The role of AGN in the colour transformation of galaxies at redshifts z~1
A. Georgakakis, K. Nandra, R. Yan, S. P. Willner, J. M. Lotz, C. M. Pierce, M. C. Cooper, E. S. Laird, D. C. Koo, P. Barmby, J. A. Newman, J. R. Primack, A. L. Coil
aa r X i v : . [ a s t r o - ph ] J a n Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 5 November 2018 (MN L A TEX style file v2.2)
The role of AGN in the colour transformation of galaxies at redshifts z ≈ A. Georgakakis ⋆ , K. Nandra , R. Yan , S. P. Willner , J. M. Lotz , , C. M. Pierce ,M. C. Cooper , , E. S. Laird , D. C. Koo , P. Barmby , J. A. Newman , J. R. Primack A. L. Coil Astrophysics Group, Blackett Laboratory, Imperial College, Prince Consort Rd , London SW7 2BZ, UK Department of Astronomy, University of California, Berkeley, CA 94720 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, MailStop 65, Cambridge, MA02138, USA Department of Physics, University of California, Santa Cruz, 1156 High Street, CA 95064, USA National Optical Astronomical Observatories, 950 N. Cherry Avenue, Tucson, AZ 85719, USA Steward Observatory, University of Arizona, 933 N. Cherry Ave., Tucson, AZ 85721-0065, USA UCO/Lick Observatory and Department of Astronomy & Astrophysics, University of California, Santa Cruz, CA 95064, USA Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA Department of Physics, University of California, Santa Cruz, 1156 High St., Santa Cruz, CA 95064, USA
ABSTRACT
We explore the role of AGN in establishing and/or maintaining the bimodal colour distri-bution of galaxies by quenching their star-formation and hence, causing their transition fromthe blue to the red cloud. Important tests for this scenario include (i) the X-ray propertiesof galaxies in the transition zone between the two clouds and (ii) the incidence of AGN inpost-starbursts, i.e. systems observed shortly after ( < Gyr) the termination of their star-formation. We perform these tests by combining deep
Chandra observations with multiwave-length data from the AEGIS survey. Stacking the X-ray photons at the positions of galaxies( . < z < . ) not individually detected at X-ray wavelengths suggests a population of ob-scured AGN among sources in the transition zone and in the red cloud. Their mean X-ray andmid-IR properties are consistent with moderately obscured low-luminosity AGN, Comptonthick sources or a mix of both. Morphologies show that major mergers are unlikely to drivethe evolution of this population but minor interactions may play a role. The incidence of ob-scured AGN in the red cloud (both direct detections and stacking results) suggests that BHaccretion outlives the termination of the star-formation. This is also supported by our findingthat post-starburst galaxies at z ≈ . and AGN are associated, in agreement with recent re-sults at low- z . A large fraction of post-starbursts and red cloud galaxies show evidence for atleast moderate levels of AGN obscuration. This implies that if AGN outflows cause the colourtransformation of galaxies, then some nuclear gas and dust clouds either remain unaffected orrelax to the central galaxy regions after quenching their star-formation. Key words:
Surveys – galaxies: active – galaxies: bulges – galaxies: evolution – galaxies:interactions – cosmology: observations
A major recent development in extragalactic astronomy is the dis-covery that most of the spheroids in the local Universe contain asuper-massive black hole (BH; e.g. Magorrian et al. 1998). The ⋆ Marie Curie fellow masses of these monsters are tightly correlated to the stellar veloc-ity dispersion of the host galaxy bulges (e.g. Ferrarese et al. 2000;Gebhardt et al. 2000), suggesting that the formation and evolutionof spheroids and the build-up of the super-massive BHs at theircentres are interconnected. This interplay may hold the key for un-derstanding some of the observed trends in galaxy evolution.One of the fundamental properties of galaxies is their bimodal c (cid:13) Georgakakis et al. distribution in rest-frame colour. This bimodality is observed to be-yond redshift z ≈ and is believed to hold important clues ongalaxy assembly (e.g. Strateva et al. 2001; Baldry et al. 2004; Bellet al. 2004; Weiner et al. 2005; Cirasuolo et al. 2007). Evolvedspheroidal galaxies are found in the “red-cloud” of the colour-magnitude diagram (CMD), star-forming systems define the “bluecloud”, while the region between these overdensities is sparselypopulated. This bimodal distribution is likely to be the combinedresult of evolution effects and external factors such as interactionsand mergers (e.g. Bell et al. 2004, 2006a, 2006b; Blanton 2006).In the simplest interpretation, the CMD is consistent with a pic-ture where the galaxy star-formation is truncated, by a mechanismthat remains to be identified, leading to the ageing of the stellarpopulation and the rapid transition from the blue to the red cloud.AGN are proposed to play a major part in this process by regulatingthe star-formation in galaxies causing their transition in colour ormaintaining them in the red cloud.Modelling work indeed suggests that the energy released byAGN is sufficient to either heat up or blow away the cold gas ofgalaxies (e.g. Silk & Rees 1998; Fabian 1999; King 2003), irre-versibly altering their evolution. Two main routes are proposed bywhich AGN feedback affects the host galaxy. The first one op-erates during the luminous and high accretion rate stage of BHgrowth, which is suggested to occur during major mergers (Hop-kins et al. 2005, 2006a, b). These catastrophic events also producenuclear starbursts that obscure the central engine for most of itsactive lifetime. When the AGN becomes sufficiently luminous itdrives outflows, which sweep away the nuclear gas and dust clouds,thereby quenching the nuclear star-formation. Following this stage,the AGN also declines as the BH runs out of accreting material andeventually switches off.In addition to the cold gas accretion mode above, simulationsalso include a second AGN feedback recipe, which is invoked tosuppress the accretion of hot gas onto the central galaxy of largedark matter haloes (e.g. Croton et al. 2006; Okamoto et al. 2007;Cattaneo et al. 2007). This mode operates in systems where the su-permassive BHs have already been built through mergers and are inplace at the central galaxy regions. In this picture, as galaxies enterthe massive dark matter haloes of groups or clusters, their cold gassupply is cut off leading to passive evolution and their migration tothe red cloud (Dekel & Birnboim 2006; Croton et al. 2006). How-ever, cooling flows in these dense environments could produce areservoir of cold gas in the central galaxy regions, the raw materialfor starburst and QSO activity. Low-level BH accretion is there-fore invoked to counterbalance cooling flows by producing low-luminosity AGN that heat the bulk of the cooling gas and therebysuppressing any subsequent star-formation in these galaxies.In the two AGN feedback prescriptions above, the outflow sce-nario during the luminous and high accretion rate phase of BHgrowth, assigns AGN a central role in establishing the bimodalcolour distribution of galaxies by directly causing their transitionfrom the blue to the red cloud. The distribution of X-ray selectedAGN in the CMD appears to be broadly consistent with this picture.AGN are preferentially associated with galaxies in the red cloud, atthe red limit of the blue cloud and in the transition zone in be-tween. These associations are suggestive of AGN-driven truncationof the star-formation (Nandra et al. 2007). However, more detailedstudy of the properties of AGN reveal a number of inconsistencieswith the merger scenario. Obscured AGN are predominantly hostedby galaxies in the red cloud (Nandra et al. 2007; Rovilos & Geor-gantopoulos 2007), with prominent bulges and little morphologicalevidence for ongoing major mergers (Grogin et al. 2003; Pierce et al. 2007). The red optical colours of these systems are mostlikely because of old stars and not dusty