Supermassive black-hole growth over cosmic time: Active
galaxy demography, physics, and ecology from
Chandra surveys

W. N. Brandt

a,1

and D. M. Alexander

b

a

Department of Astronomy and Astrophysics, The Pennsylvania State University, University Park, PA 16802; and

b

Department of Physics,

Durham University, Durham DH1 3LE, United Kingdom

Edited by Harvey D. Tananbaum, Smithsonian Astrophysical Observatory, Cambridge, MA, and approved February 4, 2010 (received for review December 7, 2009)

Extragalactic X-ray surveys over the past decade have dramatically improved understanding of the majority populations of active galactic
nuclei (AGNs) over most of the history of the universe. Here we brie

fly highlight some of the exciting discoveries about AGN demography,

physics, and ecology, with a focus on results from Chandra. We also discuss some key unresolved questions and future prospects.

active galactic nuclei

|

Chandra X-ray Observatory

|

extragalactic surveys

E

xtragalactic X-ray surveys are
powerful for studying the growing
supermassive black holes
(SMBHs) in active galactic nuclei

(AGNs) for several reasons. First, X-ray
emission is empirically found to be nearly
universal from luminous AGNs; the accre-
tion disk and its

“corona” are robust even if

their details remain somewhat mysterious.
Second, X-ray emission is penetrating and
has reduced absorption bias compared
with, for example, optical and UV emission.
This is critically important because it is now
known that the majority of AGNs suffer
from signi

ficant intrinsic obscuration. Fur-

thermore, the level of X-ray absorption bias
drops toward high redshift, because in-
creasingly penetrating rest-frame X-rays
are observed. Finally, X-ray observations
maximize the contrast between SMBH-re-
lated light and host-galaxy starlight. Having
such high contrast is crucial when studying
high-redshift objects that cannot be re-
solved spatially. X-ray surveys thus provide
the

“purest” AGN samples; most (≥80%)

of the sources even in the deepest X-ray
observations are AGNs, whereas few
(

≤10%) of the sources in the deepest opti-

cal and infrared observations are AGNs.

Relevant Capabilities of X-Ray Surveys
with Chandra
The unmatched angular resolution, low
background, broad bandpass, and
respectable

field of view of Chandra have

provided dramatic advances in our ability
to survey the X-ray emission from AGNs
over most of the history of the universe.
The deepest Chandra observations are 50

–

250 times more sensitive than those of
previous missions (the exact factor de-
pending on the bandpass considered), de-
tecting sources with photon

fluxes as low

as one count per 5 days. Source positions
measured by Chandra are typically reliable
to within 0.2

–0.5″; this is essential for ro-

bust identi

fications and follow-up work at

faint

fluxes. The surveys executed by

Chandra have each detected hundreds
to thousands of sources, allowing statisti-
cally meaningful studies of source pop-
ulations. Finally, the well-maintained data
archive allows the effective federation of
Chandra surveys by any astronomer to
address speci

fic scientific questions

of interest.

Currently approximately 35 Chandra

and XMM-Newton surveys have been per-
formed that cover most of the practically
accessible

“discovery space” of sensitivity

vs. solid angle. These include contiguous
surveys, many of which are shown in Fig. 1,
as well as the equally important non-
contiguous and often serendipitous sur-
veys [e.g., The Chandra Multiwavelength
Project (ChaMP), The High Energy Large
Area Survey with XMN-Newton (HEL-
LAS2XMM), The Serendipitous Extra-
galactic X-Ray Source Identi

fication

Program (SEXSI), and the XMM-Newton
Survey Science Centre (SSC) surveys].
Enormous progress has been made over
the past decade in obtaining identi

fication

spectra for large, representative samples
of the detected sources; this work has of-
ten used the largest ground-based tele-
scopes on Earth (e.g., Gemini, Keck,
Subaru, and the Very Large Telescope).
However, spectroscopic identi

fication re-

mains a persistent challenge and bottle-
neck, especially at faint

fluxes (I = 24–28),

and serves as one important driver for
future extremely large telescopes (ELTs).
Multiwavelength observations of the
Chandra survey sources, from the radio to
the UV, have also been critical for ad-
vancing understanding, as expected given
the broadband nature of AGN emission.
These have improved the reliability of the
X-ray source identi

fications, allowed the

derivation of high-quality photometric
redshifts, constrained AGN accretion
physics, measured host-galaxy properties,
assessed the relative importance of SMBH
vs. stellar power, and even discovered

AGNs that were missed by Chandra (e.g.,
owing to extreme obscuration).

Below we will brie

flyhighlight someof the

exciting discoveries from Chandra surveys
about the tightly related topics of AGN
demography, physics, and ecology. Our fo-
cus will be on Chandra results from the past
decade, as be

fits this 10th birthday sympo-

sium for Chandra, implicitly also recogniz-
ing the fundamental advances made by
extragalactic surveys with XMM-Newton.
Comparisons will sometimes be made with
the community

’s understanding at around

the time of the Chandra launch in mid-1999,
because these illustrate just how dramatic
the advances have been over Chandra

’s first

decade of discovery. Furthermore, owing to
limited space, our references to the liter-
ature will necessarily be limited, highly se-
lective, and incomplete; our humble
apologies in advance.

Demography
From the 1960s to the 1990s, the study of
AGN evolution was dominated by wide-

field optical surveys of rare, luminous
quasars (e.g., refs. 2 and 3). These were
found to peak in comoving number density
at z

≈ 2–3 and showed evolution consistent

with pure luminosity evolution models.
These surveys left open a major question:
how does the numerically dominant pop-
ulation of moderate-luminosity AGNs
evolve? Many astronomers expected,
before the launch of Chandra, that
moderate-luminosity AGNs would evolve
in the same manner as luminous quasars.

However, even from the Röntgensatellit

(ROSAT) soft X-ray extragalactic surveys,
hints were emerging that AGN evolution
is signi

ficantly luminosity dependent (e.g.,

Author contributions: W.N.B. and D.M.A. designed re-
search; performed research; and wrote the paper.

