Defining the intrinsic AGN infrared spectral energy distribution and measuring its contribution to the infrared output of composite galaxies

Abstract

We use infrared spectroscopy and photometry to empirically define the
intrinsic, thermal infrared spectral energy distribution (i.e., 6-100 um SED)
of typical active galactic nuclei (i.e., 2-10 keV luminosity,
Lx=10^{42}-10^{44} ergs/s AGNs). On average, the infrared SED of typical AGNs
is best described as a broken power-law at <40 um that falls steeply at >40um
(i.e., at far-infrared wavelengths). Despite this fall-off at long wavelengths,
at least 3 of the 11 AGNs in our sample have observed SEDs that are
AGN-dominated even at 60 um, demonstrating the importance of accounting for
possible AGN contribution even at far-infrared wavelengths. Our results also
suggest that the average intrinsic AGN 6-100 um SED gets bluer with increasing
X-ray luminosity, a trend seen both within our sample and also when we compare
against the intrinsic SEDs of more luminous quasars (i.e., Lx>10^{44} ergs/s).
We compare our intrinsic AGN SEDs with predictions from dusty torus models and
find they are more closely matched by clumpy, rather than continuous, torus
models. Next, we use our intrinsic AGN SEDs to define a set of correction
factors to convert either monochromatic infrared or X-ray luminosities into
total intrinsic AGN infrared (i.e., 8-1000 um) luminosities. Finally, we
outline a procedure that uses our newly defined intrinsic AGN infrared SEDs, in
conjunction with a selection of host-galaxy templates, to fit the infrared
photometry of composite galaxies and measure the AGN contribution to their
total infrared output. We verify the accuracy of our SED fitting procedure by
comparing our results to two independent measures of AGN contribution. Our SED
fitting procedure opens up the possibility of measuring the intrinsic AGN
luminosities of large numbers of galaxies with well-sampled infrared data
(e.g., IRAS, ISO, Spitzer and Herschel).

Full text

Page 1

arXiv:1102.1425v1 [astro-ph.CO] 7 Feb 2011
Mon. Not. R. Astron. Soc. 000, 1–23 (2010)Printed 9 February 2011(MN LATEX style file v2.2)
Defining the intrinsic AGN infrared spectral energy distribution and
measuring its contribution to the infrared output of composite
galaxies⋆
J. R. Mullaney1,2†, D. M. Alexander1, A. D. Goulding1,3and R. C. Hickox1
1Department of Physics, Durham University, South Road, Durham, DH1 3LE, U.K.
2Laboratoire AIM-Paris-Saclay, CEA/DSM/Irfu - CNRS, Universit´ e Paris Diderot, CE-Saclay, pt courrier 131, 91191 Gif-sur-Yvette, France
3Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, U.S.
Date Accepted
ABSTRACT
We use infrared spectroscopy and photometry to empirically define the intrinsic, thermal in-
frared spectral energy distribution (i.e., 6-100 µm SED) of typical active galactic nuclei (i.e.,
2-10keV luminosity,L2−10keV∼1042−1044ergs s−1AGNs). On average,the infraredSED of
typicalAGNs is best describedas a brokenpower-lawat ?40µm that falls steeply at ?40µm
(i.e., at far-infraredwavelengths). Despite this fall-off at long wavelengths, at least 3 of the 11
AGNs in our sample have observed SEDs that are AGN-dominated even at 60 µm, demon-
strating the importance of accounting for possible AGN contribution even at far-infrared
wavelengths. Our results also suggest that the average intrinsic AGN 6-100 µm SED gets
bluer with increasingX-ray luminosity,a trend seen bothwithin our sample and also when we
compare against the intrinsic SEDs of more luminous quasars (i.e., L2−10keV? 1044ergs s−1).
We compare our intrinsic AGN SEDs with predictions from dusty torus models and find they
are more closely matched by clumpy, rather than continuous, torus models. Next, we use our
intrinsic AGN SEDs to define a set of correction factors to convert either monochromatic in-
frared or X-ray luminosities into total intrinsic AGN infrared (i.e., 8-1000 µm) luminosities.
Finally, we outline a procedure that uses our newly defined intrinsic AGN infrared SEDs, in
conjunction with a selection of host-galaxy templates, to fit the infrared photometry of com-
posite galaxies and measure the AGN contribution to their total infrared output. We verify the
accuracy of our SED fitting procedure by comparing our results to two independentmeasures
of AGN contribution: (1) 12 µm luminosities obtained from high-spatial resolution observa-
tions of nearby galaxies and (2) the equivalent width of the 11.25 µm PAH feature. Our SED
fitting procedureopens upthe possibility of measuringthe intrinsic AGN luminosities of large
numbers of galaxies with well-sampled infrared data (e.g., IRAS, ISO, Spitzer and Herschel).
Key words: Galaxies, Seyfert, Active, Quasars, Infrared, X-rays
1INTRODUCTION
The spectral energy distribution (hereafter, SED) of a continuum
source is the description of its energy output as a function of pho-
ton frequency or wavelength. As such, it is one of the most impor-
tant measurables in astronomy, providing information on both the
physical nature of a continuum source (e.g., stars, galaxies, active
galacticnuclei, heated dust), itsinfluence onthe surrounding matter
(i.e., heating, ionisation state) and, when integrated over all wave-
lengths, its bolometric luminosity (i.e., power output). However,
⋆The intrinsic AGN infrared SEDs, host-galaxy templates and a pro-
cedure used to combine them to fit infrared photometry is available at
http://sites.google.com/site/decompir
† E-mail: james.mullaney@cea.fr
deriving an SED is a challenging task, requiring multiple observa-
tions across the whole of the electromagnetic spectrum and, in the
case of active galactic nuclei (hereafter, AGNs), one that is further
complicated by contamination from the ever-present host galaxy.
The level of host-galaxy contamination to the observed AGN
SED is typically highest at infrared wavelengths (i.e., 8-1000 µm)
where the strongly rising host-galaxy SED typically dominates
(e.g. Elvis et al. 1994; Richards et al. 2006; Netzer et al. 2007). In-
deed, this contamination is so severe that the intrinsic AGN SED at
these wavelengths remains largely unconstrained by observations
(except for rare quasar-luminosity AGNs; see later). The signifi-
cance of the uncertainties surrounding the intrinsic AGN infrared
SED is clear when we consider that the infrared portion of the
average observed (i.e., AGN + host) AGN SED represents ?50
per cent of the total radio – X-ray output (e.g., Elvis et al. 1994).

Page 2

2
J.R. Mullaney et al
Physical models of the infrared-emitting dust surrounding the AGN
can provide some insights (e.g., Siebenmorgen & Kruegel 1992;
Fritz et al. 2006; Schartmann et al. 2008), although they tend to
predict a very broad range of infrared AGN SEDs as their input
parameters are typically poorly constrained (see our §5.3); a situa-
tionthat willbe helped withabetter understanding of thetrue range
of intrinsic AGN infrared SEDs.
Well-defined intrinsic AGN infrared SEDs may also prove a
useful tool in the measurements of AGN and star-formation activ-
ity that are crucial to our understanding of the interactions between
these two processes that are believed to exist (e.g., Magorrian et al.
1998; Ferrarese & Merritt 2000; Gebhardt et al. 2000; Croton et al.
2006; Bower et al. 2006; Hopkins et al. 2006, 2007). It has al-
ready been demonstrated that the intrinsic mid-infrared emission
of AGNs is closely linked to the total AGN luminosity (e.g.,
Horst et al. 2008; Gandhi et al. 2009), while the far-infrared emis-
sion of a pure star-forming galaxy provides a proxy measure of
its star formation rate (e.g., Kennicutt 1998). Separating the total
infrared SED of a composite (i.e., AGN + star-forming) galaxy
into AGN and host components would therefore provide a mea-
sure of the levels of AGN and star-formation activity that is largely
unaffected by obscuration or absorption; similar, in principle, to
the infrared spectral decompositions carried out by, for example,
Laurent et al. (2000),Tran et al. (2001) and Lutz et al. (2004). The
benefit of using the whole infrared SED is that this technique is
not restricted to the small fraction of galaxies for which infrared
spectra are available. Furthermore, such techniques are likely to
become increasingly popular with the availability of infrared pho-
tometry measurements from the Spitzer and Herschel telescopes.
However, most studies that have used SED decomposition ap-
proaches have had to rely on either observed AGN SEDs that
likely include considerable amounts of host-galaxy contamination
at infrared wavelengths or predictions from radiative transfer mod-
els which are often poorly constrained (e.g. Polletta et al. 2007;
Fiore et al. 2008, 2009; Pozzi et al. 2010; Hatziminaoglou et al.
2010). Of course, measurements derived from decomposing the in-
frared SED should be regarded as complementary to other mea-
sures of AGN activity that rely on other observables (e.g., X-ray lu-
minosity, optical and infrared spectroscopy, mid-infrared photome-
try etc.; Kauffmann et al. 2003; Heckman et al. 2004; Hickox et al.
2007, 2009; Goulding & Alexander 2009).
In their 2007 paper, Netzer et al. derived the average intrin-
sic infrared SED of a sample of luminous PG quasars by sub-
tracting the host-galaxy component from the average observed
SED, showing that their average intrinsic AGN SED falls rapidly
longward of ∼ 20 µm. However, such luminous quasars are rare
among the AGN population and it is not clear whether the in-
trinsic quasar SED can be applied to the much larger, less lu-
minous (i.e., L2−10keV< 1044ergs s−1) population of AGNs
which represents the majority of the integrated accretion output of
AGNs in the Universe (e.g., Ueda et al. 2003; Barger et al. 2005;
Hasinger et al. 2005; Aird et al. 2010). In this paper we extend
the work of Netzer et al. (2007) to include more typical, lower lu-
minosity AGNs and demonstrate that the intrinsic AGN infrared
SED appears to be linked to AGN luminosity. To do this, we first
identify a sample of AGNs that show minimal amounts of host-
galaxy contamination at MIR wavelengths, then carefully subtract
any host-galaxy component that can still dominate at far-infrared
wavelengths (see §2, §3). We compare the derived intrinsic AGN
SEDs against other, more commonly assumed, AGN SEDs includ-
ing those derived from quasars and radiative transfer models in §5.
In §6 weintroduce a procedure that uses these intrinsicAGN SEDs,
Figure 1. The distribution of intrinsic 2-10 keV luminosities (i.e., L2−10keV)
and hydrogen column densities (i.e., NH) of the Swift-BAT sample of AGNs
which we use to define the intrinsic AGN infrared SED (circles; taken from
Winter et al. 2009). In the background (small squares) we also show the
L2−10keV–NHdistribution of galaxies in the Chandra Deep Field-South (i.e.,
CDF-S; Tozzi et al. 2006). The intrinsic properties of these local (i.e., Swift-
BAT) and high-redshift (i.e., CDF-S) samples of galaxies are well matched.
We highlight those 25 and 11 AGNs (see key) that we use to define the
range of intrinsic AGN mid-infrared and mid to far-infrared SEDs. These
subsamples cover a similar range of L2−10keV–NHparameter space spanned
by the Swift-BAT and CDF-S samples. See online manuscript for a colour
version of this plot.
together with a sample host-galaxy templates, to measure the AGN
and host-galaxy contributions to the infrared output of a sample of
local composite galaxies and demonstrate theobtained values agree
withresultsfromother, independent measures. In§7 weoutlinetwo
obvious applications of these analyses: (1) defining correction fac-
tors to convert 12 µm and X-ray luminosities to total intrinsic AGN
infrared luminosities and (2) measuring the intrinsic AGN power
using infrared photometry alone. We summarise our findings in §8.
Throughout this work we adopt H0= 71 km s−1Mpc−1,
ΩM= 0.27, and ΩΛ= 0.73.
2SAMPLE DESCRIPTION
The first step we take in defining the intrinsic infrared SEDs
of typical AGNs is to identify a sample of moderate luminosity
(i.e., L2−10keV∼ 1042−1044ergs s−1) AGNs that suffer minimal
amounts of host galaxy contamination at infrared wavelengths. By
doing this, we only focus on those SEDs that require as little ma-
nipulation as possible to extract the intrinsic AGN infrared SED. In
this section, we describe how combine information derived from
hard X-ray observations and infrared spectra to produce such a
sample. We then explain how we use IRAS photometry to extrapo-
late the SEDs of a subsample of these AGNs to FIR wavelengths.
The Burst Alert Telescope (hereafter, BAT) on-board the Swift
observatory is currently undertaking a survey of the sky at hard
X-ray energies (i.e., 14-195 keV), where the effects from ab-
sorption (at least to NH< 1024cm−2) are negligible. As a re-
sult, the second data release of the Swift-BAT survey, published

