BAT AGN Spectroscopic Survey. I. Spectral Measurements, Derived Quantities, and AGN Demographics

The BAT survey is an all-sky survey in the ultra-hard X-ray range ($\gt 10$ keV) that, as of its first 70 months²¹ of operation, has identified 1210 objects (Baumgartner et al. 2013). Because of the large positional uncertainty of BAT ($\approx 2^{\prime}$) higher angular resolution X-ray data for every source from Swift-XRT or archival data have been obtained, providing associations for 97% of BAT sources. From this catalog, we exclude BAT sources lacking a soft X-ray counterpart, and those associated with galaxy clusters and Galactic sources. In addition, we exclude M82, which is a very nearby star-forming galaxy detected by Swift-BAT without the presence of an AGN. This leaves 836 BAT-detected AGNs.

Almost all of the host galaxy AGN counterparts used in this study (99%, 633/642) are based on the published counterparts in the Baumgartner et al. (2013) paper, which was based on Swift/XRT, XMM-Newton, Chandra, Suzaku, and ASCA follow-up. In Baumgartner et al. (2013), the host galaxy counterpart of the BAT AGN was determined using the brightest counterpart above 3 keV typically from Swift/X-Ray Telescope (XRT) observations, within the BAT error radius (≈3'). In only nine cases we use updated galaxy counterparts based on subsequent published studies. These differences are associated with the updated counterparts for SWIFT J1448.7–4009 and SWIFT J1747.7–2253 provided in Masetti et al. (2013); SWIFT J0634.7–7445, SWIFT J0654.6+0700, and SWIFT J2157.4–0615 found in Parisi et al. (2014); SWIFT J0632.8+6343, SWIFT J0632.8+6343, and SWIFT J1238.6+0928 provided in the updated INTEGRAL catalog of Malizia et al. (2016), and finally SWIFT J0350.1–5019; we use the counterpart from Ricci et al. (2017) (ESO201-4). Another issue is that some BAT AGN counterparts host dual AGNs (Koss et al. 2016b); however, the median value of the(${L}_{2\mbox{--}10\,\mathrm{keV}}^{\mathrm{int}}$) ratio between the dual AGNs is 11 (Koss et al. 2012), so the majority of the emission in the BAT detection is typically coming from a single AGN. A full review of all of the AGN galaxy counterpart identifications and dual AGNs is provided in Ricci et al. (2017).

We then cross-match this sample to the Roma Blazar Catalog (BZCAT) catalog (Massaro et al. 2009) and identify 11% (96/836) of AGNs as possibly beamed sources, such as blazars or flat spectrum radio quasars where Doppler boosting may amplify the non-thermal emission including the hard X-rays. As the Roma BZCAT authors note, classifications of beamed AGNs in their catalog have significant uncertainty because only the very brightest AGNs have the necessary polarimetric observations or detections of compact cores and superluminal motions using high-resolution radio imaging combined with variability studies.

All of the AGNs in the BAT survey have been analyzed using X-ray observations spanning 0.3–195 keV. This includes homogeneous model-fitting using some of the best available soft X-ray data in the 0.3−10 keV band from XMM-Newton, Chandra, Suzaku, or Swift/XRT, and the 14–195 keV band from Swift/BAT. This analysis provides measurements of the obscuring column (${N}_{{\rm{H}}}$) and intrinsic X-ray emission (${L}_{2\mbox{--}10\mathrm{keV}}^{\mathrm{int}}$). Full details of the X-ray analysis and fitting measurements are provided in separate publications (Ricci et al. 2015, Ricci et al. 2017).

The goal of the BASS sample is to use the largest available optical spectroscopic sample of Swift-BAT sources using dedicated observations and public archival data. In many cases, there were many duplicate observations of the same source. In such cases, we select the single best spectrum for line measurements for each AGN in our catalog from a single telescope.

For line-fitting, our first criterion was full coverage of the spectral region covering all the lines from Hβ to [S ii] λ6717,6731 (i.e., 4800–7000 Å). We then selected, based on the signal-to-noise in the continuum, for fitting of absorption lines and stellar population templates. Each spectrum was visually inspected to avoid any cases that may be problematic for fitting (e.g., bad sky subtraction or noise spikes). This process reduced the number of spectra from 972 spectra including duplicates to 642 unique spectra. Of these 642 unique spectra, 33% (209/642) were from targeted observations and 67% (433/642) were part of archives or previously published papers. A summary of all the observational setups is shown in Table 1 and Figure 1. The original spectra and model fits are available online.²²

