By stacking the X-ray emission of X-ray undetected massive galaxies it is possible to gain insight into their behaviour outside of the “transformative” and “maintenance” accretion modes. Using this information one can speculate as to how these X-ray undetected galaxies are related to the X-ray detected population. Strong stacked signals are observed in the X-ray undetected SFGs but the X-ray undetected QGs produce weak stacked signals (see Table 4.7), a result that is consistent with previous observations of massive galaxies at z ∼ 2 by Olsen et al. (2013).
The hard HR and high (for X-ray undetected sources) L2−10keV of X-ray undetected
compact SFGs suggests the presence of AGN within this population, a significant frac- tion of which are heavily obscured. These obscured AGN are likely the progenitors of the heavily obscured, highly luminous AGN observed in X-ray detected compact SFGs preparing to enter the transformative mode. Like the X-ray undetected compact SFGs, the L2−10keV of X-ray undetected extended SFGs also suggests additional contributions
from AGN. The HR of X-ray undetected extended SFGs, however, is lower than the HR of the X-ray undetected compact SFGs. This implies that the stacked emission from X-ray undetected extended SFGs is generated primarily by low luminosity unobscured AGN ac- tivity as opposed to heavily obscured AGN. These low luminosity unobscured AGN might be preparing to enter maintenance mode, with the gas that is fuelling star formation within the galaxy having just begun to trickle down and accrete onto the central AGN. The strong stacked signal of X-ray undetected SFGs obviously raises questions regarding the true AGN content of the extended SFG sample. This in turn might have implica- tions regarding the proposed maintenance mode feedback that appears to be occurring in extended galaxies.
Neither of the X-ray undetected QG subsamples produces a statistically significant signal in the hard band when stacked. As such there is nothing to suggest the presence of obscured AGN activity in either of these populations. The L0.5−2keV of both populations
do suggest a contribution from unobscured low luminosity AGN in addition to (very lim- ited) star formation. In the case of X-ray undetected compact QGs these are likely the remnants of AGN that have left the transformative mode having exhausted their fuel sup- ply and removed obscuring matter from the central regions of the host galaxy. For X-ray undetected extended QGs the emission could be from AGN that have just left either the transformative or maintenance modes, again having exhausted their supply of fuel.
It is difficult to robustly determine the precise nature of the stacked X-ray emission of massive galaxies, but it is possible to interpret them in the context of the two distinct accretion modes (transformative and maintenance) observed in the massive galaxy sample. The stacked emission of X-ray undetected compact SFGs and X-ray undetected extended SFGs is indicative of a high concentration low accretion rate AGN that have yet to enter the transformative and maintenance accretion modes respectively. Conversely the stacked
4.6 Summary 149
emission of X-ray undetected compact QGs and X-ray undetected extended QGs is con- sistent with a much lower concentration of low accretion rate AGN that may be leaving the transformative and maintenance modes respectively. In fewer words, there is evidence to support AGN in the X-ray undetected SFGs that are switching on and AGN in the X-ray undetected QGs that are switching off.
4.6
Summary
In this chapter, the X-ray properties of massive (M> 1010.5 M⊙) compact and extended
galaxies at z ∼ 2 have been investigated using 4Ms CDFS observations, finding evidence for two modes of black hole accretion within these massive galaxies. Direct X-ray detec- tions were established using a likelihood ratio matching technique and then subjected to spectral fitting analysis to obtain detailed information regarding their obscuration and X-ray luminosities. The main results are listed below:
(i) It is found that 22.0 ± 2.5% (59/268) of the massive galaxies possess X-ray counter- parts. The luminosity limit of the observations for galaxies at these redshifts is such that almost all of the X-ray detections are due to AGN activity. When the sample is subdivided according to galaxy compactness and sSFR the extended QGs have the highest detection rate (47.6 ± 10.9%) of any of the subsamples. The high X-ray detection rate of extended QGs implies they have a long duty cycle and that AGN play a role in their evolution. The AGN in compact SFGs also appear to have a long duty cycle with significant amounts of cold gas available to fuel AGN, possibly the same cold gas as is fuelling the star formation. Conversely the extended SFGs have a much lower X-ray detected fraction (14.3 ± 3.0%) and thus although there is cold gas readily available for star formation the AGN are being starved of fuel. The compact QGs also have a low X-ray detection rate (19.2 ± 5.5%) which indicates they have a short duty cycle.
(ii) Through X-ray spectral analysis of X-ray detected massive galaxies evidence is found supporting two distinct modes of AGN accretion; a “transformative mode” where AGN are heavily obscured and luminous, and a “maintenance mode” in which the AGN are unobscured and less luminous. It appears that the dominant fraction of black hole growth occurs in AGN in the transformative mode as opposed to the maintenance mode.
