Host galaxy colour gradients and accretion disc obscuration in AEGIS z~1 X-ray-selected active galactic nuclei

Abstract

We describe the effect of AGN light on host galaxy optical and UV-optical colours, as determined from X-ray-selected AGN host galaxies at z~1, and compare the AGN host galaxy colours to those of a control sample matched to the AGN sample in both redshift and stellar mass. We identify as X-ray-selected AGNs 8.7 +4/-3 per cent of the red-sequence control galaxies, 9.8 +/-3 per cent of the blue-cloud control galaxies, and 14.7 +4/-3 per cent of the green-valley control galaxies. The nuclear colours of AGN hosts are generally bluer than their outer colours, while the control galaxies exhibit redder nuclei. AGNs in blue-cloud host galaxies experience less X-ray obscuration, while AGNs in red-sequence hosts have more, which is the reverse of what is expected from general considerations of the interstellar medium. Outer and integrated colours of AGN hosts generally agree with the control galaxies, regardless of X-ray obscuration, but the nuclear colours of unobscured AGNs are typically much bluer, especially for X-ray luminous objects. Visible point sources are seen in many of these, indicating that the nuclear colours have been contaminated by AGN light and that obscuration of the X-ray radiation and visible light are therefore highly correlated. Red AGN hosts are typically slightly bluer than red-sequence control galaxies, which suggests that their stellar populations are slightly younger. We compare these colour data to current models of AGN formation. The unexpected trend of less X-ray obscuration in blue-cloud galaxies and more in red-sequence galaxies is problematic for all AGN feedback models, in which gas and dust is thought to be removed as star formation shuts down. [See paper for full abstract.] Comment: Accepted to MNRAS. 19 pages, 14 figures, 1 table; table, four figures (4, 6, 11, 13) revised to reflect corrected values for one of our objects; results unchanged

Full text

Page 1

arXiv:1006.3571v2 [astro-ph.CO] 28 Jun 2010
Mon. Not. R. Astron. Soc. 000, 1–19 (2010) Printed 29 June 2010(MN LATEX style file v2.2)
Host galaxy colour gradients and accretion disc obscuration
in AEGIS z ∼ 1 X-ray-selected active galactic nuclei
C. M. Pierce,1,2⋆J. M. Lotz,3† S. Salim,3E. S. Laird,4A. L. Coil,5K. Bundy,6
C. N. A. Willmer,7D. J. V. Rosario,8J. R. Primack9and S. M. Faber8
1Department of Physics, University of California, Santa Cruz, 1156 High Street, Santa Cruz, CA 95064, USA
2School of Physics, Georgia Institute of Technology, 837 State Street, Atlanta, GA 30332-0430, USA
3National Optical Astronomical Observatories, 950 N. Cherry Avenue, Tucson, AZ 85719, USA
4Astrophysics Group, Imperial College London, Blackett Laboratory, Prince Consort Rd., London SW7 2AW, UK
5Department of Physics, University of California, San Diego, CA 92093, USA
6Department of Astronomy, University of California, Berkeley, CA 94720, USA
7Steward Observatory, University of Arizona, 933 North Cherry Avenue, Tucson, AZ 85721, USA
8UCO/Lick Observatory; Department of Astronomy and Astrophysics, University of California, Santa Cruz, 1156 High Street,
Santa Cruz, CA 95064, USA
9Santa Cruz Institute of Particle Physics, University of California, Santa Cruz, 1156 High Street, Santa Cruz, CA 95064, USA
Accepted 2010 June 3. Received 2010 June 2; in original form 2010 January 20
ABSTRACT
We describe the effect of AGN light on host galaxy optical and UV-optical colours,
as determined from X-ray-selected AGN host galaxies at z ∼ 1, and compare the
AGN host galaxy colours to those of a control sample matched to the AGN sample in
both redshift and stellar mass. We identify as X-ray-selected AGNs 8.7+4
the red-sequence control galaxies, 9.8 ± 3 per cent of the blue-cloud control galaxies,
and 14.7+4
−3per cent of the green-valley control galaxies. The nuclear colours of AGN
hosts are generally bluer than their outer colours, while the control galaxies exhibit
redder nuclei. AGNs in blue-cloud host galaxies experience less X-ray obscuration,
while AGNs in red-sequence hosts have more, which is the reverse of what is expected
from general considerations of the interstellar medium. Outer and integrated colours of
AGN hosts generally agree with the control galaxies, regardless of X-ray obscuration,
but the nuclear colours of unobscured AGNs are typically much bluer, especially for
X-ray luminous objects. Visible point sources are seen in many of these, indicating
that the nuclear colours have been contaminated by AGN light and that obscuration
of the X-ray radiation and visible light are therefore highly correlated. Red AGN
hosts are typically slightly bluer than red-sequence control galaxies, which suggests
that their stellar populations are slightly younger. We compare these colour data to
current models of AGN formation. The unexpected trend of less X-ray obscuration
in blue-cloud galaxies and more in red-sequence galaxies is problematic for all AGN
feedback models, in which gas and dust is thought to be removed as star formation
shuts down. A second class of models involving radiative instabilities in hot gas is more
promising for red-sequence AGNs but predicts a larger number of point sources in red-
sequence AGNs than is observed. Regardless, it appears that multiple AGN models
are necessary to explain the varied AGN host properties discussed in the current work.
Finally, we find that integrated optical and UV-optical colours are not strongly affected
by X-ray-selected AGNs except in rare cases (< 10 per cent) where the AGN is very
luminous, unobscured, and/or visible as a point source.
−3per cent of
Key words:
galaxies.
galaxies: active – galaxies: nuclei – galaxies: photometry – X-rays:
⋆E-mail: christina.pierce@physics.gatech.edu
† Leo Goldberg Fellow
c ? 2010 RAS