The authors declare no con

flict of interest.

This article is a PNAS Direct Submission.

1

To whom correspondence should be addressed. E-mail:
[e-mail].

7184

–7189

|

PNAS

|

April 20, 2010

|

vol. 107

|

no. 16

www.pnas.org/cgi/doi/10.1073/pnas.0914151107

ref. 4). These surveys also hinted, in-
dependently, that the X-ray

–selected

quasar space density at z

≥ 3 might not

decline in the manner seen for optically
and radio-selected quasars. The ob-
servational constraints, at the time of the
Chandra launch, admitted the possibility
that luminous AGNs dominated cosmic
reionization. There were even widely dis-
cussed claims (5) that Chandra might de-
tect

≈100 quasars at z ≥ 5 in a single deep-

field observation!

Chandra observations allow the effective

selection of AGNs, both obscured and un-
obscured, that are up to

≈100 times less

luminous than those from wide-

field optical

surveys. These AGNs are

≥500 times more

numerous. As a result, the AGN number
counts from the deepest Chandra surveys
have reached

≈7,200 deg

−2

(e.g., ref. 6; vs.

≈13 deg

−2

from the Sloan Digital Sky Sur-

vey and

≈800 deg

−2

from the deepest

ROSAT surveys). At a basic level, this is the
key demographic

“discovery space” that

was opened by Chandra surveys.

The moderate-luminosity AGNs dis-

covered in the Chandra surveys are not
measured to evolve in the same manner as
luminous quasars, indicating that AGN
evolution is luminosity dependent (e.g.,
refs. 7

–11). Lower-luminosity AGNs are

found to peak in comoving number den-

sity at later cosmic times; this general
behavior is sometimes referred to as

“cosmic downsizing” or “antihierarchical
growth.

” The details of this behavior are

still somewhat uncertain owing to multiple
thorny observational (e.g., detection in-
completeness, source identi

fication,

follow-up incompleteness, X-ray spectral
complexity) and statistical issues. Thus,
the workers in this

field often have strong,

inconsistent opinions! However, the gen-
eral consensus is that total SMBH power
production peaks at signi

ficantly lower

redshifts (z

≈ 1–1.5) than expected on the

basis of evolution studies solely of lumi-
nous quasars (z

≈ 2–3). At high redshift,

the demographic constraints now show
that there is indeed a decline in the space
density of X-ray

–detected AGNs at z ≥ 3

(e.g., refs. 9, 12, and 13). This decline has
a roughly exponential form, similar to
what is found for optically selected qua-
sars. Luminous AGNs are unlikely to have
dominated cosmic reionization, leaving
stars as the most likely agents.

The luminosity functions delivered by the

X-ray AGN demographers have been used
with versions of the elegant So

łtan argument

(14) to predict the masses of remnant
SMBHs in galactic centers as well as the
typical growth histories of SMBHs of vari-
ous masses (e.g., refs. 15

–18). The most ro-

bust points generally emerging from this
elaborate work are that standard radiatively
ef

ficient accretion can plausibly drive most

SMBH growth, and that more massive
SMBHs generally grew earlier in cosmic
time (Fig. 2). Signi

ficant uncertainties still

remain, however, in the luminosity func-
tions themselves, the local SMBH mass
function, bolometric corrections, Edding-
ton ratios, and the ef

ficiency of SMBH

accretion. Together these limit the strength
of some of the constraints that can be de-
rived from So

łtan-type arguments.

What has been the relative production

of cosmic power by SMBHs vs. stars?
Shortly before the Chandra launch, it was
claimed that SMBHs may have supplied
up to 50% of the universe

’s total energy

output since the formation of galaxies
(19). The Chandra AGN demographic
results, however, now show that SMBH
accretion has likely only supplied approx-
imately 5

–10% of this energy output;

the remaining majority comes from nu-
clear fusion in stars. We seem to live
in a remarkably economical X-ray uni-
verse, in that the observed cosmic X-ray
background (CXRB) is produced with al-
most the least cosmic effort possible. It is
not dominated by luminous obscured
quasars thundering out huge amounts
of power at z

≈ 2–4 but rather by

moderate-luminosity, obscured AGNs at
z

≈ 0.5–2.

The work of the demographers is not

finished. There is strong evidence that a
large population of intrinsically luminous
but heavily obscured (N

H

≥ 3 × 10

23

cm

−2

)

AGNs, comprising a signi

ficant fraction of

cosmic SMBH growth, is still not detected
in the Chandra surveys. This is not sur-
prising, given expectations from the low-
redshift universe. For example, many local
Compton-thick (N

H

≥ 1.5 × 10

24

cm

−2

)

AGNs that are intrinsically luminous (e.g.,
NGC 1068, NGC 6240, and Mrk 231)
would remain undetected even in the
Chandra deep

fields if placed at z ≥ 0.5–3.

Direct evidence for missed distant AGNs
comes in several forms. For example,
stacking analyses show that only

≈50–70%

of the 6

–8-keV CXRB is resolved even in

the deepest X-ray observations; the cor-
responding undetected X-ray source pop-
ulation plausibly has a sky density
of

≥2,000–3,000 deg

−2

with N

H

≥ 10

23

–

10

24

cm

−2

at z

≥ 0.5–1.5 (e.g., ref. 20).

Many compelling X-ray

–undetected AGN

candidates have been found within the
deepest Chandra observations via Spitzer
surveys, radio surveys, optical-to-mid-
infrared spectroscopy, and optical-
variability studies (e.g., refs. 21

–28). These

objects now require better character-
ization, at X-ray and other wavelengths, so
that the contribution from SMBH accre-
tion to their total luminosities can be
determined reliably.

0.1

10

10

−17

10

−16

10

−15

0.5−2 keV flux limit (erg cm

−2

s

−1

)

Solid angle (degrees

2

)

ROSAT UDS

2 Ms CDF−S; 2 Ms CDF−N

E−CDF−S

Lynx

LALA Cetus

SSA13

AEGIS Deep

CLASXS

C−COSMOS

SSA22

ELAIS−S1

XBootes

3 Ms XMM CDF−S

XMM COSMOS

Subaru XMM

XMM LSS

1

Fig. 1.