Zoom In Zoom Out Reset image size

Table 1.  Summary of Instrumental Setups

Telescope	Inst.	Total	Grating	Slit	Resolution	Total Range	Velocity Dispersion Fit Range
				Width ('')	FWHM $[\mathring{\rm A}$]	$[\mathring{\rm A} ]$	$[\mathring{\rm A} ]$
SDSS	SDSS	142		2,3	2.76	3900–8000	3900–7000
UK Schmidt	6dF	112	600V, 316R	6.7	5.75	3900–7000	3900–7000
Perkins 1.8 m	Deveny	52	300	2	5.43	3900–7500	3900–5500
KPNO 2.1 m	Goldcam	36	32	2	7.4	4200–8800	4400–5500
		16	26, 35	2	3.3	3900–8500	3900–7000
SAAO 1.9 m	300	46	7	2	5.00	4000–7500	4600–7000
Hale 200 inch	DBSP	20	600	1.5	4.4, 5.8	3900–7000	3900–5500, 8450–8700
		15	600	2	4.8, 6.8	3900–7000	3900–5500, 8450–8700
Gemini 8.1 m	GMOS	21	B600	1	4.84	4000–7000	4000–7000
		5		0.75	3.75	4300–7000	4300–7000
		3		0.75	3.72	4000–7000	4000–7000
CTIO 1.5 m	R–C	10	26, 35	2	4.30	3900–7500	3900–7000
		2	47	2	3.10	5200–7500	5200–7000
		2	36	2	2.20	4500–5500	4500–5500
Tillinghast 1.5 m	FAST	10	300	3	5.5	3700–7500
UH 2.2 m	SNIFS	5	300	2.4	5.80	3200–7000	3900–5500
APO 3.5 m	DIS	4	B400, R300	1.5	7.00	3900–7000	3900–5600
Shane 3 m	Kast	4	600, 830	2	4.0, 3.2	3900–7000	3900–7000

Download table as:  ASCIITypeset image

2.2.1. Archival Public Data

2.2.2. Targeted Spectroscopic Observations

We use the penalized PiXel Fitting software (pPXF; Cappellari & Emsellem 2004) to measure stellar kinematics and the central stellar velocity dispersion (${\sigma }_{* }$). This method operates in pixel space and uses a maximum penalized likelihood approach for deriving the line-of-sight velocity distribution (LOSVD) from kinematic data (Merritt 1997). As a first step, the pPxf code (version 5.1.9) creates a model galaxy spectrum ${G}_{\mathrm{mod}}(x)$ by convolving empirical stellar population models by a parameterized LOSVD. Then, it determines the best-fitting parameters of the LOSVD by minimizing the value of ${\chi }^{2}$, which measures the agreement between the model and the observed galaxy spectrum over the set of reliable data pixels used in the fitting process. Finally, pPxf uses the "best-fit spectra" to calculate the velocity dispersion and associated uncertainty from the absorption lines.

The pPxf code uses a large set of single stellar populations to fit each galaxy spectrum. We used the templates from the Miles Indo-U.S. Catalog (MIUSCAT) library of stellar spectra (Vazdekis et al. 2012). The MIUSCAT library of stellar spectra contains $\approx 1200$ well-calibrated stars covering the spectral region of 3525–9469 Å at a spectral resolution of 2.51 Å (FWHM). These spectra are then computed into stellar libraries with an IMF slope of 1.3, and the full range of metallicities ($M/H=-2.27$ to +0.40) and ages (0.03–14 Gyr). These templates have been observed at higher spectral resolution (FWHM = 2.51 Å) than the AGN observations and are convolved in pPxf to the spectral resolution of each observation before fitting.

We fitted the spectra in the wavelength region 3900–7000 Å, if this entire range was covered by the given spectrum. This range covers the Ca H+K λ3935, 3968 and Mg i $\lambda 5175$ absorption features. For some spectra the velocity dispersion was estimated only based on the Mg i absorption line, because of a lack of blue wavelength coverage (e.g., spectra taken with KPNO and grating 32, Perkins, and some Gemini gratings). To avoid complications with discontinuities and dispersion changes in spectra that have both blue and red setups, only the blue channel spectra (3800–5500 Å) were used to estimate velocity dispersion, focusing on the Ca H+K and Mg i absorption lines. The only exception is the dual-channel SDSS data, where the full range was fit. Whenever available, we also fitted the Ca ii triplet spectral region (8450–8700 Å).

We modified the ppxf-kinematics-example-sdss code to measure stellar kinematics in our sample. Table 3 shows the emission and bright night sky lines that were masked. The code automatically applies a mask for several bright emission lines: Hδ, Hγ, Hβ, [O iii] λ5007, [O i]$\,\lambda 6300$, [N ii] λ6583, Hα, and [S ii] λ6717,6731. We masked sky lines at λ5577, λ6300, λ6363, and λ6863. We also mask the region around the Ca H λ3968 line, because of overlap with the H λ3970 and the [Ne iii] $\lambda 3968$ emission lines (see, e.g., Greene & Ho 2005a). We have also masked the region surrounding the Na I line, since it may be affected by interstellar absorption. Finally, we masked regions affected by sky emission lines. The width of these emission-line masks was set to 2400 km s⁻¹. For the broad-line AGNs, we use a wider mask (3200 km s⁻¹) for the Balmer lines (Hα, Hβ, Hγ, Hδ), in order to mask the broad emission components. This allowed us to measure the velocity dispersion for some AGNs with broad emission in Hα, but narrow Hβ.