star-formation. Althoughthere are examples of deeply buried AGN in star-forming galaxies(e.g. Genzel et al. 1998; Cid Fernandes et al. 2001; Franceschini etal. 2003; Georgakakis et al. 2004; Alexander et al. 2005), the fre-quency of these systems among the obscured X-ray population at z ≈ is not yet clear. For example, Rovilos et al. (2007) argue thata non-negligible fraction of X-ray selected AGN at z ≈ have µ Jyradio continuum emission (1.4 GHz) that is consistent with star-formation in the host galaxy. These authors however, do not find astrong trend between X-ray obscuration and the incidence of faintradio emission, likely to be associated with starbursts.A plausible interpretation of the evidence above is that currentX-ray surveys are biased against the early phase of BH evolution.At this initial stage of their formation, AGN may be deeply buriedunder star-forming clouds and/or low luminosity because the BHmass is still small, although growing rapidly. If there is such a pop-ulation of obscured and/or intrinsically faint young AGN below thedetection threshold of current surveys, it may be possible to findthem using stacking analysis. Alternatively, AGN may not be re-sponsible for quenching the star-formation in galaxies and produc-ing the bimodality of the CMD. A key test of this scenario is theincidence of AGN in post-starburst galaxies, i.e. systems observedshortly after the termination of their star-formation ( < Gyr). Inthe local Universe ( z . . ), a number of studies point to an as-sociation between optically selected AGN and post-starbursts (e.g.Kauffmann et al. 2004; Goto 2006; Yan et al. 2006). At z ≈ , closeto the peak of the AGN density in the Universe (e.g. Barger et al.2005; Hasinger et al. 2005), there is still very limited informationon the link between X-ray selected AGN and post-starbursts.We address the issues above using data from the All-wavelength Extended Groth strip International Survey (AEGIS;Davis et al. 2007). X-ray stacking is employed to search for deeplyburied and/or low-luminosity AGN among optically selected galax-ies in the redshift interval . < z < . to explore the role of BHaccretion in establishing the bimodal colour distribution of thesesystems. We also study the X-ray properties of post-starbursts asidentified from Keck spectra at z ≈ . to investigate the linkbetween recent star-formation events and active BHs. We adopt H = 70 km s − Mpc − , Ω M = 0 . and Ω Λ = 0 . . Details about the AEGIS datasets can be found in Davis et al.(2007). In this paper we use (i) the deep
Chandra observations toexplore the X-ray properties of optical galaxies with spectroscopicredshifts from the DEEP2 survey, (ii) the
Spitzer
IRAC and MIPS µ m data to select subsamples based on mid-IR colour, and (iii)the high resolution HST /ACS imaging for morphological studies.The X-ray data are from the
Chandra survey of the ExtendedGroth Strip (EGS). The observations consist of 8 ACIS-I pointings,each with a total integration time of about 200 ks split in at least3 shorter exposures obtained at different epochs. The data reduc-tion, source detection and flux estimation are described in detailby Nandra et al. (2007, in preparation) and are based on methodspresented by Nandra et al. (2005). Briefly, standard reduction stepsare taken using the
CIAO version 3.2 data analysis software. Aftermerging the individual observations into a single event file, we con-structed images in four energy bands 0.5-7.0 keV (full), 0.5-2.0 keV(soft), 2.0-7.0 keV (hard) and 4.0-7.0 keV (ultra-hard). The countrates in the above energy intervals are converted to fluxes in the c (cid:13) , 000–000 he role of AGN in the colour transformation of galaxies at redshifts z ≈ Figure 2.
Mid-IR colour-colour plot. For clarity we use different panels. Left: red cloud galaxies in the redshift range . − . . Middle: blue cloud galaxies inthe same redshift interval. In both panels we only plot sources detected in all 4 IRAC/MIPS bands used to estimate colours. Right: expected tracks of differentgalaxy types in the range z = 0 − using observed mid-IR SEDs from the SINGS described by Dale et al. (2007). NGC 1097 Sy1: long-dashed dotted;NGC 5408 Im: short-dashed dotted; Sc spiral galaxy NGC 628: dotted; Sab spiral NGC 4450: continuous; elliptical galaxy NGC 584: long dashed. The dotsmark the position of z = 0 and the crosses correspond to z = 0 . . Figure 1.
Rest-frame U − B colour against B-band absolute magnitude forDEEP2 galaxies (small blue dots) in the range . < z < . estimatedby Willmer et al. (2006). Post-starburst galaxies are shown with the (red)triangles. The continuous line is defined by Willmer et al. (2006) to separatethe red from the blue clouds. standard bands 0.5-10, 0.5-2, 2-10 and 5-10 keV, respectively. Thelimiting flux in each of these bands is estimated to be . × − , . × − , . × − , and . × − erg s − cm − , respec-tively. The X-ray catalogue comprises a total of 1318 sources over .
63 deg to a Poisson detection probability threshold of × − .At z ≈ the 2-10 keV limit above corresponds to an X-ray lumi-nosity of about × erg s − ( Γ = 1 . ), i.e. typical of Seyferts.For the optical identification we use the DEEP2 photometric cat-alogues and the Likelihood Ratio ( LR ) method (e.g. Brusa et al.2007). A total of 903 sources have counterparts to R AB = 25 mag( LR > . ), the approximate limit of the DEEP2 photometric sur-vey. The DEEP2 redshift survey uses the DEIMOS spectrograph(Faber et al. 2003) on the 10 m Keck-II telescope to obtain redshiftsfor galaxies to R AB = 24 . mag. The observational setup uses amoderately high resolution grating ( R ≈ ), which providesa velocity accuracy of
30 km s − and a wavelength coverage of6500–9100 ˚A. This spectral window allows the identification of thestrong [O II] doublet 3727 ˚A emission line to z < . . We useDEEP2 galaxies with redshift determinations secure at the > confidence level (quality flag Q > ; Davis et al. 2007).The Spitzer
IRAC and MIPS- µ m observations cover thecentral ×
120 arcmin subregion of the EGS (Barmby et al.2006). The integration time is about 2.7 hr and 1200 s per point-ing for the IRAC and MIPS observations, respectively. The σ flux density limit for point sources is 0.9, 5.8 and 83 µ Jy at 3.6,8.0 and 24 µ m , respectively. There are about 57 400, 13 600 and6 300 detections to the limits above. These sources are matched tothe DEEP2 photometric catalogue by finding the nearest neighbourwithin a search radius of 1.0 arcsec. HST images of the EGS were taken with the Advanced Cam-era for Surveys (ACS) in the V (F606W, 2260s) and I (F814W,2100s) filters over a . × . strip (Lotz et al. 2007).The 5 sigma limiting magnitudes for a point source are V F W =28 . (AB) and I F W = 27 . (AB). For extended sources theselimits are about 2 mag brighter. A total of 15 797 galaxies were de-tected to I F W = 25 mag. These were matched to the DEEP2photometric catalogue using an 1.0 arcsec matching radius. The main sample used in this paper consists of optically selectedgalaxies with DEEP2 spectroscopy and redshifts in the interval . < z < . . This redshift range is to minimise colour dependentbiases in the selection of galaxies introduced by the magnitude limitof the DEEP2 spectroscopic survey, R AB < . mag. Red cloudsources drop below the survey limit at lower redshifts than intrin-sically bluer galaxies (e.g. Gerke et al. 2007). This effect becomesincreasingly severe at z & as the R -band straddles the rest-frameUV. At z = 0 . the sample is complete to M B = − mag. The c (cid:13) , 000–000 Georgakakis et al. lower redshift cut, z = 0 . , is to avoid biases associated with thesmall volume sampled by the AEGIS below this limit.There are 4814 sources in the main galaxy sample of whichabout 70 per cent overlap with the Spitzer