(iii) The two AGN accretion modes are directly related to the compactness of the host galaxy, with dense compact galaxies hosting AGN in the transformative mode and more diffuse extended galaxies hosting AGN in the maintenance mode. The high density environment provided by compact galaxies means more gas is readily avail- able to fuel AGN accretion within these galaxies. Conversely the central regions of extended galaxies would appear to be more diffuse, thus gas accretion occurs at
4.6 Summary 150
a lower rate. The majority of obscured AGN are best fit by a spherical obscura- tion model (model B) which implies the AGN being primarily obscured by the host galaxy as opposed to a local dust torus (see Figure 4.5). This is consistent with the idea that the observed column density of an AGN is dependent upon the density of the host galaxy.
(iv) The transformative mode is similar to the “quasar mode” postulated by Hop- kins et al. (2006), where a highly luminous, heavily obscured AGN quenches star- formation within the host galaxy through AGN feedback. The short duty cycle of X-ray detected compact QGs suggests they are a transitional population where the transformative mode AGN has just quenched star formation within the host galaxy. There is no evidence, however, supporting major mergers acting as trigger for transformative mode accretion.
(v) The maintenance mode is similar to “radio mode” feedback predicted by Croton et al. (2006) and Bower et al. (2006), where lower luminosity, unobscured AGN suppress star formation within their host galaxies by heating infalling gas. Extended galaxies containing AGN in the maintenance mode are thought to oscillate across the quiescent star forming boundary as infalling gas triggers star formation and AGN accretion. Accretion onto the AGN subsequently produces a gentle feedback which heats the gas and suppresses further star formation. The long duty cycle of extended QGs relative to the star forming SFGs suggests the suppression of star formation continues for extended periods.
(vi) The transformative mode could play a role in the size evolution of compact galaxies at z ∼ 2, when the galaxy merger rate is insufficient (Conselice, 2006; Newman et al., 2012). Two possible methods from the literature have been considered to explain how an AGN in the transformative mode may produce feedback which drives the size evolution of the host galaxy. The first is the feedback model of Fan et al. (2008) in which the AGN drives gas from the central regions of the galaxy, rapidly quenching star formation. The expulsion of gas from the central regions forces the host galaxy to undergo size increase in order to restore virial equilibrium, creating an extended QG. The second is a feedback model proposed by Ishibashi et al. (2013) in which the AGN produces a positive feedback, promoting star formation in the outskirts of the host galaxy. The size of the galaxy increases as a result and after several such episodes an extended SFG would be formed. The remnant AGN of either feedback process are expected to posses a lower luminosity and obscuration than its predecessor. This transformation is consistent with the change in X-ray luminosity and obscuration properties of AGN in compact galaxies compared to those in extended galaxies. Assuming AGN do play a role in the size evolution of massive galaxies, the low fraction of AGN among extended SFG greatly disfavours the Ishibashi et al. (2013) feedback mechanism. Therefore the Fan et al. (2008)
4.6 Summary 151
feedback mechanism appears to be the most likely to be driving size evolution in this massive galaxy sample.
152
Chapter 5
The efficacy of optical emission
line diagnostics for AGN at
0.3
≤ z ≤ 0.8
Overview
The research presented in this chapter concerns an investigation into the accuracy with which the colour-excitation (CEx) and mass-excitation (MEx) optical emission line galaxy diagnostic techniques identify obscured AGN populations. Candidate AGN are selected using these techniques from a sample of 1367 galaxies with DEEP2 spectroscopy and 0.3 ≤ z ≤ 0.8 in the 800ks EGS footprint. The accuracy with which these techniques select type 2 AGN relative to one another is measured using a combination of X-ray spectral fitting and X-ray stacking analyses. Both techniques select a high fraction of X-ray detected sources and the absorption corrected X-ray luminosities suggest that all these sources are AGN. The CEx selection technique identifies a high fraction of obscured AGN which is consistent with the obscured fraction of type 2 AGN observed in the local universe. The X-ray detected CEx SFGs on the other hand are primarily unobscured AGN with lower absorption corrected X-ray luminosities. The X-ray detected CEx SFGs appear to be AGN-SF composite galaxies. The MEx AGN selection identifies nearly all X-ray detected AGN but has a lower X-ray detected fraction and lower obscured AGN fraction than the CEx AGN subsample. Therefore it appears that the MEx AGN selection identifies numerous unobscured low luminosity AGN and possibly some SFGs in addition to obscured AGN. Stacking the X-ray undetected AGN reveals a hard signal with average stacked X- ray luminosities in excess of those expected from star formation. This is indicative of a large population of heavily obscured actively accreting AGN. The stacked signal of MEx AGN is softer which suggests that the fraction of obscured AGN in this sample is is less than in the CEx AGN. The CEx AGN selection appears to be superior to the MEx AGN selection when trying to specifically identify type 2 AGN. This is probably because
5.1 Introduction 153
unobscured AGN make galaxies appear bluer, biasing the CEx AGN selection against them. The galaxy mass is measured independently of the optical colour thus the same effect is not observed in the MEx AGN selection. Recalibrating the MEx AGN selection region should improve the accuracy with which it identifies type 2 AGN candidates. The [OIII] luminosity of X-ray undetected CEx and MEx AGN suggests that they are obscured as opposed to intrinsically less luminous than the X-ray detected CEx and MEx AGN. The T-ratio may also be used to identify heavily obscured AGN candidates but this technique is prone to contamination from SFGs.
