XMM-Newton Observations of a Complete Sample of Optically Selected Type 2 Seyfert Galaxies

Abstract

The majority of active galactic nuclei (AGNs) suffer from significant obscuration by surrounding dust and gas. The penetrating power and sensitivity of hard X-ray observations allow obscured AGNs to be probed out to high redshifts. However, X-ray surveys in the 2-10 keV band will miss the most heavily obscured AGNs in which the absorbing column density exceeds ~1024 cm–2 (the Compton-thick AGN). It is, therefore, vital to know the fraction of AGNs that are missed in such X-ray surveys and to determine if these AGNs represent some distinct population in terms of the fundamental properties of AGNs and/or their host galaxies. In this paper, we present the analysis of XMM-Newton X-ray data for a complete sample of 17 low-redshift Type 2 Seyfert galaxies chosen from the Sloan Digital Sky Survey based solely on the high observed flux of the [O III]λ5007 emission line. This line is formed in the narrow-line region hundreds of parsecs away from the central engine. Thus, unlike the X-ray emission, it is not affected by obscuration due to the torus surrounding the black hole. It therefore provides a useful isotropic indicator of the AGN luminosity. As additional indicators of the intrinsic AGN luminosity, we use the Spitzer Space Telescope to measure the luminosities of the mid-infrared continuum and the [O IV] 25.89 μm narrow emission line. We then use the ratio of the 2-10 keV X-ray luminosity to the [O III], [O IV], and mid-infrared luminosities to assess the amount of X-ray obscuration and to distinguish between Compton-thick and Compton-thin objects. The various diagnostics of AGN luminosity with heavily obscured hard X-ray emission are in broad agreement. We find that the majority of the sources suffer significant amounts of obscuration: the observed 2-10 keV emission is depressed by more than an order of magnitude in 11 of the 17 cases (as expected for Compton-thick sources). Thus, surveys in the rest-frame 2-10 keV band will be significantly incomplete for obscured AGNs. We find a strong inverse correlation between the ratio of the 2-10 keV X-ray and [O III] (or [O IV] or mid-IR) fluxes and the equivalent width of the 6.4 keV Fe Kα line. This demonstrates that the weak hard X-ray continuum emission is due to obscuration (rather than due to intrinsically weak emission). In several cases, the large amount of obscuration is not consistent with the values of absorbing column density derived from simple spectral fits to the data. We run simulations of a more physically realistic model with partial covering of the X-ray source plus Compton scattering, and show that such models are consistent with the data. We show that the distribution of obscuration in the 2-10 keV band in our sample is similar to what is seen in samples selected in the Swift BAT energy band (14-195 keV). This implies that the BAT surveys do recover a significant fraction of the local population of Compton-thick AGNs. Finally, we find no relationship between the amount of X-ray obscuration and the other properties of the AGN and its host galaxy. Hence, Compton-thick and Compton-thin sources do not seem to trace distinct populations.

