SPECTRAL ENERGY DISTRIBUTIONS AND MULTIWAVELENGTH SELECTION OF TYPE 1 QUASARS
Gordon T. Richards,
1,2
Mark Lacy,
3
Lisa J. Storrie-Lombardi,
3
Patrick B. Hall,
4
S. C. Gallagher,
5
Dean C. Hines,
6
Xiaohui Fan,
7
Casey Papovich,
7
Daniel E. Vanden Berk,
8
George B. Trammell,
8
Donald P. Schneider,
8
Marianne Vestergaard,
7
Donald G. York,
9,10
Sebastian Jester,
11, 12
Scott F. Anderson,
13
Tama
´
sBudava
´
ri,
2
and Alexander S. Szalay
2
Received 2006 January 24; accepted 2006 May 26
ABSTRACT
We present an analysis of the mid-infrared ( MIR) and optical properties of type 1 (broad-line) quasars detected by
the Spitzer Space Telescope. The MIR color-redshift relation is characterized to z 3, with predictions to z ¼ 7. We
demonstrate how combining MIR and optical colors can yield even more efficient selection of active galactic nuclei
(AGNs) than MIR or optical colors alone. Composite spectral energy distributions (SEDs) are constructed for 259
quasars with both Sloan Digital Sky Survey and Spitzer photometry, supplemented by near-IR, GALEX, VLA, and
ROSAT data, where available. We discuss how the spectral diversity of quasars influences the determination of bolo-
metric luminosities and accretion rates; assuming the mean SED can lead to errors as large as 50% for individual quasars
when inferring a bolometric luminosity from an optical luminosity. Finally, we show that careful consideration of the
shape of the mean quasar SED and its redshift dependence leads to a lower estimate of the fraction of reddened / obscured
AGNs missed by optical surveys as compared to estimates derived from a single mean MIR to optical flux ratio.
Subject h eadinggs: catalogs — galaxies: active — infrared: galaxies — quasars: general — radio continuum: galaxies —
surveys — ultraviolet: galaxies — X-rays: galaxies
Online material: machine-readable tables
1. INTROD UCTION
Access to the mid-infrared ( MIR) region opens up new realms
for quasar science as we are able to study large numbers of ob-
jects with high signal-to-noise ratio data in this bolometrically
important band for the first time. At least four distinct energy gen-
eration mechanisms are at work in active galactic nuclei (AGNs)
from jets in the radio, dust in the IR, accretion disks in the optical–
UV–soft–X-ray, and Compton upscattering in hot coronae in the
hard X-ray. All of these spectral regions need to be sampled with
high precision if we are to understand the physical processes
governing AGN emission. The Spitzer Space Telescope (Werner
et al. 2004) allows the first robust glimpse of the physics of the
putative dusty torus in AGNs out to z 23 and makes it pos-
sible to compare high-quality mid-IR data to the expectations of
the latest models (e.g., Nenkova et al. 2002; Dullemond & van
Bemmel 2005; Fritz et al. 2006).
MIR photometry from Spitzer has provided a better census of
active nuclei in galaxies than has been previously possible (e.g.,
Lacy et al. 2004). Optical surveys are biased against heavily
reddened and obscured objects, and even X-ray surveys may fail
to uncover Compton-thick sources (e.g., Treister et al. 2006).
Thus, the MIR presents an attractive window for determining the
black hole accretion history of the universe. To that end, Spitzer
will be of considerable utility in helping to decipher the nature of
the M
BH
- relation (e.g., Tremaine et al. 2002), in terms of mak-
ing a complete census of AGNs—a necessary condition for a full
understanding of the physical relationship between black holes
and their host galaxies.
High-sensitivity, high-accuracy MIR photometry also fills a
huge gap in our knowledge of the overall spectral energy dis-
tribution (SED) of AGNs, which now lacks only detailed far-IR/
centimeter and extreme-UV meas urements for a large sample of
quasars. Without the mid-IR data, we have been forced to rely on
the mean properties of a few dozen of the brightest quasars (e.g.,
Elvis et al. 1994) to estimate bolometric luminosities (and, in turn,
Eddington masses and accretion rates) for quasars. Since the 1–
100 m part of the spectrum contributes nearly 40% of the bolo-
metric luminosity, this added knowledge represents a significant
gain in our ability to explore the properties of AGNs as a function
of the bolometric luminosity.
