Abstract

We use morphological information of X-ray selected active galactic nuclei (AGN) hosts to set limits on the fraction of the accretion density of the Universe at z≈ 1 that is not likely to be associated with major mergers. Deep X-ray observations are combined with high-resolution optical data from the Hubble Space Telescope in the All-wavelength Extended Groth strip International Survey, Great Observatories Origins Deep Survey (GOODS) North and GOODS South fields to explore the morphological breakdown of X-ray sources in the redshift interval 0.5 < z < 1.3. The sample is split into discs, early-type bulge-dominated galaxies, peculiar systems and point sources in which the nuclear source outshines the host galaxy. The X-ray luminosity function and luminosity density of AGN at z≈ 1 are then calculated as a function of morphological type. We find that disc-dominated hosts contribute 30 ± 9 per cent to the total AGN space density and 23 ± 6 per cent to the luminosity density at z≈ 1. We argue that AGN in disc galaxies are most likely fuelled not by major merger events but by minor interactions or internal instabilities. We find evidence that these mechanisms may be more efficient in producing luminous AGN forumla compared to predictions for the stochastic fuelling of massive black holes in disc galaxies.

1 INTRODUCTION

It is well established that active galactic nuclei (AGN), which are signposts of accretion events on to the supermassive black holes (SMBHs) at the centres of galaxies, evolve strongly with cosmic time. Observational programmes over the last 35 years have shown that the space density of these systems has a broad peak at z≈ 1– 3 (e.g. Hasinger, Miyaji & Schmidt 2005; Aird et al. 2008; Brusa et al. 2008; Silverman et al. 2008) followed by a decline by almost 2 dex to the present day (e.g. Ueda et al. 2003; Barger et al. 2005; Hasinger et al. 2005). The physical interpretation of the observed rapid evolution of the AGN population, however, is still a matter of considerable debate. An important piece of the puzzle is the process that triggers the accretion of material on to the central SMBH.

A major galaxy merger is one of the popular mechanisms suggested to be responsible for fuelling AGN. Numerical smoothed particle hydrodynamic (SPH) simulations demonstrate that these violent events are very efficient in funnelling gas to the nuclear galaxy regions (e.g. Hernquist 1989; Barnes & Hernquist 1991, 1996), where it can be consumed by the SMBH (Di Matteo, Springel & Hernquist 2005; Springel et al. 2005). This scenario is also attractive because galaxy bulges are built parallel to the formation of SMBHs at their centres, thereby providing a physical interpretation for the observed tight correlation between bulge velocity dispersion and black hole mass in nearby galaxies (e.g. Ferrarese & Merritt 2000; Gebhardt et al. 2000). As a result, most semi-analytic cosmological simulations of galaxy formation (e.g. Somerville et al. 2008) use major mergers as the primary mechanism for growing SMBHs. In this scenario, changes with redshift of the AGN space density are intimately related to the evolution of the major merger rate in the Universe.

Observations provide some support to the simulations described above, thereby suggesting that SMBHs can be fuelled by major mergers. In the local Universe, for example, the population of ultraluminous infrared galaxies (ULIRGs), which are dominated by major mergers (e.g. Surace et al. 1998; Borne et al. 2000; Farrah et al. 2001), includes a large fraction of systems that show AGN signatures at optical and/or X-ray wavelengths (Sanders & Mirabel 1996; Franceschini et al. 2003). Additionally, a large fraction of luminous quasi-stellar objects (QSOs) at low redshift are associated with either morphologically disturbed galaxies, suggesting ongoing tidal encounters (Canalizo & Stockton 2001; Guyon, Sanders & Stockton 2006), or early-type hosts with fine structure in their optical light distribution, indicative of past interactions (e.g. Canalizo et al. 2007; Bennert et al. 2008). Moreover, the excess of optical neighbours around z < 0.4 QSOs on small scales (≈0.1–0.5 Mpc) compared to L galaxies (Serber et al. 2006) and the higher fraction of QSO pairs with separations <0.1 Mpc compared to the expectation on large (>3 Mpc) scales (Hennawi et al. 2006; Myers et al. 2007) both suggest that galaxy mergers play an important role in the evolution of QSOs.

Despite the evidence above, it is also recognized that many AGN are not fuelled by major mergers. In the nearby Universe, for example, many Seyferts, which represent intermediate and low-luminosity AGN, are associated with non-interacting spirals (e.g. Ho, Filippenko & Sargent 1997). The small-scale environment (<0.1 Mpc) of narrow-line AGN, including a large fraction of Seyferts, in the Sloan Digital Sky Survey (SDSS) suggests that major galaxy interactions cannot explain the observed activity in the bulk of the population (Li et al. 2008). It is believed that the SMBHs of these systems accrete cold gas stochastically by mechanisms such as bar instabilities or minor tidal disruptions rather than major mergers (Hopkins & Hernquist 2006). It is also suggested that minor interactions and secular evolution are important at low redshift (z≲ 1) and low luminosities (e.g. LX≲ 1043 erg s−1), while major mergers dominate at high redshifts and bright luminosities (e.g. Hasinger 2008; Hopkins & Hernquist 2006). Compared to models where major mergers drive the history of luminous accretion at all epochs (e.g. Hopkins et al. 2006; Somerville et al. 2008), the scenario above offers an alternative interpretation of the evolution of AGN based on a change with redshift of the dominant fuelling mode of the SMBH.

Although there is no doubt that AGN are fuelled by mechanisms that do not involve major mergers, as yet there are no robust limits on the relative contribution of such mechanisms to the total accretion luminosity. One way to address this issue is to study the morphology of the AGN host galaxies. In the standard view of major mergers, any pre-existing discs are destroyed to form a bulge-dominated remnant (e.g. Barnes & Hernquist 1996). Therefore, AGN hosted by either disturbed/irregular systems or spheroidal galaxies are likely associated with ongoing or past major mergers, respectively. In contrast, AGN in disc galaxies are candidates for the stochastic SMBH fuelling mode. Although recent simulations show that under certain conditions (e.g. high gas fraction) discs can survive major mergers (e.g. Springel & Hernquist 2005; Robertson et al. 2006; Governato et al. 2009; Hopkins et al. 2009), it is also claimed that the formation of a remnant with a large disc requires conditions which are often not associated with substantial SMBH fuelling (e.g. Hopkins & Hernquist 2009a). The fraction of AGN in disc galaxy hosts, therefore, remains a strong constraint on the fuelling mechanism and/or the conditions under which SMBH growth occurs.

At low redshift, for example, about 30 per cent of the powerful QSOs live in spirals (Canalizo & Stockton 2001; Guyon et al. 2006), while at z≈ 1, close to the peak of the accretion history of the Universe, about 20–30 per cent of the X-ray selected AGN are hosted by late-type disc galaxies (Pierce et al. 2007; Gabor et al. 2009). These observations are often interpreted as examples of SMBHs fuelled by minor interactions and/or secular evolution. What is uncertain, however, is the luminosity distribution of the AGN in disc hosts, i.e. what fraction of the AGN space and luminosity density is associated with this type of galaxy. This is essential to place a limit on the significance of minor interactions and/or internal instabilities to the accretion history of the Universe.

In this paper, we address this issue by estimating, as a function of host galaxy morphology, the X-ray luminosity function and the X-ray luminosity density of X-ray selected AGN at z≈ 1. Our analysis provides a lower limit to the fraction of AGN fuelled by secular evolution and minor mergers or equivalently an upper limit to the fraction of AGN in major mergers. For this exercise, we use deep Chandra surveys with available multiwaveband Hubble Space Telescope (HST) observations. The fields of choice are the 2 Ms Chandra Deep Field-North (CDF-N), the 2 Ms Chandra Deep Field-South (CDF-S) and the All-wavelength Extended Groth strip International Survey (AEGIS; Davis et al. 2007). Using data from these programmes has the advantage of publicly available follow-up observations, including deep ground-based optical photometry and extensive spectroscopy. Throughout this paper, we adopt a cosmology with forumla and ΩΛ= 0.7.

2 DATA

2.1 X-ray observations

Table 1 presents information on the Chandra surveys of the AEGIS, CDF-N and CDF-S. The X-ray observations of the Extended Groth Strip consist of eight Advanced CCD Imaging Spectrometer (ACIS-I) pointings, each with a total integration time of about 200 ks split in at least three shorter exposures obtained at different epochs. The CDF-N X-ray survey consists of 20 individual ACIS-I observations, which sum up to a total exposure time of 2 Ms. The Chandra data in the CDF-S have recently been supplemented with new ACIS-I pointings, which have increased the total exposure time in that field to about 2 Ms.

