Co-evolution of Extreme Star Formation and Quasar: hints from {\it Herschel} and the Sloan Digital Sky Survey

1.2 MB·English

by  Zhiyuan Ma

Draftversion August7,2015
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
CO-EVOLUTION OF EXTREME STAR FORMATION AND QUASAR: HINTS FROM HERSCHEL AND THE
SLOAN DIGITAL SKY SURVEY
Zhiyuan Ma1,† and Haojing Yan1,††
Draft version August 7, 2015
ABSTRACT
5 Using the public data from the Herschel wide field surveys, we study the far-infrared properties of
1 optical-selected quasars from the Sloan Digital Sky Survey. Within the common area of ∼ 172deg2,
0
we have identified the far-infrared counterparts for 354 quasars, among which 134 are highly secure
2
detections in the Herschel 250µm band (signal-to-noise ratios ≥ 5). This sample is the largest far-

g infraredquasarsampleofitskind,andspansawideredshiftrangeof0.14≤z ≤4.7. Theirfar-infrared
u spectral energy distributions, which are due to the cold dust components within the host galaxies,
A areconsistentwith being heated by activestar formation. In mostcases(&80%),their totalinfrared
6 luminositiesasinferredfromonlytheirfar-infraredemissions(L(IcRd))alreadyexceed1012L⊙,andthus
these objects qualify as ultra-luminous infrared galaxies. There is no correlation between L(cd) and
IR
] theabsolutemagnitudes,theblackholemassesortheX-rayluminositiesofthequasars,whichfurther
A
supportthattheirfar-infraredemissionsarenotduetotheiractivegalacticnuclei. Alargefractionof
G these objects (& 50–60%) have star formation rates & 300M⊙yr−1. Such extreme starbursts among
opticalquasars,however,isonlyafewpercent. Thisfractionvarieswithredshift,andpeaksataround
.
h z ≈ 2. Among the entire sample, 136 objects have secure estimates of their cold-dust temperatures
p (T), and we find that there is a dramatic increasing trend of T with increasing L(cd). We interpret
- IR
o this trend as the envelope of the general distribution of infrared galaxies on the (T, L(cd)) plane.
r IR
t Subject headings: infrared: galaxies; galaxies: starburst; galaxies: high-redshift, galaxies: evolution,
s
a (galaxies:) quasars: general
[

2 1. INTRODUCTION the IRAS S60 > 0.6Jy galaxy sample, Lawrence et al.
v (1999) have found that ∼ 20% of their 95 ULIRGs are
Ultra-Luminous InfraRed Galaxies (ULIRGs), discov-
0
AGN. The consensus is that ULIRGs that have “warm”
4 ered as a distinct class in the early 1980’s (Houck et al.
IRcolors,i.e.,whoseemissionstendtopeakatrestframe
2 1984,1985;Aaronson & Olszewski1984)bythe Infrared
mid-IR (MIR) rather than far-IR (FIR), generally host
1 Astronomical Satellite (IRAS) survey of the sky in 12
(optical) AGN, which could be the main energy sources
0 to 100µm, are believed to host extreme star formation
thatpowertheirstrongIRemissions(e.g.,de Grijp et al.
. regions that are heavily enshrouded by dust. They are
1 1985; Osterbrock & De Robertis 1985; Kim & Sanders
characterized by their exceptionally high IR luminosi-
50 ties (LIR > 1012L⊙; integrated over restframe 8 to 1th9e9i8r)I.ROemnistshioenostpheearkhaatnFdI,R“schooldu”ldUbLeImRoGsstlythpaotwhearevde
1000µm), which are believed to be predominantly due
1 by starbursts (e.g., Elston et al. 1985; Heckman et al.
to the re-radiation of star light processed by dust (see
v: Lonsdale et al. 2006 for a review), and thus imply very 1987).
Xi high star formation rates (SFR) of > 100–1000M⊙yr−1 ULAInRGinst,eersepsetciniagllyquqeusatsiaornUtLhIeRnGiss,hwahveetshtearrbaunrystsAtGhaNt
(usingtheconversionofKennicutt1998)completelyhid-
r dominatetheirIRemissions. Quasarsrepresentthemost
a den by dust. extreme process of supermassive black hole accretion,
However,ithasbeennoticedeversincetheirdiscovery
whilestarburstsarethemostextremeprocessofstarfor-
that a significant fraction of ULIRGs, especially those
mation. Itisexpectedthattheinterplayofthesetwoex-
very luminous ones, have optical signatures indicative
tremes will have important consequences. This is made
of classic AGN activities. For examples, Carter (1984)
particularly important by the quasar evolutionary sce-
notes that ten among a sample of 13 IRAS sources with
nario of Sanders et al. (1988), as such objects could be
60µm flux density f ≥ 1.2Jy are Seyfert galaxies.
60 the transitional type between non-quasar ULIRGs and
Sanders et al. (1988) show that ten galaxies among the
“fully exposed” quasars. Sanders et al. themselves be-
324 sources with f ≥ 5.4Jy from the IRAS Bright
60 lieve that AGN heating is the main mechanism for the
Galaxy Survey are ULIRGs and that they all have a
strong IR emission of ULIRG quasars. In their dis-
mixture of starburstand AGN signatures,which has led
cussion of PG quasar continuum distributions from UV
them to propose an evolutionary scenario that ULIRGs
to millimeter (mm), Sanders et al. (1989) further pro-
are the prelude to quasars. In their redshift survey of
pose that a warped galactic disk (beyond the central
∼ 10pc to a few kpc) heated by the central AGN can
1Department of Physics and Astronomy, University of explain the entire range of IR emission from 5µm to
Missouri-Columbia
†zmzff@mail.missouri.edu 1mm. However,Rowan-Robinson(1995)arguesthatthis
††[email protected] scheme could only be viable when the total luminosity
2 Ma & Yan
in IR is comparable or less than that in UV-optical; in- ing instruments, namely, the Photodetector Array Cam-
stead,hehassuccessfullymodeledtheIRAS-detectedPG era and Spectrometer (PACS, Poglitschet al. 2010) and
quasars by attributing their mid-IR emissions to AGN theSpectralandPhotometricImagingREceiver(SPIRE,
heating and the FIR emission to starburst heating, re- Griffin et al. 2010). The PACS bands are 100 (or 70)
spectively. More analysis using larger samples from the and 160µm, and the SPIRE bands are 250, 350 and
Infrared Space Observatory (ISO) observations support 500µm. Together they sample the peak of heated dust
this view. For example, Haas et al. (2003) have studied emissionfromz ≈0to6andbeyond. Therealreadyhave
64 PG quasars and have concluded that starburst heat- beenanumberofstudies ontheFIRemissionofquasars
ing is more likely the cause of the observed cold dust using Herschel observations (e.g., Serjeant et al. 2010;
(∼ 30–50K) FIR emissions among the ULIRG part of Leipski et al. 2010, 2013; Dai et al. 2012; Netzer et al.
