AEGIS: Demographics of X-ray and Optically Selected AGNs

2.1 MB·English

by  Renbin Yan

Draftversion January24,2011
PreprinttypesetusingLATEXstyleemulateapjv.08/13/06
AEGIS: DEMOGRAPHICS OF X-RAY AND OPTICALLY SELECTED AGNS
Renbin Yan1, Luis C. Ho2, Jeffrey A. Newman3, Alison L. Coil4,5, Christopher N. A. Willmer6, Elise S. Laird7,
Antonis Georgakakis8, James Aird4, Pauline Barmby9, Kevin Bundy10,11, Michael C. Cooper11,12, Marc
Davis10,13, S. M. Faber14, Taotao Fang15, Roger L. Griffith16, Anton M. Koekemoer17, David C. Koo14, Kirpal
Nandra7, Shinae Q. Park18, Vicki L. Sarajedini19, Benjamin J. Weiner6, S. P. Willner18
1 DepartmentofAstronomyandAstrophysics,UniversityofToronto,50St. GeorgeStreet,Toronto, ONM5S3H4,Canada;
[email protected]
2 TheObservatoriesoftheCarnegieInstitutionforScience,813SantaBarbaraStreet, Pasadena,CA91101,USA
1 3 DepartmentofPhysicsandAstronomy,UniversityofPittsburgh,3941O’HaraStreet, Pittsburgh,PA15260,USA
1 4 DepartmentofPhysicsandCenter forAstrophysicsandSpaceSciences, UniversityofCalifornia,SanDiego,9500GilmanDr.,La
0 Jolla,CA92093,USA
2 5 AlfredP.SloanFoundationFellow
6 StewardObservatory,UniversityofArizona,933NorthCherryAvenue, Tucson,AZ85721, USA
n 7 AstrophysicsGroup,BlackettLaboratory, ImperialCollegeLondon,PrinceConsortRd,LondonSW72AZ,UK
a 8 National ObservatoryofAthens,V.PaulouandI.Metaxa, 11532,Greece
J 9 DepartmentofPhysicsandAstronomy,UniversityofWesternOntario,1151RichmondSt.,London,ONN6A3K7,Canada
10DepartmentofAstronomy,UniversityofCalifornia,Berkeley,601CampbellHall,Berkeley,CA94720,USA
0 11 HubbleFellow
2 12 Center forGalaxyEvolution,DepartmentofPhysicsandAstronomy,UniversityofCalifornia,Irvine,CA92697,USA
13 DepartmentofPhysics,UniversityofCalifornia,Berkeley,CA94720,USA
] 14 UCO/LickObservatory,DepartmentofAstronomyandAstrophysics,UniversityofCalifornia,SantaCruz,CA95064,USA
O 15 DepartmentofPhysicsandAstronomy,UniversityofCalifornia,Irvine,CA92697
16 JetPropulsionLaboratory,CaliforniaInstituteofTechnology, MS169-327, 4800OakGroveDr.,Pasadena,CA91109,USA
C
17 SpaceTelescopeScienceInstitute, 3700SanMartinDrive,Baltimore,MD21218,USA
. 18 HarvardSmithsonianCenter forAstrophysics,60GardenStreet,Cambridge,MA02138, USAand
h 19 DepartmentofAstronomy,UniversityofFlorida,Gainesville,FL32611,USA
p Draft version January 24, 2011
-
o
ABSTRACT
r
t WedevelopanewdiagnosticmethodtoclassifygalaxiesintoAGNhosts,star-forminggalaxies,andabsorption-
s
a dominatedgalaxiesbycombiningthe[OIII]/Hβratiowithrest-frameU−Bcolor. Thiscanbeusedtorobustly
[ select AGNs in galaxy samples at intermediate redshifts (z <1). We compare the result of this optical AGN

selection with X-ray selection using a sample of 3150 galaxies with 0.3 < z < 0.8 and I < 22, selected
2 AB
from the DEEP2 Galaxy Redshift Survey and the All-wavelengthExtended Groth Strip InternationalSurvey
v
4 (AEGIS).Amongthe146X-raysourcesinthissample,58%areclassifiedopticallyasemission-lineAGNs,the
9 rest as star-forming galaxies or absorption-dominated galaxies. The latter are also known as “X-ray bright,
4 optically normal galaxies” (XBONGs). Analysis of the relationship between optical emission lines and X-ray
3 propertiesshowsthatthecompletenessofopticalAGNselectionsuffersfromdependenceonthestarformation
. rateandthequalityofobservedspectra. ItalsoshowsthatXBONGsdonotappeartobeaphysicallydistinct
7
populationfromother X-raydetected, emission-line AGNs. On the otherhand, X-rayAGN selectionalsohas
0
strong bias. About 2/3 of all emission-line AGNs at L > 1044 ergs−1 in our sample are not detected in
0 bol
1 our 200 ks Chandra images, most likely due to moderate or heavy absorption by gas near the AGN. The 2–7
: keVdetectionrateofSeyfert2satz ∼0.6suggeststhattheir columndensitydistributionandCompton-thick
v
fraction are similar to that of local Seyferts. Multiple sample selection techniques are needed to obtain as
i
X complete a sample as possible.
Subject headings: galaxies: active—galaxies: nuclei—galaxies: fundamentalparameters—galaxies:
r
a Seyfert — galaxies: statistics
1. INTRODUCTION due to three reasons.
First, at higher redshifts, it is difficult to spatially
Recently,ithasbeenrealizedthatnearlyeverymassive
isolate the nuclear regions of galaxies for AGN detec-
galaxy bulge hosts a supermassive black hole (SMBH)
tion. Due to the smaller apparent sizes and the fainter
whose mass is tightly correlated with the stellar veloc-
flux levels of distant galaxies, spatially isolated nuclear
ity dispersion or the bulge mass (Magorrianet al. 1998;
spectroscopy studies such as that of Ho et al. (1995)
Ferrarese & Merritt 2000; Gebhardt et al. 2000). The
are not feasible at high redshifts. Using the integrated
tightnessofthesecorrelationssuggeststhatthegrowthof
light, the detectability of AGNs at high-z unavoidably
the SMBH is physically linked with the evolution of the
depends on host galaxy properties such as stellar mass
host galaxy. When SMBHs grow by accretion, they will
(e.g. Moran et al. 2002) and star formation rate (SFR).
appear observationally as active galactic nuclei (AGNs).
All AGN selection methods have such dependences, dif-
A complete census of AGNs, which includes both the
fering in the galaxy property involved and on the level
rare high-luminosity quasars and the more typical low-
of sensitivity. This issue has not been fully addressed in
luminosity AGNs, is essential for our understanding of
the literature.