Distribution of some well-known extragalactic surveys by Chandra (blue dots), XMM-Newton

(red stars), and ROSAT (green square) in the 0.5

–2-keV flux-limit vs. solid-angle plane. Each of the surveys

shown has a range of

flux limits across its solid angle; we have generally shown the most sensitive flux

limit. All surveys shown are contiguous. Adapted from Brandt and Hasinger (1).

Brandt and Alexander

PNAS

|

April 20, 2010

|

vol. 107

|

no. 16

|

7185

SPECI

A

L
F

EATUR

E

:

P

ERSPECTIVE

Physics
Extragalactic Chandra surveys have also
provided insights into the processes shap-
ing the observed X-ray emission from
AGNs, ranging from accretion-disk (down
to light minutes) to

“torus” (0.1–100 light

years) physical scales. They have served as
an essential complement to detailed X-ray
studies of bright and usually nearby AGNs,
often by providing powerful statistical
constraints upon the basic emission prop-
erties of moderate-luminosity, typical
AGNs in the distant universe.

When combined with multiwavelength

AGN samples, Chandra surveys have been
important in tightening empirical con-
straints upon the universality of X-ray
emission from SMBH accretion disks and
their so-called coronae (e.g., refs. 1, 29,
and 30). This central dogma of universal
X-ray emission (cf. ref. 31), still on em-
barrassingly shaky ground from an ab ini-
tio physics point of view, underlies the
utility of all Chandra surveys for

finding

AGNs throughout the universe.

The broad coverage of the luminosity

–

redshift plane provided by AGN samples in
Chandra extragalactic surveys has allowed
substantially improved constraints to be set
upon X-ray-to-optical/UV spectral energy

distributions (SEDs; e.g., refs. 30, 32

–37).

This is the spectral region where the direct
accretion emission is dominant for relatively
unobscured AGNs, and X-ray-to-optical/
UV SED studies thus probe the inner

≈100–

1,000 gravitational radii (e.g., the relative
amounts of power emitted by the corona vs.
the underlying disk). Although there are
still some discrepancies among published
results (e.g., where

fitted parameters from

different samples disagree by much more
than is allowed by their statistical un-
certainties), some general points of con-
sensus have emerged. First, there is a clear
luminosity dependence of X-ray-to-optical/
UV SEDs for the majority population of
radio-quiet AGNs, such that the ratio of
X-ray vs. optical/UV emission declines with
rising optical/UV luminosity (Fig. 3). This
result, initially found in the 1980s with
limited samples (e.g., ref. 31), has now been
established to hold out to z

≈ 4–6 and over a

range of

≈100,000 in luminosity. The form

of the luminosity dependence is likely non-
linear, being stronger at high luminosities
than low luminosities. Further work to
constrain this nonlinearity is required, as are
ab initio physics-based calculations capable
of predicting the luminosity dependence of

X-ray-to-optical/UV SEDs (see, e.g., ref. 38
and references therein).

The majority of current studies indicate

that, after controlling for the luminosity
dependence of X-ray-to-optical/UV SEDs,
there is no remaining detectable redshift
dependence. For example, refs. 32 and 33
show that, at a

fixed luminosity, the ratio

of X-ray-to-optical/UV luminosity is con-
strained to change with redshift by

<30%

out to z = 5

–6. It seems that, despite the

large demographic changes in the AGN
population over cosmic time, the in-
dividual AGN unit is remarkably stable on
the scale of the inner accretion disk.

Obscuration-based uni

fication models

have also been re

fined using the large AGN

samples from Chandra extragalactic surveys
(e.g., refs. 8 and 39

–41). Here again the

broad coverage of the luminosity

–redshift

plane has been essential, allowing obscura-
tion dependences upon luminosity and red-
shift to be constrained in much greater detail
than was previously possible. The improved
data con

firm longstanding expectations

(e.g., refs. 42 and 43) that the fraction of
obscured AGNs drops with increasing
luminosity; that is, the covering factor of the
torus is luminosity dependent, perhaps
because more luminous AGNs can
evacuate their environments better. The
obscured AGN fraction drops in a roughly
linear manner as a function of logarithmic
2

–10-keV luminosity, falling from ≈80% at

10

42

erg s

−1

to

≈20% at 10

45

erg s

−1

. Of

course, the exact numerical values for these
fractions depend upon how obscured AGNs
are de

fined (X-ray, optical, and SED-based

classi

fication schemes do not consistently

agree, especially at low luminosities) and still
have nonnegligible systematic uncertainties
owing to spectral complexity and
missed AGNs.

After controlling for luminosity effects,

the fraction of obscured AGNs is found to
rise with redshift as (1 + z)

0.3

–0.7

(e.g., refs. 8,

40, and 41). This behavior seems to hold at
least up to z

≈ 2 where uncertainties become

large (systematic uncertainties, as men-
tioned above for the luminosity depend-
ence of the obscured fraction, are also
relevant here). The processes ultimately
shaping the torus thus seem to evolve over
cosmic time, in notable contrast to what is
found for the inner accretion disk. The in-
crease in the covering factor of the torus
with redshift is plausibly driven by the
greater availability of gas and dust in gal-
axies at earlier cosmic epochs.

Ecology
Since the launch of Chandra, it has become
well established that AGNs play a role
in the evolution of galaxies. The

finding

that many massive galaxies in the local
universe host an SMBH with a mass
broadly proportional to that of the galaxy
spheroid hints at concordant SMBH-

Fig. 2.

Average growth history of SMBHs as computed by Marconi et al. (16) using X-ray AGN luminosity

functions. The symbols along each curve indicate the points where an SMBH reaches a given fraction of its
final mass. Note that more massive SMBHs grew at earlier cosmic times. SMBHs that are now more massive
than

≈10

8

M

⊙

gained most of their

final mass by z ≈ 1.5, whereas lower-mass black holes still grew sub-

stantially at lower redshifts. [Reproduced with permission from Marconi et al. (16) (Copyright 2006, Society
of Astronomy, Italy).]

