MOIRCS Deep Survey III: Active Galactic Nuclei in Massive Galaxies at z=2-4
T. Yamada, M. Kajisawa, M. Akiyama, T. Ichikawa, M. Konishi, T. Nishimura, K. Omata, R. Suzuki, I. Tanaka, C. Tokoku, Y. K. Uchimoto, T. Yoshikawa
aa r X i v : . [ a s t r o - ph . C O ] M a y MOIRCS Deep Survey III: Active Galactic Nuclei in MassiveGalaxies at z = 2 − T. Yamada, M. Kajisawa, M. Akiyama, T. Ichikawa, M. Konishi, T. Nishimura, K. Omata, R. Suzuki, I. Tanaka, C. Tokoku, Y. K. Uchimoto, T. Yoshikawa, , [email protected] ABSTRACT
We investigate the X-ray properties of the K -band-selected galaxies at redshift2 < z < Chandra
X-ray source catalog. 61 X-ray sourceswith the 2-10 keV luminosity L X = 10 − erg s − are identified with the K -selected galaxies and we found that they are exclusively (90%) associated withthe massive objects with stellar mass larger than 10 . M ⊙ . Our results areconsistent with the idea that the M BH /M str ratio of the galaxies at z = 2 − M ⊙ ishigh, 33% (26/78). They are active objects in the sense that the black-hole massaccretion rate is ≈ BH /M str ratio with those observed in the local universe. The active duration inthe AGN duty cycle of the high-redshift massive galaxies seems large. Subject headings: galaxies: active — galaxies: formation — galaxies: evolution
1. Introduction
It is now widely accepted that the most of the massive galaxies have their central super-massive black holes (SMBHs) (e.g., Kormendy & Richstone 1995). The tight correlation Astronomical Institute, Tohoku University, Aramaki, Aoba-ku, Sendai, Miyagi, 980-8578 Subaru Telescope, National Astronomical Observatory of Japan, 650 North A’ohoku Place, Hilo, HI96720, U.S.A Institute of Astronomy, University of Tokyo, 2-21-1 Osawa, Mitaka, Tokyo, 181-0015 BH , and the spheroid mass, M sph , or the velocity dispersion ofthe host is observed for both normal and active galaxies in the nearby universe (Magorrian et al. et al. BH -M sph correlation, it is needed to study the relationshipamong stellar mass, star-formation, and active galactic nuclei (AGN) activities of the galaxiesat high-redshift when they are majorly developing their stars, structures, and central blackholes. It is also very interesting to see whether such relationship has always been hold ornot during the hierarchical growth of massive galaxies (Haehnelt & Rees 1993; Haehnelt &Kauffmann 2000; Granato et al. et al. z =0, AGN are observedexclusively in the massive galaxies with the stellar mass larger than ∼ M ⊙ (Kauffmann et al. BH /M sph ratio increases with increasing redshift (McLure et al. BH -M sph correlation at high redshift. Thetechniques to evaluate the BH mass of normal galaxies cannot be used for the distant objects.For active galaxies at high redshift, massive BHs are likely to have been formed in luminousquasars (Turner 1991) and their virial mass can be measured by using the velocity width andthe size of the broad-line regions estimated from the quasar luminosity (Kaspi et al. -10 M ⊙ for the luminous quasars at z > et al. et al. et al. (2006) analyzed the sample of 3C RRradio-loud objects and argue that the M BH -M sph ratio is larger at high redshift, ∼ × of 3 –the local at z ∼
2. The line-width of the emission lines in radio-loud objects, however, maybe affected by the gas kinematics due to the radio jet activity, which may cause possibleuncertainty in BH mass estimation.Molecular gas, which is related to star-formation but not directly to the BH activity,is a reliable probe to see the host galaxies of luminous quasars (Yamada 1994, Miolino et al. > M ⊙ ) host galaxies ofthe quasars at z =2-6 were reported (e.g., Ohta et al. et al. etal. (2005; 2006) have revealed that ∼
75% of the sub-mm-selected galaxies (SMGs) showAGN activities and the majority of the active nuclei seem to be heavily obscured by dust.Daddi et al. (2007) also studied the mid-infrared excess galaxies at z ∼ BH -M sph ratioin quasar and ULIRG phase (Alexander et al. z = 2-4 using the new extremely deep near-infrared (NIR) imaging data obtainedin the MOIRCS Deep Survey (MODS, see Section 2) to investigate the AGN activity amongthe stellar-mass based sample. Previously, only the X-ray fraction among the NIR-selectedsources at z ∼ et al. et al. z = 2 −
4, the X-ray sources detected even in the deepestobservation seems dominated by AGN population with the X-ray luminosity L X (2-10 keV)=10 -10 erg s − , which corresponds to the luminosity of the luminous Seyfert galaxies inthe local universe (Green et al. BH /M sph ratio, wemay discuss the AGN activity along the stellar mass for the galaxies at such high redshift.In Section 2, we introduce the NIR and the X-ray data and the sample we use, anddescribe the method to obtain the various properties of the galaxies, such as stellar mass,reddening, star-formation rate. We show the X-ray properties of these samples in Section 3and discuss the AGN activity among the high-redshift galaxies in Section 4. The cosmologicalparameters used here are canonical values after WMAP, namely, Ω =0.3, Ω Λ = 0 .
