UNCOVERING OBSCURED ACTIVE GALACTIC NUCLEI IN HOMOGENEOUSLY SELECTED SAMPLES OF SEYFERT 2 GALAXIES

A subset of galaxies are active, indicating that the central supermassive black hole is accreting material. Within such active galactic nuclei (AGNs), this accretion disk is in turn surrounded by an obscuring medium of dust and gas, thought to have a toroidal geometry (e.g., Antonucci 1993; Urry & Padovani 1995). In Type 1 AGNs, the system is oriented such that the line of sight intercepts the opening of this "torus," exposing the accretion disk and broad load region (BLR). Conversely, this central region is blocked in Type 2 AGNs, as the system is inclined such that the line of sight is through the obscuring medium. In these obscured AGNs, the narrow line region (NLR) is visible as well as scattered and reflected emission from the central engine.

Previous studies have shown that obscured AGNs constitute at least half of the local population (Risaliti et al. 1999; Bassani et al. 1999; Guainazzi et al. 2005b). Obscuration can result from the putative torus or even the host galaxy where dust from nuclear star formation processes (e.g., Ballantyne 2008), extranuclear dust (Rigby et al. 2006) or dilution of AGN emission by the galaxy (Trump et al. 2009) can attenuate optical signatures of AGN activity. However, in this study we focus on "toroidal" obscured sources (where the absorption is intrinsic to the AGN, on the scale of a few parsecs, enshrouding the BLR) since a substantial fraction of heavily obscured, Compton-thick (column density N_H⩾ 1.5 × 10²⁴ cm⁻²) AGNs are often invoked to explain the unresolved portions of the X-ray background at 30 keV (e.g., Gilli et al. 2007). An accurate census of these AGNs is necessary to constrain X-ray background synthesis models (e.g., Treister et al. 2009) and studying their properties are crucial in understanding the full AGN population. Unfortunately, such obscured AGNs are missed in 2–10 keV X-ray surveys as absorption and Compton-scattering severely attenuate this X-ray emission.

Selecting AGN samples based on intrinsic AGN flux proxies (F_intrinsic), which are ideally unaffected by the amount of toroidal obscuration present, is therefore necessary to uncover Compton-thick AGNs. Such diagnostics include emission lines which are primarily ionized by accretion disk photons and are formed in the NLR, making them not subject to torus obscuration, and include the optical [O iii] 5007 line (e.g., Heckman et al. 2005) and the infrared [O iv] 25.89 μm line (e.g., Meléndez et al. 2008; Diamond-Stanic et al. 2009; Rigby et al. 2009). The obscuring medium absorbs continuum photons and re-radiates them in the mid-infrared (MIR). This emission constitutes approximately 20% of the bolometric luminosity in most Type 1 and Type 2 AGNs (Spinoglio & Malkan 1989), making it another isotropic indicator of intrinsic AGN flux. Follow-up X-ray observations can then reveal which sources are potentially Compton-thick.

Several studies have used infrared emission to locate such obscured AGNs (Daddi et al. 2007; Fiore et al. 2009). These studies have selected sources with infrared emission in excess of that attributable to star formation, indicating the presence of an AGN. Using either Chandra and/or XMM-Newton spectra (Fiore et al. 2009) or stacked X-ray spectra for non-detections (Daddi et al. 2007; Fiore et al. 2009), the column densities of these sources were estimated to constrain the Compton-thick fraction. However, these studies focus on high redshift sources (z = 0.7–2.5), rely on assumptions of the source spectrum shape (to convert count rate to flux for detections or hardness ratio to flux for stacked spectra) and in the case of stacked sources, only characterize the aggregate population rather than individual sources. These analyses are useful in estimating a potential Compton-thick fraction at early times in the universe, but local sources generally have the advantage of X-ray spectra with higher signal-to-noise, where each source can be individually analyzed and the spectra better characterized to more robustly constrain obscuration diagnostics.

We have undertaken such an analysis on two homogeneous samples of local (z< 0.15) Seyfert 2 galaxies (Sy2s) selected based on intrinsic flux proxies: an MIR sample from the original IRAS 12 μm survey (Spinoglio & Malkan 1989) and an [O iii]-selected sample culled from the Sloan Digital Sky Survey (SDSS; LaMassa et al. 2009). The Chandra and/or XMM-Newton spectra of these sources were fit in a homogeneous and systematic manner. However, column densities derived from spectral fitting in the 2–10 keV band are highly model dependent and thus may not always reflect the intrinsic toroidal absorption. Other proxies are therefore necessary to identify potentially Compton-thick sources. As the 2–10 keV X-ray emission is suppressed in absorbed sources, the ratio of this emission to intrinsic flux indicators can probe the amount of obscuration. In Compton-thick sources, the ratio of the 2–10 keV X-ray flux (F_2–10 keV) to F_intrinsic is about an order of magnitude or lower than what is observed in unobscured sources (e.g., Bassani et al. 1999; Cappi et al. 2006; Panessa et al. 2006; Meléndez et al. 2008). X-ray spectral signatures, most notably the equivalent width (EW) of the neutral Fe Kα line at 6.4 keV, can also aid in uncovering heavily obscured sources. As the EW is measured against a suppressed continuum, it rises with increasing column density, reaching values of several hundred eV to over 1 keV in Compton-thick sources (e.g., Levenson et al. 2002). Similar to LaMassa et al. 2009, we use F_2–10 keV/F_intrinsic and the Fe Kα EW as obscuration diagnostics in this work.

This paper is organized as follows: we describe the sample selection in Section 2 and the spectral fitting procedures in Section 3. We then estimate the potential Compton-thick population using ancillary optical and IR data and discuss the various possible absorber geometries revealed by our obscuration diagnostics. Utilizing IR data to parameterize host galaxy characteristics, namely star formation processes, and AGN activity, such as intrinsic AGN luminosity, central black hole mass, and Eddington ratio, we investigate whether Compton-thick sources trace a unique population. We also comment on the feasibility of detecting these sources in an upcoming hard X-ray (5–80 keV) mission, the Nuclear Spectroscopic Telescope Array (NuSTAR).

Our combined sample consists of 28 MIR selected and 17 [O iii] selected Sy2s. The MIR sample is a subset of the 31 Sy2s from the original IRAS 12 μm survey (Spinoglio & Malkan 1989),⁵ which was drawn from the IRAS point-source catalog (version 2). The Sy2s in the 12 μm sample were selected via a weak color cut (i.e., 12 μm flux less than the 60 or 100 μm flux) and is complete to a flux-density limit of 0.3 Jy at 12 μm, with latitude |b|>25° which avoids Galactic contamination (Spinoglio & Malkan 1989). Archival Chandra and XMM-Newton data exist for 28 of these 31 sources.

The [O iii]-selected Sy2s were culled from the main Galaxy sample (Strauss et al. 2002) in the SDSS Data Release 4 (Adelman-McCarthy et al. 2006). Using the SDSS spectra, Sy2s were identified using the diagnostic line ratio plot of [O iii]/Hβ and [N ii]/Hα (BPT diagram; Baldwin et al. 1981) and the Kauffmann et al. (2003) and Kewley et al. (2006) demarcations which distinguish Sy2s, composite systems, and star-forming galaxies. The 20 local (z< 0.15) Sy2s with [O iii] 5007 Å flux >4 × 10⁻¹⁴ erg s⁻¹ cm⁻² that lie within the AGN locus of the BPT diagram were selected to comprise this sample. Of these 20 Sy2s, 2 had archival XMM-Newton observations and we were awarded XMM-Newton time for another 15 sources. The X-ray analysis for these 17 [O iii] selected Sy2s was presented in LaMassa et al. (2009) and is not replicated here; we utilize the results of that study (2–10 keV X-ray fluxes, Fe Kα EWs, etc.) in this work. Though both original samples are complete, since X-ray data only exist for 28/31 and 17/20 sources from the 12 μm and [O iii] samples, respectively, our resultant sample for X-ray analysis is nearly complete.

We note that this study has the advantage of analyzing samples of Sy2s selected via two different techniques which can mitigate biases from any individual selection criterion. For instance, dusty host galaxies can attenuate the optical emission lines used to identify AGNs (and thus potentially miss galaxy obscured AGNs), whereas MIR selection can isolate such sources. Star formation processes in the host galaxy can also enhance MIR, limiting its usefulness as an intrinsic indicator of pure AGN flux. However, such effects are minimal in this study as all but two MIR identified Sy2s live in the AGN locus of the BPT diagram and those two sources inhabit the composite (AGN and star-forming) locus and would not be optically identified as pure star-forming galaxies. We also note that many low-luminosity AGNs and some quasars lack the IR signature of a torus (Ho 2008; Hao et al. 2010, respectively), and thus MIR selection is not a useful tool for investigating such AGNs or their obscured counterparts. We explore the issues of selection effect biases further in Section 4.2.

We analyzed the available archival Chandra and XMM-Newton observations for the 12 μm sources with XAssist (Ptak & Griffiths 2003). This program runs the appropriate Science Analysis Systems (SAS) tasks to filter the data and clean for flaring as well as extract spectra and associated response files for user-defined sources. Table 1 lists the X-ray observations used in this analysis, including the ObsIDs and net exposure times after filtering.

Twenty-five out of 28 sources were detected at the 3σ level or greater in the 0.5–10 keV band. One (NGC 5193) was detected in the soft band (0.5–2 keV) and we were thus able to fit this part of the spectrum. We obtained upper 2–10 keV flux limits on this source and the two undetected sources (F08572+3915 and NGC 7590), discussed in detail below.