Table 1

Chandra observations details.

Survey  Obs. IDs  Exposure  Area  All  0.5–7 keV 
(Ms)  forumla   sources  sources 
(1)  (2)  (3)  (4)  (5)  (6) 
CDF-N  580, 957, 966, 967, 1671, 2232,  1.93  0.11  516  273 
2233, 2234, 2344, 2386, 2421,   
2423, 3293, 3294, 3388-3391,   
3408, 3409   
CDF-S  581 441 582 1672 2405 2239 2312  1.93  0.06  428  245 
2313 2406 2409 1431 8591-8597   
9575 9578 9593 9596 9718   
AEGIS  3305, 4357, 4365, 5841-5854,  0.19  0.63  1325  437 
6210-6223, 6366, 6391, 7169 7180,   
7181, 7187, 7188, 7236, 7237,   
7238, 7239   
Survey  Obs. IDs  Exposure  Area  All  0.5–7 keV 
(Ms)  forumla   sources  sources 
(1)  (2)  (3)  (4)  (5)  (6) 
CDF-N  580, 957, 966, 967, 1671, 2232,  1.93  0.11  516  273 
2233, 2234, 2344, 2386, 2421,   
2423, 3293, 3294, 3388-3391,   
3408, 3409   
CDF-S  581 441 582 1672 2405 2239 2312  1.93  0.06  428  245 
2313 2406 2409 1431 8591-8597   
9575 9578 9593 9596 9718   
AEGIS  3305, 4357, 4365, 5841-5854,  0.19  0.63  1325  437 
6210-6223, 6366, 6391, 7169 7180,   
7181, 7187, 7188, 7236, 7237,   
7238, 7239   

Note. Columns are (1) Survey name; (2) Chandra observation IDs used for each survey; (3) exposure time in Ms after cleaning for flares. In the case of the AEGIS survey with multiple pointings, the median exposure time per pointing is listed; (4) total surveyed area in forumla; (5) total number of sources in each survey; (6) number of 0.5–7 keV selected sources that overlap with the regions covered by the HST/ACS observations of each survey.

The Chandra observations in the three fields above have been reduced and analysed in a homogeneous way using the methodology described by Laird et al. (2008). Briefly, the reduction used the ciao data analysis software. After merging the individual observations into a single event file, we constructed images in four energy bands 0.5–7.0 keV (full), 0.5–2.0 keV (soft), 2.0–7.0 keV (hard) and 4.0–7.0 keV (ultrahard). The count rates in the above energy intervals are converted to fluxes in the standard bands 0.5–10, 0.5–2, 2–10 and 5–10 keV, respectively. For the full, hard and ultrahard bands, this conversion involves an extrapolation from 7 to 10 keV. Sources in the flux range probed by the Chandra surveys listed in Table 1 have, on average, spectra similar to that of the cosmic X-ray Background in the 2–10 keV band, forumla (e.g. Hickox & Markevitch 2006). Therefore, to convert source count rates to fluxes we adopt a power-law X-ray spectrum with index forumla absorbed by the Galactic hydrogen column density appropriate for each field. Adopting forumla would decrease the estimated fluxes in the 0.5–10 keV band by about 30 per cent.

The source detection is a two-pass process. A list of candidate sources is first constructed by running the wavdetect task of ciao at a low detection threshold (10−4). The total counts (source and background) at the position of each candidate source are then extracted within the 90 per cent encircled energy fraction (EEF) radius of the point spread function (PSF; see section 2.2. of Laird et al. 2009 for details). A local background value is determined by first excluding pixels within the 95 per cent EEF of each candidate wavdetect source and then summing up the counts in an annulus centred on the source with inner radius equal to 1.5 times the 90 per cent EEF radius and width of 50 arcsec. The probability that the candidate source is a random fluctuation of the background is estimated assuming Poisson statistics. The final catalogue is comprised sources with Poisson probabilities <4 × 10−6 in at least one of the energy bands defined above. We choose to work with sources selected in the 0.5–7 keV energy band because it provides higher sensitivity resulting in a larger sample size. As we will only use X-ray sources that overlap with the multiwaveband HST surveys (see below) of the AEGIS, CDF-N and CDF-S, in Table 1 we list the number of 0.5–7 keV selected sources that lie in the area covered by the HST observations.

The construction of the sensitivity maps is described by Georgakakis et al. (2008a). In this paper, we will use the one-dimensional representation of the sensitivity map, the X-ray area curve, which provides an estimate of the total survey area in which a source with flux fX can be detected. For simplicity, we use area curves calculated in the standard way, i.e. by assigning a single limiting flux to a detection cell, instead of the Bayesian approach developed by Georgakakis et al. (2008a). The area curves for the HST subregions of the AEGIS, CDF-N and CDF-S are shown in Fig. 1.

Figure 1

Sensitivity curves in the 0.5–10 keV band for the HST covered subregions of the CDF-S (dotted line), CDF-N (short-dashed line) and AEGIS (long-dashed line). The continuous line is the sum of the individual curves.

Figure 1

Sensitivity curves in the 0.5–10 keV band for the HST covered subregions of the CDF-S (dotted line), CDF-N (short-dashed line) and AEGIS (long-dashed line). The continuous line is the sum of the individual curves.

2.2 HST photometry

High-resolution imaging observations from the HST are used to identify X-ray sources with optical counterparts and to explore the morphology of their host galaxies.

The Advanced Camera for Surveys (ACS) aboard HST has been used to obtain deep images in the V (F606W, 2260s) and I (F814W, 2100s) filters over a 0.1 × 70.5 arcmin2 strip in the AEGIS field (Lotz et al. 2008b). The 5σ limiting magnitudes for a point source are VF606W= 28.14 (AB) and IF814W= 27.52 (AB). For extended sources, these limits are about 2 mag brighter. A total of 15 797 galaxies are detected to IF814W= 25 mag.

The HST has also surveyed the central most sensitive part of the CDF-N and CDF-S with the ACS as part of the Great Observatories Origins Deep Survey (GOODS) programme. These observations cover a combined area of about forumla in four passbands, B (F435W), V (F606W), i (F775W) and z (F850LP). The survey setup, data reduction and source detection are described by Giavalisco et al. (2004). We use the version v1.0 of the mosaicked ACS images and the version v1.1 of the ACS multiband source catalogues. The total exposure time of this data release is about 7200, 5000, 5000 and 10660 s in the B, V, i and z filters, respectively. We use the z-band selected catalogues which include a total of 32 048 and 29 599 sources in the CDF-N and the CDF-S, respectively.

The likelihood ratio (LR) method (Sutherland & Saunders 1992; Ciliegi et al. 2003; Laird et al. 2009) is adopted to identify X-ray sources with optical counterparts. In this exercise, we estimate the ratio between the probability that a source, at a given distance from the X-ray position and with a given optical magnitude, is the true counterpart and the probability that the source is a spurious alignment,  

1
formula
where q(m) is the expected magnitude distribution of the true optical counterparts, f(r) is the probability distribution function of the positional uncertainties in both the X-ray and the optical source catalogues and n(m) is the background density of optical galaxies of magnitude m.

For the positional accuracy of the X-ray sources, we adopt a Gaussian distribution with standard deviation estimated as a function of off-axis angle and total number of counts as described by Laird et al. (2009). The positional uncertainty of the HST source lists is also assumed to follow the normal distribution with full width half-maximum of 0.07 arcsec for AEGIS (Lotz et al. 2008b) and 0.12 arcsec for the CDF-N and CDF-S. In equation 1, f(r) is therefore assumed to be a Gaussian with standard deviation equal to the combined optical and X-ray source positional errors added in quadrature.

The a priori probability q(m) that an X-ray source has a counterpart with magnitude m is determined as follows. First, the X-ray and optical source catalogues are matched by simply finding the closest counterpart within a fixed search radius of 2 arcsec. Secondly, a total of 100 mock catalogues are constructed by randomizing the source positions of the HST catalogue and the cross matching is repeated. The magnitude distribution of the counterparts in the mock and the real catalogues is subtracted to determine the magnitude distribution of the true associations, q(m).