7, and 4 – H =70 kms − Mpc − . NIR magnitude values are in Vega system (Oke & Gunn 1983; Fukugitaet al 1996).
2. Data and Samples
In order to study the relationship between AGN activities and galaxy properties at highredshift, we constructed a sample of K -band selected galaxies between z =2 and 4 from theMOIRCS Deep Survey (MODS, Kajisawa et al. et al. et al. et al. J , H , and Ks band images at the sky area in the fields of Great Observatories Origins Deep Surveyfield North (GOODS-N). We cover the area by the four fields of views (FoV) of MOIRCS,referred as GT1, GT2, GT3, and GT4, from north-east to south west. Full description ofMODS result will be published in the separated papers (Kajisawa et al. in preparation).GT2, which includes the Hubble Deep Field North (HDF-N) is also the ultra-deep field ofMODS survey where more than 28h science integrations were obtained in both J and Ks bands. The total sky area (GT1-GT4) used in the analysis of this paper is 103.3 arcmin .The basic properties of the data obtained and reduced by June, 2008 are summarized inTable 1.In MODS, more than 7000 K -band-selected galaxies are detected down to K ∼ . σ )at GT1-GT4 and 24.6 in the deepest GT2 field. The MOIRCS images are carefully alignedto the optical images obtained by Advanced Camera for Survey (ACS) on Hubble SpaceTelescope (HST), as well as the infrared images taken with IRAC and MIPS instruments onSpitzer Space Telescope. ACS and MOIRCS J and H images are smoothed to be matchedwith the Ks -band images and the colors of the objects are obtained by the apertures withthe diameter of 1.2 arcsec. For the IRAC data, we first measured the flux in apertures in3.16, 2.78, 3.43, and 3.47-arcsec diameter in 3.6, 4.5, 5.8, and 8 µ m, and then made aperturecorrection by using the light profile of the Ks -band image smoothed to be matched with eachIRAC band. The multi-band photometric data are then fitted by the galaxy evolutionarymodels of Bruzual & Charlot (2003) so that we obtained the photometric redshift, stellarmass, reddening, and parameters of the star-formation histories such as ages, star-formation-rate, and the decaying parameter τ by the similar method described in Kajisawa & Yamada(2005). Kajisawa & Yamada (2005) carefully investigated the uncertainty in the evaluation ofthe stellar mass and found that the stellar mass of the most of the K -selected galaxies can bedetermined well for the given SED while there may be degeneracy in other parameters such asage, extinction, and star-formation-decaying time scale. Kajisawa et al. (in preparation) also 5 –Table 1: Summary of the NIR observationField Filter FWHM (ch1) Depth (ch1) a FWHM (ch2) Depth (ch2) a Exp. Timearcsec mag. arcsec mag. hoursGT1 J 0.59 24.9 0.59 24.8 8.0H 0.58 23.8 0.59 23.7 2.5K 0.58 23.7 0.53 23.8 8.3GT2 J 0.48 25.7 0.49 25.6 28.2H 0.46 24.4 0.46 24.2 5.7K 0.45 24.7 0.46 24.6 28.0GT3 J 0.57 24.9 0.58 24.8 6.3H 0.55 23.7 0.55 23.7 3.2K 0.59 23.8 0.60 23.7 10.7GT4 J 0.58 24.8 0.59 24.8 9.1H 0.58 23.8 0.59 23.8 4.3K 0.59 23.6 0.60 23.7 9.8 a The 3 σ limits given in Vega magnitude for the aperture with the diameter of 2xFWHM of the averagepoint source profile. As MOIRCS has the two equivalent optical channels, we obtained the detection limitsseparately for each channel. σ uncertainty is 0.1 dex for M str =10 . − M ⊙ and 0.2-0.3 dex for M str =10 . − . M ⊙ .In this paper, we used the models with Salpeter initial mass function (IMF). Reduction bythe factor of two in the stellar mass values is needed to compare with those obtained withthe Chabrier-like IMF (Chabrier et al. et al. (2003)who conducted 2M-sec observations with the X-ray satellite Chandra at the field. The on-axis sensitivity limit is ≈ . × − erg s − at 0.5-2 keV (soft) band, ≈ . × − ergs − at the 2-8 keV (hard) band, and ≈ . × − erg s − at the 0.5-8 keV (total) band,respectively. Their exposure maps show that the data is fairly homogeneous over the MODSGT1-GT4 fields. The median positional determination accuracy is ≈ . et al. et al. Ks -band sources by the following procedure. We first identify the pixelon the Ks -band image at the X-ray source position in the catalog and then find the Ks -bandselected object to which the pixel belongs. If there is no Ks -band object at the pixel, wethen extend the search pixels up to three (0.35 arcsec) to find the Ks -band counterpart. Infact, the most of the Chandra sources can be identified with the Ks -band sources in thefirst step thanks to the good positional accuracy of the Chandra data and the depth of theinfrared data. While the depth of NIR images at the field GT2 is significantly deeper thanother fields, most of the X-ray sources but a few are identified with the Ks -band sourcesabove K ∼ K -selected sample within the redshift range, 2 < z < Ks -band light samples ∼ − et al. et al. et al. K =23, which is the conservative completeness limit overGT1-GT4. 