Beyond Simple AGN Unification with Chandra-observed 3CRR Sources at 0.5 < z < 1

The 5 GHz radio core and extended radio lobe flux densities have been compiled from the literature. The radio core, the total (core+lobe) flux densities, and the total luminosity densities at 5 GHz (L_ν (5 GHz)) are presented in Table 1. Also given is the radio core fraction R_CD, which is often used as an orientation indicator in radio-loud AGN (Orr & Browne 1982; Ghisellini et al. 1993) and gives, in general, an estimate of the inclination angle accurate to within ±20° (Wills & Brotherton 1995) and, in the case of the z ≥ 1 3CRR sources, ±10° or less (Marin & Antonucci 2016). When available, we used the same reference for the radio core and extended radio lobe luminosities when calculating R_CD. Other references were checked for flux consistency. For sources with no 5 GHz data, the 8 GHz flux density was used to estimate the 5 GHz flux density, assuming a radio spectral index of α = 0.7 (typical of extended emission from radio galaxies; e.g., Dennett-Thorpe et al. 1999) for the radio lobes and α = 0.3 (a compromise between a flat-spectrum and a steep-spectrum core) for the radio core (where F_ν ∝ ν^−α ). In the medium-z 3CRR sample, log R_CD spans values from 0.15 to less than −3.5, which, according to Marin & Antonucci (2016), correspond to a range of viewing angles measured in respect to the radio jets that range between 8° (close to pole-on) and 90° (perpendicular to the jet or edge-on to the torus).

Spitzer (Werner et al. 2004) IRAC and MIPS photometry has been obtained and analyzed for the full 3CRR sample (Haas et al. 2008 for z > 1 and Ogle et al. 2006 for z < 1 sources). The IRS spectroscopy is also available for sources in the redshift range 0.4 < z < 1.4 (Cleary et al. 2007 for 0.4 < z < 1.2 and Leipski et al. 2010 for 1 < z < 1.4). All sources were observed in the far-IR during Herschel guaranteed time (PI: Barthel) with PACS and SPIRE, and their IR spectral energy distributions (SEDs; including 2MASS, WISE, Spitzer, and Herschel data) were analyzed by Podigachoski et al. (2015) for z > 1 and Westhues et al. (2016) for z < 1. The near-to-mid-IR (3–40 μm) emission, dominated by the AGN, was found to be stronger in quasars than in radio galaxies, while the far-IR component, dominated by dust heated by star formation, is comparable in strength for the two classes. The difference in the mid-IR emission is consistent with the unification scenario, where the hot dust from the inner regions is directly visible in face-on quasars but obscured in NLRGs, which are viewed edge-on to the dusty torus. At z < 1, an additional population of weak mid-IR AGN was found (LERGs and weak mid-IR sources), possibly representing a different class of objects (nonthermal and jet-dominated with low accretion power) or different evolutionary stage from the mid-IR-bright sources (Ogle et al. 2006).

The Chandra data, both new and archival, were reprocessed using the standard pipeline to apply the latest calibration products appropriate for their observation dates and assure that processing was uniform across the sample. The counts for each source were extracted from a 22 radius circle (to enclose the full point-spread function) centered on the radio core coordinates or the AGN X-ray position when the radio core position was not available (Table 1). The background counts were extracted from an annulus with inner and outer radii of 15'' and 35'', respectively, centered on the AGN, then scaled for area and subtracted to determine the net counts for each source. In a few sources, the background annulus was adjusted to exclude bright incidental X-ray sources. For 11 sources (3C 172, 175, 228, 263, 265, 268.1, 330, 334, 340, 337, and 441) for which the radio lobes showed substantial and extended X-ray emission, two circular regions with a 15'' radius lying outside the extended emission were used for background count estimation.

We use the following X-ray energy bands: broad (B = 0.5–8.0 keV), soft (S = 0.5–2.0 keV), and hard (H = 2.0–8.0 keV). The broadband net source and background counts for each source are given in Table 2 (columns (3) and (4)). The soft- and hard-band source and background counts were used to calculate hardness ratios (HRs; column (14)).