IRAC and MIPS surveysof the AEGIS and 30 per cent have I -band morphological informa-tion from the HST/ACS observations. The morphological samplewas compiled using the criteria of Lotz et al. (2007) by selectinggalaxies brighter than I F W = 25 mag with average signal-to-noise ratio per pixel > > . arcsec. About65 per cent of the sources detected on the HST/ACS images ful-fil these criteria. Almost all of them (97 per cent) overlap with theIRAC/MIPS area.The CMD of the main galaxy sample is presented in Figure 1.The blue and red clouds are defined using the relation of Willmeret al. (2006). Galaxies with U − B colours ± . about this rela-tion define the valley between the two clouds. Willmer et al. (2006)find that the redshift evolution of the red galaxy population U − B colour is not strong in DEEP2. As a result a line with fixed nor-malisation is adequate for defining the blue and red clouds over theredshift interval of our sample. The next sections discuss the prop-erties of galaxies in narrow slices of the CMD. These are definedto run parallel to the Willmer et al. (2006) line separating the twoclouds and are parametrised by their distance in U − B colour, ∆C ,from that line.The post-starburst sub-sample was drawn from the maingalaxy sample following the method described by Yan et al. (2006,2007 in prep.). The continuum-subtracted DEEP2 optical spectrawere fit with a combination of an old (K-component) and a young(A-component) stellar population SED. Post-starburst candidatesare defined as systems with f A > . , where f A is the frac-tion of light contributed by the young stellar component around4500 ˚A in the linear decomposition of the spectrum. The equivalentwidth of H β , after the removal of the stellar absorption, was thenused to identify residual/ongoing star-formation activity within thesample. Yan et al. (2006) showed that, for high redshift galax-ies, this line is a more reliable star-formation indicator than the[OII] 3727 line, which is often associated with AGN activity. Post-starbursts (K+As) are defined as systems with little H β emission, EW(H β ) < f A − (total of 44). Because of the wavelengthcoverage of the DEEP2 spectroscopy, post-starbursts can be iden-tified only in the redshift slice . − . . In this interval, theDEEP2 spectral window includes the H β line as well as the rest-frame wavelength range . λ . ˚A, which is essentialfor the decomposition of the spectrum into an old and a young stel-lar components. Figure 1 shows the position of post-starbursts inthe CMD. Most of them are in the red cloud.The mid-IR SED of galaxies results from a combination of cir-rus radiation, stellar emission, dusty star-formation, and possiblyhot dust associated with an AGN. Samples selected at µ m aregenerally dominated by dust enshrouded galaxies, both AGN andstarbursts (e.g. Yan et al. 2004). In this paper, we will use the µ m emission to identify such systems and to study them separatelyfrom quiescent galaxies. Figure 2 presents the Spitzer
IRAC/MIPScolours of DEEP2/AEGIS sources. Also shown in this plot are thecolour tracks for different galaxy types using a selection of ob-served SEDs from the SINGS program (Dale et al. 2007). Undoubt-edly there are large intrinsic variations in the mid-IR properties ofgalaxies of a given type resulting in a range of mid-IR colours.Nevertheless, the general trend in Figure 2 is for quiescent systemsto occupy the lower part of the plot and star-formation or AGNactivity to produce redder mid-IR colours. In this paper we define“ µ m –bright” sources as those with log( f /f . ) > . . Bluer Table 1.
Fraction of X-ray detections in different subsamples
Sample N TOT N X f X ( % )(1) (2) (3) (4)red . < z < .
548 36 . ± . red . < z < .
490 20 . ± . red post-starburst 34 5 . ± . valley . < z < .
105 4 . ± . valley . < z < .
126 8 . ± . valley post-starburst 9 1 ± blue . < z < . . ± . blue . < z < . . ± . blue post-starburst 10 2 ± The columns are: (1): Sample definition. Note that there is overlap between the red or the blue cloudand the valley. The redshift interval . < z < . is comparable in terms of selection effectsto post-starburst galaxies; (2): N TOT is the total number of DEEP2 galaxies in the sample; (3): N X corresponds to the number of DEEP2 galaxies with X-ray counterparts; (4): f X is the fractionof X-ray detections in the sample, i.e. f X = N X /N TOT . The errors are estimated assumingPoisson statistics.
Figure 3.
Mean hardness ratio for galaxies in the blue/red clouds and thevalley. Within each of these 3 groups, open circles correspond to all sourcesin the group, squares are for µ m bright galaxies, and triangles represent µ m faint systems. The hardness ratio uncertainties are estimated assum-ing Poisson statistics. In the case of no detection in the hard band, the σ upper limit in the hardness ratio is plotted. The horizontal lines correspondto the expected hardness ratio of a power-law X-ray spectrum with spectralindex from top to bottom Γ = 1 . , 1.4 and 1.9, respectively. Despite theerror bars, there is evidence for hardening of the mean X-ray spectrum fromthe blue to the red cloud. µ m bright sources have harder spectra in eachgroup. mid-IR colours ( log( f /f . ) < . ) or no detection at µ m defines the “ µ m –faint” sample. The latter sample is likely to in-clude “ µ m –bright” sources that are below the detection limit ofthe µ m observations. As expected most of the µ m selectedsources lie above the log( f /f . ) > . cut in Figure 2. Quies-cent galaxies are nevertheless also present in that sample. c (cid:13) , 000–000 he role of AGN in the colour transformation of galaxies at redshifts z ≈ Figure 4.
Mean hardness ratio estimated by stacking optical galaxies withindifferent slices of the CMD. ∆ C is defined as the difference between thecolour of the galaxy and the line separating the blue from the red clouds(Willmer et al. 2006; see Fig. 1). The colour slices are parallel to this lineand their position in the CMD depends on M B . The vertical dotted linesdefine the valley between the red and the blue clouds. The horizontal dashedlines are the same as in Figure 3. The line of Γ = 1 . corresponds to themean hardness ratio of the X-ray detected AGN. µ m -bright and µ m -faint systems are shown with the squares and the triangles respectively. Thehorizontal error bar of each point corresponds to the width of the CMDslice within which galaxies are stacked. The hardness ratio uncertainties areestimated assuming Poisson statistics. In the case of no detection in the hardband, the σ upper limit in the hardness ratio is plotted (arrows pointingdownward). The mean X-ray properties of sources which are not individuallydetected to the limit of the X-ray survey are explored using stack-ing analysis (e.g. Nandra et al. 2002; Laird et al. 2005). We useda fixed radius aperture to extract and to sum the X-ray photons atthe positions of optically selected galaxies. Sources that are sep-arated from an X-ray detection by less than 1.5 times the local90 per cent Encircled Energy Fraction (EEF) radius are excludedfrom the stacking. This is to avoid contamination of the stacked sig-nal from photons associated with the Point Spread Function (PSF)wings of X-ray detections. Also, in order to exclude regions wherethe
Chandra
PSF is large and the sensitivity is low, we do not ex-tract photons from pointings where a particular source is locatedmore than 9 arcmin away from the centre of the detector. This off-axis angle cutoff is chosen to maximise the stacked signal, althoughour results are not particularly sensitive to this parameter. For theextraction radius, we experimented with apertures in the range 1–3 arcsec and selected 2 arcsec as the optimal radius that maximisesthe signal-to-noise ratio of the stacked X-ray photons.To assess the significance of the stacked signal, we estimatedthe local background of a particular source by averaging the X-ray photons within a 50 arcsec radius (100 pixels) and then scalingto the area of the extraction aperture. When determining the localbackground we clipped regions around X-ray detections using a ra-dius 1.5 times larger than the 90 per cent EEF. The significance(in sigma) of the stacked signal is estimated by ( T − B ) / √ B , Figure 5.