Full text

Page 1

arXiv:0909.3044v1 [astro-ph.CO] 16 Sep 2009
XMM-Newton Observations of a Complete Sample of Optically
Selected Type 2 Seyfert Galaxies
Stephanie M. LaMassa1, Timothy M. Heckman1, Andrew Ptak1, Ann Hornschemeier2,
Lucimara Martins3, Paule Sonnentrucker1, and Christy Tremonti4
1The Johns Hopkins University,2NASA Goddard Space Flight Center,3NAT -
Universidade Cruzeiro do Sul,4Max-Planck Institute for Astronomy
ABSTRACT
The majority of Active Galactic Nuclei (AGN) suffer from significant obscu-
ration by surrounding dust and gas. The penetrating power and sensitivity of
hard X-ray observations allows obscured AGN to be probed out to high redshifts.
However, X-ray surveys in the 2-10 keV band will miss the most heavily-obscured
AGN in which the absorbing column density exceeds ∼ 1024cm−2(the Compton-
thick AGN). It is therefore vital to know the fraction of AGN that are missed in
such X-rays surveys and to determine if these AGN represent some distinct popu-
lation in terms of the fundamental properties of AGN and/or their host galaxies.
In this paper we present the analysis of XMM-Newton X-ray data for a com-
plete sample of 17 low-redshift Type 2 Seyfert galaxies chosen from the Sloan
Digital Sky Survey based solely on the high observed flux of the [OIII]λ5007
emission-line. This line is formed in the Narrow Line Region hundreds of par-
secs away from the central engine. Thus, unlike the X-ray emission, it is not
affected by obscuration due to the torus surrounding the black hole. It therefore
provides a useful isotropic indicator of the AGN luminosity. As additional in-
dicators of the intrinsic AGN luminosity, we use the Spitzer Space Telescope to
measure the luminosities of the mid-infrared continuum and the [OIV]25.89µm
narrow emission-line. We then use the ratio of the 2-10 keV X-ray luminosity
to the [OIII], [OIV], and mid-infrared luminosities to assess the amount of X-
ray obscuration and to distinguish between Compton-thick and Compton-thin
objects. The various diagnostics of AGN luminosity with heavily obscured hard
X-ray emission are in broad agreement. We find that the majority of the sources
suffer significant amounts of obscuration: the observed 2-10 keV emission is de-
pressed by more than an order-of-magnitude in 11 of the 17 cases (as expected for
Compton-thick sources). Thus, surveys in the rest-frame 2-10 keV band will be
significantly incomplete for obscured AGN. We find a strong inverse correlation
between the ratio of the 2-10 keV X-ray and [OIII] (or [OIV] or mid-IR) fluxes

Page 2

– 2 –
and the equivalent width of the 6.4 keV Fe Kα line. This demonstrates that
the weak hard X-ray continuum emission is due to obscuration (rather than to
intrinsically weak emission). In several cases the large amount of obscuration is
not consistent with the values of absorbing column density derived from simple
spectral fits to the data. We run simulations of a more physically realistic model
with partial covering of the X-ray source plus Compton scattering, and show that
such models are consistent with the data. We show that the distribution of ob-
scuration in the 2-10 keV band in our sample is similar to what is seen in samples
selected in the Swift BAT energy-band (14-195 keV). This implies that the BAT
surveys do recover a significant fraction of the local population of Compton-thick
AGN. Finally, we find no relationship between the amount of X-ray obscuration
and the other properties of the AGN and its host galaxy. Hence, Compton-thick
and Compton-thin sources do not seem to trace distinct populations.
Subject headings: galaxies: Seyfert - X-rays: galaxies
1.Introduction
In the standard model (e.g. Urry & Padovani 1995), Active Galactic Nuclei (AGN) con-
sist of an accretion disk around a supermassive black hole (SMBH) surrounded by a torus of
dust and gas. The accretion disk is a source of ultraviolet and optical continuum emission,
while X-rays are believed to originate through inverse Compton scattering of soft seed pho-
tons from the disk by hot electrons in a surrounding corona. Dust in the torus absorbs the
emission from the accretion disk and re-emits it in the mid-infrared. Type 1 (unobscured)
and Type 2 (obscured) AGN are intrinsically the same, but they display different charac-
teristics because of their orientation. In Type 2 AGN the torus is viewed nearer edge-on,
obscuring the central source, whereas in Type 1 AGN, the torus is viewed nearer face-on,
providing an unobscured view of the central source (Antonucci 1993). Obscured AGN have
long been identified by their narrow high-ionization emission lines in the optical band, pro-
duced by photoionized gas far above and below the plane of the torus. Obscured AGN can
also be recognized by the mid-infrared emission from the torus. Optical and infrared surveys
show that obscured AGN constitute the majority of the population (e.g. Risaliti et al. 1999;
Hao et al. 2005).
Hard X-rays (E > 2 keV) also offer a potentially powerful way to find obscured AGN.
So long as the obscuring column density (NH) is < 1024cm−2(the Compton-thin regime),
hard X-rays will suffer little absorption. However, observations of a local (<0.035) sample of
49 Type 2 AGN show that in fact many of these are very weak in the 2-10 keV X-ray band

