, MNRAS 000, 000–000