SMBH-galaxy co-evolution. However, such a census is
Second, there is no single method that can select a
not yet available beyond the local universe, primarily
2 Yan et al.
complete sample of AGNs. In other words, no single multi-wavebanddatasetsenabledbythe All-wavelength
method identifies allthe AGNs found by other methods. Extended Groth Strip International Survey (AEGIS)
Everymethod hasits ownbias. Besides the differentde- Collaboration and high-quality DEEP2 optical/near-IR
pendences on host galaxy properties mentioned above, spectra to compare the two major AGN selection meth-
obscuration of the AGN light also causes different ob- ods at 0.3 < z < 0.8: optical emission-line diagnostics
jects to be missed by different techniques. For exam- and X-ray selection. In particular, we pay attention to
ple, the two methods commonly regarded as the most objectsthatareinconsistentlyclassifiedbythetwometh-
complete for AGN selection are optical emission-line se- ods. This paves the way to the construction of a more
lection and X-ray selection. Dust extinction throughout complete AGN sample.
the host galaxy can dramatically reduce the observed As a by-product, a comparison between optical
optical emission-line luminosity and bias optical selec- emission-line luminosities and X-ray luminosities of
tionagainstedge-ondiskgalaxies. X-rayselection,while AGNs can also help us evaluate the absorbing column
unaffected even by the worst levels of extinction in the density distribution among AGNs. In particular, this
hostgalaxy,is biasedagainstsourcesin which the X-ray helps to constrain the fraction of Compton-thick AGNs,
emission is heavily absorbed and/or Compton-scattered which are required to explain the spectrum of the hard
by dense gas clouds much closer to the central engine. X-raybackground(Gilli et al.2007). Manystudiesonlo-
When optical AGN selection is compared with X-ray calAGNsamples(e.g.,Bassani et al.1999;Risaliti et al.
AGN selection, one type of inconsistency attracts 1999) have shown that about 50% of Type 2 AGNs are
special attention: objects generally referred to as Compton-thick. However, the Compton-thick fraction
“X-ray bright, optically normal galaxies (XBONGs)” at higher redshift is more uncertain and is hotly de-
or “optically dull” X-ray galaxies (Elvis et al. 1981; bated(Daddi et al.2007;Fiore et al.2008;Donley et al.
Fiore et al. 2000; Mushotzky et al. 2000; Barger et al. 2008; Treister et al. 2009; Georgantopoulos et al. 2009;
2001; Comastri et al. 2002; Maiolino et al. 2003; Georgakakis et al. 2010; Park et al. 2010), partly due to
Brusa et al. 2003; Szokoly et al. 2004; Rigby et al. 2006; thelackofanemission-lineselectedAGNsamplebeyond
Cocchia et al.2007;Civano et al.2007;Caccianiga et al. thelocaluniverse. Therefore,wehopetoshedsomelight
2007; Trump et al. 2009b). These galaxies are bright in on this topic with our emission-line AGN sample.
the X-ray, so bright that they are undoubtably AGNs. Throughoutthepaper,weuseaflatΛCDMcosmology
However, their optical spectra either show no emission with Ω = 0.3. We adopt a Hubble constant of H =
m 0
linesatallorelseemissionlineshavinglineratiostypical 70 km s−1 Mpc−1 to compute luminosity distances. All
ofstar-forminggalaxies. The natureofthesesourceshas magnitudes are expressed in the AB system.
been hotly debated. Some have argued they could have
an intrinsically weak narrow-line region due to large 2. DATA
covering factor or radiatively inefficent accretion flow
2.1. X-ray imaging and optical identification
(Yuan & Narayan 2004). Others have suggested that
the missing emission lines are due to dilution by host TheAEGIS-Xsurvey(Laird et al.2009,L09hereafter)
galaxy light (Moran et al. 2002) or extinction in the has obtained 200 ks exposures over the entire Extended
host galaxy (Rigby et al. 2006). We will investigate the Groth Strip (EGS; Davis et al. 2007; see §2.2.1) using
nature of this population here. However, we distinguish Chandra ACIS-I. It covers an area of 0.67 deg2 and
those hosts showing star-forming-like emission-line reaches a limiting flux of 5.3 × 10−17erg cm−2 s−1 in
spectra from those showing no emission lines, as the the 0.5–2 keV band and 3.8×10−16erg cm−2 s−1 in the
explanations are different. 2–10 keVbandatthedeepestpoint. Itprovidesaunique
Lastly, the standard method used for spectroscopic combinationofdepthandarea,bridgingthegapbetween
identification of AGNs is currently observationally too the ultradeep pencil-beam surveys, such as the Chan-
expensive to use for large galaxy samples at z >0.4. In dra Deep Fields (CDFs) and the shallower, very large-
the local universe, the classification of AGNs and star- area surveys. Its areal coverage is nearly three times
forminggalaxiesis usuallyachievedbythe use ofoptical larger than the CDF-North and South combined. Com-
emissionline ratio diagnostics(e.g., Baldwin et al.1981; pared to the Chandra imaging in the Cosmic Evolution
Veilleux & Osterbrock 1987). The commonly-used dia- Survey field (Chandra-COSMOS; Elvis et al. 2009), our
gram involves two sets of line ratios: [NII] λ6583/Hα survey is slightly deeper but covers a slightly smaller
and [OIII] λ5007/Hβ (Figure 1). However, at z > 0.4, area. A detailed comparison of area and flux limits
[NII] and Hα are redshifted out of the optical win- among these state-of-the-art X-ray surveys is given by
dow into the near infrared. Other available diagnostics Elvis et al. (2009).
involve either two emission lines separated by a large The data reduction and point source catalogs of
wavelengthinterval, such as [OII] λ3727/Hβ(Rola et al. AEGIS-X are presented by L09. The reduction basi-
1997),whichissensitivetoextinctionandalsohasalim- cally followed the techniques described by Nandra et al.
ited observable redshift range, or involve the relatively (2005) with new calculations of the point spread func-
weak lines, such as [NI] λ5197,5200/Hβ ratio, which tion (PSF; see L09). L09 used a wavelet detection algo-
limit its applicability. These factors have hindered the rithmrunwithalowthresholdof10−4 toidentifycandi-
constructionofalarge,complete,narrow-lineAGNsam- dateX-raydetections. Countsandbackgroundestimates
ple beyond z ∼0.4. werethenextractedwithinanaperturecorrespondingto
This paperfirstestablishesa newopticalemission-line the 70% encircled energy fraction (EEF) for each can-
diagnostic method that avoids the use of [NII] and Hα didate source and used to calculate the probability that
lines so that we can select emission-line AGNs at red- the source is a spurious detection. All sources with a
shifts beyond z ∼ 0.4. We then make use of the rich false-positive probability (p) less than 4×10−6 in any
X-ray vs. optical selection of AGNs 3
of the bands (soft, hard, ultrahard, or full band) are in- 2.2.2. MMT/Hectospec follow-up
cludedinthefinalcatalog. TheX-raycountratewasesti-
Because the sampling fraction of the DEEP2 Galaxy
matedusingthe90%EEFaperture. UnlikeL09,wecon-
Redshift Survey is ∼50% in the EGS field, and the sur-
vertedthecountratetofluxusingaphotonindexofΓ=
veybegan before the X-rayobservationsweretaken, not
1.9, which is more appropriate for unabsorbed sources
allX-raysourceswithopticalcounterpartsweretargeted
(e.g., Nandra & Pounds 1994; Nandra et al. 1997). We
forspectroscopy. We thereforeobtainedadditionalspec-
also assumed this power-law spectrum to measure K-
tra using the Hectospec fiber spectrograph on MMT for
corrections from the observed X-ray flux to the rest-
as many X-ray sources with optical counterparts as pos-
frame bands. Both fluxes and luminosities in this paper
sible. Theobservationsanddatareductionaredescribed
are reported for the rest-frame bands: 0.5–2 keV for the
in detail by Coil et al. (2009). The spectra have a res-
softbandand2–10keVforthehardband. Thehardness
olution of 6 ˚A and cover a wavelength range of approx-
ratio is defined as HR=(H−S)/(H+S), where S and
imately 4500−−9000 ˚A. In total, we targeted optical
H arethe observedcountsinthe 0.5–2keVand2–7keV
counterparts for 498 X-ray sources with 288 yielding re-
bands, respectively. The HRs were computed using a
liable redshifts, including 23stars and265galaxies. The
Bayesian method following Park et al. (2006) and using
redshift success rate is a strong function of the optical
the BEHR package (version 07-11-2008). They are not
magnitude (Coil et al. 2009, Figure 2).