3.1.1. Velocity Dispersion Measurements

Figure 2 summarizes the results of the stellar velocity dispersion measurements and individual measurements are found in Table 4. ${\rm{\Delta }}{\sigma }_{* }$ denotes the error in ${\sigma }_{* }$. We were able to achieve a reliable velocity dispersion measurement, with ${\rm{\Delta }}{\sigma }_{* }\lt 60\,\mathrm{km}\,{{\rm{s}}}^{-1}$, in $201$ of the $642$ (31.3%) galaxies in our sample. For these AGNs, the stellar continuum was subtracted prior to emission-line-fitting (discussed below). For most AGNs with broad Hβ, the AGN continuum contaminated the host galaxy and stellar absorption lines, limiting reliable ${\sigma }_{* }$ measurements to only 13 type 1.2–1.8 AGNs. Additionally, in $166$ sources the signal-to-noise ratio and spectral resolution were too low to robustly identify the absorption lines.

Zoom In Zoom Out Reset image size

Table 4.  Stellar Velocity Dispersion Measurements

IDa	Source	Redshiftb	σ	log ${M}_{\mathrm{BH}}/{M}_{\odot }$	${\mathrm{Flag}}_{\sigma }$ c	Ca H+Kd	Mgbe	${\sigma }_{\mathrm{CaT}}$ f	CaTg	${\sigma }_{\mathrm{lit}.}$	$\mathrm{log}{({M}_{\mathrm{BH}}/{M}_{\odot })}_{\mathrm{lit}.}$ h	Referencesi
			(km s−1)					(km s−1)		(km s−1)
1	SDSS	0.03767	152 ± 6	7.97 ± 0.30	1	1	1	⋯	⋯	⋯	⋯	⋯
2	6DF	0.05846	180 ± 38	⋯	7	1	1	⋯	⋯	⋯	⋯	⋯
4	SDSS	0.03970	142 ± 7	7.85 ± 0.30	1	1	1	⋯	⋯	⋯	⋯	⋯
5	Masetti	⋯	⋯	⋯	9	1	1	⋯	⋯	⋯	⋯	⋯
6	Perkins	⋯	⋯	⋯	9	1	1	⋯	⋯	⋯	⋯	⋯
7	SDSS	0.07334	250 ± 16	8.91 ± 0.32	1	1	1	⋯	⋯	⋯	⋯	⋯
8	Gemini	⋯	⋯	⋯	9	0	1	⋯	⋯	⋯	⋯	⋯
10	6DF	0.09561	284 ± 21	9.16 ± 0.32	2	1	1	⋯	⋯	⋯	⋯	⋯
13	Palomar	⋯	⋯	⋯	9	1	1	⋯	9	⋯	⋯	⋯
14	6DF	0.06286	326 ± 49	⋯	7	1	1	⋯	⋯	⋯	⋯	⋯

Notes.

^aSwift-BAT  70-month hard X-ray survey ID (http://swift.gsfc.nasa.gov/results/bs70mon/). ^bRedshift measured from the stellar template. ^cQuality flag: 1—excellent fit with small error ($\langle {\sigma }_{* }\rangle$ error = 11 $\mathrm{km}\,{{\rm{s}}}^{-1}$, ${\sigma }_{* }\lt 34$ $\mathrm{km}\,{{\rm{s}}}^{-1}$), 2—larger errors than flag 1 ($\langle {\sigma }_{* }\rangle$ error = 23 $\mathrm{km}\,{{\rm{s}}}^{-1}$, ${\sigma }_{* }\lt 60$ $\mathrm{km}\,{{\rm{s}}}^{-1}$), but acceptable fit, 3—bad fit with high S/N, 7—presence of broad component at Hβ or Hα, 8—very weak absorption features, 9—bad fit. ^dFlag = 1 when Ca H+K λ3935, 3968 is fitted. ^eFlag = 1 when Mg i is fitted. ^f ${\sigma }_{* }$ measured from Ca ii triplet. ^gFlag = 1 when Ca ii triplet is fitted. ^hBlack hole mass from literature: C05 (Capetti et al. 2005); C09 (Cappellari et al. 2009); D03 (Devereux et al. 2003); H05 (Herrnstein et al. 2005); K08 (Kondratko et al. 2008); K11 (Kuo et al. 2011); L03 (Lodato & Bertin 2003); M11 (Medling et al. 2011); O14 (Onken et al. 2014); RJ06 (Rothberg & Joseph 2006); T03 (Tadhunter et al. 2003); TYK05 (Trotter et al. 1998; Yamauchi et al. 2004; Kondratko et al. 2005); W06 (Wold et al. 2006); W12 (Walsh et al. 2012). ⁱReference for ${\sigma }_{\mathrm{lit}.}$: C04 (Cid Fernandes et al. 2004); F00 (Ferrarese & Merritt 2000); G05 (Garcia-Rissmann et al. 2005); G13 (Grier et al. 2013); Hy (http://leda.univ-lyon1.fr); H09 (Ho et al. 2009); L17 (Lamperti et al. 2017); M13 (McConnell & Ma 2013); NW95 (Nelson & Whittle 1995); N04 (Nelson et al. 2004); RJ06 (Rothberg & Joseph 2006); V15 (van den Bosch et al. 2015).