79 objects have the stellar mass larger than 10 M ⊙ and 219 objects have 10 . M ⊙ < M str < M ⊙ . The accuracy of the photometric redshift, evaluated by using thespectroscopic sample, is δz /(1 + z )=-0.009 ± k δz /(1 + z ) k > . δz /(1+ z )=-0.071 ± K -selected galaxies and thoseassociated with the X-ray sources, respectively. The redshift distribution at 2 < z <
3. Results
Fig.2 shows the stellar mass versus J − K color of the 2621 objects. Galaxies with K <
23, above the completeness limit over GT1-GT4, are shown by the large dots and thefainter ones are by the small dots. As previously pointed out (e.g., Kajisawa & Yamada2005; 2006), the most massive galaxies tend to have the redder colors. At M str > × M ⊙ , more than 80% of the galaxies have J − K > K -band sources.We found that 21 of 61, or 19 of 52 objects have the spectroscopic redshift values between z =2 and 4. Since the majority of the X-ray sources are detected in the hard energy band, 2-8keV, which corresponds to 6-24 keV and 10-40 keV at z =2 and 4, respectively, these X-raysources are likely to be AGNs. Indeed, their luminosity range, 10 − erg s − , is consistentwith those of the luminous Seyfert galaxies or faint quasars in nearby universe. While activestar-formation also produces luminous X-ray emission, the expected luminosity is more thanan order of magnitude lower than the measured values. The empirical fit by Ranalli et al. (2003) predicts L X (2-10 keV) = 5 . × erg s − for the star-formation rate (SFR) of 1 M ⊙ yr − . As shown below, the largest SFR estimated for the current sample is 1000 M ⊙ yr − that may produce the X-ray luminosity of 5 . × erg s − , which is about the lower boundof our sample (see Fig.8 below). We conclude that the observed X-ray emission is associatedwith the AGN activity without serious contamination by star-formation.We first note that the most of the identified X-ray sources are associated with massivegalaxies. Among the 61 total-band sources, 90% (55/61) are associated with the galaxieswith M str > . M ⊙ and 48% (29/61) are those with M str > M ⊙ . Table 2 alsosummarizes these numbers. On the other hand, for the massive galaxies with M str > M ⊙ , the fraction of the X-ray detection is notably high, 33% (26/78) and 29% (23/78) forthe total-band and the hard-band sources, respectively. These values can be interpreted asthe lower limit as we ignored the attenuation to the X-ray emission; we may missed veryCompton-thick objects with absorption larger than N H ≈ cm − .We checked by the several methods that the NIR light of the objects is indeed stellarorigins and not heavily contaminated by the AGN component and the evaluation of thestellar mass based on NIR luminosity is not significantly affected by the AGN light. If weadopt the typical X-ray to optical flux-density (per unit frequency) ratio for broad-line AGN, f ν (20 keV)/ f ν (0.5 µ m) ∼ − -10 − (Richards et al. Ks -band, namely rest-frame optical band, is estimated to be 19.5-22 mag for the sourceswith f X (2-8 keV)=10 − erg s − cm − without extinction at the optical wavelength. While 8 –the most of the X-ray sources in our sample are fainter than this flux range and their host-galaxy K -band magnitudes are in the range of K =19-22.5 mag, the contribution of the AGNlight in K band may not be negligible for more than a few cases.We therefore carefully examined the colors and the optical-mid-infrared spectral energydistributions (SED) of the K -band sources which have the X-ray counterpart. Donley et al. (2007) examined the properties of the Spitzer IRAC sources in GOODS-N which have ’power-law’ SEDs and found about a half of them are detected in X-ray to show the AGN activity.We checked how many K -selected sources may be classified with the ’power-law’ sources andfound that 13 K -selected sources are matched with such SED in Donley et al. (2007) andof