This paper builds on and extends the results from recent pa-
pers describing the Spitzer MIR color distribution of AGNs.
Lacy et al. (2004) showed that MIR colors alone can be used to
select AGNs with both high efficiency and completeness, includ-
ing both dust-reddened and optically obscured (type 2) AGNs
that may otherwise be overlooked by optical selection techniques.
We will show that the addition of optical colors and morphology
can be used to improve the MIR-only selection efficiency of type 1
quasars (including those that are moderately reddened).
A
1
Princeton University Observatory, Peyton Hall, Princeton, NJ 08544.
2
Department of Physics and Astronomy, The Johns Hopkins University,
3400 North Charles Street, Baltimore, MD 21218-2686.
3
Spitzer Science Center, California Institute of Technology, Mail Code 220-6,
Pasadena, CA 91125.
4
Department of Physics and Astronomy, York University, 4700 Keele Street,
Toronto, ON M3J 1P3, Canada.
5
Department of Physics and Astronomy, UCLA, Mail Code 154705, 475
Portola Plaza, Los Angeles, CA 90095.
6
Space Science Institute, 4750 Walnut Street, Suite 205, Boulder, CO 80301.
7
Steward Observatory, University of Arizona, 933 North Cherry Avenue,
Tucson, AZ 85721.
8
Department of Astronomy and Astrophysics, Pennsylvania State Univer-
sity, 525 Davey Laboratory, University Park, PA 16802.
9
Department of Astronomy and Astrophysics, University of Chicago, 5640
South Ellis Avenue, Chicago, IL 60637.
10
Enrico Fermi Institute, University of Chicago, 5640 South Ellis Avenue,
Chicago, IL 60637.
11
Fermi National Accelerator Laboratory, P.O. Box 500, Batavia, IL 60510.
12
School of Physics and Astronomy, Southampton University, Southampton
SO17 1BJ, UK.
13
Department of Astronomy, University of Washington, Box 351580, Seattle,
WA 9 8 1 9 5.
470
The Astrophysical Journal Supplement Series, 166:470–497, 2006 October
# 2006. The American Astronomical Society. All rights reserved. Printed in U.S.A.
Stern et al. (2005) also describe a MIR selection technique for
AGNs, making statistical arguments that the obscured AGN frac-
tion may be as high as 76%. We reconsider their argument in light
of the influence that the shape of the mean quasar spectral energy
distribution (SED) has on determining the obscured quasar frac-
tion. Such considerations allow us to demonstrate that the true ob-
scured AGN fraction must be lower than that determined by Stern
et al. (2005).
Finally, Hatziminaoglou et al. (2005) investigated the combined
optical + MIR color distribution of quasars by combining data
from the ELAIS-N1 field in the Spitzer Wide-Area Infrared Extra-
galactic Survey (SWIRE; Lonsdale et al. 2003) with data from the
Sloan Digital Sky Survey (SDSS; York et al. 2000). Using the data
from 35 SDSS quasars they determine the mean optical-MIR SED
of type 1 quasars and investigate their mass and bolometric lumi-
nosity distribution. We expand on these results by determining a
number of different ‘‘mean’’ SEDs as a function of color and lumi-
nosity for 259 SDSS quasars in the Spitzer Extragalactic First Look
Survey
14
(XFLS), SWIRE
15
ELAIS-N1/N2, and SWIRE Lockman
Hole areas. We use these SEDs to demonstrate that the diversity of
quasar SEDs must be considered when determining bolometric
luminosities and accretion rates for individual quasars—as was
emphasized in the seminal SED work of Elvis et al. (1994).
Section 2 reviews the data sets used in our analysis. In x 3we
explore the MIR color-redshift relation and MIR-optical color-
color space occupied by type 1 quasar s. In addition to showing
these relations for the data, we also show the predicted relations
derived from two quasar SEDs convolved with the SDSS and
Spitzer filters curves : one SED derived largely from broadband
photometry ( Elvis et al. 1994), the other from a mean optical +
IR spectral template (Glikman et al. 2006). Section 4 presents a
brief discussion of the determination of the type 1 to type 2 ratio
of quasars. In x 5 we discuss the radio through X-ray SED of qua-
sars and construct new MIR-optical templates from our sample.