Page 3

– 3 –
(Guainazzi et al. 2005). In the Chandra Deep Field South, only 82 out of over 300 AGN
with optical spectroscopy were “X-ray bright” (>80 counts in the 2-10 keV band). Though
this sample includes AGN up to much higher redshifts (up to z ∼ 3), the subsamples of
X-ray bright and X-ray faint targets have similar redshift distributions, indicating that the
X-ray faint nature of the sample is not due to selection effects but rather that the majority
of targets are weak X-ray emitters (Tozzi et al. 2006). The cases where the X-ray emission is
severely attenuated are interpreted as cases in which the torus is Compton-thick (NH> 1024
cm−2). More recent surveys with INTEGRAL and Swift suggest that a at least some low
redshift (z < 0.05) Type 2 AGN remain weak emitters even at substantially higher photon
energies (> 10 keV, e.g. Treister et al. 2009 for z < 0.05, but see Mel´ endez et al. 2008).
This would require even higher obscuring column densities (> 1025cm2).
How do we know that the weakness of the hard X-ray emission in these Type 2 AGN
is due to obscuration rather than an intrinsic property (e.g. Brightman & Nandra 2008)?
Fortunately, there are signatures of obscuration in the X-ray spectra themselves. The weak
X-ray emission that is seen in such cases often has a relatively flat X-ray spectrum and a
strong Fe Kα emission-line. These features are indicative of heavily reprocessed emission
such as from reflection of hard X-rays off of the torus wall. This is as expected for a highly
obscured X-ray source.
The weakness of the hard X-ray emission in many Type 2 AGN (regardless of its origin)
implies that hard X-ray surveys do not fully probe the AGN population. This is important
from several perspectives. First, phenomenological models often invoke a significant popu-
lation of obscured AGN in order to successfully match the spectral energy distribution of
the cosmic X-ray background at energies above 10 keV (Comastri 2004). Are such models
consistent with the observed properties of obscured AGN? Second, deep X-ray surveys can
detect AGN out to high redshift. They then allow us to probe the cosmic evolution of the
AGN population down to luminosities and SMBH masses well below the levels reached in
optical surveys of quasars. Are the implications drawn from such surveys affected by a bias
against Compton-thick AGN?
To address these issues, it is important to study a complete sample of AGN selected
without regard to their hard X-ray properties. Optical spectroscopic surveys can in principle
provide such a sample. Type 2 AGN in these surveys are identified by the high-excitation
narrow emission from photoionized gas in the Narrow Line Region (NLR), material located
hundreds of parsecs out along the polar axis of the torus. This emission is therefore un-
affected by the obscuration of the torus itself and to first-order is radiated isotropically.
Comparison of the luminosity of the gas in the NLR to that in hard X-rays can then be used
to estimate the amount of X-ray obscuration. The most commonly used optical emission-line