them, 9 are detected by Chandra . Other four non-Xray objects may host hidden veryCompton-thick AGN, which make the AGN fraction of the sample even larger. Thus about15% (9/61) of the X-ray-identified K -selected galaxies the NIR light may be contaminatedby the hot-dust component. In our own photometry, 7 of the 9 ’power-law’ sources (detectedin X-ray) indeed have the SEDs monotonically increased in AB magnitude toward the longerwavelength. However, the rest (85%, 52/61) of the sample can be well fitted by the stellar-population SEDs showing the turn off near the rest-frame ’1.6 µ m bump’, which supportsthat our stellar-mass evaluation for these objects is robust. Furthermore, even for the 9’power-law’ sources, we found that our stellar-mass estimation may be reasonable. We madethe similar SED fitting process with regarding the K and the IRAC magnitudes as the upperlimits, allowing the contamination by the AGN light, but the evaluated stellar mass valuesdo not change very much with a few exceptions. Fig.3 compare the stellar mass obtainedby the SED fitting to the all measured photometric values and those with the upper-limitsubstitution in K and the IRAC bands. We also show the three examples of the SED fittingresults in Fig.4a-4c. Fig.4a shows the very typical case which has the middle-range χ valuein the fitting among the sample. The magenta line is the best-fitted models using the allphotometric data points and the blue line is that using the K and IRAC-band data (opencircles) as the upper limit values. In this case, clear turn off around the 5.8 µ m-band (IRACch3) data, which corresponds to the 1.6 µ m-bump feature in the stellar light, is appeared.Fig.4b shows the case of a relatively red objects, which has the χ value at ∼
16% (10thamong the 61 objects) from the worst. While the 8 µ m data is deviated from the stellarmodel, which shows the evidence of hot dust component, the data at the wavelength shorterthan 5.8 µ m are well fitted by the stellar model including the Balmer jump feature near the H band. In this case, the difference in stellar mass from the two different fitting proceduresis ∼
50% , which is one of the largest deviation among the sample (see Fig.3). Fig.4c showsthe case of a ’power-law’ object, which indeed shows the monotonic rising of the flux towardthe longer wavelength. Even in this case, we see clear jump of the SED below the J band,which is very likely to the Balmer jump in the stellar spectrum. The obtained stellar mass is 9 –almost same for the two different fitting procedures while the other parameters of the models(age, extinction) are different.We also found that the Ks -band light distributions show the extended profile and notdominated by the point source of the AGN components. Fig. 5 shows the FWHM of the Ks -band images of the 61 X-ray sources. Majority ( > ∼ Ks -band light distributions are also extendedin many cases. From these above, we conclude that the observed NIR light are not seriouslycontaminated by the AGN light and the obtained stellar mass values are as robust as for thenon-X-ray galaxies.Fig.6 shows the X-ray luminosities of the sources as the function of the stellar mass ofthe host galaxies. We obtained the X-ray luminosity at 2-10 keV from the hard-band fluxadopting the relation f ν ∼ ν − . to obtain the K-correction factor. While the soft band(0.5-2 keV) is closer to the rest-frame 2-10 keV energy range for the objects at z = 2 −
4, thehard-band flux is less affected by the extinction and we may obtain the X-ray luminositieswhich are closer to the intrinsic values. For those which are not detected significantly in thehard-band, we used the total-band energy flux. The galaxies between z = 2 and 3 are shownby the blue dots and the galaxies at z = 3-4 by the red dots. There is no clear trend in Fig.6and thus the stellar mass and the X-ray luminosity does not seem to strongly correlate witheach other.In Fig.7, we plotted the ’specific AGN activity’, namely the X-ray luminosity dividedby the host stellar mass, against their host stellar mass. Note that this is just a differentpresentation of the data shown in Fig.6. As expected from Fig.6, notable trend that morestellar massive galaxies have smaller specific AGN activity is observed. If the M BH /M str ratiois the same among the galaxies and the BHs are similarly active (in mass accretion), thespecific AGN luminosity