Having estimated f(r) and q(m), we identify all possible counterparts to X-ray sources within a 4 arcsec radius and the LR of each one is estimated using equation 1. We consider as counterparts those galaxies with LR above a certain limit. The choice of the cut-off in LR is a trade-off between maximum number of counterparts and minimum spurious identification rate. In order to assess how secure an optical counterpart is, we use the reliability parameter defined by Sutherland & Saunders (1992), 

2
formula
where Rel i is the reliability of the i optical counterpart of an X-ray source, the index j of the summation runs over all the possible counterparts within the search radius and Q is the fraction of X-ray sources with identifications to the magnitude limit of the optical survey. It can be shown that the sum of the reliabilities of individual counterparts equals the expected number of true associations (Sutherland & Saunders 1992). Comparison of forumla with the total number of counterparts with LR > LRlimit provides an estimate of the spurious identification rate. By varying LRlimit, one can minimize the number of false associations. We chose to use LRlimit= 0.5. At this cut-off, 78 (340/437), 74 (201/273) and 81 (199/245) per cent of the 0.5–7 keV selected X-ray sources in the AEGIS, CDF-N and CDF-S, respectively, have counterparts. These fractions are estimated using the total number of X-ray sources that overlap with the HST surveys of these fields. The estimated spurious identification rate is <5 per cent.

2.3 Optical spectroscopy

Optical spectroscopy of X-ray sources in the AEGIS field is primarily from the Deep Extragalactic Evolutionary Probe 2 (DEEP2) redshift survey (Davis et al. 2003), which uses the Deep Imaging Multi-Object Spectrograph (DEIMOS) spectrograph (Faber et al. 2003) on the 10 m Keck-II telescope to obtain redshifts for galaxies to RAB= 24.1 mag. The observational setup uses a moderately high-resolution grating (R≈ 5000), which provides a velocity precision of 30 km s−1 and a wavelength coverage of 6500–9100 Å. This spectral window allows the identification of the strong [O ii] doublet 3727 Å emission line to z < 1.4. We use DEEP2 galaxies with redshift determinations secure at the >90 per cent confidence level (quality flag Q≥ 3; Davis et al. 2007). Spectroscopic observations targeting X-ray sources in AEGIS have also been carried out at the multi-mirror telescope (MMT) using the Hectospec fibre spectrograph. Full details about these observations are given by Coil et al. (2009). A total of 449 X-ray sources were targeted with five Hectospec configurations. The data were taken in queue mode in 2007 May, July and 2008 May. Total integration times were 2 h per configuration. The wavelength coverage is ∼4500–9000 Å at 6 Å resolution. The data were reduced using the hsredidl reduction pipeline.1 Redshifts were measured either from emission lines from the AGN itself or H ii regions if the host galaxy has ongoing star formation, or from the Ca H+K absorption features for lower luminosity AGN in early-type galaxies with little or no star formation. Although objects were targeted to RAB= 25, the secure redshift identification rate was 50 per cent at RAB= 22.5 mag and 15 per cent at RAB= 24.5 mag. We obtained high-confidence redshifts for 230 sources, or 51 per cent of the targeted sample. The resulting redshifts lie in the range 0 < z < 4. The original Groth Strip has been targeted by a number of spectroscopic programmes (Lilly et al. 1995; Brinchmann et al. 1998; Hopkins, Connolly & Szalay 2000; Vogt et al. 2005) that have been compiled into a single data base by Weiner et al. (2005).2 The entire AEGIS overlaps with the SDSS and therefore spectra for relatively bright galaxies and QSOs are also available (York et al. 2000). All the above diverse data sets provide high-confidence spectroscopic redshifts for 159 0.5–7 keV selected X-ray sources in the HST/ACS subregion of the AEGIS.

Optical spectroscopy in the CDF-N and in the GOODS-North is available from either programmes that specifically target the X-ray population in these fields (e.g. Barger et al. 2003; Cowie et al. 2003; Barger et al. 2005) or the Keck Treasury Redshift Survey (TKRS; Wirth et al. 2004). The TKRS uses the DEIMOS spectrograph (Faber et al. 2003) at the Keck-II telescope to observe optically selected galaxies to RAB≈ 24.4 mag in the GOODS-North. The publicly available catalogue consists of about 1400 secure redshifts. Out of the 201 X-ray sources selected in 0.5–7 keV band and optical counterparts with LH > 0.5, a total of 149 have spectroscopic redshifts.

There are many spectroscopic campaigns in the CDF-S. These include follow-up observations targeting specific populations, such as X-ray sources (Szokoly et al. 2004), K-band selected galaxies (Mignoli et al. 2005) and high-redshift candidates (e.g. Dickinson et al. 2004; Stanway et al. 2004a,b), as well as generic spectroscopic surveys of faint galaxies in the GOODS area using the FORS2 (Vanzella et al. 2005, 2008) and VIMOS (Le Fèvre et al. 2004; Popesso et al. 2009) spectrographs at the Very Large Telescope (VLT). There are 151 spectroscopic redshift estimates of the 199 0.5–7 keV selected X-ray sources with HST optical identifications.

2.4 Ground-based imaging and photometric redshifts

The spectroscopic observations are complemented by photometric redshifts calculated using ground-based imaging of the fields listed in Table 1. This provides a redshift (spectroscopic and photometric) completeness of 100 per cent for the magnitude-limited subsample used in this paper (see the next section for details on sample selection). We prefer the ground-based observations over the HST/ACS data for the photometric redshift estimation because they are deeper (AEGIS, CDF-N) and/or have more photometric bands, including imaging in the U filter.

The AEGIS is one of the fields of the deep synoptic Canada–France–Hawaii Telescope Legacy Survey (CFHTLS). These observations cover 1 deg2 in the Extended Groth Strip in five filters (ugriz). We use the CFHTLS data from the T0003 release,3 which reaches a limiting magnitude of iAB≈ 26.5 mag. We use the photometric redshifts estimated by Ilbert et al. (2006) for sources brighter than iAB≈ 25 mag. The CFHTLS data overlap fully with area surveyed by the HST, thereby providing photometric redshifts for all HST sources with iAB≲ 25 mag.

Deep multiwavelength imaging (UBVRIz) in the CDF-N has been presented by Capak et al. (2004). These observations cover about 0.2 deg2 and extend beyond the CDF-N field of view. In this study, we use the R-band selected sample comprising 47 451 sources to the limit RAB= 26.6 mag (5σ). Photometric redshifts using these data have been presented by Georgakakis et al. (2007) using the methods described by Rowan-Robinson et al. (2005, 2008).

Ground-based photometric observations in 17 narrow and broad-band filters have been obtained in an area of 1 deg2 centred on the CDF-S as part of the COMBO-17 survey (Wolf et al. 2004). Although these observations are not as deep as the HST photometry, the large number of filters provides excellent photometric redshift estimates to R≈ 24 mag (Wolf et al. 2004).

Fig. 2 compares photometric with spectroscopic redshift estimates for X-ray sources in the CDF-S, CDF-N and AEGIS. Catastrophic redshifts, defined as those with (zspeczphot)/(1 +zspec) > 0.15, represent 13, 16 and 32 per cent of the photometric redshift determinations of X-ray sources in the CDF-S, CDF-N and AEGIS, respectively. The accuracy of the photometric redshifts is estimated by the rms values of the quantity (zspeczphot)/(1 +zspec) after excluding catastrophic redshifts. This is estimated 0.04, 0.06 and 0.05 for each of the surveys above, respectively. The CFHTLS photometric redshifts of X-ray sources suffer the largest catastrophic redshift rate. This is not surprising as these redshifts have been estimated using galaxy templates only (Ilbert et al. 2006) and also use a smaller number of photometric bands compared to the CDF-N and CDF-S photometric redshifts.

Figure 2

Photometric against spectroscopic redshift estimates for the X-ray sources in the CDF-S (red dots), CDF-N (open blue circles) and AEGIS (black stars). The continuous line is the zphot=zspec relation, while the dashed lines correspond to δz=±0.2. All sources with spectroscopic redshifts are plotted without applying any magnitude limit.

Figure 2

Photometric against spectroscopic redshift estimates for the X-ray sources in the CDF-S (red dots), CDF-N (open blue circles) and AEGIS (black stars). The continuous line is the zphot=zspec relation, while the dashed lines correspond to δz=±0.2. All sources with spectroscopic redshifts are plotted without applying any magnitude limit.

Uncertainties in the photometric redshifts are expected to have only a small impact on the results. As described in the next section, in the analysis we use a magnitude-limited subsample of X-ray sources (IAB < 23 mag for the AEGIS and zAB < 24 mag for the GOODS fields), for which the spectroscopic redshift completeness is about 88 per cent.