Table 2. X-Ray Source Parameters

Name	Chandra	Net Counts	Bkgrd. Counts	F(0.5−8 keV)	log L(0.5−8 keV)	Γ	NH	f(1 keV)	Reduced	F(0.5−8 keV)	F(0.5−8 keV)	log L(0.5−8 keV)	HR
	ObsID	(0.5–8 keV)	(0.5–8 keV)	Srcflux	Srcflux		1022 cm−2	10−6	χ2	Observed	Intrinsic
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
3C 6.1	3009	1718.8 ± 41.5	1.22 ± 0.08	${33.60}_{-0.90}^{+0.80}$	${45.04}_{-0.01}^{+0.01}$	${1.76}_{-0.09}^{+0.10}$	${0.32}_{-0.12}^{+0.13}$	${14.52}_{-1.28}^{+1.44}$	0.7	${43.59}_{-5.05}^{+5.03}$	${47.33}_{-5.81}^{+7.20}$	${45.19}_{-0.06}^{+0.06}$	$-{0.29}_{-0.02}^{+0.02}$
3C 6.1	4363	811.1 ± 28.5	0.87 ± 0.07	${28.90}_{-1.00}^{+1.10}$	${44.98}_{-0.02}^{+0.02}$	${1.62}_{-0.12}^{+0.12}$	${0.26}_{-0.12}^{+0.14}$	${11.26}_{-1.23}^{+1.39}$	0.7	${39.97}_{-5.55}^{+7.38}$	${42.91}_{-7.35}^{+8.76}$	${45.15}_{-0.08}^{+0.08}$	$-{0.29}_{-0.03}^{+0.03}$
3C 22	14994	68.8 ± 8.3	0.22 ± 0.03	${4.38}_{-0.65}^{+0.72}$	${44.27}_{-0.07}^{+0.07}$	1.9	${21.24}_{-7.36}^{+16.02}$	${11.31}_{-3.64}^{+6.21}$	0.5	${12.53}_{-5.03}^{+5.62}$	${31.14}_{-10.35}^{+15.71}$	${45.13}_{-0.18}^{+0.18}$	${0.59}_{-0.08}^{+0.11}$
3C 34	16046	73.7 ± 8.6	0.32 ± 0.04	${6.18}_{-0.79}^{+0.79}$	${44.12}_{-0.05}^{+0.06}$	1.9	${12.32}_{-3.09}^{+3.09}$	${56.31}_{-13.44}^{+13.44}$	0.21	${5.70}_{-1.52}^{+1.52}$	${25.37}_{-0.33}^{+0.33}$	${44.74}_{-0.01}^{+0.01}$	${0.67}_{-0.08}^{+0.09}$
3C 41	16047	37.5 ± 6.1	0.47 ± 0.05	${2.40}_{-0.52}^{+0.46}$	${43.87}_{-0.09}^{+0.09}$	1.9	${29.45}_{-13.11}^{+13.11}$	${48.37}_{-21.75}^{+21.75}$	0.21	${2.86}_{-1.42}^{+1.42}$	${16.81}_{-0.30}^{+0.30}$	${44.71}_{-0.01}^{+0.01}$	${0.78}_{-0.08}^{+0.12}$
3C 49	14995	161.8 ± 12.7	0.24 ± 0.03	${12.30}_{-1.00}^{+1.00}$	${44.29}_{-0.04}^{+0.03}$	1.9	${6.38}_{-0.91}^{+1.11}$	${17.64}_{-2.49}^{+2.63}$	0.2	${25.16}_{-3.22}^{+3.46}$	${52.81}_{-8.09}^{+6.33}$	${44.92}_{-0.07}^{+0.05}$	${0.39}_{-0.06}^{+0.09}$
3C 55	16050	15.6 ± 4.0	0.35 ± 0.04	${17.70}_{-4.90}^{+4.20}$	${44.65}_{-0.11}^{+0.12}$	1.9	⋯	${2.27}_{-0.98}^{+0.98}$	0.4	${1.08}_{-0.40}^{+0.40}$	...	...	$-{0.12}_{-0.27}^{+0.22}$
3C 138	14996	388.9 ± 19.7	0.10 ± 0.02	193.00 ± 10.00	${45.70}_{-0.02}^{+0.02}$	1.9	<0.96	${71.07}_{-6.74}^{+7.51}$	1.2	${192.48}_{-29.60}^{+24.94}$	${211.64}_{-18.91}^{+17.16}$	${45.74}_{-0.04}^{+0.03}$	$-{0.15}_{-0.05}^{+0.05}$
3C 147	14997	151.0 ± 12.3	0.04 ± 0.01	${81.30}_{-6.60}^{+6.70}$	${44.98}_{-0.04}^{+0.03}$	1.9	<0.95	${30.61}_{-4.21}^{+4.48}$	0.3	${79.41}_{-10.78}^{+12.55}$	${91.97}_{-17.82}^{+11.84}$	${45.03}_{-0.09}^{+0.05}$	$-{0.31}_{-0.08}^{+0.07}$
3C 172	14998	31.7 ± 5.7	0.30 ± 0.05	${1.84}_{-0.37}^{+0.44}$	${43.28}_{-0.10}^{+0.09}$	1.9	${82.97}_{-42.29}^{+75.16}$	${31.92}_{-21.31}^{+82.59}$	0.5	${6.88}_{-5.21}^{+13.59}$	${135.73}_{-87.35}^{+70.93}$	${45.15}_{-0.45}^{+0.18}$	${0.73}_{-0.10}^{+0.13}$
3C 175	14999	354.9 ± 18.8	0.07 ± 0.03	${155.00}_{-9.00}^{+8.00}$	${45.62}_{-0.03}^{+0.02}$	1.9	<0.56	${54.81}_{-5.01}^{+5.41}$	0.7	${152.51}_{-17.48}^{+13.15}$	${162.99}_{-16.42}^{+14.80}$	${45.64}_{-0.05}^{+0.04}$	$-{0.32}_{-0.05}^{+0.05}$
3C 175.1	15000	88.7 ± 9.4	0.31 ± 0.04	${6.72}_{-0.75}^{+0.76}$	${44.44}_{-0.05}^{+0.05}$	1.9	${1.23}_{-0.56}^{+0.79}$	${3.73}_{-0.72}^{+0.80}$	0.2	${8.62}_{-1.77}^{+2.00}$	${11.40}_{-2.03}^{+2.00}$	${44.67}_{-0.08}^{+0.07}$	$-{0.21}_{-0.10}^{+0.10}$
3C 184	3226	47.5 ± 6.9	0.47 ± 0.05	${1.17}_{-0.17}^{+0.20}$	${43.76}_{-0.07}^{+0.07}$	1.9	<2.23	${0.24}_{-0.09}^{+0.13}$	0.3	${2.07}_{-0.27}^{+0.21}$	${0.72}_{-0.34}^{+0.35}$	${43.55}_{-0.28}^{+0.17}$	${0.30}_{-0.12}^{+0.15}$
3C 184 a	XMM	776 ± 65	⋯	${1.75}_{-0.24}^{+0.27}$	${43.94}_{-0.06}^{+0.06}$	${1.4}_{-0.2}^{+0.3}$	${48.7}_{-12.1}^{+22.0}$	${24}_{-10}^{+11}$	39.3	${17}_{-6}^{+7}$	⋯	${44.8}_{-0.2}^{+0.1}$	⋯
3C 196	15001	89.9 ± 9.5	0.07 ± 0.02	${32.60}_{-3.60}^{+3.60}$	${45.07}_{-0.05}^{+0.05}$	1.9	${2.68}_{-0.85}^{+1.17}$	${24.85}_{-5.06}^{+5.60}$	0.1	${48.60}_{-11.55}^{+11.40}$	${70.97}_{-13.87}^{+14.98}$	${45.41}_{-0.09}^{+0.08}$	$-{0.07}_{-0.10}^{+0.10}$
3C 207	2130	6462.7 ± 80.4	8.34 ± 0.21	${85.40}_{-1.10}^{+1.10}$	${45.23}_{-0.01}^{+0.01}$	${2.15}_{-0.06}^{+0.07}$	${0.29}_{-0.04}^{+0.05}$	${66.05}_{-3.53}^{+3.76}$	0.9	${141.60}_{-8.78}^{+12.41}$	${165.25}_{-12.74}^{+13.72}$	${45.52}_{-0.03}^{+0.03}$	$-{0.29}_{-0.01}^{+0.01}$
3C 216	15002	247.9 ± 15.7	0.07 ± 0.02	89.90 ± 5.80	${45.23}_{-0.03}^{+0.03}$	1.9	${0.43}_{-0.15}^{+0.18}$	${39.73}_{-4.29}^{+4.53}$	0.6	${102.39}_{-14.21}^{+9.91}$	${119.58}_{-15.37}^{+11.94}$	${45.36}_{-0.06}^{+0.04}$	$-{0.34}_{-0.06}^{+0.05}$
3C 220.3	14992	5.7 ± 2.4	0.28 ± 0.04	${0.46}_{-0.17}^{+0.22}$	${42.97}_{-0.20}^{+0.17}$	1.9	⋯	<0.45	0.1	⋯	⋯	⋯	$-{0.33}_{-0.46}^{+0.29}$
3C 225B	16058	12.6 ± 3.6	0.40 ± 0.04	${0.84}_{-0.27}^{+0.23}$	${43.07}_{-0.12}^{+0.14}$	1.9	⋯	${2.04}_{-0.93}^{+0.93}$	0.2	${0.98}_{-0.35}^{+0.35}$	⋯	⋯	$-{0.23}_{-0.30}^{+0.23}$
3C 226	15003	58.8 ± 7.7	0.21 ± 0.04	${3.67}_{-0.51}^{+0.57}$	${44.05}_{-0.06}^{+0.06}$	1.9	${16.23}_{-4.78}^{+7.57}$	${7.90}_{-2.42}^{+2.89}$	0.5	${8.99}_{-3.38}^{+3.30}$	${22.40}_{-6.90}^{+9.14}$	${44.84}_{-0.16}^{+0.15}$	${0.56}_{-0.08}^{+0.13}$
3C 228	2095	341.6 ± 18.5	0.45 ± 0.07	${14.20}_{-0.70}^{+0.80}$	${44.23}_{-0.02}^{+0.02}$	1.9	${0.11}_{-0.06}^{+0.07}$	${5.49}_{-0.51}^{+0.53}$	0.5	${15.41}_{-1.57}^{+1.87}$	${16.65}_{-1.76}^{+1.26}$	${44.30}_{-0.05}^{+0.03}$	$-{0.57}_{-0.04}^{+0.04}$
3C 228	2453	251.6 ± 15.9	0.43 ± 0.07	${13.20}_{-0.80}^{+0.80}$	${44.20}_{-0.03}^{+0.03}$	1.9	<0.24	${4.56}_{-0.39}^{+0.53}$	0.6	${13.40}_{-1.82}^{+1.70}$	${13.66}_{-1.86}^{+1.06}$	${44.21}_{-0.06}^{+0.03}$	$-{0.58}_{-0.05}^{+0.05}$
3C 247	16060	42.7 ± 6.6	0.33 ± 0.04	${2.78}_{-0.51}^{+0.44}$	${43.87}_{-0.07}^{+0.07}$	1.9	${7.56}_{-2.87}^{+2.87}$	${23.53}_{-7.42}^{+7.42}$	0.4	${3.99}_{-0.95}^{+0.95}$	${10.48}_{-0.37}^{+0.37}$	${44.44}_{-0.02}^{+0.02}$	${0.43}_{-0.13}^{+0.15}$
3C 254	2209	5087.4 ± 71.3	1.64 ± 0.11	${88.60}_{-1.20}^{+1.20}$	${45.33}_{-0.01}^{+0.01}$	${1.99}_{-0.06}^{+0.06}$	${0.08}_{-0.04}^{+0.04}$	${47.56}_{-2.48}^{+2.60}$	0.6	${128.68}_{-10.20}^{+7.87}$	${132.12}_{-6.77}^{+6.28}$	${45.50}_{-0.02}^{+0.02}$	$-{0.43}_{-0.01}^{+0.01}$
3C 263	2126	9061.1 ± 95.2	2.89 ± 0.18	${88.60}_{-0.90}^{+0.90}$	${45.19}_{-0.00}^{+0.00}$	${1.89}_{-0.03}^{+0.04}$	<0.06	${48.84}_{-0.93}^{+1.62}$	0.9	${145.38}_{-4.89}^{+4.85}$	${145.95}_{-4.42}^{+4.01}$	${45.41}_{-0.01}^{+0.01}$	$-{0.32}_{-0.01}^{+0.01}$
3C 263.1	15004	423.8 ± 20.6	0.21 ± 0.03	30.70 ± 1.50	${44.98}_{-0.02}^{+0.02}$	1.9	${0.21}_{-0.10}^{+0.12}$	${11.84}_{-0.95}^{+0.98}$	0.8	${32.01}_{-2.27}^{+3.87}$	${35.47}_{-3.13}^{+2.84}$	${45.05}_{-0.04}^{+0.03}$	$-{0.39}_{-0.04}^{+0.04}$
3C 265	2984	362.3 ± 19.1	2.68 ± 0.17	2.53 ± 0.15	${43.88}_{-0.03}^{+0.03}$	1.9	${35.68}_{-7.49}^{+11.71}$	${11.35}_{-2.23}^{+3.42}$	0.5	${10.28}_{-2.51}^{+2.57}$	${33.74}_{-9.33}^{+7.04}$	${45.01}_{-0.14}^{+0.08}$	${0.45}_{-0.05}^{+0.04}$
3C 268.1	15005	48.8 ± 7.0	0.22 ± 0.05	${2.44}_{-0.40}^{+0.45}$	${44.06}_{-0.08}^{+0.07}$	1.9	${24.00}_{-6.83}^{+11.21}$	${8.70}_{-2.73}^{+3.52}$	0.2	${9.72}_{-3.83}^{+2.97}$	${27.11}_{-11.64}^{+8.61}$	${45.10}_{-0.24}^{+0.12}$	${0.82}_{-0.06}^{+0.09}$
3C 275.1	2096	4085.1 ± 63.9	0.94 ± 0.07	${95.20}_{-1.50}^{+1.40}$	${45.09}_{-0.01}^{+0.01}$	${1.85}_{-0.05}^{+0.05}$	${0.02}_{-0.01}^{+0.01}$	${242.0}_{-11.4}^{+11.4}$	1.3	${99.79}_{-4.55}^{+4.55}$	${112.5}_{-2.44}^{+2.44}$	${45.16}_{-0.01}^{+0.01}$	$-{0.46}_{-0.01}^{+0.01}$
3C 277.2	16063	10.7 ± 3.3	0.27 ± 0.04	${0.47}_{-0.17}^{+0.14}$	${43.12}_{-0.13}^{+0.15}$	1.9	⋯	${1.45}_{-0.83}^{+0.83}$	0.4	${0.69}_{-0.38}^{+0.38}$	...	...	$-{0.44}_{-0.29}^{+0.21}$
3C 280	2210	116.6 ± 11.0	5.39 ± 0.21	${0.70}_{-0.07}^{+0.07}$	${43.54}_{-0.05}^{+0.04}$	1.9	${21.83}_{-6.99}^{+13.77}$	${1.97}_{-0.54}^{+0.88}$	0.6	${2.43}_{-0.82}^{+0.83}$	${6.10}_{-2.14}^{+1.49}$	${44.48}_{-0.19}^{+0.09}$	${0.10}_{-0.10}^{+0.08}$
3C 286	15006	118.9 ± 10.9	0.10 ± 0.02	${45.90}_{-4.00}^{+4.10}$	${45.19}_{-0.04}^{+0.04}$	1.9	<0.30	${14.31}_{-1.43}^{+2.46}$	0.5	${41.47}_{-7.34}^{+4.83}$	${42.40}_{-6.71}^{+5.27}$	${45.16}_{-0.07}^{+0.05}$	$-{0.61}_{-0.08}^{+0.06}$
3C 289	15007	55.7 ± 7.5	0.32 ± 0.04	${2.38}_{-0.40}^{+0.45}$	${44.04}_{-0.08}^{+0.08}$	1.9	${16.52}_{-5.22}^{+10.83}$	${6.88}_{-2.13}^{+3.15}$	1.0	${7.92}_{-3.24}^{+4.26}$	${21.84}_{-8.63}^{+5.98}$	${45.01}_{-0.22}^{+0.11}$	${0.70}_{-0.08}^{+0.11}$
3C 292	17488	59.6 ± 7.7	0.43 ± 0.05	${9.24}_{-1.36}^{+1.26}$	${44.33}_{-0.06}^{+0.06}$	1.9	${20.03}_{-4.60}^{+6.41}$	${98.63}_{-29.29}^{+25.28}$	0.4	${7.68}_{-2.74}^{+2.74}$	${44.29}_{-0.51}^{+0.51}$	${45.01}_{-0.01}^{+0.01}$	${0.86}_{-0.05}^{+0.08}$
3C 309.1	3105	5306.3 ± 72.8	0.69 ± 0.09	${163.00}_{-2.00}^{+3.00}$	${45.81}_{-0.01}^{+0.01}$	${1.62}_{-0.03}^{+0.05}$	<0.12	${61.91}_{-1.23}^{+2.48}$	0.7	${231.20}_{-8.39}^{+10.08}$	${232.37}_{-11.96}^{+9.02}$	${45.96}_{-0.02}^{+0.02}$	$-{0.49}_{-0.01}^{+0.01}$
3C 330	2127	128.0 ± 11.4	1.02 ± 0.11	${1.61}_{-0.14}^{+0.14}$	${43.28}_{-0.04}^{+0.04}$	1.9	${22.88}_{-10.49}^{+20.29}$	${3.38}_{-1.25}^{+2.33}$	0.5	${4.11}_{-1.48}^{+1.31}$	${11.91}_{-4.15}^{+4.13}$	${44.59}_{-0.19}^{+0.13}$	$-{0.08}_{-0.09}^{+0.08}$
3C 334	2097	7223.5 ± 85.0	8.52 ± 0.31	${133.00}_{-2.00}^{+1.00}$	${45.21}_{-0.01}^{+0.00}$	${1.90}_{-0.02}^{+0.03}$	<0.03	${49.88}_{-0.65}^{+1.10}$	0.7	${149.45}_{-4.78}^{+2.45}$	${148.32}_{-5.13}^{+3.75}$	${45.26}_{-0.02}^{+0.01}$	$-{0.57}_{-0.01}^{+0.01}$
3C 336	15008	193.9 ± 13.9	0.06 ± 0.02	${72.20}_{-5.20}^{+5.30}$	${45.48}_{-0.03}^{+0.03}$	1.9	<0.84	${29.87}_{-3.50}^{+3.69}$	0.1	${82.93}_{-9.17}^{+12.74}$	${89.72}_{-12.74}^{+9.38}$	${45.57}_{-0.07}^{+0.04}$	$-{0.47}_{-0.07}^{+0.06}$
3C 337	15009	9.8 ± 3.2	0.25 ± 0.05	${0.52}_{-0.18}^{+0.23}$	${42.94}_{-0.18}^{+0.16}$	1.9	⋯	${0.12}_{-0.10}^{+0.10}$	0.5	${1.52}_{-1.13}^{+1.03}$	...	⋯	${0.38}_{-0.22}^{+0.34}$
3C 340	15010	87.8 ± 9.4	0.20 ± 0.05	${5.82}_{-0.68}^{+0.69}$	${44.20}_{-0.05}^{+0.05}$	1.9	${6.12}_{-1.58}^{+2.20}$	${7.44}_{-1.63}^{+1.86}$	0.5	${11.15}_{-2.67}^{+2.79}$	${21.58}_{-3.17}^{+4.52}$	${44.77}_{-0.07}^{+0.08}$	${0.31}_{-0.10}^{+0.10}$
3C 343	15011	17.7 ± 4.2	0.26 ± 0.04	${1.48}_{-0.31}^{+0.36}$	${43.86}_{-0.10}^{+0.09}$	1.9	⋯	${0.57}_{-0.18}^{+0.20}$	0.2	${1.60}_{-0.83}^{+0.74}$	⋯	⋯	$-{0.44}_{-0.24}^{+0.17}$
3C 343.1	15012	47.7 ± 6.9	0.28 ± 0.04	${3.46}_{-0.50}^{+0.54}$	${43.94}_{-0.07}^{+0.06}$	1.9	${1.94}_{-1.14}^{+1.77}$	${2.28}_{-0.79}^{+0.92}$	0.4	${4.35}_{-1.38}^{+2.08}$	${6.93}_{-2.83}^{+2.35}$	${44.24}_{-0.23}^{+0.13}$	$-{0.25}_{-0.15}^{+0.12}$
3C 352	15013	135.7 ± 11.7	0.26 ± 0.04	${9.30}_{-0.85}^{+0.80}$	${44.44}_{-0.04}^{+0.04}$	1.9	${3.44}_{-0.74}^{+0.92}$	${8.57}_{-1.34}^{+1.43}$	0.5	${14.92}_{-2.18}^{+2.12}$	${25.04}_{-3.79}^{+4.89}$	${44.87}_{-0.07}^{+0.08}$	${0.07}_{-0.09}^{+0.08}$
3C 380	3124	2642.8 ± 51.4	0.23 ± 0.03	${324.00}_{-6.00}^{+7.00}$	${45.82}_{-0.01}^{+0.01}$	${1.91}_{-0.08}^{+0.08}$	${0.05}_{-{nan}}^{+0.06}$	${148.97}_{-10.45}^{+11.22}$	0.7	${423.44}_{-41.64}^{+41.57}$	${445.74}_{-34.26}^{+32.53}$	${45.96}_{-0.03}^{+0.03}$	$-{0.34}_{-0.02}^{+0.02}$
3C 427.1	2194	31.2 ± 5.7	1.84 ± 0.10	${0.52}_{-0.09}^{+0.11}$	${42.83}_{-0.08}^{+0.08}$	1.9	${26.35}_{10.09}^{+25.91}$	${0.11}_{-0.04}^{+0.05}$	0.1	${0.86}_{-0.11}^{+0.14}$	${0.39}_{-0.19}^{+0.16}$	${43.10}_{-0.30}^{+0.15}$	${0.07}_{-0.19}^{+0.17}$
3C 441	15656	1.9 ± 1.4	0.14 ± 0.04	${0.24}_{-0.14}^{+0.23}$	${42.71}_{-0.40}^{+0.30}$	1.9	⋯	<0.45	0.1	${0.89}_{-0.68}^{+0.40}$	⋯	⋯	${0.64}_{-0.06}^{+0.36}$
3C 455	15014	151.7 ± 12.3	0.26 ± 0.04	13.00 ± 1.00	${44.18}_{-0.03}^{+0.03}$	1.9	<0.81	${4.83}_{-0.71}^{+0.81}$	0.6	${12.81}_{-2.56}^{+1.90}$	${14.71}_{-2.06}^{+1.69}$	${44.23}_{-0.07}^{+0.05}$	$-{0.38}_{-0.08}^{+0.07}$