is to be constant. For the guide, we plotted the case of the localM BH /M sph ratio and the 10% Eddington accretion rate by the dotted line and 100% and 1%cases by the upper and lower dashed lines, respectively. Fig.7 shows that the specific AGNactivity is smaller than the 10%-Eddington value for the massive galaxies with M str > Table 2: Summary of the K -band and X-ray SourcesAll M str > M ⊙ . M ⊙ < M str < M ⊙ K -band Sources with 2 < z < ⊙ and larger for the less massive galaxies near M str = 10 . M ⊙ . This indicates that BHsin massive galaxies are relatively inactive, or, that more massive BHs are formed in relativelyless massive galaxies and they have the similar activities. We discuss these results furtherin Sec. 4.It is also possible that attenuation of the X-ray emission in massive galaxies causes sucha trend. In order to see whether this is the case, we checked the Spitzer
MIPS 24 µ m fluxdata of the sample since the objects with luminous infrared emission may host such obscuredAGN. The objects with large circles in Fig.7 are those detected in MIPS 24 µ m band above ∼ µ Jy. The MIPS detection fraction among the massive objects, M str > × M ⊙ ,is 39% (7/18) for those with the relatively large specific AGN activity (objects above thedotted line in Fig.7) and 52% (13/25) for the rest. Thus the 24 µ m detection rate is similarfor the two samples, which does not support that the low specific AGN activity is due to theX-ray attenuation in dust-rich environment.Fig.8a and 8b show the distribution of the X-ray luminosity and the star-formation rate(SFR) of the galaxies obtained by the SED fitting. We use the ’instantaneous’ SFR in the leftpanel (Fig.8a) and the SFR averaged over the past 300 Myrs in the right (Fig.8b). The SFRranges from a few to 1000 M ⊙ yr − and the specific SFR, namely the SFR divided by theirstellar mass, ranges from 10 − to 10 − yr − . We do not see any clear correlation between L X (2-10 keV) and SFR over 1-1000 M ⊙ yr − . If the star-formation in each galaxy and themass accretion to its BHs are closely correlated and exactly coeval all the time, there shouldbe more tight correlation between L X and SFR, while a large uncertain here is how muchfraction of the SFR is indeed directly related to the formation of the spheroid component.As also inferred from Fig.8, the specific AGN activity and the specific SFR do not showtight correlation. The dotted lines in the figures show the expected X-ray luminosity due tothe star-formation activity predicted from the empirical fit to the local galaxies by Ranalli et al. (2003). The recent models studied by Mas-Hesse et al. (2008) give the consistentresults. We see the observed X-ray luminosities are an order-of-magnitude more luminousthan the prediction, which implies that they are due to the AGN activities and not seriouslycontaminated even by the very intense star formation.In Fig.9, we also show the distribution of the K -selected as well as the X-ray-identifiedobjects along the color excess values, E(B-V) and the average age. We adopted the Calzettilaw for the dust extinction in the SED fitting. The X-ray-identified sources are distributedover the parameter plane and no very strong trend is recognized. Ages and E(B-V) values,however, may be degenerate in the SED fitting, which possibly dilutes any intrinsic trends.There is a hint that the X-ray fraction is slightly high at the middle range of E(B-V) between0.3 and 0.6 for the massive galaxies but at the same time we note that the most reddened 11 –objects have few X-ray counterparts.
4. Discussion4.1. X-ray Sources in Distant Red Galaxies
Our results in the last section largely owe to the photometric redshift estimation. Al-though the 40% of the X-ray sources have the secure spectroscopic redshift values and thecomparison between photometric and spectroscopic redshift for the K -selected galaxies isalso reasonable (Kajisawa and Yamada 2005), it is worth analyzing the similar sample basedon the simple color cut avoiding the uncertainty of the detailed photometric redshift mea-surements.For the purpose, we defined the sample of Distant Red Galaxies (DRGs), which have J − K > . et al. et al. M ∗ ∼ M ⊙ ) at 2 < ∼ z < ∼ et al. et al. et al. et al. < ∼ z < ∼ et al. et al. et al. K <
20 have z <
2, the fainter DRGs are likely to be dominated by thegalaxies at high redshift (Kajisawa et al. z > ∼
