A NEW POPULATION OF COMPTON-THICK AGNs IDENTIFIED USING THE SPECTRAL CURVATURE ABOVE 10 keV

, , , , , , , , , , , , , , , , , , , , , and

Published 2016 July 6 © 2016. The American Astronomical Society. All rights reserved.
, , Citation Michael J. Koss et al 2016 ApJ 825 85 DOI 10.3847/0004-637X/825/2/85

0004-637X/825/2/85

ABSTRACT

We present a new metric that uses the spectral curvature (SC) above 10 keV to identify Compton-thick active galactic nuclei (AGNs) in low-quality Swift/Burst Alert Telescope (BAT) X-ray data. Using NuSTAR, we observe nine high SC-selected AGNs. We find that high-sensitivity spectra show that the majority are Compton-thick (78% or 7/9) and the remaining two are nearly Compton-thick (NH ≃ (5–8) × 1023 cm−2 ). We find that the SCBAT and SCNuSTAR measurements are consistent, suggesting that this technique can be applied to future telescopes. We tested the SC method on well-known Compton-thick AGNs and found that it is much more effective than broadband ratios (e.g., 100% using SC versus 20% using 8–24 keV/3–8 keV). Our results suggest that using the >10 keV emission may be the only way to identify this population since only two sources show Compton-thick levels of excess in the Balmer decrement corrected [O iii] to observed X-ray emission ratio (${F}_{[{\rm{O}}{\rm{III}}]}/{F}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}}\gt 1$) and WISE colors do not identify most of them as AGNs. Based on this small sample, we find that a higher fraction of these AGNs are in the final merger stage (<10 kpc) than typical BAT AGNs. Additionally, these nine obscured AGNs have, on average, ≈4× higher accretion rates than other BAT-detected AGNs ($\langle {\lambda }_{\mathrm{Edd}}\rangle \;=\;0.068\pm 0.023$ compared to $\langle {\lambda }_{\mathrm{Edd}}\rangle \;=\;0.016\pm 0.004$). The robustness of SC at identifying Compton-thick AGNs implies that a higher fraction of nearby AGNs may be Compton-thick (≈22%) and the sum of black hole growth in Compton-thick AGNs (Eddington ratio times population percentage) is nearly as large as mildly obscured and unobscured AGNs.

Export citation and abstract BibTeX RIS

1. INTRODUCTION

While there has been great progress understanding the origin of the cosmic X-ray background (CXB) and the evolution of active galactic nuclei (AGNs) with XMM-Newton and Chandra (e.g., Brandt & Alexander 2015), it is clear that a significant fraction of the >8 keV background is not produced by known 2–8 keV sources (Worsley et al. 2005; Luo et al. 2011; Xue et al. 2012). This background probably originates from a high-column-density, low-redshift population (z < 1). However, the source of the bulk of the CXB's surface brightness, peaking at ≈30 keV, is still unknown. The measurement of the space density and evolution of this population of highly absorbed AGNs, as well as the derivation of their column density distribution function with luminosity and redshift, is crucial for understanding the cosmic growth of black holes. Population-synthesis models attempt to explain the CXB by introducing appropriate numbers of absorbed Seyferts (e.g., Treister & Urry 2005; Gilli et al. 2007). However, studies suggest that the number of Compton-thick AGNs (NH > 1024 cm−2 ) is a factor of 3–4 smaller than expected in the population-synthesis models (e.g., Treister et al. 2009), at least in the local universe (Ajello et al. 2012). Additional studies suggest that Compton-thick AGNs evolve differently than other obscured sources and are more likely associated with rapid black hole growth at higher redshift (e.g., Draper & Ballantyne 2010; Treister et al. 2010). These problems limit our current knowledge of the origin of the CXB at >10 keV.

In many well-studied objects, obscuration significantly attenuates the soft X-ray, optical, and UV signatures of AGNs. There are only two spectral bands, the ultra-hard X-ray (>10 keV) and the mid-infrared (5–50 μm), where this obscuring material is optically thin up to high column densities NH < 1024 cm−2 (Compton-thin). Thus, these spectral bands are optimal for less biased AGN searches (e.g., Treister et al. 2004; Stern et al. 2005; Alexander et al. 2008). Radio selection of AGNs is also largely obscuration independent, though only ∼10% of AGNs are radio loud (e.g., Miller et al. 1990; Stern et al. 2000), and finding a radio excess in radio-quiet AGNs can be difficult because of the host galaxy contribution from star formation (Del Moro et al. 2013) and significant free–free absorption from the ionized torus (Roy et al. 2000). Mid-IR selection is very effective at identifying high-luminosity AGNs, where the nuclear emission dominates, but moderate-luminosity AGNs, like those common in the local universe, are harder to identify because the host galaxy contribution is relatively larger (e.g., Cardamone et al. 2008; Eckart et al. 2010; Donley et al. 2012; Stern et al. 2012). In contrast, X-ray surveys suffer little contamination from non-nuclear emission at typical survey depths, and thus a hard X-ray survey can efficiently find both low- and high-luminosity AGNs in a uniform fashion, including even the heavily obscured, lower-luminosity AGNs, which we expect to be important contributors to the CXB.

The Burst Alert Telescope (BAT; Barthelmy et al. 2005), a large field of view (1.4 sr) coded aperture imaging instrument on the Swift satellite, has surveyed the sky to unprecedented depth. The all-sky BAT survey is a factor of ≈20 more sensitive than previous satellites such as HEAO 1 (Levine et al. 1984). BAT selection is particularly powerful because it uses the 14–195 keV band, which can pass through obscuring material of NH > 1024 cm−2 , though it is still biased against the most obscured AGNs (e.g., >1025 cm−2; Lanzuisi et al. 2015a). It is therefore sensitive to most obscured AGNs where even moderately hard X-ray surveys (∼10 keV) are severely reduced in sensitivity. The 70-month Swift/BAT survey has identified 1210 objects, of which 823 are AGNs, while the rest are overwhelmingly Galactic in origin (Baumgartner et al. 2013). Higher angular resolution X-ray data for every source were obtained with the Swift X-ray Telescope (Burrows et al. 2005) because of the large positional uncertainty of Swift/BAT (≈6') for fainter sources.

Unfortunately, due to the large number of sources spread across the sky and the limited sensitivity of Swift/XRT to obscured sources, X-ray follow-up and identification of the entire BAT catalog of ≈800 AGNs have been difficult. Survey programs typically used the first year or two of stacked data (e.g., 9-month Survey, PI R. Mushotzky; Northern Galactic Cap 22-month Survey, PI N. Brandt). After 10 yr of the mission, one can detect many more obscured AGNs, which are critical to estimating the fraction of Compton-thick AGNs and the source of the CXB. Additionally, the majority of sources had Swift/XRT coverage, which is insufficient for measuring the column density (NH ) in heavily obscured AGNs (Winter et al. 2009). Finally, accurately estimating column densities based on Swift/XRT and BAT data is problematic because of time variability and the low signal-to-noise ratio of typical BAT detections (Vasudevan et al. 2013).

To make progress in this area requires (1) an ultra-hard X-ray survey of sufficient sensitivity, angular resolution, and solid angle coverage at ≈30 keV to identify a large number of sources and (2) high-sensitivity observations to obtain the column density of the sources, their detailed X-ray spectral properties, and confirmations of their identifications. With the new focusing optics on the Nuclear Spectroscopic Telescope Array (NuSTAR; Harrison et al. 2013), the entire 3–79 keV energy range can be studied at sensitivities more than 100× better than those of previous coded aperture mask telescopes such as Swift/BAT or INTEGRAL. This enables detailed X-ray modeling of heavily obscured AGNs (e.g., Arévalo et al. 2014; Baloković et al. 2014; Bauer et al. 2014; Gandhi et al. 2014; Puccetti et al. 2014; Brightman et al. 2015; Koss et al. 2015).

In this article, we combine the all-sky nature of Swift/BAT with the unprecedented NuSTAR sensitivity over a wide energy range to develop a new technique to find previously unknown heavily obscured AGNs. The 10–100 keV spectrum becomes increasingly curved with increasing absorption. This is especially useful in selecting Compton-thick AGNs because of its effectiveness up to very high column densities (NH ∼ 1025 cm−2 ). Additionally, detection based solely on spectral curvature (SC) offers an important test of AGN torus models and is less biased against Compton-thick AGNs. In Section 2, we detail the NuSTAR sample and the SC selection. Section 3 describes the data reduction and analysis procedures for the NuSTAR observations. Section 4 focuses on the results of SC on the full BAT sample of 241 nearby AGNs (z < 0.03), a subset of 84/241 NuSTAR-observed AGNs, and the X-ray spectral and multiwavelength analyses for the nine NuSTAR-observed SC-selected AGNs. Finally, Section 5 gives a summary of our results and discusses the implications of the full survey in terms of the black hole growth. Throughout this work, we adopt Ωm = 0.27, ΩΛ = 0.73, and H0 = 71 km s−1 Mpc−1. Errors are quoted at the 90% confidence level unless otherwise specified.

2. SAMPLE SELECTION

Here we describe the simulations used to derive the SC measurement in Swift/BAT and NuSTAR data (Section 2.1). We then discuss the results of applying this SC measurement to nearby Swift/BAT AGNs (Section 2.2). Finally, we select nine high SC BAT AGNs for NuSTAR follow-up.

2.1. SC Measurement

We define a curvature parameter to estimate Compton thickness using the distinctive spectral shape created by Compton reflection and scattering (Figure 1). We hereafter call it the SC. We generated the simulated data (using the XSPEC fakeit feature) from Compton-thick obscuration using the MYTorus model (Murphy & Yaqoob 2009).

Figure 1.

Figure 1. Left: simulated Compton-thick AGNs compared to an unabsorbed power-law source with Γ = 1.9 showing the increasing SC with column density. Center: Swift/BAT count rates for these same sources normalized by the rate of an unobscured source for the four BAT channels between 14 and 50 keV. The energy bands above 24 keV show an excess, while the bands between 14 and 24 keV show a decrement. Right: NuSTAR count rates for these same sources normalized by the rate of an unobscured source in the range 8–30 keV. At energies between 14 and 30 keV, a Compton-thick source has an excess, while the 8–14 keV energy band shows a decrement compared to the count rates of an unobscured source. The weighted average of BAT and NuSTAR energy bands can be used to find Compton-thick sources.