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Four authors (M.K., B.T., S.B., and I.L.) visually inspected the fit of the stellar continuum and absorption lines, and assigned a quality flag to each spectrum. In our sample we have $128$ spectra with flag 1 and $73$ spectra with flag 2, which both designate reliable measurements. More details and examples of these quality flags are given in the Appendix A.1. Generally, the ${\sigma }_{* }$ uncertainty measured by pPxf for flag 1 fits is typically small ($\langle {\rm{\Delta }}{\sigma }_{* }\rangle \simeq 12\,\mathrm{km}\,{{\rm{s}}}^{-1}$). Flag 2 fits have somewhat worse quality fits, judged from our visual inspection, consistent with pPxf measurements ($\langle {\rm{\Delta }}{\sigma }_{* }\rangle \simeq 27\,\mathrm{km}\,{{\rm{s}}}^{-1}$), but the Ca H+K and Mg i absorption lines are still well fit.

For AGNs with reliable measurements of ${\sigma }_{* }$ we calculated the black hole mass, ${M}_{\mathrm{BH}}$, using the ${M}_{\mathrm{BH}}$–${\sigma }_{* }$ relation. We use the relation from Kormendy & Ho (2013):

$$\begin{eqnarray}&&\mathrm{log}\left(\displaystyle \frac{{M}_{\mathrm{BH}}}{{M}_{\odot }}\right)=4.38\times \mathrm{log}\left(\displaystyle \frac{{\sigma }_{* }}{200\,\mathrm{km}\,{{\rm{s}}}^{-1}}\right)+8.49.\end{eqnarray} \tag{ 1 }$$

The slope of this relation is shallower than the slope of the relation from McConnell & Ma (2013), who reported a value of 5.64, and is consistent with the slope of the relation from Gültekin et al. (2009). A small number of sources have direct measurements of black hole masses, either from reverberation mapping (39) or OH megamasers (8), which we have adopted and tabulated whenever available.

We fit emission lines in our sample of optical spectra using an extensive spectroscopic analysis toolkit for astronomy, PySpecKit, which uses a Levenberg–Marquardt algorithm for spectral fitting (Ginsburg & Mirocha 2011). All emission-line fits were visually examined by five authors (M.K., B.T., S.B., K.S., and I.L.) to verify proper fitting, and to adjust some subtle parameters. We implement separate methods for fitting sources with only narrow lines and for sources with broad lines. For narrow-line sources, we first fit and subtract a host stellar component, to remove the galaxy continuum and stellar absorption features, as described in Section 3.1. We then separately fit three spectral regions, focusing on the [O ii] (3300–4000 Å), Hβ (4650–5050 Å), and Hα (6250–6770 Å) emission lines. All measurements for narrow-line sources are listed in Tables 5–7 for the [O ii] $\lambda 3727$, Hβ, and Hα regions measurements, respectively. The emission-line classifications are provided in Table 8. For broad-line sources, the properties are listed in Table 9 for both Hβ and Hα.

Table 5.  Emission-line Measurements—[O ii] $\lambda 3727$ Spectral Region

IDa	FWHMb	[Ne v] $\lambda 3346$ c	[Ne v] $\lambda 3426$ c	[O ii] $\lambda 3727$ c	[Ne iii] $\lambda 3869$ c	[Ne iii] $\lambda 3968$ c	Flagd
	(km s−1)	(erg cm−2 s−1)	(erg cm−2 s−1)	(erg cm−2 s−1)	(erg cm−2 s−1)	(erg cm−2 s−1)
1	382 ± 4	⋯	⋯	1.5 ± 0.0	0.7 ± 0.0	$\lt 1.2$	1
2	632 ± 55	⋯	⋯	⋯	5.5 ± 0.2	$\lt 4.7$	2f
4	493 ± 2	⋯	⋯	5.5 ± 0.1	1.4 ± 0.0	$\lt 1.1$	1
5	⋯	$\lt 104.0$	$\lt 72.9$	$\lt 24.0$	$\lt 6.5$	$\lt 21.2$	9
6	801 ± 16	⋯	⋯	⋯	$\lt 38.0$	28.9 ± 0.8	2
7	671 ± 8	⋯	⋯	13.1 ± 0.1	2.8 ± 0.0	$\lt 1.9$	1
8	⋯	⋯	⋯	⋯	⋯	⋯	9
10	⋯	⋯	⋯	⋯	⋯	$\lt 1.5$	2f
13	⋯	$\lt 261.4$	$\lt 150.9$	$\lt 45.3$	$\lt 41.5$	$\lt 49.5$	9
14	⋯	⋯	⋯	$\lt 4.3$	$\lt 7.4$	$\lt 5.5$	2f

Notes.

