The Evolution of AGN Host Galaxies: From Blue to Red and the Influence of Large-Scale Structures
J. D. Silverman, V. Mainieri, B. D. Lehmer, D. M. Alexander, F. E. Bauer, J. Bergeron, W. N. Brandt, R. Gilli, G. Hasinger, D. P. Schneider, P. Tozzi, C. Vignali, A. M. Koekemoer, T. Miyaji, P. Popesso, P. Rosati, G. Szokoly
aa r X i v : . [ a s t r o - ph ] M a r Published in ApJ, 2008, 675, 1025
Preprint typeset using L A TEX style emulateapj v. 02/07/07
THE EVOLUTION OF AGN HOST GALAXIES: FROM BLUE TO RED AND THE INFLUENCE OFLARGE-SCALE STRUCTURES
J. D. Silverman, V. Mainieri,
B. D. Lehmer, D. M. Alexander, F. E. Bauer, J. Bergeron,
W. N. Brandt, R. Gilli, G. Hasinger, D. P. Schneider, P. Tozzi, C. Vignali, A. M. Koekemoer, T. Miyaji, P. Popesso,
P. Rosati, & G. Szokoly Published in ApJ, 2008, 675, 1025
ABSTRACTWe present an analysis of 109 moderate-luminosity (41 . ≤ log L . − . ≤ .
7) AGN in theExtended
Chandra
Deep Field-South survey, which is drawn from 5,549 galaxies from the COMBO-17and GEMS surveys having 0 . ≤ z ≤ .
1. These obscured or optically-weak AGN facilitate the studyof their host galaxies since the AGN provide an insubstantial amount of contamination to the galaxylight. We find that the color distribution of AGN host galaxies is highly dependent upon (1) thestrong color-evolution of luminous ( M V < − .
7) galaxies, and (2) the influence of ∼
10 Mpc scalestructures. When excluding galaxies within the redshift range 0 . ≤ z ≤ .
76, a regime dominated bysources in large-scale structures at z = 0 .
67 and z = 0 .
73, we observe a bimodality in the host galaxycolors. Galaxies hosting AGN at z & . U − V < .
7) colorsthan their z . . . < U − V < . . ≤ z ≤ .
76. The AGN fraction in this redshift and color intervalis 12.8% (compared to its ‘field’ value of 7 . .
8% at U − V ∼ . n > .
5) galaxies have the highest fractionof AGN (21%) in our sample. We explore the scenario that the evolution of AGN hosts is drivenby galaxy mergers and illustrate that an accurate assessment requires a larger area survey since onlythree hosts may be undergoing a merger with timescales . Subject headings: quasars: general, galaxies: active, galaxies: evolution, X-rays: galaxies, large-scalestructure of universe INTRODUCTION
There has been remarkable evidence found in the lastfew years that the evolution of supermassive black holes(SMBHs) and galaxies are inextricably linked. For in-stance, ultraluminous infrared and submillimeter galax-ies (Alexander et al. 2005), with prodigious rates of starformation, show a high fraction of AGN activity. Cur-rent models of galaxy evolution (e.g., Di Matteo et al. Max-Planck-Institut f¨ur extraterrestrische Physik, D-84571Garching, Germany European Southern Observatory, Karl-Schwarzschild-Strasse 2,Garching, D-85748, Germany Department of Astronomy & Astrophysics, 525 Davey Lab,The Pennsylvania State University, University Park, PA 16802,USA Department of Physics, University of Durham, South Road,Durham, DH1 3LE, UK Columbia Astrophysics Laboratory, Columbia University,Pupin Labortories, 550 W. 120th St., Rm 1418, New York, NY10027, USA Institut d’Astrophysique de Paris, 98bis Boulevard, F-75014Paris, France Istituto Nazionale di Astrofisica (INAF) - Osservatorio Astro-nomico di Bologna, Via Ranzani 1, 40127 Bologna, Italy INAF - Osservatorio Astronomico di Trieste, via G. B. Tiepolo11, 34131 Trieste, Italy Dipartimento di Astronomia, Universit´a degli Studi diBologna, Via Ranzani 1, 40127 Bologna, Italy Space Telescope Science Institute, 3700 San Martin Drive,Baltimore, MD 21218, USA Department of Physics, Carnegie Mellon University, Pitts-burgh, PA 15213, USA Based on observations made at the European Southern Ob-servatory, Paranal, Chile (ESO programs 170.A-0788, 171.A-3045,072.A-0139) z ∼ z < . z ∼
1) of the global star-formation and merger rate of galaxies Silverman et al.(Hopkins et al. 2006b; Kartaltepe et al. 2007). Recently,Nandra et al. (2007) have shown that the host galaxies ofmoderate-luminosity AGN with 0 . < z < .
4, found inthe Extended Groth Strip, have a broad range of opticalcolors that span the same region of the color-magnituderelation as luminous ( M B < − .
5) galaxies. The preva-lence of AGN host galaxies within the region separatingthe blue and red galaxy populations may lend supportfor the importance of AGN feedback since this locationis thought to represent a transitional phase in the evolu-tion of galaxies.It is of much interest to constrain observationallywhether environment plays a significant role in thegrowth of SMBHs as expected by merger-driven accre-tion models. Observational support for the importance ofmergers is mainly based on circumstantial evidence: (1)AGN predominately reside in massive, early-type galax-ies (e.g., Kauffmann et al. 2003), of which many aremassive ellipticals, well thought to be the end-product ofa major-merger between gas-rich disk galaxies, (2) ultra-luminous infrared galaxies, that are morphologically dis-turbed in almost all cases, have a high fraction ( ∼ z < .
3) AGN from the SDSS in-dicate that the fraction of galaxies harboring AGN is in-dependent of environment (e.g., Miller et al. 2003) evenin the cores of massive clusters. For the most-luminousnarrow-line AGN in the SDSS (Kauffmann et al. 2004),an AGN-fraction–density relation has been shown to ex-ist analogous to the SFR-density relation with a higherAGN fraction in low-density (i.e., ‘field’) environments.Both studies provide evidence counter to the expecta-tion that SMBH accretion is induced by galaxy mergers.Hard X-ray selected surveys, most effective at identify-ing obscured accretion at higher redshifts, also presentdisparate views on the relationship between AGN ac-tivity and their environments. Georgakakis et al. (2007)present preliminary results that demonstrate that X-rayselected AGN at z ∼ U − V .
1) hosts are found in denserenvironments, hinting at a connection to star forma-tion. On the contrary, Grogin et al. (2005) find thatAGN in the
Chandra
Deep Fields show no evidencefor an environmental dependency based on similar AGNhost morphologies and near-neighbor counts to the non-active galaxy population. The AGN fraction in clusters(Martini et al. 2006, 2007), though higher than previ-ously determined (Dressler & Gunn 1983), is not signif-icantly different than that in the ‘field’; a different pic-ture may emerge at higher redshifts since Eastman et al.(2007) find significant evolution of the AGN fraction inclusters at z ∼ .
6. There is also evidence that largerscale ( & z = 0 .
67 and z = 0 .
73 and spread acrossthe full
Chandra field-of-view (17 ′ × ′ ), which corre- sponds to a physical scale of 7.3 Mpc at z = 0 .
7. Similarstructures are evident in the CDF-N (Barger et al. 2003)that may indicate the importance of large-scale ( ∼ Chandra
Deep Field-South (E-CDF-S)is an ideal survey field to use for investigating the proper-ties of galaxies harboring AGN and the role of environ-ment due to its remarkable multi-wavelength coverage.We have completed a 1 Ms
Chandra
Legacy program(P.I.: W. N. Brandt; Lehmer et al. 2005) that coversa wide area (0.33 deg ; three times area of the CDF-S) at the depths required to detect moderate-luminosity( L X ∼ erg s − ) AGN, including those with signifi-cant obscuration, out to the quasar epoch ( z ∼ . σ z ≈ HST
Advanced Camera for Sur-veys (ACS) observations via the GEMS (Rix et al. 2004;H¨aussler et al. 2007), GOODS (Giavalisco et al. 2004)and the
HST
Ultra Deep field (UDF; Beckwith et al.2006) projects. Over 1000 spectroscopic redshifts areavailable via the CDF-S (Szokoly et al. 2004), VVDS(Le Fevre et al. 2004), K20 (Mignoli et al. 2005), andGOODS (Vanzella et al. 2005, 2006) surveys.In this paper, we investigate the location of moderate-luminosity AGN, in the E-CDF-S, on the color-magnitude diagram and their relation to the underly-ing galaxy population using our current catalog of X-ray selected AGN with either spectroscopic or photo-metric redshifts. The E-CDF-S contains two prominentredshift spikes (Gilli et al. 2003) enabling us to deter-mine the influence of ∼
10 Mpc structures on the over-all color-magnitude distribution. We discuss how ourresults fit in with the morphological properties of thesample and the impact on galaxy evolution models thatincorporate AGN feedback to quench star formation ef-fectively. Throughout this work, we assume H = 70 kms − Mpc − , Ω Λ = 0 .