7186

|

www.pnas.org/cgi/doi/10.1073/pnas.0914151107

Brandt and Alexander

spheroid growth (e.g., refs. 44 and 45),
suggesting a close connection between
AGN activity and star formation. The
optical-to-near-infrared emission from
most of the distant AGNs detected in
AGN surveys before Chandra was domi-
nated by the active nucleus, restricting the
constraints that could be set upon host
galaxies. Because the optical-to-near-in-
frared emission from a large fraction of
the Chandra-selected AGNs is dominated
by starlight, it is now possible to measure
directly the host-galaxy properties (e.g.,
morphology, color, luminosity, and stellar
mass). By combining the X-ray data with
infrared-to-radio observations, the relative
power from AGN vs. star-formation ac-
tivity can also be assessed.

High-resolution Hubble Space Telescope

(HST) imaging of z

≤ 1.5 X-ray–selected

AGNs in deep Chandra (and XMM-
Newton) surveys has shown that their host
galaxies often have concentrated optical-
light pro

files, consistent with expectations

for galaxy spheroids (e.g., refs. 46

–49);

≈40–50% seem to be early-type galaxies,

≈20–30% seem to be late-type galaxies, and
the rest are peculiar or point-like systems.
AGN host galaxies are also optically
luminous, indicating that they are massive
[M

*

≈ (0.3–3) × 10

11

M

⊙

e.g., refs. 50

–52].

First-order constraints therefore suggest
that the SMBHs are comparatively massive
and slow growing (typically M

BH

≈ 10

8

M

⊙

and L

Bol

/L

Edd

≈ 10

−2

; e.g., refs. 50, 51, 53,

and 54), implying that they accreted the
bulk of their mass at z

≥ 1.5. Current con-

straints on the host-galaxy properties and
SMBH masses of z

≥ 1.5 AGNs are, how-

ever, poor because of the faintness (at
optical-to-near-infrared wavelengths) of

the majority of the population, and deeper
imaging and spectroscopy are required for
signi

ficant results (see, e.g., refs. 52, 55, and

56 for some constraints). Small, rapidly
growing SMBHs (M

BH

≤ 10

7

M

⊙

; L

Bol

/L

Edd

> 10

−2

) at z

< 1 are detected in deep X-ray

surveys but seem to be comparatively rare
(e.g., refs. 53 and 57).

Similar to the normal-galaxy population

at z

≤ 1.5, X-ray AGN host galaxies have a

broad range of optical colors. However,
whereas the optical-color distribution of
normal galaxies is clearly bimodal, with a

“red sequence” and “blue cloud,” AGNs
preferentially reside in the red sequence,
the top of the blue cloud, and the

“green

valley

” in between (e.g., refs. 54, 58, and

59). The green valley is the expected lo-
cation for galaxies transitioning between
the blue cloud and the red sequence owing
to the quenching of star formation (pre-
dicted by most galaxy formation models to
be caused by large-scale out

flows); how-

ever, bulge-dominated systems re-
juvenated by the accretion of fresh gas
from their environments could also lie in
the green valley (e.g., ref. 41). Sensitive
spectroscopic observations could dis-
tinguish between these scenarios by re-
vealing the presence/absence of out

flow

signatures and cold accreted gas.

The AGN host galaxies show no strong

asymmetry when compared with non-AGN
systems, indicating that they reside in rel-
atively undisturbed systems. Contrary to
some early expectations, there also does
not seem to be a connection between
recent strong galaxy mergers and
moderate-luminosity AGN activity, sug-
gesting that SMBH growth is typically
initiated by secular host-galaxy processes
and/or galaxy interactions (e.g., refs. 46

–

49). These results contrast with those
found for rare, optically luminous quasars
(

≈100–1,000 times more luminous than

the typical AGNs in Chandra blank-

field

surveys), which often seem to be asso-
ciated with galaxy major mergers (e.g., ref.
60). These differences imply a change in
the catalyst that drives the fueling of lu-
minous quasars and moderate-luminosity
AGNs, as predicted by some models (e.g.,
ref. 61). However, the fraction of z

≤ 1

galaxies hosting X-ray AGN activity seems
to be enhanced in redshift

filaments

(slightly overdense regions) when com-
pared with

field-galaxy regions, suggesting

that large-scale environment may help
drive SMBH growth (e.g., refs. 59 and 62;
but see ref. 63 for potential host-galaxy
mass dependence). Differences in the
AGN fraction between

field galaxies and

galaxies in distant (proto-)clusters are also
signi

ficant and show that the bulk of

SMBH growth in the densest regions oc-
curred at z

≫ 1 (e.g., refs. 64 and 65).

The star-formation and SMBH-accretion

histories broadly track each other at least out

0.5

1.0

1.5

2.0

2.5

3.0

3.5

log SFR (M

sun

yr

-1

)

0.01

0.10

1.00

f

AG

N

(L

0.

5-

8 k

eV

>

1

0

43

er

g

s

-1

)

Starbursts

LIRGs

ULIRGs

A05b

10

43

erg s

-1

10

43.5

erg s

-1

Fig. 4.

Dependence of the AGN fraction on SFR for

AGNs with 0.5

–8-keV luminosities above 10

43

erg s

−1

;

the dark-purple curve shows the best-estimated
fraction, whereas the light-purple region indicates
the estimated uncertainty. The approximate AGN
fraction for z

≈ 2–3 submm galaxies, from ref. 69, is

shown with the black data point. Approximate SFR
ranges for starburst galaxies, luminous infrared
galaxies (LIRGs), and ultraluminous infrared galaxies
(ULIRGs) are shown along the top. A

“sliding bin”

with a minimum of 10 AGNs was used to construct
this plot; the mean width of this bin is shown in
the lower right-hand corner. The minimum AGN 0.5

–

8-keV luminosity sampled at the minimum and
maximum SFR values is also indicated. From Rafferty
et al. (73). [Reproduced with permission from ref. 73
(Copyright 2009, American Institute of Physics).]