Notes. Column (1) source name; (2) Chandra ObsID; (3) and (4) Chandra source and background counts; (5) and (6) rest-frame (K-corrected) fluxes (in 10⁻¹⁴ erg cm⁻² s⁻¹) and luminosities (in erg s⁻¹) determined from Srcflux and corrected for Galactic N_H (both quoted with 1σ errors); and (7)–(9) X-ray power-law slope Γ, intrinsic N_H (in units of 10²² cm⁻²), and normalization of the power law at 1 keV (10⁻⁶ photons cm⁻² s⁻¹ keV⁻¹; all with 1σ errors) from Sherpa spectral fitting. Sources with <30 counts were modeled with a power law (Γ = 1.9) and Galactic N_H (Table 1). Spectral fits for sources with 30−700 counts also included intrinsic absorption. A few complex spectra also included a soft excess and/or a 6.4 keV fluorescence Fe Kα line (see Table 3 for details). Fluxes and luminosities quoted in columns (11)–(13) include these features. For sources with >700 counts, Γ was freed after an initial fit with Γ = 1.9. Column (11) rest-frame ("observed") fluxes (in 10⁻¹⁴ erg cm⁻² s⁻¹) are corrected for Galactic N_H only. Column (12) rest-frame "intrinsic" fluxes are corrected for both the Galactic and intrinsic N_H, and column (13) rest-frame luminosities (in erg s⁻¹) are all derived from the best-fit Sherpa model. The fluxes of piled-up sources are corrected for pileup (see Section 4.4). Column (14) hardness ratios are calculated using the BEHR (Park et al. 2006; Section 4.1).

^a The XMM data are from Belsole et al. (2006). Note that the net counts, fluxes, and luminosities are quoted in the 2–10 keV band.

Download table as:  ASCIITypeset images: 1 2

To provide uniformly derived X-ray fluxes, the X-ray data for Chandra-observed sources were initially processed with Srcflux, a program in the Chandra Interactive Analysis of Observations (CIAO; Fruscione et al. 2006) that is particularly useful in calculating the net count rates and fluxes in low-count sources, where spectral fits are poorly constrained. Srcflux performs no spectral fits but instead fits the normalization based on the observed count rate for an assumed source spectrum and source and background regions. This results in fluxes estimated in a consistent manner, particularly for sources with highly absorbed or complex spectra and low-S/N data. We assumed a power-law spectrum with a canonical photon index Γ = 1.9 (Mushotzky et al. 1993; Just et al. 2007) and Galactic absorption characterized by the equivalent hydrogen column density from Dickey & Lockman (1990) and quoted in Table 1. The same source and background regions as described in Section 4.1 were used. The Srcflux fluxes and luminosities (K-corrected assuming a power law with Γ = 1.9) in the 0.5–8 keV range are given in Table 2 (columns (5) and (6)). We will use these values as X-ray fluxes and luminosities throughout the paper for sources with <10 counts.

We performed X-ray spectral modeling of all sources in the sample with Sherpa (Freeman et al. 2001), a modeling and fitting package in CIAO. We used the Levenberg–Marquardt optimization method with the χ² statistic including the Gehrels variance function, which allows for a Poisson distribution for low-count sources. First, a power law with a canonical photon index Γ = 1.9 and Galactic absorption was fit to binned spectra. For sources with ≥30 net counts, a second step including intrinsic absorption (N_H) at the redshift of the source was added to the fit. For sources with ≳700 net counts (mostly quasars), the power-law photon index was then freed in the final spectral fit. The results of the analysis are presented in Table 2. Significantly detected N_H, indicating absorption in excess of the Galactic column density, is most likely absorption intrinsic to the quasar associated with the nucleus and/or the host galaxy. Although unlikely, a contribution from absorption by intervening material/sources along the line of sight cannot be ruled out.

For eight archival sources with more than a few thousand counts resulting in δ > 5% pileup, the CIAO pileup model (jdpileup) was included in the spectral fits. The pileup fraction is reported in Table 3, and the pileup-corrected fluxes are presented in Table 2.

Table 3. X-Ray Parameters of Sources with Complex Spectra and/or Pileup

Name	Chandra	Type a	χ2	Fe Kα	Fe Kα	Fe Kα	Soft Excess	Soft Excess	Pileup
	ObsID			FWHM b	Pos. c	Ampl. d	Ampl. d	Γ	Fraction
3C 006.1	3009	G	0.7	⋯	⋯	⋯	⋯	⋯	0.076
3C 006.1	4363	G	0.7	⋯	⋯	⋯	⋯	⋯	0.065
3C 184	3226	G	0.4	${1.31}_{-0.55}^{+0.67}$	⋯	${1.91}_{-0.86}^{+1.10}$	⋯	⋯	⋯
3C 207	2130	Q	0.9	⋯	⋯	⋯	⋯	⋯	0.350
3C 254	2209	Q	0.7	⋯	⋯	⋯	⋯	⋯	0.296
3C 263	2126	Q	1.2	⋯	⋯	⋯	⋯	⋯	0.403
3C 265	2984	G	0.6	${0.21}_{-0.13}^{+0.13}$	${6.49}_{-0.11}^{+0.14}$	${5.56}_{-3.33}^{+3.33}$	${1.49}_{-0.35}^{+0.37}$	${2.57}_{-0.47}^{+0.51}$	⋯
3C 280	2210	G	0.7	⋯	⋯	⋯	${0.08}_{-0.04}^{+0.04}$	${3.25}_{-0.83}^{+0.93}$	⋯
3C 309.1	3105	Q	0.7	⋯	⋯	⋯	⋯	⋯	0.067
3C 330	2127	G	0.4	${1.21}_{-0.58}^{+2.29}$	⋯	${1.08}_{-0.63}^{+0.63}$	${2.06}_{-0.35}^{+0.44}$	${2.50}_{-0.56}^{+0.35}$	⋯
3C 334	2097	Q	0.8	⋯	⋯	⋯	⋯	⋯	0.085
3C 380	3124	Q	0.7	⋯	⋯	⋯	⋯	⋯	0.230
3C 427.1	2194	G	0.1	<0.005	⋯	${179.8}_{-297.0}^{+297.0}$	${0.62}_{-0.29}^{+0.29}$	${2.40}_{-0.99}^{+0.92}$	⋯

Notes. Sources with complex spectra are fitted in Sherpa with an absorbed power law (see Table 2 for the best-fit parameters), a 6.4 keV fluorescence Fe Kα line, and/or a soft excess. An ellipsis indicates that the relevant fit was not needed for that source. If the best-fit parameter value = 0, a 1σ upper limit is quoted.

^a Source type: Q = quasar, G =NLRG. ^b Fe Kα line FWHM in keV. ^c Fe Kα line position in keV. ^d Fe Kα line and soft excess amplitudes are in units of 10⁻⁶ photons cm⁻² s⁻¹ keV⁻¹.

Download table as:  ASCIITypeset image

Several NLRGs displayed complex X-ray spectra. In particular, 3C 265, 280, and 330 showed excess soft X-ray emission above the absorbed primary power law. This soft excess may be due to thermal emission from a surrounding cluster, emission from the accretion disk or inner region of the jets, intrinsic AGN emission visible due to partial covering of the AGN, or scattered emission from material close to the nucleus. For example, 3C 265, an NLRG, shows an Sy1 spectrum in visible polarized light (Véron-Cetty & Véron 2006), implying scattered intrinsic AGN emission that may extend to the X-rays. Four galaxies, 3C 184, 265, 330, and 427.1, show a strong 6.4 keV fluorescent Fe Kα line arising from the reflection of the hard X-ray power law on the (relatively) cold matter in an accretion disk or torus (Fabian et al. 2000 and references therein). Higher-S/N XMM-Newton data of 3C 184 require a soft excess and N_H $=\,{4.9}_{-1.2}^{+2.2}\times {10}^{23}$ cm⁻² (Belsole et al. 2006). The objects 3C 265 and 330 display both a soft excess and an Fe Kα line. The soft excess and Fe Kα line become pronounced in the heavily obscured sources, when the contribution of the intrinsic power law is significantly reduced.

The fits of complex spectra were built up using an iterative approach. In the initial stage, a model consisting of an absorbed power law was fitted as described in Section 4.3. If an Fe Kα line was visible in the fit residuals, the power law was then fitted over the energy range excluding the line. Next, the fitted parameters were frozen, and an additional component, the soft excess or the Fe Kα line, was added to the model. For two sources that required both the Fe Kα line and the soft excess, the soft excess component was added and fitted first. The soft excess was modeled as an unabsorbed power law with a fixed Γ = 1.9. Then the slope was freed and fitted, after which the primary intrinsic power-law normalization and N_H were freed and fitted. The Fe Kα line was modeled with a Gaussian and fitted iteratively. First, the Fe Kα line amplitude was fitted assuming an approximate peak position at 6.4 keV (rest frame), appropriate for neutral Fe Kα, and an arbitrary FWHM of 0.2 keV. Then the line amplitude and FWHM were freed and fitted simultaneously. For 3C 265, where the iron line is particularly strong, the position of the Fe Kα peak was also fitted. As a next step, the Fe Kα line parameters were frozen, and all other noniron parameters (i.e., intrinsic power-law normalization, N_H, soft excess power-law slope, and normalization) were refitted followed by another Fe Kα line-only fit. The resulting best-fit parameters for the soft excess and Fe Kα line (in the complex spectra) are given in Table 3. For pileup sources, the pileup fraction is also shown in this table. Spectral fits for all complex sources are plotted in Figure 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Here we explore the hierarchical Bayesian modeling (HBM) to constrain individual and whole-sample intrinsic luminosities and column densities and the obscured and CT AGN fractions in the sample. The HBM is a statistical method that facilitates inferences about a population based on individual objects and their observations (and vice versa). Our hierarchical model has three layers. The bottom layer is formed by the observed data (X-ray spectra) and fixed. The middle layer contains the parameters for each object, namely, their intrinsic X-ray luminosity L(0.5–10 keV) and column density N_H. The top layer describes the L(0.5–10 keV) and N_H distributions of the whole population. The HBM simultaneously finds posteriors on individual and population parameters. It "shrinks" individual parameter estimates toward the population mean, which lowers rms errors and naturally deals with large uncertainties and upper limits. The uncertainty is determined via nested sampling. The Appendix presents a detailed explanation of the method.