Standard image High-resolution image

The MYTorus -based model used for calculating the SC (Model SC hereafter) has the following form:

Here, MYTZ represents the zeroth-order transmitted continuum (POW) through photoelectric absorption and the Compton scattering of X-ray photons out of the line of sight, MYTS is the scattered/reflected continuum produced by scattering X-ray photons into the line of sight, and MYTL is the fluorescent emission-line spectrum. The torus model we use for measuring SC is viewed nearly edge-on (θinc = 80° ) with a cutoff power law (Γ = 1.9, Ec > 200 keV). The SC model assumes a half opening angle of 60° , which is equivalent to a covering factor of 0.5. The SC is calculated so that a heavily Compton-thick source in an edge-on torus model has a value of 1 (e.g., SC = 1 for NH = 5 × 1024 cm−2 ) and an unabsorbed AGN has a value of 0.

The SC can be applied to X-ray observations from any satellite with energy coverage of the "Compton hump" (≈10–30 keV). This "hump" occurs because of the energy dependence of photoelectric absorption, whereby soft X-rays are mostly absorbed, and higher-energy photons are rarely absorbed and tend to Compton scatter (see, e.g., Reynolds 1999). The SC measurement uses weighted averages of different energy ranges as compared to the count rate in an unobscured AGN. To estimate the SC for BAT data, we focus on data below 50 keV because this shows the strongest difference in curvature compared to an unobscured source. Additionally, the BAT sensitivity is significantly reduced in the 50–195 keV energy ranges. For NuSTAR, we use the 8–14 keV, 14–20 keV, and 20–30 keV energy ranges, because of the reduced sensitivity of NuSTAR above 30 keV compared to the preceding energy ranges. We note that the SC method will be biased against Compton-thick AGNs with very high column densities (NH > 5 × 1024 cm−2 ) because of the large reduction in count rates (Figure 2).

Figure 2.

Figure 2. Simulations using an edge-on MYTorus model showing the reduction in count rates for instruments observing AGNs of different column densities. The black and blue horizontal dashed lines indicate a 50% and 90% reduction, respectively. Surveys below 10 keV like ROSAT and eROSITA are strongly biased against detecting heavily obscured AGNs, and Chandra and XMM-Newton are heavily biased against detecting Compton-thick AGNs, while NuSTAR and Swift/BAT are less biased against Compton-thick AGNs with very high column densities (NH > 5 × 1024 cm−2 ) because of the large reduction in count rates.

Standard image High-resolution image

The two SC equations take the following form:

Equation (1)

where A, B, C, and D refer to the 14–20 keV, 20–24 keV, 24–35 keV, and 35–50 keV channel Swift/BAT count rates, respectively, and the total rate refers to the 14–50 keV total rate; and

Equation (2)

where E, F, and G refer to the NuSTAR 8–14 keV, 14–20 keV, and 20–30 keV on-axis count rates, respectively, and the total rate refers to the 8–30 keV total rate in the A and B telescope. Simulations of the SC measurement with Swift/BAT and NuSTAR, as well as different model parameters and column densities, are shown in Figure 3. The differences between SC measurements at a specific column density using a variety of different torus model parameters or using NuSTAR or Swift/BAT are small except at very high column densities (NH > 4 × 1024 cm−2 ). The SC measure is more sensitive to sources that are mildly Compton-thick (NH ≈ 3 × 1024 cm−2 ), as can be seen by the flattening of the slope at very high column densities. As SCBAT only uses emission above 14 keV, it is insensitive to differences between unobscured sources and mildly obscured sources (NH < 7 × 1023 cm−2 ) that fail to obscure any of the softest 14–20 keV emission. Finally, we show simulations of broadband ratios that, unlike the SC measurement, are ineffective at selecting Compton-thick AGNs because of a degeneracy with sources at lower column densities.

Figure 3.

Figure 3. Left: variations in SC with column density based on XSPEC simulations. The blue solid line shows the MYTorus model used for the SCBAT definition with Γ = 1.9, Ec = 200 keV, and θinc = 80°. SCNuSTAR (blue dashed line) only shows small differences from SCBAT (ΔSC < 0.07) at all column densities. We also show the dependency of SCBAT on inclination angle (θinc, gray circle), intrinsic power law (Γ, purple circle), opening angle with BNTorus model (θopen, pink and brown stars; Brightman & Nandra 2011), or a simple sphere model (black diamonds; Brightman & Nandra 2011). The differences between models are small (ΔSCBAT < 0.1) except at very high column densities (NH > 4 × 1024 cm−2 ). At very high column densities (NH > 4 × 1024 cm−2 ) the SC measure does not increase at larger column densities. At low column densities (NH < 5 × 1023 cm−2 ), the lack of the high-energy cutoff in the BNTorus and sphere models raises the SC because of the additional flux in the high-energy emission. A red dotted line shows the Compton-thick lower limit (SCBAT = 0.4) used for this study, and a gray dotted line shows the upper limit from the torus model (SCBAT = 1.28). Right: simulations of obscured sources using band ratios. The band ratio, unlike the SC method, is ineffective at selecting Compton-thick AGNs because of a degeneracy with sources at lower column densities.

Standard image High-resolution image

2.2. SC of Nearby BAT AGNs and NuSTAR Targets

We applied the SC to study Compton thickness in a sample of nearby (z < 0.03) BAT-detected AGNs. We use the 70-month catalog (Baumgartner et al. 2013) with a low-luminosity cut (L14−195 keV > 1042 erg s−1 ) to avoid detecting purely star-forming galaxies such as M82. This luminosity limit corresponds to the 90% BAT all-sky sensitivity limit at 100 Mpc (z = 0.023), so that our survey sensitivity is more uniform in the volume explored. This does exclude one lower-luminosity (${L}_{14\mbox{--}195\mathrm{keV}}\lt {10}^{42}$ erg s−1 ) BAT-detected Compton-thick AGN (NGC 4102). We also exclude four sources in the Galactic plane with heavy obscuration ($E[B-V]\gt 1.2$, galactic latitude $| b| \lt 10^\circ $) because of the difficulty of multiwavelength study, as well as eight BAT detections that are heavily contaminated by a secondary BAT source. The final sample includes 241 BAT-detected nearby AGNs. Of these, 35% (84/241) have been observed by NuSTAR.

We measure the SC in these 241 BAT AGNs (Figure 4). We have highlighted 10 "bona fide" Compton-thick AGNs in our sample that have been observed with NuSTAR and have been confirmed to be Compton-thick based on spectral fitting (Gandhi et al. 2014). We targeted nine northern hemisphere objects with archival optical imaging and spectroscopy from past studies (e.g., Koss et al. 2011b, 2012) and very high SC values for NuSTAR follow-up. Two of the nine targeted NuSTAR sources have measured curvatures above simulation upper limits. The majority of our sample is at much fainter fluxes than previously known Compton-thick AGNs observed by NuSTAR (e.g., Arévalo et al. 2014; Baloković et al. 2014; Bauer et al. 2014; Puccetti et al. 2014; Brightman et al. 2015). A list of likely Compton-thick AGNs based on SC is found in Table 1.

Figure 4.

Figure 4. SC measurement for our sample of 241 BAT AGNs at z < 0.03. Compton-thick AGNs confirmed with NuSTAR are shown in black, mildly obscured Seyfert 2s observed with NuSTAR are shown in red, and NuSTAR-observed Seyfert 1s are shown in green. The simulation upper limits of any torus model are shown as a horizontal gray dotted line, while a Compton-thick column is shown by a horizontal red dotted line. The SC-selected NuSTAR targets in our program (blue) were selected to have the highest SC measure along with archival optical imaging and spectroscopy. The majority of the NuSTAR program is at much fainter fluxes than previously known Compton-thick AGNs.