^aSwift-BAT 70-month hard X-ray survey ID (http://swift.gsfc.nasa.gov/results/bs70mon/). ^bFWHM measured from [O ii] $\lambda 3727$. ^cEmission-line flux ($\times {10}^{-15}$). Symbols "⋯" and "<" indicate a lack of spectral coverage and the 3σ upper limit estimation, respectively. ^dSpectral fitting quality flag: 1—a good fit with small error, 2—acceptable fit, 3—bad fit for high S/N source due to either the presence of a broad-line component or an offset in emission lines, 9—lack of spectral coverage or no emission line is detected, f—poor calibration because a single flux calibration was applied to optical spectra taken over several different nights, as in the 6DF spectra.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 6.  Emission-line Measurements—Narrow Hβ Spectral Region

IDa	FWHMb	He ii $\,\lambda 4686$ c	Hβc	${\mathrm{Flag}}_{\mathrm{bH}\beta }$ d	[O iii] λ5007c	Flage
	(km s−1)	(erg cm−2 s−1)	(erg cm−2 s−1)		(erg cm−2 s−1)
1	249 ± 1	0.3 ± 0.0	1.1 ± 0.0	n	16.1 ± 0.0	1
2	560 ± 3	$\lt 5.4$	7.5 ± 0.4	n	44.6 ± 0.3	1f
4	382 ± 1	0.3 ± 0.0	1.2 ± 0.0	n	49.4 ± 0.0	1
5	1336 ± 3	$\lt 7.7$	12.4 ± 0.1	n	139.6 ± 0.2	2
6	609 ± 27	⋯	82.0 ± 15.7	b	276.2 ± 6.5	1
7	637 ± 2	$\lt 0.9$	4.5 ± 0.1	n	38.9 ± 0.1	1
8	416 ± 6	⋯	14.2 ± 2.7	b	181.9 ± 2.6	1
10	463 ± 98	$\lt 1.5$	$\lt 2.6$	n	11.0 ± 1.8	2f
13	330 ± 3	$\lt 22.5$	$\lt 17.6$	n	43.3 ± 1.3	1
14	506 ± 7	$\lt 5.1$	$\lt 6.2$	n	39.8 ± 0.3	1f

Notes.

^aSwift-BAT 70-month hard X-ray survey ID (http://swift.gsfc.nasa.gov/results/bs70mon/). ^bFWHM measured from narrow Hβ. ^cEmission-line flux ($\times {10}^{-15}$). Symbols "⋯" and "<" indicate a lack of spectral coverage and the 3σ upper limit estimation, respectively. ^dFlag discriminating narrow Hα ("n") and broad Hα ("b"). "h" denotes a high-redshift source (see Table 10). ^eSpectral fitting quality flag: 1—a good fit with small error, 2—acceptable fit, 3—bad fit for high S/N source due to either the presence of a broad-line component or an offset in emission lines, 9—lack of spectral coverage or no emission line is detected, f—poor calibration because a single flux calibration was applied to optical spectra taken over several different nights, as in the 6DF spectra.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 7.  Emission-line Measurements—Narrow Hα Spectral Region

IDa	FWHMb	[O i]$\,\lambda 6300$ c	Hαc	${\mathrm{Flag}}_{\mathrm{bH}\alpha }$ d	[N ii] λ6583c	[S ii] $\lambda 6716$ c	[S ii] $\lambda 6731$ c	Flage
	(km s−1)	(erg cm−2 s−1)	(erg cm−2 s−1)		(erg cm−2 s−1)	(erg cm−2 s−1)	(erg cm−2 s−1)
1	221 ± 2	0.6 ± 0.0	7.6 ± 0.0	b	1.9 ± 0.0	1.4 ± 0.0	1.4 ± 0.0	1
2	510 ± 5	5.7 ± 0.3	24.4 ± 0.2	b	22.1 ± 0.3	7.8 ± 0.3	6.9 ± 0.2	2f
4	426 ± 1	1.1 ± 0.0	10.5 ± 0.0	n	5.8 ± 0.0	3.4 ± 0.0	2.5 ± 0.0	1
5	269 ± 0	1.3 ± 0.0	1.7 ± 0.1	b	1.9 ± 0.0	5.8 ± 0.0	3.9 ± 0.1	3
6	597 ± 53	$\lt 17.5$	243.4 ± 6.9	b	22.9 ± 2.6	$\lt 50.8$	$\lt 50.8$	2
7	624 ± 1	6.8 ± 0.1	24.5 ± 0.1	n	22.6 ± 0.1	10.6 ± 0.1	9.7 ± 0.1	2
8	411 ± 0	10.9 ± 0.9	58.9 ± 0.8	b	24.3 ± 0.6	17.0 ± 0.7	14.9 ± 0.8	2
10	558 ± 70	$\lt 1.9$	4.6 ± 1.5	n	3.3 ± 0.6	$\lt 2.4$	$\lt 2.4$	2f
13	449 ± 3	$\lt 5.7$	24.0 ± 0.4	n	25.0 ± 0.2	9.2 ± 0.1	8.4 ± 0.4	1
14	506 ± 52	$\lt 5.7$	12.2 ± 1.4	b	13.8 ± 2.5	4.0 ± 1.4	2.0 ± 2.2	2f

Notes.