3 SAMPLE SELECTION AND MORPHOLOGICAL CLASSIFICATION

The sample consists of X-ray sources selected in the 0.5–7 keV band and in the redshift (photometric or spectroscopic) interval z= 0.5–1.3. The median redshift of the sample is ≈0.8 (see below), thereby allowing the study of AGN hosts close to the peak of X-ray luminosity density of the Universe (e.g. Ueda et al. 2003).

Two main approaches are adopted in the literature for quantifying the optical morphology of galaxies. Visual inspection (e.g. Bell et al. 2005) and automated classification schemes, which measure morphology-sensitive structural parameters of the light distribution of galaxies. The latter methods include both non-parametric morphology estimators (Abraham et al. 1996; Lotz, Primack & Madau 2004) and model fits to the optical profile of galaxies (e.g. Peng et al. 2002; Simard et al. 2002). These techniques have the advantage of objectivity and repeatability of the measurements. However, there are also limitations. In the case of AGN, for example, the nuclear point source can be sufficiently bright to bias the estimation of structural parameters of the host galaxy light profile. The AGN light distribution (usually a point source) has to be modelled accurately and then subtracted from the image, before the morphology of the host galaxy is determined. Although there has been progress recently in this direction (e.g. Kuhlbrodt, Wisotzki & Jahnke 2004; Gabor et al. 2009; Simmons & Urry 2008), the decomposition of the nuclear point source from the host galaxy is not straightforward, especially in the case of real galaxies with complex profiles, including discs, bulges, star-forming regions, dust lanes, etc. Additionally, non-parametric morphology estimators typically produce broad morphological classes. Irregular/interacting systems, for example, may be hard to discriminate from late-type disc galaxies with bright star-forming regions superimposed on their spiral arms. It is also unclear how methods that fit models (e.g. Sersic function) to the galaxy light profile treat irregular or interacting galaxies and whether they can isolate such systems.

Eyeballing has issues of its own (e.g. Lotz et al. 2008a), but is best suited to this study which requires discrimination between spirals, bulges and irregular/interacting systems. Also, the presence of a photometrically important bulge or nuclear point source may bias automated morphological classification schemes against discs, even if they are large and obvious to the eye. We therefore choose to proceed with visual inspection of the HST images of X-ray AGN to classify them into four groups, discs (i.e. spirals), early-types (i.e. bulge-dominated), peculiar/interacting and point sources. The latter class includes systems where the central AGN outshines the host galaxy and therefore morphological classification is not possible. This type of sources poses a challenge for any method that attempts to determine the light distribution of the underlying AGN host galaxy. The morphological classes have been determined independently by three classifiers (AG, ESL, ALC). In the analysis that follows, we apply equal weights to each of the three independent classifications. We caution that faint or red disc-dominated sources may be misclassified as early-type hosts, thereby introducing a systematic bias against discs in our visual morphological classification. However, comparison with the classifications based on the Sersic index suggests that this bias is not affecting the results and conclusions (see Section 5).

We have empirically determined that reliable morphological classes are possible to IAB= 23 mag for the AEGIS and zAB= 24 mag for the deeper GOODS HST observations. Therefore, in the analysis that follows we use a magnitude-limited subsample of 454 (CDF-N: 156; CDF-S: 130; AEGIS: 168) X-ray sources brighter than IAB= 23 and zAB= 24 mag for the AEGIS and the GOODS fields, respectively. This choice of magnitude limits has the additional advantage of high spectroscopic redshift completeness, thereby minimizing the use of photometric redshifts. Secure spectroscopic redshifts are available for 398 of the 454 sources, of which seven are Galactic stars. We use photometric redshifts for 56 sources only (CDF-N: 14; CDF-S: 13; AEGIS: 29) and therefore uncertainties in the estimation of photometric redshifts have a small impact on the results. If we further constrain the sample to the redshift interval z= 0.5–1.3, the total number of X-ray sources is 266 (median redshift 0.8). About 30 per cent of the sources in that sample have been assigned different morphological classifications by the three independent classifiers. We have tested that these differences do not affect the results and conclusions of the paper. In the next section, for example, the X-ray luminosity function for different host galaxy morphological types is estimated. It is found that the X-ray luminosity functions for the three independent morphological classifications agree within the 1σ Poisson uncertainty. Examples of the HST images of X-ray sources classified as spirals, bulges and peculiars/interacting are shown in Fig. 3.

Figure 3

Examples of the HST true colour images of AGN classified as discs (top row), early-type (middle row) and peculiar/interacting (bottom row). The X-ray luminosity (0.5–10 keV) and redshift of each source are also shown. The images are about 5 × 5 arcsec in size.

Figure 3

Examples of the HST true colour images of AGN classified as discs (top row), early-type (middle row) and peculiar/interacting (bottom row). The X-ray luminosity (0.5–10 keV) and redshift of each source are also shown. The images are about 5 × 5 arcsec in size.

4 LUMINOSITY FUNCTION

The binned X-ray luminosity function is derived using the standard non-parametric 1/Vmax method (Schmidt 1968). In this calculation, we take into account the X-ray selection function, parametrized by the X-ray area curve, and the optical magnitude limit for morphological classification. The luminosity function, φ, in logarithmic luminosity bins, d log L, is estimated from the relation  

3
formula
where Vmax,i(LX, M, z) is the maximum comoving volume for which the source i with X-ray luminosity LX, absolute optical magnitude M and redshift z satisfies the sample selection criteria, i.e. redshift range, apparent optical magnitude cut-off for morphological classification and X-ray flux limit. The maximum comoving volume is estimated from the relation  
4
formula
where dV/dz is the volume element per redshift interval dz. The integration limits are z1 =zL and z2 = min (zoptical, zU), where we have defined zL, zU the lower and upper redshift limits of the sample and zoptical is the redshift at which the source will become fainter than the optical magnitude cut-off for morphological classification. Ω(LX, z) is the solid angle of the X-ray survey available to a source with luminosity LX at a redshift z (corresponding to a flux fX in the X-ray area curve). The uncertainty of a given luminosity bin is  
5
formula
In addition to the Poisson uncertainties above, differences among the morphological classes determined by the three independent classifiers are also accounted for in the error budget. We perform 1000 experiments in which the X-ray Luminosity Function (XLF) is determined by randomly choosing for each source one of the three morphological types. This calculation provides for each luminosity bin an estimate of the 1σ rms of the XLF because of uncertainties in the morphological classifications. This is then added quadratically to the Poisson errors above. We note, however, that Poisson errors typically dominate the uncertainties.

In order to convert absolute to apparent optical magnitude for the 1/Vmax calculation, optical k-corrections for both galaxy and QSO templates are used depending on the source morphological classification. For the early, disc and peculiar classes, we adopt, respectively, the E/S0, Scd and Im galaxy type templates of Coleman, Wu & Weedman (1980), while for point sources we use the SDSS composite QSO spectrum of Vanden Berk et al. (2001). For AEGIS sources, the optical k-corrections are estimated for the F814W filter, while for CDF-N and CDF-S X-ray AGN the F850LP filter is used in the calculations.

The X-ray luminosities in the 0.5–10 keV band are corrected for obscuration. The hardness ratio (HR), estimated in the 0.5–2 and 2–7 keV bands, is used to determine the column density, NH, of individual sources assuming an intrinsic power-law X-ray spectrum with index Γ= 1.9 (e.g. Nandra & Pounds 1994). These column densities are then used to correct the observed LX for absorption. We caution that for unobscured sources the assumption of Γ= 1.4 in Section 2 overestimates the fluxes and hence the luminosities by about 30 per cent or Δ log LX= 0.1. This effect therefore has only a minor impact on the estimated LX(0.5–10 keV). Additionally, unobscured sources (NH < 1022 cm−2) are only a small fraction of the X-ray population at the depths of the CDF-N/S (≈20 per cent; Akylas et al. 2006) and the AEGIS (≈40 per cent; Georgakakis et al. 2006). When estimating the X-ray k-correction of individual sources for the 1/Vmax calculation, we also take into account the intrinsic source NH by adopting an absorbed power-law spectral energy distribution with Γ= 1.9 and photoelectric absorption cross-sections as described by Morrison & McCammon (1983) for solar metallicity.