Fig. 3.

One recent example showing the correlation between X-ray-to-optical/UV

flux ratio, α

ox

= 0.3838 log

(L

2 keV

/L

2500 Å

), and the rest-frame 2,500 Å monochromatic luminosity for radio-quiet AGNs; large negative

values of

α

ox

correspond to relatively weak X-ray emission. The different plotted symbols denote the AGN

samples used in the correlation analyses, ranging from local Seyfert galaxies to the most-luminous quasars in
the universe (the small number of downward-pointing arrows denote X-ray upper limits). The

α

ox

-L

2500 Å

relations from refs. 32 and 33 are shown as dotted and solid lines, respectively, and the functional form of the
dotted line is given at the bottom of the top panel. The bottom panel shows residuals about the dotted line.
The overlaid black error bars show, in L

2500 Å

bins, the mean of the residuals and the 3

σ standard deviation of

the mean. Adapted from refs. 32 and 33, where details of the samples and

fitting analyses are provided.

Brandt and Alexander

PNAS

|

April 20, 2010

|

vol. 107

|

no. 16

|

7187

to z

≈ 2 (with an overall offset of a factor of

≈5,000), as expected if the volume-aver-
aged growth of galaxies and SMBHs was
concordant (e.g., refs. 9 and 66). Star for-
mation in galaxies also

“downsizes” in a

qualitatively similar manner to what is seen
for AGNs (e.g., ref. 67 and references
therein). The majority of individual X-ray

–

selected AGNs have star-formation sig-
natures with implied star-formation rates
(SFRs) of

≈1–1,000 M

⊙

yr

−1

(e.g., refs. 68

–

72), although the SFR vs. SMBH-accretion
ratios for individual AGNs can vary by
several orders of magnitude. Using 70-

μm

Spitzer data, Rafferty et al. (73) have stud-
ied the X-ray AGN fraction as a function of
dust-obscured SFR for systems at z

≈ 0.2–

2.5 with L

IR

≈ 10

10

–10

12

L

⊙

. They

find that

the fraction of galaxies hosting X-ray mod-
erate-to-luminous AGN activity increases
as a function of SFR, with an

≈3–40% AGN

fraction for SFRs of

≈30–1,000 M

⊙

yr

−1

(Fig. 4), showing directly that the duty cycle
of moderate-luminosity AGN activity re-
lates to the SFR of the host galaxy. The
average AGN vs. star-formation luminosity
ratios for X-ray AGNs are found to be
broadly consistent with those expected
from the local SMBH

–spheroid mass rela-

tionships, indicating a close connection
between AGN activity and star formation
across a broad range of SFR. However, it is
currently unclear whether the AGN

–star

formation connection is caused by regu-
latory feedback due to out

flows (as adopted

by some galaxy-evolution models) or some
other process.

Some Unresolved Questions and Future
Prospects
This concise review has provided a sam-
pling of some of the signi

ficant discoveries

obtained by Chandra on the growth of
SMBHs over cosmic time. However, many
important questions remain unanswered.
Below we outline several of these along
with prospects for future progress.

Demography.

How many obscured AGNs

are missed even in the deepest X-ray sur-
veys, and what is their contribution to the
growth of SMBHs? The current multi-
wavelength investigations have made great
advances in identifying X-ray undetected
obscured AGNs, but all suffer from sig-
ni

ficant uncertainties (e.g., potential AGN

misidenti

fications, poorly constrained

intrinsic AGN luminosities, and small
numbers of reliable identi

fications).

Ultradeep Chandra and XMM-Newton
exposures, such as the upcoming 4 Ms
Chandra Deep Field-South, will help to
provide improved AGN characterization.
Future sensitive

≈10–200-keV ob-

servations [e.g., with the Nuclear Spectro-
scopic Telescope Array, Astro-H, the
International X-ray Observatory (IXO), and
the Energetic X-ray Imaging Survey Tele-

scope], particularly when allied with im-
proved data from multiwavelength
facilities [e.g., ELTs, the James Webb
Space Telescope (JWST), and Herschel],
will signi

ficantly extend the current census

of SMBH growth in the most
obscured systems.

How do moderate-luminosity (L

X

≈ 10

43

erg s

−1

) AGNs evolve over the important

redshift interval of z

≈ 2–6 and beyond?

Existing deep X-ray surveys already have
the ability to detect high-redshift moderate-
luminosity AGNs, provided their level of
obscuration is not too strong, but it is often
challenging to obtain accurate spectro-
scopic and/or photometric redshifts for
these optically faint X-ray sources. Sig-
ni

ficant advances in redshift determination

can be made, for example with ultradeep
(i.e.,

>8 h) optical spectroscopy using the

largest ground-based telescopes and with
future large-area X-ray-to-millimeter ob-
servatories [e.g., IXO, ELTs, JWST, and the
Atacama Large Millimeter/Submillimeter
Array (ALMA)]. Larger X-ray survey areas
at sensitive

flux levels [e.g., from IXO, the

Extended Roentgen Survey with an Imaging
Telescope Array, and Wide Field X-Ray
Telescope (WFXT) observations] will also
be essential for setting statistically powerful
evolution constraints at the highest red-
shifts (z

≈ 4–10).

Physics.

Are there signi

ficant exceptions to

the rule of universal X-ray emission from
luminous AGNs? Most of the Chandra
AGN survey results are ultimately built
upon the idea that strong underlying X-ray
emission is universally present. However,
there are a small number of apparent X-
ray weak exceptions to this rule that may
be indicative of broader problems (e.g.,
refs. 30 and 74 and references therein).
Surveys for further exceptions are im-
portant so that any foundational cracks
may be identi

fied and patched. These

surveys may also lead to insights about
accretion disks and their coronae. Strange
objects, which persist in showing a type of
spectrum entirely out of keeping with their
luminosity, may ultimately teach us more
than a host, which radiates according to
rule (cf. ref. 75)!