Standard image High-resolution image

Table 1.  Likely Compton-thick AGNs Using Spectral Curvature

Object Typea Sey.b zc SCBAT SCNuSTAR d BAT 14–50 keVe
NGC 6232 P 2 0.015 1.93 ± 0.51 0.27 ± 0.14 1.19E-05
CGCG 164-019 P 1.9 0.030 1.64 ± 0.62 0.39 ± 0.06 1.04E-05
ESO 406–G004 ... 2 0.029 1.45 ± 0.63 ... 1.16E-05
MCG +00-09-042 ... 2 0.024 1.22 ± 0.49 ... 1.55E-05
NGC 3393 P 2 0.013 1.15 ± 0.35 0.77 ± 0.04 2.31E-05
ESO 323-32 ... 2 0.016 1.13 ± 0.82 ... 1.11E-05
NGC 4945 CTB 2 0.002 1.11 ± 0.04 1.01 ± 0.01 2.17E-04
MCG +06-16-028 P 1.9 0.016 1.09 ± 0.50 0.56 ± 0.04 1.64E-05
UGC 3157 P 2 0.015 1.03 ± 0.47 0.28 ± 0.04 1.98E-05
2MFGC 02280 P 2 0.015 0.97 ± 0.33 0.73 ± 0.07 2.15E-05
NGC 3588 NED01 P 2 0.026 0.96 ± 0.59 0.26 ± 0.04 1.07E-05
NGC 7212 NED02 P 2 0.027 0.94 ± 0.55 0.35 ± 0.04 1.41E-05
NGC 1106 ... 2 0.015 0.92 ± 0.45 ... 1.65E-05
ESO 565–G019 ... 2 0.016 0.87 ± 0.53 ... 1.52E-05
CGCG 229-015 ... 1.5 0.028 0.85 ± 0.51 ... 1.43E-05
NGC 1194 ... 2 0.014 0.81 ± 0.22 0.56 ± 0.03 3.44E-05
2MASX J07262635-3554214 ... 2 0.029 0.81 ± 0.39 0.37 ± 0.03 1.71E-05
NGC 1068 CTB 2 0.004 0.80 ± 0.17 0.66 ± 0.02 4.15E-05
NGC 3079 P 1.9 0.004 0.77 ± 0.17 0.87 ± 0.04 3.31E-05
UGC 12282 ... 2 0.017 0.75 ± 0.41 ... 1.62E-05
ESO 426–G002 ... 2 0.022 0.75 ± 0.30 ... 2.25E-05
ESO 005–G004 ... 2 0.006 0.61 ± 0.24 ... 2.72E-05
UGC 12741 ... 2 0.017 0.61 ± 0.31 ... 2.23E-05
MCG +04-48-002 ... 2 0.014 0.52 ± 0.11 0.45 ± 0.04 7.29E-05
NGC 6240 CTB 1.9 0.025 0.50 ± 0.18 0.50 ± 0.02 6.98E-05
Fairall 51 ... 1.5 0.014 0.49 ± 0.18 ... 5.13E-05
Mrk 3 CTB 1.9 0.014 0.48 ± 0.05 0.35 ± 0.01 1.25E-04
Circinus Galaxy CTB 2 0.001 0.48 ± 0.03 0.69 ± 0.01 3.50E-04
NGC 612 ... 2 0.030 0.47 ± 0.13 0.39 ± 0.03 4.61E-05
NGC 7582 CTB 2 0.005 0.45 ± 0.08 0.39 ± 0.02 8.71E-05
NGC 3281 ... 2 0.011 0.44 ± 0.08 ... 9.96E-05
ESO 297-018 ... 2 0.025 0.42 ± 0.09 ... 6.92E-05
NGC 3081 ... 2 0.008 0.42 ± 0.09 ... 8.50E-05
NGC 1365 ... 2 0.006 0.40 ± 0.08 0.26 ± 0.01 7.92E-05
ARP 318 ... 2 0.013 0.72 ± 0.57 ... 1.22E-05
UGC 07064 ... 1.9 0.025 0.72 ± 0.55 ... 1.06E-05
HE 1136-2304 ... 1.9 0.027 0.69 ± 0.51 ... 1.66E-05
NGC 452 ... 2 0.017 0.68 ± 0.43 ... 1.62E-05
NGC 7479 ... 1.9 0.008 0.68 ± 0.40 ... 1.89E-05
MCG –01-30-041 ... 1.8 0.019 0.67 ± 0.46 ... 1.65E-05
NGC 2788A ... ... 0.013 0.66 ± 0.38 ... 1.75E-05
CGCG 122-055 ... 1.5 0.021 0.63 ± 0.66 ... 1.02E-05
Mrk 1310 ... 1.5 0.019 0.62 ± 0.49 ... 1.54E-05
NGC 7465 ... 2 0.007 0.59 ± 0.52 ... 1.39E-05
NGC 1125 ... 2 0.011 0.58 ± 0.36 ... 1.83E-05
NGC 3035 ... 1.5 0.015 0.57 ± 0.40 ... 1.93E-05
ESO 553–G043 ... 2 0.028 0.57 ± 0.45 ... 1.53E-05
NGC 3786 ... 1.9 0.009 0.54 ± 0.38 0.09 ± 0.05 1.55E-05
ESO 317–G038 ... 2 0.015 0.54 ± 0.65 ... 1.16E-05
UGC 11397 ... 2 0.015 0.53 ± 0.42 ... 1.83E-05
ESO 374–G044 ... 2 0.028 0.52 ± 0.46 ... 1.64E-05
ESO 533–G050 ... 2 0.026 0.51 ± 0.70 ... 1.11E-05
PKS 2153-69 ... 2 0.028 0.51 ± 0.48 ... 1.48E-05
NGC 7679 ... 1.9 0.017 0.50 ± 0.44 ... 1.71E-05
Mrk 622 ... 1.9 0.023 0.49 ± 0.55 ... 1.31E-05
UGC 03995A ... 2 0.016 0.48 ± 0.38 0.36 ± 0.02 2.06E-05
NGC 7319 ... 1.9 0.023 0.48 ± 0.21 0.40 ± 0.07 3.29E-05
MCG –01-05-047 ... 2 0.017 0.47 ± 0.26 0.28 ± 0.03 2.71E-05
Mrk 590 ... 1.5 0.026 0.44 ± 0.41 ... 1.71E-05
NGC 6552 ... 2 0.027 0.44 ± 0.34 ... 1.83E-05
ESO 549–G049 ... 1.9 0.026 0.44 ± 0.31 ... 2.18E-05
NGC 5643 CTB 2 0.004 0.43 ± 0.65 0.47 ± 0.05 2.18E-05
2MASX J07394469-3143024 ... 2 0.026 0.42 ± 0.25 0.25 ± 0.02 2.80E-05
NGC 4992 ... 2 0.025 0.42 ± 0.13 0.30 ± 0.02 5.02E-05
CGCG 367-009 ... 2 0.027 0.41 ± 0.27 0.10 ± 0.03 2.32E-05
2MASX J18263239+3251300 ... 2 0.022 0.40 ± 0.44 0.07 ± 0.01 1.78E-05
NGC 7130 CTB 1.9 0.016 0.40 ± 0.55 0.86 ± 0.08 1.56E-05
CGCG 420-015 CTB 1.9 0.029 0.31 ± 0.28 0.44 ± 0.03 3.08E-05
NGC 5728 CTB 2 0.009 0.28 ± 0.14 0.61 ± 0.02 8.56E-05
NGC 424 CTB 1.9 0.012 0.23 ± 0.25 0.58 ± 0.05 2.58E-05
2MASX J00253292+6821442 ... 2 0.012 -0.1 ± 0.40 0.56 ± 0.05 1.59E-05

Notes. List of likely Compton-thick AGNs based on SC from our BAT sample of 241 z < 0.03 AGNs. The errors indicated here are 1σ. The upper section lists any likely Compton-thick BAT AGNs (34/241; SCBAT > 0.4 and an error lower bound of SCBAT > 0.3). The middle section lists the remaining sources with SCBAT > 0.4, which are less likely to be Compton-thick (33/241) because of the large error bars. Finally, the bottom section shows four sources with SC > 0.4 in NuSTAR, but SCBAT < 0.4.

aType of AGN: P—SC-selected BAT AGNs observed in our NuSTAR program; CTB—"bona fide" Compton-thick AGNs confirmed to be Compton-thick based on spectral fitting (Gandhi et al. 2014); U—above upper limit in SCBAT for torus simulations. bAGN type based on optical spectra from Koss et al. (2016) or NED. cMeasured redshift from NED. dWhere ellipses are shown, no public NuSTAR observations exist. eBAT 14–50 keV total count rate in counts s−1. Total count rate errors are small (<3 × 10−6 counts s−1).

Download table as:  ASCIITypeset images: 1 2

The brightest three targets from the NuSTAR SC program (NGC 3079, NGC 3393, and NGC 7212) have already been claimed to be Compton-thick, but we observed them for the first time with NuSTAR to confirm this. NGC 3079 was suggested to be Compton-thick based on a large Fe Kα equivalent width (>1 keV) and prominent emission above 10 keV with BeppoSAX (Iyomoto et al. 2001). For NGC 3393, BeppoSAX observations in 1997 suggested column densities of NH = 3 × 1023 cm−2 , but the large Fe Kα line equivalent width (>1 keV), high ratio of [O iii] to soft X-ray flux, and a $\gt 20\;\mathrm{keV}$ excess suggested a Compton-thick AGN (Salvati et al. 1997). Risaliti et al. (2000) suggested that NGC 7212 is Compton-thick from analysis of a low signal-to-noise ratio ASCA spectrum, based on a flat continuum and a prominent Fe Kα line.

The remaining six targets (2MFGC 02280, CGCG 164-019, MCG +06-16-028, NGC 3588 NED0201, NGC 6232, and UGC 3157) have never been suggested to be Compton-thick in past literature. They were first observed with Swift/XRT for BAT counterpart identification. 2MFGC 02280 (SWIFT J0251.3+5441) was observed for 10.9 ks, with no counterpart detected above 3σ in the BAT error circle (Baumgartner et al. 2013). The remaining five sources were observed for 8–13 ks and confirmed to be the brightest X-ray source within the BAT error circle. Most of these AGNs are just above the BAT detection limit of 4.8σ (NGC 3588 S/N = 5.0, CGCG 164-019 S/N = 5.1, MCG +06-16-028 S/N = 6.1, NGC 6232 S/N = 5.1, UGC 3157 S/N = 5.4), with a more significant detection of 2MFGC 02280 (S/N = 8.9). Of these, only NGC 3588 NED02 was observed using Chandra, in a study searching for dual AGNs in close mergers (Koss et al. 2012).

3. DATA AND REDUCTION

3.1. NuSTAR

Table 2 provides details, including dates and exposure times, for the nine NuSTAR observations of SC-selected BAT AGNs. We analyzed these sources, as well as 75 other low-redshift ($z\lt 0.03$) AGNs in the NuSTAR public archive, for a total of 84 NuSTAR observations. We have chosen not to include the Swift/BAT data in the model fits because our selection method will bias our fits to Compton-thick obscuration and the Swift/BAT data were taken over a period of 6 yr (2004–2010).

Table 2.  X-Ray Observation Log

  NuSTAR Observations XRT Observations
Object Name BAT ID z Observation ID UT Date ${t}_{\mathrm{eff}}$ Observation ID UT Date t
(1) (2) (3) (4) (5) (6) (7) (8) (9)
2MFGC 02280 SWIFT J0251.3+5441 0.0152 60061030002 2013 Feb 16 15 00080255001 2013 Feb 16 6
CGCG 164-019 SWIFTJ1445.6+2702 0.0299 60061327002 2013 Sep 13 24 00080536001 2013 Sep 13 6
MCG +06-16-028 SWIFTJ0714.2+3518 0.0157 60061072002 2013 Dec 03 23 00080381001 2013 Dec 03 7
NGC 3079 SWIFTJ1001.7+5543 0.0037 60061097002 2013 Nov 12 21 00080030001 2013 Nov 12 6
NGC 3393 SWIFTJ1048.4-2511 0.0125 60061205002 2013 Jan 28 15 00080042001 2013 Jan 28 7
NGC 3588 NED02 SWIFTJ1114.3+2020 0.0262 60061324002 2014 Jan 17 23 00080533001 2014 Jan 17 5
NGC 6232 SWIFTJ1643.2+7036 0.0148 60061328002 2013 Aug 17 18 00080537001 2013 Aug 17 6
NGC 7212 NED02 SWIFTJ2207.3+1013 0.0267 60061310002 2013 Sep 01 24 00080283001 2013 Sep 02 3
UGC 3157 SWIFTJ0446.4+1828 0.0154 60061051002 2014 Mar 18 20 00041747001 2010 Oct 22 10

Note. (1) Full NED object name for BAT counterpart. (2) Swift/BAT name. (3) Redshift. (4) and (5) NuSTAR observation ID and start date (YYYY-MM-DD), respectively. (6) Effective exposure time (ks). This is the net value after data cleaning and correction for vignetting. (7) and (8) Swift/XRT observation ID and start date (YYYY-MM-DD), respectively. (9) Net on-axis, flaring-corrected exposure time (ks).