K-corrected.
Forsourcesdetected(p<4×10−6)insomebandsbut
not others, if the false-positive probability (p) in an un- Combining good redshifts from both surveys, out of
detectedbandislessthan0.01,westillmakeafluxmea- 895 unique X-ray sources with secure optical counter-
surement for that band. Throughout this paper, unless parts, we have redshifts for 493 objects, including 25
otherwisenoted,detectionreferstohavingp<4×10−6. stars and 468 galaxies. If we limit to the 426 sources
Incertaincases,wetreatsourcesdetectedinotherbands with IAB < 22 (the limit of our main sample used in
but having 4×10−6 < p < 0.01 in the 2–7 keV band this paper), we have 354 secure redshifts out of 388 X-
as “detections” to increase the size of the sample with ray sources targeted, correspondingto a redshift success
2–10 keV flux measurement. rate of 91% and an overallcompleteness rate of 83%.
The AEGIS X-ray source catalog presented by L09 The rest-frame U −B colors of galaxies in both spec-
contains 1325 sources. All but 8 are inside the troscopic samples were derived using the K-correction
boundaries of the DEEP2 CFH12K photometry cata- code described by Willmer et al. (2006). Stellar masses
log (Coil et al. 2004). Based on a maximum-likelihood were derived by Bundy et al. (2006) from fitting spec-
matching method (Sutherland & Saunders 1992), 895 tral energy distributions to Palomar/WIRC J- and Ks-
sourcesareuniquely matched1 to the DEEP2photomet- band photometry and CFH12K BRI photometry. For
ric catalog with a Likelihood Ratio (LR) greater than galaxies without Bundy et al. (2006) stellar mass esti-
0.5 (L09), corresponding to a 68.0% optical identifica- mates, we substituted the stellar masses computed from
tion rate. The estimated contamination rate is ∼6%. If absolute MB magnitude and rest-frame B − V color
limitedtoR <24.1,whichistheDEEP2surveylimit, using the prescription given by Bell et al. (2003) and
AB
calibrated to the Bundy et al. (2006) stellar mass scale
the optical identification rate is 53.5%.
withcolor-andredshift-dependentcorrections(Lin et al.
2007; Weiner et al. 2009).
2.2. Optical spectroscopy
2.2.1. DEEP2 2.3. Emission-line measurements
DEEP2is a galaxyredshiftsurvey using the DEIMOS Wemeasuredtheemissionlinefluxesineachspectrum
spectrographontheKeck-IItelescope(Davis et al.2003; afterfittingandsubtractingthestellarcontinuum. Asin
J.A.Newmanetal. inprep). Itcoveredfourwidelysep- Yan et al. (2006), each spectrum was fitted by a linear
arated fields totaling 3 deg2 on the sky down to a limit- combination of two templates after blocking the wave-
ing magnitude of R = 24.1. In the EGS, the survey lengths corresponding to emission lines. The templates
AB
obtained ∼17,775 spectra with 12,651 yielding reliable were constructed using the Bruzual & Charlot (2003)
redshifts. DEEP2 spectra typically cover approximately stellarpopulationsynthesis code. One template features
6500− −9200 ˚A with a resolution of R ∼ 5000. The a young stellar population observed 0.3 Gyr after a 0.1
high resolution enables good subtraction of atmospheric Gyr starburst with a constant SFR. The other template
emission lines and thus yields better sensitivity for line isanold7Gyrsimplestellarpopulation. Bothtemplates
detection in the target spectra. DEEP2 employed a slit are modeled assuming solar metallicity and a Salpeter
width of 1′′, which corresponds to a physical transverse initial mass function. After subtracting the stellar con-
scale of 7.1 kpc at z =0.7. tinua, we simply summed the flux around the emission
Out of the 895 optically identified X-ray sources, 375 lines and divided by the median continuum level of the
wereobservedaspartoftheDEEP2survey,yielding249 originalspectrum (before subtraction) in two bracketing
successful redshifts (66.4%) including 5 stars and 244 sidebands to get equivalent widths (EWs). The defini-
galaxies. tions for the central bands and sidebands are the same
as in Yan et al. (2006).
1 Two X-ray sources in the catalog are matched to the same Emission line luminosity estimates require accurate
DEEP2object,bothwithLR>0.5. Thesetwoarenotconsidered flux calibration and correction for slit losses. Both are
as valid matches and are removed; visual inspection of both X- difficult to measure to better than 10% accuracy. On
ray and optical images suggests that both X-ray sources are in
theotherhand,broadbandphotometryusuallyhasmuch
fact components of a single extended X-ray-emitting halo around
acompactgroupofgalaxies. smallererrors. Usingbroadbandphotometry,wecanap-
4 Yan et al.
plyK-correctionstogettherest-frametotalfluxinanar- Bayesian method (Helene 1984; Kraft et al. 1991, L09).
tificialfilter correspondingto our sidebandsin EWmea- However,unlike L09,who usedapower-lawpriorfor the
surements. Combining this with the EW measurements sourcefluxdistribution,weusedaconstant,non-negative
yieldstotallinefluxandluminosity,avoidingtheneedfor prior(Kraft et al.1991)withminimalassumptionsmade
spectrophotometricfluxcalibration(Weiner et al.2007). about the source flux distribution. This gives more con-
The accuracy of this method relies on the assumption servative upper limits than the power-law prior. As we
thatthe emissionline EWdoesnotvaryspatiallywithin do not require an X-ray detection, we do not suffer from
a galaxy. This assumption may not be accurate for Eddingtonbias. Ourupperlimitscorrespondtothe95%
AGNs, since the narrow-line regions subtend smaller confidence limit (2σ). In detail, the method applies the
scales than the stars in the host galaxies, leading to Bayes’theorem,
slightly larger[OIII]EWs inthe centralregionsof AGN
P(s|N,b)∝L(N|s,b)π(s), (1)
hosts. However, for galaxies whose angular sizes are
small compared to the seeing, the smearing of light by where L(N|s,b) is the conditional probability of observ-
the atmosphere will make the emission line EW more ing N counts given the expected source counts s and
uniform across the galaxy. X-ray sources are frequently expectedbackgroundcountsb;itfollowsthePoissondis-
massive galaxies. The median angular FWHM of our tribution,
main sample of X-ray sources (0.3 < z < 0.8) in the R-
band is 1′.′27, which is slightly larger than the slit width L(N|s,b)= (s+b)Ne−(s+b). (2)
and the seeing. Therefore, this procedure will only miti- N!