We used simple absorbed power-law models to fit the spectra for the detected sources, which may not accurately represent the complex geometry of these systems. However, our main goal is to apply a systematic and homogeneous analysis of the spectra in a similar manner as LaMassa et al. (2009) to derive an observed X-ray flux, and where possible, EW of the Fe Kα line. More extensive X-ray modeling of several sources has been investigated in detail in the literature (e.g., see Brightman & Nandra 2010 for more detailed X-ray modeling of the extended 12 μm sample) and we do not intend to replicate previously published work. In Appendix A, we discuss individual sources, compare our derived parameters with those quoted in the literature, and comment on the impact more complex models have on such parameters. We find that in 18/23 sources, we recover consistent (within 1σ) observed X-ray fluxes and Fe Kα EW values as more complex models. This work also represents the first analysis for a handful of data sets (i.e., Chandra spectrum of IC 5063, XMM-Newton 2004 EPIC spectra of NGC 7172, and XMM-Newton EPIC spectra of NGC 7674).

3.1. Fitting Spectra from Multiple Observations

3.2. Spectral Models

We initially fit all spectra with an absorbed power-law model. Most spectra (18/26) had an adequate number of detected photons to be grouped by a minimum of 15 counts bin⁻¹ without loss of spectral information and were thus analyzed with χ²; the remaining 8 (NGC 3982, NGC 4501, TOLOLO 1238-364, NGC 4968, NGC 5135, NGC 5953, NGC 6890, and NGC 7130) were analyzed with the Cash statistic (C-stat; Cash 1979) and binned by 2–3 counts as XSpec handles slightly binned spectra better than unbinned when using C-stat (Teng et al. 2005). With the exception of seven sources (NGC 1667, NGC 3982, NGC 4501, TOLOLO 1238-364, NGC 4968, NGC 5953, and Arp 220), a second power-law component was needed to accommodate the data (i.e., phabs₁*(pow₁ + phabs₂*pow₂)). The two power-law indices (Γ) were tied together and the normalizations and absorption components were fit independently. Such a model represents a partial covering geometry with the first power law denoting the soft scattered and/or reflected AGN continuum and the second component describing the absorbed transmitted emission.

Residuals below 2 keV were present in many of the sources, suggesting emission in excess of the scattered AGN continuum. This excess is likely due to thermal emission from hot gas related to star formation processes, and consistent with LaMassa et al. (2009), we used a thermal model (APEC in XSpec) with abundances fixed at solar to fit this emission. According to the f-test, addition of this component improved the fit at greater than the 3σ level over the best-fit single or double power-law model for 16/25 sources.⁶ In Table 2, we present the best-fit parameters from the APEC plus power-law models, along with the χ² values from the single absorbed power-law fit, and where applicable, the double absorbed power-law fit. We required a lower limit on the first absorption component (N_H,1) to be equal to the Galactic absorption. In some cases, the best-fit absorption was equal to the Galactic N_H and we subsequently froze N_H,1 to the Galactic value for these sources. We were only able to obtain an upper limit on N_H,1 for three sources (Mrk 463, NGC 6890, and NGC 7130), as the lower error bound pegged at the Galactic absorption; the upper 90% limit is thus listed in Table 2. We also quote the 90% upper limit on kT for the six cases where the lower error on the temperature pegged at the limit of 0.1 keV (F05189-2524, NGC 3982, the Chandra observation of NGC 4388, NGC 4968, NGC 5347, and the Chandra observation of NGC 7582). We included Gaussian components to accommodate the Fe Kα emission when present (see below) and additional Gaussian components for other emission features in NGC 1068 and NGC 7582 (see Appendix A). In Table 3, we list the best-fit parameters for the absorbed single/double power-law fit for NGC 5953 and the nine sources which according to the f-test are not statistically significantly improved (⩾3σ) by adding the APEC component and are therefore better described by the simpler single/double power-law model (NGC 424, the Chandra observation of NGC 4388, NGC 4968, NGC 5135, NGC 5347, NGC 6890, IC 5063, the Chandra observation of NGC 7582 and NGC 7674).

Table 3. Power-law Model Parameters

Galaxy	NH,1	Γ	NH,2	χ2
	(1022 cm−2)		(1022 cm−2)	(dof)
NGC 0424a	0.07+0.03−0.03	2.97+0.27−0.26	16.9+6.0−3.0	273.8 (173)
NGC 4388 (Chandra)	0.22+0.24−0.15	0.38+0.39−0.36	23.3+3.5−3.1	121.7 (94)
NGC 4968b,c	0.08	1.94+0.14−0.13	...	337.8 (270)
NGC 5135b	0.05	2.75+0.11−0.10	118+82−60	200.8 (134)
NGC 5347b	0.02	1.41+0.24−0.22	56.2+31.9−22.9	36.9 (24)
NGC 5953b,c,d	0.03	2.10+0.63−0.65	...	39.9 (21)
NGC 6890c	0.21+0.11−0.09	3.86+0.75−0.64	18.9+16.5−11.0	171.3 (150)
IC 5063b,e	0.06	1.48+0.26−0.25	20.5+1.4−1.4	135.6 (119)
NGC 7582 (Chandra)e	<0.23	1.63+0.50−0.40	18.8+2.9−2.1	117.5 (83)
NGC 7674b	0.04	2.86+0.12−0.11	36.9+12.4−7.7	129.6 (74)

Notes. ^aBest-fit parameters between Chandra and XMM-Newton observations are consistent. ^bBest-fit N_H was same as Galactic value and therefore frozen at this value. ^cUsed C-stat. ^dOnly detected in soft band. ^eUsed the pileup model.

Download table as:  ASCIITypeset image

We list the observed 2–10 keV X-ray flux from these best-fit models in Table 4. For the cases where addition of the APEC model improved the fit, we excluded this component when deriving the X-ray flux. The flux was averaged among multiple observations when these observations were consistent. For variable sources, the flux is listed independently for each observation. For Arp 220, only the absorption varied between the XMM-Newton and Chandra observations, which had a negligible impact on the flux. We therefore averaged the XMM-Newton and Chandra fluxes for this source. We note that NGC 7582 has a higher observed Chandra flux, compared to the XMM-Newton fluxes, despite the smaller Chandra spectral extraction area; aperture effects could contribute to the lower Chandra flux (compared with XMM-Newton) for NGC 4388.

Table 4. 2–10 keV X-ray Flux and Luminosity for 12 μm Sample

Galaxy	F2–10 keV	log L2–10 keV	Comments
	(10−13 erg s−1 cm−2)	(erg s−1)
NGC 0424	11.5+6.4−3.7	41.56+0.19−0.17
NGC 1068	54.2+110−32.0	41.27+0.48−0.39
NGC 1144	33.4+36.6−12.9	42.77+0.32−0.21
NGC 1320	3.84+2.83−1.39	40.83+0.24−0.20
NGC 1386	1.55+0.42−0.50	39.48+0.10−0.17	Chandra observation
NGC 1667	0.43+0.11−0.11	40.33+0.10−0.13
F05189-2524	23.5+5.5−4.9	43.00+0.09−0.10	Chandra observations
F08572+3915	<1.26	<42.02
NGC 3982	0.56+1.28−0.39	39.28+0.52−0.52	Chandra observation
NGC 4388	74.6+88.5−38.5	42.01+0.34−0.32	Chandra observation
	86.9+28.6−17.7	42.08+0.12−0.10	XMM-Newton 2002 Jul observation
	244+76−47	42.53+0.12−0.09	XMM-Newton 2002 Dec observation
NGC 4501	1.07+0.73−0.51	40.17+0.23−0.28	Chandra observation
TOLOLO 1238-364	1.21+0.31−0.26	40.50+0.10−0.11
NGC 4968	2.08+0.26−0.26	40.65+0.05−0.06
M-3-34-64	32.5+3.1−3.1	42.31+0.04−0.04
NGC 5135	2.31+0.98−1.68	40.99+0.15−0.56
NGC 5194	1.04+2.28−0.73	38.95+0.50−0.53
NGC 5347	2.58+1.20−1.41	40.55+0.17−0.34
Mrk 463	2.95+1.84−0.82	42.23+0.21−0.14	Chandra observation
NGC 5506	725+68−79	42.75+0.04−0.05	2001 & 2004 Jul observations
	1113+59−59	42.94+0.02−0.02	2002, 2004 Aug, 2008 and 2009 observations
NGC 5953	<0.51	<39.73
Arp 220	1.07+0.18−0.16	40.88+0.07−0.07
NGC 6890	1.20+4.01−0.88	40.22+0.6400.57
IC 5063	134+73−47	42.55+0.19−0.19
NGC 7130	2.07+2.09−1.04	41.06+0.30−0.30
NGC 7172	517+43−40	42.96+0.03−0.03	2007 observation
	234+19−18	42.61+0.03−0.03	2002 and 2004 observations
NGC 7582	21.1+1.7−1.8	41.05+0.03−0.04	2005 XMM-Newton observation
	38.6+3.0−3.1	41.32+0.03−0.04	2001 XMM-Newton observation
	164+263−87	41.95+0.42−0.33	Chandra observations
NGC 7590	<2.72	<40.17
NGC 7674	5.71+3.05−1.69	42.03+0.19−0.15

Download table as:  ASCIITypeset image

In Figure 1, we plot the spectra with the best-fit models. As many sources have multiple observations, we plot only one spectrum per observation, generally using the PN spectrum for XMM-Newton observations unless the MOS spectrum had better signal-to-noise. Though we report the flux of the Chandra observations only for NGC 1386, F05189-2524, and Mrk 463, we plot both the XMM-Newton and Chandra spectra to illustrate how the XMM-Newton spectra helped to constrain the fit.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.3. Pileup

3.4. Upper Limits

3.5. Fe Kα

With the observed X-ray flux and Fe Kα EW constrained, we can determine the distribution of the amount of 2–10 keV attenuation associated with the obscuring torus. As both the 12 μm and [O iii] samples were selected on intrinsic AGN properties, such a percentage might represent an unbiased estimate for the global AGN population. Similar to LaMassa et al. (2009), we also explore if the fitted column densities agree with the proxies we use for AGN obscuration: if the emission is seen primarily via scattering and/or reflection, do the fitted N_H values recover the intrinsic absorption? The obscuration flux diagnostics and Fe Kα EWs also provide clues as to the obscuration geometry in these sources. We compare host galaxy and AGN properties with Compton-thick diagnostics to determine if sources with heavy absorption trace unique populations from their less obscured counterparts. Finally, as higher energy (>10 keV) observations are necessary to confirm a source as Compton-thick, we comment on the detectability of these Sy2s by NuSTAR, an upcoming hard X-ray mission.