Download table as:  ASCIITypeset image

The raw data were reduced using the NuSTARDAS software package (version 1.3.1) jointly developed by the ASI Science Data Center and the California Institute of Technology. NuSTARDAS is distributed with the HEAsoft package by the NASA High Energy Astrophysics Archive Research Center. We extracted the NuSTAR source and background spectra using the nuproducts task included in the NuSTARDAS package using the appropriate response and ancillary files. Spectra were extracted from circular regions 40'' in radius, centered on the peak of the centroid of the point-source images. The background spectra were extracted from a circular region lying on the same detector as the source. We also applied the same reduction procedure to the other 75 low-redshift ($z\lt 0.03$) NuSTAR -observed AGNs in the public archive for SCNuSTAR measurements of a total of 84 NuSTAR -observed nearby BAT-detected AGNs.

3.2. Soft X-Ray Observations of SC-selected AGNs

Most NuSTAR observations were accompanied by a short observation (3–7 ks) with Swift/XRT within 24 hr, although for one source, UGC 3157, the only available observation was from 4 yr earlier. These observations provided mostly simultaneous coverage in the soft X-rays (<3 keV), where NuSTAR is not sensitive. All the Swift/XRT data were collected in Photon Counting mode. We built Swift/XRT spectra using the standard point-source processing scripts from the UK Swift Science Data Centre in Leicester (Evans et al. 2009). Table 2 provides the complete list of observations. The Swift/XRT observations of 2MFGC 02280 and NGC 3588 NED02 did not yield a detection below 3 keV, so we use them here only to place an upper limit on the soft X-ray emission.

In addition to the NuSTAR and Swift/XRT data, there are archival XMM-Newton (for NGC 3079 and NGC 7212 NED02) and Chandra data available (for NGC 3393 and NGC 3588 NED02). Studies of NGC 3079, NGC 7212, and NGC 3393 found no signs of variability (Hernández-García et al. 2015; Koss et al. 2015), so we use these spectra because of their much higher sensitivity than Swift/XRT. For NGC 3588 NED02, the source is very faint with a total of 9 counts, all above 3 keV, in the Swift/XRT observation, compared to 72 counts in the Chandra observation. We fit the Chandra, NuSTAR, and Swift/XRT spectra between 3 and 8 keV with a power law and a cross-normalization factor for the Chandra data. We find that Chandra is consistent with no variability (1.2 ± 0.5).

We processed the XMM-Newton observations using the Science Analysis Software, version 13.5.0, with the default parameters of xmmextractor. NGC 3079 had two XMM-Newton observations, which we combined using epicscombine for a total exposure time of 9.4 ks after filtering. After filtering, NGC 7212 had an exposure time of 9.3 ks. We also used epicscombine to combine the MOS1 and MOS2 instruments into a single spectrum before fitting. We reduced and combined the two Chandra observations of NGC 3393, with a total exposure of 99.9 ks, and NGC 3588 NED02, with a total exposure of 9.9 ks, following Koss et al. (2015).

3.3. X-Ray Spectral Fitting

For the three brightest sources with XMM-Newton and Chandra data, NGC 3079, NGC 7212 NED02, and NGC 3393, we binned to a minimum of 20 photons per bin using the HEAsoft task grppha. We use statistic cstat (Wachter et al. 1979) in XSPEC for the remaining six sources, which is more appropriate than χ2 in the case of Poisson-distributed data (Nousek & Shue 1989). In the case of unmodeled background spectra, cstat applies the W statistic.19 While the W statistic is intended for unbinned data, bins containing zero counts can lead to erroneous results,20 so we group the Swift/XRT and NuSTAR data by a minimum of 3 counts per bin, respectively (e.g., Wik et al. 2014), using the HEAsoft task grppha.

We let the Swift/XRT and NuSTAR FPMA and FPMB cross-normalizations vary independently within 5% based on the most recent calibrations (Madsen et al. 2015). We also use varying cross-normalizations of 0.93 ± 0.05 for Chandra, 1.05 ± 0.05 for XMM-Newton pn, and 1.02 ± 0.05 for the combined MOS observation based on recent NuSTAR calibrations.

Conventional spectral fitting and error estimation can sometimes underestimate the likely range of model parameters. Additionally, several past studies have found degeneracies between photon index (Γ) and column density (NH ) for NuSTAR -observed Compton-thick AGNs (e.g., Gandhi et al. 2014). We therefore use Markov chain Monte Carlo methods built into XSPEC for error estimation. We use the default Goodman–Weare algorithm to sample parameter space, constructing a chain of parameter values. The algorithm works by holding multiple sets of parameters, called walkers, for each step in the chain and generating walkers for the next step using those from the current step.

3.4. X-Ray Spectral Modeling

For spectral modeling, we follow the strategies of past studies of a single AGN observed with NuSTAR (e.g., Gandhi et al. 2014) and use the MYTorus and BNTorus models. Since the soft X-ray data (<3 keV) from Swift/XRT typically have a small number of counts (≈0–30) and therefore lack the statistics to model this emission in detail, we use two components corresponding to scattered AGN emission on larger scales in the host galaxy and a thermal plasma. In the absence of high signal-to-noise ratio and spectral resolution soft X-ray data, these are only meant as a simple prescription to describe the spectral shape in this regime. The scattered emission was simulated using a single power law (of photon index Γ and normalization tied to that of the AGN) and with a scattering fraction, fscatt, relative to the intrinsic power law. We also include a low-energy component with a thermal plasma component (APEC; Smith et al. 2001), fixing the abundance to solar, similar to past studies (e.g., Guainazzi et al. 1999).

The MYTorus -based model (Model M hereafter) has the following form:

The common parameters of MYTZ, MYTS, and MYTL (NH and θinc) are tied together. NH is defined along the equatorial plane of the torus. There is a fitted constant (${{\mathtt{f}}}_{{\mathtt{refl}}}$) between the zeroth-order continuum and the scattered/reflected and fluorescent emission-line spectrum. The intrinsic (unprocessed) photon indices and normalizations are tied to those of the zeroth-order continuum (POW). The torus opening angle (θtor) is fixed at 60° in the current version of MYTorus.

Our second choice of physically motivated model (Model T hereafter) uses the BNTorus model and has the following form:

BNTorus self-consistently includes photoelectric absorption, Compton scattering, and fluorescent line emission due to the obscuration of an intrinsic power-law continuum by a biconical torus (Brightman & Nandra 2011), so no ${{\mathtt{f}}}_{{\mathtt{refl}}}$ factor is used in this model. NH is defined along the line of sight and is independent of θinc.

4. RESULTS

Here we discuss our results measuring the SC for the full BAT sample of 241 nearby AGNs, as well as a subset of 84/241 NuSTAR-observed nearby BAT AGNs (Section 4.1). We then discuss results for the nine NuSTAR-observed SC-selected AGNs, focusing on simple and complex X-ray spectral fits (Section 4.2). Next, we examine other measurements of Compton thickness using the mid-IR and [O iii] emission (Section 4.3). We conclude with a comparison of the host galaxy properties and accretions rates compared to all BAT-detected nearby AGNs (Section 4.4).

4.1. Spectral Curvature

The SC measure may be used to measure the overall fraction of Compton-thick sources and whether the curvature of these sources is consistent with torus models. Using SCBAT , about 28% ± 5% (67/241) of AGNs fall within the Compton-thick region (Figure 4). However, the large error bars of SCBAT at low fluxes (<3 × 10−5 counts s−1 ) may overestimate this fraction because of the larger number of sources below the limit. A more conservative estimate using only bright sources (>3 × 10−5 counts s−1 ) finds a Compton-thick fraction of 22% ± 8% (18/79), which is lower than, but not statistically different from, the fraction for the whole sample. We note that both of these fractions are broadly consistent with a recent publication of the intrinsic fraction of Compton-thick AGNs in the entire BAT sample (27 ± 4%, Ricci et al. 2015). The SCBAT identifies most well-known Compton-thick AGNs in the "bona fide" sample (7/10), with the remaining three identified within their 1σ error.

We can also estimate the local Compton-thick space density. Assuming a luminosity threshold of L2–10 keV > 1043 erg s−1 and a conversion factor between 14–195 keV and 2–10 keV of 2.67 (Rigby et al. 2009), we find 74 BAT AGNs. We estimate that 16 are Compton-thick (SCBAT > 0.4). Given that the sample is within z < 0.03 excluding the 10° within the Galactic plane and confused sources, and the 91% completeness at this luminosity threshold (Baumgartner et al. 2013), this implies a volume of 7.0 × 106 Mpc3. The Compton-thick number density is therefore (2.3 ± 0.3) × 10−6 Mpc−3 above L2–10 keV > 1043 erg s−1.

The higher sensitivity of NuSTAR allows a more precise SC measurement than Swift/BAT because of the higher sensitivity. We measure SCNuSTAR for those nearby AGNs with NuSTAR observations (35%, 84/241; Figure 5). While the SC measurements are designed to be independent of the telescope, NuSTAR is studying a somewhat softer energy range because of its reduced sensitivity above 30 keV (14–50 keV vs. 8–30 keV). Thus, the two SC measurements may have systematically different average values. However, for the sample of 84 overlapping sources SCBAT = 0.29 ± 0.07 and SCNuSTAR = 0.27 ± 0.03, showing no evidence of a significant difference, at least on average. We can also compare our SC measurements for "bona fide" NuSTAR-observed Compton-thick AGNs that were confirmed based on spectral fitting (Gandhi et al. 2014). We find that the Compton-thick fraction estimated using SCNuSTAR is also similar (21% ± 7%, 16/84) to the SCBAT measurement.

Figure 5.

Figure 5. Top: plot of the SCNuSTAR measurement for nearby (z < 0.03) BAT AGNs. The nine BAT AGNs targeted based on 14–50 keV curvature are shown in blue. Well-known "bona fide" Compton-thick AGNs observed with NuSTAR are shown in black. The simulation limits for an edge-on torus are shown as a gray dotted line, while the lower limit of a Compton-thick column is shown by a red dotted line. Bottom: band ratios of BAT AGNs observed with NuSTAR (8–24 keV/3–8 keV).