gatetheerrorcausedbytheslitlossbutnoteliminateit,
π(s) is the prior probability distribution of the expected
and the emission line luminosity may be overestimated
source counts; following Kraft et al. (1991), we adopt a
forAGNs. However,weexpecttheresultingbiasandun-
step function for π(s), which is a constant when s > 0
certaintyintheemission-lineluminositytobefarsmaller
and is zero otherwise. Bayes’ theorem yields the pos-
thanthoseintheX-rayluminosityorotheruncertainties
terior distribution function, P(s|N,b), for the expected
involved in the analysis, such as X-ray variability and
sourcecountsgiventheobservedcountsandtheexpected
unknown extinction corrections.
background.
We used the DEEP2 CFH12K photometry catalog
Here we assume that the error in the estimated back-
(Coil et al. 2004) and Blanton & Roweis’s (2007) k-
ground is small, thus b is known. For each source, the
correctcodev4 1 4toderivetherest-frameabsolutemag-
totalcountswereestimatedinanellipticalaperturethat
nitudesforthesidebandsaroundHβand[OIII]. Wethen
contains 70% of the energy in the simulated PSF. The
converted the magnitudes to fluxes and multiplied them
aperture was obtained using the PSF look-up table pro-
bytheEWstocomputethetotallinefluxesandluminosi-
vided by L09. The mean background counts for each
ties. The uncertainties for the emission-line EWs were
sourcewerederivedthe samewayasbyL09butwerees-
scaled up according to the differences from repeated ob-
timatedfortheaperturewithEEF(enclosedenergyfrac-
servations, following Balogh et al. (1999) and Yan et al.
tion)of70%insteadofthe90%EEFaperture. Giventhe
(2009). The EW uncertainties are propagated into the
observedcountsandtheexpectedbackgroundcounts,we
luminosity uncertainties, along with uncertainties from
estimated the upper limit of the source counts using the
photometry and K-corrections. Throughout the paper,
formulae given in Kraft et al. (1991). The count limit
we call an emission line detected if the EW is at least
was then converted to a flux limit and K-corrected by
twice as large as its uncertainty. For non-detections, the
assuming a power-law photon index of Γ=1.9.
luminosity upper limits are taken to be twice the uncer-
tainties, i.e., 2σ. 3. A NEW EMISSION LINE RATIO DIAGNOSTIC
2.4. X-ray upper limits for optical sources 3.1. The traditional method
Besides studying the X-ray selected sample, we have The classification of AGN and star-forming galaxies
alsoinvestigatedtheX-raypropertiesofthe opticallyse- is usually achieved by the use of optical emission line
lected sample. Therefore, we estimated the X-ray flux ratio diagnostics (Baldwin et al. 1981, hereafter BPT;
upper limits for sources not detected in X-ray. When Veilleux & Osterbrock 1987). One most commonly-used
extracting the X-ray counts at positions of the optical diagraminvolvestwosetsofline ratios: [NII] λ6583/Hα
sources, we found that many of them have significant and [OIII] λ5007/Hβ (Figure 1). In such a diagram,
countsabovethe background. Inmanycases,the proba- a.k.a. the BPT diagram, the star-forming galaxies pop-
bility that the counts arise from background fluctuation ulate a sequence from the upper left to the lower center
is less than 0.1%. The number of such cases is much whichisthe consequenceofacorrelationbetweenmetal-
higher than the expected number of false-positive cases, licity and ionization parameter. With increasing metal-
suggesting that many of them could be truly associated licity,the ionizationparameterandhence [OIII]/Hβ de-
with the optical sources. It is very tempting to include crease. Galaxiestothe rightandupper rightofthe star-
them as X-ray detections to increase the sample size. formingsequence are AGN hosts,whichinclude Seyferts
However, because the X-ray PSF is usually larger than ([OIII]/Hβ > 3) and low-ionization nuclear emission-
thePSFintheopticalimages,theX-rayfluxcouldcome line regions (LINERs, [OIII]/Hβ < 3). AGNs produce
from other untargetted galaxies nearby or even galax- higher[OIII]/Hβ and[NII]/Hαratiosthanstar-forming
ies beyond the optical photometry detection threshold. galaxiesbecause oftwo reasons. First, AGNs havemuch
Therefore,weonlyquoteX-rayfluxupperlimitsforthese higher ionization parameters so that they are capable
sources although the X-ray flux could be significant. of doubly ionizing more oxygen atoms. Secondly, their
We estimate the X-ray flux upper limits using the photonsaremoreenergeticthanthosefrommassivestars
X-ray vs. optical selection of AGNs 5
and can generate a more extended partially ionized re-
gion, where [NII] is produced. The exact line ratios de-
pend on the elementalabundances andthe strengthand
shape of the ionizing spectra. For a more detailed dis-
cussion,seeStasin´ska et al.(2006). Nonetheless,theline
ratio diagnostics provide a powerful way to distinguish
AGN-dominatedgalaxiesfromstar-formation-dominated
galaxies.
Two empirical demarcations are commonly used for
definingAGNsandillustratedinFigure1. Kewley et al.
(2001) provided a demarcation based on extreme star-
burst models. It is a fairly conservative limit for defin-
ing AGNs. Kauffmann et al. (2003) proposeda more in-
clusive demarcation. Galaxies that lie between the two
demarcations are often referred to as composite galax-
ies that have both AGN and star formation. Since
SDSS fiber apertures include both the nucleus and the
host galaxy, a lot of galaxies in this region are prob-
ably composites. However, this name is misleading in
two ways. First, galaxies outside this intermediate re-
gion could also be composite galaxies. Second, some
galaxies inside this region do not have to be compos-
ites, and there are evidences against the composite as-
Fig. 1.— Most commonly used line ratio diagnostic diagram
sumption. Using HST/STIS observations, Shields et al.
(BPT)forasampleoftheSloanDigitalSkySurvey(SDSS)galax-
(2007)showedthatinmanysuchobjectsidentifiedinthe ies with 0.05 < z < 0.1, r < 19.77, and all four emission lines
Palomar survey (Ho et al. 1995), the line ratios do not detected at >2σ level. The two curves indicate the AGN demar-
become more AGN-like on smallerapertures (10–20pc), cations of Kauffmannetal. (2003) (solid line) and Kewleyetal.