To apply HBM to our sample, we first used Bayesian inference in analyzing the X-ray spectra assuming flat, uninformative priors for L(0.5–10 keV) and N_H, which were then updated using Bayes' theorem to posterior priors, taking into account the parameter distributions of the whole population. The Bayesian X-ray analysis (BXA) module was used (Buchner et al. 2014) for Sherpa (Fruscione et al. 2006), assuming an AGN with intrinsic obscuration and taking into account Compton scattering and iron fluorescence (BNTORUS model; Brightman & Nandra 2011) with an added warm-mirror power law (same as the scatterd-light component in Section 4.4). All normalizations had wide log-uniform priors, and the intrinsic photon index was assigned a Gaussian prior centered at 1.95 with a standard deviation of 0.15. The warm-mirror normalization can reach up to 10% of the intrinsic AGN power-law component. The above setup is described by, e.g., Buchner et al. (2014). The analysis gives preliminary posterior probability distributions for the parameters in the middle layer, i.e., the individual posterior HBM L(0.5–10 keV) and N_H, which are shown in Figures. The effect of the HBM is that weak observations are informed by well-constrained observations, which indicate probable parameter values. For example, extremely high luminosities are suppressed. The HBM median values of intrinsic L(0.5–10 keV) and N_H for each source are given in Table 4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 4. Source Properties from BXA X-Ray Spectral Fit

Name	log L(2–10 keV) a	fscat b	log NH c	P(NH < 1022)	P(1022 < NH < 1024)	P(NH > 1024)
	(erg s−1)		(cm−2)
3C 006.1*	${45.08}_{-0.02}^{+0.03}$	${7.17}_{-3.44}^{+1.74} \%$	${21.99}_{-0.10}^{+0.08}$	56%	44%	0%
3C 022	${45.32}_{-0.17}^{+0.25}$	${1.24}_{-0.71}^{+1.01} \%$	${23.55}_{-0.17}^{+0.20}$	0%	93%	6%
3C 034	${44.78}_{-0.11}^{+0.13}$	${0.20}_{-0.17}^{+0.65} \%$	${23.11}_{-0.10}^{+0.09}$	0%	100%	0%
3C 041	${44.98}_{-0.18}^{+0.21}$	${0.32}_{-0.26}^{+0.55} \%$	${23.61}_{-0.16}^{+0.17}$	0%	95%	4%
3C 049	${44.93}_{-0.08}^{+0.07}$	${0.07}_{-0.05}^{+0.30} \%$	22.79 ± 0.06	0%	100%	0%
3C 055	${45.34}_{-0.36}^{+0.22}$	${0.63}_{-0.36}^{+1.00} \%$	${25.16}_{-0.65}^{+0.56}$	1%	1%	97%
3C 138	${45.66}_{-0.03}^{+0.04}$	${1.27}_{-1.21}^{+5.14} \%$	${21.84}_{-0.13}^{+0.12}$	89%	10%	0%
3C 147	${45.03}_{-0.06}^{+0.08}$	${2.54}_{-2.45}^{+4.56} \%$	21.96 ± 0.17	59%	41%	0%
3C 172	${45.27}_{-0.26}^{+0.23}$	${0.05}_{-0.03}^{+0.14} \%$	24.01 ± 0.13	0%	47%	52%
3C 175	45.58 ± 0.03	${0.46}_{-0.43}^{+4.02} \%$	${21.39}_{-0.18}^{+0.15}$	99%	0%	0%
3C 175.1	44.63 ± 0.07	${0.25}_{-0.22}^{+2.12} \%$	22.19 ± 0.14	9%	91%	0%
3C 184	${44.99}_{-0.27}^{+1.10}$	${2.02}_{-1.92}^{+2.27} \%$	${23.74}_{-0.23}^{+1.71}$	0%	62%	37%
3C 196	45.41 ± 0.08	${0.13}_{-0.10}^{+1.30} \%$	22.46 ± 0.11	0%	100%	0%
3C 207*	45.34 ± 0.01	${0.18}_{-0.15}^{+0.79} \%$	21.25 ± 0.02	100%	0%	0%
3C 216	${45.29}_{-0.05}^{+0.04}$	${0.51}_{-0.48}^{+3.72} \%$	${21.28}_{-0.40}^{+0.21}$	98%	0%	1%
3C 220.3	${43.04}_{-0.27}^{+1.32}$	${1.06}_{-1.00}^{+4.03} \%$	${21.17}_{-0.80}^{+3.85}$	64%	2%	32%
3C 225B	${43.33}_{-0.29}^{+1.41}$	${1.42}_{-1.34}^{+3.99} \%$	${21.83}_{-1.23}^{+3.49}$	50%	2%	46%
3C 226	44.93 ± 0.13	${1.36}_{-0.92}^{+1.24} \%$	23.24 ± 0.13	0%	100%	0%
3C 228	44.24 ± 0.03	${0.42}_{-0.38}^{+3.17} \%$	${20.53}_{-0.34}^{+0.30}$	99%	0%	0%
3C 247	44.45 ± 0.14	${2.05}_{-1.76}^{+3.20} \%$	${22.94}_{-0.18}^{+0.15}$	0%	100%	0%
3C 254*	45.37 ± 0.01	${0.37}_{-0.32}^{+2.65} \%$	${20.04}_{-0.03}^{+0.06}$	100%	0%	0%
3C 263*	45.86 ± 0.01	${9.97}_{-0.04}^{+0.03} \%$	22.80 ± 0.01	0%	100%	0%
3C 263.1	${44.99}_{-0.02}^{+0.04}$	${0.32}_{-0.29}^{+2.69} \%$	${20.85}_{-0.48}^{+0.38}$	97%	0%	2%
3C 265	${45.03}_{-0.08}^{+0.10}$	${2.59}_{-0.65}^{+0.74} \%$	${23.52}_{-0.05}^{+0.07}$	0%	100%	0%
3C 268.1	${45.13}_{-0.13}^{+0.12}$	${0.09}_{-0.07}^{+0.39} \%$	23.42 ± 0.11	0%	100%	0%
3C 275.1	45.09 ± 0.01	${0.24}_{-0.21}^{+2.09} \%$	${20.55}_{-0.14}^{+0.12}$	100%	0%	0%
3C 277.2	${44.57}_{-1.27}^{+0.53}$	${1.77}_{-1.38}^{+4.01} \%$	${24.61}_{-3.72}^{+0.97}$	29%	3%	67%
3C 280	${45.85}_{-1.11}^{+0.15}$	${0.22}_{-0.09}^{+2.91} \%$	${25.10}_{-1.46}^{+0.64}$	0%	19%	80%
3C 286	${45.21}_{-0.05}^{+1.06}$	${1.08}_{-1.03}^{+5.87} \%$	${20.47}_{-0.34}^{+4.46}$	75%	0%	24%
3C 289	${45.07}_{-0.11}^{+0.16}$	${0.09}_{-0.07}^{+0.33} \%$	23.29 ± 0.11	0%	99%	0%
3C 292	${45.07}_{-0.11}^{+0.12}$	${0.15}_{-0.13}^{+0.38} \%$	23.33 ± 0.09	0%	100%	0%
3C 309.1*	45.82 ± 0.01	${0.22}_{-0.20}^{+1.95} \%$	${20.14}_{-0.11}^{+0.17}$	100%	0%	0%
3C 330	${44.29}_{-0.13}^{+0.20}$	5.31 ± 2.09%	${23.48}_{-0.15}^{+0.22}$	0%	91%	9%
3C 334*	45.19 ± 0.01	${0.20}_{-0.18}^{+1.60} \%$	${20.19}_{-0.12}^{+0.18}$	100%	0%	0%
3C 336	45.51 ± 0.05	${0.31}_{-0.27}^{+3.06} \%$	${21.48}_{-0.35}^{+0.20}$	99%	0%	0%
3C 337	${43.73}_{-0.25}^{+0.47}$	${0.21}_{-0.18}^{+1.88} \%$	${22.88}_{-0.36}^{+0.56}$	2%	84%	13%
3C 340	${44.88}_{-0.11}^{+0.09}$	${1.93}_{-1.12}^{+1.58} \%$	22.94 ± 0.10	0%	100%	0%
3C 343	${43.85}_{-0.14}^{+0.22}$	${0.57}_{-0.54}^{+4.08} \%$	${21.43}_{-0.87}^{+0.78}$	80%	4%	14%
3C 343.1	44.23 ± 0.11	${0.57}_{-0.53}^{+3.50} \%$	22.25 ± 0.16	7%	92%	0%
3C 352	44.89 ± 0.08	${0.32}_{-0.29}^{+1.83} \%$	22.55 ± 0.08	0%	100%	0%
3C 380*	45.82 ± 0.01	${0.19}_{-0.17}^{+1.72} \%$	${20.33}_{-0.21}^{+0.29}$	100%	0%	0%
3C 427.1	${44.42}_{-0.80}^{+0.70}$	${1.09}_{-0.94}^{+5.15} \%$	${24.05}_{-0.73}^{+1.42}$	0%	49%	50%
3C 441	${43.22}_{-0.76}^{+1.06}$	${0.23}_{-0.21}^{+2.43} \%$	${22.98}_{-1.14}^{+1.76}$	20%	54%	24%
3C 455	${44.16}_{-0.04}^{+0.05}$	${0.47}_{-0.43}^{+4.08} \%$	${20.98}_{-0.54}^{+0.38}$	98%	0%	1%

Notes. An asterisk indicates a piled-up source.

^a Intrinsic 2−10 keV luminosity corrected for absorption. ^b Strength of the scattered power law relative to the intrinsic continuum power law (3σ range). ^c Intrinsic column density. The last three columns show the HBM probability that the intrinsic column density is unobscured (N_H < 10²² cm⁻²), obscured Compton-thin (N_H = 10²²–10²⁴ cm⁻²), or obscured CT (N_H > 10²⁴ cm⁻²).

Download table as:  ASCIITypeset image

The quasars and NLRGs in the medium-z 3CRR sample have comparable (to within ∼1.5 dex) extended 178 MHz radio luminosities (Section 2; Figure 1, left), which implies similar intrinsic AGN luminosities. In contrast, the 2–8 keV luminosities, uncorrected for intrinsic absorption, hardly overlap (Figure 1, right), where the NLRGs show 10–1000 times lower hard X-ray luminosities than quasars, suggesting higher obscuration in NLRGs. The widely different apparent luminosities are consistent with the unification model, where the nuclei of NLRGs are thought to be viewed edge-on through a dusty, torus-like structure and thus observed through higher amounts of obscuration than the quasars.

The X-ray HR, defined as $\mathrm{HR}\equiv \tfrac{(H-S)}{(H+S)}$, where H and S are the 2–8 and 0.5−2 keV counts, respectively, is often used as a measure of intrinsic N_H and particularly useful in lower-count sources, where spectral fitting is not possible. A higher (harder) HR indicates higher obscuration, and a lower (softer) HR indicates lower obscuration. A few sources in our sample have low counts, so we determined the HRs using the Bayesian estimation of HRs (BEHR) method (Park et al. 2006), which accounts for the Poissonian nature of the data and correctly deals with non-Gaussian error propagation, appropriate for both the low- and high-count regimes. These HRs are provided in Table 2 (column (14)), and their distribution is presented in Figure 6. All quasars (plotted in blue) have soft HR < 0, with the mean HR = −0.36 ± 0.15 consistent with an AGN power law with ${\rm{\Gamma }}={1.5}_{-0.33}^{+0.32}$ and low obscuration. The quasar with the hardest HR (=−0.07) in the sample, 3C 196, has intermediate obscuration of N_H = 3 × 10²² cm⁻² and is classified as a type 1.8 based on its optical spectrum. In contrast, the NLRGs (plotted in red) span a wide range of HRs, −0.6 < HR < 0.9, implying a large range of intrinsic obscuration.