^aSwift-BAT 70-month hard X-ray survey ID (http://swift.gsfc.nasa.gov/results/bs70mon/). ^bFWHM measured from narrow Hα. ^cEmission-line flux ($\times {10}^{-15}$). Symbols "⋯" and "<" indicate a lack of spectral coverage and the 3σ upper limit estimation, respectively. ^dFlag discriminating narrow Hα ("n") and broad Hα ("b"). "h" denotes high-redshift source (see Table 10). ^eSpectral fitting quality flag: 1—a good fit with small error, 2—acceptable fit, 3—bad fit for high S/N source due to either the presence of a broad-line component or an offset in emission lines, 9—lack of spectral coverage or no emission line is detected, f—poor calibration because a single flux calibration was applied to optical spectra taken over several different nights, as in the 6DF spectra.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 8.  Strong Emission Line Classification

IDa	Counterpart Name	[N ii]/Hα	[S ii]/Hα	[O i]/${\rm{H}}\alpha$	He ii	[O iii]/[O ii]
1	2MASXJ00004876−0709117 (B)b	Seyfert	Seyfert	Seyfert	Seyfert	Seyfert
2	Fairall1203 (B)	Seyfert	Seyfert	Seyfert	No He ii detection	⋯c
4	2MASXJ00032742+2739173	Seyfert	Seyfert	Seyfert	Seyfert	Seyfert
5	2MASXJ00040192+7019185 (B)	Optically elusive	Optically elusive	Optically elusive	Optically elusive	Optically elusive
6	Mrk335 (B)	H ii	Optically elusive	Optically elusive	Seyfert	Optically elusive
7	2MASXJ00091156−0036551	Seyfert	Seyfert	Seyfert	No He ii detection	Seyfert
8	Mrk1501 (B)	Seyfert	Seyfert	Seyfert	Seyfert	⋯
10	2MASXJ00210753−1910056	AGN Limitd	Optically elusive	Optically elusive	No He ii detection	Optically elusive
13	2MASXJ00253292+6821442	AGN Limit	AGN Limit	Optically elusive	No He ii detection	⋯
14	2MASXJ00264073−5309479 (B)	AGN Limit	AGN Limit	Optically elusive	No He ii detection	Optically elusive

Notes.

^aSwift-BAT 70 month hard X-ray survey ID (http://swift.gsfc.nasa.gov/results/bs70mon/). ^bThe symbol (B) indicates a broad-line source. ^cThe symbol ⋯ indicates a lack of wavelength coverage, in reference to the "no wave" classification listed in Figures 10 and 11. ^dAGN Limit refers to objects that have an Hβ upper limit either in the Seyfert or in the LINER region.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 9.  Properties Derived from Spectral Decomposition of Broad Hβ and Hα Sources

IDa	$\mathrm{log}{L}_{5100}$	$\mathrm{log}{L}_{\mathrm{bol}}$ b	${\mathrm{FWHM}}_{\mathrm{bH}\beta }$	${\mathrm{EW}}_{\mathrm{bH}\beta }$	$\mathrm{log}{M}_{\mathrm{BH}}/{M}_{\odot }$ c	$\mathrm{log}L/{L}_{\mathrm{Edd}}$ c	${\mathrm{Flag}}_{b{\rm{H}}\beta }$ d	bHαd	${\mathrm{FWHM}}_{\mathrm{bH}\alpha }$	${\mathrm{EW}}_{\mathrm{bH}\alpha }$	$\mathrm{log}{M}_{\mathrm{BH}}/{M}_{\odot }$ e	${\mathrm{Flag}}_{b{\rm{H}}\alpha }$ f	$\mathrm{log}L/{L}_{\mathrm{Edd}}$ g	$\mathrm{log}{({M}_{\mathrm{BH}}/{M}_{\odot })}_{\mathrm{lit}.}$ h
	(erg s−1)	(erg s−1)	(km s−1)	(Å)		5100 Å		(erg cm−2 s−1)	(km s−1)	(Å)
1	⋯	⋯	⋯	⋯	⋯	⋯	⋯	9.5 ± 0.1	3680 ± 25	14.79	6.42	1	−0.00	⋯
6	43.90	44.80	2065	99.42	7.29	−0.67	1	1929.3 ± 8.7	1706 ± 669	297.48	6.86	2	−0.61	${7.23}_{0.04}^{+0.04}$
8	44.43	45.28	5295	100.66	8.45	−1.34	1	1076.6 ± 2.0	4704 ± 33	492.98	8.26	2	−0.67	${8.07}_{0.12}^{+0.17}$
16	44.66	45.50	2451	69.27	7.93	−0.61	1	⋯		⋯	⋯	⋯	⋯	${8.49}_{0.10}^{+0.12}$
18	⋯	⋯	⋯	⋯	⋯	⋯	⋯	319.3 ± 1.9	11519 ± 66	421.97	8.84	2	−1.46	⋯
28	⋯	⋯	⋯	⋯	⋯	⋯	⋯	53.7 ± 0.3	5469 ± 110	18.83	6.93	2	−0.42	⋯
33	⋯	⋯	⋯	⋯	⋯	⋯	⋯	110.2 ± 1.2	2010 ± 26	35.70	6.02	1	0.66	⋯
36	43.43	44.37	5446	133.99	7.82	−1.63	1	⋯		⋯	⋯	⋯	⋯	⋯
39	44.74	45.58	5205	126.00	8.64	−1.24	1	⋯		⋯	⋯	⋯	⋯	${8.46}_{0.08}^{+0.09}$
43	42.83	43.85	4214	57.64	7.21	−1.54	2	491.4 ± 4.2	3536 ± 1881	225.14	6.90	2	−0.93	⋯