(2001)(dashedline). Weusethesedemarcationstodividegalaxies
contrary to the expectation for composites that smaller
intothreegroups: AGNhosts (upperright),star-forminggalaxies
apertures will include less star formation contribution. (lowerleft),andtransitionregion(middle). Theirdistributionsin
Therefore, we choose to refer to the region in between ournewdiagnosticdiagramareshowninFig.2.
the two demarcations as the “Transition Region”. For
theregionsaboveandbelow,werefertothemas“AGN”
(2006). As discussed above, we separate this sample
and “star-forming”, respectively. Note, our definition of
into three classes: star-forming galaxies, transition re-
“Transition Region” is different from that of the “Tran-
gion galaxies,and AGNs. Figure 2 replaces the horizon-
sition Objects” as defined by Ho et al. (1997) based on
tal axis with the rest-frame U − B color.2 The AGN
a set of line ratios including [OI]/Hα and using nuclear
hosts (panel c) are still in the upper right portion of
spectra.
the diagram, separated from the star-forming galaxies.
3.2. Our new diagnostic method Thetransitionregiongalaxiesoverlapprimarilywiththe
star-forming galaxies in U −B color. If we limit atten-
As mentioned in §1, it is observationally very expen-
tiontopure AGNs,the U−B colorprovidesaneffective
sive to apply the traditionalAGN diagnostics at z >0.4 alternative to the [NII]/Hα ratio for selecting a sample.
due to the inaccessibility of [NII] and Hα in the visible
In choosing an empirical demarcation in the new di-
window. Weiner et al.(2007) proposeda “pseudo-BPT”
agnostic diagram, our preference is to limit contamina-
diagramusingrest-frameH-bandmagnitude,M ,tore-
H tion to the AGN sample. This new method certainly
placethe[NII]/Hαratio. Forstar-forminggalaxies,M ,
H cannot select all galaxies containing AGNs down to the
a proxy for stellar mass, correlates with the metallicity-
demarcation of Kauffmann et al. (2003) without being
indicating [NII]/Hα ratio. Thus, this method can dis-
heavily contaminated by star-forming galaxies. We thus
tinguish Seyferts in relatively massive hosts from low
place the line just above the area populated by pure
mass,lowmetallicitystar-forminggalaxies. However,be-
star-forminggalaxies,whichstillallowsustoretainmost
causethecorrelationbetweenstellarmassand[NII]/Hα
AGNs classified by the Kewley et al. (2001) limit. The
breaks down for AGN host galaxies, the separation be-
demarcation we use is given by
tweenstar-forminggalaxiesand AGNs is not very clean.
Inspired by Weiner et al. (2007), we propose a more ef-
fectiveclassificationmethodemployingtheopticalU−B log([OIII]/Hβ)>max{1.4−1.2(U −B),−0.1}, (3)
color of galaxies in place of the [NII]/Hα ratio. Below
where max{a,b} denotes the greater value of a and b,
we first demonstrate this method using a galaxy sample
and U −B is the rest-frame color in the AB magnitude
fromSDSS; then we explainwhy it worksandwhy color
system. This is illustrated by the lines in Fig. 2. The
works better than stellar mass.
horizontal cut is a bit arbitrary: red galaxies with low
The galaxy sample used in Fig. 1 is selected from the [OIII]/Hβ ratiocouldbe eithertransitionregionobjects
Sloan Digital Sky Survey (SDSS, York et al. 2000) —
or very dusty star-forminggalaxies. There are relatively
DataReleaseSix(Adelman-McCarthy et al.2008)byre-
few of them. We leave the fine tuning for future work.
quiring 0.05 < z < 0.1, r < 19.77, and all four emission
lines detected at more than 2σ significance. The emis- 2Therest-frameU−BcolorforSDSSgalaxieswasderivedusing
sionlines are measuredin the same wayas by Yan et al. Blanton&Roweis(2007)’sk-correctcodev4 1 4.
6 Yan et al.
Fig. 2.— Same sample of galaxies as in Fig. 1, but replacing [NII]/Hα ratio on the horizontal axis with the rest-frame U −B color.
Thethreepanelsarefor(a)star-forminggalaxies,(b)transitionregiongalaxies,and(c)AGNhosts,classifiedaccordingtotheirpositions
inFig.1. Thesolidlinesindicateourempiricaldemarcation forAGNselection,whichremovesnearlyallpurestar-forminggalaxies (only
0.66%ofallstar-forminggalaxiesinthesamplecrossthisline).
With this demarcation (Eqn. 3), we find 95.7% (7138 AGN and low-metallicity hot stars. Low-metallicity
outof7459)oftheAGNsselectedusingtheKewley et al. stars are only forming in very blue star-forming galax-
(2001)demarcationarestillclassifiedasAGNs usingthe ies. Therefore,whenlimitedtoredorgreengalaxies,the
newmethod. Ifweincludeallobjectsinthetransitionre- high [OIII]/Hβ ratios have to be due to an AGN.
gionas AGNhosts, the completeness ofthe new method Therefore, the rest-frame U −B color can be used to
drops to 54.3% (9757 out of 17969). About 1.9% (190 track AGN activity in a similar fashion as the BPT di-
out of 9947) of the new “AGNs” were classified as star- agram, because it correlates positively with the bulge
forminggalaxiesundertheoldmethod;theseweconsider mass and metallicity, and correlates negatively with the
contamination. star formation rate. Other colors or spectral index (e.g.,
D (4000)) could also be employed provided they satisfy
n
3.3. The principle and the bias these criteria.
Thenewclassificationmethodisbasedonthefactthat WhatkindofAGNswillourmethodmiss? Firstofall,
nearly all BPT-identified AGNs are found in red galax- this method is not intended to select broad-line AGNs
ies or those with intermediate colors between red and (Type 1) in which the broadband color is not domi-
blue(hereafter,greengalaxies),buthardlyanyarefound natedbythehostgalaxy. Amongnarrow-lineAGNs,one
in very blue galaxies. There are several reasons. First, might worry that this method will miss those Seyferts
blue galaxiesare less massiveand havesmaller bulge-to- with high L[O III]/MBH which are shown to live in blue
diskratiosthanredgalaxies. Smallerbulgeshostsmaller star-forminggalaxies(Kauffmann et al.2007). However,
black holes (Magorrian et al. 1998; McLure et al. 2006). since our demarcation is tilted, we will not miss the ma-
The bluer a galaxy is, the less massive its black hole is, jority of these Seyferts: their “blue” colors are redder
and at a fixed Eddington ratio, the less luminous the than our limit. We tested this using our SDSS sam-
AGN will be. Thus, a lower fraction of the AGNs in pledescribedin§3.2. We derivedL /M following
[O III] BH
blue galaxies will be found above the observational flux Kauffmann et al. (2007) and selected the 5% of AGNs
threshold than those in green or red galaxies. Second, (defined using the Kewley demarcation) with the high-
star formationin blue galaxiescouldoverwhelmweak or estvaluesofL /M ,77%ofthesearestillclassified
[O III] BH
moderateAGNssothatthecombinedlineratiosstillput asAGNsinournewdiagram. Thisfractionislowerthan
them in the star-formingsequence on the BPT diagram. thatforallAGNs since these galaxiesdofallonthe blue
Both of these effects tend to hide AGNs in blue galax- edge of the AGN sample. Nonetheless, our new method
ies. There may also be other physical effects that we do can recover the great majority of them.