4.1. Obscuration Diagnostics

As fitted column densities are model dependent and could be unreliable, we use other proxies to investigate the amount of toroidal absorption in these systems, including the ratio of the observed X-ray flux to the inherent AGN flux. We consider three diagnostics for intrinsic AGN power (F_intrinsic): the [O iii]λ5007 line, the [O iv]25.89 μm line, and the mid-infrared (MIR) continuum. The [O iii] and [O iv] lines are primarily ionized by the central engine, and as they form in the NLR, are not subject to torus obscuration. The MIR emission results from the reprocessing of the AGN continuum by the dusty obscuring medium. We use the flux at 13.5 μm, averaged over a 3 μm window, as F_MIR since this region is free from strong emission lines and absorption features. These fluxes are published in LaMassa et al. (2009, 2010) and are not replicated here. As these proxies are to first order unaffected by the obscuring medium, whereas the 2–10 keV X-ray flux is attenuated due to absorption and possibly Compton-scattering, the ratio of the X-ray flux to these tracers of intrinsic AGN power can probe the amount of obscuration present and has been used extensively in previous studies (e.g., Bassani et al. 1999; Heckman et al. 2005; Cappi et al. 2006; Panessa et al. 2006; Meléndez et al. 2008; LaMassa et al. 2009). We list the values of these obscuration diagnostic flux ratios in Table 6. There are, however, several limitations to using the [O iii] and MIR fluxes in tracing the intrinsic AGN flux: the [O iii] flux could be heavily affected by dust in the host galaxy and star formation processes can contaminate the MIR flux (see LaMassa et al. 2010 for a comparison between $F_{{\rm MIR}}/F_{[{\rm O\,\mathsc{iii}}]}$ between the two samples). In LaMassa et al. (2010), we noted that applying the standard R = 3.1 reddening correction utilizing the Balmer decrement introduced errors that did not better recover the intrinsic [O iii] emission for the 12 μm sample, likely due to uncertainties in the Hβ measurements from the literature. Due to uncertainties in correcting the [O iii] and MIR fluxes for contamination, we use the observed parameters, with the caveat that these may not accurately probe intrinsic AGN emission for some sources. We discuss the implications of such biases below.

Table 6. Obscuration Diagnostic Ratios

Galaxy	log($\frac{F_{2\hbox{--}10\;{\rm keV}}}{F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}}$)	log($\frac{F_{2\hbox{--}10\;{\rm keV}}}{F_{[{\rm O\,\mathsc{iv}}]}}$)	log($\frac{F_{2\hbox{--}10\;{\rm keV}}}{F_{{\rm MIR}}}$)
NGC 0424	0.22	0.66	−2.20
NGC 1068	−0.40	−0.51	...
NGC 1144	1.89	1.83	−1.11
NGC 1320	0.44	0.25	−2.32
NGC 1386	−0.71	−0.72	−2.63
NGC 1667	−0.17	0.18	−2.81
F05189-2524	1.19	<0.53	−1.53
F08572+3915	>2.02	...	>−2.52
NGC 3982	−0.55	0.14	−2.63
NGC 4388	0.97	0.45	−0.94
	1.04	0.52	−0.87
	1.48	0.97	−0.42
NGC 4501	0.46	0.60	−2.12
TOLOLO 1238-364	−0.58	−0.09	−2.99
NGC 4968	0.07	−0.10	−2.57
M-3-34-64	0.33	0.55	−1.56
NGC 5135	−0.21	−0.39	−2.26
NGC 5194	0.31	−0.42	−2.89
NGC 5347	0.77	0.61	−2.19
Mrk 463	−0.28	−0.27	−2.36
NGC 5506	1.76	1.53	−0.37
	1.94	1.71	−0.19
NGC 5953	>0.03	>−0.49	>−2.91
Arp 220	1.77	<−0.92	−2.84
NGC 6890	−0.29	0.17	−2.53
IC 5063	0.88	1.14	−1.03
NGC 7130	−0.22	0.12	−2.33
NGC 7172	3.15	2.17	0.22
	2.80	1.83	−0.13
NGC 7582	0.61	0.38	−1.38
	0.88	0.65	−1.12
	1.50	1.27	−0.49
NGC 7590	>0.97	>0.88	>−1.82
NGC 7674	−0.10	0.19	−2.29

Download table as:  ASCIITypeset image

We plot the distributions of $F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$, $F_{2\hbox{--}10\;{\rm keV}}/\break F_{[{\rm O\,\mathsc{iv}}]}$, and F_2–10 keV/F_MIR in Figure 2 where the red dashed histogram represents the [O iii] sample, the dark blue histogram denotes the non-variable 12 μm sources and the cyan histogram reflects the variable 12 μm sources, using the average X-ray flux among the multiple observations for each source. A wide range of values is evident in all three plots. We compared our values with Sy1 sources, with the average flux ratio and spread delineated by the gray shaded regions in Figure 2. The Sy1 comparison sample are culled from: (1) Heckman et al. (2005) (heterogeneous [O iii]-bright sample, log (〈$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}\rangle$) = 1.59 ± 0.49 dex), (2) Diamond-Stanic et al. (2009) (drawn from the revised Shapley–Ames catalog, log (〈$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}\rangle$) = 1.92 ± 0.60 dex), and (3) Gandhi et al. (2009) (where F_MIR is calculated at 12.3 μm with VISIR Lagage et al. (2004) observations of Sys selected from Lutz et al. (2004) and those with existing or planned hard (14–195 keV) X-ray observations, log (〈F_2–10 keV/F_MIR〉) = −0.34 ± 0.30 dex). We note that Gandhi et al. (2009) report absorption-corrected X-ray luminosity whereas the other Sy1 comparison samples utilize the observed luminosity. This correction shifts the F_2–10 keV/F_MIR Sy1 ratios to higher values, though such a correction could be expected to be minimal for Type 1 AGNs which are thought to be largely unobscured. Also, not correcting [O iii] flux for reddening and MIR flux for starburst contamination could possibly result in obscuration diagnostic ratios that are larger or smaller, respectively, and though this affects several individual galaxies with large amounts of dust and/or greater star formation activity, no such systematic trends for the sample as a whole are evident. Yaqoob & Murphy (2010) have demonstrated that the ratio of F_2–10 keV/F_MIR is more sensitive to the X-ray spectral slope and covering factor of the putative torus, rather than column density, indicating that a low ratio does not necessarily imply a Compton-thick source. However, we find all three obscuration diagnostics to agree: the majority of Sy2s have ratios an order of magnitude or lower than their Sy1 counterparts, which may indicate Compton-thick absorption.

Zoom In Zoom Out Reset image size

This trend is further illustrated by Figure 3, which plots the observed X-ray luminosity as a function of intrinsic AGN luminosity proxies, with the best-fit relationship for Sy1s overplotted. Here, the red triangles represent the [O iii]-sample, the blue diamonds denote the non-variable 12 μm sources and the cyan diamonds reflect the variable 12 μm sources, with the individual X-ray fluxes (see Table 4) plotted for each variable source and connected by a solid line. The relationship for the Sy1 sources was calculated by multiple linear regression (i.e., REGRESS routine in IDL) for the Heckman et al. (2005) and Diamond-Stanic et al. (2009) samples; for the MIR relationship, we utilized the best-fit parameters from Gandhi et al. (2009) for their Sy1 subsample. The majority of Sy2s lie well below the relations for Sy1s, demonstrating that these Type 2 AGNs have weaker observed X-ray emission.

Zoom In Zoom Out Reset image size

As the X-ray and optical and IR observations were not carried out simultaneously, it is possible that variability in the source could be responsible for the disagreements between the X-ray flux and intrinsic flux proxies. Such a scenario can be realized if the X-ray observations are made after the central source has "shut-off" (postulated to explain the discrepancy between the Type 1 optical spectrum yet reprocessing-dominated X-ray spectrum for NGC 4051, see Matt et al. 2003b, and references therein), or the converse, where optical observations are made during a sedentary state and X-ray observations catch the source in the active state (e.g., Guainazzi et al. 2005a). Though we cannot rule out variability as contributing to the discrepancy between the X-ray luminosity and intrinsic AGN luminosity proxies for any individual source, such an effect cannot be responsible for the overall trend in this sample. Variability in Sy1 samples contributes to the dispersion in L_2–10 keV/L_isotropic ratios, yet they exhibit systematically higher X-ray luminosity (normalized by intrinsic AGN power) than Sy2s (Figures 2 and 3). This is confirmed by two-sample tests where we employed survival analysis (ASURV Rev 1.2; Isobe & Feigelson 1990; LaValley et al. 1992; Feigelson & Nelson 1985 for univariate problems) to account for upper limits in the X-ray flux. The Sy1 and Sy2 obscuration diagnostic ratios differ at a statistically significant level (⩽1 × 10⁻⁴ probability that they are drawn from the same parent population), which would not be expected if variability was the main driver for the discrepancy between intrinsic AGN flux and observed X-ray flux.