We present an overall mean SED along with mean SEDs for sub-
sets of optically luminous/dim, MIR luminous/dim, and optically
blue/red quasars in order to explor e how different optical/MIR
properties are related to the overall SED. Section 6 discusses the
implications of our new SED templates on the determination of
bolometric luminosities and accretion rates. Our conclusions are
presented in x 7.
Throughout this paper we will distinguish between normal
type 1 quasars, dust-reddened/extincted type 1 quasars, and type 2
quasars. By ‘‘type 1 quasars,’’ we mean those quasars having broad
lines and optical colors/ spectral indices that are roughly consistent
with a Gaussian spectral index distribution of
¼0:5 0:3
( f
/
). Red dened type 1 quasars are those quas ars that have
broad lines but have spectral indices that are redder than about
¼1 (e.g., Gregg et al. 2002). Optical surveys can find such
quasars up to E(B V ) 0:5butareincreasinglyincomplete
above E(B V ) 0:1 (Richards et al. 2003). By type 2 quasars,
we mean those that lack rest-frame optical/UV broad emission
lines and have nuclei that are completely obscured in the optical
such that the optical colors are consistent with the host galaxy.
Throughout this paper we use a CDM cosmology with H
0
¼
70 km s
1
Mpc
1
,
¼ 0:7, and
m
¼ 0:3, consistent with the
WMAP cosmology (Spergel et al. 2003, 2006).
2. THE DATA
We investigate the mid-IR and optical properties of type 1
quasars that are detected in both the SDSS and in all four bands
of the Spitzer Infrared Array Camera ( IRAC; Fazio et al. 2004).
The Spitzer data are taken from the XFLS and SWIRE ELAIS-N1,
ELAIS-N2, and Lockman Hole areas, which have (R.A., decl.)
centers of (259N5, 59N5), (242N75, 55N0), (249N2, 41N029), and
(161N25, 58N0), respectively.
We begin with SDSS-DR3 type 1 quasars cataloged by
Schneider et al. (2005), the majority of which were selected by
the algorithm given by Richards et al. (2002). This catalog in-
cludes matches to the FIRST (Becker et al. 1995) survey with the
VLA, ROSAT ( Voges et al. 2000), and 2MASS (Skrutskie et al.
1997). For a definition of the SDSS photometric system, see
Fukugita et al. (1996); Adelman-McCarthy et al. (2006) provide
a description of the latest SDSS data release (DR4). All SDSS
magnitudes have been corrected for Galactic extinction accord-
ing to Schlegel et al. (1998).
The 46,420 SDSS quasars of Schneider et al. (2005) are
matched to IRAC detections in the XFLS (main
_
4band.cat;
Lacy et al. 2005b) and the SWIRE ELAIS-N1, -N2, and Lockman
Hole (SWIRE2
_
N1
_
cat
_
IRAC24
_
16jun05.tbl, SWIRE2
_
N2
_
cat
_
IRAC24
_
16jun05.tb l, SWIRE2
_
Lockman
_
cat
_
IRAC24
_
10Nov05.tbl; Surace et al. 2005) areas of sky. The IRAC band-
passes are generally referred to as channels 1 through 4 or as the
3.6, 4.5, 5.8 , and 8.0 m bands, respectively. For a quasar spec-
trum with MIR spectral index of
¼1(f
/
), the effec-
tive wavelengths of the IRAC bandpasses are actually closer to
3.52, 4.46, 5.67, and 7.70 m. The SWIRE catalogs also include
24 m photometry from the Multiband Imaging Photometer for
Spitzer (MIPS; Rieke et al. 2004). In the XFLS field, 24 m
sources are cataloged by Fadda et al. (2006) and we include
matches from that catalog as well. As the limits of the mid-IR
catalogs are much deeper than the SDSS spectroscopic survey,
we consider only objects detected in all four IRAC bands. Within
a matching radius of 1B0 there are 44 SDSS-DR3 quasar matches
in the XFLS area, 29 in the ELAIS-N1 area, 44 in the ELAIS-N2
area, and 142 in the Lockman Hole area. All but one of the op-
tically selected SDSS quasars has four-band IRAC coverage in
the regions of overlap between the SDSS and Spitzer data; see
Figures 1 and 2. The ex ception is SDSS J104413.47 +580858. 9
(z ¼ 3:7), which has only a limit in IRAC channel 3.