5 RESULTS

Fig. 4 plots the X-ray luminosity distribution of different AGN host galaxy morphological types. In agreement with previous studies, early-type bulge-dominated galaxies are the largest group, contributing 42 ± 4 per cent of the overall AGN population. These systems are either major merger remnants observed after the peak of the activity (e.g. Hopkins et al. 2006) or galaxies that have recently accreted cold gas, through minor interactions, which has resulted in the re-activation of the SMBH at their centres (e.g. Khalatyan et al. 2008). Alternative models for the fuelling of the SMBH in early-type galaxies include recycled gas from dying stars (e.g. Ciotti & Ostriker 2007) or accretion from a cooling hot halo (e.g. Croton et al. 2006; Cattaneo & Teyssier 2007). Georgakakis et al. (2008b) have shown that the asymmetry parameter (Abraham et al. 1996) distribution of X-ray AGN is broader compared to optically selected early-type galaxies, suggesting low-level morphological disturbances. This finding favours the scenario where AGN in early-type hosts are associated with past merger events (minor or major).

Figure 4

The X-ray luminosity distribution of X-ray selected AGN in the redshift interval 0.5 < z < 1.3 as a function of host galaxy morphology.

Figure 4

The X-ray luminosity distribution of X-ray selected AGN in the redshift interval 0.5 < z < 1.3 as a function of host galaxy morphology.

Isolated disc galaxies represent about 28 ± 3 per cent of all AGN in Fig. 4. These are systems in which the accretion on to the SMBH is likely to be triggered either by secular processes or minor tidal disruptions. Peculiars are about 20 ± 3 per cent of the AGN in our sample. These systems include ongoing major merger events. Their fraction may indicate the time-scale that major mergers can be identified morphologically (e.g. ≲ 109 yr; Lotz et al. 2008a). We caution however, that this fraction provides only an upper limit to this time-scale as the group of peculiars may also include irregular galaxies that are not necessarily the product of major mergers. Finally, point-like sources in which the light from the central engine outshines the host galaxy and does not allow morphological classification are a minority in our sample, 8 ± 2 per cent. As expected, these sources are more numerous at bright X-ray luminosities. We note that the errors in the fractions above include uncertainties in the morphological classification of individual sources (see the previous section) in addition to the Poisson errors.

The XLF, split into different host galaxy morphological types, is plotted in Fig. 5 and presented in Table 2. The XLF of all AGN in the sample is in good agreement with the recent determination of Barger et al. (2005). Fig. 6 and Table 3 present the fractional contribution of different morphological types to the total XLF. Early-type hosts are a major component of the XLF over almost the entire luminosity range, while point sources become dominant at LX(0.5–10 keV) > 1044 erg s−1. Disc hosts represent between 25 and 30 per cent of the total XLF. The fraction of Peculiar hosts to the total XLF shows a weak decreasing trend with increasing luminosity, which is, however, within the uncertainties of the data.

Figure 5

The 0.5–10 keV X-ray luminosity function for different host galaxy types. Stars (green) correspond to point sources, open triangles (cyan) are peculiar systems, open circles (red) are for early-type hosts and filled circles (blue) correspond to disc galaxies. For clarity, triangles are offset horizontally by −0.1 dex in log LX, while stars and filled circles are offset by +0.1 dex. The filled squares (black) represent the total X-ray luminosity function, independent of host galaxy type. This is compared with the dashed line, which is the Barger et al. (2005) XLF at z= 0.8 (the median redshift of the sample) estimated in the 2–8 keV band and converted to the 0.5–10 keV band assuming a power-law X-ray spectrum with photon index Γ= 1.9. The dotted curve is the ‘maximal evolution’ XLF prediction of the stochastic accretion mode of Hopkins & Hernquist (2006) at z= 1. This curve is adapted from fig. 6 of Hopkins & Hernquist (2006) adopting the Marconi et al. (2004) bolometric corrections. The continuous (blue) line is the updated version of the XLF for the stochastic accretion mode (Hopkins, private communication) estimated using a revised AGN light curve, which results in more rapid evolution of the AGN emission after the peak of the accretion (see the text for more details).

Figure 5

The 0.5–10 keV X-ray luminosity function for different host galaxy types. Stars (green) correspond to point sources, open triangles (cyan) are peculiar systems, open circles (red) are for early-type hosts and filled circles (blue) correspond to disc galaxies. For clarity, triangles are offset horizontally by −0.1 dex in log LX, while stars and filled circles are offset by +0.1 dex. The filled squares (black) represent the total X-ray luminosity function, independent of host galaxy type. This is compared with the dashed line, which is the Barger et al. (2005) XLF at z= 0.8 (the median redshift of the sample) estimated in the 2–8 keV band and converted to the 0.5–10 keV band assuming a power-law X-ray spectrum with photon index Γ= 1.9. The dotted curve is the ‘maximal evolution’ XLF prediction of the stochastic accretion mode of Hopkins & Hernquist (2006) at z= 1. This curve is adapted from fig. 6 of Hopkins & Hernquist (2006) adopting the Marconi et al. (2004) bolometric corrections. The continuous (blue) line is the updated version of the XLF for the stochastic accretion mode (Hopkins, private communication) estimated using a revised AGN light curve, which results in more rapid evolution of the AGN emission after the peak of the accretion (see the text for more details).

Table 2

AGN XLF and luminosity density as a function of morphological type.

log LX log φ(LXlog [LX φ(LX)] 
(erg s−1(×10−6 Mpc−3(×1037 erg s−1 Mpc−3
Disc Early Peculiar Point Disc Early Peculiar Point 
41–42 −3.60 ± 0.19 −3.57 ± 0.14 −3.43 ± 0.23 −5.89 ± 0.97 1.08 ± 0.19 1.17 ± 0.13 1.12 ± 0.16 −0.91 ± 0.97 
42–43 −4.03 ± 0.11 −3.89 ± 0.10 −4.35 ± 0.13 −5.16 ± 0.34 1.49 ± 0.13 1.62 ± 0.13 1.01 ± 0.2 0.42 ± 0.25 
43–44 −4.48 ± 0.15 −4.27 ± 0.09 −4.70 ± 0.15 −5.30 ± 0.18 1.93 ± 0.14 2.17 ± 0.10 1.68 ± 0.13 1.41 ± 0.19 
44–45 −5.42 ± 0.19 −5.50 ± 0.24 −6.08 ± 0.40 −5.15 ± 0.14 1.82 ± 0.20 1.72 ± 0.25 1.58 ± 0.40 2.26 ± 0.16 
45–46 – – – −6.20 ± 0.43 – – – 1.85 ± 0.43 
log LX log φ(LXlog [LX φ(LX)] 
(erg s−1(×10−6 Mpc−3(×1037 erg s−1 Mpc−3
Disc Early Peculiar Point Disc Early Peculiar Point 
41–42 −3.60 ± 0.19 −3.57 ± 0.14 −3.43 ± 0.23 −5.89 ± 0.97 1.08 ± 0.19 1.17 ± 0.13 1.12 ± 0.16 −0.91 ± 0.97 
42–43 −4.03 ± 0.11 −3.89 ± 0.10 −4.35 ± 0.13 −5.16 ± 0.34 1.49 ± 0.13 1.62 ± 0.13 1.01 ± 0.2 0.42 ± 0.25 
43–44 −4.48 ± 0.15 −4.27 ± 0.09 −4.70 ± 0.15 −5.30 ± 0.18 1.93 ± 0.14 2.17 ± 0.10 1.68 ± 0.13 1.41 ± 0.19 
44–45 −5.42 ± 0.19 −5.50 ± 0.24 −6.08 ± 0.40 −5.15 ± 0.14 1.82 ± 0.20 1.72 ± 0.25 1.58 ± 0.40 2.26 ± 0.16 
45–46 – – – −6.20 ± 0.43 – – – 1.85 ± 0.43 

Note. Columns are (1) X-ray luminosity interval; (2) logarithm of XLF in units of ×10−6 Mpc−3 for discs, early-type, peculiar and point sources; (3) luminosity density in units of ×1037 erg s −1 Mpc−3 for the morphological classes above.

Figure 6

The 0.5–10 keV X-ray luminosity function for different host galaxy types normalized to the total XLF. The vertical axis measures the contribution of different galaxy types to the XLF. The symbols are the same as in Fig. 5.

Figure 6

The 0.5–10 keV X-ray luminosity function for different host galaxy types normalized to the total XLF. The vertical axis measures the contribution of different galaxy types to the XLF. The symbols are the same as in Fig. 5.