not yet understand. Nonetheless, this observational fact LINERs will be missed by this method if they live in
allows us to use host galaxycolor to reproduce the BPT bluegalaxies,whichwillmakethemLINER-star-forming
selection of AGNs. composites. On the BPT diagram, they will belong to
On the other hand, nearly all red/green galaxies that the transition region. We will not be complete for this
have high [OIII]/Hβ ratios are found to be AGNs. This category. The fraction missed depends on where we put
is because high [OIII]/Hβ ratios require high ionization the demarcation and which method we regard as giving
parameters which are produced in two types of sources:
X-ray vs. optical selection of AGNs 7
the “correct” classification for each galaxy, if it can be
well defined. By lowering the demarcation to include
more transition region galaxies, we would have more
“contaminations” from the BPT star-forming sequence.
However, some of these “contaminations” could also be
AGN–SF composites. Regarding to SF–AGN compos-
ites, our method has a similar selection bias as the BPT
diagrams, but with different detailed dependences.
Locally, we do not find low-metallicity AGNs
(Stasin´ska et al. 2006). If there exist at higher-z, they
couldbemissedbybothourmethodsandthetraditional
BPT diagram.
There is also much evidence now suggesting that
many luminous AGNs havepreviouslyhad a starforma-
tion episode (Kauffmann et al. 2003; Jahnke et al. 2004;
Sa´nchez et al.2004; Silverman et al.2008), so they have
bluer host galaxies than their inactive counterparts.
Would we miss the AGNs in these galaxies? The an-
sweris“No”. Mostpost-starburstsinSDSSandDEEP2
have redder U −B colors than the median star-forming
galaxy (Yan et al. 2009), because the U band covers the
blue side of the Balmer break. We therefore do not ex-
pect to miss AGNs in most post-starbursts, at least up
to z ∼0.8. anFdigb.e3tw.—eenCo[rNreIlIa]t/iHonαsbaentdweUen−[NBII(]l/oHwαeranpdanMel)∗.(uTphpeerspaamnpelle)
The principal advantage of this method is that it
is from the SDSS with selection described in §3.4. The contours
requires fewer emission lines than the traditional one. denote equal density levels, 1/Vmax weighted so they reflect the
Hence, it can be applied to higher redshift galaxies and distribution of a volume-limited sample. The contours are loga-
rithmicallyspaced with each level inwardrepresenting a factor of
suffers less from incompleteness due to missing line de-
twoincrementinnumberdensity. Thehighlightedthickercontour
tections, especially in low signal-to-noise (S/N) spectra.
encloses88%(upper panel)or90%(lowerpanel)ofallgalaxiesin
avolume-limitedsample.
3.4. Comparison to other methods
One might expect that stellar masses would perform
equally well as a replacement for [NII]/Hα. However, [NII]abundanceor[NII]/Hαratio. Therefore,colorpro-
this is not the case. Figure 3 shows how mass compares vides a better alternative to [NII]/Hα than stellar mass
with rest-frame U − B color in their correlations with does.
[NII]/Hα. The sample used is selected from SDSS DR6 Lamareille et al. (2004; 2010) investigated the use of
with 0.02 < z < 0.1, M∗ > 109M⊙, and by requiring [OII] λ3727/Hβ EWratioasanalternativeto[NII]/Hα.
both [NII] and Hα are detected at more than 2σ signif- This enables the application to redshifts as high as our
icance. The stellar masses are derived as a by-product method, though it also requires that [OII] lines are cov-
inK-correction(Blanton & Roweis2007). Weimposeda ered in the spectra. This method also has troubles dif-
redshift-dependentstellarmasscutsothatatallredshifts ferentiating the transition region from the star-forming
the sample is complete to a certainstellar mass limit for galaxies. Under their method, 84.5% of all Seyferts and
all colors. With this cut, for each galaxy, we computed LINERs(above the Kewley et al.2001demarcation)are
themaximumvolumeoverwhichagalaxywiththatstel- classified as AGNs, smaller than our completeness of
larmasswouldbeincludedinthesample,Vmax. Figure3 95.7%. The contamination from the star-forming se-
plots the 1/Vmax-weighted distribution in [NII]/Hα ver- quence (below the Kauffmann et al. 2003 demarcation)
susM∗ and[NII]/HαversusU−B spaces. Thesereflect to the AGN categoryis 2.6%, largerthan our number of
the distributions ofallgalaxiesin a volume-limitedsam- 1.9%.
ple down to M∗ >109M⊙ that have a reliable [NII]/Hα
measurement. 3.5. Intermediate-z test
In Figure 3, the rest-frame U − B color shows a
The SDSS galaxy sample we used is at z < 0.1. We
much better correlation with [NII]/Hα ratio than stel-
have shown above that the new method works at this
lar mass does. The difference is especially dramatic at
redshift. Does it still work at higher redshifts?
log([NII]/Hα) > −0.2, where the trend for stellar mass
We can test our new AGN/star-forming classification
becomes horizontal. These high [NII]/Hα galaxies are
method between redshifts 0.2 and 0.4, where the tradi-
mostly AGN-host galaxies that are old and have very
tional line ratio diagnostics are still available within the
loworzeroSFRs(asindicatedbytheirredcolors). Both
opticalwindow. Weusedthespectroscopicdataobtained
[NII]/Hαandcolorincreasewithmetallicityinpassively
intheEGSbytheDEEP2surveyandwithHectospecto
evolving galaxies (see Groves et al. 2004 for the depen-
test the method. The results are shown in Fig. 4. The
dence of [NII]/Hα on metallicity for AGNs), and hence
traditionalmethodidentified40emissionlineAGNsthat
they remain correlated as metallicity changes. However, are above the Kewley et al. (2001) demarcation.3 The
althoughstellarmasscorrelateswithmetallicityforstar-
forminggalaxies,stellarmassstopsgrowingoncethestar 3 For four objects near the demarcation with arrows pointing
formation stops. Thus, it no longer correlates with the acrossit,thelimitsonlineratiosstronglysuggestthattheybelong
8 Yan et al.
new method identified 36 of them and missed 4, cor- 5. z >0.3.
responding to a 90% completeness. It also picked up 8
objectsfromthetransitionregionwithnocontamination In total, there are 3150 galaxies and 146 X-ray sources
fromstar-forminggalaxies. Itisalsoencouragingtonote in this sample.
thatasaboutmanyX-raysources(11outof12)areiden- In our analysis, we primarily focus on Type 2 AGNs
tified as AGNs in the new diagram as in the traditional sincetheircomovingnumberdensityismuchhigherthan
diagram (10 out of 12). thatofType1AGNs,andthelatterareusuallyidentified
Because galaxies are bluer at higher redshift (Blanton easily in both optical spectra and X-ray data. However,
2006), in principle, our demarcation should shift blue- wewilluseType1AGNsasareferencesample. Wevisu-
ward slightly. By comparing the color–magnitude dia- allyidentifiedType1AGNcandidatesamongthespectra
gram of the DEEP2 sample at z ∼ 0.9 with that of an fromDEEP2andthe MMT/Hectospecfollow-upsurvey.