7, and Ω M = 0 . DATA
We select a parent sample of galaxies in the E-CDF-Susing published catalogs from the COMBO-17 andGEMS surveys. The
Chandra observations, covering theequivalent sky area, enable identification of those galax-ies that harbor moderate-luminosity AGN in a mannerthat is least biased against obscuration. Follow-up op-tical spectroscopy of these X-ray sources with the VLTfacilitates the identification of AGN. As detailed below,our selection is tuned to generate a sample of galaxieshosting AGN for which the optical emission is dominatedby the host galaxy; thus no further removal of AGN light(i.e., cleaning) is required for this study. By restrictingourselves to an initial optically-selected sample of galax-ies, we aim to measure the fraction of galaxies harboringAGN as a function of their intrinsic properties.
Parent galaxy population volution of AGN host galaxies 3COMBO-17 provides a highly complete sample ofgalaxies over the full E-CDF-S (Wolf et al. 2004) withwell-known intrinsic properties (i.e., magnitudes and col-ors). The survey provides reliable object classifications(e.g., Galaxy, QSO, Star) and photometric redshifts byfitting synthetic template optical spectra to the observedmagnitudes over a broad wavelength range (3500–9300˚A). The source catalog contains 8,565 objects with aper-ture magnitudes R ap ≤
24, a limit at which there arephotometric redshift errors of δ z / (1 + z ) < .
1. Thismagnitude limit ensures that our sample is representa-tive of galaxies of all colors with M V . −
21 and z <
HST /ACS imaging of the field in both theF606W and F850LP filters (hereafter referred to V and z , respectively) from GEMS. Here, we use the S´ersicindices ( n ) given in H¨aussler et al. (2007) to discriminatebetween bulge and disk-dominated galaxies.We select a sample of 5,549 galaxies with photomet-ric redshifts of 0 . ≤ z ≤ . z -bandselected catalog of Caldwell et al. (2005) that has beencross-referenced to objects in COMBO-17 ( R ap ≤ M V . −
21) galaxies and hence those withAGN. Only two AGN have been unambigously identifiedat z < . L X ∼ erg s − . The redshiftcutoff at 1.1 is the limit for which COMBO-17 no longerprovides accurate source classification, and redshifts aresusceptible to large uncertainties. We primarily use therest-frame optical magnitudes ( M U , M V ; Vega magni-tudes), publicly available from COMBO-17 (Wolf et al.2004), derived from synthetic galaxy templates to mea-sure intrinsic luminosity and color. To compare withtheoretical models ( § u and r (SDSS) also provided by COMBO-17. The errors(1 σ ; Wolf et al. 2004) on rest-frame magnitudes are typ-ically 0.11 ( M U ) and 0.15 mag ( M V ). Throughout thiswork, a representative error (1 σ ) on color (U-V) is de-termined to be 0.19 from the quadrature of the errors ofthe individual rest-frame magnitudes; this is most likelyan overestimate since these magnitude errors are corre-lated. Bell et al. (2004) report a typical error on U–V tobe ∼ . AGN identification
We compile a sample of AGN, based on their high X-ray luminosities, in the E-CDF-S by matching the 762
Chandra point sources given in Lehmer et al. (2005) tothe available optical and near-infrared catalogs. We donot consider here the fainter X-ray sources solely de-tected in the 1 Ms CDF-S to maintain a fairly uni-form sensitivity across the entire E-CDF-S. Redshifts are available for 362 (48%) of the X-ray sources through ei-ther spectroscopic or photometric techniques. We signifi-cantly improve upon the 97 spectroscopic redshifts avail-able in the literature (Szokoly et al. 2004; Le Fevre et al.2004; Mignoli et al. 2005; Vanzella et al. 2005, 2006)with 95 additional redshifts acquired by observationswith VIMOS (Le Fevre et al. 2003) on the VLT throughthe ESO programs 072.A-0139 (P.I.: J. Bergeron; 65 red-shifts; Silverman et al. 2007) and 171.A-3045 (GOODS;P.I.: C. Cesarski; 30 redshifts; Popesso et al., in prepara-tion). Our VIMOS observations are unique since we haveacquired spectra over a wide wavelength range (3600–9500 ˚A) for many targets by observing with both the LR-blue and MRred grisms. With integration times reach-ing 5 hr, we are able to detect faint spectral features(e.g., [O II], Mg II, Fe II) in many of the host galaxiesof obscured or optically-faint AGN that enable a red-shift measurement. A detailed discussion of the VIMOSobservations will be provided in a subsequent paper (Sil-verman et al. 2007, in preparation). The X-ray lumi-nosity of each AGN is determined from their observedbroad-band (0.5-8.0 keV) flux, given in Lehmer et al.(2005), with a k -correction based on a power-law spec-trum (photon index Γ = 1 .
9) and no correction for in-trinsic absorption. Here, we restrict the luminosity rangeto 41 . ≤ log L . − . keV ≤ .
7. The lower luminositylimit is chosen to provide a robust sample free of anygalaxies having significant X-ray binary or diffuse X-rayemission; luminous starburst galaxies in the
Chandra
Deep Fields are typically a factor of ∼
10 fainter (seeFig. 4 of Bauer et al. 2002). We also minimize potentialluminosity-dependent effects since lower luminosity AGNwould only be detected over a small fraction of the red-shift range ( z = 0.4–1.1) considered here. The Chandra observations of the E-CDF-S are capable of detectingAGN at this lower luminosity limit up to z ∼ .
85. Ourupper limit is motivated by Silverman et al. (2005) whoshow that the optical emission, associated with X-rayselected AGN at log νl ν < . E = 2 keV is pri-marily due to their host galaxy since there is a strongdeparture of these AGN from the known l opt − l X rela-tion for more luminous X-ray selected AGN. We convertthis limit to a broad-band (0.5–8.0 keV) X-ray luminosityof 10 . erg s − assuming the same spectrum as givenabove. Furthermore, our adopted upper limit is simi-lar to that of Nandra et al. (2007) who demonstrate thatthe host galaxies of moderate-luminosity AGN, at X-rayfluxes equivalent to those detected in the E-CDF-S, con-tribute the majority of the total optical emission. Asdetailed below, we have measured upper limits to theAGN contribution, using the HST /ACS imaging of thefield, and show that the optical emission, for the ma-jority of our sample, is dominated by the host galaxy.Optical spectra provide a final check on the AGN contri-bution with most sources clearly lacking any faint broademission lines or a rising blue continuum.A cross-correlation of the X-ray and optical catalogshas indicated that 109 of the 5,549 galaxies have AGNwith L X ≈ . –10 . erg s − . Of these, spectro-scopic redshifts were available for 54 of these sources,16 (7 at z >
1) of which were from our VIMOS observa-tions. Rest-frame magnitudes for 12 galaxies were red-erived (C. Wolf, private communication) using COMBO- Silverman et al.17 tools since their spectroscopic redshifts differed sub-stantially from their photometric redshifts (∆ z > . z ∼
1, a regime where photometric redshifts forgalaxies in COMBO-17 show a higher dispersion (seeFigure 6 of Wolf et al. 2004), and lack strong contin-uum features (e.g., 4000 ˚A break) most useful for ad-equate photometric redshift estimates. Our AGN sam-ple is primarily radio-quiet since only nine have a radio-loudness ( R ) greater than 10 ( R ≡ L /L ˚ A ) basedon 20cm VLA detections (Tozzi et al. in preparation)that reach a flux limit of 8.5 µJy , optical emission at-tributed to the AGN (25%; see the following section)based on our R -band optical magnitudes, and spectralindices ( L ν ∝ ν − γ ; γ radio = 0 . γ opt = 0 . R ap > AGN contribution to host-galaxy emission
We are confident that the optical emission, from these109 galaxies hosting AGN, is dominated by star lightand not strongly influenced by the AGN based on thefollowing arguments. In Figure 1, we show that thereis no correlation between X-ray luminosity and eitherrest-frame magnitude ( M V ; Pearson correlation coeffi-cient r = − . U − V ; r = − . L X &
44) AGN andany objects classified by COMBO-17 as a “QSO”.We determine conservative upper limits upon the AGNcontribution to the total (galaxy + AGN) light usingthe
HST /ACS V and z -band images provided byGEMS. These images, available from MAST (Multi-mission Archive at Space Telescope), have been throughthe mark ′′ pixel − . A few AGN arenot included in our analysis due to their proximity to theACS field edge. Extended optical emission is clearly evi-dent for all 109 galaxies hosting AGN. Optical counts aremeasured without subtracting the background in circularapertures of two different sizes positioned at the centroidof the optical emission: (1) a small aperture with a ra-dius of 3 pixels (0.09 ′′ ) that contains 50% of the flux froman unresolved point source (see Jahnke et al. 2004b) andcorresponds to a physical scale of 0.48 ( z = 0 .