Hardness ratio against redshift. The curves are the expectedtracks for the ARP 220 dusty starburst (Ptak et al. 2003), the Comptonthick AGN NGC 1068 (Matt et al. 1999) and NGC 6240 (Ptak et al. 2003)and a simple absorbed power-law model with
Γ = 1 . and N H =2 × cm − . The short-dashed line corresponds to power-law X-rayspectrum with Γ = 1 . , typical of unabsorbed AGN. The filled (red) circlecorresponds to red µ m -bright galaxies with ∆ C > − . . The open(blue) circle is for blue µ m -bright galaxies with ∆ C < − . . where T and B are the total (source + background) and backgroundcounts respectively. The 2 arcsec radius includes only a fractionof the source photons at each position. We account for the re-maining flux when estimating count-rates and fluxes by applyinga mean aperture correction determined by averaging the exposure-time weighted PSF corrections for individual sources. Table 1 summarises the fraction of X-ray detections among galaxysamples selected at several positions in the CMD. The X-ray iden-tification rate is higher in the red cloud, in agreement with recentstudies on the host galaxy properties of X-ray selected AGN (Nan-dra et al. 2007).In this section, we search for evidence of obscured and/or low-luminosity systems below the X-ray detection threshold by stack-ing optically selected galaxies in different regions of the CMD. Ofparticular interest is the valley, where AGN are suggested to playan important role in galaxy evolution. The results for different sub-samples are summarised in Table 2 and are plotted in Figure 3.There is some evidence for a progressive hardening of the stackedsignal from the blue to the red cloud. This is further demonstratedin Figure 4 which plots the hardness ratio of galaxies in CMDslices which run parallel to the Willmer et al. (2006) line separatingthe blue from the red clouds. Although the errorbars of individualpoints are large, there is a systematic trend whereby the mean X-ray spectrum of µ m -bright sources becomes harder, reaching Γ ≈ . , for ∆ C > − . , i.e. for sources around the valley c (cid:13) , 000–000 Georgakakis et al.
Figure 6. µ m luminosity. The con-tinuous line is the L X − νL relation for star-forming galaxies adaptedfrom Ranalli et al. (2003). The dashed lines correspond to the 1 sigma rmsenvelope around this relation. The (red) triangle corresponds to the mean νL and L X for µ m -bright galaxies with ∆ C > − . . Similarly,the (red) filled circle is for µ m -bright galaxies with optical colours bluerthan ∆ C < − . . The uncertainties in the luminosities of these two pop-ulations are smaller than the size of the points and are not plotted for clarity.The open (black) circles are AEGIS X-ray AGN selected in the hard X-rayband (2-10 keV). Also shown is the local sample of low-luminosity AGN(open blue squares), which includes Compton thick candidates (crossedblue squares), from Terashima et al. (2002). The mid-IR luminosity of in-dividual AEGIS X-ray sources, νL µm , is estimated from the µ m fluxdensity by adopting the SED of NGC 1068 for the k-correction. The conclu-sions are not sensitive to this assumption. Also, for AEGIS X-ray sourcesthe L X (2 −
10 keV) is calculated from the 2-10 keV observed-frame fluxusing a typical intrinsic AGN spectrum of
Γ = 1 . (Nandra & Pounds1994). This effectively produces absorption-corrected fluxes for sourceswith column densities N H < cm − at z ≈ . For the Terashimaet al. (2002) low-luminosity AGN we use the obscuration corrected lumi-nosity listed by these authors, except in the case of Compton thick candi-dates, identified by the equivalent width of the FeK line, adopting the crite-rion EW >
900 eV . The µ m luminosity of the Terashima et al. (2002)sources is estimated using the IRAS µ m flux density. The mean µ m luminosity for the subsamples used in the stacking is estimated by taking theaverage of the νL µm of individual sources. The mean L X (2 −
10 keV) is estimated using the X-ray flux determined by the stacking analysis andthe mean redshift of the sources in each subsample. and in the red cloud. Bluer galaxies have hardness ratio upper lim-its ( σ ) consistent with softer mean X-ray spectra, Γ > . . Thecorresponding upper limits for the µ m -faint subsample do notprovide strong constraints, but also suggest relatively soft spectralproperties. In any case, there is no evidence that the mean hardnessratio of this population increases with ∆ C at the same level as for µ m -bright galaxies.In order to improve the statistics we split the µ m -bright sample at ∆ C = − . and stacked separately sourcesbluer/redder than this limit. The resulting mean hardness ratio isplotted as a function of redshift in Figure 5. The mean hardnessratio of the µ m -bright population with ∆ C > − . is harderthan the ARP 220 starburst template and in better agreement withthe obscured AGN models. In contrast, the hardness ratio of galax- ies bluer than ∆ C < − . is consistent with star-formation (i.e.X-ray binaries and hot gas), although we cannot exclude the possi-bility of low-luminosity unobscured AGN.The obscured AGN interpretation for the µm -bright popu-lation with ∆ C > − . is also supported by Figure 6, which plotsthe 2-10 keV X-ray luminosity against the µ m luminosity, incomparison with the relation between these two quantities for star-forming galaxies adapted from Ranalli et al. (2003). Sources with ∆ C > − . in this figure are X-ray luminous compared to thisrelation, suggesting that their stacked X-ray signal is dominated byAGN. The same conclusion applies if galaxies are split into finercolour bins, e.g. upper blue cloud ( − . < ∆ C < − . ), valley( − . < ∆ C < − . ) and red cloud ( ∆ C > +0 . ). Thesesubsamples have mean X-ray and µm luminosities similar tothe overall ∆ C > − . population and therefore are also X-rayluminous compared to the expectation from star-formation. Theseresults are in contrast to the mean X-ray luminosity of µm -bright galaxies with ∆ C < − . , which is consistent with thestar-formation relation. These suggests that X-ray binaries and hotgas dominate the X-ray emission of this population. An interesting trend in Table 1 is the higher fraction of X-raysources in post-starbursts. In the red cloud in particular, where themajority of post-starbursts are found, 15 per cent of these systemsare associated with X-ray sources. In contrast the X-ray identifica-tion rate is only 2 per cent for the overall optical galaxy populationand about 4 per cent for galaxies in the red cloud. In order to as-sess the significance of the excess, we resampled the optical galaxypopulation to construct subsamples with size equal to the numberof red-cloud post-starbursts and with similar M B and U − B dis-tributions. The fraction of X-ray identifications in each subsampleis registered and the experiment was repeated 10 000 times. TheX-ray detection rate of the random subsamples is lower than thatof post-starbursts in 98 per cent of the experiments. The excess ofX-ray sources in this population is therefore significant at the 98per cent level. Post-starbursts are also a non-negligible componentof the X-ray population. In the redshift interval . . z . . , ± % (5/24) of the X-ray sources in the red cloud have post-starburst optical spectra. For comparison the fraction of red cloudgalaxies that are post-starbursts is 7 per cent (34/490). These re-sults, although limited by small number statistics, tentatively sug-gest a link between AGN and the post-starburst stage of galaxy evo-lution at z ≈ . . The link is also supported by the mean stackedX-ray properties of post-starbursts not individually detected to thelimit of AEGIS Chandra survey. The post-starburst population ismarginally detected in both the soft and the hard spectral bands at asignificance level & σ (Table 2, column 7). Although the stackedsignal is not dominated by a single source just below the X-ray de-tection threshold, small number statistics are a concern. The resultsin Table 2, taken at face value, are consistent with Γ ≈ . . Thisis much harder than the X-ray spectrum of star-forming galaxies(Figure 5) or unobscured AGN ( Γ ≈ . ; Nandra & Pounds 1994)suggesting that absorption is suppressing the soft X-ray emissionfrom the central engine in these systems. In the previous sections, we found evidence for AGN activityamong µ m -bright galaxies with ∆ C > − . (which in-cludes post-starbursts). This cut includes the reddest part of the c (cid:13) , 000–000 he role of AGN in the colour transformation of galaxies at redshifts z ≈ Figure 7.