Standard image High-resolution image

We also look for individual sources whose SCBAT and SCNuSTAR measurements differ significantly (Figure 6). The SCBAT is based on time average spectra over 6 yr between 2004 November and 2010 August, which may differ from the NuSTAR observations, which occurred after 2012. We look for sources above 2.6σ, where we expect only one source from statistical noise with a sample size of 84. We find five sources whose change (SCBAT –SCNuSTAR ) was greater than 2.6σ. The majority of these are Seyfert 1s, which are already known to be variable above 10 keV based on NuSTAR (MCG –06-30-015, MCG –05-23-016, NGC4151; Parker et al. 2014; Baloković et al. 2015; Keck et al. 2015). Another source with a significant difference is Circinus. However, analysis of the NuSTAR and Swift/BAT spectra found that the 14–20 keV Swift/BAT energy band was significantly contaminated by a nearby ULX, which would explain the lower value of SCBAT compared to SCNuSTAR (Arévalo et al. 2014). The final source is NGC 6232, one of the program sources, whose SCNuSTAR was 3.1σ below SCBAT . Further NuSTAR observations would be necessary to confirm whether this source is variable.

Figure 6.

Figure 6. Comparison of SC measurement for all 84 BAT AGNs with NuSTAR observations. Compton-thick AGNs confirmed with NuSTAR are shown in black, mildly obscured Seyfert 2s observed with NuSTAR are shown in red, and NuSTAR-observed Seyfert 1s are shown in green. A horizontal black dotted line is plotted through zero. The two measurements show no significant difference on average (SCBAT = 0.29 ± 0.05 and SCNuSTAR = 0.27 ± 0.02).

Standard image High-resolution image

We can also test whether the NuSTAR SC or a broadband ratio (8–24 keV/3–8 keV) is more efficient at finding Compton-thick AGNs. For the well-known Compton-thick AGNs observed with NuSTAR, all (10/10) are found in the Compton-thick region of the 8–30 keV SC measurement. This is an improvement over SCBAT because of the greater NuSTAR sensitivity. By contrast, only 2/10 well-known Compton-thick AGNs are found in the Compton-thick region based on the 8–24/3–8 keV ratio (Figure 5).

We compare the SC measurement from NuSTAR to the one we used with Swift/BAT for selection of the nine sources. The five sources with a Swift/BAT SC above the simulation upper limits for an edge-on torus (NGC 6232, CGCG 164-019, NGC 3393, MCG +06-16-028, UGC 3157) are now within the simulation results based on the NuSTAR SC. We find that 4/9 of the selected sources in our sample are firmly in the Compton-thick region at the 3σ level (NGC 3079, NGC 3393, MCG +06-16-028, 2MFGC 02280). Three fainter sources are just below the Compton-thick cutoff, but are consistent within errors of being Compton-thick (NGC 6232, NGC 7212 NED02, CGCG 164-019). Finally, NGC 3588 NED02 and UGC 3157 are below the cutoff at the 3σ level.

4.2. X-Ray Spectral Fitting

A plot of all the NuSTAR spectra before model fitting can be found in the top panel of Figure 7. We highlight the spectral features by showing the best-fit power-law model in the bottom two panels of Figure 7. Fitting the 3–10 keV NuSTAR spectra with a power-law model indicates Γ < 1 for all sources. This suggests that complex models are required. Additionally, a prominent excess is found at 6.4 keV, matching the Fe Kα emission line. In order to measure the Fe Kα equivalent width, we add a Gaussian component. We find a large equivalent width (>1 keV) for all nine sources. The high value of the equivalent width of the Fe Kα lines is consistent with Compton-thick AGNs (e.g., Krolik & Kallman 1987; Levenson et al. 2002). At higher energies (>10 keV) an excess is seen between 10 and 25 keV for 8/9 AGNs; the remaining source, NGC 6232, shows an excess between 10 and 20 keV. The hard photon index (Γ < 1), high equivalent width fluorescent Fe Kα lines, and Compton hump suggest a strong reflection component (e.g., Matt et al. 2000), which requires more complex models to accurately measure the column density.

Figure 7.

Figure 7. Top: observed NuSTAR spectra of nine galaxies selected based on their SC from Swift/BAT. Spectra for the two focal plane modules have been co-added. We have rebinned the spectra for each galaxy to have similar numbers of points in each panel at levels between 2.5σ (NGC 6232) and 15σ (NGC 3079). The typical background is shown by the filled gray symbols. Almost all sources are above the background level between 3 and 30 keV, except for NGC 6232, which is only above the background between 5 and 20 keV. Bottom left: ratio of the spectra and a simple power-law model fitted to the 3–10 keV spectra (symbols as in the panel above). We find that all sources show spectra consistent with a prominent Fe Kα line (a decrement at 3–5 keV, an excess at 6–7 keV, and a decrement at 7–10 keV). Bottom right: ratio of the spectra and a simple power-law model fitted to the 10–70 keV spectra (symbols as in the panel above). We find that all sources show an excess at 10–25 keV, with the one exception, NGC 6232, showing an excess between 10 and 20 keV, where it is significantly detected.

Standard image High-resolution image

4.2.1. Individual Torus Model Fits

We next self-consistently fit the X-ray absorption and scattering adopting a toroidal structure with fluorescent lines to derive the covering factor and torus inclination. A summary of the best fits to the intrinsic absorbing column density (NH), the intrinsic photon index (Γ), APEC model (kT), diffuse scattering component (fscatt), and intrinsic X-ray luminosity (LX) can be found in Table 3. Figure 8 shows the X-ray spectral fits with the best-fit torus-based model (Model M; described in Section 3.4) for the sample of nine NuSTAR-observed BAT AGNs. For Model T, none of our fits constrained the torus inclination angle, so we fix it here to edge-on (θinc = 85°). For the faintest source in our program, NGC 6232, Model M and Model T fits to the intrinsic absorbing column density (NH ) are poorly constrained, so we fix Γ = 1.9, θinc = 85°, and θtor = 60°, consistent with typical AGNs observed edge-on.

Figure 8.

Figure 8. X-ray spectra of the nine NuSTAR-observed Swift/BAT AGNs selected based on SC. The best-fit MYTorus -based model (Model M ; described in Section 3.4) is shown binned to match the unfolded data. NuSTAR is shown in black (FPMA) and red (FPMB) crosses, while blue crosses represent the soft X-ray data. The soft X-ray data are from XMM-Newton for NGC 3079 and NGC 7212 NED02, Chandra for NGC 3393 and NGC 3588 NED02, and Swift/XRT for the remaining sources. The data are shown grouped to a minimum significance of 3.5σ per bin for visual purposes. The sum of the model is represented by a solid black line. The model components are represented by dashed lines indicating the zeroth-order transmitted continuum through photoelectric absorption (MYTZ, red), the scattered/reflected component (MYTS, green), and fluorescent emission-line spectrum (MYTL, dark blue). At softer energies (<3 keV), there is a model component for scattered AGN emission on larger scales in the host galaxy (fscatt, orange) and in some models there is a thermal plasma component (APEC, pink).

Standard image High-resolution image

Table 3.  Best-fit Models for the NuSTAR + Swift/XRT Phenomenological versus Physically Motivated Torus Models

Object Mod.a χ2/dofb Γ c NHd θinc e θtorf kTg ${L}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}/\mathrm{int}}$ h ${L}_{14\mbox{--}50\;\;\mathrm{keV}}^{\mathrm{obs}/\mathrm{int}}$ h fscatti
        (1024 cm−2) (°) (°) (keV) (1041 erg s−1 ) (1041 erg s−1 )  
2MFGC 02280 M 226/263 ${2.1}_{-0.1}^{+0.1}$ ${3.16}_{-0.30}^{+0.35}$ ${70.8}_{-4.63}^{+4.94}$ 60 ND 0.8/204.6 20.5/113.7 ND
  T 228/264 ${1.9}_{-0.1}^{+0.0}$ ${1.97}_{-0.19}^{+0.71}$ 85 ${48.6}_{-5.95}^{+10.5}$ ND 0.8/46.7 21.0/38.1 ND
CGCG 164-019 M 295/377 ${2.0}_{-0.1}^{+0.3}$ ${3.02}_{-0.86}^{+2.36}$ ${61.3}_{-0.95}^{+0.51}$ 60 ${0.5}_{-0.5}^{+1.4}$ 5.9/57.7 43.7/43.0 ${0.01}_{-0.01}^{+0.00}$
  T 303/378 ${1.6}_{-0.2}^{+0.3}$ ${0.55}_{-0.43}^{+0.24}$ 85 ${51.3}_{-25.3}^{+31.1}$ ${0.1}_{-0.1}^{+0.4}$ 6.1/30.2 41.7/46.8 ${0.05}_{-0.05}^{+0.00}$
MCG +06-16-028 M 465/566 ${1.7}_{-0.0}^{+0.3}$ ${1.71}_{-1.11}^{+4.59}$ ${67.9}_{-7.94}^{+17.6}$ 60 ${0.3}_{-0.3}^{+0.4}$ 2.2/45.3 29.2/55.0 ND
  T 469/567 ${1.8}_{-0.2}^{+0.3}$ ${1.21}_{-0.25}^{+0.26}$ 85 ${62.3}_{-16.5}^{+19.3}$ ${0.1}_{-0.0}^{+0.1}$ 2.2/49.3 28.6/46.6 ND
NGC 3079 M 543/287 ${1.7}_{-0.0}^{+0.0}$ ${2.45}_{-0.11}^{+0.23}$ ${88.3}_{-28.3}^{+1.61}$ 60 ${0.8}_{-0.0}^{+0.0}$ 0.1/Rj 4.6/R ${0.36}_{-0.36}^{+0.00}$
  T 534/288 ${1.4}_{-0.0}^{+0.1}$ ${1.84}_{-0.35}^{+0.28}$ 85 ${77.9}_{-7.16}^{+3.72}$ ${0.8}_{-0.0}^{+0.0}$ 0.1/5.8 4.2/12.2 ND
NGC 3393 M 1175/852 ${1.8}_{-0.1}^{+0.1}$ ${2.06}_{-0.33}^{+0.24}$ ${87.9}_{-27.9}^{+2.01}$ 60 ${0.3}_{-0.0}^{+0.0}$ 1.3/R 38.0/R ${0.04}_{-0.04}^{+0.00}$
  T 1197/853 ${2.1}_{-0.0}^{+0.1}$ ${2.16}_{-0.01}^{+0.32}$ 85 ${26.0}_{-0.0}^{+7.30}$ ${0.2}_{-0.0}^{+0.0}$ 1.4/80.6 32.1/47.2 ND
NGC3588NED01 M 386/483 ${1.7}_{-0.0}^{+0.0}$ ${0.60}_{-0.03}^{+0.07}$ ${89.3}_{-10.4}^{+0.61}$ 60 ND 7.3/75.1 51.7/94.5 ND
  T 385/484 ${1.8}_{-0.1}^{+0.0}$ ${0.57}_{-0.03}^{+0.05}$ 85 ${77.9}_{-4.99}^{+6.02}$ ND 7.4/67.2 51.9/75.2 ND
NGC 6232 M 139/176 1.9 ${1.23}_{-0.23}^{+0.97}$ 85 60 ${0.4}_{-0.4}^{+0.6}$ 0.4/10.6 4.9/10.0 ${0.01}_{-0.01}^{+0.00}$
  T 144/176 1.9 ${3.31}_{-1.31}^{+5.39}$ 85 60 ${0.1}_{-0.1}^{+1.8}$ 0.5/22.6 6.8/21.3 ${0.01}_{-0.01}^{+0.00}$
NGC 7212 NED02 M 174/135 ${1.9}_{-0.0}^{+0.1}$ ${2.64}_{-0.45}^{+0.46}$ ${61.1}_{-0.15}^{+0.15}$ 60 ${0.8}_{-0.0}^{+0.0}$ 9.0/R 76.2/R ${0.02}_{-0.02}^{+0.00}$
  T 227/136 ${2.0}_{-0.1}^{+0.2}$ ${0.90}_{-0.09}^{+0.13}$ 85 ${45.6}_{-6.65}^{+31.3}$ ${0.8}_{-0.1}^{+0.0}$ 8.9/119.0 70.2/89.6 ${0.01}_{-0.01}^{+0.00}$
UGC 3157 M 452/564 ${1.7}_{-0.0}^{+0.3}$ ${0.57}_{-0.18}^{+1.21}$ ${87.0}_{-27.0}^{+3.0}$ 60 ${0.0}_{-0.0}^{+1.9}$ 4.6/26.8 23.8/33.2 ${0.01}_{-0.01}^{+0.00}$
  T 454/565 ${1.8}_{-0.2}^{+0.0}$ ${0.55}_{-0.13}^{+0.09}$ 85 ${59.9}_{-22.9}^{+24.0}$ ${0.0}_{-0.0}^{+1.9}$ 4.6/25.9 23.9/28.7 ND