We have demonstrated that the majority of Sy2s in our samples are underluminous in X-ray emission as compared to Sy1s, but is this trend due to obscuration or inherent X-ray weakness? The EW of the neutral Fe Kα line can differentiate between these two possibilities and is thus another obscuration diagnostic. In heavily obscured sources, the AGN continuum is suppressed, whereas the Fe Kα line is viewed in reflection, leading to a large Fe Kα EW (several hundred eV to several keV, e.g., Ghisellini et al. 1994; Levenson et al. 2002). In Figure 4, Fe Kα EW is plotted as a function of obscuration diagnostic ratios. A clear anti-correlation is present which is statistically significant according to survival analysis (Isobe et al. 1986 for bivariate problems): we obtain Spearman's ρ values of −0.648, −0.657, and −0.645 for Fe Kα EW versus $F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$, $F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$, and F_2–10 keV/F_MIR, respectively. These best-fit correlations are overplotted for each relation in Figure 4. The decrease of observed X-ray flux, normalized by intrinsic AGN flux, with increasing Fe Kα EW indicates that obscuration is responsible for attenuating X-ray emission in these Sy2s. These results are consistent with the three-dimensional diagnostic diagram of Bassani et al. (1999) which shows a correlation between Fe Kα EW and column density which anti-correlates with $F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm corr}}$ (where $F_{[{\rm O\,\mathsc{iii}}],{\rm corr}}$ is the redenning corrected [O iii] flux).

Zoom In Zoom Out Reset image size

Not only do a majority of this combined Sy2 sample exhibit trademarks of Compton-thick obscuration (an order of magnitude lower F_2–10 keV/F_isotropic ratios than Sy1s and large Fe Kα EW values), but a wide range of these diagnostic values are evident. No clear separation exists between Compton-thick and Compton-thin subpopulations. Also, though the diagnostic flux ratios generally point to the same sources as having Compton-thick obscuration, not all three ratios agree for a handful of sources (e.g., F05189-2524, NGC 5347, Arp 220, NGC 4388, and NGC 7582): some ratios indicate a Compton-thin source whereas others suggest Compton-thick. As the various intrinsic AGN indicators exhibit some scatter in inter-comparisons (see, e.g., LaMassa et al. 2010), a spread in F_2–10 keV/F_isotropic values is expected. For F05189-2524, NGC 5347, and Arp 220, this discrepancy could be due to dust in the host galaxy affecting the [O iii] line, as mentioned above and/or large amounts of dust in the host galaxy boosting the MIR flux. The 2005 XMM-Newton observation of NGC 7582 has an $F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$ value marginally higher than the nominal Compton-thick/Compton-thin boundary, so the three flux ratio diagnostics may be considered to agree. However, the biases discussed previously in the observed [O iii] flux and MIR flux cannot account for the disagreement of the diagnostic flux ratios in the Chandra and July XMM-Newton observations of NGC 4388 and the 2001 XMM-Newton observation of NGC 7582, where $F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$ points to the sources being Compton-thick at these stages, but the other ratios suggest a Compton-thin nature. Similarly, an Fe Kα EW of 1 keV is often cited as the nominal boundary for a Compton-thick source based on observations (e.g., Bassani et al. 1999; Comastri 2004; Levenson et al. 2006), yet NGC 1068, the archetype for a Compton-thick Sy2 (Matt et al. 1997), has a measured EW of 0.60^+0.05_−0.05 keV (in agreement with Pounds & Vaughan 2006 but not Matt et al. 2004, see Appendix A). Hence, though the diagnostics presented here can help in uncovering the possible Compton-thick nature of a Type 2 AGN, nominal boundaries should be considered approximate, especially since a continuum of both diagnostic flux ratios and Fe Kα EWs is present.

4.2. Implications for the Local AGN Population

4.3. Investigating Obscuration Geometry

4.3.1. Fitted Column Densities

Here, we explore the relationship between obscuration diagnostics and the column densities (N_H) derived from spectral fitting. In Figure 5, we plot the fitted column densities as a function of F_2–10 keV/F_isotropic and Fe Kα EW for the 12 μm and [O iii] samples; as the [O iv] line is the least affected by the host galaxy contaminations mentioned above, we use $F_{[{\rm O\,\mathsc{iv}}]}$ as F_isotropic. With the exception of several sources, the fitted N_H values approximately trace the degree of absorption implied by the obscuration diagnostics. However, a handful of sources lay several orders of magnitude below this trend, and are labeled in Figure 5. This result is consistent with the findings of Cappi et al. (2006), where several Sy2s have fitted N_H values an order of magnitude below that suggested by $F_{2\hbox{--}10 \;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$. Both F_2–10 keV/F_isotropic and the Fe Kα EW diagnostics point to the same sources as being anomalous, with several sources missing from the Fe Kα plot due to having an unconstrained EW or upper limit on the EW, respectively. All of these sources required only a single power-law model (with a thermal component in many cases) to adequately fit the spectrum. The low observed X-ray fluxes and high Fe Kα EW values indicate that the emission is primarily seen in scattering and/or reflection, rather than transmission through the obscuring medium. Hence, such fitted N_H values are likely associated with the line-of-sight absorption to the scattered/reflected component, suggesting that simple models of a foreground screen extincting the central source do not always recover the intrinsic absorption.

Zoom In Zoom Out Reset image size

Partial covering models, parameterized in this work by a double absorbed power law with the two photon indices tied together, can also misrepresent the inherent column density. For example, such a model fairly fit the spectra for NGC 1068 (χ² = 450.4 with 247 degrees of freedom), yet the best fit N_H was ∼9 × 10²² cm⁻² whereas the lower limit on this column density from higher energy observations is 10²⁵ cm⁻² (Matt et al. 1997). Though a partial covering model could more realistically represent the geometry of the system, assuming a certain percentage of transmitted light through the obscuring medium with the rest scattered into the line of sight, it could be subject to the same limitations discussed above for single absorbed power-law models.

Published N_H distributions could potentially be biased, skewed to lower values, though checks based on obscuration diagnostics can help mitigate this problem. For example, Akylas et al. (2006) analyzed the X-ray spectra for 359 sources from XMM-Newton and the Chandra Deep Field-South (CDFS), deriving intrinsic column densities from fitted N_H values though adopting a column density of 5× 10²⁴ cm⁻² for the cases where Γ< 1, a signature of Compton-thick obscuration. However, as Cappi et al. (2006) note, this criterion could indicate a Compton-thick source while the Fe Kα EW and flux diagnostics suggest Compton-thin (e.g., NGC 4138 and NGC 4258) or vice versa (e.g., NGC 3079). Tozzi et al. (2006) use a reflection model (PEXRAV in XSpec) for Compton-thick sources in the CDFS, which are defined as those AGNs with a better fit statistic using PEXRAV than an absorbed power-law model. However, as Murphy & Yaqoob (2009) point out, such a model describes reflection off of an accretion disk, which is not physically relevant for the putative torus obscuration. Derived N_H values could then potentially be suspect for some sources. Other diagnostics are therefore crucial in checking the reliability of fitted N_H values. For example, Krumpe et al. (2008) find the ratio of X-ray to optical flux, as well as the non-detection of an Fe Kα line in the stacked spectrum of 14 type II QSOs (AGNs with intrinsic L_2–10 keV⩾ 10⁴⁴ erg s⁻¹), to verify their distribution of moderately absorbed, but not Compton-thick, column densities.

4.3.2. Variable Sources

It is intriguing to note that all X-ray variable sources in this study are on the high end of the obscuration flux diagnostics (see Figures 2 and 3). These high F_2–10 keV/F_isotropic flux ratios may indicate that the X-ray emission from the central source is seen directly. However, the high Fe Kα EW for the Chandra and 2002 July observations of NGC 4388 (0.29^+0.11_−0.08 keV and 0.62^+0.10_−0.10 keV, respectively) and for the XMM-Newton observations of NGC 7582 (0.58^+0.04_−0.04 keV and 0.31^+0.05_−0.05 keV) are higher than predicted for transmission-dominated spectra, where the EW with respect to the primary transmitted emission is <0.18 keV (Matt 2002). Piconcelli et al. (2007) propose a double absorption geometry to account for the variability in NGC 7582: a "thick" absorber which attenuates just the central source, attributed to the putative torus, and a "thin" absorber which enshrouds the primary and reflected emission and is located externally to the torus. They postulate that this inner, "thick" absorber is inhomogeneous, accounting for the observed X-ray variability. A similar scenario may be present for NGC 4388 and be responsible for both sources switching from transmitted-dominated to reflection-dominated states (or vice versa). The Fe Kα EWs for the two other variable sources, NGC 5506 and NGC 7172, as well as the flux ratio diagnostics are consistent with Compton-thin sources, implying the central source is consistently viewed directly.