4) to 0.74( z = 1 .
1) kpc, and (2) a larger aperture with a radiusof 25 pixels (0.75 ′′ ; covering a physical scale of 4.03–6.13kpc). The size of the larger aperture is set to that im-plemented for flux extraction by COMBO-17 (Wolf et al.2004). The counts within the small aperture provide afirm upper limit to the AGN contribution since we makeno attempt to remove emission of stellar origin. For com- Fig. 1.— (a) Rest-frame absolute magnitude ( M V ) and (b)optical color ( U − V ) as a function of X-ray luminosity for oursample of 109 AGN (filled symbols). The type of marker denotesAGN within three separate redshift intervals (circles: z = 0.4–0.63;squares: z = 0.63–0.76; triangles: z = 0.76–1.1). Open symbolsshow higher luminosity AGN not included in our sample. A typicalerror bar of size ± σ is placed in the upper right corner in bothpanels. parison, we measure counts in equivalent regions for theentire sample of 5,549 galaxies and 67 QSOs, which wereidentified by COMBO-17 and are not included in ourparent galaxy sample.The number distribution of the AGN host galaxiesas a function of the ratio of counts between these twoapertures is shown in Figure 2 for both the V -band(Fig. 2a) and z -band (Fig. 2b). Based on the V -band measurements, the mean ratio of counts of ourAGN sample (solid histogram) is 0 .
14 and 81% of themhave a ratio less than 0.2. The distribution is similar tothat of the galaxies (dotted histogram), though shiftedslightly by 0.04 (the difference of their median values),most likely due to the presence of optically-faint AGN.The host galaxy distribution is noticeably offset fromthat of the optically-selected QSOs, identified by spectraltemplate fitting with no preference for point-like sources(Wolf et al. 2004), that have a mean ratio of 0.48, and60% of them have a ratio greater than 0.5 most indica-tive of unresolved point sources. A number of low countratio QSOs are evident due to the fact that COMBO-17 can recognize lower luminosity Seyfert-1 galaxies asQSOs if strong AGN features are present. Similar re-volution of AGN host galaxies 5
Fig. 2.—
AGN contribution to the total (AGN + galaxy) light.The abscissa is the ratio of source counts in
HST /ACS images be-tween a circular aperture of radius 0.09 ′′ and 0.75 ′′ . In panel a, the V -band count-ratio distribution is shown for 104 AGN (solid his-togram), 5521 galaxies (dotted), and 67 (dashed) optically-selectedQSOs from COMBO-17. In panel b , the same is shown for the z -band (103 AGN, 5513 galaxies, 66 QSOs). QSOs from COMBO-17illustrate the typical count ratios for unresolved point sources. Thegalaxy number distribution in both panels has been scaled downto match the numbers of AGN host galaxies. The majority of theAGN host galaxies have &
80% of their total optical emission (inboth the V and z bands) outside the circular aperture of ra-dius 0.09 ′′ and have count ratios similar to the non-AGN galaxypopulation. sults are found for the z -band measurements, whichsuggests that color gradients are dominated by the hostgalaxies and not the underlying AGN. Since our sam-ple contains a large fraction of obscured AGN, there isa slightly larger contribution from AGN emission in the z band compared to the V band, whereas the op-posite is true for the quasars.The maximum amount that an AGN could shift itshost-galaxy rest-frame U − V color bluewards is esti-mated. We assume that an AGN contributes all theflux in the small ( r = 0 . ′′ ) aperture, the count ratioas detailed above is 0.2, and the background signal isnegligible. This corresponds to an AGN contributing anadditional amount of flux equal to 0.25 times the totalhost galaxy flux. We further consider an AGN to addflux to the U -band only. While rest-frame UV emissionis substantially diminished at λ < V band distribution in panel a shifted higher thanthat in panel b . We note that the rest-frame U -band is observed with the V band for objects with z ∼ . ∼ §
6, we utilizeoptical spectra, which are available for 46 AGN, to fur-ther confirm that the AGN host galaxies provide most ofthe optical light (see Fig. 8 and 9), thus supporting ourgeneral conclusions. REST-FRAME COLORS OF MODERATE-LUMINOSITYAGN
We were motivated by recent studies (Sanchez et al.2004; B¨ohm et al. 2006; Nandra et al. 2007) to investi-gate further the color-magnitude (rest-frame U − V versus M V ) relation of galaxies hosting moderate-luminosity X-ray selected AGN. Since the host galaxies in our samplecontribute most of the optical emission, as demonstratedin §
2, we do not need to remove the AGN componentfrom the total (host + AGN) optical emission. We usethe rest-frame U − V (i.e., M U − M V ) color up to z ∼ U − B color. We note that Wolf et al. (2004) caution thatthe rest-frame measurement of M V at z > . M U and M B rest-frame absolute magnitudes, thus removing the pos-sibility that our results are dependent on the assumedspectral template (i.e., k -correction).In Figure 3a, we plot the rest-frame colors ( U − V )as a function of absolute V -band magnitude ( M V ) forour galaxies, including those hosting AGN in the red-shift interval 0 . ≤ z ≤ .
1. We confirm past re-sults (Barger et al. 2003; B¨ohm et al. 2006; Nandra et al.2007) with better statistics: (1) a high fraction (80%) ofmoderate-luminosity AGN reside in the most-luminous( M V < − .
7) galaxies, (2) the rest-frame colors of AGNhost galaxies have a broad distribution, over the range0 < U − V < .
5, with no apparent evidence for a colorbimodality, as is distinctively evident in the underlyingpopulation of galaxies (Bell et al. 2004), and (3) the ma-jority (60%) of AGN host galaxies have bulge-dominatedmorphologies (S´ersic index n > .
5; Blanton et al. 2003;McIntosh et al. 2005) as marked by small blue dots.Here, we further find that 31–44% of the AGN with M V < − . . < U − V < .
8) between blue and redgalaxies (dashed line in Fig. 3a; see Sanchez et al. 2004),thus associating them with star-forming galaxies. Thisis not surprising since we have a fair number of AGNat z > .
8, an epoch for which the mean star-formation Silverman et al.rate of galaxies has increased by an order of magnitudecompared to the present value (e.g., Hopkins & Beacom2006; Noeske et al. 2007; Zheng et al. 2007). These re-sults are not strongly color biased since both red and blueluminous ( M V < − .
7) galaxies, out to z ∼
1, mainlyfall above the magnitude limits ( R ap .
24; Bell et al.2004) shown by the blue lines in Figure 3a ( z = 0 . z = 1 .
0: dash-dotted). We note that some red( U − V > . − . < M V < − . z > . M V < − .