Page 3

in Tueller et al. (2008), provides a homogeneous sample of 154
local (i.e., z < 0.1) X-ray AGNs. A key motivation for using the
AGNs selected from the Swift-BAT sample to define the intrinsic
AGN infrared SED is that their X-ray properties (i.e., 2-10 keV X-
ray luminosities and absorbing columns, hereafter, L2−10keVand
NH) cover largely the same ranges as those AGNs detected in
deep Chandra surveys (e.g., CDF-N and CDF-S; see fig. 1). As
such, this sample provides the ideal local analogue to the high red-
shift AGNs detected in these fields as well as typical local AGNs
(i.e., L2−10keV∼ 1042− 1044ergs s−1and NH< 1024cm−2; see
Winter et al. 2009). Furthermore, in cases where infrared photom-
etry is available, the Swift-BAT AGNs cover the same range of
infrared luminosities (integrated over 8-1000 µm; hereafter LIR;
LIR= 6×109−7×1011L⊙) as those star-forming galaxies that
make up the majority of the local galaxy population detected by
IRAS (i.e., LIR∼ 1010−1012L⊙) and z < 2 galaxies detected in
deep Spitzer/Herschel fields (e.g. GOODS; P.I.s: M. Dickinson, D.
Elbaz).
In order to identify a sample of galaxies whose infrared out-
put is dominated by a moderate luminosity AGN, we cross-match
the Swift-BAT sample with the archive of low resolution spectra
obtained by the infrared spectrograph (hereafter, IRS) on-board
Spitzer. Of the 104 Swift-BAT AGNs with measured L2−10keV
(see Winter et al. 2009), 36 have publicly available archival low
resolution IRS spectra covering (observed frame) ∼ 6− 35 µm.
These spectra were reduced and extracted following the proce-
dures outlined in Mullaney et al. (2010), Goulding et al. (2010) and
Goulding (2010; PhD thesis). Following the diagnostics presented
in Goulding & Alexander (2009) and Tommasin et al. (2010), we
quantify the contribution of the host galaxy to these MIR spectra
using the equivalent widths of the 11.25 µm PAH feature (here-
after, EW PAHλ11.25). We decide to use the 11.25 µm feature
to measure the host galaxy contribution rather than the typically
stronger one at 7.7 µm as this latter feature is often blended with
another PAH emission line at 8.6 µm. We measure EW PAHλ11.25
using the spectral fitting code PAHFIT (Smith et al. 2007) and as-
sume that strongly AGN dominated systems have EW PAHλ11.25<
0.03 µm, corresponding to a < 10 per cent host-galaxy contribu-
tion at 19 µm (see Tommasin et al. 2010).1Of the 36 X-ray AGNs
that have archival low resolution IRS spectra, we identify 25 whose
EW PAHλ11.25satisfy this criterion (see table 1 for alist of these 25
AGN-dominated galaxies; also fig. 2). Approximately half of these
AGNs (i.e., 12/25) are optically classified as either Type 1, 1.2 or
1.5 while the rest (i.e., 13/25) are classified at either Type 1.8, 1.9
or 2 (optical classifications taken from Tueller et al. 2008; see table
1 for these classifications and other information regarding these 25
AGN dominated galaxies). We visually inspect the infrared spectra
of each of these 25 AGNs to confirm that they are, indeed, domi-
nated by a featureless AGN continuum.
To provide constraints on the intrinsicAGN SED at FIRwave-
lengths, weextrapolatebeyond thelow resolution IRSspectra using
60 µm and 100 µm photometry available from the IRAS archives
(accessed through the NASA extragalactic database; NED). To en-
sure reliable extrapolation to FIR wavelengths, we only consider
photometry greater than 0.2 Jy and 1.0 Jy at 60 µm and 100 µm,
respectively (the approximate limiting sensitivity of the IRAS faint
source catalogue at these wavelengths). Of the 25 AGN-dominated
galaxies identified above, 20 satisfy these criteria. However, we
note that aperture effects must be taken into consideration when
1PAHFITisavailable fromhttp://tir.astro.utoledo.edu/jdsmith/research/pahfit.php
collating photometric and spectral data obtained by the IRAS and
Spitzer telescopes. The apertures used to measure the IRAS pho-
tometry reported in the point source and faint source catalogues
were between 2′and 9′in diameter, compared to the 3.6′′and 10.5′′
(shortestdimension) aperturesoftheshort-low andlong-low slitsof
theSpitzer-IRS instrument, respectively. Therefore, there isconsid-
erable scope for the IRAS photometry to contain a great deal more
host galaxy flux than the small aperture IRS spectra. To mitigate
such aperture effects, we only extrapolate to the 60 µm and 100 µm
photometry values when the flux density in the IRS spectrum, in-
tegrated over the 12 and 25 µm IRAS band passes, matches (to
within 30 per cent, including errors) the IRAS photometry at these
wavelengths. Our choice of 30 per cent accuracy between the IRAS
and IRS fluxes is based on the approximate systematic uncertainty
of the IRAS point source and faint source catalogues when com-
pared to the Revised Bright Galaxy Sample (Sanders et al. 2003).
We identify 11 AGN-dominated galaxies that satisfy these crite-
ria (indicated by a “Y” in column 16 of table 1), which we use
to define the intrinsic AGN SED. The 6-100 µm SEDs of these
AGN-dominated galaxies, including their 60 µm and 100 µm pho-
tometries, are presented in appendix A.

Page 4

Figure 2. The 25 AGNs from the Swift-BAT sample that are strongly AGN dominated at mid-infrared wavelengths, separated in terms of their optical
classification (here, Types 1.2 and 1.5 are classed as Type 1 AGN, Types 1.8 and 1.9 are classed as Type 2 AGNs). We have indicated the positions of the
most common, prominent emission lines, PAH features and silicate emission/absorption features. The criterion we use to determine whether the mid-infrared
SED is AGN-dominated is that the equivalent width of the 11.25 µm PAH feature is less than 0.03 µm, which is equivalent to < 10 per cent host-galaxy
contribution at 19 µm (e.g. Tommasin et al. 2010). Within this subsample we see a wide range of mid-infrared SED shapes. However, in the majority of cases
(i.e., at least 20/25; asterisked) the underlying AGN continuum can be described as an absorbed broken power law with a higher (i.e., more positive) spectral
index longward of a break at, roughly, 19 µm. In general, Type 2 AGNs have steeper SEDs (i.e., high spectral indices) at λ ? 19 µm although, in this respect,
there is considerable overlap between the two classes. Type 2 AGNs are more likely to show evidence of strong silicate absorption at ∼10 µm, while a larger
proportion of Type 1 AGNs have silicate in emission at ∼10 µm and ∼18 µm. The galaxy names have been truncated to enable them to fit on the plot. See
table 1 for the full galaxy names.