Page 2

2 C. M. Pierce et al.
1 INTRODUCTION
Galaxy colours can tell us much about the stellar popu-
lations and star formation history (SFH) of a galaxy. Of
particular interest are the possible connections between the
SFH and possible energetic feedback related to black hole
growth. However, in order to properly interpret the mea-
sured colours of a galaxy we must understand the ori-
gins of the colours. Radiation associated with an accretion
disc around active galactic nuclei (AGNs) produces large
amounts of blue light, causing potentially unexpected effects
on the measured colours of the host galaxies.
Star formation, AGN feedback and possible connections
between them have been discussed by several authors in
recent years (e.g. Silk & Rees 1998; Granato et al. 2004;
Hopkins et al. 2005a, 2005b, 2006, 2008a, 2008b; Scanna-
pieco, Silk, & Bouwens 2005; Bower et al. 2006; Cattaneo
et al. 2006; Croton et al. 2006; Dekel & Birnboim 2006;
Ciotti & Ostriker 2007). We will compare the results pre-
sented here to a few specific scenarios. One such scenario
described AGNs caused by interactions or mergers between
gas-rich disc galaxies having similar masses (Hopkins et al.
2008a, 2008b). Another explored black hole growth initiated
by instabilities in isolated giant elliptical galaxies (Ciotti
& Ostriker 2007). Both of these scenarios suggest timelines
for various observable features, such as nuclear obscuration
and the shutting down of star formation. We also compare
our results to a scenario that described how feedback from
low-luminosity AGNs may prevent additional star forma-
tion (Croton et al. 2006) following an initial burst associated
with, for example, a merger or interaction.
Ciotti & Ostriker (2007) described simulations related
to the circumstances surrounding significant growth of su-
permassive black holes (SMBHs) found in giant elliptical
galaxies. In their simulations, gas emitted from central stars
led to radiative instabilities and a collapse of metal-rich gas
in the nuclear regions of the galaxy. New star formation
claimed about half of the gas, and about half was ejected
from the nucleus; less than 1 per cent of the gas contributed
to the growth of the central SMBH. Ciotti & Ostriker (2007)
found that both the AGN and the starburst were heavily ob-
scured (Compton-thick; NH > 1024cm−2) during this stage
and probably only observable in far-infrared bands. Radia-
tive feedback from the AGN then caused the expansion of
a central hot bubble, first briefly revealing the AGN as a
traditional quasar, and then eventually shutting down both
star formation and black hole growth. After that time, the
galaxy exhibited an E+A spectrum (a post-starburst galaxy;
see Yan et al. 2006) and low X-ray luminosities.
Hopkins et al. (2008b) showed an optical CMD (their
fig. 24) featuring AGN and quasars previously presented
by Nandra et al. (2007) and S´ anchez et al. (2004), respec-
tively, and compared the observed host galaxy colours to
the colours expected from a merger scenario and from an
activation scenario involving secular (that is, in this case,
non-merger) processes. They noted that AGN host galax-
ies typically exhibited colours that placed them on the red
sequence or in the upper (redder) region of the blue cloud.
Their merger scenario predicts that the AGN host galax-
ies start in the blue cloud and transition on to the red se-
quence, while the secular processes are expected to affect the
host galaxy colours in the opposite manner. Hopkins et al.