To construct the most detailed quasar spectral energy distribu-
tions (SEDs) possible, we include data available at other wave-
lengths. We include matches to MIPS 70 m sources in the XFLS
(FLS70
_
sn7
_
jul05.txt; Frayer et al. 2006) and in the SWIRE
(SWIRE2
_
EN1
_
70um_23nov05.tbl, SWIRE2
_
EN2
_
70um
_
23nov05.tbl, SWIRE3
_
Lockman
_
70um
_
23nov05.tbl; Surace
et al. 2005) areas. No MIPS 160 m data are included as the
flux density limits of these data in the XFLS and SWIRE areas
are much brighter than expected flux densities of even the bright-
est SDSS-DR3 quasars in these fields. For the SDSS quasars in
the ELAIS fields we have extracted 15 mphotometryfromthe
Rowan-Robinson et al. (2004) catalog. We also extract J /H/K and
radio information from this catalog if that information was not
otherwise available.
Some of these areas of sky have been observed by GALEX
(Martin et al. 2005), and the data were released as part of GALEX
GR1. Quasars are readily detected by GALEX (see Bianchi et al.
2005 and Seibert et al. 2005); thus, we also include GALE X pho-
tometry where available. Matc hing of the GALEX catalogs and
the SDSS DR3 quasar sample is described by Trammell et al.
(2005). The effective wavelengths of the GALEX NUV and
FUV bandpasses (hereafter referred to as n and f magnitudes)
are 2267 and 1516 8. GALEX photometry has been corrected
for Galactic extinction assuming A
n
/E(B V ) ¼ 8:741 and
A
f
/E(B V ) ¼ 8:376 (Wyder et al. 2005). A total of 55 and 88
14
See http://ssc.spitzer.ca ltech.edu/fls/.
15
See http://swire.ipac.caltech.edu/swire/.
SEDs OF TYPE 1 QUASARS 471
of the DR3 quasars have GALEX detections in the f and n bands,
respectively.
In the radio, we have matched to the deeper VLA data taken in
the XFLS area by Condon et al. (2003), which catalogs 5 de-
tections with fluxes higher than 115 Jy (about an order of mag-
nitude deeper than FIRST). Deep VLA data also exists for the
ELAIS and Lockman Hole areas, but only over a small area of
sky (e.g., Ciliegi et al. 1999, 2003).
Most of our objects are fainter than the 2MASS (Skrutskie
et al. 1997) limits, but we have supplemental near-IR data for a
few. Near-IR (JHK
s
) magnitudes for SDSS J1716+5902 were ob-
tained on 2003 September 9 UT using the GRIM II instrument on
the Apache Point Observatory 3.5 m telescope. Dithered images
were obtained and reduced in the standard fashion, using running
flat-fielding and sky-subtraction (e.g., Hall et al. 1998) with all
available good images in a given filter for each object. Four
other sources (SDSS J171732.94+59474 7.5, SDSS J171736.90+
593011.4, SDSS J171748.43+594820.6, and SDSS J171831.73+
595309.4) were observed at Palomar Observatory.
Finally, to better characterize the optical + MIR color distri-
bution of type 1 quasars, we include 87 broadline quasars that are
fainter than the SDSS spectroscopic magnitude limit, but that
Fig. 2.—Location of SDSS-DR3 quasars in the SWIRE ELAIS N1 (left)andN2(right) fields. Red points indicate four-band IRAC sources. Blue points indicate
MIPS 70 m sources. Open triangles indicate SDSS-DR3 quasars. Green circles indicate SDSS-DR3 quasars with IRAC detections in all four bands.
Fig. 1.—Location of SDSS-DR3 quasars in the XFLS (left) and SWIRE Lockman Hole (right) fields. Red, yellow, and blue points represent IRAC, IRAC
verification, and MIPS70 sources, respectively. Open triangles represen t SDSS-DR3 quasars. Green circles represent SDSS-DR3 quasars with IRAC detections in all
four bands. Open pentagons indicate GALEX-detected SDSS quasars.