2. Above K = 22 .
8, the completeness limit of the colorselection over GT1-GT4 fields, 182 DRGs are identified.Fig.10 shows the number counts of the whole DRGs as well as DRGs with the X-raycounterparts. Their color-magnitude distribution is plotted in Fig.11. The vertical dottedline in Fig.10 shows the detection limit for DRG, K =22.8. Note that the decrease of thenumber density of DRGs is seen at around K =22, which is more rapid than that of theentire K -band selected objects. This indicates that there are less red low mass galaxies athigh redshift (Kajisawa et al. K ∼ . str ∼ M ⊙ at z ∼
3. At
K <
21, 30-35% ofthe DRGs (19/65 for the hard band, 23/65 for the total band) are identified with the X-raysources while it decreases to 21% at
K < . K < z = 2 -4 Among the Ks -band selected galaxies at z = 2-4, the X-ray emissions are exclusivelyassociated with the massive objects, with M str > . M ⊙ . As the Chandra f (2-8 keV)=1 . × − erg s − cm − corresponds to L X (2-10 keV) = 3 × erg s − to 1 . × erg s − at z = 2 to 4, which corresponds to theluminosity of the AGN such as luminous Seyfert galaxies in nearby universe. If we assume thebolometric fraction of the X-ray emission at 2-10 keV, ǫ X and the ratio of the bolometric tothe Eddington luminosity of the AGN, R edd , the observed X-ray luminosity can be connectedto the black hole mass by the relationship, L X = 4 πGm p cM BH ǫ X R edd = 2 × ( M BH / M ⊙ )( ǫ X / . R edd / . ergs − . The absence of luminous AGN in the less massive galaxies indicates that the associatedBH mass or the activity of the AGN is lower than the limit. Unfortunately from the currentdata we cannot discriminate which is the case; we need further spectroscopic observationsto evaluate virial BH mass to break this degeneracy. Instead, we here discuss the two caseswith rather extreme assumptions to see which the more natural interpretation is.We first consider the case that SMBH mass does not depend on the host stellar mass,namely similar massive (or less massive) BHs exist in massive or less massive host galaxies.In this case, the X-ray non-detection in less massive galaxies below M str ∼ × M ⊙ isdue to the lower activity (less mass accretion rate) of their BH. However, this is the oppositetrend seen for the X-ray detected sample above ∼ × M ⊙ as shown in Fig.7 and thusnot favorable; it will be strange that the activity of AGN falls off suddenly at ∼ M ⊙ .We then consider the second case that the M BH /M sph ratio is constant and does notevolve from the local value, M BH /M sph ≈ .
002 and not depends on the host stellar mass.We also assume M str ≈ M sph . In this case, the decrease of specific AGN activity among theX-ray detected sample above ∼ M ⊙ shown in Fig.7 can be interpreted as the decreaseof relative mass accretion rate; we see that the Eddington ratio of SMBH in the host withM str = 3 × M ⊙ (or the corresponding M BH = 6 × M ⊙ ) is R edd ≈ R edd =0.1-0.01, for the more massive objects. On the other hand, few X-ray AGNis detected for galaxies with the stellar mass less than 10 M ⊙ (Fig.2), which should hostSMBH with M BH ∼ × M ⊙ with the assumption above. This BH mass expected from 13 –the no-evolution hypothesis, which may be the upper limit values if we consider the lowerspheroidal mass fraction, is comparable with the ’BH mass detection limit’ for the current Chandra
X-ray data, ≈ × M ⊙ and ≈ × M ⊙ for R edd = 0 . z = 2 and z = 4,respectively, and 2 × and 1 × M ⊙ for R edd = 1. Thus the latter interpretation seemsmore natural. Our results are at least consistent with that, at z = 2 −
4, the galaxies withstellar mass with ∼ M ⊙ or less do not host the black hole which is significantly (i.e.,more than an order of magnitude) heavier than that expected from the local relationship,i.e., M BH ∼ M ⊙ unless the accretion rate is very low, R edd < − .McLure et al. (2006) argue for the sample of 3C RR radio-loud objects that M BH /M sph ratio is an increasing function with redshift and the ratio is ∼ × larger at z ∼
2. These3C RR galaxies are very massive, M str ∼ . − M ⊙ and their BH mass is estimated to beM BH ∼ − M ⊙ . If the trend is also true for the galaxies in our sample, the galaxies withM str ∼ M ⊙ may have the BH with M BH ∼ M ⊙ , which is not favored in our results.The trend that the AGN activity seen only among the massive galaxies is very similarwith that seen for the local galaxies. Kauffmann et al. (2003) investigated the host propertiesof the ∼ ∼ . < z < . str > ∼ M ⊙ . Whilewe have to consider the X-ray detection limits carefully, the current results are consistentwith that AGN are observed only among the massive galaxies even at z = 2 − z ∼ z ∼ et al. > M ⊙ ) galaxies have been formed at z ∼ < ∼ a few × M ⊙ ) do not appear tohost the BH as massive as those located in the present-day massive ( > M ⊙ ) galaxies.In this sense, it is unlikely that massive BH had already been formed in the less-massiveprogenitors of their present-day host galaxies. We may argue that massive BHs have beendeveloped as the host galaxies have grown their stellar mass, and not significantly prior tothat. 14 – It is also surprising that a very large fraction ( ∼ active AGN. If the M BH /M str ratio for the massive galaxies does not differ from the localvalue, ≈ . R edd > .