Notes. Best-fitting model parameters for the 0.5–70 keV spectrum. Parameters without errors have been fixed. The models are detailed in Section 3.4. The errors correspond to 90% confidence level for a single parameter. ND refers to model components that are not detected.

aM—MYTorus ; T—BNTorus model. bThe sources with XMM-Newton data, NGC7212 NED02 and NGC 3079, were binned to 20 counts per bin using ${\chi }^{2}$ statistics, and the remaining sources were binned to 3 counts per bin using Poisson statistics. cThe power-law photon index of the direct and scattered component. dColumn density for the direct component. In the MYTorus models the column density is equatorial rather than line of sight; however, for torus θinc > 60° (which all the MYTorus models converge to here) the difference between the equatorial and line-of-sight column density is less than 3%. eBest-fitting torus inclination angle to the observer. fBest-fitting torus opening angle. The MYTorus model assumes a 60° opening angle. gTemperature of the best-fitting APEC component. A nondetection is listed when the 90% confidence upper limit is less than 0.01 keV. hObserved compared to intrinsic emission. iScattered fraction normalized to the intrinsic direct component. A nondetection is listed when the 90% confidence upper limit is less than a fraction of 0.01. jThe intrinsic luminosity cannot be estimated because the source is reflection dominated (reflected/transmitted > 5) and the transmitted component is not detected.

Download table as:  ASCIITypeset image

Our results confirm that the two brightest sources above 10 keV, NGC 3079 and NGC 3393, are Compton-thick in both Model M and Model T. This is in agreement with past studies with other X-ray telescopes (e.g., Iyomoto et al. 2001; Koss et al. 2015). Model M also suggests that the reflection/scattering component is dominant over the transmitted component. For both NGC 3079 and NGC 3393, the large-scale scattered component contributes the majority of the emission between 1 and 5 keV, with the APEC model dominant below 1 keV. For NGC 3079, the large scattering fraction (fscatt < 0.36) may indicate additional soft components that are poorly fit. For NGC 7212 NED02, Model M suggests a Compton-thick source, whereas Model T suggests a source that is nearly Compton-thick (${N}_{{\rm{H}}}\;=\;{0.90}_{-0.09}^{+0.13}\times {10}^{24}$ cm−2). For NGC 7212 NED02, Model M yields a better fit in terms of ${\chi }^{2}/\mathrm{dof}$, and the difference in column with Model T is likely associated with the strength of the large-scale diffuse scattered component, which is dominant below 3 keV and better fit with Model M . Finally, 2MFGC 02280 was also found to be Compton-thick by both Model M and Model T , consistent with an earlier NuSTAR study (Brightman et al. 2015). 2MFGC 02280 does not have a detection below 3 keV in Swift/XRT and lies in a high Galactic column region (NH = 4 × 1021 cm−2), so there are almost no constraints on the APEC model or diffuse scattering.

CGCG 164-019, MCG +06-16-028, and NGC 6232 also have model fits suggestive of being Compton-thick. Both Model M and Model T suggest Compton-thick levels of obscuration for MCG +06-16-028 and NGC 6232, with relatively large uncertainties in the column density and weak constraints on the APEC model and diffuse component. We note that for NGC 6232, the spectral index was fixed at Γ = 1.9 and the source may not be Compton-thick if Γ < 1.7. CGCG 164-019, like NGC 7212 NED02, has Model M suggesting a Compton-thick source, whereas Model T suggests a source that is only heavily obscured (${N}_{{\rm{H}}}\;=\;{0.55}_{-0.43}^{+0.24}\times {10}^{24}$ cm−2). Some of the differences in column can be attributed to the higher Γ in Model M (${\rm{\Gamma }}\;=\;{2.0}_{-0.0}^{+0.3}$) than Model T (${\rm{\Gamma }}\;=\;{1.6}_{-0.2}^{+0.3}$). Both Model M and Model T yield similar χ2, so it is difficult to say more. Again the five detected counts below 3 keV in the Swift/XRT observations limit how well we can constrain the models.

The sources NGC 3588 NED02 and UGC 3157 have fits suggesting they are Compton-thin but heavily obscured (${N}_{{\rm{H}}}\approx 5\times {10}^{23}$ cm−2) despite large Fe Kα equivalent widths (>1 keV). Past studies have noted, however, that dusty Compton-thin gas can boost the Fe Kα equivalent widths (Gohil & Ballantyne 2015). Swift/XRT detects no counts below 3 keV in NGC 3588 NED02 and only 4 counts in UGC 3157, limiting the model constraints. Finally, we note that a recently published compilation paper on Compton thick AGN in the BAT sample (Ricci et al. 2015) found broad agreement with our analysis (NGC 6232, NGC 7212 NED02, 2MFGC 02280, NGC 3079, NGC 3393, and MCG +06-16-028 were Compton-thick).

4.2.2. Summary of X-Ray Fits

In general, the Model M and Model T fits provide similar quality of fit ${\chi }^{2}/\mathrm{dof}$ to the data with no systematic trend toward higher column density or power-law index for either model. We also report some overall properties from the SC-selected sample using the results from Model M and Model T. The mean power-law index is Γ = 1.88 ± 0.07. The mean column is NH = (1.93 ± 0.38) × 1024 cm−2. We find that the mean observed-to-intrinsic luminosity at 2–10 keV and 14–50 keV is ${L}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}/\mathrm{int}}\;=\;0.05\pm 0.02$ and ${L}_{14\mbox{--}50\;\;\mathrm{keV}}^{\mathrm{obs}/\mathrm{int}}$ = 0.59 ± 0.08, respectively. For this calculation, we used the Model T observed-to-intrinsic luminosity estimates for the reflection-dominated sources as determined from Model M . The torus inclination is found to have values between 60° and 90° (edge-on), while the torus half opening angle spans 26°–78°.

Interestingly, our individual spectral fits broadly agree with our Compton-thick selection based on SC. Using SC, we found that NGC 3079, NGC 3393, MCG +06-16-028, and 2MFGC 02280 were all Compton-thick. The three fainter sources that lie just below the Compton-thick cutoff but within 1σ error of being Compton-thick are all sources that might be Compton-thin using Model T. NGC 3588 NED02 and UGC 3157 are the only two sources significantly below the SC Compton-thick cutoff and are also Compton-thin when fitting the spectra.

Finally, we tried fitting a so-called MYTorus "decoupled" model, where the NH of the MYTS and MYTL components are allowed to vary compared to the zeroth-order transmitted continuum. In this "decoupled" model, the geometry can be thought of as a patchy torus where the global column density experienced by the scattered/reflection and fluorescent emission is different from the line-of-sight column density. However, we found that the quality of fit was not significantly better for any of the sources and the two-component NH was poorly constrained given the quality of the data.

In summary, we find that torus models suggest that Compton-thick column densities are preferred for most (78% or 7/9) of the sources selected based on their SC values, with the remaining two sources being heavily obscured (NH > 5 × 1023 cm−2). We note that our study is limited in that 6/9 sources have low-quality Swift/XRT or Chandra data with only a handful of counts below 3 keV. Higher-quality spectroscopy would be required to place stronger constraints on the large scattered component, which may affect the column density measurements. While much brighter Compton-thick AGNs such as NGC 1068 exhibit additional complexities and components that are important to model (e.g., Bauer et al. 2014), testing for these is currently not possible because of the faintness of these sources and the comparatively low photon statistics from the short, ≈20 ks observations.

4.3. Other Measures of Compton Thickness

Another way to identify Compton-thick AGNs is to use additional intrinsic luminosity indicators (e.g., L6 μm, L[O III] ) to compare to the observed X-ray luminosity. A high mid-IR/X-ray ratio and/or high [O iii] /X-ray ratio may indicate a Compton-thick AGN. A summary of all the indicators including the results from the X-rays is provided in Table 4.