the sample. However, they have also pointed out that 2014), however the current collection of quasars that
AGN heating should be the main power source for those have individual Herschel detections are still very scarce
extreme ones that qualify as “Hyper-luminous Infrared in number and few have spanned a sufficient redshift
Galaxies”(HyLIRG; usually defined by LIR >1013L⊙). range (for examples, Leipski et al. (2013) present 11 ob-
There are also two pieces of important, albeit in- jects at z > 5; Dai et al. (2012) include 32 objects at
direct, evidence supporting that the FIR emissions of 0.5 ≤ z ≤ 3.6; Netzer et al. (2014) report ten within a
quasar ULIRGs are likely due to starbursts. First, a narrow window at z ≈4.8).
significant fraction of such objects have a large amount In this paper, we present a large sample of optical
of molecular gas (e.g., Sanders et al. 1988; Evans et al. quasars that are detected by the Herschel, and provide
2001; Scoville et al. 2003; Xia et al. 2012), which is our initial analysis of their FIR properties. The quasars
a strong indicator of active star formation. Second, arefromtheSloanDigitalSkySurvey(SDSS;York et al.
most quasar ULIRGs have polycyclic aromatic hydro- 2000), and the FIR data are from the public releases
carbon (PAH) features, which are also strongly indica- of four major wide field surveys by Herschel, namely,
tive of on-going star formation (e.g., Schweitzer et al. the HerschelAstrophysicalTerahertzLargeAreaSurvey
2006;Shi et al.2007;Hao et al.2007;Netzer et al.2007; (H-ATLAS; Eales et al. 2010), the Herschel Multi-tiered
Cao et al.2008). Whilenoneofthesearesufficienttoas- Extragalactic Survey (HerMES; Oliver et al. 2012), the
sertthatstarburstsdominatethestrongFIRcontinuaof HerschelStripe82Survey(Viero et al.2014,HerS;),and
quasar ULIRGs, it is clear that they at least contribute the PACS Evolutionary Probe (PEP; Lutz et al. 2011).
significantly. We describe the data and the sample constructionin §2,
The picture above is largely based on the ULIRGs in and present our analysis of the FIR dust emission in §3.
the nearby universe where they can be studied in de- Theimplicationsofourresultsaredetailedin§4,andwe
tail. It would not be surprising if any of it changes concludewithasummaryin§5. Thecatalogofoursam-
at high redshifts, as both quasars and ULIRGs evolve ple is available as online data in its entirety. All quoted
strongly. The number density of quasars rises rapidly magnitudes in the paper are in the AB system. We
from z = 0 to z = 1, and reaches the peak at z ≈ 2–3 adoptthefollowingcosmologicalparametersthroughout:
(e.g., Osmer 2004). Similarly, while ULIRGs are very Ω =0.27, Ω =0.73 and H =71kms−1Mpc−1.
M Λ 0
rare objects today, they are much more numerous in
earlier epochs. The deep ISO surveys have revealed 2. DATADESCRIPTIONANDSAMPLECONSTRUCTION
a large number of IR-luminous galaxies, among which Inbrief,webuiltoursamplebysearchingforthecoun-
> 10% are ULIRGs and many are at z > 1 (e.g., terparts of the SDSS quasars in the Herschel wide field
Rowan-Robinsonet al. 2004). The discovery of the so- surveydata. Forthesakeofsimplicity,hereafterwerefer
called“submillimeter (submm) galaxies”(SMGs) at 450 to these objects as “IR quasars”. We describe below the
and 850µm (see Blain et al. 2002 for a review) added a data used in our study and the constructed IR quasar
new population to the ULIRG family, as most of them sample.
are at z ≈ 2–3 and have LIR > 1012L⊙ dominated by
the emission from cold dust. Furthermore, Spitzer ob- 2.1. Parent quasar samples
servations suggest that high stellar mass (> 1011M⊙) The parent quasar samples that we used are based on
and otherwise “normal” star-forming galaxies at z ∼ 2
the SDSS Data Release 7 and 10 quasar catalogs (here-
arelikely all ULIRGs (e.g.,Daddi et al.2005), whichin-
after DR7Q and DR10Q, respectively), which are sum-
creasestheULIRGnumberdensityathighredshiftstoa
marized as follows.
more dramatic level than expected. On the other hand,
molecular gas has also been detected in quasar ULIRGs DR7Q: As detailed in Schneider et al. (2010), this
from z & 1 to 6 (Solomon & Vanden Bout 2005, and quasar catalog is based on the SDSS DR7. It con-
the references therein; see also e.g., Wang et al. 2010, cludes the quasar survey in the SDSS-I and SDSS-II
2011a,b for the recent results at z ∼ 6), lending sup- over 9380deg2, and supersedes all previously released
port to the starburst-powered interpretation of the FIR SDSS quasar catalogs. It includes 105783 quasars be-
emission of such objects at high redshifts as well. tween z = 0.065 and 5.46 (the median at z = 1.49),
If quasars and dust-enshroudedstarbursts do co-exist, allwith absolutei-bandmagnitudes(M )brighterthan
i
it is important to investigate their co-evolution, which −22mag.
would require a large sample over a wide redshift range.
Herschel SpaceObservatory(Pilbratt et al.2010)hasof- DR10Q: This quasar catalog is derived from the on-
fered an unprecedented opportunity to investigate this going Baryon Oscillation Spectroscopic Survey (BOSS)
problem at the FIR wavelengths that were largely unex- as part of the SDSS-III. While it was released in the
plored by the previous studies. Herschel had two imag- SDSS DR10, its main targetselection was basedon the
Herschel-detected SDSS Quasars 3
111...050 0.2
1.0 10 0.1
R
0.5 Dz
00..88 0.0
−
0.0
R10 −0.5 DR7z −0.1 ∆z=−0.001±0.048
Dag−001...660 ∆mu=0.00±0.42 ∆mg=−0.01±0.29 ∆mr=−0.01±0.28 −0.20 1 2 3 4 5 6
m 1.5
− 16182022242628 zDR10
7 100...044 1.5
R
D
g 0.5 0 1.0
ma 0.0 DR1Mi 0.5
−000...225 − 0.0
7 −0.5
−1.0 ∆mi=−0.01±0.28 ∆mz=−0.01±0.28 DRMi −1.0 ∆Mi=0.05±0.13
−001...005 −1.5
001..400161820202.224260.228106.4182022240.2066.428 0.8 0.61.0 0−.831−28−25−221.0
magDR10 MDiR10
Figure 1. Comparisonofthephotometry, theredshiftsandtheabsolutemagnitudesfromtheDR7QandtheDR10Qfortheoverlapped
population in these two quasar catalogs (16356 objects in total). For this particular subset, we adopt the DR10Q values in this paper
unless noted otherwise. For the sake of consistency, we use the absolute magnitudes k-corrected to z =2 for the DR7Q quasars as well,
adaptingthevalues fromthecatalogpresentedinShenetal.(2011).