7) tominimize any color bias. In the redshift interval 0 . 63 (see Fig. 4a), the AGN tend to have red colors( U − V > . 7) with a mean color h U − V i = 0 . ± . . < z < . 1; Fig. 4c) is h U − V i = 0 . ± . 08 with 29% (9 of 31) of the AGN having U − V > . 7. A similar fraction (27%; 211 of 790) of lu-minous galaxies at these redshifts has these colors. Thevariance of each color distribution is similar ( s ∼ . z > . 76 have blue colors ( U − V < . 7) sinceblue galaxies dominate the luminous population at theseredshifts (Wolf et al. 2003; Bell et al. 2004). The nullhypothesis, from a Kolmogorov-Smirnov (K-S) test (seePress et al. 1993), has a probability of 2.4% that the dis-tribution of host galaxy colors in Figure 4a and 4c couldbe drawn from the same parent population. We haveimplemented further K-S tests to determine whether theAGN distributions could be drawn from the underlyinggalaxy population. In Table 1, we give the results thatshow that the color distribution of AGN host galaxies inFigure 4a ( P K − S = 0 . 79) and 4c ( P K − S = 0 . 21) resem-bles that of the overall galaxy population. We conclude,based on these tests, that the host galaxies of moderate-luminosity AGN follow a similar passive evolution, oraging, as the underlying galaxy population migrates fromblue to red colors with cosmic time . We note that AGNwith spectroscopic redshifts, shown by a small blue dotin Figure 3b, confirm this trend. LARGE-SCALE INFLUENCES As previously mentioned, recent studies(Di Matteo et al. 2005; Croton et al. 2006;Hopkins et al. 2006a) have attributed the trunca-tion of star formation and eventual redward migrationof galaxies to merger-induced AGN feedback that effec-tively populates the red sequence with massive galaxies(Bell et al. 2004; Faber et al. 2006). This scenario mayexplain the fair number of AGN host galaxies in oursample residing in the “green valley” (see Fig. 3a) ifthey are preferentially located in overdense regions.In contrast to studies that characterize the environ-ment in terms of density local to AGN, we are utilizingthe fortuitous structures in the E-CDF-S to search forlarge-scale effects. To do so, we specifically isolate a red-shift interval (0 . ≤ z ≤ . 76) dominated by two red-shift spikes (Fig. 3b, vertical dotted lines at z = 0 . 67 and z = 0 . 73, each with δz < . 02; Gilli et al. 2003) evidentin the 1 Ms CDF-S area, which appear to extend over thelarger E-CDF-S area (Silverman et al. 2007). A largerfraction of AGN activity in galaxies within these large-scale structures was reported by Gilli et al. (2003) albeitwith limited significance (2 σ ). Ideally, one would like toconsider even narrower redshift intervals ( δz ∼ . 02) toselect galaxies cleanly within these spectroscopic redshiftspikes, but we are restricted here by photometric red-shift errors. Approximately 80% of the galaxies withinthis redshift interval have errors in redshift ( σ z ) less than0.07 , a factor of two smaller than the chosen bin width(∆ z = 0 . . < U − V < . 0) placing them in the “green valley”. Many of theAGN have spectroscopic redshifts, shown by the smallblue dots, that confirm their presence within the narrowredshift spikes (∆ z < . 02) and their rest-frame U − V colors that place them in the “green valley” (67%; 12of 18 AGN). We plot in Figure 4b the U − V color dis-tribution of both AGN and galaxies with M V < − . We find that thehost galaxies of moderate-luminosity AGN are preferen-tially located at intermediate colors ( U − V ∼ . betweenthe red and blue galaxy populations, and skewed towardblue colors . A K-S test gives a probability of 6.6% thatthe AGN distribution could be randomly drawn from theunderlying galaxy population and suggests that the dis-tribution of these 36 AGN is different from that of the669 underlying galaxies.There does not appear to be any strong selection ef-fects that could be responsible for the color distributionof AGN host galaxies within this redshift interval. Essen-tially, all of the AGN have optical luminosities well abovethe limit at z=0.8 shown in Figure 3a by the solid, blueline. By isolating luminous ( M V < − . 7) host galaxies,there is clearly no color bias. This is an important pointsince the 4000 ˚A break is moving through the R -band fil-ter at z ∼ . M U and and M V fall within the observablewindow, and multiple filters sample the galaxies’ spectralenergy distribution below and above the 4000 ˚A break.Both surface brightness dimming and the use of a fixedphotometric aperture can induce systematic changes incolor with redshift. An increase in the contrast between abulge and a disk, due to surface brightness dimming as afunction of redshift, would typically redden its rest-framecolor; this effect is of most concern when comparing col-ors over a wider redshift baseline. Bell et al. (2004) ad-dress aperture effects associated with COMBO-17 pho- Photometric redshift errors are based on equation 5 ofWolf et al. (2004) and a magnitude limit of 23.5. The results of additional tests between the color distributionsof AGN hosts (AGN-AGN) in the three redshift intervals are re-ported in Table 1. We note that their significance is limited giventhe lack of a large comparison sample that effectively improves theAGN-galaxy comparisons. volution of AGN host galaxies 7 Fig. 3.— (a) Rest-frame optical color ( U − V ) as a function of rest-frame absolute magnitude ( M V ) for 109 galaxies hosting AGNcompared to their parent sample of 5,549 galaxies with 0 . ≤ z ≤ . 1. Galaxies in the parent sample are marked with small black filledcircles; those hosting X-ray selected AGN are highlighted by large red circles. The slanted blue lines denote the approximate limits forgalaxies with R ap ≤ 24 at z = 0 . z = 1 (dash-dotted). The division between red and blue galaxies implemented by Bell et al.(2004) is shown by the dashed line. AGN hosts classified as bulge-dominated (S´ersic index 2.5–8) galaxies by the HST /ACS morphology(H¨aussler et al. 2007) are marked with blue filled circles at the centers of the red circles. AGN with X-ray hardness ratios larger than0.2 (see § ± σ ) bar is shown in the bottom right corner. (b)Rest-frame U − V color versus redshift for the 2,044 most luminous ( M V < − . 7) galaxies. The vertical solid lines denote the redshiftinterval 0 . ≤ z ≤ . 76 with the dotted lines marking the redshift spikes at z = 0 . 67 and z = 0 . 73. In this panel only, smaller blue dotsmark the AGN with spectroscopic redshifts. Fig. 4.— Rest-frame color ( U − V ) histogram of galaxies that host AGN (solid line) with absolute magnitude M V < − . . < z < . 63 (a), 0 . ≤ z ≤ . 76 (b), 0 . < z < . TABLE 1Statistical comparison of color distributions Note . < z < . . ≤ z ≤ . . < z < . Silverman et al.tometry and state that an induced color gradient is onlyevident at z . . 4, below the redshift range of our sam-ple, with a potential color offset of ∼ . z ∼ . 7, coupled with oursmall AGN sample and sensitivity to cumulative biasesthough our statistical test suggests otherwise. AGN FRACTION We measure the fraction of galaxies that host AGN as afunction of color and large-scale environment. We followthe technique discussed in § Chandra observations of the E-CDF-S (see Fig. 17 of Lehmer et al.2005). The necessity of this approach is demonstrated inFigure 5, which shows the limiting X-ray luminosity as afunction of redshift for the entire galaxy sample and themeasured X-ray luminosities of those galaxies harboringAGN. In addition to the Malmquist bias, there is almostan order-of-magnitude spread in limiting luminosity overthe full redshift range. To illustrate these effects, thefraction of galaxies that could host a detectable AGNwith log L X = 41 . ≈ ≈ 80% when considering a limiting luminosity of 42.5.To account for this selection effect, we determine the con-tribution of each AGN separately to the total fraction.The AGN fraction ( f ; see equation 1 below) and asso-ciated error ( σ ; see equation 2) are a sum over the fullsample of AGN ( N ) with N gal , i representing the numberof galaxies capable of hosting the i th detectable AGNwith X-ray luminosity L i X . f = N X i =1 N gal , i (1) σ ≈ N X i =1 N , i (2)We have calculated the fraction of galaxies harboringAGN in bins of color for the entire sample (Fig. 6a). Us-ing a bin width of ∆( U − V ) = 0 . 