5.1
Introduction
Tight relations have been observed between host galaxy bulge mass/velocity dispersion and SMBH mass (Magorrian et al., 1998; Ferrarese and Merritt, 2000; Gebhardt et al., 2000; McLure et al., 2006) which suggest there is a physical link between AGN and their host galaxies. To truly comprehend the extent of possible SMBH-galaxy co-evolution requires a complete census of AGN and their host galaxies throughout the universe. AGN are identified using their unique intrinsic properties; X-ray surveys have long proved fruitful in this endeavour because few phenomena can replicate the hard X-ray emission of AGN (Boyle et al., 1993; Brandt et al., 2001). Additionally X-rays suffer less attenuation due to obscuring gas than longer wavelength radiation allowing AGN to be detected through all but the highest column density obscurers (NH > 1024cm−2). Deep X-ray observations
can resolve the majority of AGN activity (Worsley et al., 2005), but heavily obscured and Compton thick AGN can be missed, as evidenced by the Hard X-ray background deficit (Comastri et al., 2005; Gilli et al., 2007). Thus efforts have been made to provide a more complete census of AGN using alternative techniques. Studying the ratio of high- and low-excitation emission lines is one such method.
Emission line ratio diagnostics were first presented by Baldwin et al. (1981). The aim of this research was to separate galaxies based on their different excitation mechanisms. This type of analysis is now referred to as “BPT” analysis in deference to the authors of the original paper. The work by Veilleux and Osterbrock (1987) further established this emission line diagnostic technique’s usefulness in identifying type 2 AGN. A type 2 AGN, according the unified model of AGN (Antonucci, 1993), is obscured by a local dust torus with an edge-on configuration, attenuating the broad line emission that is typically used to identify AGN at optical wavelengths. The presence of the dust torus in the line of sight to the AGN also means it is liable to attenuate X-ray emission if the column density is sufficiently high, lowering the X-ray detected fraction of type 2 AGN. Due to the configuration of the torus, however, the AGN is still capable of exciting gas clouds at larger scales triggering high-excitation narrow line emission such as [OIII] (Baldwin et al., 1981; Busko and Steiner, 1988). The BPT technique isolates these type 2 AGN by essentially comparing the strength of high-excitation line emission (e.g. [OIII]) to the strength of
5.1 Introduction 154
lower-excitation emission lines (e.g. Hβ). Sources which exhibit excess high-excitation emission line activity are likely to contain type 2 AGN. Adding a secondary emission line pairing as a proxy for metallicity (e.g. [NII] and Hα) improves the accuracy of this technique by separating galaxies that produce significant [OIII] line emission through star formation (low metallicity) from those which do not (high metallicity).
The BPT technique in principle allows obscured AGN populations that are often missed by other techniques to be identified. The redshift range of BPT selection is limited to z < 0.4 because longer wavelength emission lines (e.g. [NII] and Hα) leave the optical window. It is possible to detect these emission lines at higher redshifts using NIR spec- troscopy, as was done by Trump et al. (2011, 2013), but this is expensive and NIR spectra are still of lower quality than optical spectra. An alternative route was taken by Weiner et al. (2007), replacing the redshift limiting [NII]/Hα line pairing with H-band absolute magnitude measurements as these values are typically positively correlated. This “pseudo- BPT” analysis extends the effective redshift range to z ≤ 0.8, isolating AGN in a similar manner to the original BPT technique. The H-band magnitude is only a crude proxy of the [NII]/Hα ratio and produces a fairly coarse cut between SFG and AGN, with many AGN misclassified as SFG because their [OIII]/Hβ ratios are scattered below the AGN boundary. The promise exhibited by this technique, however, inspired the development of more refined methods.