4.4. Are Compton-thick Sources Unique?

Here we investigate whether Compton-thick sources differ from their Compton-thin counterparts in terms of host galaxy and AGN properties. In particular, we examined whether systematic differences exist in intrinsic AGN power, Eddington ratio (L_bolometric/L_Eddington), central black hole mass (M_BH), the AGN contribution to the ionization field, and star formation activity. The results are summarized in Figures 6–11 and in Table 7, where we utilized survival analysis to calculate Spearman ρ values and the associated probabilities that the obscuration diagnostics are uncorrelated with host galaxy properties: P < 0.05 indicates that the quantities are significantly correlated (⩾2σ level). The values of the relevant host galaxy parameters used in this analysis are presented in LaMassa et al. (2010).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 7. Correlation of AGN Properties and Star Formation Activity with Obscuration Diagnostics

$L_{[{\rm O\,\mathsc{iv}}]}$ vs.	ρ	Pa
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$	0.273	0.053
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$	0.185	0.206
F2–10 keV/FMIR	0.349	0.015
Fe Kα EW	−0.302	0.048
$L_{[{\rm O\,\mathsc{iv}}]}/M_{BH}$ vs.
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$	0.110	0.439
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$	0.056	0.703
F2–10 keV/FMIR	0.280	0.050
Fe Kα EW	−0.219	0.151
MBH vs.
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$	0.148	0.295
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$	0.115	0.432
F2–10 keV/FMIR	−0.012	0.932
Fe Kα EW	−0.234	0.124
$F_{[{\rm O\,\mathsc{iv}}]}/F_{[{\rm Ne\,\mathsc{ii}}]}$ vs.
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$	0.036	0.804
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$	0.016	0.913
F2–10 keV/FMIR	0.269	0.067
Fe Kα EW	−0.130	0.404
PAW EW 17 μm vs.
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$	0.135	0.346
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$	0.062	0.675
F2–10 keV/FMIR	−0.055	0.700
Fe Kα EW	−0.192	0.213
α20–30 μm vs.
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$	0.272	0.057
$F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$	0.014	0.922
F2–10 keV/FMIR	0.060	0.675
Fe Kα EW	−0.235	0.127

Note. ^aProbability that the null hypothesis, that the two quantities are uncorrelated, is correct. Quantities are statistically significantly correlated if P < 0.05.

Download table as:  ASCIITypeset image

To test whether Compton-thick sources have unique AGN properties, we searched for correlations between toroidal obscuration and intrinsic AGN luminosity, accretion rate, and central black hole mass (M_BH).⁸ As discussed previously, the [O iv] 25.89 μm line serves as a robust proxy of intrinsic AGN flux as it is mainly ionized by the central engine and not affected by host galaxy reddening as the [O iii] line is (e.g., Meléndez et al. 2008; Diamond-Stanic et al. 2009; Rigby et al. 2009). We therefore utilize $L_{[{\rm O\,\mathsc{iv}}]}$ as L_isotopic in Figures 6 and 7 and Table 7. According to survival analysis, a marginal statistically significant correlation exists between L_isotropic and two of the Compton-thick flux ratios ($F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iii}}],{\rm obs}}$ and $F_{2\hbox{--}10\;{\rm keV}}/F_{[{\rm O\,\mathsc{iv}}]}$), with a marginal significant anticorrelation between L_isotropic and Fe Kα EW. Figure 6, however, shows these dependencies to be weak with a wide scatter, especially considering the error bars which cannot be accommodated in the survival analysis calculation. We find no correlations between implied column density and accretion rate (using $L_{[{\rm O\,\mathsc{iv}}]}/M_{{\rm BH}}$ as a proxy for the Eddington ratio) and M_BH (Figures 7 and 8); survival analysis does indicate a weak significant relationship between Eddington parameter and F_2–10 keV/F_MIR, but this is likely driven by the dependence on $L_{[{\rm O\,\mathsc{iv}}]}$. As the weak correlation between luminosity and obscuration is tenuous at best, we conclude that Compton-thick sources do not have systematically different AGN properties from their less obscured counterparts.

Could there be a relation between the obscuration shrouding the central engine and the large amounts of dust and gas necessary for starburst activity? As Levenson et al. (2004, 2005) point out, NGC 5135 and NGC 7130 (both members of the 12 μm sample) are starburst galaxies that likely harbor Compton-thick AGNs. The combined [O iii] and 12 μm samples provide us an opportunity to test if such a relation is generic. We use infrared quantities to illuminate the relative importance of starburst versus AGN activity: $F_{[{\rm O\,\mathsc{iv}}]}/F_{[{\rm Ne\,\mathsc{ii}}]}$, which probes the hardness of the ionization field as $F_{[{\rm O\,\mathsc{iv}}]}$ is largely ionized by the AGN whereas [Ne ii] 12.81 μm is excited by star formation processes; the EW of the 17 μm polycyclic aromatic hydrocarbon (PAH) feature (Genzel et al. 1998), which includes the emission features between 16.4 and 17.9 μm; and the MIR spectral index α_{20–30 μm}⁹ (Deo et al. 2009).¹⁰ A higher value of $F_{[{\rm O\,\mathsc{iv}}]}/F_{[{\rm Ne\,\mathsc{ii}}]}$ indicates the dominance of AGN activity, whereas larger PAH EW at 17 μm and α_{20–30 μm} values denote higher levels of starburst activity. As Figures 9–11 and Table 7 illustrate, a correlation between column density and hardness of incident ionization field/star formation activity do not exist. These results suggest that the gas responsible for starburst processes likely originates in regions of the galaxy not associated with the putative torus, and similarly that gas from the interstellar medium does not contribute significantly to toroidal AGN obscuration in hard (2–10 keV) X-rays.

4.5. NuSTAR: Detection at Higher Energies

We have analyzed archival Chandra and XMM-Newton observations for two nearly complete homogeneous samples of Sy2 galaxies: one selected from the SDSS on the basis of observed [O iii] flux and an MIR sample from the original IRAS 12 μm sample. The combined sample provided us with 45 Sy2s with existing Chandra and/or XMM-Newton data. Of these, three were not detected above the background (F08572+3915, NGC 5953, and NGC 7590) and four exhibited evidence of variability among multiple X-ray observations (NGC 4388, NGC 5506, NGC 7172, and NGC 7582).

We probed the amount of absorption present in these sources by comparing the 2–10 keV X-ray flux with optical and MIR proxies of intrinsic AGN power (F_intrinsic: the fluxes of the [O iii]λ 5007 and [O iv]25.89 μm emission lines and the MIR continuum flux at 13.5 μm) and by investigating X-ray spectral signatures of obscuration (i.e., Fe Kα EW). We compared such obscuration diagnostics with fitted column densities and explored the implications of these diagnostics on the AGN geometry. We also investigated whether a connection exists between the column density of the obscuring medium and host galaxy characteristics. Our results are summarized as follows.

• 1.  
  The majority of the combined sample have F_2–10 keV/F_intrinsic values an order of magnitude or lower than the mean values for Sy1s. The statistically significant anti-correlation between F_2–10 keV/F_intrinsic and Fe Kα EW indicates that these lower diagnostic flux ratios are due to obscuration rather than inherent X-ray weakness in Sy2s. Thus, a majority of these sources are potentially Compton-thick, consistent with the results of previous studies (e.g., Risaliti et al. 1999).
• 2.  
  A wide range of obscuration diagnostic values are present, indicating a continuum of column densities and/or inclination angles, rather than a clear segregation into Compton-thick and Compton-thin subpopulations. Though the diagnostics do generally point to the same sources as likely heavily absorbed, disagreement does exist for a handful of Sy2s. Such a discrepancy is to be expected based on the various biases affecting the observed intrinsic flux proxies and the inherent spread in such isotropic flux indicators (e.g., see LaMassa et al. 2010). Hence, nominal Compton-thick boundaries should be considered approximate.
• 3.  
  Though recent work (Georgantopoulos & Akylas 2010; Trouille & Barger 2010) shows the luminosity functions for X-ray selected and [O iii]-selected AGNs to be consistent, the various selection techniques favor differ classes of objects. Heavily obscured sources, present in optically selected samples, are missing from 2 to 10 keV X-ray samples. Sample selection based on intrinsic flux proxies are therefore necessary to include the Compton-thick population, especially since highly absorbed sources constitute the majority of our homogeneously selected samples.
• 4.  
  Though fitted column densities generally tend to trace the absorption implied by obscuration diagnostics, several glaring inconsistencies are present. Such discrepancies are most extreme when the hard X-ray spectrum is best fit by a single absorbed power law, implying that the simple geometry of a foreground screen attenuating the central source does not recover the intrinsic absorption. This could result from scattering and/or reflected emission dominating over the transmitted continuum, where the fitted column density reflects line-of-sight absorption rather than the obscuration enshrouding the AGN. Such a result indicates that published N_H distributions derived from single absorbed power-law models can be similarly biased, systematically underrepresenting the intrinsic column density of Type 2 AGNs. Other diagnostics are therefore crucial in checking the validity of fitted column densities.
• 5.  
  The X-ray variable Sy2s populate the less obscured range of the flux ratio obscuration diagnostics. Two of these sources (NGC 4388 and NGC 7582) do show evidence of switching to a reflection-dominated state, as indicated by the change in the Fe Kα EWs. Piconcelli et al. (2007) suggest that this change could reflect an inhomogeneous thick absorber covering the central source, with a thin absorber attenuating both the reflected and transmitted emission. The other two variable sources (NGC 5506 and NGC 7172) show signs of Compton-thin absorption, suggesting that the central source is viewed directly.
• 6.  
  We do not find compelling evidence that Compton-thick sources have unique AGN properties (intrinsic AGN luminosity, accretion rate, and central black hole mass) or star formation activity. Though three out of four obscuration diagnostics are significantly correlated with intrinsic AGN luminosity, the significance is marginal and the relationships display a wide scatter. Evidence linking more obscured sources to more luminous central engines is therefore tenuous at best. No correlation exists between toroidal AGN obscuration and the relative amount of ionization due to the central engine compared to star formation processes ($F_{[{\rm O\,\mathsc{iv}}]}/F_{[{\rm Ne\,\mathsc{ii}}]}$, EW of the 17 μm PAH feature, α_{20–30 μm}) and AGN absorption. Though several starburst galaxies do seem to host Compton-thick AGNs (e.g., Levenson et al. 2004, 2005), such a relation is not present globally. Hence, we conclude that the gas responsible for star formation processes is not associated with the toroidal obscuration hiding the central engine.
• 7.  
  Based on simulated high-energy (10–40 keV) spectra using the best-fit modeling of the 2–10 keV spectra, we estimate that the majority of this sample (36 out of 45) will be detected if observed by NuSTAR. The more heavily obscured sources which have not been detected by BAT surveys could likely be identified by NuSTAR as this future mission will probe to lower flux levels (∼2 × 10⁻¹⁴ erg s⁻¹ cm⁻² versus ∼3.1 × 10⁻¹¹ erg s⁻¹ cm⁻²). These observations would confirm whether the heavily absorbed sources are indeed Compton-thick.