4, we find that thefraction rises from ∼ 0% at the bluest colors to ∼ U − V ∼ . U − V ∼ . ∼ ∼ − 5% at U − V ∼ . M V < − . 7) galaxieshosting AGN that we measure for various subsamples ofgalaxies. We note that the AGN fraction in the ‘field’(i.e., 0 . ≤ z < . 63 and 0 . < z ≤ . 1) is 5 . ± . . < U − V < . 0, the AGN fraction within this narrowredshift range (solid line) is 12 . ± . 9% (21 AGN, 173 Fig. 5.— The full band X-ray luminosity as a function of redshiftfor the 5,549 galaxies in our parent sample. The filled circles arethe X-ray detections (AGN); the open circles are upper limits. Thespread of a factor of six in the value of the limits at a given redshiftis produced by the variations in survey sensitivity with location;for a given redshift, the objects with the lowest limits are located atthe center of each ACIS pointing and those with the highest limitsare located at the edge of the ACIS array. The horizontal linedenotes the minimum luminosity for inclusion in our AGN sample. galaxies), that reaches ∼ 15% at U − V ∼ . 8, while thosein the ‘field’ (dotted line) have a fraction of 7 . ± . . σ ( > ∼ relative fraction as afunction of color (Fig. 6b; dotted line); the AGN fractionappears to be rather flat ( ∼ . . U − V . . 5) that includes the red sequenceand the valley but then drops off for blue galaxies beyondthe top ( U − V < . 2) of the ‘blue cloud’. DISCUSSION Global evolution of host-galaxy colors It is useful to ask what would be the observed colordistribution of the hosts of AGN as a function of red-shift up to z ∼ M B < − 21) galax-ies at z . . z & . 6, a transition occurs where the majority of lu-minous galaxies are blue (i.e., late-type, star-forming)as a result of (1) the order-of-magnitude increase inthe mean star-formation rate (e.g., Hopkins & Beacom2006), across all mass scales (e.g., Noeske et al. 2007;Zheng et al. 2007), and (2) strong depletion of galax-ies along the red sequence (Bell et al. 2004, 2007) up to z ∼ 1. In Figure 7, we show the distribution of rest-framevolution of AGN host galaxies 9 TABLE 2Fraction of luminous ( M V < − . ) galaxies hosting AGN Index Redshift a U − V S´ersic Fraction Notesrange range indices %1 I+II+III — — 6 . ± . < . . ± . . ± . > . . ± . . ± . . ± . 57 II 0.5–1.0 — 12 . ± . . ± . . ± . . ± . < . . ± . > . . ± . a I: z = 0.4–0.63; II: z = 0.63–0.76; III: z = 0.76–1.1 Fig. 6.— Fraction of galaxies ( M V < − . 7) hosting AGN asa function of their rest-frame optical color ( U − V ) for the entiresample (a). In panel b , we have split the sample into those in(solid) and out (dashed) of the redshift interval 0 . ≤ z ≤ . σ ) are shown for all values with a slight displacement ofthe dashed set for visual purposes only. A suggestive enhancement( ∼ two × with significance at the 2 . σ level) of AGN activity is ev-ident for host galaxies residing in the “green valley” that is mainlyattributed to activity in large-scale structures. U − V colors for redshift-selected galaxy populations inour sample. The blue peak ( U − V ∼ 0) of the luminous( M V < − . 7) galaxy distribution is most prominent inthe highest redshift bin ( z = 0.76–1.1; dotted histogram),while the red peak is more fully populated in the lowestredshift bin ( z = 0.4–0.63; dashed histogram). Fig. 7.— Color distribution of all luminous ( M V < − . z = 0.4–0.63; solid: z = 0.63–0.76; dotted: z = 0.76–1.1). All have been normalized bythe number of objects in each interval. We plausibly expect to see a similar redshift de-pendence in the color distribution of galaxies hostingAGN with redshift. Observational evidence does ex-ist to support this scenario. Host-galaxy studies, pi-oneered by HST , show that early-type galaxies repre-sent the majority of the hosts of QSOs at low red-shifts ( z < . 3; e.g., Bahcall et al. 1997). At similarredshifts, Kauffmann et al. (2003) demonstrate that thehosts of over 20,000 narrow line AGN from the SDSS aremainly early-type, bulge-dominated systems. The ma-jority of the most luminous, type II QSOs from the SDSS(Zakamska et al. 2006) and radio galaxies (McLure et al.2004) appear to also have similar early-type hosts. Athigher redshifts ( z > . z > . M V < − . 7) AGN host galaxies in the ‘field’ (i.e.,outside the redshift interval containing prominent red-shift spikes) appears to support the above scenario. In0 Silverman et al.Figure 4, we see that the distribution (solid histogram)has both a red (Fig. 4a) and a blue (Fig. 4c) peak forthe 0 . < z < . 63 and 0 . < z < . z . . U − V > . 7; Figure 3b) and lie on or close to the redsequence. Optical spectra are available for five of thereddest ( U − V > . 0) AGN host galaxies and each has astrong 4000 ˚A break characteristic of an old stellar pop-ulation. At higher redshifts ( z & . aremore prevalent in blue (i.e., star-forming) galaxies. Nineof these high-redshift AGN with U − V < . U − V < . 2) end. This result sug-gests that AGN activity in the ‘field’ is primarily depen-dent on mass , given the strong dependency on observedluminosity and weaker in the luminous, blue (i.e., star-forming) galaxy population. In § Enhanced AGN activity in large-scale structures The enhancement of galaxies hosting AGN, in large-scale structures reported in Gilli et al. (2003) and sub-stantiated further in this investigation, appears to in-dicate the physical scale most nuturing for accretiononto SMBHs. The fraction of galaxies hosting AGN(Fig. 6b) is significantly enhanced within a redshift in-terval (0 . ≤ z ≤ . 76) dominated by two prominentredshift spikes ( z = 0 . 67 and z = 0 . 73) that are over-populated with both AGN (Gilli et al. 2003) and galax-ies (Cimatti et al. 2002; Adami et al. 2005). The dimen-sions of these structures are estimated to be ∼ 10 Mpcin transverse extent and ∼ 37 ( z = 0 . 67) and ∼ z = 0 . 73) Mpc in depth (R. Gilli, private communica-tion). These two structures have been shown (Gilli et al. A handful appear to be associated with a less prominent red-shift spike identified by Gilli et al. (2003) at z ∼ . Using equation 2 of Bell et al. (2005), we confirm that esti-mates of stellar mass of AGN host galaxies in our sample, basedon their rest-frame colors, represent the high end of the galaxymass distribution since we find a mean of ∼ . × M ⊙ and90% of our sample with M & . × M ⊙ , in agreement withoptically-selected AGN host galaxies (Sanchez et al. 2004). Silverman et al. (2007) demonstrate that both structures doextend beyond the central 1 Ms region and provide new measuresof their angular size. z = 0 . z = 0 . 73 ’wall’ is more compact witha central dense core due to the presence of a cluster(Cimatti et al. 2002; Gilli et al. 2003), and appears tohave significant surrounding substructure (Adami et al.2005) that may eventually collapse into a massive Virgo-type cluster in the local universe. The difference ingalaxy populations within these two structures appearsto reflect their dynamic state following the well-knownSFR-density relation: galaxies in the z = 0 . 67 structure(Fig. 3b) are mainly blue while the z = 0 . 73 structure isdominated by red (i.e., evolved) galaxies.Most remarkably, the colors of the host galaxies ofthese moderate-luminosity AGN , within the redshiftinterval z = 0.63–0.76, preferentially fall in the “greenvalley” (Fig. 4b). The color distribution is asymmet-ric, with many hosts closer to the red sequence than theblue galaxy peak. As evident in Figure 3b, this colorprofile is composed of a broad distribution of galaxiesresiding in the z = 0 . 67 redshift spike and a more com-pact, red ( U − V ∼ . 8) group in the z = 0 . 