data set over the widest area to date. The redshift
12
distributions of DR7Q, DR10Q and the merged cata-
SDSSDR7
10 log are shown in Figure 2. For simplicity, hereafter we
USELFLAG=1
refer to the merged sample as the “SDSS quasar sam-
)
30 8 SDSSDR10 ple”andthequasarsthereinasthe“SDSSquasars”. We
1
× UNIFROM>0 note that DR7Q and DR10Q are statistically different
( 6
ts DR7+DR10 samples, and therefore any statistical results from this
n 4 merged catalog should be inferred with caution. This is
u
Co exacerbated by the fact that the quasars were not se-
2 lected uniformly within either DR7Q or DR10Q, as de-
tailed in Schneider et al. (2010) and Paˆris et al. (2014),
0
0 1 2 3 4 5 6 respectively. For example, while most of the quasars in
DR7Q were selected using the algorithms as described
Redshift
in Richards et al. (2002) (with USELFLAG=1 in the cat-
alog), a non-negligible fraction of them were selected
Figure 2. Redshiftdistributionof the SDSS DR7Q(red curves)
and DR10Q catalogs (yellow curves). The solid ones are for the early in the SDSS campaign when such algorithms had
entire catalogs, while the dashed ones are for the “homogeneous” not yet been fully developed. DR10Q is even more
subsamplesasdescribedinSchneideretal.(2010)andPˆarisetal.
non-uniform in this sense, because only about half of
(2014).
its quasars (called the “CORE” sample) were selected
SDSS DR8. The detailed description of the catalog can uniformly (with UNIFORM>0 in the catalog) through the
be found in Paˆris et al. (2014). It includes new quasars XDQSO method (Bovy et al. 2011) and the other half
with M [z = 2] < −20.5mag from the SDSS-III, where (called the “BONUS” sample) were selected using vari-
i
M [z =2] is the absolute magnitude k-correctedto z = ous different methods (see Ross et al. 2012). Neverthe-
i
2 (for details, see Paˆris et al. 2014). It also includes a less,the non-uniformitydoesnotaffectourcurrentwork
largenumberofknownquasarsofsimilarcharacteristics sincewelimitourstudytotheFIRpropertiesofoptical-
(mostly from SDSS-I and II) that were re-observed by selected quasars, whose being selected did not use any
BOSS. In brief, the catalog contains 166583 quasars FIR information and thus should not favor or against
over6373deg2,withredshiftsrangingfrom0.05to5.86. any given FIR property.
2.2. Herschel Data
The quasars from these two catalogs are largely in-
dependent, however there are still 16356 of them being We utilized all high-level,publicly availabledata from
duplicates, which we define as the ones falling within a the wide field Herschel surveys. In all cases we adopted
matching radius of 0.4′′. Figure 1 shows the compari- the latest catalogs released by the survey teams to con-
son of the photometry, the redshift measurements and structtheFIRspectralenergydistributions(SEDs). The
the absolutemagnitudesfromthese twocatalogsforthis basic characteristics of these surveys are summarized in
overlapped population, which all agree reasonably well. Table 1, and their relevant details are briefly described
For the sake of simplicity, we adopt the DR10Q values below.
in this work for these duplicates unless noted otherwise.
2.2.1. HerMES
In the end, we produced a merged sample of 256010
unique quasars, which represents the largest optical HerMESwasthelargestHerschel guaranteedtimekey
quasar sample selected based on the most homogeneous program, and surveyed 100deg2 in six levels of depth
4 Ma & Yan
Table 1
SummaryofHerschel widefielddataandIRQSOsample
Survey Field Level Coverage 5σ limit Alla Match SNR S250>56.6
(deg2) (mJy) <3′′ >5 mJy
HerMES COSMOSb L1 5.00 31.4 216 47 7 3
GOODS-Northb L2 0.63 27.1 17 4 1 1
Bootes-HerMES L5 11.29 31.3 426 50 18 2
EGS-HerMESb L5 3.12 29.7 53 6 3 0
(Groth-Strip) L3 30.9 3 2 0
ELAIS-N1-HerMES L5 3.74 30.4 22 3 3 0
Lockman-SWIREb L5 19.73 37.0 276 35 12 5
(Lockman-East-ROSAT) L3 31.5 1 0 0
(Lockman-North) L3 31.4 2 1 0
FLS L6 7.30 35.4 89 17 5 3
XMM-LSS-SWIRE L6 21.61 48.5 402 25 3 3
(UDS) L4 33.8 3 1 0
(VVDS) L4 33.7 8 2 0
HerS Stripe82 ∼L7 80.76 51.8 4519 132 58 53
HATLAS SDP N/A 19.28 34.1 690 18 18 12
Total 172.46 6710 354 134 82
Note. — (1)Thedetection limitsarebasedontheSPIRE250µmsourcecount histogramswherethecount dropsto50%ofthepeak
value. The covered areas are calculated based on the pixel counts in the 250µm maps. (2) The HerMES program has several “nested
fields”, which are the sub-fields of deeper observations embedded in the shallower but wider parent fields. In this table, the nested fields
are labeled by parenthesis and are placed after their parent fields. The numbers of matched quasars in these nested fields and the wider
regionsoutsidearequotedseparately.
a“All”liststhenumberofSDSSquasarsfallwithinthecoveredregion.
bThePEPdataareavailableinthesefields.
andspatialcoveragecombinations(“L1”to“L6”),using and extracted the sources using the 250µm positions as
both SPIRE and PACS. Its latest data release, “DR2”, the priors. For the PACS bands, aperture photometry
contains the SPIRE maps and the source catalogs in all wasperformedonthe100and160µmmapsattheSPIRE
three bands (Wang et al. 2013). The PACS data have 250µm positions.
not yet been released. In this work, we used the wide-
field data and excluded those galaxy cluster fields. 2.2.3. HerS
The HerMESDR2 includes two sets ofsourcecatalogs HerS observed∼ 80deg2 in the SDSS Stripe 82 region
based on different methods, one using the SUSSEXtrac- (Abazajian et al. 2009; Annis et al. 2014) using SPIRE,
tor(SXT)pointsourceextractorandtheotherusingthe reaching the nominal 5σ limit of 50.4mJy in 250µm.
iterativesourcedetectiontoolStarFinder(SF)combined Both the image maps and the source catalogs have been
withthe De-blendedSPIREphotometryalgorithm(DE- released (Viero et al. 2014). We adopted the band-
SPHOT). We adopted the band-merged version of the mergedcatalog,whichwasbasedontheSF250µmdetec-
latter (denoted as “xID250”inDR2), which has the DE- tions and the DESPHOT photometry (Roseboom et al.