Table 4.  Compton-thick AGN Indicators

Object Fe Kα EW Γ NuSTAR MYTorus ${F}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}}$/${F}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{pred}\;6\;\mu {\rm{m}}}$ IR SED ${F}_{2\mbox{--}10\;\mathrm{keV}}^{\mathrm{obs}}$/${F}_{[{\rm{O}}{\rm{III}}]}$
  > 1 keV < 1 SC Model < 20 E(BV) < 1
2MFGC 02280 T T T T T T N
CGCG 164-019 T T T T T N N
MCG +06-16-028 T T T T T N N
NGC 3079 T T T T T N T
NGC 3393 T T T T T T T
NGC 3588 NED02 T T N N N T N
NGC 6232 T T N T T N N
NGC 7212 NED02 T T N T T T ...
UGC 3157 T T N N T T N

Note. Results of various tests of Compton thickness. N indicates that the object is Compton-thin, T indicates that the object was classified as Compton-thick, and an ellipse indicates that the test could not be performed on the object because of a lack of data.

Download table as:  ASCIITypeset image

4.3.1. IR Emission

The 6 μm AGN emission provides an additional assessment of the intrinsic AGN luminosity (e.g., Gandhi et al. 2009). Moderate-luminosity AGNs, however, can have the majority of their mid-IR emission from host galaxy contributions (e.g., Stern et al. 2012) rather than the AGN. We therefore first measure the WISE colors to test whether the sources show colors indicative of AGNs and are dominated by AGN emission in the mid-IR. We find that only three sources have $W1-W2\gt 0.8$, indicating that the AGN likely dominates the mid-IR emission (MCG +06-16-028, CGCG 164-019, and NGC 7212 NED02). It is therefore important to fit the spectral energy distributions (SEDs) to measure the intrinsic 6 μm AGN emission. The observed photometry includes optical (griz from Koss et al. 2011b), near-infrared (NIR; Two Micron All Sky Survey [2MASS] JHK), mid-IR (WISE, 3.4–22 μm), and GALEX or Swift far-ultraviolet and near-ultraviolet photometry when available. We follow the photometry procedure of Koss et al. (2011b) using Kitt Peak or 2MASS data to measure the optical and NIR photometry. We use the Assef et al. (2010) 0.03–30 μm algorithm to model the strength of the AGN emission in the mid-IR using empirical AGN and galaxy templates. The template SEDs (Figure 9) suggest a Compton-thick level of obscuration for most of the sample (5/9, 55%) based on the NH /E(BV) = 1.5 × 1023 cm−2 conversion from reddening to column density (Maiolino et al. 2001).

Figure 9.

Figure 9. SEDs of the nine SC-selected galaxies. The observed photometry includes optical (griz), NIR (JHK), and mid-IR (3.4–22 μm), as well as FUV and NUV photometry when available. We used the Assef et al. (2010) 0.03–30 μm algorithm to model the strength of the AGN emission in the IR using empirical AGN and galaxy templates. The plot points represent observed data (green circles) and predicted SED model flux (open triangles). The total best-fit SED template line (solid black) was made by combining the AGNs (dashed blue) and old (E, dashed red), intermediate (Sbc, dashed green), and young (Im, dashed cyan) galaxy stellar populations.

Standard image High-resolution image

For NGC 3079, MCG +06-16-028, and NGC 6232, almost no obscuration is detected ($E[B-V]\lt 0.11$) in the best-fit SEDs. This is inconsistent with the X-ray spectral fitting and the lack of broad lines in the optical spectra. To understand the fitting better, we produced 1000 resampled SEDs based on the measured photometry but resampled by the photometric errors (assuming Gaussian noise) and find that the SED fits do include a small percentage of solutions with Compton-thick obscuration (<5%). The high AGN obscuration (NH > 1024 cm−2 ), combined with significant host galaxy star formation as shown by the bright UV and IR emission as seen in Herschel (Meléndez et al. 2014), makes SED techniques unable to accurately model the AGN SED in some cases due to the lack of photometric bands where the AGN is dominant. Further mid-IR studies using high spatial resolution imaging of nuclear emission (e.g., Asmus:2014:1648; Asmus et al. 2014) are required to resolve these degeneracies.

We estimate the intrinsic 6 μm AGN emission from the template fitting and compare it to the observed 2–10 keV X-ray emission. We compare it to the ratio obtained from large-sample studies of AGNs (Stern 2015). This ratio is luminosity dependent, with a predicted X-ray to mid-IR ratio from 0.87 for our least luminous mid-IR source (NGC 6232) to 0.44 for the most luminous mid-IR source (NGC 7212 NED02). Assuming Model M and Γ = 1.9, the observed 2–10 keV X-ray emission is diminished by a factor of ≈20 at Compton-thick obscuration. We find that the observed X-ray emission is fainter than expected from the mid-IR (Figure 10) in all our sources by an average factor of 72 ± 29. The smallest difference is NGC 3588 NED02 (factor of 13 lower X-ray emission), and the largest difference is NGC 3079 (factor of 298 lower). NGC 3588 NED02 is also the only AGN whose mid-IR to X-ray ratio (${L}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}}$ /L6 μm ) is consistent with a Compton-thin AGN. Thus, the intrinsic mid-IR AGN emission confirms that the observed X-ray emission is consistent with Compton-thick AGNs for most of the sample.

Figure 10.

Figure 10. L6 μm from SED fitting compared to ${L}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}}$ (red dot) and L2–10 keV (blue dot) based on X-ray model fitting. A black dashed line indicates the relation from Stern (2015), and a gray region has been shaded within a factor of five of the mean ratio ( ±0.7 dex). Sources below the blue dashed line are likely to be Compton-thick. All the sources except for NGC 3588 NED02 are below the Compton-thick line based on their mid-IR to observed X-ray emission. Finally, we find that even with the X-ray spectral fitting, the majority of the intrinsic estimates of the source luminosity are below the values expected from the L6 μm , suggesting that they may still be underestimated.

Standard image High-resolution image

4.3.2. Optical Spectroscopy

We compare the optical spectroscopy of these sources with other BAT-detected AGNs (Koss et al. 2016, in preparation). We have optical spectra for all sources except NGC 7212 NED02. We first apply AGN emission-line diagnostics (e.g., Veilleux & Osterbrock 1987; Kewley et al. 2006) using the [N ii]/Hα diagnostic. All the AGNs are in the Seyfert or LINER region, except NGC 3588 NED02, which falls in the composite region, but this has been found with many other AGNs in close mergers (e.g., Koss et al. 2011a). The Hβ line for 2MFGC 02280 is not detected, but the limit places it most likely in the Seyfert or LINER region. The ratio of the observed 2–10 keV X-ray to Balmer decrement corrected [O iii] line strength provides a measure of Compton thickness (Bassani et al. 1999). We find that only NGC 3393 and NGC 3079 show an excess in the Balmer decrement corrected [O iii] versus X-ray luminosity ratio consistent with a Compton-thick AGN (${F}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}}$/${F}_{[{\rm{O}}{\rm{III}}]}\lt 1$).

AGNs are known to have a wide range in [O iii] to X-ray ratios. One possibility for the low values of [O iii] is a low scattering fraction from a "buried" AGN (Ueda et al. 2007) with a small opening angle and/or having an unusually small amount of gas responsible for scattering. Additionally, if these AGNs have high Eddington ratios, they should have relatively weak [O iii] as found by the "Eigenvector 1" relationships (e.g., Boroson & Green 1992). Finally, AGN "flickering" on shorter timescales than the light-travel time to the ionized regions can cause some AGNs to have much stronger X-ray emission since it has just begun to ionize the narrow-line region (Schawinski et al. 2015). Noguchi et al. (2010) found that optical emission-line studies are biased against "buried" AGNs that have a small scattering fraction or a small amount of narrow-line region gas. AGNs with a low ratio of [O iii] to X-ray luminosity (L[O III] /L2–10 keV ) tend to be "buried" AGNs. We use the estimated ratio obtained from large studies of AGNs (Berney et al. 2015). In our sample, we find that only 2MFGC 02280 is consistent with a "buried" AGN in that the X-ray emission is significantly outside the scatter of the mean [O iii] to X-ray ratio (L[O III] /${L}_{14\mbox{--}195\;\mathrm{keV}}^{\mathrm{obs}}$ ). NGC 3588 NED02 does have a higher X-ray to [O iii] ratio, but this ratio has been found to be elevated in many merging AGN galaxies (Koss et al. 2010). In summary, the majority of the sample does not show evidence of having uniquely high intrinsic X-ray to [O iii] values.

4.4. Host Morphology and Accretion Rates

We investigate whether our sources have unique host morphologies or accretion rates compared to the rest of the nearby BAT AGN. Tricolor gri filter images for the nine sources selected based on Swift/BAT SC are shown in Figure 11. We find that 22% (2/9) of the sample is in close mergers (<10 kpc). In both sources, faint tidal tails and radial velocity differences of less than 500 km s−1 (from NED) between the sample galaxy and its possible companion suggest an ongoing major merger rather than a chance association. NGC 3588 NED02 has a separation of 4.2 kpc (8farcs1). NGC 7212 NED02 is in a galaxy triple with a separation of 9.8 kpc (18farcs3) from NGC 7212 NED03 and a separation of 22 kpc (41'') from NGC 7212 NED01. This fraction is higher than typically found for BAT AGNs; Koss et al. (2010) found that 8% (11/144) of BAT AGNs are in close mergers (<10 kpc), though consistent within Poisson errors. With such a small sample size, this difference is not significant based on a Fisher exact test.

Figure 11.

Figure 11. Tricolor optical images in gri displayed with an arcsinh scale for nine sources selected based on Swift/BAT SC. Images are 1' on a side except for NGC 3079, which is 8' on a side. A red dashed circle indicates the BAT-detected counterpart based on soft X-ray data from Swift/XRT. For NGC 3588 NED02 and NGC 7212 NED02, Chandra data confirm that the majority of the hard X-ray emission is coming from the galaxy nucleus and not the merging counterpart (>95% at 2–8 keV). The high fraction of sources in close (<10 kpc) mergers (22%, 2/9) and/or highly edge-on (b/a < 0.22) galaxies (22%, 2/9) suggests a possible connection of high levels of obscuration to the final merger stage and host galaxy inclination.

Standard image High-resolution image

To investigate this further, we use the Koss et al. (2010) study, which had high-quality imaging and optical spectroscopy to study the companion separation. We find that the other two NuSTAR-observed AGNs in close mergers, NGC 6240 and 2MASX J00253292+6821442, are also in the SC Compton-thick regime (SCNuSTAR = 0.49 ± 0.01 and SCNuSTAR = 0.56 ± 0.05; NGC 6240 and 2MASX J00253292+6821442). The likelihood of finding all four sources being Compton-thick is <1% based on a Fisher test.