Zoom In Zoom Out Reset image size

Figure 7 shows the dependence of the observed HR on N_H compared to trends expected from modeling. The intrinsic N_H was obtained from X-ray spectral fitting (Sections 4.3 and 4.4) for sources with at least 30 counts. Most of the sources lie on the track of the pure absorbed power-law models with photon index 1.5 < Γ < 2.2 and 10²⁰ < N_H/cm⁻² < 10²⁵. The exceptions are 3C 172, 184, 265, 280, 330, and 427.1, for which the HRs are softer than predicted from an absorbed power law with the measured N_H. Apart from 3C 172, for which a low S/N (32 counts) does not allow for a complex fit, these are the sources with complex spectra discussed in Section 4.4. These sources' spectra include an additional soft excess component (besides the heavily obscured power law and the Fe Kα line) that is possibly due to scattered nuclear light or extended X-ray emission from gas surrounding the nucleus, galaxy cluster, or radio/X-ray jet.

Zoom In Zoom Out Reset image size

The Chandra data of 3C 184 had too few counts (∼48) to justify a complex fit, but the higher-S/N XMM-Newton data require a soft excess, a high column density (N_H $=\,{4.9}_{1.2}^{+2.2}\times {10}^{23}$ cm⁻²), and an Fe Kα line (Belsole et al. 2006).

The observed (uncorrected for N_H) broadband 0.5–8 keV X-ray luminosities are plotted against HRs in Figure 8. These are compared with a pure absorbed power-law model (Γ = 1.9; red dotted curve) and other absorbed power-law models (Γ = 1.5, 2.2) with an added soft excess component of varying strength (0.1%, 1%, and 5% of intrinsic light; blue and green curves). The quasars have high observed L_X and soft HRs, indicating low obscuration. The NLRGs show a broad range of HRs and lower observed L_X, indicating a varying degree of intrinsic N_H and amount of scattered/extended light emission. The majority of medium-z NLRGs lie on models that include an absorbed power law and a soft excess of varying strength, which makes their HRs softer than the ones expected from a pure absorbed power-law model. Figure 9 is a modified version of Figure 8 where the observed 0.5–8 keV X-ray luminosity is normalized to the total radio luminosity at 178 MHz (a surrogate for intrinsic AGN luminosity). Quasars show L(0.5−8 keV)/L(178 MHz) > 1 and soft HRs; NLRGs have L(0.5−8 keV)/L(178 MHz) < 1 and a range of HRs. A group of five soft NLRGs (3C 6.1, 175.1, 228, 263.1, and 455) has almost quasar-like L(0.5−8 keV)/L(178 MHz) ∼ 1, indicating low obscuration. These will be discussed further in Sections 6.2 and 6.3.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The mean quasar HR of the medium-z 3CRR sample (−0.36 ± 0.15) is comparable to that of the high-z 3CRR sample (−0.44 ± 0.20). However, the median is harder (−0.34 versus −0.51), implying flatter primary power-law slopes (Γ = 1.5 versus 1.9) and/or higher N_H in the medium-z quasars, which may reflect the fact that low N_H is easier to measure at lower redshifts as the softer X-rays move into the Chandra-observed band. Piled-up quasars, present at medium-z, will also contribute to the harder mean and median HRs. For NLRGs, the mean HR (0.14 ± 0.43) is comparable, within the uncertainties, to the high-z NLRG mean (0.10 ± 0.45), while the median is softer (0.10 versus 0.26), implying a higher fraction of NLRGs with low N_H in the medium-z sample (discussed in Section 6.2).

The median 2–8 keV luminosity, uncorrected for intrinsic column density, is 6× lower for NLRGs than quasars in the medium-z sample (10^44.4 versus 10^45.2 erg s⁻¹, respectively), while it was ∼100× lower in the high-z 3CRR sample (Wilkes et al. 2013), suggesting a higher number of NLRGs with low obscuration in the medium-z sample.

The ratio of the observed broadband 0.5–8 keV X-ray luminosity (uncorrected for N_H) to the total radio luminosity at 178 MHz (L_X/L_R, where L_R = ν L_ν (178 MHz) is calculated from the 178 MHz flux densities in Laing et al. 1983), which is a measure of gas obscuration, is plotted in Figure 10(a) as a function of the radio core fraction R_CD (an orientation indicator). Sources with lower obscuration have higher L_X/L_R ratios and show larger values of R_CD, i.e., are preferentially seen at lower viewing angles in respect to the radio jet (i.e., face-on to the torus). Sources with higher obscuration (lower L_X/L_R) have lower R_CD and thus are preferentially viewed perpendicular to the radio jet (i.e., edge-on to the torus). To show this explicitly, the intrinsic column density N_H (estimated from X-ray spectral fits in Sections 4.3 and 4.4) is plotted as a function of the radio core fraction R_CD in Figure 10(b). The strong relation between N_H (and L_X/L_R) and R_CD implies that obscuration is strongly dependent on orientation and increases with increasing viewing angle. This relation is consistent with the orientation-dependent obscuration invoked by the unification model and agrees with our results for the high-z 3CRR sample (Wilkes et al. 2013). However, at intermediate viewing angles, $-3\lt \mathrm{log}$ R_CD < −2, NLRGs with a broad range of N_H exist. These include typical obscured NLRGs with N_H > 10²² cm⁻² and a peculiar class of NLRGs, not present in the high-z 3CRR sample, with low intrinsic column densities, N_H ≲ 10²² cm⁻². These low-N_H NLRGs cannot be explained by a simple unification model dependent solely on orientation and suggest that a second parameter (clumpy torus, different obscurer, or different L/L_Edd ratio) is needed. We will focus on the low-N_H NLRGs next.

Zoom In Zoom Out Reset image size

One-fourth of the NLRGs (3C 6.1, 175.1, 228, 263.1, and 455), or 14% of the medium-z 3CRR sample, have low N_H (10²¹–10²² cm⁻²), similar to the unobscured BLRGs and quasars. As a result of low obscuration, these NLRGs have soft, quasar-like HRs (HR < 0) and the highest L_X/L_R among the NLRGs (Figure 10(a)). These low-N_H NLRGs have intermediate core fractions (−2.7 < log R_CD < −2) and thus are likely viewed at angles skimming the edge of the accretion disk or torus. No such sources were present in the high-redshift 3CRR sample, where all NLRGs had higher intrinsic column densities of log N_H/cm⁻² > 22.7 and log R_CD < −2. Although it is easier to measure low N_H values in sources at medium-z than at high-z (as the softer-energy X-rays move into the Chandra-observed band), the spectra also become more complex, often including an additional soft excess component. The low-N_H NLRGs have enough counts (90–1700) to model the soft excess, but none of them required one. We hence conclude that the low intrinsic column densities in these NLRGs are measured correctly and are not underestimated due to the lack of soft excess modeling in low-S/N spectra.

The low-N_H NLRGs show relatively low mid-IR (30 μm) emission when compared to their radio emission. The L(30 μm)/L(178 MHz) ratios are the lowest in the sample (Figure 11(a)), ∼10 times lower than in quasars. Because the X-ray emission is also weaker by a factor of 10 relative to radio emission (see Figure 10(a)), the L(30 μm)/L(2–8 keV) ratios are comparable to those of quasars (see Figure 11(b)). The SEDs of low-N_H NLRGs show no IR or big blue bump (see Figures 4, 5, and 7 of Westhues et al. 2016), and the specific star formation rates are close to those of normal galaxies (Westhues et al. 2016).

Zoom In Zoom Out Reset image size

Three of the low-N_H NLRGs were observed with the Hubble Space Telescope (3C 6.1, 228, and 263.1). The optical images show compact host galaxies with no visible dust lanes (McCarthy et al. 1997). The optical SDSS spectra are red (3C 175.1, 228, 263.1, and 455). The object 3C 6.1 shows a weak optical continuum dominated by the host galaxy (visible 4000 Å absorption feature) with an 8 Gyr old stellar population (Smith et al. 1979). The object 3C 455 has conflicting optical types (type 1 or 2) in the literature, but we classify this source as a type 2 based on the spectrum presented by Gelderman & Whittle (1994), which shows a weak continuum and no broad Hβ emission line (however, since ${\rm{H}}\alpha$ was not observed an intermediated type, ∼1.8 or 1.9 cannot be excluded).

Possible scenarios that can explain low column densities, a lack of broad emission lines, and weak IR emission in the low-N_H NLRGs are the following.

• 1.  
  These are "true" type 2 objects (Panessa & Bassani 2002; Tran 2003; Shi et al. 2010; Merloni et al. 2014), which show no detectable broad lines and have low X-ray absorption. In such sources, the BLR has faded due to recent weakening of the continuum or has not formed due to very low L/L_Edd ≪ 10⁻² (Nicastro 2000). In the latter scenario, such low L/L_Edd ratios would result in more than 100–1000× weaker 0.5–8 keV luminosities (as accretion disk SEDs strongly depend on L/L_Edd; see, e.g., Czerny et al. 1996, Figure 1), but the values of L_X/L_R only a few to 10× lower than in quasars (Figure 10(a)) rule out this scenario.
• 2.  
  The obscuration is nonstandard, caused not by a torus but by a dust lane or host galaxy disk misaligned with the dusty torus (as in the red 2MASS AGN; Kuraszkiewicz et al. 2009a, 2009b), which would result in N_H ≤ 10²² cm⁻². Such a low column density cannot significantly obscure the intrinsic X-ray emission or the IR emission from the dusty torus but is sufficient to hide the AGN's optical+UV continuum and the BLR. In this scenario, the weak IR emission in low-N_H NLRGs cannot be easily explained unless the dusty torus is absent.
• 3.  
  The L/L_Edd ratio is low. The low-N_H NLRGs are found at intermediate viewing angles (−3 < log R_CD < −2), together with NLRGs that have higher column densities of 10^22.5 < N_H/cm⁻² < 10^23.5 (Figure 10(b)). Therefore, a scenario is needed in which clouds with a large range of column densities may exist at such viewing angles. Fabian et al. (2008) showed that the distribution of column densities of the gas and dust clouds surrounding an AGN is a function of L/L_Edd, and only clouds with N_H/cm⁻² ≥ 5 × 10²³ × L/L_Edd can withstand the AGN's radiation pressure, while the lower-N_H clouds are blown away. At low L/L_Edd ∼ 0.01, clouds with column densities ranging from ∼10²² cm⁻² to CT can exist, whereas at high L/L_Edd ∼ 1, only those with CT column densities will survive. Applying the scenario to our sample inplies that the NLRGs with low N_H must have low L/L_Edd, while the NLRGs with high N_H (viewed at similar intermediate angles) have high L/L_Edd. The scenario is further confirmed by the finding that the X-ray luminosities, uncorrected for intrinsic absorption, are comparable for the low- and high-N_H NLRGs having the same intermediate viewing angles (0.2 < L(0.5−8 keV)/L(178 MHz) < 1; Figure 10(a)), despite significantly different column densities. Thus, we conclude that the low-N_H NLRGs indeed have lower intrinsic X-ray luminosities and hence lower L/L_Edd than the high-N_H NLRGs.
• 4.  
  The IR emission is weak due to low L/L_Edd. At high L/L_Edd (strong big blue bump), only a torus that is compact and CT can withstand the intense UV radiation and strong winds. Dust in such a compact geometry will strongly radiate in the near-to-mid-IR, producing an SED with a strong IR bump (Pier & Krolik 1992). At lower L/L_Edd, where the big blue bump is weaker and provides less illuminating flux for the torus, the torus may become clumpy and extended, resulting in a weaker IR bump (Kuraszkiewicz et al. 2003; Nenkova et al. 2008a, 2008b; Hönig et al. 2010; Siebenmorgen et al. 2015). Figure 12 shows the dependence of the 30 μm luminosity on the 2–8 keV intrinsic luminosity (estimated from the HBM model), which is related to the L/L_Edd ratio (e.g., Czerny et al. 1996, Figure 1). Both luminosities are normalized by the extended radio luminosity L(178 MHz) to remove any redshift dependence on the IR and X-ray luminosities. There is a strong correlation between L(30 μm)/L(178 MHz) and L(2–8 keV)/L(178 MHz) with a 0.01% probability of occurring by chance in both the generalized Kendall rank and Spearman rank tests. The correlation indicates that higher L/L_Edd sources (=higher intrinsic L(2–8 keV)) have stronger mid-IR luminosities. The low-N_H NLRGs have relatively low L/L_Edd (i.e., L(2–8 keV)/L(178 MHz) < 1), so their weak mid-IR emission can be explained as due to low L/L_Edd.Two sources, 3C 220.3 and 343, do not lie on the overall correlation in Figure 12. They have relatively low L/L_Edd but show strong mid-IR emission. The object 3C 220.3 is lensing a background submillimeter galaxy (Haas et al. 2014), which results in amplification of its IR luminosity. We suggest that perhaps 3C 343 may also be lensing a background galaxy. Another outlier is 3C 172, with high L/L_Edd and low mid-IR emission. The low-IR emission can be explained by either extreme CT obscuration of N_H > 10²⁵ cm⁻² or low amounts of dust due to a 1000× lower than Galactic dust-to-gas ratio. The former explanation is not supported by our low-S/N X-ray spectral modeling, which gives N_H ∼ 10²⁴ cm⁻². The latter is in conflict with typical AGN dust-to-gas ratios, which are generally 1–100 times lower than Galactic (Maiolino et al. 2001; Marchese et al. 2012; Burtscher et al. 2016) with a few AGN having this ratio a factor of a few times higher (Trippe et al. 2010; Ordovás-Pascual et al. 2017).