(2008b) concluded that the colours of observed AGN host
galaxies support the merger scenario more strongly than the
secular activation scenario.
Croton et al. (2006) presented a complementary sce-
nario in which the energy emitted by AGNs with low accre-
tion rates, in high mass galaxies, sufficiently heats gas in the
disc to prevent star formation, resulting in rapidly aging stel-
lar populations. We can use observations of AGNs and their
host galaxies to test the validity of the predictions suggested
by Croton et al. (2006), Ciotti & Ostriker (2007), Hopkins et
al. (2008a, 2008b) and others. The AGNs typically involved
in the simulations are relatively luminous, such as quasars
or Seyfert galaxies, many of which should be identifiable by
their X-ray emissions. X-ray luminosities can also be used to
estimate the growth rate and obscuration of the black holes.
Comparing X-ray characteristics to galaxy colours allows a
test on the connections and timescales suggested by various
models.
As a stellar population ages, its optical colours shift
from blue to red, corresponding to a decrease in the tem-
perature and energy output of the stars, so that unbiased
galaxy colours indicate the dominant age of the stellar pop-
ulations in a galaxy. Recent models, such as those described
above, make specific predictions about the connection be-
tween black hole growth and the ages of stellar populations
within a galaxy. The prediction that energetic feedback as-
sociated with black hole growth may halt star formation can
be tested by comparing observations of galaxy colours to the
black hole growth rate and obscuration of the accretion disc
around the black hole.
The colours most commonly used to estimate the age
of a galaxy’s dominant stellar population are optical (e.g.
U −B, u−r, B −V ) and UV-optical (e.g. NUV−R). Many
authors (e.g. Baldry et al. 2004; Bell et al. 2004; Faber et
al. 2007) have demonstrated that galaxies form a bimodal
distribution in a variety of optical colours, and Faber et al.
(2007) further detected a net flow of galaxies from the ‘blue
cloud’ to the ‘red sequence’. However, young stars contribute
significantly to the ultraviolet continuum, suggesting that
UV-optical colours may provide a better indication of recent
star formation (Kennicutt 1998). Wyder et al. (2007) stud-
ied the UV-optical colour NUV−R and found not only a bi-
modal distribution but also a significant population between
the two colour extremes, now known as the ‘green valley’.
Various studies have since presented observations character-
izing the green valley as a possible transition region between
the blue cloud and the red sequence (e.g. Martin et al. 2007;
Schiminovich et al. 2007), particularly for AGN host galax-
ies, in which feedback may significantly affect the star for-
mation rates (e.g. Schawinski et al. 2007; Georgakakis et al.
2008b).
A full description of an AGN host galaxy depends on
measurements of observable features such as colours and the
distribution of light in the galaxy. However, the spectrum
emitted by an AGN differs significantly from that emitted
by its host galaxy and can dominate optical observations.
This is most clearly seen in quasi-stellar objects (QSOs), but
may also be evident in less luminous AGNs. Fig. 1 shows this
difference in the optical region of the spectrum by showing
spectral templates for an AGN (QSO; upper panel) and an
Sb galaxy (lower panel) from Kinney et al. (1996). Whether
considering an optical colour (e.g. U − B) or a UV-optical
c ? 2010 RAS, MNRAS 000, 1–19