What is the nature of the luminosity

dependence of the X-ray-to-optical/UV
SEDs of AGNs? This (likely nonlinear)
luminosity dependence still needs to be
determined more reliably, because the
current measurements of it quantitatively
disagree and thus cannot effectively guide
the development of physical disk

–corona

models. A key aspect of future work must
be the reduction and realistic quanti

fica-

tion of systematic errors, including AGN
misclassi

fication, detection-fraction

effects, absorption effects, host-galaxy
light contamination, AGN variability, and
luminosity dispersion. It is also critical to

investigate further what practicable ob-
servables of AGN SEDs in the X-ray-to-
optical/UV bandpass provide the most
insight into their accretion processes, the
roles of SMBH mass and Eddington frac-
tion, and possible residual dependences of
X-ray-to-optical/UV SEDs upon redshift.

Ecology.

What are the host-galaxy proper-

ties of typical AGNs at z

> 1.5? Although

much has been revealed about the hosts
of z

< 1.5 X-ray–selected AGNs, com-

paratively little is known about the (po-
tentially more rapidly growing) hosts of
z

> 1.5 X-ray AGNs. Currently, the biggest

hindrance to addressing this question
is the lack of rest-frame optical-to-
near-infrared observations with the requi-
site combination of high sensitivity and
angular resolution. This situation should
signi

ficantly improve with HST/Wide Field

Camera 3 and JWST rest-frame optical-
to-near-infrared observations in
the future.

What is the physical meaning of the

color-magnitude diagram results for AGN
host galaxies? It is currently unclear the
extent to which the green-valley and red-
sequence colors for typical X-ray AGNs at
z

< 1.5 are due to the quenching of star

formation, the rejuvenation of bulge-
dominated systems, dust extinction, biases
in sample construction, or something else.
Spatially resolved spectroscopy of in-
dividual sources can be used to search
for the large-scale out

flow signatures ex-

pected to quench star formation (e.g., refs.
76 and 77), and millimeter spectroscopy
(e.g., with existing facilities or ALMA in
the future) can provide constraints on the
presence of cold molecular gas.

What are the effects of cosmic envi-

ronment, from voids to superclusters, on
the growth of SMBHs? Given the different
evolution of AGNs in (proto-)clusters from
those detected in blank-

field X-ray surveys

(e.g., refs. 64 and 65), it is clear that en-
vironment must play some role in the
growth of SMBHs. X-ray surveys are re-
quired with suf

ficient areal coverage and

sensitivity to identify the AGNs that
dominate the X-ray luminosity function
across the full range of redshifts and en-
vironments. This can be accomplished
with large investments of Chandra and
XMM-Newton time as well as future fa-
cilities (e.g., IXO and WFXT).

ACKNOWLEDGMENTS. We thank all of our collab-
orators on Chandra extragalactic surveys for edu-
cational interactions over the past decade; and M.
Brusa and an anonymous referee for helpful feed-
back on this article. Supported by Chandra Award
SP8-9003A (to W.N.B.), National Aeronautics and
Space Administration Astrophysics Data Analysis
Program (ADP) Grant NNX10AC99G (to W.N.B.), the
Royal Society (D.M.A.), and a Philip Leverhulme
Prize (to D.M.A.).

7188

|

www.pnas.org/cgi/doi/10.1073/pnas.0914151107

Brandt and Alexander

1. Brandt WN, Hasinger G (2005) Deep extragalactic X-ray

surveys. Annu Rev Astron Astrophys 43:827

–859.

2. Boyle BlJ (2001) The evolution of QSOs. Advanced

Lectures on

the

Starburst-AGN

Connection,

eds

Aretxaga I, Kunth D, Muzica R (World Scienti

fic,

Singapore), pp 325

–372.

3. Osmer PS (2004) The evolution of quasars. Coevolution

of Black Holes and Galaxies, ed Ho LC (Cambridge Univ
Press, Cambridge,UK), pp 324

–340.

4. Miyaji T, Hasinger G, Schmidt M (2000) Soft X-ray AGN

luminosity function from ROSAT surveys. I. Cosmological
evolution and contribution to the soft X-ray background.
Astron Astrophys 353:25

–40.

5. Haiman Z, Loeb A (1999) X-ray emission from the

first

quasars. Astrophys J 521:L9

–L12.

6. Bauer FE, et al. (2004) The fall of active galactic nuclei

and the rise of star-forming galaxies: A close look at
the Chandra Deep Field X-ray number counts. Astron J
128:2048

–2065.

7. Hasinger G, Miyaji T, Schmidt M (2005) Luminosity-

dependent evolution of soft X-ray selected AGN. New
Chandra and XMM-Newton surveys. Astron Astrophys
441:417

–434.

8. La Franca F, et al. (2005) The HELLAS2XMM survey. VII.

The hard X-ray luminosity function of AGNs up to z = 4:
More absorbed AGNs at low luminosities and high
redshifts. Astrophys J 635:864

–879.

9. Silverman JD, et al. (2008) The luminosity function of

X-ray-selected active galactic nuclei: Evolution of
supermassive black holes at high redshift. Astrophys J
679:118

–139.

10. Yencho B, Barger AJ, Trouille L, Winter LM (2009) The

OPTX project. II. Hard X-ray luminosity functions of
active galactic nuclei for z

≲ 5. Astrophys J 698:380–396.

11. Aird J, et al. (2010) The evolution of the hard X-ray

luminosity function of AGN. Mon Not R Astron Soc
401:2531

–2551.

12. Fontanot F, et al. (2007) The luminosity function of

high-redshift quasi-stellar objects. A combined analysis
of GOODS and SDSS. Astron Astrophys 461:39

–48.

13. Brusa M, et al. (2009) High-redshift quasars in the

COSMOS survey: The space density of z

> 3 X-ray

selected QSOs. Astrophys J 693:8

–22.