Page 5

Table 1. Properties of the 25 AGNs in the Swift-BAT sample that are strongly AGN dominated at 6-35 µm.
Name TypeRA Dec.Dist.
(Mpc)
(5)
S12
(Jy)
(6)
S25
(Jy)
(7)
S60
(Jy)
(8)
S100
(Jy)
(9)
SIRS
12
(Jy)
(10)
SIRS
25
(Jy)
(11)
νLν(60µm)
log(L⊙)
(12)
LIR
L2−10keV
log(ergs s−1)
(14)
EW PAHλ11.25
(µm)
(15)
Selected
log(L⊙)
(13) (1) (2)(3) (4)(16)
Mrk348
NGC985
NGC1275
ESO548-G081
3C120
NGC2110
EXO055620-3820.2
Mrk3
MCG-01-24-012
Mrk417
NGC3783
NGC4507
ESO506-G027
MCG-03-34-064
MCG-06-30-015
IC4329A
NGC5506
NGC5548
Mrk290
ESO103-035
3C390.3
Mrk509
IC5063
NGC7213
NGC7314
2
1
2
1
1
2
1
2
2
2
1
2
2
12.1964
38.6574
49.9507
55.5155
68.2962
88.0474
89.5083
93.9015
140.1927
162.3789
174.7572
188.9026
189.7275
200.6019
203.9741
207.3304
213.3119
214.4981
233.9682
279.5847
280.5375
311.0406
313.0097
332.3177
338.9426
31.9570
-8.7876
41.5117
-21.2444
5.3543
-7.4562
-38.3346
71.0375
-8.0561
22.9644
-37.7386
-39.9093
-27.3078
-16.7286
-34.2956
-30.3096
-3.2075
25.1368
57.9026
-65.4276
79.7714
-10.7235
-57.0688
-47.1667
-26.0502
63.4
185.3
74.2
61.0
141.2
32.6
144.9
56.9
83.1
140.1
38.51
49.6
106.3
69.8
32.5
67.7
28.71
72.5
126.1
55.9
244.3
147.3
47.7
22.01
19.01
0.308 ± 0.031
0.207 ± 0.031
1.060 ± 0.014
0.249 ± 0.057
0.289 ± 0.043
0.349 ± 0.028
0.529 ± 0.032
0.713 ± 0.050
0.080 ± 0.025
0.132 ± 0.033
0.840 ± 0.059
0.517 ± 0.078
0.148 ± 0.030
0.940 ± 0.040
0.380 ± 0.034
1.080 ± 0.054
1.290 ± 0.027
0.401 ± 0.040
<0.1
0.612 ± 0.043
0.128 ± 0.018
0.316 ± 0.028
1.110 ± 0.023
0.606 ± 0.049
0.268 ± 0.035
0.835 ± 0.025
0.523 ± 0.017
3.440 ± 0.027
0.097 ± 0.075
0.624 ± 0.094
0.840 ± 0.021
0.685 ± 0.034
2.900 ± 0.036
0.3810 ± 0.0088
0.227 ± 0.054
2.490 ± 0.050
1.59 ± 0.24
0.268 ± 0.022
2.970 ± 0.045
0.809 ± 0.027
2.210 ± 0.054
4.170 ± 0.056
0.769 ± 0.032
<0.1
2.360 ± 0.031
0.2870 ± 0.0089
0.702 ± 0.022
3.940 ± 0.030
0.742 ± 0.036
0.579 ± 0.048
1.29 ± 0.12
1.38 ± 0.11
6.990 ± 0.042
0.601 ± 0.048
1.38 ± 0.21
4.13 ± 0.21
0.322 ± 0.035
3.77 ± 0.15
0.654 ± 0.059
0.164 ± 0.043
3.26 ± 0.19
4.69 ± 0.70
0.498 ± 0.050
6.200 ± 0.040
1.090 ± 0.076
2.03 ± 0.10
8.420 ± 0.060
1.070 ± 0.086
0.171 ± 0.029
2.31 ± 0.12
0.204 ± 0.033
1.360 ± 0.068
5.870 ± 0.038
2.67 ± 0.16
3.74 ± 0.22
1.55 ± 0.20
1.89 ± 0.15
7.20 ± 0.47
1.49 ± 0.18
1.94 ± 0.29
5.7 ± 1.4
0.56 ± 0.12
3.36 ± 0.44
1.10 ± 0.18
0.71 ± 0.15
4.90 ± 0.54
6.28 ± 0.94
0.83 ± 0.19
6.20 ± 0.14
1.10 ± 0.22
1.66 ± 0.22
8.87 ± 0.11
1.61 ± 0.16
<0.6
1.05 ± 0.26
<0.6
1.52 ± 0.23
4.25 ± 0.21
8.18 ± 0.41
14.2 ± 1.3
0.1
0.1
0.7
0.2
0.3
0.3
0.3
0.6
0.1
0.1
0.6
0.6
0.2
0.7
0.2
0.9
1.0
0.2
0.1
0.5
0.1
0.3
0.9
0.2
0.1
0.4
0.3
2.8
0.3
0.8
0.8
0.6
2.3
0.4
0.3
1.9
1.5
0.4
2.5
0.5
2.0
3.3
0.6
0.2
2.1
0.3
0.8
3.1
0.5
0.5
9.91
10.87
10.78
9.54
10.63
9.83
10.02
10.28
9.85
9.70
9.88
10.25
9.94
10.67
9.25
10.16
10.03
9.94
9.63
10.05
10.28
10.66
10.32
9.30
9.32
10.5
11.3
11.2
10.2
11.1
10.2
11.2
10.8
10.4
10.7
10.5
10.7
10.5
11.1
9.9
10.9
10.5
10.6
<10.5
10.7
<11.2
11.2
10.8
9.9
9.8
42.6
44.3
43.9
42.8
44.0
42.5
43.7
42.9
43.0
43.3
43.1
43.0
43.2
42.8
42.6
43.6
42.8
43.5
43.2
42.7
44.1
44.0
42.7
42.3
42.2
0.002
0.016
-
0.015
0.008
0.023
-
-
0.006
0.004
0.002
0.013
-
0.005
0.004
-
0.011
0.017
-
-
0.002
0.023
0.000
0.019
0.015
-
-
-
-
Y
Y
-
Y
Y
-
-
Y
-
Y
-
Y
Y
-
-
Y
-
Y
Y
-
-
1.8
1.2
1.2
1.9
1.5
1
2
1
1.2
2
1.5
1.9
NOTES:(1) Common name, (2) Optical Class from Tueller et al. (2008) (3) and (4) Galaxy co-ordinates (J2000) from Tueller et al. (2008), (5) Distance in Mpc,1denotes that a redshift-independent distance measure
taken from NED is used, otherwise calculated from the redshift reported in Tueller et al. (2008) (6) to (9) IRAS photometry measurements, taken from NED (10) and (11) 12 µm and 25 µm flux densities derived
from the low resolution IRS spectra, (12) Monochromatic luminosity at 60 µm in units of log(L⊙) (for comparison with Netzer et al. 2007). We use L⊙= 3.83×1033ergs s−1. (13) Total infrared luminosity in units
of log(L⊙), calculated using the equations in Table 1 of Sanders & Mirabel (1996), (14) Rest-frame 2-10 keV luminosity, corrected for absorption (taken from Winter et al. 2009) (15) Equivalent width of the the
11.25 µm PAH (i.e., EW PAHλ11.25) feature in µm. (16) Selected for extrapolation to 100 µm based on whether the IRAS photometry at 12 and 25 µm matches that derived from the IRS spectra.