Another interesting morphological feature is that 22% (2/9) of the galaxies from the program (NGC 3079 and 2MFGC 02280) are nearly edge-on. Koss et al. (2011b) derived the axis ratio (b/a) from the major and minor axes derived from isophotal r-band photometry for both of these galaxies (NGC 3079; b/a = 0.15 and 2MFGC 02280; b/a = 0.21). By comparison, 6% of BAT AGNs have b/a < 0.22 (Koss et al. 2011b). The frequency of edge-on galaxies selected using SC versus the other BAT AGNs is not statistically significant because of the small sample size based on a Fisher test (11% chance), implying that larger samples are needed.

We compare the black hole mass and Eddington ratio of our sources to the other nearby BAT-detected AGNs. We use the velocity dispersion measurements for measurements of black hole mass (from Koss et al. 2016, in preparation) and the median and median absolute deviation (MAD) to compare the populations because of the spread over several orders of magnitude. We find that the typical black hole mass of our sample is a factor of four smaller than typical BAT-detected AGNs (MBH = (1.3 ± 0.4) × 107 M versus MBH = (5.1 ± 0.4) × 107M), where the error refers to the MAD 1σ error.

We also estimated the bolometric luminosity Lbol from the X-ray luminosity (${L}_{14\mbox{--}195\;\mathrm{keV}}^{\mathrm{obs}}$ ) using the bolometric corrections from Vasudevan & Fabian (2009). Including the absorption-corrected 14–195 keV emission based on the NuSTAR spectral fitting of our sources, the typical Eddington ratio of our sources is about a factor of four larger (λEdd = 0.068 ± 0.023 compared to λEdd = 0.016 ± 0.004), where the error refers to the 1σ error in the MAD (Figure 12). A Kolmogorov–Smirnov test indicates a <1% chance that the distribution of Eddington ratios for SC-selected BAT AGNs is from the same parent distribution as the other BAT AGNs. This indicates that the SC-selected BAT AGNs have, on average, higher accretion rates.

Figure 12.

Figure 12. Eddington ratio of our heavily obscured AGNs compared to other nearby (z < 0.03) BAT-detected AGNs. We find that these sources typically are in the upper left quadrant of the sample with higher accretion rates and smaller black holes than typical BAT-detected AGNs.

Standard image High-resolution image

5. SUMMARY

We define a new SC measure of Compton thickness using weighted averages of different energy bands in the low-sensitivity Swift/BAT survey. We then select nine AGNs for NuSTAR follow-up to study their possible Compton-thick nature.

  • i.  
    We find that all nine targeted sources are consistent with Compton-thick AGNs in the majority of indicators (e.g., Fe Kα EW > 1 keV, Γ < 1, X-ray spectra fitting, mid-IR indicators), confirming the effectiveness of the SC method to identify new Compton-thick AGNs. Using NuSTAR spectroscopy, the majority of the nine targets are consistent with Compton-thick AGNs using MYTorus models (78%, 7/9), and the remaining two are nearly Compton-thick (NH ≃ (5–8) × 1023 cm−2 ). The observed 2–10 keV emission compared to the 6 μm emission is also consistent with a Compton-thick AGN for most sources (8/9, 89%; ${F}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}}$/${F}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{pred}\;6\;\mu {\rm{m}}}\gt 20$).
  • ii.  
    Our results suggest that the >10 keV emission may be the only way to identify this population of Compon-thick AGNs other than through detailed SED fitting. We find that only two sources show evidence of an excess in the Balmer decrement corrected [O iii] versus X-ray luminosity consistent with a Compton-thick AGN (${F}_{[{\rm{O}}{\rm{III}}]}/{F}_{2\mbox{--}10\;\;\mathrm{keV}}^{\mathrm{obs}}\gt 1$). As expected for lower-luminosity AGNs, we find that most sources (6/9, 67%) would not be identified as AGNs using WISE colors, though detailed SED fitting with the mid-IR would identify most sources.
  • iii.  
    We find the SCBAT and SCNuSTAR measurements to be consistent on average (SCBAT = 0.29 ± 0.07 vs. SCNuSTAR = 0.27 ± 0.03). This suggests that this measure can be used with other satellites with >10 keV coverage such as Astro-H, or high-redshift AGNs (z > 3) observed with Chandra or Athena, where the bands are shifted to cover rest frame 10–30 keV.
  • iv.  
    We find that the SC measure is much more effective at selecting Compton-thick AGNs than band ratios (8–24 keV/3–8 keV), finding 10/10 well-known Compton-thick AGNs compared to only 2/10 using band ratios.
  • v.  
    We find that these heavily obscured AGNs have smaller black holes ($\langle {M}_{\mathrm{BH}}\rangle \;=\;(1.3\pm 0.36)\times {10}^{7}$M vs. $\langle {M}_{\mathrm{BH}}\rangle \;=\;(5.1\pm 0.39)\times {10}^{7}$M) and higher accretion rates than other BAT-detected AGNs ($\langle {\lambda }_{\mathrm{Edd}}\rangle \;=\;0.068\pm 0.023$ compared to $\langle {\lambda }_{\mathrm{Edd}}\rangle \;=\;0.016\pm 0.004$).

We find that the four NuSTAR-observed sources in very close mergers (<10 kpc) are all found to be Compton-thick, suggesting a physically plausible link between increased gas supply and obscuration, which might be natural in the early stages of a merger (e.g., Sanders et al. 1988; Hopkins et al. 2005). Based on simulations, the timescale within 10 kpc for major mergers is relatively short, on the order of 100–200 Myr (Van Wassenhove et al. 2012), so finding even a small number of galaxies may be significant. Another interesting morphological feature is that 2/9 program sources are in extremely edge-on galaxies (b/a < 0.25), suggesting that galaxy-wide extinction may be important for some sources. This compares to only 6% of BAT AGNs. The likelihood of this occurring by chance is 11%, implying that larger samples are needed.

Based on the robustness of SC in identifying Compton-thick AGNs, we measure the fraction of Compton-thick nearby BAT AGNs (z < 0.03) as ≈22% (SCBAT = 22%, SCNuSTAR = 21%). The Compton-thick number density is (2.3 ± 0.3) × 10−6 Mpc−3 above L2–10 keV > 1043 erg s−1 . This is a conservative estimate since Swift/BAT likely misses reflection-dominated AGNs. This number is significantly higher than previous work with Swift/BAT, which reported only a handful of Compton-thick AGNs corresponding to fractions of a few percent (e.g., Tueller et al. 2008; Winter et al. 2009; Burlon et al. 2011; Vasudevan et al. 2013). This 22% fraction is in line with estimates of the intrinsic Compton-thick fraction in X-ray background population-synthesis models (5%–52% of obscured AGNs; for review see Ueda et al. 2014).

These Compton-thick AGNs show high Eddington ratios consistent with other well-known Compton-thick AGNs in the BAT sample already observed with NuSTAR (e.g., Circinus, ${\lambda }_{\mathrm{Edd}}\;=\;0.2;$ NGC 4945, λEdd = 0.1–0.3; NGC 1068, λEdd = 0.5–0.8—Arévalo et al. 2014; Bauer et al. 2014; Puccetti et al. 2014) and also in recent results from the XMM-COSMOS survey (Lanzuisi et al. 2015b). This suggests that the sum of black hole growth in Compton-thick AGNs (Eddington ratio times population percentage) may be nearly as much as the rest of the population of mildly obscured AGNs and unobscured AGNs. A highly obscured (NH > 1024 cm−2 ), high-Eddington population (λEdd > 0.1) like these AGNs could be important for resolving discrepancies based on considerations of the Soltan argument (e.g., Brandt & Alexander 2015; Comastri et al. 2015). Additionally, the high Eddington ratios with relatively weak [O iii] to X-ray ratio, despite being Compton-thick, are consistent with the "Eigenvector 1" relationships (e.g., Boroson & Green 1992). In further studies, we will use much larger samples of BAT-detected AGNs with measured black hole masses and accretion rates to study which populations have most of the black hole growth (Koss et al. 2016, in preparation).

We acknowledge financial support from Ambizione fellowship grant PZ00P2_154799/1 (M.K.), the Swiss National Science Foundation (NSF) grant PP00P2 138979/1 (M.K. and K.S.), the Center of Excellence in Astrophysics and Associated Technologies (PFB 06), by the FONDECYT regular grant 1120061 and by the CONICYT Anillo project ACT1101 (E.T.), NASA Headquarters under the NASA Earth and Space Science Fellowship Program, grant NNX14AQ07H (M.B.), NSF award AST 1008067 (D.B.), Caltech NuSTAR subcontract 44A-1092750 and NASA ADP grant NNX10AC99G (W.N.B.), and the ASI/INAF grant I/037/12/0 011/13 and the Caltech Kingsley visitor program (A.C.). M.K. also acknowledges that support for this work was provided by the National Aeronautics and Space Administration through Chandra Award Number AR3-14010X issued by the Chandra X-ray Center, which is operated by the Smithsonian Astrophysical Observatory for and on behalf of the National Aeronautics Space Administration under contract NAS8-03060. This work was supported under NASA Contract No. NNG08FD60C and made use of data from the NuSTAR mission, a project led by the California Institute of Technology, managed by the Jet Propulsion Laboratory, and funded by the National Aeronautics and Space Administration. We thank the NuSTAR Operations, Software and Calibration teams for support with the execution and analysis of these observations. This research has made use of the NuSTAR Data Analysis Software (NuSTARDAS) jointly developed by the ASI Science Data Center (ASDC, Italy) and the California Institute of Technology (USA). This research made use of the XRT Data Analysis Software (XRTDAS), archival data, software, and online services provided by the ASDC. This work made use of data supplied by the UK Swift Science Data Centre at the University of Leicester. The scientific results reported in this article are based on data obtained from the Chandra Data Archive (Obs ID = 4078, 4868, 12290, 13895). This work is based on observations obtained with XMM-Newton (Obs ID = 0110930201, 0147760101, 0200430201), an ESA science mission with instruments and contributions directly funded by ESA Member States and NASA.

Facilities: NuSTAR - The NuSTAR (Nuclear Spectroscopic Telescope Array) mission, Swift - , XMM - , Sloan - , CXO -

Footnotes

Please wait… references are loading.
10.3847/0004-637X/825/2/85