Gini against M diagrams. The regions of the parameter space occupied by different galaxy types are demarcated with the dashed lines. For claritysources at different parts of the CMD are plotted in different panels: upper left sources with ∆ C > +0 . , upper right − . < ∆ C < +0 . (valley),lower left − . < ∆ C < − . (upper blue cloud). In all these panels open triangles are µ m bright sources and the dots are for µ m faint galaxies. Inthe lower right panel we show the position of X-ray detected AGN on the Gini– M diagram. blue cloud, the valley, and the red cloud. The optical morphologyof these sources should be a powerful diagnostic of their nature.Are they interacting/merging systems? Do they show evidence forstar-formation? For classifying galaxies into different morpholog-ical types, we use the HST /ACS data to estimate the Gini coeffi-cient, which measures the clumpiness of a source, and the secondmoment of the brightest 20% pixels of the galaxy, M (Lotz, Pri-mack & Madau 2004), which measures the central concentration ofa galaxy. Different Hubble types are separated in the Gini– M di-agram and the morphological classification based on these two non-parametric estimators remains robust at high redshift. An additional parameter, the asymmetry, is also used to search for disturbancesthat might indicate recent/ongoing interactions (e.g. Abraham et al.1996; Conselice et al. 2000). This parameter measures the degreeto which the light of the galaxy is rotationally symmetric.Figure 7 plots the Gini– M parameters for galaxies with ∆ C > − . . While the µm -bright population with ∆ C > − . includes systems classified as ongoing mergers, these areabout 7 per cent of the population. The majority of the µm -bright sources in Figure 7, about 70 per cent, have spiral or irreg-ular morphology, Sb or later. Blue star-forming regions in thesegalaxies are indeed resolved by the HST . The µm emission of c (cid:13) , 000–000 Georgakakis et al.
Table 2.
Stacking resultsSample N < z > band
T B
S/N photon rate HR f X log L X ( σ ) ( − ) ( − ) ( erg / s )(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) red cloud ( ∆ C > . ) all 881 0.67 soft 741 389.9 17.9 . ± . − . ± .
09 1 . ± . . ± . . ± . µ m -bright 177 0.69 soft 169 80.8 9.8 . ± . . ± .
14 1 . ± . . ± . . ± . µ m -faint 513 0.66 soft 451 235.2 14.1 . ± . − . ± .
12 1 . ± . . ± . . ± . valley ( − . < ∆ C < +0 . ) all 188 0.68 soft 155 81.8 8.1 . ± . − . ± .
21 1 . ± . . ± . . ± . µ m -bright 89 0.70 soft 85 41.0 6.9 . ± . − . ± .
22 2 . ± . . ± . . ± . µ m -faint 61 0.68 soft 42 26.1 3.1 . ± . < . ± . < < < blue cloud ( ∆ C < . ) all 3331 0.66 soft 2009 1469.1 14.1 . ± . − . ± .
15 0 . ± . . ± . . ± . µ m -bright 572 0.68 soft 538 269.0 15.8 . ± . − . ± .
12 1 . ± . . ± . . ± . µ m -faint 1811 0.66 soft 1018 823.7 6.8 . ± . < − .
24 0 . ± . < < < ∆ C > − . all 1270 0.67 soft 1017 560.5 19.3 . ± . − . ± .
08 1 . ± . . ± . . ± . µ m -bright 332 0.69 soft 300 153.6 11.8 . ± . . ± .
12 1 . ± . . ± . . ± . µ m -faint 648 0.66 soft 531 294.4 13.8 . ± . − . ± .
13 1 . ± . . ± . . ± . ∆ C < − . all 2942 0.66 soft 1733 1298.5 12.1 . ± . < − .
49 0 . ± . < . < . < µ m -bright 417 0.67 soft 397 196.2 14.3 . ± . − . ± .
16 1 . ± . . ± . . ± . µ m -faint 1676 0.66 soft 938 764.6 6.3 . ± . < − .
21 0 . ± . < . < < < > +0 . < < . ± . . ± . . − . for all samples except for post-starbursts; (2): number of sources used for stackingafter excluding galaxies that lie close to or associated with X-ray detections; (3): mean redshift of the sample; (4): spectral band where X-ray photons arestacked: soft: 0.5–2 keV and hard: 2–7 keV; (5): Total counts (source + background) within the extraction radius; (6): Background counts; (7): significance ofthe detected signal estimated from the relation ( T − B ) / √ B and expressed in background standard deviations; (8): Photon count rate in units of − s − corrected for aperture effects and the Chandra response. In the case of non-detection the σ upper limit is listed; (9): Hardness ratio defined as (H-S)/(H+S)where H, S are the hard and soft band photon count rates respectively. The errors are estimated assuming Poisson statistics; (10): Rest-frame X-ray flux inunits of − erg s − cm − estimated in the 0.5-2 and 2-10 keV spectral bands adopting the mean power-law spectral index consistent with the HR. In thecase of non-detection the σ upper limit is listed; (11): Logarithm of rest-frame luminosity in units of erg s − . these systems is also likely associated with young stars. Moreover,visual inspection of the HST /ACS images further shows that many µm -bright sources in the E/S0/Sa region of the Gini- M dia-gram also have disks in addition to the dominant bulge, i.e. earlytype spirals. Similarly, visual inspection of the HST/ACS imagesof the X-ray detected AGN, also plotted in Figure 7, suggests thatabout half of these systems are associated with spirals. The mor- phological evidence above suggests that disks represent a substan-tial fraction of the µm -bright galaxies with ∆ C > − . andabout half of the X-ray detected AGN. This suggests that (i) somelevel of star-formation is likely taking place in the host galaxy and(ii) recent major mergers are unlikely to have played a central rolein the evolution of these sources.Although major mergers do not appear to be frequent among c (cid:13) , 000–000 he role of AGN in the colour transformation of galaxies at redshifts z ≈ µm -bright galaxies with ∆ C > − . tidal disruptions, or mi-nor mergers may play a role in their evolution. Figure 8 comparesthe distribution of the asymmetry parameter for the µ m -brightand the µ m -faint galaxies in the ∆ C > − . part of the CMD.The panels in this figure correspond to different Hubble types basedon the Gini- M classification of Figure 7. In the E/S0/Sa andSb/bc panels, the µ m -bright sample is offset to higher asym-metries compared to the µ m -faint galaxies. For later Hubbletypes, Sc/d/Irr, there is little difference between the two subsam-ples. A Kolmogorov-Smirnov tests shows that the likelihood ofthe observed differences if the µm -bright and µ m -faint sam-ples were drawn from the same parent population is < − , × − , and 0.15 for the E/S0/Sa, Sb/bc and Sc/d/Irr classes,respectively. For comparison, X-ray detected AGN, most of whichare in the E/S0/Sa part of the Gini- M diagram, are also offset tohigher asymmetries compared to early-type µ m -faint galaxies.This evidence suggests that minor gravitational encounters play arole in the evolution of both the µ m -bright systems with early-type optical morphology (E/S0/Sa and Sb/bc) and some of X-raydetected AGN associated with disks. Alternatively clumpy star-formation in the disk of galaxies can also produce the same effect,i.e. higher asymmetry parameter distribution. Differentiating be-tween the two interpretations is not easy, and it is likely that clumpystar-formation and minor galaxy interactions are linked (e.g. Bell etal. 2005).For late type galaxies (Sc/d/Irr) there is no statistically sig-nificant difference in the asymmetry parameter distribution be-tween µ m -bright/faint sources. The two populations however,have distinct M B distributions, suggesting differences in their stel-lar mass. This is shown in Figure 9, where late-type µm -brightgalaxies with ∆ C > − . have mean absolute B -band magni-tude M B ≈ − . mag. This is more luminous than the average M B ≈ − . mag for µm -faint sources of similar morpholog-ical type. The two populations also have similar rest-frame colourdistributions suggesting similar mass–to–light ratios, M/L. Adopt-ing log M / L = +0 . , consistent with the colours of the galaxiesin the sample (Bell & de Jong 2001), the above mean optical lumi-nosities translate to stellar masses of ≈ × and × M ⊙ for µ m -bright and faint sources respectively.A plausible interpretation of the results above is that AGNactivity requires a massive BH and some gas to fuel it (Kauff-mann et al. 2003). Massive galaxies often host large BHs, whilestar-formation is an indication of gas availability. The µm -bright galaxies with ∆ C > − . are typically luminous, M B . − mag, and therefore most likely massive. The morphologicalevidence above also indicates star-formation in a large fraction ofthese systems. This is contrary to µm -faint galaxies, which areeither quiescent and/or less massive than µm -bright sources.Galaxy encounters, not necessarily major mergers, may still be re-quired in the picture above to disturb the gas to the galaxy centreand also to trigger the formation of stars. The X-ray hardness ratio & − . estimated in sections 5.1 and 5.2for µ m -bright galaxies with ∆ C > − . and post-starburstscan be attributed to low-luminosity Compton thin AGN, Comp-ton thick sources (e.g. Figure 5), or radiatively inefficient accre-tion flows (e.g. Brand et al. 2005). Figure 6 shows that the average µ m -bright but X-ray undetected galaxy with ∆ C > − . hasX-ray luminosity about a factor 10 below the typical X-ray detected Figure 8.