Page 6

3DEFINING A SET OF HOST-GALAXY TEMPLATES
The sample of AGNs identified in the previous section show
minimal amounts of host-galaxy contamination at MIR wave-
lengths. However, because a typical host-galaxy SED rises strongly
at longer infrared wavelengths, an SED that is strongly AGN-
dominated at MIR wavelengths may not necessarily be AGN-
dominated at FIR wavelengths. Indeed, Netzer et al. (2007) showed
that at 60 µm the average quasar SED is 80 to 90 per cent host-
galaxy dominated, despite having only weak PAH features at MIR
wavelengths. In this section we describe the construction of a
set of host-galaxy templates that represent the full diversity of
host-galaxy SEDs expected for typical AGNs (i.e., the full range
of IRAS colour-colour space for LIR= 1010−1012L⊙galaxies).
These templates are used to account for the host-galaxy contribu-
tion to our AGN-dominated sample in §4.
In fig. 3 (left panel) we plot the [100 µm/60 µm] vs.
[25 µm/12 µm] colours of all LIR= 1010− 1012L⊙ galaxies in
the Revised Bright Galaxy Survey. From this figure it is clear that
the majority of the colour-colour space spanned by these galax-
ies is well sampled by the Brandl et al. (2006) sample of starburst
galaxies. However, we note that there is a region of this colour-
colour space that isnot sampled by these starbursts. Galaxies inthis
un-sampled region of the colour-colour space have SEDs that rise
steeply between 60 µm and 100 µm, but are relatively flat between
12 µm and 25 µm. As such, they likely represent a population of
cold galaxies. To construct our host-galaxy templates we use a se-
lection of Brandl et al. (2006) starbursts and a selection of galaxies
that lie in this “cold” region of the IRAS colour-colour space. This
ensures that we cover the full range of SEDs displayed by LIR=
1010−1012L⊙galaxies. We note that the both the colour-colour
and [60 µm/25 µm]-LIRparameter spaces are particularly well-
sampled in the infrared luminosity range spanned by the AGNs se-
lected for decomposition (i.e., log(LIR/L⊙)= 10.2−11.2 L⊙; fig. 3,
right panel; see §4.2).
We extracted and reduced the low resolution IRS spectra of all
the Brandl et al. (2006) starbursts and four additional “cold” galax-
ies (namely, NGC 1667, NGC 5734, NGC 6286 and NGC 7590),
following the same procedures as those referred to in §2. All of the
Brandl et al. (2006) starbursts were observed in IRS staring mode,
whilethe four “cold” galaxies were observed inmapping mode. We
note that these differences in the observing modes will have littleor
noeffect on our host-galaxy templates asour selectionprocess min-
imisesapertureeffects(seelater).Weextrapolated theseMIR SEDs
to 100 µm using IRAS photometry as described in §2. To mitigate
aperture effects we only use those SEDs in which the IRS spectra
agree with the 12 µm and 25 µm IRAS photometry to within 30 per
cent. Out of the 16 Brandl et al. (2006) starbursts identified as ei-
ther pure starburstsor starburst+LINER, we identify 10 that satisfy
these criteria: Mrk 52, Mrk 520, NGC 660, NGC 1222, NGC 2623,
NGC 3256, NGC 4194, NGC 4818, NGC 7252, NGC 7714. All
four of the “cold” galaxies also satisfy these criteria. To produce
our host-galaxy templates we group these SEDs in terms of their
overall shapeand therelativestrength oftheir PAHfeatures(seefig.
4). The full range of host-galaxy SED shapes is well characterised
by five groups of SEDs, referred to as “SB1” through “SB5”, with
all four “cold” SEDs contained within a single group (i.e., “SB1”;
see table 2 and fig. 4). Five host-galaxy templates are obtained by
normalising each SED within each group at 90 µm and calculating
their mean average. Since the SEDs within each group are so sim-
ilar, the wavelength at which they are normalised has little effect
on these host-galaxy templates. Beyond 100 µm we extrapolate the
average SEDs as a modified black body (i.e., Fν= FBB
FBB
ν
is the blackbody specific flux and ν is photon frequency; we
adopt β=1.5). We verify that this is a reasonable assumption using
AKARI 140 µm and 160 µm photometry data available for 13 of
the 14 host-galaxy SEDs in our sample.2These templates are used
to remove any host-galaxy contribution to the infrared SED of the
AGN-dominated galaxy sample. We publish the host-galaxy tem-
plates in columns 5-9 of table 3 (which is available in its entirety
online at http://sites.google.com/site/decompir).
ν
νβ, where
4 CHARACTERISING THE INTRINSIC INFRARED AGN
SED
First we use the IRS spectra of our AGN sample (defined in §2.1)
to explore the range of AGN-dominated infrared SEDs at 6-35 µm.
We then extract the intrinsic 6-100 µm SEDs of a subsample of
theseAGNsbycarefullydecomposing theobserved SEDsintotheir
host-galaxy (defined in §3) and AGN components.
4.1AGN dominated SEDs at 6 µm to 35 µm
The IRS spectra presented in fig. 2 provide a clear picture of the
range of MIR SEDs produced by strongly AGN-dominated sys-
tems. As explained in §2, the weak PAH features in this sample
suggest that at least 90 per cent of the continuum emission at 19 µm
is produced by the AGN. Considering that the typical host-galaxy
continuum falls toward shorter infrared wavelengths (see fig. 4),
it is reasonable to assume that the AGN dominates the contin-
uum emission at ∼6 µm – 20 µm. At these wavelengths, all 25
AGN-dominated SEDs show clear evidence of a continuous, un-
derlying power-law continuum that is thought to be produced by
multiple dust components spanning a range of temperatures (e.g.,
Buchanan et al. 2006).
The spectral indices of the underlying AGN power-law con-
tinua span the range 0.7 ? α1? 2.7 (mean: α1=1.6), with a ten-
dency for Type 2 AGNs to have higher (more positive) spectral in-
dices compared to Type 1 AGNs.3However, there is considerable
overlap between the types (i.e., Type 1s: 0.7 ? α1? 1.7; Type 2s:
0.8 ? α1? 2.7), and the difference in the mean SEDs is not sta-
tistically significant. This large range of spectral indices is consis-
tentwiththosefound inprevious studies(e.g.Buchanan et al.2006;
Wu et al.2009).Infour cases(i.e.,oneType 1:EXO055620-3820.2
and three type 2s: ESO506-G027, NGC 5506, ESO103-035), we
see evidence of strong silicate absorption at 9.7 µm. In a further
eight cases (i.e., five type 1s: ESO548-G081, 3C120, NGC 3783,
Mrk 290, NGC 7213 and three type 2s: NGC 1275, NGC 2110,
NGC 4507) there isstrong evidence of silicateemission at ∼10 µm
which is always accompanied by another silicate emission feature
at ∼ 18 µm (see Sturm et al. 2005 for a dedicated study of the sili-
cate emission features seen in the MIR spectra of AGNs).
In at least 20 of the 25 AGN-dominated MIR spectra there is
a definite break in the power-law continuum at 15−20 µm; mean
break position of ∼ 19 µm. In previous studies this has been at-
tributed to a dominating warm (i.e., ∼ 170 K) dust component that
is heated by the AGN (e.g., Weedman et al. 2005; Buchanan et al.
2AKARI data were retrieved from the NASA/IPAC Infrared Science
Archive at http://irsa.ipac.caltech.edu/.
3Here, we use FAGN
ν
νFAGN
ν
∝ λα. To convert to νFAGN
ν
use β = α−1, where
∝ λβ

Page 7

Figure 3. IRAS colour-colour (left panel) and colour-luminosity (right panel) plots of all sources in the Revised Bright Galaxy Survey (Sanders et al. 2003)
with infrared luminosities covering the roughly the same range as the Swift-BAT sample from which our AGN-dominated galaxies were selected (i.e., LIR=
1010−1012ergs s−1). We use a subsample of 10 starburst SEDs from Brandl et al. (2006) plus four other “cold” galaxies to define our host galaxy templates
(circled stars). These extra four quiescent-galaxy SEDsare needed to sample the full range of host-galaxy SEDs covered by LIR=1010−1012ergs s−1galaxies.
The lack of “cold” galaxies in the Brandl et al. (2006) sample is shown most clearly in the colour-colour plot (left hand panel). The horizontal black line in the
right hand panel indicates the range of LIRof the AGNs selected for decomposition (see §4.2). The objects with [60 µm/25 µm] ? 4 are likely AGN dominated
and were therefore not considered as suitable host-galaxy templates.
2006). In all cases where we see a break in the continuum power
law, the AGN SED longward of λBrkhas a lower spectral index
(i.e., power-law index, 0 ? α2? 1.5, mean= 0.7; again, consis-
tent with Buchanan et al. 2006 and Wu et al. 2009) than at shorter
wavelengths. We again find that there is significant overlap in the
range of spectral indices at λ > 19 µm between Type 1 and Type
2 AGNs. We note that the power-law indices shortward and long-
ward of λBrkare uncorrelated. The number of AGNs in our sample
showing a break around 19 µm could be as high as 24, since it
can be disguised by the presence of the silicate emission feature at
18 µm. The only AGN in our sample where the 6–35 µm SED can
be unambiguously described as a single power law is NGC5506;
the spectral index of NGC5506 over 6–35 µm is α = 1.2±0.2,
which is consistent with the ranges of both α1and α2.
Because thetypical host-galaxy continuum emission increases
strongly toward longer wavelengths, it is difficult to ascertain
whether the power-law emission at longer wavelengths (i.e., λ ?
25 µm) arises from the AGN alone, or if the host galaxy also makes
a significant contribution. Indeed, in four cases (i.e., NGC7213,
NGC4507, NGC2110, 3C120), the SED longward of ∼ 25 µm
shows evidence of a turn-up in νFνwhich is consistent with be-
ing due to an underlying host-galaxy component. The host-galaxy
component must be accounted for at longer wavelengths in order
to define the intrinsic AGN infrared SED, which we address in the
next subsection.
4.2 Intrinsic AGN infrared SED at 6 µm to 100 µm
We now extend the intrinsic AGN infrared SEDs to longer wave-
lengths (i.e., λ ? 25 µm) using IRAS 60 µm and 100 µm photome-
try for guidance. Here, we only consider those 11 AGN-dominated
SEDsthat we have shown suffer from minimal aperture effects (see
§2). The approach we take is to decompose the observed SEDs into
their host-galaxy and intrinsic AGN components through simulta-
neously fitting the IRS spectra and IRAS photometry. An alterna-
tive approach would be to simply subtract a normalised host-galaxy
component from the observed SEDs to leave the emission intrinsic
tothe AGN. However, calculating the appropriate normalisation for
the host-galaxy component is a non-trivial matter that is compli-
cated by the fact that the host-galaxy SED may be modified by
dust-absorption (i.e., the observed SED is not a simple linear sum
of ahost-galaxy andanAGN component; separateabsorption terms
must also be considered). Later, and only for illustrative purposes,
we subtract the host-galaxy SEDs normalised via our SED fitting
routine from the sample of 11 observed AGN SEDs (see fig. 5).
One downside of the approach that we adopt is that it requires
us to make a priori assumptions about the form of the intrinsic
AGN infrared SED used in our fits. Our analyses of the AGN-
dominated MIR spectra in the previous subsection provides us with
a well defined set of parameters that describe the variety of intrinsic
AGN SEDs at 6 to ∼25 µm (i.e., 0.7 ? α1? 2.7, 0? α2? 1.5 and
15 µm? λBreak?20 µm). At longer wavelengths, where there are
fewer observational constraints, we can only estimate the form of
the intrinsic SED based on reasonable physical assumptions then
test whether models incorporating these SEDs reproduce observa-
tions. With this in mind, we assume that the intrinsic SED falls as
a modified black body beyond a given wavelength, λBB. In our fits,
λBBis allowed to take any value between 20 µm and 100 µm. Our
choice of a modified black body spectrum is loosely based on the
results of radiative transfer models of the dust surrounding AGNs,