01 as wecan see in Fig.7. While the observed X-ray luminosity and SED-fitting-based SFR are notstrongly correlated and the BH formation (mass accretion) and the star formation seems notcoeval exactly (Fig.8), it can be said that the active duration in the AGN duty cycle of thehigh-redshift massive galaxies is large.At the same time, we already noted that the negative dependence of the specific AGNactivity on the stellar mass is observed. We also investigated how the specific AGN activitydepends on the averaged stellar age obtained in the SED fitting. The results are shown inFig.12. While they are scattered out and no tight correlation exists, we see a broad trendthat the AGN activity is high in the young host galaxies with the stellar age less than 10 yr and relatively low in the galaxies which are older than 10 yr; 8 of 11 galaxies with T AGE < yr have R edd = 0 . − T AGE > yr have R edd = 0 . − . > K -band-selected very massive ( ∼ M ⊙ ) galaxies at z = 2 −
4, the BH growth is still ongoing but as the star-formation activitypeaked earlier in more massive galaxies, as inferred from their relatively older age and lowerspecific SFR, their AGN activities may also be peaked earlier and only relatively lower-levelactivity is seen at the observed epoch.
5. Summary
We investigated the X-ray properties of the K -band selected galaxies at 2 < z < -10 M ⊙ in 103 arcmin in GOODS-N. 61 X-ray sources, most ofwhich must be AGN, are identified with the galaxies and we found that they are exclusively( > > . M ⊙ . On the otherhand, a very large fraction, 30-35% of the galaxies with stellar mass larger than 10 M ⊙ shows the AGN activity with the X-ray luminosity larger than ∼ erg s − . We also foundthat the specific AGN activity, namely the X-ray luminosity divided by the stellar mass issmaller in the more massive galaxies. The absence of luminous AGN in the less massivegalaxies implies that the associated BH mass is smaller or the activity of the AGN is lower.Our results are consistent with the idea that the M BH /M str ratio in z = 2 − REFERENCES
Adelberger, K. L., & Steidel, C. C. 200 5, ApJ, 627, L1Akiyama, M. 2005, ApJ, 629, 72Alexander, D. M., et al. 2003, AJ, 126, 539Alexander, D. M., Bauer, F. E., Chapman, S. C., Smail, I., Blain, A. W., Brandt, W. N., &Ivison, R. J. 2005a, ApJ, 632, 736Alexander, D. M., Smail, I., Bauer, F. E., Chapman, S. C., Blain, A. W., Brandt, W. N., &Ivison, R. J. 2005b, Nature, 434, 738Alexander, D., Liu, R., & Gilbert, H. R. 2006, ApJ, 653, 719Alexander, D. M., et al. 2008, AJ, 135, 1968Barger, A. J., Cowie, L. L., & Wang, W.-H. 2008, ApJ, 689, 687Bruzual, G. & Charlot, S. 2003, MNRAS, 344, 1000Calzetti, D., Armus, L., Bohlin, R. C., Kinney, A. L., Koornneef, J., & Storchi-Bergmann,T. 2000, ApJ, 533, 682Caputi, K. I., Dunlop, J. S., McLure, R. J., & Roche, N. D. 2005, MNRAS, 361, 607Chabrier, G. 2003, ApJ, 586, L133Conselice, C. J., et al. et al. et al. et al. et al. et al. et al. et al. et al. et al. et al. et al. et al. et al.
This preprint was prepared with the AAS L A TEX macros v5.2.