73 redshiftspike. It appears that the AGN follow a SFR-densitytype relation similar to the galaxies. In Figure 9, weshow example optical spectra and HST images of fourAGN that have z = 0.63–76 and colors placing themin the “green valley”. All have no evidence for an em-bedded AGN (e.g., H β or Mg II emission lines) and arelatively mild 4000 ˚A break. The spectrum of source δ andweak [OII] emission. E+A galaxies are thought to haveundergone a starburst phase ∼ ∼ . 01 Gyr) or ongo-ing star formation. This object is only one of fifteen,over the full redshift range, that fall within the “greenvalley” and has a spectrum characteristic of an E+Agalaxy. Therefore, most of these galaxies have not hada major starburst episode within 1–2 Gyr in their past.Most spectra seem to be characteristic of a galaxy, withgradual ongoing star formation, which is best fit by theType 2 and 3 average galaxy templates shown in Figure3 of Wolf et al. (2003), effectively separating blue andred galaxies. A larger sample is required to measure sta-tistically the fraction of AGN hosts with poststarburstsignatures that have been shown to be common in AGNfrom the SDSS (Vanden Berk et al. 2006). We concludethat within this narrow redshift interval, containing twooverdense structures, the enhancement of AGN activitywithin the “green valley” signifies an important link be-tween the evolution of SMBHs and their host galaxiesthough not overwhelmingly in poststarburst systems. Optical morphology of AGN hosts It is worth noting that Georgakakis et al. (2007) recently re-ported a color-dependency for AGN residing in denser environ-ments with rest-frame colors ( U − B ∼ . 8) coincident with bluegalaxies. volution of AGN host galaxies 11 Fig. 8.— Examples of blue ( U − V < . 4) galaxies hosting AGN at z > . HST /ACS z -band images are shown with a log scaling. The optical spectra that cover a wavelength range of 6000–8500 ˚A (rightcolumn) have been binned by 2 pixels. The flux scale is erg cm − s − ˚A − . The source numbers follow those presented in the Lehmer et al.(2005) X-ray catalog. We highlight the fact that AGN hosts show signs of abimodal distribution in their colors but not in their mor-phological properties (B¨ohm et al. 2006). To illustrate,we have marked those AGN in Figure 3a that have a ra-dial surface brightness profile (i.e., S´ersic index, n ), pro-vided by H¨aussler et al. (2007), characteristic of a bulge-dominated galaxy (2 . < n < n = 1) while that of a bulge-dominated galaxy (e.g., early-type) is the r / -law (deVaucoulers profile; n = 4). We find that there are manybulge-dominated galaxies with bluer colors (i.e., ongo-ing star formation) than a typical red-sequence galaxyand there are a few disk-dominated galaxies that arered. The HST /ACS images of the red disk-dominatedgalaxies hosting AGN show that most (5 out of 7 with U − V > . 1) of them are highly-inclined or edge-on spi-rals that probably have significant dust extinction. Toinvestigate further, we plot in Figure 10, the S´ersic in-dex versus rest-frame color ( U − V ) for our luminous( M V < − . 7) galaxies and AGN. We further requirehere that the science flag provided by H¨aussler et al. (2007) is 1 for all galaxies that guarantees that GAL-FIT (Peng et al. 2002) converged; this condition is metfor 71 of the 85 luminous host galaxies with morpho-logical parameters given in H¨aussler et al. (2007). Therest-frame color distribution of this cleaned sample isequivalent to that used in previous analyses since theremoved hosts span the full range of color. Symboltypes for the AGN correspond to the three redshift in-tervals used throughout this work. First, it is evidentthat the requirement for AGN host galaxies to be bulge-dominated (e.g., Kauffmann et al. 2003) continues up to z ∼ n > . of galaxies hosting AGN as a function of n sharply rises above this value (Fig. 11a; Table 2). Second,AGN hosts preferentially become bluer with redshift butretain their bulge-dominated morphology. We concludethat the blue colors of AGN hosts up to z ∼ The method to determine the AGN fraction (Table 2; Index8-12), using our cleaned sample (science flag=1), is described in § n replacing U − V . Fig. 9.— Examples of AGN hosts with 0 . < U − V < . z ∼ . HST /ACS V -band images are shown with a log scaling. Optical spectra are described in the caption of Figure 8. Thesource numbers follow those presented in the Lehmer et al. (2005) X-ray catalog. It is clear that the fraction of bulge-dominated galaxieshosting AGN is higher for those with blue colors ( U − V < . § 5, and find that 21 . ± . 0% of the galaxies with n > . U − V < . . ± . § n < . 5) is much lower (1 . ± . HST /ACS colorimages of six AGN host galaxies in the GOODS areawith n > . U − V color. Clearly,the most prominent feature is the bright bulge for allhost galaxies. The bluest host galaxy J033213.2 − − − − n < . 5) AGN host galaxies.Bright, star-forming regions are ubiquitous in the formof knots that appear to lie within spiral arms in mostexamples.We remark that a more thorough analysis of the mor-phological structures of these host galaxies and their re-lation to the general galaxy population is required andbeyond the scope of this paper. We highlight the fol-lowing complications that are relevant for interpretingthese results and will be taken into consideration in afuture analysis: (1) S´ersic fits using de Vaucoulers pro-files have been shown to have significant uncertainties(e.g., H¨aussler et al. 2007), (2) the presence of an em-volution of AGN host galaxies 13 Fig. 10.— Morphology–color relation of luminous ( M V < − . z = 0.4–0.63 (blue circles), z = 0.63–0.76 (red boxes), z = 0.76–1.1(green triangles). Errors (1 σ ) associated with n are those reportedin H¨aussler et al. (2007). The bar in the upper left corner is themean ± σ error on U − V for the AGN host galaxies. The horizon-tal line divides those galaxies that have either a disk-dominated( n < . 5) or bulge-dominated ( n > . 5) morphology as definedin Blanton et al. (2003). The vertical lines highlight the “greenvalley”. bedded, optically-faint AGN is likely to affect the mea-sure of structural parameters, such as n , (3) accuratediscrimination between bulge and disk-dominated galax-ies is difficult since S´ersic indices are known to have sig-nificant overlap for these populations (e.g., Sargent et al.2006), and (4) one must consider redshift-dependent sur-face brightness effects. A fair number of AGN host galax-ies have intermediate S´ersic indices (1 . < n < 3; Fig-ure 10) and evidence of faint disks as described above. Acareful consideration of these issue will enable us to assessthe dependence of host-galaxy morphology with environ-ment (Figure 11b). There is suggestive evidence thatS´ersic indices of hosts within large-scale structures (Fig-ure 10; red squares) are shifted to lower values ( n ∼ Are AGN driving the evolution of the generalgalaxy population? Numerical simulations (Springel et al. 2005b) demon-strate that accreting SMBHs may be a necessary ingre-dient in the formation of massive elliptical galaxies fromthe mergers of gas-rich spiral galaxies. AGN feedbackcan potentially suppress star formation (Di Matteo et al.2005) and drive (i.e., accelerate) galaxies onto the red se-quence (Croton et al. 2006; Hopkins et al. 2006a). Theshort timescales ( < Fig. 11.— Fraction of galaxies ( M V < − . 7) hosting AGN asa function of their morphology (i.e., S´ersic index n ) for the entiresample (a). In panel b , we have split the sample into those in (solid)and out (dashed) of the redshift interval 0 . ≤ z ≤ . 