SPHOT multi-band (250, 350 and 500µm) photometry 2010).
atthepositionsoftheSF250µmsources. The5σ detec-
tion limits in 250µm range from ∼ 30 to 48mJy in our 2.2.4. PEP
fields of interest.
PEPwasalsoaHerschel guaranteedtimekeyprogram,
whichusedPACStosurveysixwell-studiedextragalactic
2.2.2. H-ATLAS fieldsandalsoanumberofgalaxyclusters. Boththeim-
agemapsandthesourcecatalogshavebeenmadepublic
H-ATLASwasthelargestHerschel open-timekeypro-
gram. Itsurveyed550deg2usingbothSPIREandPACS. through Data Release 1 (DR1) of the team. These data
are not listed in Table 1. All the PEP fields are within
Currently, this program has released the image maps
the HerMES fields, however they only covered a small
and the source catalogs in the field observed during the
fraction.
Herschel ScienceDemonstrationPhase(H-ATLASSDP;
Whenever possible, we used their 100 and 160µm
Ibar et al. 2010; Pascale et al. 2011; Rigby et al. 2011),
measurements in the COSMOS, GOODS-North, EGS,
whichcovers∼19.3deg2 and has reachedthe 5σ limit of
Lockman Hole fields to supplement the HerMES SPIRE
33.4mJy in 250µm.
photometry to better constraint the FIR SEDs of the
The catalog that we adopted is the band-merged one
detected quasars. These measurements were taken
with both the SPIRE and the PACS photometry. For
from the SF “blind extraction” catalogs as described in
the SPIRE bands, the source extraction was done using
Magnelli et al. (2009).
theMulti-bandAlgorithmforsourceeXtraction(MADX;
Maddoxetal.inprep.),whichemployedalocalizedback-
2.3. Herschel-detected quasars
groundremovalandPSFfilteringprocedurestothemap
Herschel-detected SDSS Quasars 5
250µm SNR are indeed consistent with the expectation
1.50
SNR250>5 fromEquation(2), andthe vastmajorityofthe matches
c) 0.48 ∆ = 1σpos within 3′′ fall within the 1σpos curve. Furthermore, the
se ∆ = 2σpos separation shows a double-peak feature roughly divided
arc 0.36 at3′′, indicating thatthe matches beyondthis pointare
( likelyaffectedbyotherfactors,suchasthe sourceblend-
n
io 0.24 ing problem (see below).
t
a Wang et al. (2013) have addressed the source posi-
r
pa 0.12 tional accuracy in the HerMES DR2 catalogs through
Se end-to-end simulations. For the SF catalogs in 250µm,
0.0 they find that the real matches between the input and
00.0 50.2 10 0.415 0.260 0.81030510.0 the output have the positional offsets peak at around
SNR250 Counts 5′′. We note that this is derived using the real matches
atallSNRlevels. IfaSNR>5thresholdis applied,this
Figure 3. Distribution of the separation between the SDSS po- peak is likely to shift to a smaller value. In any case,
sitionsandthematched SPIRE250µmpositionswhenexpanding the matching radius of 3′′ that we adopted is a rather
the matching radius to 5′′. The left panel shows the separation
as a function of the SNR of the 250µm detection. The solid and conservative choice.
thedashedlinesarethetheoretical∼1σ and∼2σpositionalun- InanumberoffieldsthePACSdataarealsoavailable.
certainties (σpos). Thecirclesinyellowcolorarethesources with The H-ATLAS band-mergedcatalogin the SDP field al-
SNR250 >5, while the light blue ones represent the rest. Among ready includes the PACS photometry (see §2.2.2) and
the yellow circles, the open and the solid ones are those deemed
by our visual inspection to be “blended” (i.e., affected by source thus no further action was taken. The HerMES COS-
blending) and“clean” (freeof sourceblending), respectively. The MOS, GOODS-N, EGS, and Lockman Hole fields have
horizontalthickblacklinegoingthroughbothpanelsindicatesthe PACS data from the PEP program (see §2.2.4), and we
adoptedmatchingradiusof3′′.
matched the coordinates in the PEP 100 and 160µm
Our IR quasar sample was derived by matching the “blind” catalogs to both the SPIRE “xID250” catalogs
positions of the SDSS quasars to those of the Herschel and the SDSS quasars, againusing a matching radius of
sources in the band-merged, 250µm-based catalogs as 3′′.
described above.
2.3.2. Source blending
2.3.1. Matching radius Due to the large beam sizes of the instruments, Her-
The matching was performed using TOPCAT/STILTS4. schel images still suffer from a severe source blending
Weusedamatchingradiusof3′′,whichisjustifiedbelow. problem. To evaluate how this problem could impact
our sample, we visually inspected the Herschel and the
While the Herschel instruments have large beam sizes
5, the source centroids can still be determined to high SDSSimagesofallthematchesinthe testcaseshownin
Figure 3, i.e., using a larger matching radius of 5′′. We
accuracy. The positional uncertainty of a given source
searched for the signatures of possible source blending,
depends on its signal-to-noise ratio (SNR) (see, e.g.,
suchasthepresenceofclosecompanionsintheSDSSim-
Ivison et al. 2007), which follows
ages,andthe offsetofthesourcecentersamongtheHer-
0.6θ schel bands, etc. For this test, we only used the sources
∆α=∆δ = , (1)
SNR that have SNR > 5 in 250µm. The result is also shown
in Figure 3. As it turns out, enlarging the matching ra-
where ∆α and ∆δ are the nominal 1σ uncertainties of diusfrom3′′ to5′′ willinclude47moreobjects,however
RA and Dec, respectively, and θ is the beam size. The
only 16 of these 47 objects (34%) are “clean” cases. For
SPIRE 250µm beam size is θ = 18′′, which means that
comparison, 83% of the sources within the matching ra-
the1σ positionaluncertaintyofagiven250µmsourceis dius of 3′′ are “clean”. This lends further support to
15.27′′ our choice of the matching radius. The drawback of us-
σpos = (∆α)2+(∆δ)2 = . (2) ing this conservative value is that it could exclude some
SNR
p objects that have genuine Herschel detections but are
Amatchingradiusof3′′ thuscorrespondsto∼1σ un- blended with close neighbor(s). While it is possible to
certaintyforobjectswithSNR>5,or∼0.59σ forthose include such sources after using the methods described
with SNR > 36. To further validate our choice, we per- in Yan et al. (2014) to de-blend, we defer this improve-
formed a test using an enlarged matching radius of 5′′, ment to our future work.