19 – N u m be r Redshift all K<23log(Ms)>10.5Chandra detection
Fig. 1.— The redshift distribution of the sample galaxies with
K <
23. Those with stellarmass larger than 10 . M ⊙ as well as those with X-ray detection are shown by the lowerhistograms. 20 – J - K ( V ega ) Stellar Mass (Msun, Salpeter IMF)K-selected K<23 0 0.5 1 1.5 2 2.5 3 3.5 4 1e+08 1e+09 1e+10 1e+11 1e+12 J - K ( V ega ) Stellar Mass (Msun, Salpeter IMF)K>23 0 0.5 1 1.5 2 2.5 3 3.5 4 1e+08 1e+09 1e+10 1e+11 1e+12 J - K ( V ega ) Stellar Mass (Msun, Salpeter IMF)Chandra detected 0 0.5 1 1.5 2 2.5 3 3.5 4 1e+08 1e+09 1e+10 1e+11 1e+12 J - K ( V ega ) Stellar Mass (Msun, Salpeter IMF)Chandra Hard-band detected
Fig. 2.— The stellar mass and J − K color of the galaxies in GT1-GT4 with photometricredshift 2 < z <
4. Spectroscopic redshift are used if available. Galaxies with
K < K& I R A C - uppe r- li m i t S t e ll a r M a ss ( M s un ) best-fit Stellar Mass (Msun)2 21 22 23 24 25 26 27 28 3000 5000 10000 20000 50000 100000 AB m agn i t ude Observed Wavelength (Angstrom) X-ray object at z_spec=2.981best-fit SED with all data, Ms=7.63e10Msunwith K-IRAC upper limits, Ms=1.19e11Msun Fig. 4.— Examples of the SED fitting results of the galaxies associated with the X-raysources. Results of fitting with all the bands (magenta) and that using the K and IRACbands as upper limits (blue) are shown. (a) A typical example with the χ value middleof the sample. (b) Example of objects with relatively red SEDs. The χ value is the 10thlargest (16% from the worst). (c) An example of the ’power-law’ SED case. 23 – 22 23 24 25 26 27 28 AB m agn i t ude Observed Wavelength (Angstrom)X-ray object at z_spec=3.406best-fit SED with all data, Ms=9.35e10Msunwith K-IRAC upper limits, Ms=6.39e10Msun Fig. 4.— 24 – 20 22 24 26 28 3000 5000 10000 20000 50000 100000 AB m agn i t ude Observed Wavelength (Angstrom)X-ray object at z_phot=2.11best-fit SED with all data, Ms=2.39e11Msunwith K-IRAC upper limits, Ms=2.28e11Msun Fig. 4.— 25 – IRAC power-law source 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 19 20 21 22 23 M o ff a t F W H M ( a r cs e c ) Kmag (MAG_AUTO, Vega)2 Fig. 5.— The FWHM of the MOIRCS Ks -band images of the galaxies at 2 < z < Chandra X-ray observations. Objects and the size of the stellar images inthe different field of MODS are shown in the different color. The sources with ’power-law’SED (see text) are marked with the squares. 26 –Fig. 6.— X-ray luminosity at 2-10 keV as a function of the stellar mass of the Ks -bandcounterpart. The blue dots show the objects at 2 < z < < z < BH /M str =0.002, ǫ X =0.01, and R edd = 0 . 1. The upper andlower dashed lines are for R edd = 1 and 0.01, respectively. Large circles indicate the MIPS24 µ m-band detection. 28 – Fig. 8.— X-ray luminosity as a function of the star-formation rate obtained in the SEDfitting by the models. The instantaneous SFR is used in the left panel (a) while the SFRaveraged over the past 300 Myr are used in the right panel (b). The dotted line is theexpected X-ray luminosity from the star-formation component from the empirical fit byRanalli et al. (2003). 29 – Fig. 8.— (b) 30 – A v e r age age ( y ea r) E(B-V) E(B-V) with Chandra detection 1e+07 1e+08 1e+09 0 0.2 0.4 0.6 0.8 1 E(B-V) with Chandra Hard-band detection 1e+10 20 40 60 80 100 120 140 160 180 0 0.2 0.4 0.6 0.8 1 N u m be r E(B-V) E(B-V) with Chandra detection 1e+07 1e+08 1e+09 0 0.2 0.4 0.6 0.8 1 E(B-V) with Chandra Hard-band detection Ms=1e10-1e11Msun Ms>1e11Msun Fig. 9.— The upper panels show the number distribution of the mass-selected galaxies withand without X-ray detection along the color excess values in the Calzetti extinction modelwhich are obtained in the SED fitting. The lower panel show the average ages (see text) ofthe sample galaxies. Left and right panels show the sample with 10 M ⊙ < M str < M ⊙ and M str > M ⊙ , respectively. 31 – N u m be r D en s i t y ( / a r c m i n ^ ) Kmag (Vega) J-K>2.3J-K>2.3 with Chandra detectionJ-K>2.3 with Chandra Hard-band detectionK=22.8 (DRG limit for wide-field) Fig. 10.— The K -band number counts of all the DRGs as well as those identified with theX-ray sources. The counts of the objects associated with X-ray sources are shown by thelower histograms. 32 – J - K ( V ega ) Kmag (Vega) 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 17 18 19 20 21 22 23 24 J - K ( V ega ) Kmag (Vega)J-K>2.3 with Chandra Hard-band detection 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 17 18 19 20 21 22 23 24 J - K ( V ega ) Kmag (Vega) J-K>2.3 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 17 18 19 20 21 22 23 24 J - K ( V ega ) Kmag (Vega)J-K>2.3 with Chandra detection Fig. 11.— The color-magnitude diagram of the DRGs in the MODS field. The objectsdetected in the hard-band and total-band are shown by the large and small squares, respec-tively. 33 –Fig. 12.— The specific AGN activity versus the mean stellar ages obtained in the SEDfitting. The blue dots show the objects at 2 < z < < z << z <