Page 3

AGN host galaxy colour gradients3
Figure 1. Spectral templates from Kinney et al. (1996) demon-
strating the difference in optical colours between an AGN and a
galaxy not hosting an AGN. Approximate wavelength ranges of
the NUV, U, B and r bands are indicated. The U and B bands
do not overlap; the vertical line indicates the separation between
them. Upper panel: QSO spectral template. Lower panel: Spectral
template of an Sb galaxy.
colour (e.g. NUV−R), it is clear that the AGN would ap-
pear significantly bluer than the galaxy. The effect on the
measured colour of an AGN host galaxy depends on factors
such as the AGN luminosity and obscuration of the AGN
accretion disc.
Using a low-redshift (0.03 < z < 0.07) sample of optical
spectroscopically-selected AGN host galaxies, from which
AGNs exhibiting broad spectral lines (typical of the most ex-
treme AGNs, which are most likely to affect optical measure-
ments) were specifically excluded, Kauffmann et al. (2007)
compared optical colours (g −r) from the central regions of
the galaxies, based on observations from the Sloan Digital
Sky Survey, to integrated (that is, total) UV-optical colours
(NUV−r) from Galaxy Evolution Explorer (GALEX; Martin
et al. 2005) observations. They showed that light from stars
in the outer regions of these AGN host galaxies dominates
the observed UV-optical colours, indicating that the colours
are not strongly affected by light from an AGN. However,
many current studies that use optical colours to help char-
acterize the host galaxy stellar populations focus on more
luminous AGNs (such as those selected by X-ray or radio
techniques) at higher redshifts (z ∼ 1; e.g. Nandra et al.
2007; Bundy et al. 2008; Coil et al. 2009; Silverman et al.
2009). Thus we also need to understand the possible effects
of high-luminosity, z ∼ 1 AGNs on the measured colours of
their host galaxies.
The current work and a companion study (Pierce et
al. 2010) address this using complementary methods. Pierce
et al. (2010) added an AGN spectral template (the QSO
template shown in Fig. 1) to three non-AGN templates (an
elliptical galaxy, an Sb galaxy and a starburst galaxy). They
scaled the AGN template to contribute a set of specified frac-
tions of the flux from the resulting system and measured
the original and resulting optical and UV-optical colours,
finding that the AGN template significantly affected the
measured colours. The results from Kauffmann et al. (2007)
could be considered a lower limit and the results from Pierce
et al. (2010) could be considered an upper limit to the ex-
pected effect of an AGN. In addition, to test the potential
effect of an AGN on morphology measurements of the host
galaxies, Pierce et al. (2010) added a series of optical point
sources to optical images of galaxies at z ∼ 0.5 not known
to host AGNs. They compared the measured morphologies
of the original and altered galaxy images and found that
high AGN fractions can significantly bias the morphology
measurements, but that such AGNs are often identifiable
from the optical images due to the visibility of the AGN
as a central point source. Additional previous work with X-
ray-selected AGN at redshifts 0.5 < z < 1.5, in the Great
Observatories Origins Deep Survey (GOODS), found that
unobscured AGN hosts are similar to obscured AGN hosts
in NUV−R, but slightly bluer (Ammons 2009).
The current study was initially undertaken in order to
determine the extent to which luminous AGNs at z ∼ 1 af-
fect the measured colours of their host galaxies. As a result,
we have discovered criteria for identifying the AGNs that
are most likely to cause colour contamination. Using Hubble
Space Telescope/Advanced Camera for Surveys (HST/ACS)
V and I band images, we measure the outer and nuclear
galaxy colours of a sample of X-ray-selected AGNs at red-
shifts 0.2 < z < 1.2. Measuring galaxy colours in this
manner essentially restricts to the nuclear regions any ef-
fect caused by the AGNs, while colours measured for the
outer regions are expected to be free of any influence from
the AGN. In the special case of QSOs, the outer regions
would also be overwhelmed by AGN light, but the sample
considered here does not contain any known QSOs.
The AGN host galaxy aperture colours are compared
to aperture colours of a control sample consisting predom-
inantly of galaxies not hosting AGNs. We create a con-
trol sample because most AGN host galaxies at redshifts
0.2 < z < 1.2 exhibit characteristics (most importantly,
mass and colour) that differ from the characteristics typi-
cal of most galaxies at such redshifts. From the AGN host
galaxy HST/ACS images, we determine whether or not the
AGN is apparent as an optically visible point source and find
that our results correlate with the X-ray obscuration of the
AGN, as determined by the X-ray hardness ratio (see Sec-
tion 2.3), indicating a connection between the optical obscu-
ration and the X-ray obscuration. The X-ray hardness ratios
are also found to correlate with the outer colours, facilitat-
ing comparisons to predictions from the models described
above.
The data used for this study come from the All-
wavelength Extended Groth Strip International Survey
(AEGIS; Davis et al. 2007), a multiwavelength survey cov-
ering bands from hard X-ray through radio. Many authors
have already used these observations to study topics such
as galaxy SEDs (Konidaris et al. 2007; Symeonidis et al.
2007), various aspects of star formation (Ivison et al. 2007;
Lin et al. 2007; Noeske et al. 2007a, 2007b; Weiner et al.
2007), AGN selection techniques and host galaxy character-
istics (Georgakakis et al. 2007; Gerke et al. 2007; Nandra et
al. 2007; Pierce et al. 2007; Park et al. 2008), connections
between AGN feedback and galaxy colours (Bundy et al.
2008; Georgakakis et al. 2008b), galaxy groups and cluster-
ing (Fang et al. 2007; Georgakakis et al. 2008a; Coil et al.
2009; Jeltema et al. 2009) and a variety of additional top-
ics (Barmby et al. 2006; Conselice et al. 2007; Huang et al.
c ? 2010 RAS, MNRAS 000, 1–19