Normalised distribution of the asymmetry parameter for µ m -bright (hatched red histogram) and µ m -faint galaxies (open blue his-togram) with ∆ C > − . . The different panels correspond to differentmorphological types defined using the Gini- M diagram shown in Figure7. X-ray detected AGN in the AEGIS are plotted in the top panel. Figure 9.
Normalised distribution of the absolute B -band magnitude for µ m -bright (hatched red histogram) and µ m faint galaxies (open bluehistogram) with late-type optical morphology (Sc/d/Irr) and ∆ C > − . . AGN. The X-ray undetected galaxies lie in the same region of theparameter space with moderately obscured low-luminosity AGNand Compton thick sources, suggesting that the observed stackedsignal may be associated with one or the other type of AGN ac-tivity, or possibly with a mix of both. If the mid-IR luminosity of µm -bright galaxies with ∆ C > − . is, on average, domi-nated by star-formation, and not by dust heated by the central en-gine, then the moderately obscured low luminosity AGN scenariois most probable. We note however, that star-formation in the host c (cid:13) , 000–000 Georgakakis et al. galaxy is also likely to contribute or even dominate the mid-IRemission of both the Terashima et al. comparison sample and theX-ray sources in the AEGIS survey. In the case of Compton thickactivity, the intrinsic AGN luminosity is expected to be 100–1000times brighter than the observed one (e.g. Iwasawa et al. 1997).Deeply buried AGN with luminosities approaching those of QSOs( L X ≈ erg s − ) among the red galaxies in this study can-not be ruled. Such objects are postulated to match the spectrumof the diffuse X-ray background (e.g. Gilli et al. 2007), but, cur-rently, few have been securely identified (e.g. Georgantopoulos &Georgakakis 2007; Tozzi et al. 2006). Selection methods that in-clude mid-IR wavelengths, like the one used here ( µm -brightand ∆ C > − . ) , are capable of detecting this population (e.g.Lacy et al. 2004; Stern et al. 2005; Martinez-Sansigre et al. 2006;Donley et al. 2007). X-ray stacking analysis has indeed confirmedthat at least some of these methods produce samples with hardmean X-ray spectra that are consistent with heavily obscured AGN(e.g. Daddi et al. 2007; Fiore et al. 2007; Georgantopoulos et al.2007).Brand et al. (2005) also reported a hard stacked X-ray sig-nal for red galaxies in the range . < z < . selected in theXBootes field of the NOAO Deep Wide-Field Survey (Kenter etal. 2005). These authors suggested that the detected signal can beinterpreted as unabsorbed emission from a radiatively inefficientaccretion (such as ADAFs; Narayan & Yi 1994, 1995). We explorethis possibility by comparing the mean Eddington ratio of µm -bright galaxies with ∆ C > − . with that of M 87, the best stud-ied example of a system undergoing radiatively inefficient accre-tion. The central supermassive BH of this galaxy has been shownto have very low radiative efficiency, corresponding to an Edding-ton ratio of ≈ − (Di Matteo et al. 2003). The mean opticalabsolute magnitude of µm -bright galaxies with ∆ C > − . is M B ≈ − . mag, which corresponds to a mean stellar mass of × M ⊙ for a mass–to–light ratio of log M / L = +0 . , con-sistent with the average colours of the galaxies with ∆ C > − . (Bell & de Jong 2001). Assuming that the bulk of the stellar massis associated with the galaxy bulge and using the relation betweenBH mass and bulge dynamical mass of H¨aring & Rix (2004), weestimate a BH mass of × M ⊙ . This is an upper limit as theestimated stellar mass of the µm -bright population may have alarge contribution from the disk of these galaxies, which has beenignored in this exercise. The mean 2-10 keV X-ray luminosity ofthis population (Table 2) corresponds to a bolometric luminosity of L bol = 3 . × erg s − , adopting the bolometric correction ofElvis et al. (1994). This is a lower limit in the case of Compton thickactivity. The BH mass and the L bol estimates above translate to a lower limit for the Eddington ratio of ≈ × − , more than 3 dexhigher than the corresponding quantity for M 87. If the hard X-raysignal of µm -bright galaxies with ∆ C > − . is associatedwith a radiatively inefficient accretion mode, they have to be sig-nificantly less extreme than M 87. For comparison, X-ray detected AGN at z ≈ have Eddington ratios in the range − − − (e.g. Bundy et al. 2007; Babic et al. 2007). The mean Eddingtonratio of the µm -bright galaxies with ∆ C > − . lies at thelow-end of the interval above, suggesting that they are dominatedby low luminosity and/or Compton Thick AGN. In the nearby Universe, z . . there has been evidence for anassociation between optically selected AGN, recent star-formationand galaxies in the transition zone from the blue to the red cloud.Kauffmann et al. (2003) found that luminous AGN in the SDSS arehosted by early-type bulge-dominated galaxies that have stoppedforming stars, but only in the recent past. Martin et al. (2007)extended this study by constructing the near-UV/optical CMD ofgalaxies at z < . using data from the GALEX Medium Imag-ing Survey and the SDSS. The density of Seyfert-2s in that sample,identified by diagnostic emission line ratios, peaks in the valleybetween the two clouds. Our work extends these results to higherredshift, z ≈ . , by finding evidence for obscured AGN amonggalaxies in the vicinity of the valley of the CMD and in the redcloud. Similar results also apply to the X-ray detected AGN popu-lation. X-ray sources with ∆ C > − . have an average hardnessratio of − . ± . , while for those with ∆ C < − . the meanhardness ratio is − . ± . .The incidence of AGN among transition zone galaxies sug-gests that BH accretion may play a role in the evolution of thesesystems. This link has been highlighted recently by Hopkins et al.(2007) who show that the buildup rate of the red cloud mass func-tion is in good agreement with predictions based on the QSO lumi-nosity function. This calculation assumes either a fixed Eddingtonaccretion rate for QSOs and the local BH-host mass empirical re-lation or a more physically motivated model for the QSO evolutionbased on simulations where BH accretion is triggered by gas-richgalaxy mergers (Hopkins et al. 2006a, b). Hopkins et al. (2007)also found that the observed rate of galaxy mergers and their dis-tribution in stellar mass at different redshifts are broadly consistentwith the growth rate of the red cloud mass function. The evidenceabove suggests an association between AGN activity, the transitionof galaxies to the red cloud, and mergers. These catastrophic eventscan lead to the morphological transformation of galaxies and at thesame time offer an efficient way for channeling gas to the nuclearregions of the galaxy, triggering the growth of the central BH.Although our analysis broadly supports an association be-tween AGN and galaxies on the move to the red cloud, the evidencefor a causal link between the two is less clear. The X-ray spectra ofAGN in the red cloud (both detections and stacking results) imply atleast moderate amounts of obscuration in a large fraction of thesesystems. This is counter-intuitive to a picture where AGN-drivenoutflows blow away the gas and dust clouds from the nuclear galaxyregions and suggests that some gas and dust clouds either remain ator relax to the centre of the galaxy after the quenching of the star-formation to form a torus. Additionally, major mergers are not sup-ported by the data. Minor