Page 8

Table 2. Galaxy SEDs that used to construct our host galaxy templates.
Host Galaxy Template
(1)
Galaxy
(2)
D (Mpc)
(3)
log(LIR/L⊙)
(4)
SB1NGC 1667
NGC 5734
NGC 6286
NGC 7590
NGC 7252
Mrk 52
NGC 4818
NGC 7714
NGC 1222
NGC 3256
NGC 4194
NGC 520
NGC 660
NGC 2623
60.51
59.28
79.78
21.58
64.67a
33.50b
9.37
38.16
32.26
35.35
40.33
30.22
12.33
77.43
10.96
11.06
11.32
10.16
10.77c
10.25c
9.75
10.72
10.60
11.56
11.06
10.91
10.49
11.54
SB2
SB3
SB4
SB5
NOTES: (1) Host-galaxy template, (2) Galaxy name, (3) Distance in Mpc
taken from Sanders et al. (2003),acalculated using the redshift reported
in NED using the same cosmology as Sanders et al. (2003),bredshift-
independent distance reported in NED (4) 8–1000 µm infrared luminosity,
taken from Sanders et al. (2003),ccalculated using the distance in column
3 and the prescription outlines in table 1 of Sanders & Mirabel (1996).
although we note that the precise form of this fall-off does not have
a significant impact on the fits. When needed, we add silicate emis-
sion features at 10 µm and 18 µm to the intrinsic AGN component.
These are modelled with broad (i.e., FWHM∼3 µm) gaussians. In
our models, we account for any absorption to both the AGN and
host-galaxy components using a Draine (2003) extinction curve.
We proceed by simultaneously fitting the IRS spectra and the
IRAS photometry of the 11 AGNs in our sample with a combina-
tion of our host-galaxy templates (see §3) and the estimated intrin-
sic AGN SED described above. Any strong emission lines (exclud-
ing PAH features) are masked to ensure that they do not adversely
effect the fit. Each of the 11 infrared SEDs are fit five times, each
time using a different host-galaxy template (defined in §3; see also
fig. 4). We use χ2-minimisation to obtain the best fitting parame-
ters for each choice of host-galaxy template. Plots showing the fits
to the data (including the host-galaxy and intrinsic AGN compo-
nents) and the resulting residuals are presented in appendix A.
Because the observed SEDs consist, in part, of very high
signal-to-noise IRS spectra, none of the fits to the data are formally
good (i.e., their reduced χ2>> 1). Therefore, we cannot use χ2
statistics to unambiguously determine whether any of the five dif-
ferent AGN–host-galaxy template solutions provide a good charac-
terisation of the data. Instead, we use the following criteria to deter-
mine which of the five models provide suitable fits to the observed
data: does it (a) pass within 2σ of the 60 µm and 100 µm photom-
etry and (b) reproduce the general shape of the IRS spectrum. For
each of the 11 SEDs in our sample, we can identify at least one of
thefive model fitsthat satisfyboth these criteria.When both criteria
are met we select, by eye, those model fits that produce the small-
est residuals at the wavelengths of the strongest PAH features, i.e.,
7.7 µm, 11.25 µm and 12.8 µm. We use this additional criterion to
ensure that any contribution from the host-galaxy well accounted-
for by the fitted host-galaxy component. Where there is no signif-
icant difference in the PAH residuals of two or more such fits we
identify each of them as being suitable (in appendix A we have
highlighted all selected fits). In only two cases, namely NGC 5506
Figure 4. The five average host-galaxy templates derived from the
Brandl et al. (2006) starburst sample and the four “cold” galaxies selected
from the Revised Bright Galaxy Sample (Sanders et al. 2003; see fig. 3).
Each template is the mean-average of a group of SEDs that have similar
overall shape and relative PAH strengths (see §3 for a description of how
we construct these host-galaxy templates). Each SED is made up of a low-
resolution IRS spectrum that we extrapolate to far-infrared wavelengths us-
ing IRAS photometry at 60 µm and 100 µm. We also show the mean flux at
65 µm, 140 µm and 160 µm measured from AKARI data (when available)
as a check that the modified black-body extrapolation beyond 100 µm is a
reasonable approximation of the real SEDs. We use these average templates
to model any host galaxy components in our sample of AGN-dominated
galaxies to derive the intrinsic AGN infrared SED.
andESO103-035, doweseelargeresidualsat thePAHwavelengths
for all five fits. In both these cases, we assume that all of the fits are
as good as each other and select all five for further consideration.
In all selected cases (including NGC 5506 and ESO103-035), the
model fits lie within 10 per cent of the measured flux density for
at least 85 per cent of the IRS data points, demonstrating the abil-
ity of our SED fitting approach to reproduce the continuum shape
and PAH strengths of the AGNs in our sample. As a final note to
the fitting procedure we would like to point out that, while we are
confident that selecting the fits that closely reproduce the PAH fea-
tures is the most appropriate procedure to take, our main results do

Page 9

not change if, instead, we use the results derived from using all our
host-galaxy templates (which cover the full range of SED shapes
of local, LIR= 1010−1012L⊙galaxies; see fig. 3). The strongest
effect that taking this alternative approach has on our results is to
increase the scatter of intrinsic AGN SEDs at FIR wavelengths by
a factor of ∼ 2−3; as we shall see, this does not represent a large
increase over the range of intrinsic AGN SED shapes produced by
selecting only the suitable host-galaxy components identified using
the selection criteria outlined above.
For illustrative purposes, we show in fig. 5 the SEDs obtained
by subtracting the suitable host galaxy components (i.e., those ex-
tractedfrom thesuitablemodel fitsidentifiedabove and highlighted
in appendix A) from the observed SEDs. As we have selected only
those fitsthat reproduce the general shape of the observed SED, the
SEDspresented infig.5areequivalent totheintrinsicAGNcompo-
nents produced during the fitting procedure and shown in appendix
A. For guidance, we have highlighted the SED that is produced by
the fit with the smallest overall residuals between 6 µm and 35 µm
(i.e., the “best-of-the-best”), although we stress that all plotted in-
trinsic SEDs are viable.
For all 11 galaxies in our sample, the choice of host-galaxy
template has almost no effect on the derived intrinsic SED short-
ward of λBrk. On the other hand, in at least five cases the SED at
longer wavelengths is strongly dependent on the choice of host-
galaxy template, meaning the exact form of the intrinsic AGN SED
at these wavelengths is uncertain. However, in all cases the intrin-
sic SED falls rapidly (in νFν) beyond 15-60 µm, irrespective of
the choice of host-galaxy component. Despite this fall-off at long
wavelengths, the total emission at 60 µm is still dominated by the
AGN in at least three, possibly four, cases (namely, MCG-03-34-
064, ESO-103-035, IC5063 and, possibly, Mrk 3; there is one uns-
elected host-galaxy component that dominates at 60 µm in this last
case), irrespective of the choice of host-galaxy SED (see fig. 5).
In fig. 6 we plot the full range of possible intrinsic AGN SED
components extracted from our models. Shown in this plot are the
intrinsic SEDs from all suitable fits identified using the criteria out-
lined above. A notable feature of this plot is the large range of in-
trinsic SED shapes extracted from the fits to the 11 observed SEDs
in our sample. For example, when normalised at 19 µm the flux
the intrinsic AGN flux extracted from our model fits spans over an
order of magnitude at 60 µm and 100 µm. While the range mid-
infrared (i.e., λ ? 25 µm) SEDs is well constrained by the AGN-
dominated IRSspectra, itisnot clear how muchof thescatter atFIR
wavelengths (i.e.,λ?25µm)isdue todifferencesinthetrueintrin-
sic SED and how much is introduced by the different host-galaxy
templates we use. Therefore, the range shown here will cover a
larger spreadthan thetrueintrinsicAGN SEDsatλ?25 µm. Infig.
6 we have discriminated between AGNs with X-ray luminosities
above and below the median of the sample (i.e., above and below
log(L2−10keV) =42.9). We identify a clear trend for more intrinsi-
cally luminous AGNs to have intrinsic infrared SEDs that fall more
rapidly at longer wavelengths, although stress that there is some
overlap between the SEDs of AGNs with X-ray luminosities above
and below the median of our sample (indeed, the most rapidly
falling SED is that of the lowest luminosity AGN in our sample,
NGC 2110). We also find that, on average, the mid-infrared SEDs
of the more X-ray luminous AGNs in our sample have weaker sil-
icate absorption and stronger silicate emission, although this could
be related to the fact that all of the low X-ray luminosity AGNs are
classed as either Type 1.8, 1.9 or 2 AGNs which, as we have seen,
tend to have stronger silicate absorption features (see §4.1). The
evidence for a link between the shape of the intrinsic infrared SED
and L2−10keVis strengthened by differences between the average
intrinsic SEDs of the moderate luminosity AGNs studied here and
those of more luminous quasars (e.g., the average quasar SED pre-
sented in Netzer et al. 2007; see §5.2). There is weak evidence of
a relation between the shape of the SEDs shortward and longward
of λBrk, with SEDs with lower values of α1falling more rapidly
at FIR wavelengths. However, we note that there are at least two
objects in our sample that deviate strongly from this trend (namely,
Mrk 3 and NGC 5506; see fig. 5).
We also plot in fig. 6 the average intrinsic SEDs calculated by
taking the mean of the intrinsic AGN components extracted from
the fits. Because of the considerable overlap in shape of the intrin-
sic SED at MIR wavelengths between Type 1 and Type 2 AGNs
(see §4.1 and fig. 2) and the small number of intrinsic SEDs in our
sample (i.e., 11) we do not discriminate between AGN types when
producing these averages. Although not included in this plot, we
note that the average of the SEDs obtained by subtracting the host-
galaxy components from the observed data (i.e., those shown in fig.
5) is consistent with the average SED shown in fig. 6. This average
SED can be expressed as:
Fν∝



λ1.8
λ0.2
at
at
at
6 µm < λ < 19 µm
19 µm < λ < 40 µm
λ > 40 µm
ν1.5FBB
ν
(1)
and shows evidence of weak silicate absorption (equivalent to an
absorbing column of NH∼5×1021cm−2, or τ9.7∼0.2, using fig.
10 of Draine 2003 to convert NHto τ9.7) but little or no silicate
emission. For clarity, we plot the average mean-average SEDs of
log(L2−10keV) >42.9 and log(L2−10keV) <42.9 AGNs separately
in the right-hand panel of fig. 7. As expected from the trend identi-
fied above, the average intrinsic SED of the more luminous AGNs
in our sample (i.e., log(L2−10keV) >42.9) falls more rapidly at
longer wavelengths than that of the lower luminosity AGNs (i.e.,
log(L2−10keV)<42.9). For example, when normalised at 19 µm the
lower luminosity AGNs emit, on average, 2-3 times more flux at
60 µm and 100 µm than the higher luminosity AGNs in our sam-
ple. The parameters describing the average SEDs of the high and
low luminosity AGNs in our sample are largely the same as shown
in Eqn. 1, although the spectral indices at λ > 19 µm are somewhat
different, i.e., α2=0.0 and 0.4, respectively. This relative difference
between the FIR SEDs of high and low luminosity AGNs in our
sample could be the result of higher luminosity AGNs being capa-
ble of heating a larger fraction of their surrounding dust to higher
temperatures. A consequence of this increased heating would be
relatively stronger emission at MIR wavelengths; hence the appar-
ent flux deficit at FIR wavelengths when normalised at 19 µm (see
also §5.2). All three average intrinsic AGN SEDs described here
are published in columns 2-4 of table 3 (which is available in its
entirety online at http://sites.google.com/site/decompir).
Finally, we explore whether using a more straightforward ap-
proach to normalise the host galaxy component has any effect on
the extracted intrinsic AGN SEDs. To extract the average intrin-
sic SED of more luminous, quasar AGNs, Netzer et al. (2007) nor-
malised the host galaxy component such that it constitutes the ma-
jorityof theobserved fluxat60µmand100µm.Ifwetakeasimilar
approach and assume that 90% of the flux at 100 µm is emitted by
the host-galaxy, we obtain largely the same overall intrinsic SED
shapes as shown in fig. 5 and fig. 6, but are left with strong features
at 7.7 µm where the PAH features are over-estimated by the host-