Page 4

4C. M. Pierce et al.
2007; Kassin et al. 2007; Moustakas et al. 2007; Wilson et
al. 2007; Aird et al. 2010; Laird et al. 2009; Sato et al. 2009).
We begin with a description of the data used for the
current study (Section 2) and then describe the selection of
AGN and control samples (Section 3). Nuclear and outer
optical colours and integrated UV-optical colours are pre-
sented in Section 4, followed by a discussion of the results
and their implications in Section 5. Finally, in Section 6 we
provide a summary of the main scientific results. Through-
out, we use {h, ΩΛ, ΩM} = {0.7, 0.7, 1−ΩΛ} and AB mag-
nitudes, unless otherwise noted. In addition, uncertainties
accompanying numerical fractions represent 1σ uncertain-
ties, calculated following Gehrels (1986).
2 DATA
2.1 Optical images
High spatial resolution HST/ACS images, with a point
spread function (PSF) FWHM of ∼ 0.1 arcsec, are avail-
able for 0.197 deg2of the Extended Groth Strip (EGS).
This region was observed in the F606W (V ) and F814W
(I) passbands to 5σ limiting magnitudes of V = 28.14 mag
and I = 27.52 mag for a point source; the limiting magni-
tudes are slightly brighter for extended objects (Davis et al.
2007).
We measure the observed V and I band light in a series
of apertures with radii 0.15 arcsec, 0.2 arcsec and 1.5 arc-
sec, and then calculate the observed optical colours (V −I)
within a central region of radius 0.15 arcsec (the ‘nuclear’
colours) and an annulus having an inner radius of 0.2 arcsec
and an outer radius of 1.5 arcsec (the ‘outer’ colours). The
central region encloses the nuclear point sources that are
visibly present in several of the AGN host galaxies, as well
as ∼80 per cent of the HST/ACS V band PSF. The outer
radius of the annulus is large enough to fully enclose most of
the galaxies in our sample, and the size of the inner radius
allows a small separation between the outer annulus and the
central region. From the observed aperture colours, we esti-
mate the K-corrected, rest-frame aperture colours (U−B)out
and (U −B)nuc, following methods described by Willmer et
al. (2006) and Weiner et al. (2009). Redshift evolution of the
colour gradients presented in Section 4.1 is minimal and does
not significantally influence our results, so that we are not
concerned with the differences between the physical scales
of the regions measured using the same angular scales at
different redshifts.
For each of the X-ray-selected AGNs (Section 3.1), the
V and I band images were inspected by-eye to assess the
visibility of the AGN as a central (or offset) point source.
Criteria used to distinguish between a point source and a re-
gion of star formation include size, shape and distinctness of
edges; our ‘point sources’ are required to have sizes similar
to the ACS PSF (which is independent of redshift), circu-
lar shape and distinct edges. Each of the three AGN host
galaxies shown in Fig. 2 represent one of our three point
source visibility classifications – ‘definite point source’, ‘pos-
sible point source’ and ‘no clear point source’. These postage