76 as done inFigure 6. AGN clearly reside in bulge-dominated (Panel a ; n & b ; solid line)possibly having a higher fraction with faint disks that effectivelysoftens their S´ersic index ( n ∼ galaxies with intermediate colors are thought to be evolv-ing rapidly. Strong observational evidence supportingthis scenario has been elusive. We now explore whetherour AGN sample with well determined host colors andluminosities offers further insight into the AGN-galaxyconnection.The color distribution of AGN host galaxies in ourstudy further substantiates the aformentioned model ofgalaxy evolution. The fraction of galaxies harboringAGN is significantly enhanced (Fig. 6) in the “greenvalley”, a region on the color-magnitude diagram wheregalaxies are thought to migrate from blue to red. Thishigher incidence is associated with galaxies residing in aredshift interval 0 . ≤ z ≤ . 76 dominated by two large-scale structures that are overdense and have an ongoingassembly of substructure (i.e., galaxy groups and clus-ters). Within this same redshift interval, the color dis-tribution of the underlying galaxy population (Fig. 7) hasrapidly evolved as shown by (1) a significant reduction inthe numbers of blue galaxies, a dominant population athigher redshifts ( z > . . < z ≤ . 1; dottedline). This narrow redshift interval (0 . ≤ z ≤ . Fig. 12.— Color HST /ACS postage-stamp images of bulge-dominated ( n > . 5) host galaxies located in the GOODS region. Colorscorrespond to ACS B (blue), V (green), and z (red) bandpass images. In each postage-stamp image, we indicate the source name(top), the redshift and rest-frame U − V color of the galaxy (lower right), and a vertical line of length 1 . ′′ for scaling reference. The imageshave been sorted by rest-frame U − V color such that the bluest source is located in the upper left panel and the reddest source is shownin the lower right panel. with an elapsed time of 0.73 Gyr, offers a compressedwindow of the more passive galaxy evolution that occursover longer timescales of ∼ . < z < . 1) as illus-trated in Figure 3b.In addition, numerical simulations demonstrate thatthe color evolution of a merger event slows upon ap-proach to the red sequence (Springel et al. 2005a). InFigure 14, we show the color-magnitude relation forgalaxies and AGN within the redshift interval 0 . ≤ z ≤ . 76 with the evolutionary tracks of mergers, includingSMBH feedback, with virial velocity v vir =113, 160, 226,and 320 km s − overplotted (Springel et al. 2005a). Thefact that these models do not account for dust redden-ing should not severely impact our subsequent findings.Here, we use SDSS photometric bands ( u , r ) to utilizethese models and have converted the photometry to theVega system. To account roughly for redshift evolutionfrom z ∼ . M r by − M ⋆B in theliterature (see Table 5 of Faber et al. 2006). The evolu-tionary tracks have ages up to ∼ with the first It is just a coincidence and most likely of no physical signifi- data point corresponding to 1 Gyr after the initial star-burst phase. Based on these model curves, over 1 Gyr haselapsed since the starburst phase concluded for almost allAGN hosts. The majority of AGN hosts cover a broadrange of age between 1–4 Gyr and virial velocity between113 and 226 km s − . As shown in Figure 14, a largefraction of the merger sequence ( t ≈ t . cance that these large-scale structures in the CDF-S are present at z ∼ . 7, a redshift singled out by Springel et al. (2005a), as a pos-sible formation epoch of elliptical galaxies based an elapsed timeof 5.5 Gyr to complete a merger sequence. volution of AGN host galaxies 15 Fig. 13.— Color HST /ACS postage-stamp images of disk-dominated ( n < . 5) host galaxies located in the GEMS area. Colors correspondto ACS V (blue), ( V + z )/2 (green) and z (red) bandpass images. The labels are described in Figure 12. van Dokkum et al. 2005; Bell et al. 2006) in rich envi-ronments that can effectively move them to redder colorsand high luminosities/masses since there are few massive( v vir > 226 km s − ) and luminous progenitors that couldpopulate the luminous end ( M r . − . 5) of the red se-quence. We conclude that the color distribution of AGNhosts further substantiates a coevolution scenario due tomergers and interactions that are effectively nurtured in ∼ 10 Mpc scale structures.The role of major mergers in triggering AGN hasmet observational scrutiny even up to z ∼ 1, wherethe merger rate is expected to be higher. Morpho-logical studies (Grogin et al. 2005; Pierce et al. 2007)have yet to find X-ray selected AGN at z > . HR = ( H − S ) / ( H + S ) indicative of X-ray absorption( HR > − . 2; Γ = 1 . N H = 10 cm − at z = 0).The hardness ratio is a measure of the relative numbersof observed X-ray counts in the soft ( S ; 0.5–2.0 keV)and hard ( H ; 2–8 keV) energy bands. We see that hardsources are present over a wide range of rest-frame colors. This is not unexpected since AGN are known to usu-ally have parsec-scale molecular tori (Antonucci 1993)with substantial absorbing columns that bear no rela-tion to the presence of star formation. The lack of merg-ers may reflect the limited sample used to date, since apair of galaxies undergoing a major merger at z ∼ . HST for less than ∼ v vir > 160 km s − ) galaxies on a timescale of ∼ v vir > 160 kms − and τ < . ≤ z ≤ . 67 with M r = − . 47 and u − r = 0 . 32. In Figure 14 b , we showthe HST V -band image that exemplifies a complexmorphology in the nuclear region and tidal features onscales of ∼ 10 kpc (1 ′′ = 7.1 kpc at z = 0 . 69) that mostlikely arise from a major merger of massive galaxies. Asecond example (J033213.2 − HST surveys of deep extra-galactic fields, such as COSMOS (Scoville et al. 2006)6 Silverman et al. Fig. 14.— (a) Color-magnitude relation of galaxies (0 . ≤ z ≤ . 76) associated with large-scale structures and AGN-influencedmodel evolutionary tracks from Springel et al. (2005a). Galaxieshosting AGN are further marked by an open circle. Theoreticaldata for merging galaxies is overplotted for galaxies with virialvelocities as shown. The large filled circles start from 1 Gyr afterthe inital starburst phase and each spacing corresponds to 0.5 Gyr.(b) HST V -band image of source z phot = 0 . L X =1 . × erg s − ) that is the only AGN host galaxy shown above( M r = − . u − r = 0 . n = 3 . ± . with an area coverage of 1.8 deg (5.4 × the area of theE-CDF-S) will provide improved statistics to assess ade-quately the role of mergers in triggering SMBH accretion.Finally, we note an alternative to the merger scenarioin which galaxies may be more favorable to AGN activitysimply due to the presence of a massive bulge and diskthat provide the two required ingredients (i.e., a SMBHand a reservoir of gas for accretion; Kauffmann et al.2006). This is possible since a high fraction ( ∼ n > . U − V < . 7) host moderate-luminosity AGN and there is a slight enhancement ofAGN activity in the ‘field’ for galaxies in the “green val-ley” (Fig. 6b; dotted line). HST images of host galaxiesresiding within the “green valley” (Fig. 9) appear to havebulges and faint disks. If this were the case, however, we would then not expect such a strong increase in the AGNfraction as a function of environment. SUMMARY We identified a sample of 109 X-ray selected AGNin the E-CDF-S with moderate luminosities (41 . ≤ log L . − . ≤ . 7) to investigate the rest-frame col-ors of their host galaxies. These AGN have been selectedfrom a parent sample of 5,549 galaxies from COMBO-17and GEMS with 0 . ≤ z ≤ . 1. Optical spectra areavailable for 48% of the sample; these provide assuranceof the accuracy of the photometric redshifts and showthat no strong AGN signatures are present, confirmingthe stellar nature of their rest-frame colors.We find that the broad distribution of host-galaxycolors of moderate-luminosity AGN is due to both (1)the strong color evolution of the underlying luminous( M V < − . 7) and bulge-dominated ( n > . 5) galaxypopulation and (2) an enhancement of AGN activity inlarge-scale structures. We draw the following three mainpoints: • The host galaxies of X-ray selected AGN have acolor bimodality when excluding a redshift interval0 . ≤ z ≤ . 76, which contains two redshift spikesat z = 0 . 67 and z = 0 . 73. Galaxies hosting AGNat z . . z & . 