The colour-excitation (CEx) technique, pioneered by Yan et al. (2011), replaces the [NII]/Hα ratio with the rest-frame U − B colour. Yan et al. (2011) find the rest-frame U − B colour correlates positively with both bulge mass and metallicity and thus is a suitable replacement for the [NII]/Hα ratio. The CEx selection is empirically calibrated using BPT classified sources at low redshifts (z ∼ 0.01) to split galaxies into AGN and star forming galaxies (SFG). The selection is then applied to galaxies at 0.3 ≤ z ≤ 0.8 in the EGS field and matched to X-ray detected sources from the EGS 200ks survey (Laird et al., 2009). At higher redshifts, the CEx AGN exhibit a high X-ray detection rate and the majority of the X-ray detected sources are unambiguous AGN (LX > 1042erg
s−1). Comparison of the expected and detected fraction of X-ray sources suggests that
the column densities of the CEx AGN are similar to those of type 2 AGN observed in the local universe. Consequently there are probably numerous heavily obscured and Compton thick AGN within the CEx AGN subsample that are producing some X-rays but lie below the detection threshold of the EGS 200ks survey. This technique is liable however to miss AGN in blue galaxies, as evidenced by the presence of X-ray detected AGN-SF composites in the CEx SFG sample.
The mass-excitation (MEx) technique, developed by Juneau et al. (2011) adopts a dif- ferent tactic, replacing the [NII]/Hα ratio with galaxy stellar mass. This approach hinges on the assumption that the stellar mass of a galaxy is correlated with the [NII]/Hα ratio by the empirically derived mass-metallicity relation (e.g. Tremonti et al. 2004; Savaglio et al. 2005; Kewley and Ellison 2008). Similar to the CEx technique, the MEx technique is
5.2 Data 155
calibrated using low redshift optical sources (z ∼ 0.01) in order to optimise the agreement with BPT AGN and SFG galaxy classifications and minimise contamination. The MEx technique also introduces a classification designed to target the BPT-transition galaxies. The MEx technique is then applied to data taken in the GOODS-N and EGS fields and matched to 2Ms and 200ks X-ray observations respectively. The MEx AGN and MEx transition galaxy regions identify the majority of unambiguous X-ray AGN (LX > 1042erg
s−
1 or HR> −0.1) in the sample. The MEx transition galaxy region, in particular, ap- pears to preferentially select AGN-SF composite galaxies that were simply classified as SFGs by the CEx technique. The stacked X-ray emission of X-ray undetected MEx AGN and transition galaxies possesses a hard, flat signal consistent with some of these galax- ies possessing an actively accreting SMBH. Furthermore T-ratio analysis (Bassani et al., 1999) identifies 33 absorbed AGN candidates, of which 13 are not detected individually in the X-ray observations. Stacking the 13 X-ray undetected absorbed AGN candidates produces a hard signal, again consistent with accretion onto an SMBH.
The work presented in this chapter aims to investigate the efficiency with which the CEx and MEx AGN selection techniques identify obscured AGN within a sample of galaxies at 0.3 ≤ z ≤ 0.8 that lie in the EGS 800ks footprint (Nandra et al. in prep). The obscured AGN activity within the MEx and CEx AGN subsamples is assessed using X-ray spectral fitting for X-ray detected galaxies and X-ray stacking for X-ray undetected galaxies. These are compared to the obscured AGN activity within SFGs, QGs and a control sample of X-ray detected galaxies from the EGS 800ks footprint that have not been covered by optical spectroscopic surveys. A comprehensive analysis of the similarities and differences of the CEx and MEx selection techniques is then undertaken. Finally the role of [OIII] luminosity as an intrinsic measure of the AGN X-ray luminosity is investigated. Throughout this chapter a standard, flat ΛCDM cosmology with ΩΛ = 0.7
and H0 = 70 km s−1 Mpc−1 is assumed.
5.2
Data
5.2.1 Optical data
The optical data used in this chapter was originally presented in Yan et al. (2011). The majority of the optical spectroscopy was taken as part of the DEEP2 galaxy redshift survey using the DEIMOS spectrograph on the Keck-II telescope (Davis et al., 2003). The spectral range of these observations is approximately 6500-9200 Angstroms with a resolution of R ∼ 5000. Follow-up observations were made for 498 X-ray sources from the EGS 200ks point source catalogue that had optical counterparts because they were missed in the original DEEP2 survey. These observations yielded 288 reliable redshifts for the X-ray sources; 265 galaxies. The rest-frame U − B colours of the galaxies were derived