SDSSgalaxysampleatz ∼0.1,wefoundthedivisionbe- We measured the FWHM of the broad lines (Hα, Hβ,
tween red and blue galaxiesshifts by 0.14 mag in U−B or MgII) and classified those with FWHM greater than
between these redshifts (Cooper et al. 2008; Yan et al. 1000 km s−1 as Type 1 AGNs. There are 21 Type 1
2009),whichisconsistentwiththepassiveevolutionpre- AGNs in our sample. All but one are detected in the
diction(van Dokkum & Franx2001). Thecorresponding X-ray. The oneundetectedobjectis notfarfromthe de-
shift between z =0.3 and z =0.1 is about 0.04. For the tection threshold in the hard band, with a false-positive
sample we will discuss later, which has 0.3 < z < 0.8 probability of 1.8×10−3.
andamedianredshiftofz ∼0.55,the shiftisabout0.08 In Figure 6, we present the new emission-line diagnos-
from z ∼ 0.1. These shifts are quite small and insignif- tic diagram for all non-Type-1 galaxies in our sample.
icant for our results. Considering our sample covers a Thedistributionissimilartothatat0.2<z <0.4(right
wideredshiftrange,forsimplicity,wedonotapplythese panel of Fig. 4). For galaxies with either [OIII] or Hβ
shifts. undetected, we place them attheir loweror upper limits
for[OIII]/Hβ. In89%ofsuchcases(726outof815),the
4. X-RAY SELECTION VS. OPTICAL SELECTION upperlimitsontheundetectedlinesaretightenoughthat
A commonly used method to identify X-ray AGNs is theydonotintroduceanyambiguityintheclassifications
to use a pure luminosity cut of L2−10keV >1042 erg s−1. ofthe objects. The remaining11%areclassifiedas“am-
This is a very conservative threshold which is based on biguous”,which represents3%ofour sample andcan be
the fact that no local star-forming galaxies have X-ray safely neglected. Figure 6 also plots those galaxies with
luminosityaboveit4. However,lowerluminositysources neither [OIII] nor Hβ detected at the bottom. Effec-
could also be bona fide AGNs, which are equally, if not tively, we classify all galaxiesinto three main categories:
more,interesting. The advantageofaselectionbasedon star-forming, AGNs, and quiescent. The X-ray detected
X-ray luminosity is its rough correspondence to a bolo- sources (excluding Type 1 AGNs) are found in all three
metric luminosity selection (Elvis et al. 1994). However, categories. We use solid symbols to denote sources with
this can be compromised by intrinsic absorption of the LX(2–10 keV) > 1042erg s−1 and open symbols to de-
X-ray luminosity. Below, we compare the X-ray selec- note fainter sources. The fainter sources either have a
tion to optical emissionline selection in the DEEP2 and detectedLX(2–10 keV)>1041erg s−1,or,inthecasesof
Hectospec sample. hard band undetection, have a minimum LX(2–10 keV)
We limited our sample to objects that have both extrapolated (assuming Γ ≤ 1.9) from the 0.5–2 keV
[OIII]andHβ coveredinthe spectra,whichcorresponds band greater than 1041erg s−1.
roughly to a redshift range of 0.3 < z < 0.8. The red- The differences betweenthe two selectionmethods are
shift range is approximate due to the slightly varying apparent in this figure. First, consider the optically se-
wavelength coverage of the DEEP2 spectra. We made a lected Type 2 AGNs: these are points above the demar-
magnitude cut at I < 22 so that our redshift success cation. Manyofthem(78%ofallopticallyselectedType
AB
rate is above 90% for both red and blue galaxies, and 2 AGNs) are not detected in X-rays. Some are detected
the spectra have a sufficient S/N for stellar continuum but are fainter than the commonly used 1042erg s−1
subtraction. As shown in Figure 5, few X-ray sources at threshold. The majority (51%) of non-Type-1 X-ray
0.3<z <0.8areexcludedbythiscut. Tosummarize,all sources with LX(2–10 keV) > 1042erg s−1 are also op-
sources in our sample have to satisfy all of the following tically classified as AGNs. However, 22% of them are
criteria: found in the star-forming part of the line ratio diagram
and 25% are found to have no detectable line emission.
1. be within the X-ray footprint; As mentioned in §1, these cases are often referred to as
2. I <22; XBONGs or optically-dull X-ray galaxies. Often, the
AB
termXBONGisusedtoreferbothtogalaxieswithnode-
3. have reliable redshifts from either the DEEP2 sur-
tectablelineemissionandtothosewithlineratiosoftyp-
vey or the Hectospec follow-up;
icalstar-forminggalaxies. Wesuggesttreatingthesetwo
4. have [OIII] and Hβ well covered in the spectra, cases separately as the galaxies are two distinct types;
not badly affectedby CCDgaps orverybrightsky in the remainder of this paper, we only use XBONG to
lines; refer to the class with no detectable line emission, and
thus our XBONGs are nearly all red-sequence galaxies.
to the category across the demarcation. Thus, we assigned them To understand the reasonfor the discrepancy between
thoseclassifications.
optical selection and X-ray selection, we investigate a
4 UsingthecalibrationbyRanallietal.(2003),onewouldneed
anSFRof200M⊙ yr−1 toproduceenoughX-rayluminosityfrom few classes of objects, grouped according to the differ-
non-AGNsourcestocrossthisthreshold. ences in X-ray and optical classifications. To simplify
X-ray vs. optical selection of AGNs 9
Fig. 4.— Left: standard line ratio diagnostic diagram for a sample of sources in the EGS with 0.2 < z < 0.4. Arrows indicate the
2σ upper and lower limits for galaxies in which one of the four lines is not significantly detected. The solid and dashed curves show the
demarcationusedbyKauffmannetal.(2003)andKewleyetal.(2001)toseparateAGNhostsfromstar-forminggalaxies. Weusethetwo
curves to classify all galaxies into three categories: red squares indicate AGN host galaxies; small gray crosses are star-forminggalaxies;
andbluetrianglesaregalaxies inbetween, whichisusuallyconsidered ascompositeobjects. Largedarkcrosses indicatesources detected
in the X-ray. Right: the same galaxies now plotted on the U −B vs. [OIII]/Hβ diagram. The AGN hosts are still at the upper right
portion,separatedfromstar-forminggalaxies. Thesolidlinesmarkourempiricalcuts,whichweredesignedonthebasisoflower-redshift,
SDSSdata.