RICHARDS ET AL.472 Vol. 166
TABLE 1
SDSS-Spitzer Quasar Photometry I
Name (SDSS J) z
em
L
bol
a
log (ergs s
1
)
L
opt
b
log (ergs s
1
)
L
ir
c
log (ergs s
1
)BC
a
X-Ray
log (counts s
1
)
f
(AB mag)
n
(AB mag)
u
(AB mag)
g
(AB mag)
r
(AB mag)
i
(AB mag)
z
(AB mag)
105705.39+580437.4 .......... 0.140 45.06 44.49 44.78 10.60 0.708 18.31 0.08 18.15 0.04 17.92 0.03 17.61 0.05 17.25 0.02 16.83 0.02 16.56 0.04
171902.28+593715.9 .......... 0.178 45.21 44.74 44.93 9.41 1.221 18.10 0.01 17.99 0.01 17.49 0.02 17.50 0.02 17.36 0.02 17.06 0.02 17.20 0.02
160655.34+534016.8 .......... 0.214 45.13 44.45 44.91 11.87 ... ... ... 18.85 0.03 18.71 0.02 18.22 0.02 17.86 0.02 17.91 0.03
163111.28+404805.2........... 0.258 45.68 45.27 45.19 9.84 0.551 ... ... 16.98 0.01 17.05 0.02 17.08 0.01 17.10 0.01 16.86 0.01
171207.44+584754.4 .......... 0.269 45.49 45.06 45.12 12.29 1.235 17.97 0.01 18.08 0.01 17.83 0.02 17.93 0.02 17.88 0.02 17.94 0.02 17.51 0.02
171033.21+584456.8 .......... 0.281 45.15 44.46 44.95 10.44 ... 20.59 0.04 20.06 0.02 19.58 0.03 19.25 0.03 18.70 0.02 18.52 0.02 18.10 0.03
105644.52+572233.4 .......... 0.286 45.08 44.53 44.78 10.12 ... ... ... 19.36 0.03 19.26 0.02 18.88 0.02 18.69 0.02 18.32 0.02
104739.49+563507.2 .......... 0.303 45.19 44.66 44.88 9.82 ... ... ... 19.16 0.04 19.02 0.04 18.72 0.04 18.57 0.03 18.20 0.03
155936.13+544203.8 .......... 0.308 45.42 44.87 45.15 11.75 ... ... ... 18.55 0.03 18.42 0.04 18.27 0.02 18.38 0.03 17.87 0.03
105626.96+580843.1 .......... 0.342 45.29 44.66 45.03 8.40 ... ... 21.50 0.20 19.45 0.03 18.95 0.02 18.57 0.03 18.46 0.02 17.88 0.02
Note.—Table 1 is available in its entirety in the electronic edition of the Astrophysical Journal Supplement. A portion is shown her e for guidance regarding its form and content.
a
Bolometric (100 m to 10 keV) luminosity and bolometric correction (from 5100 8).
b
1–0.1 m integrated luminosity.
c
100–1 m integrated luminosity.
TABLE 2
SDSS-Spitzer Quasar Photometry II
Name (SDSS J) J (Vega) H (Vega) K (Vega)
S
3:6
(Jy)
S
4:5
(Jy)
S
5:8
(Jy)
S
8:0
(Jy)
S
15
(mJy)
S
24
(mJy)
S
70
(mJy)
Radio
(mJy)
L
rad
log (ergs s
1
Hz
1
)
105705.39+580437.4 ........... 14.99 0.08 14.21 0.09 13.48 0.07 2351.5 5.6 2366.9 7.3 2838.0 15.1 6273.4 16.1 ... 16.61 0.02 98.0 0.6 ... <29.69
171902.28+593715.9 ........... 15.89 0.09 15.02 0.09 14.15 0.06 2925.1 293.1 4095.1 409.8 5365.6 541.1 7193.8 720.4 ... 26.91 0.04 22.9 4.0 0.23 29.28
160655.34+534016.8 ........... 16.37 0.10 15.33 0.11 14.32 0.07 1396.7 4.6 1657.2 5.7 2047.6 14.1 2973.1 10.9 7.72 14.80 0.02 37.7 1.6 ... <30.09
163111.28+404805.2............ 16.24 0.10 15.45 0.12 14.48 0.09 2729.7 5.0 3632.2 6.6 4686.4 15.1 6218.9 11.9 ... 16.90 0.03 ... ... <30.26
171207.44+584754.4 ........... 16.26 0.10 15.36 0.09 14.61 0.10 2024.6 203.1 2411.9 242.2 3162.9 321.5 4353.7 437.9 ... 13.34 0.07 ... 0.14 29.45
171033.21+584456.8 ........... 16.90 0.18 15.68 0.11 14.96 0.10 589.2 59.9 708.4 71.8 709.8 78.1 1571.6 159.5 ... 6.06 0.07 44.0 8.0 ... <30.34
105644.52+572233.4 ........... 16.71 0.10 16.33 0.27 15.10 0.12 1161.4 4.4 1280.1 5.2 1417.1 13.2 1742.5 9.7 ... 3.18 0.02 ... ... <30.36