Table 3

Fractional contribution of different morphological types to the total AGN XLF and luminosity density.

log LX (erg s−1Fraction of φtotal(LXFraction of LX φtotal(LX
Disc Early Peculiar Point Disc Early Peculiar Point 
41–42 0.28 ± 0.12 0.30 ± 0.10 0.41 ± 0.22 0.001 ± 0.001 0.29 ± 0 0.13 0.36 ± 0.11 0.33 ± 0.12 0.002 ± 0.006 
42–43 0.34 ± 0.08 0.46 ± 0.11 0.16 ± 0.05 0.02 ± 0.01 0.35 ± 0 0.11 0.49 ± 0.15 0.12 ± 0.03 0.03 ± 0.01 
43–44 0.29 ± 0.10 0.47 ± 0.10 0.17 ± 0.06 0.04 ± 0.02 0.27 ± 0 0.09 0.48 ± 0.11 0.15 ± 0.04 0.08 ± 0.03 
44–45 0.25 ± 0.11 0.21 ± 0.12 0.05 ± 0.05 0.47 ± 0.15 0.19 ± 0 0.09 0.15 ± 0.09 0.11 ± 0.10 0.53 ± 0.20 
45–46 – – – 1.00 ± 1.00 – – – 1.00 ± 1.00 
41–45 0.30 ± 0.09 0.35 ± 0.07 0.34 ± 0.15 0.010 ± 0.005 0.23 ± 0.05 0.31 ± 0.06 0.13 ± 0.05 0.33 ± 0.12 
log LX (erg s−1Fraction of φtotal(LXFraction of LX φtotal(LX
Disc Early Peculiar Point Disc Early Peculiar Point 
41–42 0.28 ± 0.12 0.30 ± 0.10 0.41 ± 0.22 0.001 ± 0.001 0.29 ± 0 0.13 0.36 ± 0.11 0.33 ± 0.12 0.002 ± 0.006 
42–43 0.34 ± 0.08 0.46 ± 0.11 0.16 ± 0.05 0.02 ± 0.01 0.35 ± 0 0.11 0.49 ± 0.15 0.12 ± 0.03 0.03 ± 0.01 
43–44 0.29 ± 0.10 0.47 ± 0.10 0.17 ± 0.06 0.04 ± 0.02 0.27 ± 0 0.09 0.48 ± 0.11 0.15 ± 0.04 0.08 ± 0.03 
44–45 0.25 ± 0.11 0.21 ± 0.12 0.05 ± 0.05 0.47 ± 0.15 0.19 ± 0 0.09 0.15 ± 0.09 0.11 ± 0.10 0.53 ± 0.20 
45–46 – – – 1.00 ± 1.00 – – – 1.00 ± 1.00 
41–45 0.30 ± 0.09 0.35 ± 0.07 0.34 ± 0.15 0.010 ± 0.005 0.23 ± 0.05 0.31 ± 0.06 0.13 ± 0.05 0.33 ± 0.12 

Note. Columns are (1) X-ray luminosity interval; (2) fractional contribution to the XLF for discs, early-type, peculiar and point sources; (3) fractional contribution to the luminosity density for the morphological classes above.

The contribution of different morphological types to the luminosity density, φLX, at z≈ 1 is shown in Figs 7 and 8. The results are also presented in Tables 2 and 3. Early-type hosts dominate the luminosity density up to LX(0.5–10 keV) = 1044 erg s−1. Above this limit, point-like sources, i.e. QSOs, become the dominant component of φLX. Disc and peculiar host galaxies represent about 11–35 per cent of the total AGN luminosity density.

Figure 7

The 0.5–10 keV X-ray luminosity density for different host galaxy types. The symbols and curves are the same as in Fig. 5.

Figure 7

The 0.5–10 keV X-ray luminosity density for different host galaxy types. The symbols and curves are the same as in Fig. 5.

Figure 8

The fractional contribution of different AGN host galaxy morphological types to the 0.5–10 keV X-ray luminosity density. The symbols are the same as in Fig. 5.

Figure 8

The fractional contribution of different AGN host galaxy morphological types to the 0.5–10 keV X-ray luminosity density. The symbols are the same as in Fig. 5.

We have checked how sensitive the above results are to the adopted method for determining the morphology of the AGN host galaxies. The classifications based on visual inspection are compared with those determined by fitting a single Sersic function to the AGN host galaxy light distribution (Griffith et al. in preparation) using the galfit code (Peng et al. 2002). A cut-off in the Sersic index, n= 2.5, is applied to discriminate between disc (n < 2.5) and bulge (n > 2.5) dominated galaxies. The Sersic model fits are reliable for systems where the central AGN does not dominate over the host galaxy light. This requirement translates to an empirically determined (visual inspection of the HST images) X-ray luminosity cut of LX(0.5–10 keV) ≈ 1044 erg s−1. The comparison between the visual and Sersic index classifications is therefore restricted to sources with LX(0.5–10 keV) < 1044 erg s−1. We only consider sources that are visually classified as ‘early’ or ‘disc’. The class of ‘peculiars’ is excluded because they comprise a large fraction of disturbed interacting/merging systems with complex optical light distributions. ‘Point’ sources are also excluded to avoid uncertainties in the Sersic function fits associated with the presence of bright emission from the nuclear galaxy regions. The Sersic indices of both ‘peculiars’ and ‘point sources’ are hard to interpret in the context of disc- or bulge-dominated light profiles. It is found that although there is a discrepancy between the two classification methods (eyeballing versus Sersic index) for about 30 per cent of the bulge- and disc-dominated sources in the sample the main results, e.g. the XLF of discs and bulges, are consistent within the 1σ errors and do not depend on the adopted method for determining the morphology of AGN hosts.

6 DISCUSSION

This paper estimates the space and luminosity density of X-ray selected AGN as a function of their host galaxy morphologies in an attempt to place constraints on the importance of major mergers, minor interactions and secular evolution to the accretion density of the Universe at z≈ 1. In the standard view of major gaseous mergers, the discs of the participating galaxies are destroyed to form an elliptical remnant. In this picture, disc hosts therefore indicate AGN fuelled by minor interactions or internal instabilities.

It is found that disc galaxies represent about 25–34 per cent of the X-ray AGN number density and 19–35 per cent of the luminosity density of these systems (i.e. accretion density). These fractions are also almost independent of X-ray luminosity up to LX(0.5– 10) ≈ 1045 erg s−1. Recently, Hopkins & Hernquist (2006) have developed a model for the stochastic fuelling of SMBHs via e.g. bar instabilities or minor interactions. In their model, this accretion mode produces relatively low-luminosity AGN, while the bulk of the more luminous active SMBH are the result of major mergers. Fig. 5 compares our results with the maximal X-ray luminosity function of the Hopkins & Hernquist (2006) stochastic accretion mode (dotted line). This model predicts a large number of low-luminosity AGN, which can potentially explain the entire XLF below ≈1043 erg s−1. Also plotted in Fig. 5 is an updated version of the XLF for the stochastic accretion mode (Hopkins, private communication; continuous line) in which the AGN light curve decays more rapidly than what is assumed by Hopkins & Hernquist (2006), in agreement with recent simulations (e.g. Younger et al. 2008) and observations of the Eddington ratio distribution of low-redshift AGN in the SDSS (e.g. Hopkins & Hernquist 2009b). More quantitatively, in the work by Hopkins & Hernquist (2006) the AGN light curve declines with time, t, as Lt−β with β= 0.6, while in the revised curve shown in the figure β= 1.5. The updated model reproduces the overall shape of the observed AGN XLF of disc galaxies but lies below the observations. Also, both models of the stochastic AGN accretion mode shown in Fig. 5 underpredict the number density of AGN in disc galaxies at bright luminosities, >1044 erg s−1. This suggests that bar instabilities and minor interactions are more efficient in producing powerful AGN than the Hopkins & Hernquist (2006) model predicts.