mergers or tidal disruptions may still playa role in the evolution of µm -bright galaxies with ∆ C > − . .Recent studies on the evolution of the star-formation since z ≈ also suggest that a decline in the major mergers is not the dominantphysical mechanism driving the evolution. Gas exhaustion in diskgalaxies that form stars in a quiescent mode may be more important(e.g. Bell et al. 2005; Wolf et al. 2005; Melbourne, Koo & Le Floch2005; Noeske et al. 2007a, b). The gas depletion scenario is alsoconsistent with the obscured AGN activity observed in galaxies inthe transition zone and in the red cloud. c (cid:13) , 000–000 he role of AGN in the colour transformation of galaxies at redshifts z ≈ In the local Universe, z < . , an increasing body of evidencepoints to a link between black-hole accretion and post-starbursts.Yan et al. (2006) performed a systematic search for post-starburstsin the SDSS using methods similar to those described here. Alarge fraction of the sources in that sample ( > per cent) showweak emission lines consistent with low-luminosity and/or ob-scured AGN, e.g. LINERs, Seyfert-2s, transition objects. In a com-plementary study Goto (2006) selected SDSS galaxies with deep H δ absorption and emission line signatures typical of AGN. Thesesources represent about 4 per cent of the AGN in a volume limitedsample, much higher than in the overall SDSS population (0.2 percent).Our results extend the association between AGN and post-stabursts to high redshift, z ≈ . . Firstly, there is an increasedfraction of X-ray detected AGN among the post-starburst galaxypopulation. Secondly, about 20 per cent of the red-cloud AGN inthe redshift interval . . z . . , where post-starbursts canbe identified, are hosted by such galaxies. Most of the X-ray de-tected post-starbursts are obscured, with an average hardness ratioof ≈ − . . Thirdly, stacking the X-ray photons at the positionsof post-starbursts that are not individually detected at X-ray wave-lengths, reveals a hard mean X-ray spectrum and suggests obscuredAGN activity in the bulk of this population.The low redshift, post-starburst sample of Yan et al. (2006)is comparable in terms of selection to that presented here. Mostof the emission line sources in that sample belong to the LINERclass. These low-luminosity AGN often show hard X-ray spectraand, in many cases, are associated with moderate column densi-ties, N H ≈ − cm − , or even Compton thick systemswhere the X-ray emission is dominated by reflected radiation (e.g.Terashima et al. 2002; Gonz´alez-Mart´ın et al. 2006). In the lattercase, the intrinsic AGN power is likely to be significantly higherthan the observed one. The evidence above suggests a similar na-ture for the z ≈ . post-starbursts for which our analysis shows ahard X-ray spectrum in the mean. While the incidence of AGN inpost-starbursts is consistent with models where BH accretion is re-sponsible for quenching the star-formation in galaxies, the fact thanmany of these AGN are obscured means that gas and dust cloudsremain in the nuclear galaxy regions, possibly in a form of a torus,after the AGN-driven blow-out phase. A striking result from this study is that the red cloud includes alarge population of obscured AGN, both above and below the X-ray detection threshold of the AEGIS
Chandra survey. This extendsrecent studies, which have shown that a large fraction of the X-ray detected
AGN at z ≈ , and certainly, the majority of the obscuredones, are hosted by red galaxies (e.g. Georgakakis, Georgantopou-los & Akylas 2006; Nandra et al. 2007; Rovilos & Georgantopoulos2007).The incidence of a large population of obscured AGN in thered cloud suggests that the BH accretion outlives the terminationof star-formation. This is also supported by the X-ray properties ofpost-starburst galaxies in our study. It is not clear why such a largefraction of red cloud galaxies and post-starbursts show high columndensities. If these sources represent systems after the AGN driventermination of star-formation then one would expect the central en-gine to be unobscured. The X-ray obscuration in these systems maytherefore represent cold gas that either has not been blown out, or has not been processed and which relaxes to the galaxy centre toform a torus surrounding the central engine (e.g. Hopkins et al.2006a). Alternatively, gas exhaustion and not necessarily outflowsdriven by accreting BHs, may be the mechanism behind the evolu-tion of both AGN and galaxies. In this picture the AGN may remainobscured as long as there are dust and gas clouds to feed the cen-tral BH. Also, some of the red cloud µ m bright sources may bedusty systems (hence the red colour) observed before the quench-ing of the star-formation.The large fraction of obscured AGN in the red cloud is con-trary to the distribution of optically selected, low redshift AGN inthe GALEX near-UV/optical CMD (Martin et al. 2007). The num-ber density of Seyfert-2s in that study peaks in the valley. This dis-crepancy may be related to selection effects. For example, dilutionof the emission lines by the host galaxy stellar population makesthe identification of the AGN optical spectral signatures difficult(e.g. Comastri et al. 2002; Severgnini et al. 2003; Georgantopoulos& Georgakakis 2005). Also, contamination of the GALEX near-UV band by scattered AGN light will move red galaxies into thevalley. Alternatively, the GALEX near-UV bands may be more sen-sitive to low level star-formation compared to optical wavebands.In any case, this potential discrepancy highlights the need to studyin more detail the overlap and the differences between optical andX-ray AGN selection methods in terms of host galaxy properties. We explore the role of AGN in establishing the bimodal colour dis-tribution of galaxies by quenching the star-formation of blue star-forming systems causing their transformation to red, evolved galax-ies. The main conclusions from this study are summarised below1. There is evidence for AGN activity among galaxies in thetransition zone between the red and blue clouds of the CMD. Alarge fraction of these accreting BHs are obscured, suggesting thatif AGN outflows are related to the colour transformation of galax-ies, at least some nuclear gas and dust gas clouds are either notaffected or can efficiently reform after the truncation of the star-formation.2. Morphological analysis suggests that major mergers do notdominate the evolution of this population. Minor interactions how-ever, may play a role.3. We find an association between BH accretion and post-starbursts at z ≈ . , in agreement with studies on the propertiesof AGN hosts at low redshift, z ≈ . .4. AGN activity outlives the termination of the star-formation.A large fraction of active BHs are present in red cloud galaxies.This is in contrast to optically selected AGN, which lie predomi-nantly in the valley between the two clouds. We thank the referee, Rachel Somerville, for providing constructivecomments and suggestions that significantly improved this paper.This work has been supported by funding from the Marie-CurieFellowship grant MEIF-CT-2005-025108 (AG), STFC (ESL) andChandra grant GO5-6141A (DCK). JML acknowledges supportfrom the NOAO Leo Goldberg Fellowship, the NASA grants HST-GO-10314.13-A and HST-AR-10675-01-A from the Space Tele-scope Science Institute, which is operated by the AURA, Inc., un-der NASA contract NAS5-26555, NASA grant NAG5-11513 to P. c (cid:13) , 000–000 Georgakakis et al.