and Figure 3 demonstrates the results. The separations In summary, our conservative choice of the matching
between the 250µm and the SDSS positions versus the radius has resulted in a sample that is free of significant
blending problem. This has also simplified the photom-
4 TOPCAThttp://www.starlink.ac.uk/topcat; etry in the SPIRE 350 and 500µm bands. While the
STILTShttp://www.starlink.ac.uk/stilts
source catalogs that we adopted from the survey teams
5 TheFWHMbeamsizesare18′′,25′′ and36′′ fortheSPIRE
250, 350 and 500µm, respectively, and 6–7′′ and 11–14′′ for the were all derived using PSF-fitting based on the 250µm
PACS100and160µm,respectively. detections,thephotometryin350and500µmwouldstill
6 We note that Smithetal. (2011) use a somewhat differ- be prone to large errors introduced by blending due to
ent relation between the astrometric uncertainty and the SNR, theirmuchlargerbeamsizes(25′′and36′′,respectively).
2σ′.p′4osfo=rS0N.6R55=θ/5SNinR2,5w0µhimch. Omueranressuthltasts1hoσwunnicnerFtiagiunrtye3wionudlidcabtee Ourrelativelycleansample allowsus to assumethat the
that such a matching radius would be slightly too stringent, and fluxesmeasuredinthesetworedderbandsaresolelycon-
thereforeweadheredtoourchoice. tributed by the source detected in 250µm.
6 Ma & Yan
We used the cmcirsed code8 of Casey (2012) to per-
80
form the MBB fitting, which allowed us to derive the
70 β =1.0
IRluminosity, the dust temperature,andthe dust mass.
60 β =1.5 This procedure was only carriedout for the objects that
K) 50 β =2.0 have photometry in all the three SPIRE bands (187 ob-
( 40 jectsintotal,amongwhich103areintheSNR5subsam-
k
a ple), because the MBB fit would become unconstrained
Tpe 30 with less bands.
20
The FIR emission due to MBB can be written as
10
0 S(λ)=N (1−e−(λλ0)β)(λc)3 , (3)
0 10 20 30 40 50 60 70 80 mbb ehc/(λkTmbb)−1
T (K)
mbb whereT isthecharacteristictemperatureoftheMBB,
mbb
N is the scaling factor that is related to the intrinsic
mbb
Figure 4. Relation between the temperature of the modified luminosity, β is the emissivity, and λ is the reference
blackbody spectrum Tmbb and the temperature inferred from the wavelength where the opacity is unity0. As most of our
Wein’sdisplacementlawTpeak,forthethreecaseswhentheemis-
sivityβis1.0,1.5and2.0,respectively. Thedashedlinerepresents quasars only have three SPIRE bands available, we had
theequalityifthesetwoquantitieswerethesame. to limit the degrees of freedom. We adopted the default
2.3.3. IR quasar sample summary
emissivity of β = 1.5, which is the value typically as-
Our Herschel-detected SDSS quasar sample, derived sumedforcolddust(Casey2012). Bydefault, cmcirsed
using a matching radius of 3′′ as described above (Ta- sets λ0 = 200µm. We adopted λ0 = 100µm, follow-
ble 1), contains 3547 objects in total. Most of our ing Draine (2006). While the exact choice of λ0 only
studies are based on a subsample of those, called the marginally affects the estimates of the total IR luminos-
“SNR5” sample, which only includes 134 quasars that ity and the dust mass, it will significantly impact the
have SNR ≥ 5 in 250µm and thus is the sample of the estimate of the dust temperature. We will further dis-
most robust Herschel detections. Based on this SNR5 cuss this effect in Appendix A.
sample,wefurtherimposedacutinthe250µmfluxden- We note that the above form is for generalopacity. In
sity, S250 ≥ 56.6mJy, and formed a “bright” subsample theopticalthincase,atλ≫λ0,theterm(1−e−(λλ0)β)re-
of 82 objects for various discussions below. This flux ducesto(λ0)β,whichisoftenadoptedinthesubmm/mm
density threshold was adopted based on the lowest flux λ
regime. Throughoutthiswork,weusedthegeneralopac-
density of the SNR5 objects in the shallowest HerMES
ity form as in Equation (3).
field L6-XMM-LSS-SWIRE. The cmcirsed code has the capability of superposing
3. DUSTEMISSIONMODELING a power-law component (PL) to the MBB spectrum to
accommodate the possible warm dust component whose
We inferred the dust emission properties of the IR
effectcouldbepresentinthemid-IRregime(typicallyat
quasarsbyfittingtheirFIRSEDs. Thethermalemission <50µm in restframe). Thirteen quasars (seven of them
over the full IR regime can be viewed as the collective
are in the SNR5 sample) have PACS data in addition to
result of all heated dust components of various temper- thethreebandSPIREdata,andweutilizedthiscapabil-
atures, and the FIR part is dominated by the coldest
itywhenfittingtheseobjects. Inthiscase,theMBB+PL
component. In this paper, we focus on this FIR part in
model then reads
the IR quasars and hence our conclusions are pertain-
itningcttotytpheesirocfomldo-dduelsst tcoomfiptotnheentFs.IRWSeEDuss,edontewobedinisg- S(λ)=Nbb(1eh−c/e(λ−k(Tλλm0b)bβ))−(λc1)3 +Nplλαe−(λλc)2, (4)
a single-temperature, modified blackbody (MBB) spec-
trum and the other being three different sets of star- where α is the PL slope, Npl is the normalization of the
burst templates. Fitting an MBB spectrum to the FIR PL part, and λc is the turnover wavelength (for detail,
SEDis alwaysvalid,regardlessofthe exactdust heating see Casey2012). Again,tolimitthedegreesoffreedom,
sources (i.e., due to photons from either star formation α had to be fixed, and we adopted the default value of
or AGN activity). However, it has the drawback that α=2.0 (Casey 2012).