Alternatively, one might argue that a fraction of the X-ray AGN in disc hosts represent major merger events. Recent simulations indeed show that under certain conditions discs can survive or reform after a major merger. Some authors show that the formation of a remnant with a large disc requires very high cold gas fractions (>50 per cent in mass) in the progenitors (e.g. Springel & Hernquist 2005; Hopkins et al. 2009), which may be typical of high-redshift (z≳ 2) objects. Others show that cold gas flowing on to the remnant from Mpc scales can quickly reform a disc even if the progenitors have moderate amounts of cold gas reservoirs (Governato et al. 2009). At the same time, however, it is also argued that the conditions that lead to the formation of a disc merger remnant are not associated with substantial SMBH growth (e.g. Hopkins & Hernquist 2009a). The dark matter halo of the simulation of Governato et al. (2009), for example, has a mass of <1012 M, while X-ray AGN have been shown to reside in haloes more massive than that (e.g. Coil et al. 2009; Gilli et al. 2009; Hickox et al. 2009; Silverman et al. 2009), where cold gas flows are expected to be suppressed because of shock heating (e.g. Croton et al. 2006; Dekel & Birnboim 2006; Cattaneo et al. 2007). Additionally, in the simulation of Governato et al. (2009) the black hole mass of the merger remnant at z= 0 is expected to be 4 × 107 M, assuming the relation between SMBH mass and bulge mass of Häring & Rix (2004), while X-ray AGN at z≈ 1 are believed to be associated with black holes more massive than that (e.g. Babić et al. 2007; Bundy et al. 2008; Coil et al. 2009; Hickox et al. 2009). AGN in disc hosts therefore pose a strong constraint on the fraction of SMBHs that are fuelled by mechanisms other than major mergers. Cosmological simulations of galaxy formation and SMBH growth will need to reproduce the observed luminosity distribution of AGN in spirals by invoking either minor interactions, bar instabilities or the reformation of discs in major mergers.

If AGN in disc galaxies are systems where the accretion on the SMBH is triggered by secular evolution or minor tidal disruptions then their estimated fraction to the XLF represents only a lower limit to the contribution of these SMBH fuelling processes to the accretion density at z≈ 1. Although bulge-dominated galaxies are likely to be the products of a major merger, the observed nuclear activity in these systems may not be associated with this past event. The AGN in some early-type galaxies, for example, may be fuelled by cold gas accretion via minor tidal interactions. The simulations of Khalatyan et al. (2008) demonstrate that this process can activate the SMBH of early-type galaxies, thereby producing AGN with luminosities similar to those probed here. Ciotti & Ostriker (2007) also showed that secular evolution in isolated elliptical galaxies could drive recycled gas to the central regions, thereby producing recursive nuclear starbursts and accretion on to the SMBH. Similarly, not all QSOs (LX > 1044 erg s−1; point sources in our sample) are likely to be the products of major mergers. Studies of such sources in the nearby Universe have revealed that some of them are hosted by non-interacting spirals (e.g. Canalizo & Stockton 2001), while more recent observations suggest that about 30 per cent of QSOs at low redshift are in disc galaxies (e.g. Guyon et al. 2006). Therefore, the fraction of AGN associated with early-type galaxies, peculiars and point sources is an upper limit to the contribution of major mergers to the accretion density.

Recent studies on star-forming galaxies also downplay the contribution of major mergers to the evolution of these systems since z≈ 1 (Bell et al. 2005; Melbourne, Koo & Le Floc'h 2005; Elbaz et al. 2007; Buat et al. 2008). It is found that about half of the starburst population at these redshifts is associated with spirals, underlining the importance of minor interactions and bar instabilities in galaxy evolution. If the buildup of the stellar mass of galaxies and the growth of SMBH at their centres are related (e.g. Gebhardt et al. 2000; Ferrarese & Merritt 2000), then the above findings support the idea that a potentially large fraction of the AGN at z≈ 1 are triggered by processes other than major mergers.

The authors wish to thank the anonymous referee for providing constructive comments and suggestions that significantly improved this paper and Philip F. Hopkins for making available unpublished results on the XLF for the stochastic fuelling of SMBH. The authors acknowledge the use of data from the TKRS (http://www2.keck.hawaii.edu/science/tksurvey/) and the ESO/GOODS project which is based on observations carried out at the VLT at the ESO Paranal Observatory under program IDs: 170.A-0788; 074.A-0709; 275.A-5060 and 171.A-3045. Support for this work was provided by NASA through the Spitzer Space Telescope Fellowship Program. The source catalogues used in the paper are available at http://www.astro.noa.gr/~age.

Based on observations obtained with MegaPrime/MegaCam, a joint project of Canada–France–Hawaii Telescope (CFHT) and CEA/DAPNIA, at the CFHT which is operated by the National Research Council (NRC) of Canada, the Institut National des Sciences de l'Univers of the Centre National de la Recherche Scientifique (CNRS) of France and the University of Hawaii. This work is based in part on data products produced at TERAPIX and the Canadian Astronomy Data Centre as part of the CFHTLS, a collaborative project of NRC and CNRS.