things, we do not separate the bright and faint X-ray 4.1. Unambiguous AGNs
sourcesintotwoseparatecategories,sincetheluminosity
Most(83%)opticallyclassifiedAGNs thatarealsode-
threshold is somewhat arbitrary. We have three optical tected in X-rays have L (2–10 keV) > 1042 erg s−1,
X
classifications—AGN,SF,andquiescent—andtwoX-ray
which confirms their identity as AGNs. The correla-
classifications—detections and non-detections.
tionsbetweentheemissionlineandX-rayluminositiesof
First, we will consider the objects that are classified
this population establish prototype relations for AGNs.
as AGNs by the optical selection and are detected in X- Both[OIII]andhardX-rayaregoodindicatorsforAGN
ray. Forbrevity,wewillcallthem“unambiguousAGNs”. bolometric luminosity (Heckman et al. 2004 for [OIII];
Second, we will discuss X-ray detected sources that are Elvis et al. 1994 for X-ray). Because [OIII] originates
found in the star-forming branch of our diagnostic dia-
fromthenarrow-lineregion,whichisoutside theobscur-
gram. FollowingMoran et al.(1996)andLevenson et al.
ing dusty torus, it is usually regarded as an isotropic
(2001),wecallthem“X-ray-loudcompositegalaxies”,re-
luminosity indicator. While X-rays can penetrate dust
flecting their AGN-star-formingcomposite nature,as we
easily,they canbe absorbedbya highcolumndensity of
willshowbelow. Third,wewillinvestigateX-raysources
neutralgasinthetorus. AmongType1AGNs,forwhich
with no detectable line emission (XBONGs). Lastly, we
we have an unobstructed view of the accretion disk, X-
willdiscussopticallyselectedAGNswithnoX-raydetec- rayluminosity is foundto correlatewith [OIII]luminos-
tion, which we refer to as “optical-only AGNs”. We will
ity (Mulchaey et al. 1994; Heckman et al. 2005). There-
use these names for each class throughout the remain- fore, comparing X-ray with [OIII] can reveal the level
der of this paper. Their definitions are summarized in
of X-ray absorption (Maiolino et al. 1998; Bassani et al.
Table 1, along with the number of objects in each class.
1999).
Note, because of the preferential follow-up of the X-ray Figure 7 compares [OIII] emission with X-ray in both
sources, the numbers in this table should not be used
flux and luminosity. Type 1 AGNs have a slightly
to estimate the fraction of emission-line AGNs that are
detected in the X-ray. For that analysis, we limit the larger median log(LX(2–10 keV)/L[OIII]) ratio (1.83
dex) and a narrower distribution than our Type 2
sample to only sources targeted in the DEEP2 survey,
AGNs (median=1.42 dex, see Table 2). However, the
where no preference was given to X-ray sources.
As will be shown below, the union of X-ray selected LX(2–10 keV)/L[OIII] difference between Type 1 and
Type 2 AGNs depends on how the sample is selected.
AGNsandopticallyselectedAGNsprovidesamuchmore
complete AGN sample. Table. 4 lists the IDs, coordi- Heckman et al. (2005) showed that in a hard-X-ray-
selectedsampleoflocalAGNs,Type1andType2AGNs
nates, redshifts, optical colors, emission line properties,
and the classifications of all the X-ray sources and the exhibit LX(2–10 keV)/L[OIII] ratios indistinguishable
optical-only AGNs in our sample, along with Type 1 fromeachother. However,inasampleselectedby[OIII]
AGNs. luminosity,manyType2’scanbepresentwithlowX-ray
10 Yan et al.
TABLE 1
Classification DefinitionsforNon-Type-1 Objects
EmissionLine(3129) X-rayDetected (126) X-rayUndetected (3003)
AGN(291) UnambiguousAGNs(64) Optical-onlyAGNs(227)
SF(1799) X-ray-loudcompositegalaxies(28)
Quiescent(950) XBONGs(32)
Ambiguous(89) Ambiguous(2)
Note. —Thenumbersinparentheses indicatethesamplesizeineachclass.
TABLE 2
Median logLX(2–10keV)/L[OIII] RatiosforUnambiguousAGNs
Class HX-selected [OIII]-selected
OurSample Heckmanetal.(2005) LaMassaetal.(2009)
Type1 1.83(0.29) 1.69(0.41)
Type2 1.42(0.50) 0.68(1.16) 0.46(1.01)
Combined 1.57(0.53) 1.22(1.01)
Note. —Numbersinparentheses arethe scaledmedianabsolute deviations SMAD of eachsample. Medianabsolute deviation (MAD)
isarobustestimatorofdistributionwidthforsmallsamples. WefollowBeersetal.(1990)todefineSMAD=MAD/0.6745,whichis1for
anormaldistributionwithstandarddeviation=1. Statistics fromHeckmanetal.(2005)andLaMassaetal.(2009)have beencomputed
fromtheirdatatables.
Fig. 5.— Magnitude and redshift distribution of X-raysources Fig. 6.—Newemission-linediagnosticdiagramforDEEP2galax-
(large points) that have optical counterparts and successful red- ies with 0.3<z <0.8. The solidline shows the proposed demar-
shifts from the DEEP2 survey and/or the Hectospec follow-up cation between star-forming galaxies (below and to the left) and
program,comparedwiththeoveralldistributionofgalaxies(small AGNs (above and to the right). Galaxies with either [OIII] or
gray points) with redshift successfully obtained in these two sur- Hβ undetected areplacedattheir2σ upper(bluepoints)orlower
veys. TheredpointsareX-raysourcesinoursample. Thedashed (red points) limits, respectively (arrows are omitted for clarity).
lines indicate the magnitude limit and the approximate redshift Innearlyallcases,theselineratiolimitsdonotaffecttheclassifi-
limits. Thelatterisapproximateduetotheslightlyvaryingwave- cation of objects. Quiescent galaxies without detectable emission
lengthcoveragesofDEEP2spectra. Objectsarerejectedfromthe linesarealsoshownatthebottom. X-raysources,whicharefound
sample if their spectra do not cover both [OIII] and Hβ, which inallthreecategories, aremarkedaslargesymbols. Solidsquares
couldalsobeduetoCCDgaps,badcolumns,etc. Wealsoexclude arebrightsourceswithL2−10keV >1042ergs−1;opensquaresare
allobjectsatz<0.3. faintsources withL2−10keV <1042ergs−1;andcrosses areX-ray
sources withuncertain luminosityclasses—they areusuallyunde-
tectedinthehardband. Onlyobjectsinthefirstcategorymaybe
definitelyclassifiedasAGNs,thoughmanyinthesecondclasswill
luminosity, presumably due to absorption. The Type 1 beAGNs.
AGNs then have a median ratio much larger than that
of Type 2 and have significantly less variation in the ra-
tio than Type 2 AGNs. The unambiguous AGNs are
effectively a hard-X-ray-selectedsample; there is a small
difference between Type 1 and Type 2 but not nearly as