the SED must be properly sampled in order to obtain Recalling that the FIR emission is dominated by the
well constrained results. Fitting starburst templates, on cold-dust component, we can obtain the total IR lumi-
the other hand, is only appropriate if the FIR emission nosity of this component as
is dominated by heating from star formation, and the 1000µm
motivation of using these models was to test if the FIR L(cd) ≡Lmbb ≡ Lmbb(λ)dλ (5)
emissionsareconsistentwithbeingcausedbystarforma- IR IR Z8µm
tion. These two types of SED fitting approach provide
independent check to each other, and we will show later by integrating the best-fit model Lmbb(λ) from 8 to
that they also lead to insights into the heating sources. 1000µm. For the 13 objects that have PACS data, we
also calculated their total IR luminosities as
3.1. SED fitting using MBB model 1000µm
L ≡Lmbb+pl ≡ Lmbb+pl(λ)dλ. (6)
7Wenotethatthereare18morematchedobjectsfromHerMES IR IR Z8µm
DR2catalogs discardedduetoSNR250<1 8 http://herschel.uci.edu/cmcasey/sedfitting.html
Herschel-detected SDSS Quasars 7
11.05
14
01.83
12
01.61
)
L⊙ X =MBB(+PL) X =SK07
(
XIR 10
L
g
o
L 01.44
13
01.22
11
X =CE01 X =DH02
01.00
00.0 1 02.2 3 04.4 0 01.6 2 03.8 4 15.0
Redshift
Figure 5. Derived IR luminosities (LX ) of the IR quasars in our sample, where “X” denotes one of the four models in use, namely,
IR
MBB(+PL), SK07, CE01, and DH02. The errors between the MBB results and those from the starburst templates are not directly
comparable because of the different approaches adopted in evaluating the errors. The colored symbols represent the SNR5 sample,
while the grey squares represent the rest. Among the SNR5 objects, the red diamonds indicate those that are in the bright subsample
(S250≥56.6mJy),whiletheyellowcirclesindicatethosethatarenot. ThedarkgreensolidcirclesintheMBBpanelaretheobjectswith
PACS data available and hence a power-law component was added to the MBB spectrum in the fitting. To illustrate the impact of the
surveylimit,thelimitsofLXIR correspondingtothefiducialfluxdensitylimitofS250=56.6mJyareshownasthedashedlinesinthethree
panelsforthestarbursttemplatefits(SK07,CE01andDH02). Ineachofthesecases,thelimitisderivedfromtheentirelibrarybyusing
thetemplatewiththelowestpossibleLXIR atagivenredshift. InthepanelfortheMBBfit,thelimitsaregivenusingthreedifferentTmbb
of15,25,and35K,shownasthedashed, thedot-dashedandthedottedlines,respectively.
140 Wealsocalculatedthe“peak”temperatureinferredfrom
Wien’s displacement law, which is given by
SK07
120
CE01 T =b/λ , (7)
100 peak peak
DH02
ts 80 MBB wherethecoefficientb=2.898×103µmK. Ascompared
oun 60 toTmbb, Tpeak is lesssensitiveto thespecifics ofthe dust
C emission models in use and therefore could be a better
40 proxy to the dust temperature when comparing results
derived based on different types of templates. For this
20
reason, while we mostly use T ≡ T in this paper,
dust mbb
0 we also use T in some occasions. The relation be-
0 2 4 6 8 10 peak
tween T and T is shown in Figure 4. In case of
χ2 peak mbb
β = 1.5, the relation can be roughly described by the
following broken linear function:
Figure 6. Distributionofχ2 ofthebest-fitmodelsforalltheIR
quasarsinoursample,usingthefourdifferentmethodsasdetailed 0.7653T +0.9529, T <35K
inthetext. T ≃ mbb mbb (8)
peak (cid:26)0.5485T +8.5153, T >35K.
mbb mbb
WeemphasizethatL(cd)(hererepresentedbytheMBB
fit)istheluminosityofItRhecold-dustcomponentoverthe 3.2. SED fitting using starburst templates
entire IR range(8–1000µm). While the bulk of its emis- We also fitted the FIR SEDs using three different li-
sionisintheFIRregime,thiscold-dustcomponentemits brariesofstarburstgalaxiesseparately,namely,thetheo-
beyond FIR and thus it is necessary to integrate over reticalmodelofSiebenmorgen & Kru¨gel(2007,hereafter
8–1000µm to capture its total IR luminosity. Note that SK07; 7220 templates), and the empirical templates of
this is not the total IR luminosity (LIR) of the galaxy Chary & Elbaz (2001, hereafter CE01; 105 templates)
that includes the contributions of all dust components andDale & Helou(2002,hereafterDH02;64templates).
over 8–1000µm. In other words, we have L(cd) <L . For a given quasar, the restframe templates were red-
The best-fit temperature from cmcirsedIwRas takeInRas shifted according to the quasar’s redshift and convolved
the temperature of the cold dust, i.e., T ≡ T . with the Herschel passband response curves, and then
dust mbb
8 Ma & Yan
1.0
SB=SK07 SB=CE01 SB=DH02
bbR 1
mI
L
g
o 0.8
L
− 0
BR
SI
L
og −1
L 0.6
−2
SB=SK07 SB=CE01 SB=DH02
bR 0.4
bI 1
mF
L
g
o
L
− 0
R 0.2
BI
SF
L
og −1
L
0−.02
00.0 1 2 03.2 4 0 01.4 2 3 04.6 0 1 02.8 3 4 15.0
Redshift
Figure 7. Comparisons of IR luminositiesderived usingthe MBB model and those based on the three sets ofstarburst templates. The
symbolsarethesameasinFigure5. TheupperpanelsshowthecomparisonstoLSB (i.e.,computedoverthefullIRrangeof8–1000µm),
IR
whilethebottompanelsshowthecomparisontoLSB (computedovertheFIRregimeof60–1000µm),where“SB”isoneofSK07,CE01
FIR
andDH02. See§4.1fordetails.
were compared to the observed IR quasars SEDs. The strong dependent of dust temperature (M ∝ L T−5;
d IR
best-fittemplatewaschosenbyminimizingtheχ-square: see Casey 2012), it should be used with caution. For
example, in our analysis in §4, we will only use those
χ2 = N fobs−fth 2 . (9) thAatphpalyviengTmabbn/o∆mTimnabbl≥ga3s.-to-dust ratio, the gas mass
(cid:18) σ (cid:19)
Xi=1 obs (Mgas) can also be obtained. We adopted the nom-
inal Milky Way gas-to-dust-mass ratio of 140 (e.g.