Zoom In Zoom Out Reset image size

In summary, a simple unification model where obscuration changes only with orientation cannot fully describe the observed multiwavelength properties of the medium-z 3CRR sample, and a range of L/L_Edd ratios extending to low values is required to explain the existence and properties of the low-N_H NLRGs. In contrast, the multiwavelength properties of the high-z 3CRR sample were explained by pure unification, suggesting that L/L_Edd had a narrower range and possibly higher values in comparison with the medium-z sample, allowing orientation effects to dominate the observed properties of the sample.

6.4.1. Compton-thick Candidates

The luminosity of the [O iii] λ5007 emission line (hereafter L([O iii]) was found to track the radio and intrinsic X-ray luminosities for both the type 1 and type 2 AGN (Mulchaey et al. 1994; Jackson & Rawlings 1997). It is often used as an indicator of intrinsic AGN luminosity (Risaliti et al. 1999; Panessa et al. 2006) and has little or no inclination dependence at high luminosities (Jackson & Rawlings 1997; Grimes et al. 2004). The observed hard X-ray luminosity, on the other hand, is strongly dependent on obscuration (especially at high N_H), so the ratio of L([O iii])/L(2–8 keV) is often used to discriminate between Compton-thin and CT sources (Risaliti et al. 1999; Panessa et al. 2006). Figure 13 shows the ratio L([O iii])/L(2–8 keV) plotted against the radio core fraction R_CD. The L([O iii]) values are from Grimes et al. (2004) and shown in Table 1. Seventeen sources have actual [O iii] measurements, and for the remainder, L([O iii]) was estimated from either the [O ii] λ3727 emission line or the 151 MHz radio luminosity (3C 292 and 427.1). The dotted line in Figure 13 shows the dividing line between Compton-thin and CT sources reported by Juneau et al. (2011), and seven sources, 3C 184, 220.3, 225B, 277.2, 280, and 441 (all NLRGs, with L([O iii]) estimated from L([O ii])) and 3C 427.1 (an LERG), appear to be CT. The HBM analysis (Section 4.5, Table 4, Figure 3) gives CT probabilities ranging from 24% to 80%. Objects 3C 220.3, 225B, 277.2, and 441 have too few counts (<15) to model the X-ray spectrum to confirm the high N_H, but HBM implies CT obscuration (see Figure 3, Table 4). The low-S/N Chandra spectrum of 3C 184 (48 counts) shows a strong Fe Kα line (Figure 2), implying heavy obscuration (the reflection component becomes stronger as the intrinsic power law weakens with increasing obscuration), while the higher-S/N XMM data are fitted with high N_H, a strong Kα line, and a soft excess (Belsole et al. 2006). The Chandra X-ray spectrum of 3C 280 (117 counts) is modeled with a strong soft excess and intermediate N_H (Figure 2, Table 2).

Zoom In Zoom Out Reset image size

Five of the above CT candidates have measured L(30 μm) (Westhues et al. 2016), and all except 3C 427.1 have log L(30 μm)/L(2–8 keV) > 1.8 (Figure 11(b)). The Spitzer/IRS spectra of 3C 184 and 441 show strong 9.7 μm silicate absorption (an indicator of large amounts of dust) with τ_9.7 > 0.3. Despite being a CT candidate, 3C 280 has no 9.7 μm silicate absorption (Georgantopoulos et al. 2011). All of the above CT candidates, except for 3C 427.1, have log R_CD < −3, indicating inclination angles larger than 80° (i.e., orientation edge-on to the torus).

For 3C 427.1, neither the [O iii] nor the [O ii] luminosity was measured directly, and L(151 MHz) was used to estimate L([O iii]). To confirm this source's CT nature, we consider other CT indicators. The object 3C 427.1 has the lowest L(0.5−8 keV)/L(178 MHz) in the sample (Figure 10(a)), suggesting low observed L_X, which may be due to either CT obscuration, low L/L_Edd, or X-rays being recently turned off (the source is an LERG that harbors a low-luminosity AGN). The L(30 μm)/L(178 MHz) ratio is the lowest in the sample (Figure 11(a)). Low mid-IR emission is typical for LERGs (Westhues et al. 2016), where low L/L_Edd results in weaker big blue bump emission, which provide less illuminating flux for the circumnuclear dust emitting in the IR (see Figure 12 showing a relation between L/L_Edd and L(30 μm)/L(178 MHz)). Alternatively, the mid-IR emission could be suppressed by heavy obscuration, N_H ≳ 10²⁵ cm⁻², resulting in a strong 9.7 μm silicate absorption that cannot be checked in this source for lack of a Spitzer/IRS spectrum. However, the presence of a strong Fe Kα line (Figure 2) implies that 3C 427.1 is indeed heavily obscured.

6.4.2. Compton-thick and Borderline Compton-thick Candidates with Low [O iii] Emission

Table 5 summarizes the various CT indicators for each of the CT candidates discussed above and shows that these indicators do not always agree. We analyze the reasons and give recommendations for their use.

The distribution of the most widely used CT indicator L([O iii])/L(2–8 keV), where L(2–8 keV) is X-ray luminosity not corrected for intrinsic N_H, is plotted in Figure 14(a). Most (7/9 = 78%) of the CT candidates in our medium-z 3CRR sample lie at log L([O iii])/L(2–8 keV) ≥ −0.25, the dividing line between the Compton-thin and CT sources from Juneau et al. (2011). Exceptions are 3C 55 and 172, which, together with the three borderline CT sources (3C 330, 337, and 343) show Compton-thin L([O iii])/L(2–8 keV). Interestingly, the sources that make the CT cut cover a full range of the sample's intrinsic L_X (log L_X = 43–46; see Table 4), which means that they also cover the full range of L/L_Edd in the sample, suggesting that L([O iii])/L(2–8 keV) is independent of L/L_Edd.

Zoom In Zoom Out Reset image size

The mid-IR (30 μm) luminosity, similarly to L([O iii]), is used as a measure of intrinsic AGN luminosity; hence, L(30 μm)/L(2–8 keV) can also be used as an indicator of CT obscuration. We plot this ratio as a function of N_H in Figure 15 and find that most of the sources with N_H > 10²³ cm⁻² have log L(30 μm)/L(2–8 keV) > 1. The distribution of L(30 μm)/L(2–8 keV) in Figure 14(b) shows that a value >1.8 finds most CT sources in the sample: five out of seven (71%) CT candidates with measured L(30 μm) and one borderline CT source (3c 343). Relaxing this criterion to >1.2 finds six out of those seven CT sources (86%) but also picks three highly obscured, non-CT NLRGs. The object 3C 172 is the only CT source with log L(30 μm)/L(2–8 keV) < 1 due to the unusually weak mid-IR emission (see Section 6.4.2). The L(30 μm)/L(2–8 keV) > 1.8 is therefore a robust CT indicator, but it does not find CT sources exclusively. The ratio may be enhanced by emission from lensed background galaxies (as in 3C 220.3; Haas et al. 2014). The fact that 3C 427.1, a low-L/L_Edd source, does not make the cut suggests that the Eddington ratio also plays a role.

Zoom In Zoom Out Reset image size

The L(0.5−8 keV)/L(178 MHz) or L(2–8 keV)/L(178 MHz) ratios may also be used to indicate heavy obscuration, where this time the total radio luminosity at 178 MHz is a measure of intrinsic AGN luminosity. The log L(0.5−8 keV)/L(178 MHz) < 0 finds all of the CT and heavily obscured (borderline CT) sources but also includes one Compton-thin NLRG (Figure 10(a)). Highly obscured, non-CT sources will make this cut if their L/L_Edd is low (resulting in low L_X).

Seven out of nine (78%) CT sources have low radio core fractions, log R_CD ≤ −3, i.e., are highly inclined with viewing angles θ > 80°. This low R_CD value may be used to find CT sources; however, other heavily absorbed sources with N_H > few × 10²³ cm⁻² (Figure 10(b)) also show a similarly low R_CD.

Out of the nine CT candidates, four have IRS/Spitzer spectra, where three show strong 9.7 μm silicate absorption (optical depth τ_9.7 > 0.3), while one (3C 280), despite being mid-IR bright, does not. Strong silicate absorption is a good indicator of heavy dust obscuration, but a lack thereof does not rule out that the source is heavily obscured by gas. For example, the nearby canonical CT galaxy NGC 1068 lacks 9.7 μm silicate absorption, and only half of the nearby (z < 0.05) CT AGN show τ_9.7 > 0.5 (Goulding et al. 2012). The strength of the 9.7 μm silicate absorption is also affected by dust lying farther out in the galaxy or in a galaxy merger environment, where the AGN residing in mergers or postmergers show the strongest silicate absorption (Goulding et al. 2012).

We summarize as follows.

• 1.  
  Of the CT indicators studied here, log L([O iii])/L(2–8 keV) ≥ −0.25 is the most robust. It is available for both the radio-quiet and radio-loud sources, finds exclusively (78%) CT sources, and does not depend on L/L_Edd.
• 2.  
  The log L(30 μm)/L(2–8 keV) > 1.8 identifies 71% of CT sources in the sample but possibly only the ones with high L/L_Edd ratios. Lowering this criterion to >1.2 finds more CT sources (86%), regardless of their L/L_Edd ratio. However, either criterion includes heavily obscured sources that are not CT. This CT indicator is affected by L/L_Edd and any gravitational lensing.
• 3.  
  The log L(0.5–8 keV)/L(178 MHz) < 0 is an indicator of heavy (both CT and borderline CT) obscuration available for radio-loud sources. It is affected by L/L_Edd.
• 4.  
  A low radio core fraction log R_CD ≤ −3 finds 78% of the CT sources in our sample, together with the highly obscured but non-CT objects. It is a good indicator of high obscuration, both CT and -thin, but only available for sources in which R_CD can be measured.
• 5.  
  Strong 9.7 μm silicate absorption (τ_9.7 > 0.5) is an indicator of heavy dust absorption, including by dust lying at larger host galaxy scales and related to mergers. However, sources in which CT obscuration originates from dustless circumnuclear gas will not have strong silicate absorption (as 3C 280).

Out of all of the CT indicators studied above, log L([O iii])/L(2–8 keV) ≥ −0.25 is the most reliable CT indicator that finds exclusively CT sources, does not depend on the L/L_Edd ratio, and is available for both the radio-quiet and radio-loud sources. All other indicators pick up a small fraction of highly obscured but non-CT sources and depend on L/L_Edd, lensing, or the location of the obscurer. None of the indicators find all of the CT sources in the sample, so we recommend examining all that are available.