This work was supported by grant 10-ADAP10-0167 and Spitzer grant RSA1287640. The authors thank the anonymous referee whose comments and suggestions improved the manuscript. This research has made use of data obtained from the High Energy Astrophysics Science Archive Research Center (HEASARC), provided by NASA's Goddard Space Flight Center.

NGC 424. The default Chandra aperture extraction spectrum and XMM-Newton spectra were consistent and fit simultaneously using a double absorbed power law with a Gaussian component to accommodate the Fe Kα emission; including a thermal component did not statistically significantly improve the fit. We note that fitting the 3–8 keV continuum with a power law plus a Gaussian led to a tighter constraint on the Fe Kα emission and a more realistic EW value. Matt et al. (2003a) analyzed the Chandra and XMM-Newton spectra independently using a slightly more complicated model, including Gaussian components at 0.55 and 0.90 keV to account for emission features, possibly from the O viii recombination line and the O viii recombination continuum or Ne IX recombination line, respectively. They also added a component for cold reflection, the PEXRAV model in XSpec. However, their 2–10 keV flux (1.6 × 10⁻¹² erg s⁻¹ cm⁻²) and Fe Kα EW values (∼0.88 keV) are consistent with the values we obtained (1.15^+0.64_−0.37 × 10⁻¹² erg s⁻¹ cm⁻² and 1.33^+0.36_−0.39 keV, respectively) using a simpler model.

NGC 1068. The PN and MOS1 XMM-Newton spectra for both observations showed evidence of pileup according to epatplot and were therefore not included in the spectral fit; archival Chandra data also exist for NGC 1068, but were not used in this study due to the effects of pileup. Also, as several strong emission features were present below 1 keV (which are not important for the purposes of this study), we fitted the spectrum from 1 keV to 8 keV. The MOS2 spectra were best fit by an absorbed double power law with a thermal model and Gaussian components at 2.0 keV, 2.43 keV, 6.4 keV (neutral Fe Kα), 6.66 keV (likely ionized Fe Kα), and 6.95 keV. The neutral Fe Kα line was detected at a statistically significant level. Pounds & Vaughan fitted the 3.5–15 keV XMM-Newton spectra with two continuum components, a cold reflection model (PEXRAV) and a series of Gaussian emission features from 6 to 8 keV (where nine of these features were detected at a significant level). Based on this fit, they find a 3–15 keV flux of 63 × 10⁻¹³ erg s⁻¹ cm⁻², which is consistent with our 2–10 keV flux of 54.2⁺¹¹⁰_−32.0 × 10⁻¹³ erg s⁻¹ cm⁻², and a neutral Fe Kα EW of 0.60 ± 0.10 keV, which agrees with our neutral Fe Kα EW value (0.60 ± 0.05 keV). However, as noted in the main text, Matt et al. (2004) obtained an EW value of 1.2 keV, using a PEXRAV model, power law, and a series of Gaussians for emission line features.

NGC 1144. The XMM-Newton spectra were best fit by a double absorbed power law with a thermal component and a Gaussian component for the Fe Kα emission. When fitting the 3–8 keV continuum to obtain a local fit for the Fe Kα component, a double power law was needed to accommodate the spectrum shape; the power-law indices of the two components were tied together with the normalizations and an absorption component attenuating the second power law allowed to vary. Winter et al. (2008) fit NGC 1144 with the partial covering model in XSpec (which is akin to a double absorbed power-law model with the photon indices tied together, which we have done) and a blackbody component for the soft emission. They derived comparable 2–10 keV flux (3 × 10⁻¹² erg s⁻¹ cm⁻²) and Fe Kα EW (0.22 keV) values as ours (33.4^+36.6_−12.9 × 10⁻¹³ erg s⁻¹ cm⁻² and 0.25^+0.06_−0.06 keV, respectively).

NGC 1320. The XMM-Newton spectrum to be best fit by an absorbed double power-law component with a thermal component and a Gaussian component to accommodate the Fe Kα emission. Greenhill et al. (2008) used a cold reflection model (PEXRAV) as well as two thermal components (using MEKAL whereas we used APEC) for the soft emission; a Gaussian had also been included to model the Fe Kα emission. We derive consistent Fe Kα EWs (3.02^+0.46_−0.50 keV versus 2.20^+0.44_−0.43 keV), yet their 2–10 keV flux is over an order of magnitude higher than ours (3.84^+2.83_−1.39 × 10⁻¹³ erg s⁻¹ cm⁻² versus ∼43 × 10⁻¹³ erg s⁻¹ cm⁻²).

NGC 1386. As noted in the main text, the spectral parameters between the XMM-Newton and Chandra observations were consistent, except for a lower multiplicative factor for the Chandra observation, indicating extended emission contaminated the XMM-Newton observation. Though the Chandra data were grouped by a minimum of 5 counts bin⁻¹, the XMM-Newton spectra were grouped by a minimum of 15 counts, so we used χ² statistics in this analysis. To derive the 2–10 keV flux, we fitted spectra from both observatories simultaneously to better constrain the Chandra spectrum, using a double absorbed power law with a thermal component and a Gaussian feature at 6.4 keV to accommodate the Fe Kα emission, yet we report the Chandra flux only. We fit the Chandra spectrum independently, both globally and locally, to derive the Fe Kα parameters, binning the data by a minimum of three counts and using C-stat. Levenson et al. (2006) fit the Chandra 4–8 keV nuclear spectrum with a reflection model (PEXRAV) and a Gaussian component at the Fe Kα energy. We obtain consistent 2–10 keV flux values (1.55^+0.74_−0.33 × 10⁻¹³ erg s⁻¹ cm⁻² versus 2.1±0.1 × 10⁻¹³ erg s⁻¹ cm⁻²) and Fe Kα EWs (2.30^+1.00_−0.78 versus 2.3 ± 1.5 keV).

NGC 1667. The XMM-Newton spectra were best fit with a single absorbed power law plus a thermal component. To constrain the Fe Kα EW in the local continuum fit (i.e., 3–8 keV), the spectra were binned by a minimum of 2 counts versus the 15 counts used for the global fit; C-stat was utilized in this local fit. Bianchi et al. (2005) fit the spectra with a reflection model (PEXRAV) with a soft excess, including a line at ∼0.9 keV. Our Fe Kα EWs are consistent (0.86^+0.66_−0.50 versus <0.60 keV), however, our 2–10 keV flux values disagree by about a factor of two (0.43^+0.15_−0.11 × 10⁻¹³ erg cm⁻² s⁻¹ versus 10⁻¹³ erg cm⁻² s⁻¹).

F05189-2524. Since the spectral parameters were consistent between the XMM-Newton and Chandra observations, other than a lower constant multiplicative factor for the Chandra spectra, we fit these spectra simultaneously to constrain the 2–10 keV flux. However, we only report the flux from the Chandra observation as we have demonstrated that extended emission contaminates the XMM-Newton field of view. The spectra were best fit by a double absorbed power law plus a thermal component, though the temperature was not constrained. The Chandra spectra were fit independently to model the Fe Kα emission. Though this feature was not detected, we derived an upper limit on the EW of 0.17 keV, consistent with the results of Ptak et al. (2003) who fit the 2002 Chandra observation with a single power law plus thermal component (MEKAL). We also obtain similar 2–10 keV fluxes (23.5^+5.5_−4.9 × 10⁻¹³ erg cm⁻² s⁻¹ versus 37 × 10⁻¹³ erg cm⁻² s⁻¹).

NGC 3982. As noted in the main text, the parameters, flux, and Fe Kα EW values listed are from the model fits to Chandra spectrum only as extended emission is present in the XMM-Newton field of view. The spectrum was best-fit by an absorbed power-law model with a thermal component, using the C-statistic on data grouped by 3 counts. We used ZGAUSS to test for the presence of an Fe Kα line, but the EW was unconstrained in both the global and local fit, where in the latter, it was necessary to group by 1 count bin⁻¹ to fit the continuum. Guainazzi et al. (2005b) fit the Chandra spectrum in a similar fashion (single absorbed power law with a thermal component) and obtained an upper limit on the 2–10 keV X-ray flux of 0.5 × 10⁻¹³ erg s⁻¹ cm⁻², which is consistent with our value of 0.56^+1.15_−0.35 erg s⁻¹ cm⁻². They also report a 1σ detection on the Fe Kα EW of 8 ± 5 keV, which may indicate that this parameter is unconstrained.