stamps have been created from HST/ACS I band images,
and all use the same logarithmic scale, bias and contrast.
2.2 Spectroscopic redshifts
Spectroscopic redshifts are available from the DEEP2 Red-
shift Survey (Davis et al. 2003, 2007) and an MMT Obser-
vatory survey of X-ray-selected AGN host galaxies (Coil et
al. 2009). Redshifts were individually verified and assigned
quality codes pertaining to the redshift reliabilities, result-
ing in ∼3700 DEEP2 and 82 MMT ‘high quality’ (reliability
confidence levels ? 95 per cent) redshifts in the HST/ACS-
imaged region of the EGS; 80 of these galaxies have high
quality spectroscopic redshifts from both the DEEP2 and
the MMT surveys.
2.3 X-ray images
The Chandra X-ray Observatory (Chandra) Advanced CCD
Imaging Spectrograph observed a strip of eight pointings
(0.67deg2total) along the EGS for approximately 200 ks
per pointing (Nandra et al. 2005; Davis et al. 2007). Laird
et al. (2009) presented the AEGIS-X survey (the 1325 X-
ray sources detected in the EGS) and described the meth-
ods used to reduce and analyse the observations, includ-
ing calculation of the hardness ratios HR1. For the AEGIS-
X survey catalog, X-ray sources are defined as X-ray de-
tections that have at least a 5σ detection significance. At
z = 1, the on-axis flux limit2for hard-band3-selected sources
(f = 3.8×10−16erg cm−2s−1) corresponds to an X-ray lu-
minosity L2−10 keV = 2.4 × 1042erg s−1, slightly in excess
of the minimum luminosity used to define our sample of X-
ray-selected AGNs (Section 3.1). Thus, although the X-ray
AGN sample is complete to L2−10 keV= 1042erg s−1at red-
shifts z < 0.7, it may miss AGNs with L2−10 keV< 4×1042
erg s−1at z = 1.2 (cf. Fig. 3).
2.4Stellar masses
Bundy et al. (2006) estimated galaxy stellar masses in the
four DEEP2 fields by fitting spectral energy distributions
(SEDs) based on B, R, I and KS-band images to models
created using the Bruzual & Charlot (2003) stellar popu-
lation synthesis code. The robustness of these stellar mass
estimates for X-ray-selected AGN host galaxies was tested
by Bundy et al. (2008), and they found that although the fits
between the observed and model SEDs are generally better
for galaxies not hosting AGNs, they are typically good even
for AGN host galaxies. The work by Bundy et al. (2006)
provides stellar mass estimates for 3382 AEGIS galaxies in
the HST/ACS-imaged region, including 54 of our 56 X-ray-
selected AGNs (Section 3.1).
Estimates provided by C. N. A. Willmer supplement the
1HR ≡ (H−S)/(H+S); H=2–7 keV counts; S=0.5–2 keV counts.
HR indicates the amount of attenuation experienced by the lower
energy X-rays due to gas and dust in our line-of-sight. At z = 0,
HR > −0.25 (< −0.25) indicates high (low) attenuation of low-
energy X-rays. Due to redshifting of the energy bands, high-z
sources may have harder X-ray spectra (more obscuration) than
observed.
2The flux limit increases with the distance from the centre of
the Chandra pointing.
3X-ray energy bands: full (0.5–7 keV), soft (0.5–2 keV), hard
(2–7 keV) and ultra-hard (4–7 keV)
c ? 2010 RAS, MNRAS 000, 1–19