The strong dependence of L_X/L_R (where L_X is uncorrected for N_H) and N_H on R_CD (Figure 10, Section 6.1) implies that obscuration in the medium-z 3CRR sample is orientation-dependent, increases with viewing angle, and, to first order, is consistent with the standard unification model. However, at intermediate viewing angles, sources with a large range of N_H between 10^21.3 and 10^23.5 cm⁻² are present, suggesting that another parameter independent of orientation (possibly L/L_Edd) contributes to the spread in N_H.

The number of sources as a function of N_H can provide constraints on the covering factor of the obscuring material. If we assume that the 3CRR sources have a geometry in which the obscuring material lies in the plane perpendicular to the radio jet, and the sources lie randomly oriented on the sky, the probability of finding a source lying in a cone of angle ϕ is $P(\theta \lt \phi )=1-\cos \phi$ (Barthel 1989). Because 14 out of the 44 (32%) sources in the sample are quasars with N_H < 10^21.5 cm⁻², this gives an estimate of the half-opening angle of the obscuring material (torus) of 47° ± 3°. For comparison, 60° ± 8° was found in the high-z sample.

Nine NLRGs are CT candidates characterized by the following CT indicators: L([O iii])/L(2–8 keV) ≥ −0.25, L(30 μm)/L(2–8 keV) > 1.8, L(0.5–8 keV)/L(178 MHz) < 0, low radio core fraction (log R_CD < −3), and/or strong 9.7 μm silicate absorption. In the unification model, these sources are viewed at the highest inclination angles through optically thick material lying in the plane of the torus/accretion disk. The CT candidates represent 20% (9/44) of the total sample, which implies that CT material covers an angle of 12° ± 3° above and below the equatorial plane of the obscuring structure, as shown in Figure 16. The remaining Compton-thin NLRGs (with 10^22.5 < N_H/cm⁻² < 1.5 × 10²⁴) cover 21° ± 2°. Intermediate column density (10^21.5 < N_H/cm⁻² < 10^22.5) sources, including five low-N_H NLRGs (Section 6.2) and 3C 196, a red broad-line radio galaxy with relatively high N_H = 2.7 × 10²² cm⁻², cover 10° ± 4°. Figure 16 shows the geometry of the obscuring material found from these simple estimates, together with the high-z sample (Wilkes et al. 2013), and a summary is given in Table 6. In both samples, the covering factor for CT material is similar (same percentage of CT sources in both samples), but the opening angle of the torus is smaller for the sample at medium-z than at high-z (47° versus 60°), implying that the Compton-thin (10^21.5–10⁻²⁴ cm⁻²) part of the obscuring material (torus or accretion disk wind) is more "puffed up" in the medium-z 3CRR sample.

Zoom In Zoom Out Reset image size

Fabian et al. (2008) showed that the long-lived gas and dust clouds in the vicinity of an AGN have a range of column densities that depend on L/L_Edd, where N_H/cm⁻² > 5 × 10²³ × L/L_Edd. Ricci et al. (2017) studied a sample of local AGN (both type 1 and type 2 with median z = 0.037) from the all-sky hard X-ray (14−195 keV) Swift/BAT survey, for which reliable estimates of BH mass, intrinsic column densities, X-ray luminosities, and L/L_Edd were obtained. They found the fraction of CT sources in their hard X-ray–selected sample to be ∼23% ± 6%, independent of L/L_Edd, and similar to the fraction in the medium- and high-z 3CRR samples. However, the fraction of Compton-thin but obscured sources strongly decreases with L/L_Edd in their sample from 0.8 for L/L_Edd < 0.01 to 0.2 for L/L_Edd > 0.1. Ricci et al. (2017) therefore suggested a "radiation-regulated unification" model, where the covering factor of the Compton-thin gas (10²² < N_H/cm⁻² < 10²⁴) increases with decreasing L/L_Edd, while the covering factor of the CT gas stays the same. In this model, for lower L/L_Edd, the obscuring structure (torus/accretion disk wind) is more puffed up (see their Figure 4). Our results for the medium-z sample imply the presence of a puffed-up torus in the low-N_H NLRGs, suggesting that L/L_Edd extends to lower values than those in the high-z sample.

The distributions of the intrinsic N_H in the medium- and high-z 3CRR samples are presented in Figure 17. The high-z sample (on the right) shows a bimodal distribution, where quasars have N_H < 10^22.5 cm⁻², consistent with low obscuration at face-on inclination angles, while the NLRGs show N_H > 10^22.5 cm⁻², implying higher obscuration at higher inclination angles, consistent with unification schemes. There are two quasars with moderate column densities (10^22.5 < N_H /cm⁻² < 10²³) and hard HRs (0 < HR < 0.5) in this sample. In the medium-z sample, the distributions of quasars and NLRGs overlap. Although the quasars show N_H < 10^22.5 cm⁻², similar to quasars at high redshifts, the NLRGs have a much broader range of column densities that extend to lower, quasar-like values in the low-N_H NLRGs. These NLRGs possibly have low L/L_Edd, which allows clouds with low column density to form in the vicinity of the central engine (Section 6.3). Such low L/L_Edd NLRGs are missing from the high-z sample.

Zoom In Zoom Out Reset image size

Although a simple unification model was sufficient to explain the X-ray data and the bimodal N_H distribution in the high-z sample, this is not the case in the medium-z sample. An additional parameter, a range of L/L_Edd, is required to explain the large range of N_H in NLRGs seen at intermediate inclination angles, skimming the edge of the torus or accretion disk atmosphere/wind. As a result, the broad range of L/L_Edd smears the N_H distribution for NLRGs, removing the bimodality that was found in the high-z sample. Turning this argument around, because the unification model was sufficient for the high-z 3CRR sample, producing a bimodal and narrow N_H distribution, the L/L_Edd ratio must have a narrower range and higher values compared to the medium-z sample, allowing orientation effects to dominate the properties of the high-z sample. To test this hypothesis, we compiled spectra of the high-z 3CRR quasars (from the SDSS archive; Barthel et al. 1990; M. Vestergaard & D. Stern 2021, private communication) and measured the black hole masses from the widths of the C iv and Mg ii emission lines. The masses (measured in 12 out of 20 high-z quasars) are in the range of M_BH = 10^7.7 to −10^9.0 M_☉. The radio–to–X-ray SEDs, compiled using data from the NED, provided estimates of bolometric luminosities. The inferred L/L_Edd ratios are indeed high, >0.3, implying that orientation dominates the observed properties of the high-z sample; therefore, simple unification suffices.

The distribution of the intrinsic 0.5–8 keV X-ray luminosity (obtained from HBM modeling; Section 4.5) of the medium- and high-z 3CRR populations is presented in Figure 18. The medium-z sample peaks at lower L_X (mean log L_X/erg s⁻¹ = 44.97 ± 0.09) and has a broader intrinsic L_X distribution (σ = 0.51) extending to ∼10 times lower L_X values than the distribution for the high-z sample (the high-luminosity tail in the medium-z sample is due to a simplistic treatment of piled-up sources for which N_H < 10²¹ cm⁻² was assumed). The high-z sample shows a narrower distribution (σ = 0.27) peaking at higher L_X values (mean log L_X/erg s⁻¹ = 45.48 ± 0.06 erg s⁻¹). Because the intrinsic L_X depends on L/L_Edd (e.g., Czerny et al. 1996, Figure 1), we interpret the difference as due to a broad range of L/L_Edd in the medium-z sample extending to lower values, while the high-z sample has a higher L/L_Edd with a narrower range. The different distributions of intrinsic N_H in the two samples are also consistent with this scenario (Section 7.2).

Zoom In Zoom Out Reset image size

Obscuration in AGN is highly anisotropic and strongly wavelength-dependent. Hence, the "obscured fraction," defined as the ratio of the number of obscured AGN (either optically classified type 2s or those with N_H > 10²² cm⁻² in X-ray studies) to all AGN and its dependence on luminosity and/or redshift differs for samples selected at different wave bands. Optical surveys at low redshift (z < 0.05) and low (Seyfert) luminosity find obscured fractions of ∼0.65–0.75 (Lawrence & Elvis 1982; Huchra & Burg 1992; Maiolino & Rieke 1995), implying that there are two to three times more type 2s than type 1s in the local universe. High-luminosity, radio-selected, and hence unbiased by orientation samples with z > 0.3 find an optical obscured fraction of ∼0.6, consistent with a torus half-opening angle of ∼53° in unification models (Willott et al. 2000) and a luminosity dependence (Grimes et al. 2005) consistent with the "receding torus model." The X-ray surveys, sensitive to gas rather than dust obscuration and probing deeper into the obscured AGN population, find a wide range of obscured fractions, ∼0.1–0.8, decreasing with luminosity and increasing with redshift (La Franca et al. 2005; Treister & Urry 2006; Hasinger 2008; Burlon et al. 2011; Sazonov et al. 2012), although Ricci et al. (2017), using a local Swift/BAT-selected sample, showed that the dependence is primarily with L/L_Edd.

The obscured fraction in the medium-z 3CRR sample studied in this paper is 0.68 when the optical classification (based on the presence or absence of the broad emission lines in optical spectra) is used. However, if the classification is based on X-rays, where N_H = 10²² cm⁻² is assumed to divide obscured from unobscured sources, then four out of the five low-N_H NLRGs will qualify as X-ray unobscured, and 3C 196, a quasar with N_H > 10²² cm⁻², will qualify as X-ray obscured, yielding an obscured fraction of 0.61. The ratio of X-ray unobscured (N_H < 10²² cm⁻²) to Compton-thin obscured (10²² < N_H < 1.5 × 10²⁴) to CT (N_H > 1.5 × 10²⁴) sources is then 1.9:2:1.

In the high-z 3CRR sample, the obscured fraction is lower. It is 0.42 if optical classification is used and 0.5 if X-ray classification is used, the difference being due to two quasars with N_H > 10²² cm⁻² classified as obscured in X-rays. The ratio of X-ray unobscured to Compton-thin obscured to CT sources is 2.5:1.4:1.

The difference between optical and X-ray obscured fractions comes from four low-N_H, low L/L_Edd NLRGs in the medium-z 3CRR sample and two high-N_H, high-L/L_Edd quasars in the high-z sample. In the former case, the X-ray obscured fraction is lower in comparison with the optical obscured fraction, while in the latter case (high-L/L_Edd sample), it is higher.

As shown above, the obscured fraction is an inaccurate tool for measuring the level of obscuration in a sample. The obscured fraction depends not only on the sample's wavelength selection, luminosity, and redshift but also on whether optical or X-ray classification is used. It also depends on the sample's L/L_Edd range, which defines the geometry of the obscuring material (a more puffed-up torus for lower L/L_Edd; see Section 7.1) and number of sources with inconsistent optical and X-ray types.

The obscured AGN fraction in the medium- and high-z samples differs slightly depending on whether the source classification is based on optical spectra or X-ray data. Merloni et al. (2014) studied AGN with a wide range of redshifts (0.3 < z < 3.5) in the XMM-COSMOS survey and found that setting the dividing line between type 1 and type 2 at N_H = 10^21.5 cm⁻² rather than 10²² cm⁻² gives a better correspondence between optical and X-ray type. However, even then, ∼30% of AGN in their sample have conflicting optical and X-ray classifications. At dust extinctions A_V = 5–6 mag, the broad emission lines Hβ and Hα are totally obscured. This corresponds to column densities N_H = (0.9 to −1.1) × 10²² cm⁻² for a Galactic dust-to-gas ratio. A small (factor of a few) divergence from the Galactic dust-to-gas ratio will result in inconsistent X-ray and optical classifications around the dividing type 1/type 2 column density of N_H = 10²² cm⁻².

Merloni et al. (2014) found that the AGN with conflicting optical and X-ray types can be divided into two classes:

• 1.  
  optical type 1 and X-ray type 2 sources, which are high-luminosity broad-line AGN with X-rays absorbed by dust-free material lying at subparsec scales, and
• 2.  
  optical type 2 and X-ray type 1 sources, which are low-luminosity, unobscured AGN where the broad lines are probably diluted by the host galaxy.

The radio-selected 3CRR sample can give further insight into the nature of sources with inconsistent optical and X-ray classifications.