Notes.

^aSwift-BAT 70-month hard X-ray survey ID (http://swift.gsfc.nasa.gov/results/bs70mon/). ^b ${L}_{\mathrm{bol}}$ is estimated from L₅₁₀₀, following Trakhtenbrot & Netzer (2012). ^cFollowing Trakhtenbrot & Netzer (2012), with ${L}_{\mathrm{bol}}$ estimated from L₅₁₀₀. ^dSpectral fitting quality flag for broad Hβ: 1—a good fit with small error, 2—acceptable fit. ^eBlack hole mass derived from broad Hα following Greene & Ho (2005b). ^fSpectral fitting quality flag for broad Hα: 1—a good fit with small error, 2—acceptable fit. ^gEddington ratio derived from the Swift-BAT survey (14–195 keV) and ${M}_{\mathrm{BH}}$ following Greene & Ho (2005b). ^hBlack hole mass from Bentz & Katz (2015, http://www.astro.gsu.edu/AGNmass/).

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

For 6DF spectra, the survey applied a single calibration to all spectra to convert measured counts to the correct spectral shape, but a nightly flux calibration was not applied. For 6DF emission-line measurements, we have calculated a rough flux calibration factor for 6DF spectra based on 12 overlapping spectra in distant AGNs ($z\gt 0.05$) with the SDSS. This factor assumes 1 ct is equal to $7.69\times {10}^{-17}\,\mathrm{erg}\,{{\rm{s}}}^{-1}\,{\mathrm{cm}}^{-2}\,{\mathrm{AA}}^{-1}$.

In each of the three spectral fitting regions, we adopt a power-law fit (1st order) to account for the (local) AGN continuum and a series of Gaussian components to model the emission lines. We define an emission-line detection when we reach a ${\rm{S}}/{\rm{N}}\gt 3$ over the line with respect to the noise of the adjacent continuum, and otherwise list upper detection limits. To estimate errors in line fluxes and widths (in terms of FWHM), we use a simple re-sampling procedure that adds noise based on the error spectrum and reruns the fitting procedure 10 times. The (fractional) flux uncertainty for the [O iii] emission line is typically less than 1%. We use the narrow Balmer line ratio (Hα/Hβ) to correct for dust extinction, assuming an intrinsic ratio of ${R}_{{\rm{V}}}=3.1$ and the Cardelli et al. (1989) reddening curve. In the case of a Hβ non-detection, we assume the 3σ upper limits for the extinction correction. When neither Hβ nor Hα are detected we present the fluxes as measured.

For the Hβ spectral region, we fit the He ii $\lambda 4686$, Hβ, [O iii] $\lambda 4959$, and [O iii] λ5007 lines. The widths of the narrow lines are tied with an allowed variation of ±500 km s⁻¹. The central wavelength of the NLR is defined by a joint fit of all the narrow lines where the wavelength separation of all lines is tied. The intensity of [O iii] $\lambda 4959$ relative to [O iii] λ5007 is fixed at the theoretical value of 2.98 (Storey & Zeippen 2000) and the intensity of [N ii] $\lambda 6548$ relative to [N ii] λ6583 is set to the theoretical value of 2.96 (Acker et al. 1989). For the Hβ complex, we use the 4660–4750 Å (except around He ii) and 5040–5200 Å regions for continuum determination. Within the Hα spectral region, we fit the [O i]$\,\lambda 6300$, [N ii] $\lambda 6548$, Hα, [N ii] λ6583, and [S ii] $\lambda 6716$, and [S ii] $\lambda 6731$ lines. Here too, the widths of the narrow lines are tied, with an allowed variation of ±500 km s⁻¹, and the systemic redshift is determined from all narrow lines. In the case of a non-detection of the narrow Hα or [S ii] λ6717,6731 line (i.e., due to a very strong broad Hα component or weak narrow emission lines), we use the FWHM of [O iii] λ5007 to constrain the widths of the narrow lines in the Hα region. The relative strengths of [N ii] $\lambda 6548$ and [N ii] λ6583 lines are fixed at 1:2.94. To estimate the continuum for the Hα complex, we use the wavelength regions 5800–6250 Å and 6750–7000 Å. Finally, within the [O ii] spectral region, we fit the [Ne v] $\lambda 3346$, [Ne v] $\lambda 3426$, [O ii] $\lambda 3727$, [Ne iii] $\lambda 3869$, and [Ne iii] $\lambda 3968$ lines. The continuum around the [O ii] spectral region is usually more complicated to fit due to a nonlinear shape or because it lies near the blue edge of the wavelength coverage. To fit this blue continuum, we use the region between 3300 and 4000 Å, except for small regions surrounding the emission lines themselves (±1000 km s⁻¹).