14. Soltan A (1982) Masses of quasars. Mon Not R Astron

Soc 200:115

–122.

15. Marconi A, et al. (2004) Local supermassive black holes,

relics of active galactic nuclei, and the X-ray background.
Mon Not R Astron Soc 351:169

–185.

16. Marconi A, et al. (2006) Local supermassive black holes

and relics of active galactic nuclei. Mem Soc Astron Ital
77:742

–749.

17. Merloni A, Heinz S (2008) A synthesis model for AGN

evolution: Supermassive black holes growth and
feedback modes. Mon Not R Astron Soc 388:1011

–1030.

18. Shankar F, Weinberg DH, Miralda-Escudé J (2009)

Self-consistent models of the AGN and black-hole
populations: Duty cycles, accretion rates, and the
mean radiative ef

ficiency. Astrophys J 690:20–41.

19. Fabian AC, Iwasawa K (1999) The mass density in black

holes inferred from the X-ray background. Mon Not R
Astron Soc 303:L34

–L36.

20. Worsley MA, et al. (2005) The unresolved hard X-ray

background: The missing source population implied by
the Chandra and XMM-Newton deep

fields. Mon Not R

Astron Soc 357:1281

–1287.

21. Alonso-Herrero A, et al. (2006) Infrared power-law

galaxies in the Chandra Deep Field-South: Active
galactic nuclei and ultraluminous infrared galaxies.
Astrophys J 640:167

–184.

22. Daddi E, et al. (2007) Multiwavelength study of

massive galaxies at z

≈ 2. II. Widespread Compton-

thick active galactic nuclei and the concurrent
growth of black holes and bulges. Astrophys J 670:
173

–189.

23. Donley JL, Rieke GH, Pérez-González PG, Rigby JR,

Alonso-Herrero A (2007) Spitzer power-law active
galactic nucleus candidates in the Chandra Deep
Field-North. Astrophys J 660:167

–190.

24. Donley JL, Rieke GH, Pérez-González PG, Barro G

(2008) Spitzer

’s contribution to the AGN population.

Astrophys J 687:111

–132.

25. Alexander DM, et al. (2008) Reliable identi

fication of

Compton-thick quasars at z

∼2: Spitzer mid-infrared

spectroscopy of HDF-oMD49. Astrophys J 687:835

–847.

26. Fiore F, et al. (2008) Unveiling obscured accretion in

the Chandra Deep Field-South. Astrophys J 672:94

–101.

27. Fiore F, et al. (2009) Chasing highly obscured QSOs in

the COSMOS

field. Astrophys J 693:447–462.

28. Treister E, et al. (2009) Heavily obscured AGN in star-

forming galaxies at z

∼2. Astrophys J 706:535–552.

29. Mushotzky

RF

(2004)

How

are

AGN

found?

Supermassive Black Holes in the Distant Universe, ed
Barger AJ (Kluwer Academic Publishers, Dordrecht,
The Netherlands), pp 53

–88.

30. Gibson RR, Brandt WN, Schneider DP (2008) Are

optically selected quasars universally X-ray luminous?
X-ray-UV relations in Sloan Digital Sky Survey quasars.
Astrophys J 685:773

–786.

31. Avni Y, Tananbaum H (1986) X-ray properties of

optically selected QSOs. Astrophys J 305:83

–99.

32. Steffen AT, et al. (2006) The X-ray-to-optical properties

of

optically

selected active

galaxies over

wide

luminosity and redshift ranges. Astrophys J 131:
2826

–2842.

33. Just DW, et al. (2007) The X-ray properties of the most

luminous quasars from the Sloan Digital Sky Survey.
Astrophys J 665:1004

–1022.

34. Kelly BC, Bechtold J, Siemiginowska A, Aldcroft T,

Sobolewska M (2007) Evolution of the X-ray emission
of radio-quiet quasars. Astrophys J 657:116

–134.

35. Kelly BC, Bechtold J, Trump JR, Vestergaard M,

Siemiginowska A (2008) Observational constraints on
the dependence of radio-quiet quasar X-ray emission
on black-hole mass and accretion rate. Astrophys J
176(Suppl):355

–373.

36. Green PJ, et al. (2009) A full year

’s Chandra exposure

on Sloan Digital Sky Survey quasars from the Chandra
Multiwavelength Project. Astrophys J 690:644

–669.

37. Young M, Elvis M, Risaliti G (2010) The X-ray energy

dependence of the relation between optical and X-ray
emission in quasars. Astrophys J 708:1388

–1397.

38. Noble SC, Krolik JH (2009) GRMHD prediction of

coronal variability in accreting black holes. Astrophys
J 703:964

–975.

39. Ueda Y, Akiyama M, Ohta K, Miyaji T (2003)

Cosmological evolution of the hard X-ray active
galactic nucleus luminosity function and the origin of
the hard X-ray background. Astrophys J 598:886

–908.

40. Treister E, Urry CM (2006) The evolution of obscuration

in active galactic nuclei. Astrophys J 652:L79

–L82.

41. Hasinger G (2008) Absorption properties and evolution

of active galactic nuclei. Astron Astrophys 490:
905

–922.

42. Lawrence A, Elvis M (1982) Obscuration and the

various kinds of Seyfert galaxies. Astrophys J 256:
410

–426.

43. Lawrence A (1991) The relative frequency of broad-lined

and narrow-lined active galactic nuclei: Implications for
uni

fied schemes. Mon Not R Astron Soc 252:586–592.

44. Tremaine S, et al. (2002) The slope of the black hole

mass versus velocity dispersion correlation. Astrophys J
574:740

–753.

45. Häring N, Rix H-W (2004) On the black hole mass-bulge

mass relation. Astrophys J 604:L89

–L92.

46. Grogin NA, et al. (2005) AGN host galaxies at z

∼0.4 –

1.3:

Bulge-dominated

and

lacking

merger-AGN

connection. Astrophys J 627:L97

–L100.