We note again that the starburst model fitting was also
Draine et al. 2007) for this work.
done for the objects that have photometry in only two
or even one SPIRE band. During the fitting, we did 4. RESULTSANDDISCUSSIONS
not re-scale the templates. By construction, these tem-
plates all have associated L values (over 8-1000µm), The major physical properties obtained in §3 are all
IR
andthevalueofthebest-fittemplatewasadoptedasthe given in the online table accompanying this paper. A
totalIRluminosityoftheobjectinquestion. Forclarity, summary of the information contained in this table is
we will refer to it as LSB, where “SB” can be replaced given in Appendix B (see Table B1). Some examples of
IR
by “SK07”, “CE01” or “DH02” when appropriate. The the SED fitting are also provided in Appendix B. Here
error of LSB was estimated by taking the difference be- wediscussindetailtheseresults,somepotentialselection
IR
tween the L value of the best-fit template and that of effects, and their implications.
IR
the one of the second smallest χ2. As compared to the
formal likelihood method, this simple approach has the 4.1. IR luminosity
advantage that it works consistently when the param- For clarity, the various flavors of IR luminosities dis-
eter space is discrete, and that it includes the possible cussed in §3 are summarized here:
systematic errors intrinsic to the template set.
WeshouldemphasizethatLSBthusderivedisthetotal • L : the general designation of the total IR lumi-
IR IR
IR luminosity of the galaxy. If using the starburst mod- nosity over 8–1000µm;
els is appropriate, we should have LSB > Lmbb, where
IR IR
Lmbb ≡L(cd). WhileL(cd)cannotbeseparatedfromLSB • L(cd): the contributionofthecold-dustcomponent
IR IR IR IR IR
in any of these starburst models, we will show that LSB to the total IR luminosity over 8–1000µm, which
IR is also referred to as the total IR luminosity of the
can help interpret the origin of L(cd).
IR cold-dust component;
3.3. Dust mass and gas mass • Lmbb: the MBB best-fit to the FIR SED (as repre-
IR
The MBB fit by cmcirsed also resulted in estimate sented by the SPIRE data points) integrated over
of dust mass (hereafter M ). As this quantity is a 8–1000µm, and by definition L(cd) ≡Lmbb;
dust IR IR
Herschel-detected SDSS Quasars 9
111..005
0≤z<1 1≤z<2 2≤z<3 z≥3
01.84
01.63
01.42
001..281
01.00
−0.302 −30 −28 −26 −0.224 −22 −30 −28 −0.246 −24 −22 −30 −0.268 −26 −24 −22 −0.380 −28 −26 −24 −1.202
Mi[z=2]
11.05
0.6 0≤z<1 1≤z<2 2≤z<3 z≥3
) 01.84
⊙
L
( 01.63
cd()LIR 01.42
g
o 01.21
L 0.4
01.00
07.0 8 9 100.2 11 8 90.4 10 11 8 0.69 10 11 0.88 9 10 11.10
LogMBH (M⊙)
15
1.0
0≤z<1.7 z≥1.7
14
00..82
0.6 13
0.4 12
0.2 11
00..00 10
00..00 43 4004..22 45 46 4700..44 00..6643 44 45 00..4886 47 11..00
LogL2–10keV (erg/s)
X
Figure 8. NocorrelationbetweenL(cd) andtheabsolutemagnitudesofquasars(upperpanels),theirblackholemasses(middlepanels),
IR
ortheX-rayluminosities(lowerpanels). Theresultsareshowninfourredshiftbinsaslabeled,andtheerrorbarsareomittedforclarity.
• Lmbb+pl: the MBB+PL best-fit to the FIR SED expect. This is more clearly shown in the upper panels
IR
(as represented by the SPIRE and the PACS data of Figure 7, where LSB are compared against Lmbb. On
IR IR
points) integrated over 8–1000µm, which is the average,Lmbb valuesarelowerby0.13,0.23and0.25dex
IR
measurement of LIR using the MBB+PL model; as compared to LSB from the SK07, the CE01 and the
IR
• LSB: the measurement of L using the star- DH02 models, respectively. As explained earlier, this
IR IR is due to the fact that the MBB model only includes
burst models (“SB”is one of“SK07”,“CE01”and
“DH02”, depending on the model set in use), and the cold-dust component (Lmbb ≡ L(cd)), whereas the
IR IR
effectively is the best-fit SB template integrated starburst models also include all other components of
over 8–1000µm. higher temperatures and thus give the total IR luminos-
ity (L ). This is also demonstrated by the 13 objects
IR
Forone ofour purposeslater,we alsodefine the corre-
that have PACS data (dark green points), for which we
sponding quantities inthe FIR insteadofoverthe entire
carriedoutthe MBB+PLfitandthusobtainedthetotal
IR range, i.e., by integrating over 60–1000µm only. We IR luminosity in the form of Lmbb+pl. As one can see,
designate these quantities with the subscript of “FIR”, IR
there are no offsets between Lmbb+pl and LSB.
e.g., L(cd), Lmbb, LSB , etc. IR IR
FIR FIR FIR To further demonstrate this point, the lower panel of
The derived LmIRbb and LSIRB (referred to as LXIR) val- Figure7showssimilarcomparisonsasintheupperpanel,
ues of our IR quasars are shown in Figure 5, and the butbetweenLmbb andLSB . The agreementsareexcel-
distributions ofthe best-fitχ2 areshownin Figure6, re- lent and no syFsItRematic oFffIsRets are found. This can be
spectively. WenotethatderivingLmbb (andhenceL(cd)) understood as follows. The emission from the hot-dust
IR IR
wasnotalwayspossiblebecausetheMBBfitrequiresthe components should be minimal in the FIR regime, and
FIRSEDbeingwellsampledbythe threeSPIREbands, henceintegratingthestarburstmodelsintheFIRshould
while obtaining LSB could always be done because of onlycapturethecontributionfromthecold-dustcompo-
IR
the nature of the method (see §3.1 & 3.2). We also note nent, i.e., LSB =L(cd) =Lmbb.
FIR FIR FIR
that the SNR5 subsample, as expected, has the smallest As we have been emphasizing, the MBB fit is inde-
errors in LX . In addition, the majority of the objects pendent of the heating source, while the SB fit being
IR
outside of the SNR5 sample still have χ2 ≤10 and thus valid hinges upon the heating source being star forma-
are also deemed as having reliable LX measurements. tion. Therefore,the agreementbetween Lmbb and LSB
IR FIR FIR
Regardless of the exact model set in use, the majority strongly suggests that the SB fits are valid and that the
of our IR quasars have obtained good fits and the de- FIRemissionsintheseIRquasarsaredue tostarforma-
rived IR luminosities also agree to the extend that we
10 Ma & Yan
tion. The corollary then is that Lmbb (≡ L(cd)) is due be positively correlated with the AGN activity, i.e., the
IR IR
to star formation, because it is the total luminosity of strongertheAGNis,thelargerL(cd) shouldbe. Figure8
the samecold-dustcomponentthatgivesriseto the FIR IR
showsL(cd) versusthequasarabsolutei-bandmagnitude
emission. In the rest of this paper, the quoted IR lumi- IR
nosity is Lmbb unless explicitly stated otherwise (mostly (normalized to z = 2, adapted from Shen et al. (2011)
IR and Paˆris et al. (2014) for DR7 and DR10, respectively,
in§4.5,4.6andAppendixD),andweuseLmIRbb andL(IcRd) and all are based on PSF magnitudes after the Galactic
interchangeably depending on the context. While Lmbb extinction correction) in four redshift bins. Apparently,
IR
could underestimate the true LIR (see the top panel of no such a correlation can be seen. In the lowest redshift
Figure 7) by a factor of 1.35 (as compared to LSK07) to bin,thedistributionoftheobjectsiscompletelychaotic.