Page 10

Figure 5. The intrinsic mid-infrared spectra (light blue lines) and far-infrared photometry (light blue points) of the 11 AGN-dominated sources in our sam-
ple after subtracting suitable host-galaxy components. The suitability of the host-galaxy components is assessed from the SED fits described in §4.2 (see
also appendix A). We highlight the SED produced by subtracting the most suitable host-galaxy component (i.e., the “best-of-the-best”; see §4.2; dark blue
line/points). The original observed data is shown in red. The intrinsic AGN infrared SEDs shown here cover a broad range of shapes, although all fall rapidly at
wavelengths longwards of 15-60 µm. Despite this fall-off at long wavelengths, there are at least four observed AGNs SEDs (namely, Mrk 3, MCG-03-34-064,
ESO-103-035 and IC5063) that are AGN-dominated even at 60 µm, irrespective of the choice of subtracted host-galaxy template.
galaxy component, indicating that thisapproach islessreliablethan
our adopted approach.4
5 COMPARISON WITH PREVIOUSLY DEFINED AGN
INFRARED SEDS AND RESULTS FROM DUSTY
TORUS MODELS
We have produced a set of intrinsic AGN infrared SEDs by decom-
posing the observed SEDS of local, AGN-dominated galaxies into
4We also get broadly similar results if we assume that 80% or 99% of the
100 µm flux is due to the host-galaxy.
well-definedhost-galaxy and intrinsicAGN components (see§4.2).
The full variety of these intrinsic AGN SEDs is shown in fig. 6. In
this section we explore how these intrinsic SEDs compare against
other commonly assumed AGN SEDs,including those produced by
radiativetransfer modelsof thedust surrounding theAGN.Forsim-
plicity, the comparisons made in this section are against the mean
average and range of all the intrinsic SEDs in our sample (i.e., we
do not differentiate between the high and low luminosity AGNs in
our sample; see previous section).

Page 11

Figure 7. Left: The mean-average and range of intrinsic AGN infrared SEDs from this study plotted against the observed infrared SEDs of Mrk 231 and
NGC 1068 (i.e., two galaxies commonly used to represent the typical AGN infrared SED). When all are normalised at 19 µm, the SEDs of NGC 1068 and
Mrk 231 lie above the intrinsic AGN SEDsat FIR wavelengths, which we attribute to host-galaxy contamination in the two comparison SEDs(see §5.1). Right:
The mean average intrinsic SEDs of (a) all 11 AGNs in our sample, (b) log(L2−10keV)<42.9 AGNs and (c) log(L2−10keV)>42.9 AGNs, plotted against the
average quasar SEDs of Richards et al. (2006) and the average intrinsic quasar infrared SED of Netzer et al. (2007). Again, all SEDs are normalised at 19 µm.
Note that the Richards et al. (2006) and Netzer et al. (2007) SEDs extend to ∼ 95 µm and ∼65 µm, respectively. At this normalisation, the Richards et al.
(2006) average quasar SED is well matched to the average intrinsic SED at λ ?30 µm, although it lies above our average intrinsic SED at shorter wavelengths.
At λ ?19 µm the average intrinsic quasar SED of Netzer et al. (2007) lies slightly above all the average intrinsic SEDs from this study, but lies below them at
λ ?19 µm. This plot clearly shows the trend reported in §4.2 for more luminous AGNs to have bluer intrinsic infrared SEDs.
5.1 NGC 1068, Mrk 231
Of all known local AGNs, NGC 1068 and Mrk 231 are com-
monly used to characterise the infrared SEDs of AGNs. In fig. 7
we illustrate how our range of intrinsic AGN SEDs compare with
these canonical AGNs.5When normalised to 19 µm, the SEDs
of Mrk 231 and NGC 1068 are, respectively, ∼15 and ∼8 times
higher at 100 µm than the average SED calculated from our AGN
templates. Assuming that the average intrinsic AGN emission con-
tinues to fall as a modified black body to 1000 µm, the Mrk 231
and NGC 1068 SEDs contain roughly 1.5-3 times the power at
infrared wavelengths than the average intrinsic AGN SED (when
all are normalised at 19 µm). This excess power is likely the
result of host galaxy contamination in these two cases, as sug-
gested by Telesco et al. (1984); Downes & Solomon (1998) and
Le Floc’h et al.(2001).Indeed, fitstotheinfraredSEDsof Mrk231
and NGC 1068 using the approach outlined in §4.2 are consistent
with them being host-galaxy dominated at FIR wavelengths. Fi-
nally, we note that at wavelengths ? 20 µm, where the host-galaxy
contributions are small, the SEDs of NGC 1068 and Mrk 231 are
largely consistent with our intrinsic AGN SEDs.
5The NGC 1068 and Mrk 231 SEDs shown here represent the infrared
emission produced by the entire galaxy (i.e., including emission from
the host galaxy). We note that, whilst mid-infrared flux density measure-
ments of the resolved AGN core of NGC 1068 (e.g., Gandhi et al. 2009;
Prieto et al. 2010) are now available, we choose to include emission from
the host galaxy to allow easier comparison with previous studies which
have, in general, done the same.
5.2 Infrared Quasar SEDs
There have been a number of attempts to constrain the aver-
age quasar SED, although only a handful have included coverage
to FIR wavelengths (e.g. Elvis et al. 1994; Richards et al. 2006;
Netzer et al. 2007). One of the most prominent, recent studies is
that of Richards et al. (2006), which combined photometric data
covering the radio to X-ray regimes for a sample of optically-
selected, broad-line quasars. In fig. 7 we compare the average in-
frared SED of all quasars from that study to the AGN templates
derived here. The average quasar SEDs from Richards et al. (2006)
turns over at approximately the same FIR wavelengths as the aver-
age of our intrinsic AGN SEDs, although we note that host-galaxy
contamination at FIR wavelengths has not been removed from
this average quasar SED. This host-galaxy contamination will tend
to push the position of the turn-over to longer wavelengths. The
Richards et al. quasar SED is flatter than the average intrinsic SED
of more moderate luminosity AGNs, which could be due to the in-
creased quasar luminosities heating the surrounding dust to higher
temperatures compared to more moderate luminosity AGNs. De-
spite being flatter at short wavelengths, the Richards et al. (2006)
SED contains only ∼10 per cent more flux over 8-1000 µm than
the average intrinsic SED when both are normalised to 19 µm and
extrapolated to longer wavelengths assuming a modified blackbody
SED.
Recently, Netzer et al. (2007) removed the host-galaxy com-
ponent from the average infrared SED of a sample of PG-quasars
to produce an average, intrinsic quasar infrared SED. This SED
is also investigated in fig. 7. When both are normalised at 19 µm

Page 12

Figure 6. The full range of possible intrinsic AGN infrared SEDs produced
by our SED fitting procedure described in §4.2. We include in this plot
the intrinsic SEDs extracted from the fits deemed suitable using the cri-
teria outlined in §4.2 (i.e., those highlighted in appendix A). Each intrin-
sic SED is normalised at 19 µm to demonstrate the range of SED shapes.
There is a clear systematic difference between the intrinsic SEDs of AGNs
with X-ray luminosities above and below the median of our sample (i.e.,
log(L2−10keV)=42.9). Also included in this plot is the mean-average intrin-
sic SEDs of all 11 AGNs in our sample, labelled to illustrate the parameter-
isation outlined in Eqn. 1.
the average intrinsic quasar SED emits more strongly at λ <19 µm
than the average intrinsic SED of more moderate luminosity AGNs
calculated above. However, this situation is reversed at λ >19 µm,
with the former falling more rapidly at longer wavelengths. This
strengthens our findings described in §4.2, where we report a simi-
lar trend between the higher (i.e., log(L2−10keV) >42.9) and lower
(i.e., log(L2−10keV) <42.9) luminosity AGNs within our sample.
We note that the intrinsic quasar falls even more rapidly than the
average intrinsic SED of the higher luminosity AGNs in our sam-
ple. The intrinsic quasar SED has stronger silicate emission fea-
tures than the average intrinsic SED of more moderate luminosity
AGNs, although this may be due to the Netzer et al. (2007) being
comprised of Type 1 AGNs which tend to have stronger silicate
emission features (see §4.1). As we suggest in §4.2 the increased
flux at short wavelengths and relative deficit of flux at longer wave-
lengths in the average quasar SED may be related to the higher lu-
minosities of quasars, with more luminous objects capable of heat-
ing more dust to higher temperatures. However, we can only claim
this in an average sense as the average quasar SED is consistent
with the spread of intrinsic infrared SEDs extracted from the 11
AGN-dominated sources above. The four intrinsic AGN SEDs that
bear the closest similarity to the average quasar intrinsic SED are
3C120, IC4239A, Mrk 509 and NGC 4507; three are type 1 AGNs
and one is a type 2 AGN, respectively. All four of these SEDs have
intrinsic 2-10 keV luminosities greater than the median of our sam-
ple, (i.e., log(L2−10keV) >42.9).
For comparison, we fit the Netzer et al. (2007) intrinsic quasar
SED using DECOMPIR (see appendix B), obtaining:
Figure 8. The range of the intrinsic AGN infrared SEDs (as shown in
fig. 7, left panel) plotted against the range of SEDs predicted by contin-
uous (left-hand plot) and clumpy (right-hand plot) torus models described
in Fritz et al. (2006) and Schartmann et al. (2008), respectively (shaded re-
gions in the left and right panels, respectively). Each set of SEDs has been
normalised at 19 µm. In general, a clumpy torus model provides a better
representation of the intrinsic AGN SED than the continuous torus models
which over-predict the range of intrinsic AGN infrared SEDs by at least one
order of magnitude at ∼60 µm with this normalisation. However, none of
the clumpy torus models predict 6-20 µm SEDs as steep as some of those
in our AGN-dominated sample (e.g., Mrk 3)
Fν∝
?
λ1.2
at
at
6 µm < λ < 20 µm
λ > 20 µm
ν1.5FBB
ν
(2)
where the symbols are the same as those defined in §4.2. These
parameters confirm the shallower SED at MIR wavelengths and
shorter-wavelength turnover of the Netzer et al. (2007) intrinsic
quasar SED, compared to the average AGN SED defined in this
study. When we extrapolate the Netzer et al. (2007) intrinsic quasar
SED to longer wavelengths as a modified blackbody and normalise
both SEDs at 19 µm, then we find that the total power radiated at
infrared wavelengths is ∼80 per cent that of the average intrinsic
infrared SED of typical AGNs calculated here.
5.3 Radiative transfer (dusty torus) models
Finally, we compare the range of intrinsic, infrared AGN SED tem-
plates with predictions from radiative transfer models of the dust
surrounding the active nucleus, which is thought to be the main
source of infrared emission from AGNs (although there is some
evidence to suggest that infrared emission from AGNs could also
be produced in more extended, dusty regions; e.g. Schweitzer et al.
2008; Mor et al. 2009). We compare our templates with the SEDs
produced by both (a) continuous and (b) clumpy distributions of
dust (specifically, the models described in Fritz et al. 2006 and
Schartmann et al. 2008, respectively). The range of input param-
eters of the former (i.e., continuous) model are given in table 1 of
Fritz et al. (2006). The input parameters of the clumpy torus model
is presented in table 1 of Schartmann et al. (2008); we used the
SEDs produced by varying both orientation angle (see their fig. 4)
and dust mass (see their fig. 10). The SEDs predicted by these torus
models are shown in fig. 8, together with the average and range of
intrinsic SEDs defined in this study.
The continuous dust-distribution models produce a much
broader range of SED shapes than the intrinsic AGN infrared SEDs
defined here. For example, when normalised at 19 µm these mod-
els predict a range of flux densities spanning almost three orders
of magnitude at 60 µm, compared to a spread of (at most) two or-
ders of magnitude for the intrinsic SEDs defined in this study. We