NGC 4388. This source varied between the XMM-Newton observations from 2002 July to December, increasing in flux and decreasing in Fe Kα EW. Both observations were best fit by a double absorbed power law (allowing the normalization of the second power-law component to vary between the two observations), a thermal component (necessary to fit the soft emission), and a Gaussian component to accommodate the Fe Kα line. The spectral shape of the local continuum (3–8 keV, to constrain the Fe Kα EW) for the December observation and Chandra observation required a base model of a double power law with an absorption component attenuating the second power law; a single power-law base model was sufficient to fit the local continuum for the July observation. Beckmann et al. (2004) fit the XMM-Newton spectra with a single absorbed power law and a Raymond–Smith thermal plasma model; they also detect Fe Kα and a possible Fe Kβ line at ∼6.89 keV. Our derived Fe Kα EWs are consistent (0.62^+0.10_−0.10 keV versus 0.57 keV and 0.18^+0.03_−0.02 keV versus 0.22 keV for the July and December observations, respectively); they do not report a 2–10 keV flux or luminosity. The Chandra flux for NGC 4388, from the 2001 June observation, is consistent with the XMM-Newton 2002 July flux, though the Fe Kα EW increased, which could be due to the presence of extended emission the XMM-Newton field of view. Iwasawa et al. (2003) found the nucleus from the Chandra observation to be moderately affected by pileup, but we did not see evidence of this when we applied the jdpileup model to the spectrum in Sherpa. We obtain consistent Fe Kα EW values as Iwasawa et al. (0.29^+0.11_−0.10 versus 0.44 ± 0.09 keV), though a somewhat higher 2–10 keV flux (74.6^+88.5_−38.5 × 10⁻¹³ erg s⁻¹ cm⁻²) than their reported 2–7 keV flux (27 × 10⁻¹³ erg s⁻¹ cm⁻²), though these values are likely consistent given the error bars on their flux and the more limited energy range over which they integrated.

NGC 4501. We report the parameters from the Chandra spectral fit as the XMM-Newton observation is contaminated by extended emission. The spectrum was best fit by an absorbed power law with a thermal component and we utilized the C-statistic as the data were grouped by 3 counts bin⁻¹. Brightman & Nandra (2008) also find the XMM-Newton field of view to be contaminated by extended emission. They fit the Chandra spectrum with a reflection component, on top of an absorbed thermal power-law model. Similar to our work, they do not detect the Fe Kα emission line in the global spectral fit.

TOLOLO 1238-364. The Chandra spectrum was best fit by an absorbed power law with a thermal component and a Gaussian to accommodate the Fe Kα emission. The data were binned by a minimum of 2 counts and we therefore employed the C-statistic. Ghosh et al. (2007) fit this spectrum with an absorbed power law plus thermal brehmsstrahlung model after binning by a minimum of 20 counts which washes out the Fe Kα feature. They detected the line at low signal-to-noise after re-binning by constant width, but obtain an unconstrained EW whereas we detected this feature.

NGC 4968. The two XMM-Newton observations for this source were best fit by a single absorbed power-law model with a Gaussian component at the Fe Kα energy, using the C-statistic on data binned by 2–3 counts; we saw no evidence for variability between the two observations. Bianchi et al. (2005) fitted these spectra that were fit independently with a reflection model (PEXRAV) with fixed photon index (Γ = 1.7), a power-law component for the soft excess and Gaussian for the Fe Kα line. We obtained consistent 2–10 keV flux (2.08^+0.26_−0.26 × 10⁻¹³ erg cm⁻² s⁻¹ versus 2.7 ± 0.08, 2.3 ± 0.08 × 10⁻¹³ erg cm⁻² s⁻¹) and Fe Kα EW values (3.06^+0.99_−0.78 keV versus 1.9 ± 0.9, 3.2 ± 1.1 keV) using the simpler power-law model.

M-3-34-64. The XMM-Newton spectra were fit by a double absorbed power law with a thermal component and a Gaussian at the Fe Kα line energy. Miniutti et al. (2007) fit this source with a reflection model, with the soft emission accommodated by a power-law model with two thermal components and a photoionized gas model and Gaussian components at 6.4 and 6.8 keV. Our observed 2–10 keV flux values approximately agree (32.5^+3.1_−3.1 × 10⁻¹³ erg s⁻¹ cm⁻² versus 21±2 × 10⁻¹³ erg s⁻¹ cm⁻²), as well as our Fe Kα EWs derived from the global fits (0.17^+0.04_−0.03 keV versus 0.11 ± 0.02 keV), though our local continuum fit results in a higher EW (0.31^+0.05_−0.04 keV).

NGC 5135. Chandra observations of NGC 5135 reveal two X-ray point sources near the nucleus of the galaxy. The northern source was identified by Levenson et al. (2004) as the active nucleus, so we restrict our analysis to this source, using an extraction region of 12. They find the AGN spectrum to be best fit by a model consisting of two thermal components, a Gaussian component at ∼2 keV and at the Fe Kα energy, and an absorbed power law. We grouped the data by a minimum of 3 counts bin⁻¹, employed the C-statistic and find that a double absorbed power law with a Gaussian component to accommodate the Fe Kα emission reasonably fits the spectrum. Despite the different models used between Levenson et al. (2004) and us, we obtain consistent 2–10 keV fluxes (2.31^+0.98_−1.68 × 10⁻¹³ erg cm⁻² s⁻¹ versus 2.10^+0.19_−0.68 × 10⁻¹³ erg cm⁻² s⁻¹) and Fe Kα EWs (2.44^+0.94_−0.82 keV versus 2.4^+1.8_−0.5 keV).

NGC 5194. The nuclear region of NGC 5194 contains several X-ray emitting features: the AGN and diffuse emission to the north and south (see Terashima & Wilson 2001). Chandra is necessary to isolate the Seyfert nucleus and we therefore present the results of the Chandra analysis only, and do not include the archival data from XMM-Newton. Similar to Terashima & Wilson (2001), we extracted a source region centered on the optical center of the galaxy with a 15 radius from the Chandra data. The data were binned by a minimum of 3 counts and we utilized the C-statistic. The spectra were best fit by a double absorbed power law with a thermal component and a Gaussian component to accommodate the Fe Kα emission. Terashima & Wilson (2001) fit the observation with an absorbed power-law model and a reflection model, both of which yield consistent fluxes (∼1.2 × 10⁻¹³ erg s⁻¹ cm⁻²) and high EW values (3.5^+2.7_−1.6 keV and 4.8^+4.3_−2.5 keV, respectively) which agree with the values we obtain (1.04^+2.28_−0.73 erg s⁻¹ cm⁻² and 4.64^+1.42_−1.47 keV).

NGC 5347. The Chandra spectrum is best fit by a double absorbed power law. To fit the local continuum to test for the presence of the Fe Kα line, we rebinned the spectrum by a minimum of 3 counts, utilized the C-statistic and detected the line at the 3σ level. Levenson et al. (2006) applied a reflection model (PEXRAV) to the higher energy range of the spectrum (4–8 keV) and fit the lower energy portion with power laws, a thermal component and line emission. Our 2–10 keV flux and Fe Kα EW values agree, 2.58^+1.20_−1.41 × 10⁻¹³ erg s⁻¹ cm⁻² versus 2.2 ± 0.4 erg s⁻¹ cm⁻² and 1.35^+0.53_−0.44 keV versus 1.3 ± 0.5 keV, respectively.

Mrk 463. Extended emission was evident in the XMM-Newton field of view as indicated by the lower observed flux from the Chandra spectrum. Indeed, the XMM-Newton is likely contaminated by the double nucleus (see Bianchi et al. 2008), which the Chandra observation is able to resolve. However, other than the constant multiplicative factor, the spectral parameters were consistent among the Chandra and three XMM-Newton spectra. The observations were consequently fit simultaneously to better constrain the Chandra parameters, though only the flux for the Chandra spectrum (for the Eastern source) is reported. To test for the presence of the Fe Kα line, the Chandra spectrum was rebinned by a minimum of 3 counts and the C-statistic was employed in the local (3–8 keV) fit; the line was detected at greater than the 99% confidence level according to our simulations. Bianchi et al. (2008) also employed a double absorbed power-law model to fit the spectrum and we obtain consistent 2–10 keV flux values and Fe Kα EWs, 2.95^+1.84_−0.82 × 10⁻¹³ erg s⁻¹ cm⁻² versus 4.1 ± 1.8 × 10⁻¹³ erg s⁻¹ cm⁻² and 0.20^+0.16_−0.13 keV versus 0.21^+0.15_−0.12 keV, respectively.

NGC 5506. The XMM-Newton MOS1 spectra showed evidence of mild pileup above 6 keV according to the SAS tool epatplot for all eight observations and were therefore not fit. For four of these (the 2001, 2002, 2008, and 2009 observations), the MOS2 spectra were also slightly piled and we excluded them from fitting. Archival Chandra data for this source do exist, but were not included in this study because they were severely affected by pileup. We find the spectra to be best fit by a double absorbed power law with a thermal component and several Gaussian components to fit Fe K emission, at energies ∼6.4 keV (neutral Fe Kα), ∼6.7 keV, and ∼6.95–7.0 keV; however, we note that for the local continuum fits, sometimes only two components were needed. The source varies by a factor of ∼1.5 in flux on the time scale of approximately several months, though the Fe Kα EW remains relatively constant. For the purposes of our analysis, we use the average flux and Fe Kα EW among the 2001 and 2004 July observations (〈EW〉 = 0.14^+0.05_−0.06 keV) and among the 2002, 2004 August, 2008, and 2009 observations (〈EW〉 = 0.13^+0.05_−0.04 keV; i.e., Figures 4–11). Guainazzi et al. (2010) studied the Fe Kα emission of these observations in depth, using a suite of physically motivated models (i.e., a combination of relativistically/non-relativistically broadened Fe Kα emission with relativistic/non-relativistic Compton reflection); the Fe Kα EWs are consistent among these various fits. We derive EW values that agree with Guainazzi et al. (2010) using simpler modeling of the local (4–8 keV) spectrum with a power law and Gaussian components.