Page 4

– 4 –
is [OIII]λ5007 (e.g. Bassani et al. 1999; Heckman et al. 2005). Several indicators of the
intrinsic AGN luminosity also exist in the mid-infrared band where the effects of obscura-
tion will be minimal. A close analog to the [OIII]λ5007 line is the luminosity of the [OIV]
25.89µm line from the NLR (e.g. Mel´ endez et al. 2008, Diamond-Stanic et al. 2009, Rigby
et al. 2009). The luminosity of the mid-infrared continuum from the torus is also a useful
indicator, although the torus is not believed to emit in a fully isotropic fashion (e.g. Pier &
Krolik 1993; Nenkova et al. 2008).
In this paper we present X-ray analysis of a nearly complete sample of 17 Type 2
Seyfert galaxies (Sy2) selected from the Sloan Digital Sky Survey (SDSS) based on their
high [OIII]λ5007 flux. Using pointed XMM-Newton observations of these sources, we derive
their X-ray spectral characteristics and calculate their hard X-ray luminosities. We then use
the SDSS data to measure the luminosities of [OIII]λ5007 lines and Spitzer Space Telescope
Infrared Spectrograph (IRS) data to measure the luminosities of the mid-infrared continuum
and of the [OIV] 25.89µ line. We use these data to assess the amount of X-ray obscuration.
We also use measurements of the Fe Kα line as an additional tool for probing highly obscured
X-ray emission. Finally, we use the SDSS and Spitzer data to derive the basic intrinsic
properties of the AGNs and their host galaxies. This allows us to determine whether the
Compton-thick and Compton-thin populations differ in any systematic way.
2.XMM-Newton Observations and Data Reduction
2.1.Sample Selection
Our Type 2 AGN are drawn from a parent sample of the spectra of roughly 480,000
galaxies in the SDSS Data Release 4 = DR4 (Adelman-McCarthy et al. 2006). This par-
ent sample of “SDSS Main Galaxies” (Strauss et al. 2002) is complete over the SDSS DR4
footprint and is selected solely on the basis of the galaxy r-band apparent magnitude (14.5
< r < 17.77). The diagnostic line ratio plot of [NII]/Hα vs. [OIII]/Hβ in Kauffmann et al.
(2003 hereafter K03) and Kewley et al. (2006) was used to identify Type 2 AGN based on
the SDSS spectra. Our sample was then defined as the 20 such objects having an observed
[OIII]λ5007 flux above 4×10−14erg cm−2s−1, corresponding to a luminosity of ∼10
at the median distance for our sample (∼150 Mpc). XMM-Newton data exist for 17 of the
20 objects in our sample, so the present sample is nearly complete.
41erg/s
We note that our sample is highly complementary to the sample of “Type II Quasars”
investigated by Zakamska et al. (2003; 2008) and Ptak et al. (2006). These objects were
selected on the basis of a high [OIII]λ5007 luminosity (rather than on the basis of a flux

Page 5

– 5 –
limit as we have done). The Type II Quasars were drawn from the full set of SDSS spectra,
rather than from the homogeneous SDSS Main Galaxy sample. These objects define the
extreme high luminosity end of the Type 2 AGN population, while our sample deals with
typical Type 2 AGN.
2.2. Data Analysis
We were awarded XMM-Newton time to observe 15 members of the 20 from this sample
for a nominal 23 ks per target before filtering. Two additional targets from this sample
were previously observed with XMM-Newton and were added: 0325-0608 (Mrk 609, PI:
Aschenbach) and 1218+4706 (PI: Page), bringing the sample total to 17. Table 1 lists the
targets along with their ObsID number. The luminosity distances listed in Table 1 are based
on the optical spectroscopic redshift, adopting a cosmology of Ho= 71 km s−1Mpc−1, ΩM
= 0.27 and ΩΛ= 0.73.
The XMM-Newton data were reduced using XAssist (Ptak & Griffiths 2003), which
runs the Science Analysis System (SAS) tasks to filter the raw data, generate light curves,
clean the data for flaring, and extract spectra and associated response files (rmfs and arfs)
for user-defined sources. Table 1 lists the good exposure time for each detector. For three
sources, 1018+3722, 1111+0228, and 1437+3634, flaring was significant, so we applied min-
imal background filtering. The net X-ray counts, soft counts (0.5 - 2 keV) and hard counts
(2 - 10 keV) are listed in Table 2.
For most of the sources, there were sufficient counts to group the PN, MOS1, and MOS2
data by >5 counts/bin (and in a majority of these cases, >10 counts/bin) and we used the
χ2statistic for the spectral analysis. The following targets were grouped by 2-3 counts/bin
and were therefore analyzed using the C-statistic (XSpec handles spectral fitting via C-stat
better when the data are slightly binned rather than unbinned, e.g. Teng et al 2005): 0053-
0846, 1123+4703, and 1346+6423. One target, 1218+4706, fell on the chip gap in the PN
detector, so we analyzed only the MOS1 and MOS2 data for this source.
The PN, MOS1 and MOS2 data were fit simultaneously for each source in XSpec,
with all parameters tied together except for a constant multiplicative factor, which was
frozen at 1 for the PN spectra and allowed to vary between 0.8 to 1.2 for MOS1 and
MOS2, as the responses among detectors should generally not vary by greater than ∼20%
(http://xmm.vilspa.esa.es/docs/documents/CAL-TN-0052.ps.gz); though in some cases, these
limits were extended slightly to better fit the data when necessary (i.e. when the error on
the statistical count rate is >20%). We first fit all targets with a simple absorbed power law.