Madau. JAN is supported by NASA through Hubble Fellowshipgrant HF-011065.01-A, awarded by the Space Telescope ScienceInstitute, which is operated by the Association of Universities forResearch in Astronomy, Inc., for NASA, under contract NAS 5-26555.Support for GO program 10134 was provided by NASAthrough NASA grant HST-G0-10134.18-A from the Space Tele-scope Science Institute, which is operated by the Association ofUniversities for Research in Astronomy, Inc., under NASA contractNAS 5-26555.The authors wish to recognise and acknowledge the very sig-nificant cultural role and reverence that the summit of Mauna Keahas always had within the indigenous Hawaiian community. Weare most fortunate to have the opportunity to conduct observationsfrom this mountain. This work is based in part on observationsmade with the
Spitzer
Space Telescope, which is operated by theJet Propulsion Laboratory, California Institute of Technology un-der a contract with NASA. Support for this work was provided byNASA through an award issued by JPL/Caltech.
REFERENCES
Abraham R. G., et al., 1996, ApJS, 107, 1Alexander D. M., et al., 2005, Nature, 434, 738Alexander D. M., et al., 2003, AJ, 126, 539Babic A., Miller L., Jarvis M. J., Turner T. J., Alexander D. M., Croom S.M., 2007, A&A, 474, 755Baldry I. K., Glazebrook K., Brinkmann J., Ivezic Z., Lupton R. H., NicholR. C., Szalay A. S., 2004, ApJ, 600, 681Barger A. J., et al., 2005, AJ, 129, 578Barmby P., et al., 2006, ApJ, 642, 126Bell E., Phleps S., Somerville R. S., Wolf C., Borch A., Meisenheimer K.,2006a, ApJ, 652, 270Bell E., et al., 2006b, ApJ, 640, 241Bell E., et al., 2005, ApJ, 625, 23Bell E., et al., 2004, ApJ, 608, 752Blanton M. R., 2006, ApJ, 648, 268Brand K., et al., ApJ, 2005, 626, 723Brusa M., et al., 2007, ApJS, 172, 353Bundy K., et al., 2007, ApJ, submitted, arXiv0710.2105Cattaneo A., et al., 2007, MNRAS, 377, 63Cid Fernandes R., et al., 2001, ApJ, 558, 81Conselice C. J., Bershady M. A., Jangren A., 2000, ApJ, 529, 886Comastri A., et al., 2002, ApJ, 571, 771Croton D. J, et al., 2006, MNRAS, 365, 11Daddi E., et al., 2007, ApJ, submitted, arXiv:0705.2832Dale D. A., et al., 2007, ApJ, 655, 863Davis M., et al., 2007, ApJ, 660, 1LDekel A., Birnboim Y., 2006, MNRAS, 368, 2Di Matteo T., Allen S. W., Fabia A. C. Fabian, Wilson A. S., Young A. J.,ApJ, 2003, 582, 133.Donley J. L., et al., 2005, ApJ, 634, 169Elvis M., 1994, ApJS, 95, 1Faber et al. 2003, SPIE, 4841, 1657Fabian A. C., 1999, MNRAS, 308L, 39Ferrarese L. & Merritt D., 2000, ApJ, 539, L9Franceschini A., et al., 2003, MNRAS, 343, 1181Fiore F., et al., 2007, ApJ, submitted, arXiv:0705.2864Gebhardt K., et al., 2000, ApJ, 539, L13Genzel R., et al., 1998, ApJ, 498, 579Georgakakis A., Georgantopoulos I. & Akylas A., 2006, MNRAS, 366 ,171Georgakakis A., Hopkins A. M., Afonso J., Sullivan M., Mobasher B., CramL. E., 2004, MNRAS, 354, 127Georgantopoulos I., et al., 2007, MNRAS, submittedGeorgantopoulos I. & Georgakakis A., 2007, A&A, 466, 823 Georgantopoulos I. & Georgakakis A., 2005, MNRAS, 358, 131Gerke B. F., et al., 2007, MNRAS, 376, 1425Gilli R., Comastri A., Hasinger G., 2007, A&A, 463, 79Gonz´alez-Mart´ın O., Masegosa J., M´arquez I., Guerrero M. A., Dultzin-Hacyan D., 2006, A&A, 460, 45Goto T., 2006, MNRAS, 369, 1765Grogin N. A., 2005, ApJ, 627, 97LGrogin N. A., et al., 2003, ApJ, 595, 685H¨aring N. & Rix H. W., 2004, ApJ, 604L, 89Hasinger G., Miyaji T., Schmidt M., 2005, A&A, 441, 417Hopkins P. F., et al., 2005, ApJ, 630, 705Hopkins P. F., Hernquist L., Cox T. J., Di Matteo T., Robertson B., S. Volker,2006a, ApJS, 163, 1Hopkins P. F., Somerville R. S., Hernquist L., Cox T. J., Robertson B., LiY., 2006b, ApJ, 652, 864Hopkins P. F., Bundy K., Hernquist L., Ellis R. S., 2007, ApJ, 659, 976Iwasawa K., Fabian A. C., Matt G., 1997, MNRAS, 289, 443Kauffmann G., et al., 2004, MNRAS, 353, 713Kauffmann G., et al., 2003, MNRAS, 346, 1055Kenter A., et al., 2005, ApJS, 161, 9King A., 2003, ApJ, 596, 27LCirasuolo M., et al., 2007, MNRAS, submitted, arXiv:astro-ph/0609287Lacy M., et al., 2004, ApJS, 154, 166Laird E. S., Nandra K., Adelberger K. L., Steidel C. C., Reddy N. A., 2005,MNRAS, 359, 47Lotz J. M., et al., 2007, ApJ, submitted, arXiv:astro-ph/0602088Lotz J. M., Primack J., Madau P., 2004, AJ, 128, 163Magorrian J., et al., 1998, AJ, 115, 2285Martin C. D., et al., 2007, ApJS, in press, astro-ph: 0703281Martinez-Sansigre A., et al., 2006, MNRAS, 370, 1479Matt G., et al., 1999, A&A, 341, 39LMelbourne J., Koo D. C., Le Floch E., 2005, ApJ, 632L, 65Nandra K., et al., 2007, ApJ, 660, 11LNandra K., et al., 2005, MNRAS, 356, 568Nandra K., Mushotzky R. F., Arnaud K., Steidel C. C., Adelberger K. L.,Gardner J. P., Teplitz H. I., Windhorst R. A., 2002, ApJ, 576, 625Nandra K. & Pounds K. A., 1994, MNRAS, 268, 405Narayan R., Igumenshchev I. V., & Abramowicz M. A. 2000, ApJ, 539, 798Narayan R., & Yi I., 1994, ApJ, 428, L13Noeske K. G. et al., 2007a ,ApJ, 660L, 47Noeske K. G. et al., 2007b ,ApJ, 660L, 43Okamoto T., Nemmen R. S., Bower R. G., 2007, MNRAS, submitted,arXiv:0704.1218Pierce C. M., et al., 2007, ApJ, 660, 19LPtak A., Heckman T., Levenson N. A., Weaver K., Strickland D., 2003, ApJ,592, 782Quintero A. D., et al. 2004, ApJ, 602, 190Ranalli P., Comastri A., Setti G., 2003, A&A, 399, 39Rovilos E., Georgakakis A., Georgantopoulos I., Afonso J., Koekemoer A.M., Mobasher B., Goudis C., 2007, A&A, 466, 119Rovilos E. & Georgantopoulos I., 2007, A&A, 475, 115Severgnini P., et al., 2003, A&A, 406, 483Silk J., & Rees M. J. 1998, A&A, 331, 1LStern D., et al., 2005, ApJ, 631, 163Strateva I., 2001, AJ, 122, 1861Terashima Y., Iyomoto N., Ho L. C., Ptak A. F., 2002, ApJS, 139, 1Tozzi P., et al., 2006, A&A, 451, 457Weiner B. J., 2005, ApJ, 620, 595Willmer C. N. A., et al., 2006, ApJ, 647, 853Wolf C., et al., 2005, ApJ, 630, 771Yan R. et al., 2006, ApJ, 648, 281Yan L. et al., 2004, ApJS, 154, 60 c (cid:13)000