Arp 220. The XMM-Newton and Chandra spectra were best fit by a single absorbed power-law model with a thermal component; only the absorption varied between the XMM-Newton and Chandra observations, though this had a negligible impact on the observed flux. We discarded the XMM-Newton spectra from 2005 from our fitting due to low signal-to-noise, though we note the best-fit parameters were consistent with the other XMM-Newton spectra. To test for the presence of Fe Kα emission in the nuclear region, we utilized the Chandra data only and rebinned by a minimum of 3 counts, employing the C-statistic, but the line was not detected. We note, however, that an emission line for ionized Fe Kα at E = 6.51 keV was detected, consistent with Iwasawa et al. (2005). The Chandra data were first analyzed by Clements et al. (2002) who report a double nucleus and a halo of extended emission. We obtain consistent fluxes between their 3'' extraction area (which encompasses the double nuclei) and our 45 extraction region: 1.07^+0.18_−0.16 × 10⁻¹³ erg s⁻¹ cm⁻² versus 1.0×10⁻¹³ erg s⁻¹ cm⁻².

NGC 6890. The XMM-Newton spectra were grouped by a minimum of 3 counts bin⁻¹ and were fit with the C-statistic. The data were best fit by a double absorbed power-law model and the Fe Kα line was marginally detected at the ∼93% confidence level. Shu et al. (2008) fit this source with a single power law and did not detect the Fe Kα line, though this is likely due to their choice of binning the data by a minimum of 20 counts which would eradicate the weak Fe Kα feature. We obtain consistent 2–10 keV flux values given the error range on our derived flux: 1.20^+4.01_−0.88 × 10⁻¹³ erg s⁻¹ cm⁻² versus 0.69 × 10⁻¹³ erg s⁻¹ cm⁻².

IC 5063. The Chandra spectrum was moderately affected by pileup, ∼14% according to the jdpileup model in Sherpa. We therefore included a pileup model in the XSpec spectral fits, allowing only the grade migration parameter (α) to be free and obtained an α value of 0.37; excluded the pileup component when calculating the observed 2–10 keV flux. The jdpileup model indicated no pileup in the local 4–8 keV spectral fit, so it was modeled without a pileup component in XSpec. The broadband spectrum was best fit by a double absorbed power law and our simulations indicate that the Fe Kα emission feature is significant at greater than the 2.5σ level. This Chandra spectrum has not been previously analyzed.

NGC 7130. The Chandra spectrum of NGC 7130 was best fit by a double absorbed power law with a thermal component for the soft emission and a Gaussian feature at the Fe Kα energy; we used the C-statistic with the data binned by 2 counts. This source was studied in detail by Levenson et al. (2005) where they fit the AGN spectrum with a double absorbed power law with a Gaussian component as well as two thermal components. Though they added an extra thermal component, our derived 2–10 keV flux and Fe Kα EW values are consistent (2.07^+2.09_−1.04 × 10⁻¹³ erg cm⁻² s⁻¹ versus 1.6^+0.3_−0.4 × 10⁻¹³ erg cm⁻² s⁻¹ and 0.82^+0.48_−0.33 keV versus 1.8^+0.7_−0.8, respectively).

NGC 7172. The XMM-Newton observations were best fit by a double absorbed power law with a thermal component and Gaussian feature at the Fe Kα energy. The normalization of the second power law was consistent between the 2002 and 2004 observations, yet had to be fit independently for the 2007 observation. We only fit the PN and MOS2 spectra for the 2007 observation as the MOS1 spectrum showed evidence for milder pileup at higher energies from the task epatplot. Our results indicate that the source increased by about a factor of two in flux between 2004 and 2007. The Fe Kα emission features were fit independently among the three observations, though we use the average Fe Kα EW for the 2002 and 2004 observations in the plots (〈EW〉 = 0.16^+0.06_−0.04 keV, i.e., Figures 4–11) as the 2–10 keV flux values are consistent between the two observations and the EW values are not widely discrepant in both the local and global fits, indicating that the variations in the independent Gaussian fits are likely not significant. Noguchi et al. (2009) fit the 2007 observation with a double absorbed power-law, thermal component (using MEKAL whereas we used APEC) and two Gaussian components, one at the Fe Kα energy and the other at 1.7 keV; we obtain consistent 2–10 keV fluxes (517⁺⁴³₋₄₀ × 10⁻¹³ erg cm⁻² s⁻¹ versus 423×10⁻¹³ erg cm⁻² s⁻¹) and Fe Kα EWs (0.10^+0.02_−0.01 keV versus 0.07 ± 0.01 keV). Shu et al. (2007) fit the spectra from the 2002 XMM-Newton observation by a double absorbed power law with a Gaussian component at the Fe Kα energy, similar to our analysis, though they do not include a thermal component. Our 2–10 keV flux values are consistent (234⁺¹⁹₋₁₈ × 10⁻¹³ erg cm⁻² s⁻¹ versus 220 × 10⁻¹³ erg cm⁻² s⁻¹), but our Fe Kα EWs are not (0.14^+0.03_−0.03 keV from the global continuum fit versus 0.04 ± 0.03 keV). This discrepancy could be due to constraints placed on their modeling of the Fe Kα line: they froze the energy at 6.4 keV, whereas we allowed this parameter to be free and it is not clear whether they placed a similar constraint on σ. Regardless of this discrepancy, both EW values are consistent with a Compton-thin source. Analysis of the EPIC data for the 2004 observation has not been previously published, making our analysis the first for this data set.

NGC 7582. This source varied between the 2000 Chandra observations and later XMM-Newton observations, as well as between the 2001 and 2005 XMM-Newton observations; the Chandra and XMM-Newton observations were fit independently. The Chandra spectra were moderately affected by pileup (∼30%–49% according to the jdpileup model in Sherpa) and were therefore fit with a pileup model component in XSpec, with only α, the grade migration parameter, allowed to vary; the pileup component was discarded before calculating the flux and the Fe Kα EW. The broadband Chandra spectra were best fit by a double absorbed power law whereas the XMM-Newton spectra required a thermal component and Gaussian components at the Fe Kα energy and at ionized Fe Kα energies (∼6.72 keV for the 2005 observation and ∼6.97 keV for the 2001 observation),¹¹ ∼1.84 keV and ∼2.47 keV. The Fe Kα feature was detected at a statistically significant level for all observations according to our simulations. NGC 7582 dimmed between the Chandra observation and each subsequent XMM-Newton observation and the Fe Kα EW increased. This decrease in flux with increase in Fe Kα EW could reflect a variation in the obscuring medium, where the obscuration enhanced over time. Indeed, such an interpretation is favored by Piconelli et al. (2007), who postulate the existence of multiple absorption components in this system: a higher column-density absorber (possibly mildly Compton-thick) attributed to the putative torus and a lower-column density absorber acting as a screen to both the reflected and transmitted radiation. Though Piconelli et al. (2007) fit the XMM-Newton observations with a more complex model (PEXRAV) we obtain consistent fluxes (21.1^+1.7_−1.8 ×  10⁻¹³ erg s⁻¹ cm⁻² versus 23.5 × 10⁻¹³ erg s⁻¹ cm⁻² for the 2005 observation and 38.6^+3.0_−3.1 × 10⁻¹³ erg s⁻¹ cm⁻² versus 40.2 × 10⁻¹³ erg s⁻¹ cm⁻² for the 2001 observation), though lower Fe Kα EW values (2005: 0.58^+0.04_−0.04 keV versus 0.77^+0.05_−0.04 keV; 2001: 0.31^+0.05_−0.05 keV versus 0.62^+0.07_−0.08). Dong et al. (2004) fit the Chandra spectra independently with a double absorbed power law, yet obtain an observed flux about half of ours (164⁺²⁶³₋₈₇ × 10⁻¹³ erg s⁻¹ cm⁻² versus ∼75 × 10⁻¹³ erg s⁻¹ cm⁻²), though this discrepancy could result from us modeling pileup whereas they did not; we obtain consistent Fe Kα EW values (0.15^+0.07_−0.07 keV versus an averaged ∼0.19 keV).

NGC 7674. The global XMM-Newton spectra were best fit with a double absorbed power law. To fit the local continuum between 3 and 8 keV, the data were binned to a minimum of 3 counts and the C-statistic was utilized; the Fe Kα feature was detected a statistically significant level. Analysis of the broadband XMM-Newton spectra for this source has not been previously published.

Using ARF, RMF, and background files provided by the NuSTAR team, which include separate ARF and background files for a 45'' and a 101'' point-spread function (useful for weak and strong sources, respectively), we generated a simulated NuSTAR spectrum in the 10–40 keV energy range as described in the main text. If the net count rate was >10⁻² s⁻¹, we utilized the simulated spectrum corresponding to the larger PSF, otherwise we used the spectrum generated with the smaller PSF. In Tables 8 and 9, we list the simulated source count rate and corresponding exposure time for the target to be detected at greater than the 5σ level above the background, which was either ∼8 × 10⁻⁴ s⁻¹ or ∼4 × 10⁻³ s⁻¹, depending on the PSF. For the sources where this derived exposure time is under 5 ks, we instead list the exposure time necessary for at least 100 counts to be detected; if this exposure time is also under 5 ks, we adopt a minimum exposure time of 5 ks. For the three 2–10 keV non-detections from the 12 μm sample, we list the exposure times using a spectrum simulated from the PLCABS model in XSpec, with N_H = 10²⁴ cm⁻², the number of scatterings set to 5, Γ = 1.8 and the normalization adjusted such that the 2–10 keV flux equals the 3σ upper limit. Again, we note that such a flux estimate is quite optimistic and the corresponding derived exposure times necessary for detection should be considered as lower limits and are listed as such in Table 8.