Page 13

Figure 9. The intrinsic 12 µm AGN luminosities derived from our fits to the
IRAS photometry plotted against the core 12 µm luminosities obtained us-
ing high spatial resolution (i.e., sub-arcsecond) mid-infrared observations of
23 nearby AGNs, taken from Horst et al. (2008) and Gandhi et al. (2009).
In the latter, the host galaxy contribution is spatially resolved, providing
an uncontaminated measure of the intrinsic AGN infrared luminosity at
12 µm. After fitting the large-aperture (i.e., arcminute) IRAS photometry
and excluding any host galaxy contamination we reproduce, on average,
the intrinsic luminosities derived from high spatial resolution observation
to within a factor of ∼2, compared to a factor of ∼4 when the 12 µm IRAS
photometry is not corrected for host-galaxy contamination. Points associ-
ated with Compton-thick AGNs have been marked with a “C” and show a
similar level of scatter as the Compton thin sources.
note that the model SEDs that match the range of intrinsic AGN
SEDs do not correspond to a particular region of input parame-
ter space (i.e., torus opening angle, optical depths, inner and outer
torus radii, radial dust density distribution, angular dust density dis-
tribution, SED of incident radiation). Therefore, our intrinsic AGN
SEDs are unable to directly constrain acceptable input parameters
of these continuous dust-distribution models.
The SEDs produced by the clumpy torus models provide a
closer match to the observed intrinsic AGN infrared SEDs, al-
though they too over-predict the range of FIR fluxes spanned by
our intrinsic SEDs and tend to peak (in νFν) at longer wavelengths.
The closer match to the clumpy torus models is consistent with
other evidence in support for such models (e.g., Horst et al. 2006;
Gandhi et al. 2009; Ramos Almeida et al. 2009). However, none
of the clumpy torus models described in Schartmann et al. (2008)
predict 6-20 µm SEDs as steep as some of those in our AGN-
dominated sample (e.g., Mrk 3). As a consequence, at this normal-
isation the clumpy torus models over-predict the emission of many
of our sources at λ ? 19 µm.
6 TESTING THE AGN AND HOST-GALAXY
TEMPLATES
A principal motivation behind defining the intrinsic infrared SEDs
of moderate luminosity AGNs is to use these as templates to fit the
infrared SEDs of composite galaxies and calculate the AGN and
host galaxy contributions to their total infrared output. This is espe-
ciallyuseful inthemajorityof cases when only broad-band infrared
photometry measurements are available to constrain the infrared
SED. In this section we explore whether fits to the broad-band pho-
tometry incorporating the host-galaxy and AGN templates defined
Figure 10. EW PAHλ11.25plotted against the AGN contribution at 19 µm
derived from our fits to IRAS photometry of 78 sources in the 12 µm sample
of Seyfert galaxies (Rush et al. 1993; Tommasin et al. 2008, 2010). Sources
where the IRAS 12 µm photometry measurement is at least 2 times higher
than the flux at the same wavelength in the high resolution IRS spectrum
in which the EW PAHλ11.25is measured have been indicated. Such cases
are likely to suffer from significant aperture effects which will typically
lead to more of the host galaxy flux being included in the IRAS photome-
try measurements. Also shown in this plot is the EW PAHλ11.25– percent
AGN contribution relation taken from Tommasin et al. (2010) (black line)
with a ±25 per cent margin shown in grey. We note a general agreement
between these two measures of the host-galaxy contribution, with ∼ 76 per
cent (i.e., 59/78) of AGN contributions measured from the photometry fits
lying within 25 per cent of that expected using EW PAHλ11.25.
herecan reproduce theintrinsicAGN luminositiesand AGN contri-
butions derived using other, independent approaches (i.e., high res-
olution imaging of nearby AGNs and emission line diagnostics).
To perform these fits, we have developed an IDL procedure, DE-
COMPIR,which wedescribe inappendix B and ispublicly available
online at http://sites.google.com/site/decompir.
6.1Comparison with high spatial resolution observations of
local Seyferts.
A number of recent studies (e.g. Krabbe et al. 2001; Horst et al.
2006, 2008; Gandhi et al. 2009) have shown that the intrinsic in-
frared luminosities of nearby AGNs can be reliably measured from
high spatial resolution observations (i.e., ∼ 0.5′′, typically probing
physical scales of ? 100 pc). At such high resolutions, host galaxy
contamination is minimised, leaving only the emission intrinsic to
the active nucleus. However, such analyses are only viable for a
small number of systems that are nearby enough that the central
regions can be sufficiently well resolved. On the other hand, if we
can use SED decomposition to exclude the host galaxy contribu-
tion to the broad-band SEDs, then we could potentially measure
the intrinsic 12 µm luminosities for all AGNs with well sampled
infrared SEDs. Here, we demonstrate that the SED decomposition

Page 14

procedure outlined in this study can, indeed, be used to measure the
intrinsic AGN 12 µm flux.
We use DECOMPIR (see appendix B) to fit the four-band IRAS
photometries of 23 AGNs with core 12 µm luminosities reported
in Horst et al. (2008) or Gandhi et al. (2009) and X-ray luminosi-
ties within the range spanned by the AGNs used to define our in-
trinsic AGN SEDs (i.e., L2−10keV= 1042−1044ergs s−1). As we
are limited to only four independent flux density measurements we
only allow the normalisations of the AGN and host galaxy compo-
nents to vary when fitting the observed SEDs. All other parameters
are fixed to their average values derived from the full sample of
11 AGN dominated galaxies described above (also see appendix
B), but we note that our main results do not change if we use the
average SED of either the high (i.e., log(L2−10keV)>42.9) or low
(i.e., log(L2−10keV)<42.9) X-ray luminosity AGNs in our sample.
Each SED was fit five times, once for each of our host-galaxy tem-
plates (i.e., as recommended in appendix B). To calculate the in-
trinsic 12 µm AGN flux, we integrate the intrinsic AGN compo-
nent derived from the best fitting solution of these five fitted SEDs
(i.e., that with the lowest associated χ2value) over a narrow (i.e.,
1 µm) top-hat passband centred at 12 µm, which is a similar re-
sponse function as those filters used for the observations described
in Horst et al. (2008) and Gandhi et al. (2009).
In fig. 9 we plot the intrinsic 12 µm flux derived from our best
fitting SEDs against those obtained using high spatial resolution
observations as reported in Horst et al. (2008) and Gandhi et al.
(2009). We find that the intrinsic luminosities derived using our
SED fits are well matched to those obtained using high resolution
observations of the AGN core. On average, the intrinsic 12 µm
AGN luminosity derived from our fits lie within a factor of two
of the core luminosity at this wavelength, compared to within a
factor of four when the 12 µm IRAS photometry is not corrected
for host-galaxy contamination. A linear regression to these points
gives:
log
?
LFit
12
1043erg s−1
?
=(0.13±0.07)+
(0.96±0.09)log
?
LHiRes
12
1043erg s−1
?
(3)
which has a slope consistent with a 1:1 relationship. For compari-
son, we also plot the 12 µm luminosities calculated directly from
the 12 µm IRAS photometry, noting that these often overestimate
the intrinsic 12 µm luminosities, sometimes by more than a factor
of 5. Finally, we note that the five Compton-thick sources in this
sample (i.e., NGC 1068, IC 3639, Swift J0601.9-8636, NGC 5728
and NGC3281; Della Ceca et al. 2008) display a similar degree of
scatter as the full sample. This demonstrates the power of using
infrared SED fitting to identify even heavily obscured AGNs.
6.2Comparison with results derived from emission line
diagnostics
A number of infrared emission line diagnostics have recently been
identified as providing reliable proxy measures of the AGN and
host-galaxy contributions to the infrared and bolometric output
of composite galaxies (e.g. Genzel et al. 1998; Tommasin et al.
2008, 2010; Goulding & Alexander 2009). A key test for our fits
is whether they can broadly reproduce the intrinsic AGN luminosi-
ties derived from these diagnostics.
We use DECOMPIR to fit the IRAS photometries of 78 galax-
ies from the IRAS 12 µm sample for which accurate emission line
fluxes, measured from high resolution IRS spectra, are available
(see Tommasin et al. 2008, 2010 for a complete list of all the AGNs
used for this section of our study, together with their measured
emission line fluxes). Again, as we are dealing with only four pho-
tometry measurements, we only allow the normalisations of the
AGN and host-galaxy components to vary when fitting the ob-
served data. All other parameters are fixed to their average values
derived from the full sample of 11 AGN-dominated galaxies (see
§4.2; again, we note that our results do not change if we use the av-
erageSEDsderived fromthehigh (i.e.,log(L2−10keV)>42.9) or low
(i.e., log(L2−10keV)<42.9) X-ray luminosity AGNs in our sample).
The IRAS photometry of each of the 78 objects in our sample are
fitted five times, each time using a different choice of host-galaxy
template (defined in §3). Here, we report the AGN contribution to
the 19 µm infrared emission measured from the best fitting of these
five independent fits (i.e., that with the lowest associated χ2value).
In fig. 10 we plot the AGN contribution at 19 µm derived
from our fits to the IRAS photometry against EW PAHλ11.25mea-
sured from high resolution IRS spectra; see Tommasin et al. (2008,
2010). We find that the AGN contribution measured from our fits
to the IRAS photometry are in good general agreement with those
based on EWPAHλ11.25, especially at high levelsof AGN contribu-
tion (i.e., ?50 per cent). Of the entire sample of 78 AGNs, ∼76 per
cent (i.e., 59/78) have AGN contributions measured from our broad
band SED fits that lie within ±25 per cent of those derived from
EW PAHλ11.25. At lower AGN contributions where the intrinsic
AGN component is more difficult to constrain, our fits typically es-
timate the AGN contribution to within ∼50 per cent of that derived
from EW PAHλ11.25. We note, however, that at such low AGN con-
tributions even emission line diagnostics have considerable diffi-
culty in estimating the intrinsic AGN contribution, as demonstrated
by the significant amounts of scatter in the [Ne V]/[Ne S3.2.2 II]
– EW PAHλ11.25correlation shown in fig. 4 of Tommasin et al.
(2010) (see also right-hand panel of fig. 8., Goulding & Alexander
2009)
7APPLICATION OF OUR TEMPLATES AND
INFRARED PHOTOMETRY FITTING PROCEDURE
In the previous section, we quantified the accuracy of our SED
fitting routine by comparing the results from these fits to results
obtained by other well-established approaches In this section, we
demonstrate two additional applications of our AGN intrinsic AGN
SEDs and fitting approach: (a) defining the correction factors to
convert 12 µm and 2–10 keV luminosities to total AGN infrared
luminosities (i.e., LAGN
IR
) and (b) measuring the intrinsic AGN lu-
minosities of large samples of composite galaxies for which only
infrared photometry measurements are available.
7.1
νLν(12 µm):LAGN
IR
and L2−10keV:LAGN
IR
correction factors
In cases where the infrared SED is either poorly constrained or
dominated by emission from the host galaxy, measuring the total
intrinsic AGN contribution can prove extremely difficult. Further-
more, to accurately estimate key parameters such as star-formation
rates from infrared emission requires that any AGN contribution be
accounted for. However, often, the only accurate way to measure
the levels of AGN activity in composite galaxies is via X-ray or
high resolution infrared observations. Using the intrinsic AGN in-
frared SEDs defined here coupled with the correlation between the
intrinsic 12 µm luminosity and the 2–10 keV luminosity defined
in Gandhi et al. (2009), we can now define a correction factor to