8, a distinct, bluepopulation of host galaxies is prevalent with colorssimilar to the star-forming galaxies. • The fraction of galaxies hosting AGN has a promi-nent peak in the “green valley” that is primarilydue to enhanced AGN activity in large-scale struc-tures. Within the redshift interval 0 . ≤ z ≤ . ∼ 15% at U − V ∼ . 8. Over the color interval 0 . < U − V < . . ± . . σ ) higher than that measured in the‘field’ (i.e., over all other redshifts; 7 . ± . • We find that AGN continue to preferentially residein luminous bulges up to z ∼ 1. A large fraction(75%) of AGN host galaxies with M V < − . n > . 5, even those at z ∼ n > . U − V < . 7) are mosthospitable for AGN activity based on a measuredAGN fraction of 21 . ± . 0% (four times the fractionof the ‘field’ sample).The overabundance of AGN associated with the red-shift spikes found in the E-CDF-S and their special lo-cation in the color-magnitude relation highlight the im-portance of environment, on large scales ( ∼ 10 Mpc),to influence the evolution of AGN and their host galax-ies. The richness of these structures (i.e., galaxy over-densities, group/clusters in early stages of formation) al-ludes to mergers as a dominant mechanism to triggerAGN activity, quench star formation and to drive sub-sequent migration of galaxies from the blue cloud to thered sequence. We compare the color-magnitude relationof our AGN host galaxies with evolutionary tracks ofmerging galaxies from Springel et al. (2005a) that incor-porate AGN feedback. Our AGN host galaxies have col-ors and morphologies (i.e., bulge-dominated) indicativevolution of AGN host galaxies 17of evolved systems that had undergone a starburst phase ≈ HST imagingsurveys such as COSMOS will statistically sample theserare events with timescales less than 1 Gyr.The E-CDF-S is a unique survey field with a fortunatealignment of large-scale structures for such studies. Ourfindings exemplify the complexities that must be disen-tangled to determine underlying relationships betweenAGN and their host galaxies. Much optical and near-infrared followup is forthcoming in the E-CDF-S thatwill further our understanding of the connection betweencoevolution of AGN and galaxies.We are especially grateful to the referee for providinginsightful comments that strengthened the overall con- tent of this work. We also thank L. Guzzo, K. Iwa-sawa, A. Merloni, K. Nandra, and I. Strateva for help-ful discussions and suggestions. We also recognize thecontribution of the GEMS group that provided an earlyversion of their catalog, V. Springel for supplying themodel galaxy tracks, and C. Wolf for computing up-dated photometry. Support for this work was providedby NASA through Chandra Award Number G04-5157A(BDL, WNB, DPS). RG, CV and PT acknowledge par-tial support by the Italian Space agency under the con-tract ASI–INAF I/023/05/0.”Some of the data presented in this paper were ob-tained from the Multimission Archive at the Space Tele-scope Science Institute (MAST). STScI is operated bythe Association of Universities for Research in Astron-omy, Inc., under NASA contract NAS5-26555. Supportfor MAST for non- HST data is provided by the NASAOffice of Space Science via grant NAG5-7584 and byother grants and contracts.”Facilities: VLT(VIMOS). REFERENCESAdami, C. et al. 2005, A&A, 805, 818Alexander, D. M., Smail, I., Bauer, F. E., Chapman, S. C., Blain,A. W., Brandt, W. N., Ivison, R. J. 2005, Nature, 434, 7034Antonucci, R. 1993, ARA&A, 31, 473Barger, A.J. et al. 2003, AJ, 126,632Bahcall, J. N., Kirhakos, S., Saxe, D. H. 1997, ApJ, 479, 642Balogh, M. L., Baldry, I. K., Nichol, R., Miller, C., Bower, R.,Glazebrook, K. 2004, ApJ, 615, 101Baldry, I. K., Balogh, M. L., Bower, R., Glazebrook, K., Nichol, R.C., Bamford, S. P., Budavari, T. 2006, MNRAS, 373, 469Bauer, F.E., Alexander, D.M., Brandt, W.N., Hornschemeier, A.E.,Vignali, C., Garmire, G., Schneider, D. P. 2002, AJ, 124, 2351Beckwith, S. V. W., et al. 2006, AJ, 132, 1729Bell, E. F., et al. 2004, ApJ, 608, 752Bell, E. F., et al. 2005, ApJ, 625, 23Bell, E. F., et al. 2006, ApJ, 640, 241Bell, E. F., et al. 2007, ApJ, 663, 834Best, P. N., Kauffmann, G., Heckman, T. M., Brinchmann, J.,Charlot, S., Ivezic, Z., White, S. D. M. 2005, MNRAS, 362, 25B¨ohm, A., Wisotzki and the GEMS team 2006, ASP conferenceSeries, astro-ph/0612310Boyle, B. J. & Terlevich, R. J. 1998, MNRAS, 293, L49Blake, C. et al. 2004, MNRAS, 355, 713Blanton, M. R., et al. 2003, ApJ, 594, 186Brandt, W. N., Hasinger, G. 2005, ARA&A, 43, 827Cassata, P. et al. 2007, ApJS, in press, astro-ph/0701483Caldwell, J. A. R. et al. 2005, submitted to ApJS, astro-ph/0510782Cimatti, A. et al. A&A, 392, 395Croton, D. J. et al. 2006, MNRAS, 365, 11Di Matteo, T., Springel, V., Hernquist, L. 2005, Nature, 433, 604Dressler, A., Gunn, J. E. 1983, ApJ, 270, 7Eastman, J., Martini, P., Sivakoff, G., Kelson, D. D., Mulchaey, J.,Tran, K.-V. 2007, ApJ, in press, arXiv:0706.0209Faber, S. M. et al. 2006, ApJ, in press, astro-ph/0506044Ferrarese, L., Merritt, D. 2000, ApJ, 539, L9Franceschini, A., Hasinger, G., Miyaji, T., Malquori, D. 1999,MNRAS, 310, L5Gavignaud, I. et al. 2006, A&A, 457, 79Gebhardt, K. et al. 2000, ApJ, 539, L13Georgakakis, A. et al. 2006, ApJ, 660, L15Giacconi, R., et al. 2002, ApJS, 139, 369Giavalisco, M., et al. 2004, ApJ, 600, L93Gilli, R. et al. 2003, ApJ, 592, 721Grogin, N. A. et al. 2005, ApJ, 627, 97H¨aussler, B. et al. 2007, ApJS, in press, arXiv:0704.2601V2Hopkins, A. M., Beacom, J. F. 2006, ApJ, 651, 142Hopkins, P. F. et al. 2005, ApJ, 630, 705Hopkins, P. F. et al. 2006, ApJS, 163, 1Hopkins, P. F., Somerville, R. S., Hernquist, L., Robertson, B., Li,Y. 2006, ApJ, 652, 864 Jahnke, K., Kuhlbrodt, B., Wisotzki, L., MNRAS, 352, 399Jahnke, K. et al. 2004b, ApJ, 614, 568Kartaltepe, J. S. et al. 2007, ApJS, in press, arXiv:0705.2266Kauffmann, G. et al. 2003, MNRAS, 346, 1055Kauffmann, G. et al. 2004, MNRAS, 353, 713Kauffmann, G. et al. 2007, MNRAS, in press, astro-ph/0609436Li, C., Kauffmann, G., Wang, L., White, S. D. M., Heckman, T.,Jing, Y. P. 2006, MNRAS, 373, 457Lehmer, B. D. et al. 2005, ApJS, 161, 21Lehmer, B. D. et al. 2007, ApJ, 657, 681Le Fevre, O., Vettolani, G., Maccagni, D., and the VVDSconsortium, 2003, The Messenger 111, 18Le Fevre, O., et al. 2004, A&A, 428, 1043Martini, P., Kelson, D. D., Kim, E., Mulchaey, J., Athey, A. 2006,ApJ, 644, 116Martini, Mulchaey, J., Kelson, D. D. 2007, ApJ, 664, 761McIntosh, D.H. et al. 2005, ApJ, 632, 191McLure, R. J. et al. 2004, MNRAS, 351, 347Merloni, A., Rudnick, G., Di Matteo, T. MNRAS, 354, 37Mignoli, M. et al. 2005, A&A, 437, 883Miller, C. J., Nichol, R. C., Gomez, P.L., Hopkins, A. M., Bernardi,M 2003, ApJ, 597, 142Nandra, K. et al. 2006, ApJ, 660, L11Noeske, K. G. et al. 2007, ApJ, 660, L43Peng, C.Y., Ho, L.C., Impey, C. D., Rix, H.-W. 2002, AJ, 124, 266Pierce, C. et al. 2006, ApJ, 660, L19Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P.1993, Numerical Recipes in C: The Art of Scientific Computing,2nd edn.(Cambridge: Cambridge University Press)Rix, H. -W. et al 2004, ApJS, 152, 163Rovilos, E., Georgakakis, Georgantopoulos, I., Koekemoer, A.,Mobasher, B., Goudis, C. 2007, A&A, 466, 119Sanchez, S. F. et al. 2004, ApJ, 614, 586Sanders, D. B., Mirabel, I.F. 1996, ARA&A, 34, 749Sargent, M T. et al. 2006, ApJS, in press, astro-ph/0609042Scoville, N. et al. ApJS, in press, arXiv:asto-ph/0612306Silverman, J. D., et al. 2005, ApJ, 618, 123Springel, V., Di Matteo, T., Hernquist, L. 2005a, ApJ, 620, L79Springel, V., Di Matteo, T., Hernquist, L. 2005b, MNRAS, 339,289Strateva, I. et al. 2001, AJ, 122, 1861Sturm, E. et al. 2006, ApJ, 642, 81Szokoly, G. P., et al. 2004, ApJS, 155, 271Tremaine, S. et al. 2002, ApJ, 574, 740Vanden Berk, D. E. et al. 2006, AJ, 131,84Vanzella, E. et al. 2005, A&A, 434, 53Vanzella, E. et al. 2006, A&A, 454, 423van Dokkum, P. G. 2005, ApJ, 130, 2647Wisotzki, L., Jahnke, K., Sanchez, S. F., Schramm, M., 2006,NewAR, 50, 8298 Silverman et al.