47. Pierce CM, et al. (2007) AEGIS: Host galaxy morphologies

of X-ray-selected and infrared-selected active galactic
nuclei at 0.2

≤ z < 1.2. Astrophys J 660:L19–L22.

48. Gabor JM, et al. (2009) Active galactic nucleus host-

galaxy morphologies in COSMOS. Astrophys J 691:
705

–722.

49. Georgakakis A, et al. (2009) Host-galaxy morphologies

of X-ray selected AGN: Assessing the signi

ficance of

different black-hole fuelling mechanisms to the
accretion density of the Universe at z

∼1. Mon Not R

Astron Soc 397:623

–633.

50. Babi

ć A, et al. (2007) Low accretion rates at the AGN

cosmic downsizing epoch. Astron Astrophys 474:
755

–762.

51. Alonso-Herrero A, et al. (2008) The host galaxies and

black holes of typical z

∼ 0.5–1.4 AGNs. Astrophys J

677:127

–136.

52. Brusa M, et al. (2009) Black-hole growth and starburst

activity at z = 0.6

–4 in the Chandra Deep Field-South:

Host-galaxy properties of obscured AGN. Astron
Astrophys 507:1277

–1289.

53. Ballo L, et al. (2007) Black-hole masses and Eddington

ratios of AGNs at z

< 1: Evidence of retriggering for a

representative sample of X-ray-selected AGNs. Astrophys
J 667:97

–116.

54. Hickox RC, et al. (2009) Host galaxies, clustering,

Eddington ratios, and evolution of radio, X-ray, and
infrared-selected AGNs. Astrophys J 696:891

–919.

55. Alexander DM, et al. (2008) Weighing the black holes

in z

≈ 2 submillimeter-emitting galaxies hosting active

galactic nuclei. Astron J 135:1968

–1981.

56. Yamada T, et al. (2009) MOIRCS Deep Survey III. Active

galactic nuclei in massive galaxies at z = 2

–4. Astrophys

J 699:1354

–1364.

57. Shi Y, et al. (2008) Black-hole accretion in low-mass

galaxies since z

∼1. Astrophys J 688:794–806.

58. Nandra K, et al. (2007) AEGIS: The color-magnitude

relation for X-ray-selected active galactic nuclei.
Astrophys J 660:L11

–L14.

59. Silverman JD, et al. (2008) The evolution of AGN host

galaxies: From blue to red and the in

fluence of large-

scale structures. Astrophys J 675:1025

–1040.

60. Canalizo G, Stockton A (2001) Quasi-stellar objects,

ultraluminous infrared galaxies, and mergers. Astrophys
J 555:719

–743.

61. Hopkins PF, Hernquist L (2009) A characteristic division

between the fueling of quasars and Seyferts: Five
simple tests. Astrophys J 694:599

–609.

62. Gilli R, et al. (2003) Tracing the large-scale structure in

the Chandra Deep Field-South. Astrophys J 592:
721

–727.

63. Georgakakis A, et al. (2007) AEGIS: The environment of

X-ray sources at z

≈ 1. Astrophys J 660:L15–L18.

64. Lehmer BD, et al. (2009) The Chandra Deep Protocluster

Survey: Evidence for an enhancement of AGN activity in
the SSA22 protocluster at z = 3.09. Astrophys J 691:
687

–695.

65. Martini P, Sivakoff GR, Mulchaey JS (2009) The

evolution of active galactic nuclei in clusters of
galaxies to redshift 1.3. Astrophys J 701:66

–85.

66. Merloni A, Rudnick G, Di Matteo T (2004) Tracing

the cosmological assembly of stars and supermassive
black holes in galaxies. Mon Not R Astron Soc 354:
L37

–L42.

67. Damen M, et al. (2009) The evolution of the speci

fic

star formation rate of massive galaxies to z

∼1.8 in the

Extended Chandra Deep Field-South. Astrophys J 690:
937

–943.

68. Alexander DM, et al. (2005) The X-ray spectral prop-

erties of SCUBA galaxies. Astrophys J 632:736

–750.

69. Alexander DM, et al. (2005) Rapid growth of black

holes in massive star-forming galaxies. Nature 434:
738

–740.

70. Pope A, et al. (2008) Mid-infrared spectral diagnosis of

submillimeter galaxies. Astrophys J 675:1171

–1193.

71. Murphy EJ, et al. (2009) Balancing the energy budget

between star formation and active galactic nuclei in
high-redshift infrared luminous galaxies. Astrophys J
698:1380

–1397.

72. Silverman JD, et al. (2009) Ongoing and co-evolving

star formation in zCOSMOS galaxies hosting active
galactic nuclei. Astrophys J 696:396

–410.

73. Rafferty DA, et al. (2009) Supermassive black-hole

growth in starburst galaxies: Constraints from the
deepest Chandra

fields. The Monster’s Fiery Breath:

Feedback in Galaxies, Groups, and Clusters, eds Heinz
S, Wilcots EM (AIP Press, Melville, NY), pp 150

–153.

74. Leighly K, Halpern JP, Jenkins EB, Casebeer D (2007)

The intrinsically X-ray-weak quasar PHL 1811. II.
Optical and UV spectra and analysis. Astrophys J
173(Suppl):1

–36.

75. Eddington A (1922) President

’s introductory address:

Stellar socialism vs. stellar individualism. Mon Not R
Astron Soc 82:432

–438.

76. Nesvadba NPH, Lehnert MD, De Breuck C, Gilbert AM,

van Breugel W (2008) Evidence for powerful AGN
winds at high redshift: Dynamics of galactic out

flows

in radio galaxies during the

“Quasar Era”. Astron

Astrophys 491:407

–424.

77. Alexander DM, Swinbank AM, Smail I, McDermid R,

Nesvadba

N

(2010)

Searching

for

evidence

of

energetic feedback in distant galaxies: A galaxy wide
out

flow in a z ∼2 ultraluminous infrared galaxy. Mon

Not Roy Ast Soc, in press.

Brandt and Alexander

PNAS

|

April 20, 2010

|

vol. 107

|

no. 16

|

7189