• 1.  
  The high-z sample has two quasars (optical type 1) with a high column density of N_H = 10^22.7−23 cm⁻² and HR > 0 (X-ray type 2). These sources (3C 68.1 and 325) have high L/L_Edd > 0.3 and intermediate viewing angles (−3 < log R_CD < −2), where our line of sight is skimming the edge of the accretion disk or torus. In these high-L/L_Edd sources, the X-rays are possibly obscured by gas in the strong, outflowing accretion disk wind (Luo et al. 2015; Ni et al. 2018), while the BLR is visible directly. Because of strong UV radiation pressure (high L/L_Edd), the low-N_H gas and dust clouds are blown away.
• 2.  
  The medium-z 3CRR sample has five NLRGs (optical type 2) with low column density N_H < 10²² cm⁻² and quasar-like HR < 0 (X-ray type 1). They have intermediate viewing angles, skimming the edge of the torus/accretion disk. These NLRGs have low L/L_Edd, which allows low-N_H clouds to survive in the vicinity of the nucleus and results in a puffed-up torus (see Section 6.3), which can hide the BLR.

In both the high- and medium-z 3CRR samples, AGN with conflicting optical/X-ray types have intermediate radio core fractions (−3 < log R_CD < −2), where viewing angles are skimming the edge of the accretion disk or torus. In this regime, the torus and accretion disk are most vulnerable to changes in the L/L_Edd ratio. We find that sources classified as optical type 1 and X-ray type 2 (X-ray-obscured quasars) have high L/L_Edd ratios, where the strong accretion disk winds obscure the X-rays. The optical type 2 and X-ray type 1 sources (unobscured NLRGs) are low-L/L_Edd AGN, where the edge or atmosphere of the puffed-up dusty torus provides obscuration for both the X-rays and the BLR.

Inference on the obscured fraction in this sample is challenging because of substantial uncertainties of measurement. For several observations, short exposures give large uncertainties on the line-of-sight obscuration, which can additionally be degenerate with the intrinsic AGN luminosity. We want to incorporate these uncertainties to produce realistic estimates of the obscured fraction. A self-consistent framework to do this is an HBM. We begin by writing down the Bayes theorem for an individual object:

$$\begin{eqnarray*}&&p(\theta | D)=\displaystyle \frac{p(D| \theta )\times p(\theta )}{p(D)}.\end{eqnarray*}$$

The posterior probability distribution p(θ∣D) of the parameters θ = (L_X, N_H) is primarily shaped by the likelihood function p(D∣θ), given by the Poisson count probability (Cash 1979) comparing the detected counts c_i to the assumed X-ray spectral model m_i propagated through the detector response:

$$\begin{eqnarray*}&&p(D| \theta )=\displaystyle \prod _{i}\mathrm{Poisson}({c}_{i};{m}_{i}(\theta )).\end{eqnarray*}$$

The second ingredient is p(θ), which normally is the prior of the Bayesian computation, describing the prior knowledge of the parameters θ. In an HBM, we estimate p(θ) simultaneously from the observations themselves.

To this end, we define only the shape of p(θ), assuming population distributions with hyperparameters (parameters of the prior distribution) H:

$$\begin{eqnarray*}&&p(\theta ,H)=p({L}_{{\rm{X}}},H)\times p({N}_{{\rm{H}}},H).\end{eqnarray*}$$

As an example, the population distribution could be described by Gaussian distributions whose parameters would be the hyperparameters H.

For the column density N_H, we assume a log-uniform distribution within three bins ("unobscured," 20–22; "Compton-thin obscured," 22–24; and "Compton-thick," 24–26). The relative ratio is defined by the obscured fraction, f_obsc, and the fraction of obscured AGN that are CT, f_CT:

$$\begin{eqnarray*}p({N}_{{\rm{H}}},{f}_{\mathrm{obsc}},{f}_{\mathrm{CT}})=\left\{\begin{array}{ll}1-{f}_{\mathrm{obsc}} & \mathrm{if}\,{N}_{{\rm{H}}}\lt {10}^{22}\,{\mathrm{cm}}^{-2}\\ {f}_{\mathrm{obsc}}\times (1-{f}_{\mathrm{CT}}) & \mathrm{if}\,{N}_{{\rm{H}}}={10}^{22-24}\,{\mathrm{cm}}^{-2}\\ {f}_{\mathrm{CT}} & \mathrm{if}\,{N}_{{\rm{H}}}\gt {10}^{24}\,{\mathrm{cm}}^{-2}.\end{array}\right.\end{eqnarray*}$$

For the luminosity distribution, we adopt the flexible beta distribution (adopting a Gaussian or Student-t distribution instead does not change the results significantly):

$$\begin{eqnarray*}&&p({L}_{{\rm{X}}},\mu ,\sigma ,a,b)={\rm{b}}{\rm{e}}{\rm{t}}{\rm{a}}({\rm{l}}{\rm{o}}{\rm{g}}{L}_{X};\mu ,\sigma ,a,b)\times p(\mu ,\sigma ,a,b).\end{eqnarray*}$$

For the population hyperparameters, we adopt a uniform prior on the mean logarithmic luminosity μ, a log-uniform prior on the population dispersion scale σ, and uniform priors on the shape parameters a and b.

We can then estimate the parameters θ for all sources (their L_X, N_H) simultaneously with the hyperparameters H = (f_obsc, f_CT, μ, σ, a, b). That means we explore a (6 + N × 2)-dimensional parameter space,

$$\begin{eqnarray*}&&p(H,{\theta }_{0},{\theta }_{1}...,{\theta }_{N})=\displaystyle \prod _{i}p(D| {\theta }_{i})\times p({\theta }_{i},H),\end{eqnarray*}$$

and, as usual in Bayesian inference, derive marginalized probability distributions on the physical parameters (e.g., N_H), but also the population distribution in N_H.

In this analysis, the source parameters influence the population distributions. At the same time, if the population distribution is well constrained by the majority of sources, a source with poor observational constraints can benefit from the population distribution, as it gives a prior where the parameters are most likely. Thus, the HBM strengthens weak observations (the population informs the inference of individual objects down the hierarchy) and allows inference of the population distribution (individual objects inform the population distribution up the hierarchy). The model is illustrated in Figure 19.

Zoom In Zoom Out Reset image size

In practice, we compute the hierarchical model in two steps. First, we compute p(θ_i ∣D) for each object under uninformative (flat) priors. This is a simple X-ray spectral analysis with the BXA (Buchner et al. 2014) module for Sherpa (Fruscione et al. 2006), assuming an AGN with obscuration (BNTORUS model; Brightman & Nandra 2011) with a warm-mirror power law added (similar to the scattered AGN light component in Section 4.4). All normalizations have wide log-uniform (uninformative) priors, and the intrinsic photon index is assigned a Gaussian prior centered at 1.95 with a standard deviation of 0.15. The warm-mirror normalization can reach up to 10% of the intrinsic AGN power-law component. This standard setup is described in, e.g., Buchner et al. (2014). The X-ray spectral analysis produces posterior distributions in p(θ_i ∣D) that are described by equally probable posterior samples θ_ij (as may be familiar from Markov Chain Monte Carlo analyses). In our case, we select M = 1000 posterior samples for each source. These samples cluster where the posterior is most probable and thus can be used as weight points in Monte Carlo integrations.

To constrain the population parameters, we then evaluate the population distribution at the object posterior sample locations,

$$\begin{eqnarray*}&&p(H)=\displaystyle \prod _{i}\displaystyle \frac{1}{M}\displaystyle \sum _{j}p({\theta }_{{ij}},H),\end{eqnarray*}$$

and only need to explore a six-dimensional parameter space. This is well defined because the θ_ij samples indicate where the population distributions have the most weight. If the samples from one object are clustered distant from another object's samples, then the population distribution must spread to cover both. On the other hand, because the population distribution is a probability distribution normalized to unity, extremely wide distributions give low probabilities at any specific location. Therefore the population distribution will prefer to cover the samples. If uncertainties (cluster widths) become large, both narrow and broad population distributions are similarly probable. Thereby, this formalism self-consistently carries forward the uncertainties from each source analysis into the uncertainties on population parameters.

We first analyze the spectra of sources independently with BXA and report the N_H and L_X constraints under flat priors. The probability distributions of N_H and L_X for each source are presented in Figures 3 and 4, respectively, as gray contours.

We correct for significantly piled-up (>20%) sources, where naive spectral analysis may be biased by assuming that these are unobscured luminous AGN, in the following way. We assume a log-uniform column density probability $\mathrm{log}{N}_{{\rm{H}}}\,=20-21$. For the luminosity distribution, we take the lower limit of the luminosity derived from spectral analysis as a lower limit on the true luminosity and assume a log-uniform distribution extending to very high luminosities ($\mathrm{log}{L}_{{\rm{X}}}\,=\mathrm{log}{L}_{{\rm{X}},\min }-47$). The population model will truncate the high-luminosity end based on other sources (see Figures 3 and 4, where gray contours show probability distributions from spectral analysis, and red contours show updated probability distributions after reweighting by the HBM analysis). This suppresses the extremely high luminosities of the piled-up sources (marked with an asterisk) and some CT, high-luminosity secondary solutions (e.g., in 3C 184 and 280) based on the CT fractions of the well-constrained (by HBM) sources.

Figure 5 shows the constraints on both the luminosity and column density for all sources in the medium-z 3CRR sample.

We now use the HBM to estimate the intrinsic luminosity distribution and obscured fractions of the population. The luminosity distribution for the medium-z 3CRR sample is shown in gray in Figure 18. It is centered at μ = 45.1 ± 1.2 and σ = 4.5 ± 1.0 wide. The shape of the distribution is described by a = 7.7 ± 1.8 and b = 3.6 ± 2.6, indicating a right-skewed, steeply falling distribution.

We perform the same analysis for the high-z 3CRR sample analyzed in Wilkes et al. (2013) and find consistent results in the spectral analysis and reported obscured and CT fractions. The X-ray luminosity distribution for the high-z sample is shown in yellow in Figure 18.

Finally, to investigate the dependence of obscured and CT fractions on orientation (i.e., R_CD), we modify the HBM to allow three different groups (log R_CD < −3, −3 < log R_CD < −2, and −2 < log R_CD < 0) to have different obscured fractions while still enforcing the same luminosity distribution for all groups. The results are presented in Figure 20, with the total obscured fraction in the top panels and the CT fraction in the bottom panels. The obscured fraction increases to >70% for more intermediate and edge-on viewing angles (log R_CD < −2). It is remarkably low (≲20%) for face-on sources. The CT fraction is small until the lowest log R_CD < −3 values are reached; in those edge-on sources, it reaches $\sim 60{\rm{ \% }}$. Within the uncertainties, the obscuration results from the two samples are consistent with each other.

Zoom In Zoom Out Reset image size

The obscured fractions are shown in Figure 20. We obtain an upper limit on the Compton-thick fraction of all AGN in both medium- and high-z samples of 20% and an obscured fraction of 55% ± 10%. The low Compton-thick fraction (found from the HBM) is due to the few secure CT candidates (see Figure 5). However, when information from multiwavelength data is included in estimating the number of CT sources (Section 7.4), the CT fraction for the medium-z sample and the most edge-on inclination sources (log R_CD < −3) increases and approaches those found for the high-z sample. All obscured and CT fractions, derived with the HBM, are listed in Table 7.

Table 7. Obscured Fractions Derived with the HBM

	Obscured Fraction	CT Fraction
	Medium-z Sample
Full sample	${0.57}_{+0.08}^{+0.09}$	${0.07}_{+0.04}^{+0.07}$
−2 < log RCD < 0	${0.19}_{+0.08}^{+0.11}$	${0.05}_{+0.04}^{+0.07}$
−3 < log RCD < −2	${0.81}_{+0.11}^{+0.08}$	${0.11}_{+0.07}^{+0.12}$
log RCD < −3	${0.92}_{+0.10}^{+0.05}$	${0.34}_{+0.17}^{+0.21}$
	High-z Sample
Full sample	${0.51}_{+0.07}^{+0.08}$	${0.26}_{+0.06}^{+0.07}$
−2 < log RCD < 0	${0.12}_{+0.06}^{+0.08}$	${0.06}_{+0.04}^{+0.07}$
−3 < log RCD < −2	${0.78}_{+0.10}^{+0.09}$	${0.30}_{+0.10}^{+0.11}$
log RCD < −3	${0.92}_{+0.12}^{+0.06}$	${0.74}_{+0.17}^{+0.12}$

Download table as:  ASCIITypeset image