IR
1.70-1.78 (as compared to LCE01 or LDH02), we prefer Inthe binsathigherandhigherredshifts,wetendtosee
IR IR
to be conservative due to the lack of observational con- only those objects that are more and more IR luminous,
straints in the mid-IR for our entire sample. This, how- whichissimplyduetotheselectioneffectinaflux-limited
ever, is not necessarily a drawback because the mid-IR survey. Even among these the most luminous ones, no
emission,unlike the FIRone,couldbe seriouslycontam- correlationamongL(cd) andM canbevouchedfor. Fur-
IR i
inated by the AGN contribution (see Appendix C). thermore,Figure 8 alsoshows L(cd) versus the backhole
OurIRquasarshaveLX valuesrangingfrom∼1010.5 IR
IR mass (MBH) for the quasars that have these estimates
to1013.8L⊙(afterdiscardingtwoobjectswhoseSEDsare (taken from Shen et al. (2011) for the DR7Q quasars).
barely constrained). Mostof them (&80%) areULIRGs Similarly,nocorrelationexists. Finally,thebottompan-
(LIR >1012L⊙),andsomeofthem(&15–23%)areeven els of Figure 8 show L(cd) versus the hard-band X-ray
HyLIRGs (LIR >1013L⊙). As Figure 5 indicates, there luminosityintherestframIRe2–10keV(L2–10keV)foralim-
is a trend of LX versus redshifts. Obviously, the lack of X
IR ited number of quasars that we can derive this quantity
IRquasarswithlowLX athighredshiftsiscausedbythe basedonthe dataavailablein the literature11. As there
IR
selection effect due to the survey limit. For illustration, areonly 40suchquasars,we split theminto two redshift
Figure 5 shows the LX selection limit corresponding to bins, 0<z ≤1.7 and z >1.7, respectively, so that each
IR
a 250µm flux density limit of S250 =56.6mJy (which is bin receives approximately the same number of objects
whatweadoptedtoselectthebrightsubsamplefromthe for statistics (19 and 21 objects, respectively). They all
SNR5 sample). Interestingly, there seems to be a deficit have L2–10keV >1043erg/s, which is well above the con-
of very luminous IR quasars at z < 1, which reflects a X
ventional X-ray AGN selection threshold of 1042erg/s,
genuine IR luminosity evolution that is broadly consis-
above which the X-ray luminosity is believed to be pre-
tent with the evolution of ULIRGs, i.e., there are more
dominantlydue to AGN. Therefore,L2–10keV is astrong
ULIRGs at z >1 than at lower redshifts. X
indicator of the AGN activity. Again, no correlationbe-
Our conclusion that the FIR emission of IR quasars
tween L(cd) and L2–10keV can be seen. This is also very
are powered by star forming activity in dust-rich en- IR X
consistent with the recent results of Symeonidis et al.
vironments has also been suggested by previous stud-
(2014) and Azadi et al. (2015) in the similar L2–10keV
ies at high redshifts (e.g., Wang et al. 2011b). If X
regime.
this is indeed the case, using the standard L to
IR
Therefore, while we do not have direct evidence to as-
SFR conversion of Kennicutt (1998), i.e., SFR =
IR
1.0×10−10LIR/L⊙ for a Chabrier initial mass function sert that AGN has no contribution to L(IcRd), we do have
(IMF)9, the SFR of the HyLIRGs in our sample would evidence (albeit still indirect) against that AGN con-
be ∼1.0–6.3×103M⊙yr−1 10. tribution can be dominant. This further strengths our
There might be concerns whether such extreme SFRs conclusion of L(cd) being due to star formation based
IR
are physical. The SFR could indeed be overestimated on the earlier argument of Lmbb = LSB . As an addi-
FIR FIR
in two ways. First, one could argue that AGN heating tionalcheckofconsistencyinourconclusion,wehavealso
is still an important contributor to L(cd) and hence the tested the AGN/starburst decomposition approach, us-
IR
SFR cannot be calculated without subtracting this con- ing the method of Mullaney et al. (2011) on the objects
tribution. While there is no viable model quantitatively that have data in the PACS and/or the Spitzer MIPS
showingthat this couldbe the case (in fact, allavailable bands. The results are detailed in Appendix C. Note
models assume the opposite), we cannot yet assert that thatalltheAGN/starburstdecompositionschemesavail-
this is impossible. Our argument of Lmbb = LSB pre- able in the literature to date (including Mullaney et al.
FIR FIR (2011)) assume that the AGN contribution drops off in
sented earlier is only a necessary condition that L(cd) is
IR theFIR,andhencethedecompositionisnotentirelyap-
due to the heating from star formation but not a suffi-
propriate in asserting whether AGN contribute signifi-
cient one, and therefore we cannot rule out such a pos-
cantly to L . Nevertheless, our result shows that the
FIR
sibility based on this argument alone. However, we can
starburst-contributed IR luminosities as derived in the
demonstratethatAGNheatingisveryunlikelydominant
(cd) (cd)
inLIR . Ifitisdominant,itisexpectedthatLIR should 11 L2X–10keV were derived based on the data from the Chandra
SourceCatalogRelease1(Evans etal.2010)andthe3XMM-DR5
catalog (Rosenetal. 2015). Briefly, a power-law in the form of
Sa9lpeTthereIcMonFv,ewrshioicnhwwoausldadboeptaedfaicntoKreonfni∼cu1tt.7(1h9i9g8h)e.r if using a Iν ∝ ν−α was fit to the flux densities at different energy bands,
andthetotalenergyinrestframe2–10keVwascalculatedbyinte-
10 The most conservative SFR estimates would be using LmFIbRb gratingthebest-fitpower-lawoverthisenergyrange. Thebest-fit
(over 60 to 1000µm) instead of Lmbb (over 8 to 1000µm), which α has a median of ∼ 0.7, which correspond to the photon index
IR
wouldreducetheSFRvalues byafactorof∼1.5. Γ∼1.7.