Page 15

Table 3. Average AGN and host-galaxy SEDs and the wavelength-dependent correction factors used
to calculated LAGN
IR
and LQSO
IR.
FAGN
ν
Hi. Lum
λ
Mean Lo. Lum
FSB1
ν
(Jy)
FSB2
ν
(Jy)
FSB3
ν
(Jy)
FSB4
ν
(Jy)
FSB5
ν
(Jy)(µm) (Jy)(Jy) (Jy)
LAGN
IR
νLAGN
ν
(10)
LQSO
IR
νLQSO
ν
(11)(1) (2)(3)(4)(5) (6)(7) (8)(9)
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
10.0
10.5
11.0
11.6
12.0
12.6
1.0
1.1
1.3
1.4
1.4
1.4
1.4
1.5
1.7
2.0
2.3
2.5
2.8
3.0
1.0
1.1
1.2
1.3
1.5
1.6
1.7
1.9
2.2
2.4
2.6
2.7
2.8
2.9
1.0
1.1
1.3
1.4
1.4
1.4
1.3
1.3
1.5
1.8
2.1
2.5
2.8
3.1
1.0
1.5
1.9
4.1
3.3
2.8
1.1
0.9
0.8
1.1
2.0
2.6
2.3
3.2
1.0
1.3
2.0
4.3
3.9
3.7
2.0
1.8
1.8
2.3
3.3
4.5
4.3
5.6
1.0
1.4
2.1
4.0
3.6
3.5
2.7
2.7
2.9
3.8
4.7
6.0
6.2
7.8
1.0
1.4
2.0
4.3
3.7
3.4
1.9
1.5
1.5
2.2
3.0
4.2
4.3
5.9
1.0
1.5
2.4
4.7
4.5
3.7
1.8
1.5
1.5
1.9
2.6
4.3
4.7
6.7
3.1
3.0
2.9
2.8
2.9
3.1
3.3
3.3
3.0
2.7
2.6
2.4
2.3
2.2
1.9
1.9
2.0
2.0
1.9
1.9
1.7
1.6
1.5
1.4
1.5
1.5
1.6
1.7
NOTES: (1) Wavelength, (2)-(4) Flux densities of the average intrinsic AGN infrared SEDs defined
in §4, normalised at 6 µm. Hi. Lum and Lo. Lum refer to the average SEDs of log(L2−10keV)> 42.9
and log(L2−10keV)< 42.9 AGNs in our sample, respectively. (5)-(9) Flux densities of the five host
galaxy templates defined in §3, normalised at 6 µm, (10) Conversion factors to convert νLAGN
intrinsic AGN infrared luminosities (i.e., LAGN
IR
) at all wavelengths covered by our templates, based
on the average AGN SED defined in §4. (11) The same as column 10, but for the average quasar
template of Netzer et al. (2007). This table is available in its entirety at higher wavelength resolutions
at http://sites/google.com/site/decompir.
ν
to total
convert 2–10 keV luminosities to total intrinsic AGN infrared lu-
minosities. This correction factor enables any AGN contribution to
the infrared output of galaxies for which, for example, only X-rays
or monochromatic infrared observations provide a measure of the
intrinsic AGN luminosity.
In fig. 11 we show the range of LAGN
all 11 AGNs for which we measure the intrinsic AGN infrared
SED, plotted against νLν(12 µm). Here, we have assumed that the
intrinsic AGN infrared SED continues as a modified black body
to 1000 µm. These corrections therefore only apply to the ther-
mal component of the intrinsic AGN SED as we do not consider
any emission from non-thermal (e.g., synchrotron) processes. We
choose νLν(12 µm) as a reference point as it has been shown to
be strongly correlated with the intrinsic luminosity of the AGN
(e.g., Gandhi et al. 2009). However, as we have defined the intrin-
sic AGN SED at 6-1000 µm, an equivalent figure/analyses could
be made/performed using any wavelength in this range as a refer-
ence point (see table 3). The LAGN
IR
these 11 AGNs span the range ∼1.3 to ∼3.5, with an average ratio
of ∼ 2.2 (derived from the average SED defined in §4). We do not
seeanycorrelationbetweenLAGN
IR
our sample. The average LAGN
IR
defined in Netzer et al. (2007) is slightly lower (i.e., ∼ 1.6) than
that of the lower luminosity AGNs considered here, although this
difference is not significant based on the range of ratios spanned by
our sample of intrinsic AGN SEDs. However, we note that the at
other wavelengths the Netzer et al. (2007) quasar bolometric cor-
rection differs considerably from the average bolometric correction
of more modest luminosity AGNs defined here (see table 3). We
also show in fig. 11 the LIR/νLν(12 µm) ratios for NGC 1068 and
IR
/νLν(12 µm) ratios for
/νLν(12 µm) ratios displayed by
/νLν(12µm) andνLν(12µm)for
/νLν(12 µm) ratio for quasar SEDs
Mrk 231, the latter of which is significantly higher than the range
of intrinsic ratios due to the additional host-galaxy components in
these SEDs (see §5.1).
Using the νLν(12 µm):L2−10keV correlation defined in
Gandhi et al. (2009), we convert our LAGN
LAGN
IR
2.2±1.3 (to cover the full range of ratios displayed by our sample)
we obtain:
IR
/νLν(12 µm) ratios into
/νLν(12 µm) =
/L2−10keVcorrection factors. Assuming LAGN
IR
log
?
LAGN
IR
1043erg s−1
?
=(0.53±0.26)+
(1.11±0.07)log
?
L2−10keV
1043erg s−1
?
(4)
For a L2−10keV= 1043ergs s−1AGN, typical of the range of
L2−10keVconsidered here, equation 4 gives L2−10keV/LAGN
This compares to a median ratio of L2−10keV/LAGN
quasar sample of Elvis et al. (1994), although in this case there will
be some contribution to LAGN
IR
from the host galaxy that has not
been accounted for, hence the lower limit (see also Alexander et al.
2005; we have assumed LIR∼ 1.3LFIR). We attribute the factor of
?7differencebetween theseratiostoboth theunaccounted-for host
galaxy contribution and the different fractions of the total bolomet-
ric luminosities of AGNs (i.e., LBol) that are emitted at X-ray ener-
gies by quasars and more modest AGNs (see Vasudevan & Fabian
2007). For example, L2−10keV= 1043ergs s−1AGNs emit approx-
imately 5 per cent of their bolometric power in the 2–10 keV band,
compared to just ∼ 1 per cent for a L2−10keV= 1045ergs s−1
quasar (see fig. 3 of Vasudevan & Fabian 2007; i.e., changes in
αOXwithluminosity; Avni & Tananbaum 1982; Wilkes et al.1994;
Vignali et al. 2003; Steffen et al. 2006). Using these conversion
factors, wefindthat LIR/LBolaresimilar for both moderate andhigh
IR
∼0.3.
IR
?0.04 for the