104739.49+563507.2 ........... 16.63 0.16 16.24 0.24 15.51 0.18 572.2 2.6 671.2 2.6 886.3 10.7 1566.3 6.5 ... 8.61 0.02 ... ... <30.41
155936.13+544203.8 ........... 16.61 0.20 16.77 0.10 14.92 0.16 1093.4 2.6 1437.4 3.8 1997.8 8.6 3268.1 8.7 ... 14.59 0.02 ... 3.40 30.96
105626.96+580843.1 ........... 16.74 0.17 16.25 0.27 15.42 0.16 1362.5 5.2 1660.5 4.9 1975.0 15.5 2303.3 9.0 ... 4.23 0.02 ... ... <30.53
Note.—Table 2 is available in its entirety in the electronic edition of the Astrophysical Journal Supplement. A portion is shown her e for guidance regarding its form and content.
![Fig. 9.—Mean quasar SEDs compared to various survey flux limits. The black solid and dashed lines are the z ¼ 0 Elvis et al. (1994) mean radio-quiet and radio-loud SEDs, normalized to i ¼ 16:4 (1 mJy). The colored dashes show the flux density limits for the Spitzer XFLS (red), Spitzer Boötes (brown; Stern et al. 2005), 2MASS (green), SDSS (imaging, blue), SDSS (low-z quasar spectroscopy, orange), andGALEX (cyan) surveys. The lighter gray lines show the radio-quiet SED at z ¼ 0:5, 1.5, and 2.5 normalized to the Boötes field 8 m flux limit [dashed gray line is z ¼ 1:5 SMC-reddened by E(B V ) ¼ 0:25]. The dotted line shows a spectral index of ¼ 0:73 between 8 m and the SDSS r band.](/figures/fig-9-mean-quasar-seds-compared-to-various-survey-flux-12jrt9tr.webp)
![Fig. 11.—Mean quasar SEDs and the data used to construct them. The points show the data; the lines show themean SEDs. From left to right: black crosses, MIPS70; black points, MIPS24/ ISO15; red points, S8:0; green points, S5:8; cyan points, S4:5; blue points, S3:6; gray points, K; yellow points, H; cyan points, J; black points, z; magenta points, i; red points, r; green points, g; blue points, u; orange points, n; purple points, f ; black crosses, X-ray. The dotted lines show the range of ox with luminosity for the luminosity extremes of our sample (Strateva et al. 2005). From top to bottom (at 1014 Hz) the thick solid lines are color coded as follows: green, optically luminous SDSS quasars; red, optically red SDSS quasars; cyan, all SDSS quasars; blue, optically blue SDSS quasars; black, Elvis et al. (1994) radio-quiet mean SED; gray, Hatziminaoglou et al. (2005) mean SED (normalized to Elvis et al. 1994 at 1 m); orange, optically dim SDSS quasars. The near-IR luminous/dim composites are nearly identical to the optical composites. The thin green, cyan, and orange curves show the host galaxy contribution assuming (LBol/LEdd) ¼ 1 (see eq. [1]) for the elliptical galaxy composite spectrum of Fioc & Rocca-Volmerange (1997). The dashed cyan curve shows the effect of ignoring the host contribution. The inset at the upper right zooms in on theMIR region of the spectrum. The three curves are normalized at 1.3 m, and the y-axis shows relative luminosity ( L ), while the x-axis is in microns. The bottom panel shows the 1 error array (dex) for our overall mean SED. Those frequencies where over 100 quasars contribute to the mean SED are shown as thicker line segments and are relatively unaffected by the gap-repair process.](/figures/fig-11-mean-quasar-seds-and-the-data-used-to-construct-them-3qjowc2a.webp)