REFERENCES

Abraham
R. G.
Van Den Bergh
S.
Glazebrook
K.
Ellis
R. S.
Santiago
B. X.
Surma
P.
Griffiths
R. E.
,
1996
,
ApJS
 ,
107
,
1
Aird
J.
Nandra
K.
Georgakakis
A.
Laird
E. S.
Steidel
C. C.
Sharon
C.
,
2008
,
MNRAS
 ,
387
,
883
Akylas
A.
Georgantopoulos
I.
Georgakakis
A.
Kitsionas
S.
Hatziminaoglou
E.
,
2006
,
A&A
 ,
459
,
693
Babić
A.
Miller
L.
Jarvis
M. J.
Turner
T. J.
Alexander
D. M.
Croom
S. M.
,
2007
,
A&A
 ,
474
,
755
Barger
A. J.
et al.,
2003
,
AJ
 ,
126
,
632
Barger
A. J.
Cowie
L. L.
Mushotzky
R. F.
Yang
Y.
Wang
W.-H.
Steffen
A. T.
Capak
P.
,
2005
,
AJ
 ,
129
,
578
Barnes
J. E.
Hernquist
L. E.
,
1991
,
ApJ
 ,
370
,
L65
Barnes
J. E.
Hernquist
L.
,
1996
,
ApJ
 ,
471
,
115
Bell
E. F.
et al.,
2005
,
ApJ
 ,
625
,
23
Bennert
N.
Canalizo
G.
Jungwiert
B.
Stockton
A.
Schweizer
F.
Peng
C. Y.
Lacy
M.
,
2008
,
ApJ
 ,
677
,
846
Borne
K. D.
Bushouse
H.
Lucas
R. A.
Colina
L.
,
2000
,
ApJ
 ,
529
,
L77
Brinchmann
J.
et al.,
1998
,
ApJ
 ,
499
,
112
Brusa
M.
et al.,
2009
,
ApJ
 ,
693
,
8
Buat
V.
et al.,
2008
,
A&A
 ,
483
,
107
Bundy
K.
et al.,
2008
,
ApJ
 ,
681
,
931
Canalizo
G.
Stockton
A.
,
2001
,
ApJ
 ,
555
,
719
Canalizo
G.
Bennert
N.
Jungwiert
B.
Stockton
A.
Schweizer
F.
Lacy
M.
Peng
C.
,
2007
,
ApJ
 ,
669
,
801
Capak
P.
et al.,
2004
,
AJ
 ,
127
,
180
Cattaneo
A.
Teyssier
R.
,
2007
,
MNRAS
 ,
376
,
1547
Cattaneo
A.
et al.,
2007
,
MNRAS
 ,
377
,
63
Ciliegi
P.
Zamorani
G.
Hasinger
G.
Lehmann
I.
Szokoly
G.
Wilson
G.
,
2003
,
A&A
 ,
398
,
901
Ciotti
L.
Ostriker
J. P.
,
2007
,
ApJ
 ,
665
,
1038
Coil
A. L.
et al.,
2009
, preprint (arXiv:0902.0366)
Coleman
G. D.
Wu
C.-C.
Weedman
D. W.
,
1980
,
ApJS
 ,
43
,
393
Cowie
L. L.
Barger
A. J.
Bautz
M. W.
Brandt
W. N.
Garmire
G. P.
,
2003
,
ApJ
 ,
584
,
L57
Croton
D. J.
et al.,
2006
,
MNRAS
 ,
365
,
11
Davis
M.
et al.,
2003
, in
Guhathakurta
P.
, ed., SPIE Conf. Ser., Vol.
4834
, Discoveries and Research Prospects from 6- to 10-m Class Telescopes. SPIE, Bellingham, p.
161
Davis
M.
et al.,
2007
,
ApJ
 ,
660
,
L1
Dekel
A.
Birnboim
Y.
,
2006
,
MNRAS
 ,
368
,
2
Di Matteo
T.
Springel
V.
Hernquist
L.
,
2005
,
Nat
 ,
433
,
604
Dickinson
M.
et al.,
2004
,
ApJ
 ,
600
,
L99
Elbaz
D.
et al.,
2007
,
A&A
 ,
468
,
33
Faber
S. M.
et al.,
2003
, in
Iye
M.
,
Moorwood
A. F. M.
, eds, SPIE Conf. Ser. Vol.
4841
,
Instrument Design and Performance for Optical/Infrared Ground-based Telescopes
 .
SPIE
, Bellingham, p.
1657
Farrah
D.
et al.,
2001
,
MNRAS
 ,
326
,
1333
Ferrarese
L.
Merritt
D.
,
2000
,
ApJ
 ,
539
,
L9
Franceschini
A.
et al.,
2003
,
MNRAS
 ,
343
,
1181
Gabor
J. M.
et al.,
2009
,
ApJ
 ,
691
,
705
Gebhardt
K.
et al.,
2000
,
ApJ
 ,
539
,
L13
Georgakakis
A.
et al.,
2006
,
MNRAS
 ,
371
,
221
Georgakakis
A.
Rowan-Robinson
M.
Babbedge
T. S. R.
Georgantopoulos
I.
,
2007
,
MNRAS
 ,
377
,
203
Georgakakis
A.
Nandra
K.
Laird
E. S.
Aird
J.
Trichas
M.
,
2008a
,
MNRAS
 ,
388
,
1205
Georgakakis
A.
et al.,
2008b
,
MNRAS
 ,
385
,
2049
Giavalisco
M.
et al.,
2004
,
ApJ
 ,
600
,
L93
Gilli
R.
et al., 2009,
A&A
 ,
494
,
33
Governato
F.
et al.,
2009
,
MNRAS
 , in press (arXiv:0812.0379)
Guyon
O.
Sanders
D. B.
Stockton
A.
,
2006
,
ApJS
 ,
166
,
89
Häring
N.
Rix
H.-W.
,
2004
,
ApJ
 ,
604
,
L89
Hasinger
G.
,
2008
,
A&A
 ,
490
,
905
Hasinger
G.
Miyaji
T.
Schmidt
M.
,
2005
,
A&A
 ,
441
,
417
Hennawi
J. F.
et al.,
2006
,
AJ
 ,
131
,
1
Hernquist
L.
,
1989
,
Nat
 ,
340
,
687
Hickox
R. C.
Markevitch
M.
,
2006
,
ApJ
 ,
645
,
95
Hickox
R. C.
et al.,
2009
,
ApJ
 ,
696
,
891
Ho
L. C.
Filippenko
A. V.
Sargent
W. L. W.
,
1997
,
ApJS
 ,
112
,
315
Hopkins
P. F.
Hernquist
L.
,
2006
,
ApJS
 ,
166
,
1
Hopkins
P. F.
Hernquist
L.
,
2009a
,
ApJ
 ,
694
,
599
Hopkins
P. F.
Hernquist
L.
,
2009b
,
ApJ
 ,
698
,
1550
Hopkins
A. M.
Connolly
A. J.
Szalay
A. S.
,
2000
,
AJ
 ,
120
,
2843
Hopkins
P. F.
Hernquist
L.
Cox
T. J.
Di Matteo
T.
Robertson
B.
Springel
V.
,
2006
,
ApJS
 ,
163
,
1
Hopkins
P. F.
Cox
T. J.
Younger
J. D.
Hernquist
L.
,
2009
,
ApJ
 ,
691
,
1168
Ilbert
O.
et al.,
2006
,
A&A
 ,
457
,
841
Khalatyan
A.
Cattaneo
A.
Schramm
M.
Gottlöber
S.
Steinmetz
M.
Wisotzki
L.
,
2008
,
MNRAS
 ,
387
,
13
Kuhlbrodt
B.
Wisotzki
L.
Jahnke
K.
,
2004
,
MNRAS
 ,
349
,
1027
Laird
E. S.
et al.,
2009
,
ApJS
 ,
180
,
102
Le Fèvre
O.
et al.,
2004
,
A&A
 ,
428
,
1043
Li
C.
Kauffmann
G.
Heckman
T. M.
White
S. D. M.
Jing
Y. P.
,
2008
,
MNRAS
 ,
385
,
1915
Lilly
S. J.
Tresse
L.
Hammer
F.
Crampton
D.
Le Fevre
O.
,
1995
,
ApJ
 ,
455
,
108
Lotz
J. M.
Primack
J.
Madau
P.
,
2004
,
AJ
 ,
128
,
163
Lotz
J. M.
Jonsson
P.
Cox
T. J.
Primack
J. R.
,
2008a
,
MNRAS
 ,
391
,
1137
Lotz
J. M.
et al.,
2008b
,
ApJ
 ,
672
,
177
Marconi
A.
Risaliti
G.
Gilli
R.
Hunt
L. K.
Maiolino
R.
Salvati
M.
,
2004
,
MNRAS
 ,
351
,
169
Melbourne
J.
Koo
D. C.
Le Floc'h
E.
,
2005
,
ApJ
 ,
632
,
L65
Mignoli
M.
et al.,
2005
,
A&A
 ,
437
,
883
Morrison
R.
McCammon
D.
,
1983
,
ApJ
 ,
270
,
119
Myers
A. D.
Brunner
R. J.
Richards
G. T.
Nichol
R. C.
Schneider
D. P.
Bahcall
N. A.
,
2007
,
ApJ
 ,
658
,
99
Nandra
K.
Pounds
K. A.
,
1994
,
MNRAS
 ,
268
,
405
Peng
C. Y.
Ho
L. C.
Impey
C. D.
Rix
H.-W.
,
2002
,
AJ
 ,
124
,
266
Pierce
C. M.
et al.,
2007
,
ApJ
 ,
660
,
L19
Popesso
P.
et al.,
2009
,
MNRAS
 ,
494
,
443
Robertson
B.
Bullock
J. S.
Cox
T. J.
Di Matteo
T.
Hernquist
L.
Springel
V.
Yoshida
N.
,
2006
,
ApJ
 ,
645
,
986
Rowan-Robinson
M.
et al.,
2005
,
AJ
 ,
129
,
1183
Rowan-Robinson
M.
et al.,
2008
,
MNRAS
 ,
386
,
697
Sanders
D. B.
Mirabel
I. F.
,
1996
,
ARA&A
 ,
34
,
749
Schmidt
M.
,
1968
,
ApJ
 ,
151
,
393
Serber
W.
Bahcall
N.
Ménard
B.
Richards
G.
,
2006
,
ApJ
 ,
643
,
68
Silverman
J. D.
et al.,
2008
,
ApJ
 ,
679
,
118
Silverman
J. D.
et al.,
2009
,
ApJ
 ,
695
,
171
Simard
L.
et al.,
2002
,
ApJS
 ,
142
,
1
Simmons
B. D.
Urry
C. M.
,
2008
,
ApJ
 ,
683
,
644
Somerville
R. S.
Hopkins
P. F.
Cox
T. J.
Robertson
B. E.
Hernquist
L.
,
2008
,
MNRAS
 ,
391
,
481
Springel
V.
Hernquist
L.
,
2005
,
ApJ
 ,
622
,
L9
Springel
V.
et al.,
2005
,
Nat
 ,
435
,
629
Stanway
E. R.
Bunker
A. J.
McMahon
R. G.
Ellis
R. S.
Treu
T.
McCarthy
P. J.
,
2004a
,
ApJ
 ,
607
,
704
Stanway
E. R.
et al.,
2004b
,
ApJ
 ,
604
,
L13
Surace
J. A.
Sanders
D. B.
Vacca
W. D.
Veilleux
S.
Mazzarella
J. M.
,
1998
,
ApJ
 ,
492
,
116
Sutherland
W.
Saunders
W.
,
1992
,
MNRAS
 ,
259
,
413
Szokoly
G. P.
et al.,
2004
,
ApJS
 ,
155
,
271
Ueda
Y.
Akiyama
M.
Ohta
K.
Miyaji
T.
,
2003
,
ApJ
 ,
598
,
886
Vanden Berk
D. E.
et al.,
2001
,
AJ
 ,
122
,
549
Vanzella
E.
et al.,
2005
,
A&A
 ,
434
,
53
Vanzella
E.
et al.,
2008
,
A&A
 ,
478
,
83
Vogt
N. P.
et al.,
2005
,
ApJS
 ,
159
,
41
Weiner
B. J.
et al.,
2005
,
ApJ
 ,
620
,
595
Wirth
G. D.
et al.,
2004
,
AJ
 ,
127
,
3121
Wolf
C.
et al.,
2004
,
A&A
 ,
421
,
913
York
D. G.
et al.,
2000
,
AJ
 ,
120
,
1579
Younger
J. D.
Hopkins
P. F.
Cox
T. J.
Hernquist
L.
,
2008
,
ApJ
 ,
686
,
815