Page 5

AGN host galaxy colour gradients5
Figure 2. AGN host galaxies representing our three point source visibility classifications. From left to right, the galaxies have a ‘definite
point source’, a ‘possible point source’ and ‘no clear point source’. All three I band images were created using the same bias, contrast
and logarithmic scale.
stellar mass estimates from Bundy et al. (2006). Willmer de-
rived the masses from rest-frame optical B −V colours, fol-
lowing Bell & de Jong (2001), Bell et al. (2005), Willmer et
al. (2006), Lin et al. (2007) and Weiner et al. (2009). We test
the reliability of these stellar mass estimates for AGN host
galaxies by comparing the masses determined from V − I
aperture colours using three different inner radii (0.0 arc-
sec, 0.1 arcsec and 0.2 arcsec) and a common outer radius
(1.5 arcsec). The difference in the resulting mass estimates
is minimal, indicating that the colour-derived estimates are
also robust.
Both stellar mass estimates are available for 390 of our
460 control galaxies (Section 3.2), for which the median dif-
ference is log(M∗,Bundy)−log(M∗,B−V) = 0.045. The masses
derived by Bundy et al. (2006) typically have lower uncer-
tainties than the masses derived using the B − V colours
(∼ 0.3 dex). The stellar mass estimates for the AGN hosts
(where available) and 96.5 per cent of the control sample
galaxies are from Bundy et al. (2006); the remaining stellar
mass estimates that we use come from Willmer.
2.5UV-optical colours
Applying the methods described by Salim et al. (2005,
2007) to AEGIS galaxies, Salim et al. (2009) combined
GALEX and Canada–France–Hawaii Telescope Legacy Sur-
vey u∗g′r′i′z′observations with KS-band photometry to es-
timate the rest-frame UV-optical colour NUV−R, exclud-
ing galaxies for which the DEEP2 spectra fit a template
for Type-1 (unobscured) AGNs; NUV−R colour estimates
are available for 91 per cent (51/56) of our X-ray-selected
AGN sample (Section 3.1). Of the five AGNs for which we
do not have NUV−R estimates, two exhibit U − B colours
and MB magnitudes consistent with Type-1 AGNs (cf. fig.
1 of Nandra et al. 2007) and the remaining three have U −B
colours and MB magnitudes consistent with the majority of
the AGNs in our sample. These AGNs exhibit approximate
X-ray luminosities 3 × 1042erg s−1< L2−10 keV < 3 × 1044
erg s−1and X-ray hardness ratios −0.59 <HR < 0.88.
Estimating the UV-optical colours includes fitting the
galaxy SED to a library of model SED templates and de-
termining the strength of the match (the goodness of fit,
χ2; Salim et al. 2007). Galaxies for which χ2< 10 are
considered to have ‘reliable’ UV-optical colours, and 94 per
cent (48/51) of the UV-optical colour estimates for our X-
ray sample are thereby deemed reliable. The NUV−R and
V −I colours of the three AGN hosts that have ‘unreliable’
UV-optical colours are similar to the colours of the other
AGN host galaxies in our sample. Although our results do
not depend upon the inclusion or exclusion of these three
AGNs, we will present them for comparison along with the
galaxies that have reliable colours. Throughout the analy-
ses described here, references to blue, green and red UV-
optical colours indicate NUV−R < 3, 3 <NUV−R < 4.5
and NUV−R > 4.5, respectively.
3 SAMPLES
3.1X-ray-selected AGNs at 0.2 < z < 1.2
Luminous, high-energy X-ray sources are believed to be
AGNs because star formation processes are only expected to
account for lower energy and/or less luminous X-ray emis-
sions (e.g. Grogin et al. 2003, 2005; Barger et al. 2005; Laird
et al. 2005). The current work uses L2−10 keV> 1042erg s−1
as the criterion for X-ray-selected AGNs (Grogin et al. 2005;
Barger, Cowie & Wang et al. 2007), an order of magnitude
higher than a conservative cut used by Laird et al. (2005) to
exclude AGNs from their sample. Thus, the X-ray-selected
AGN sample used here should be fairly pure, though it may
exclude some low-luminosity or Compton-thick AGNs.
Georgakakis et al. (2009) used the Likelihood Ratio
method (e.g. Ciliegi et al. 2003) to identify DEEP2 coun-
terparts for 131 AEGIS X-ray sources. Using these matches,
we follow the method described by Teng et al. (2005) to con-
vert the X-ray flux in each of the four X-ray energy bands
to 2–10 keV fluxes, assuming a power-law slope Γ = 1.4
(e.g. Peterson 1997), which assumes uniform X-ray obscura-
tion for all of our AGNs. This calculation provides estimates
of the observed 2–10 keV X-ray fluxes, which may be less
than the intrinsic X-ray fluxes of the more heavily obscured
AGNs. In order to determine the extent to which the choice
c ? 2010 RAS, MNRAS 000, 1–19