For sources with broad Hβ, we use the fitting procedure described in detail in Trakhtenbrot & Netzer (2012; see Appendix C1 therein), and here we provide only a brief description of the key spectral components. The AGN spectrum is first fitted with a linear (pseudo-) continuum, based on two narrow (±10 Å) continuum bands, typically around 4440 Å (or 4720 Å) and 5110 Å. Next, a broadened and shifted iron emission template (Boroson & Green 1992) is fitted to and subtracted from the continuum-free spectrum. Then, we fit the remaining emission lines with a set of Gaussian profiles. In particular, the narrow components of Hβ, [O iii] $\lambda 4959$, and [O iii] $\lambda 5007$ are fitted with a single Gaussian profile, while the broad components of He ii and Hβ are described by two Gaussian components (each). As described in Trakhtenbrot & Netzer (2012), the widths of the narrow components are tied among different emission lines, primarily to allow a robust decomposition of the narrow and broad components that make up the Hβ emission-line profile. The FWHM of the broad Hβ line is measured from the (reconstructed) best-fit model of the broad component.

For sources with broad Hα, we use a fitting procedure that involves several progressively complicated steps, depending on the complexity of the emission lines. We start with allowing two Gaussian components for the Hα emission lines: a narrow component (with $\mathrm{FWHM}\lt 1000\,\mathrm{km}\,{{\rm{s}}}^{-1}$) and a broad component ($\mathrm{FWHM}\gt 1000\,\mathrm{km}\,{{\rm{s}}}^{-1}$), both with the wavelength centered at the Hα line. A visual inspection is made (M.K., B.T., K.O., and I.L.) to determine whether a more complex fit is required, this time using multiple Gaussians that are also allowed to be shifted. If the fit quality is still poor, we use the width of the [O iii] line (from the Hβ complex fitting procedure) as an additional constraint on the narrow components in the Hα spectral region. Examples of emission-line fits are given in the Appendix.

For sources with broad Balmer lines, we estimated the black hole masses (${M}_{\mathrm{BH}}$) through virial, "single-epoch" prescriptions, which are in turn based on the ${R}_{\mathrm{BLR}}$–L relation obtained through reverberation mapping of low-redshift AGNs (e.g., Kaspi et al. 2000; Bentz et al. 2006). The virial mass estimators we used for broad Hβ are known to suffer from systematic uncertainties of about 0.3 dex (see, e.g., Shen 2013; Peterson 2014, and references therein).

For sources with broad Hβ we used the same prescription used in Trakhtenbrot & Netzer (2012), which uses the continuum and line-emission parameters for virial estimates of ${M}_{\mathrm{BH}}$:

$$\begin{eqnarray}{M}_{\mathrm{BH}}({\rm{H}}\beta ) & = & 1.05\times {10}^{8}{\left(\displaystyle \frac{{L}_{5100}}{{10}^{46}\mathrm{erg}{{\rm{s}}}^{-1}}\right)}^{0.65}\\ & & \times {\left[\displaystyle \frac{\mathrm{FWHM}({\rm{H}}\beta )}{{10}^{3}\mathrm{km}{{\rm{s}}}^{-1}}\right]}^{2}\,{M}_{\odot },\end{eqnarray} \tag{ 2 }$$

where L₅₁₀₀ is the the monochromatic luminosity at rest-frame 5100 Å, $\lambda {L}_{\lambda }$(5100 Å), measured from the best-fit model of the Hβ region. As mentioned above, FWHM(Hβ) is measured from the entire (best-fit) broad profile. Although the fitting procedure is executed automatically for the large data set studied here, we note that we visually inspected all the Hβ fitting results (including more than one spectrum per source), and applied minor manual adjustments to provide satisfactory fits to the data. We also stress that we did not apply the Hβ-fitting code to spectra with poor absolute flux calibration (i.e., those from the 6DF survey). A summary of these fits is found in Figure 3.

Zoom In Zoom Out Reset image size

For sources with broad Hα lines, we used the prescription of Greene & Ho (2005b):

$$\begin{eqnarray}{M}_{\mathrm{BH}}({\rm{H}}\alpha ) & = & 1.3\times {10}^{6}{\left(\displaystyle \frac{{L}_{{\rm{H}}\alpha }}{{10}^{42}\mathrm{erg}{{\rm{s}}}^{-1}}\right)}^{0.57}\\ & & \times {\left[\displaystyle \frac{\mathrm{FWHM}({\rm{H}}\alpha )}{{10}^{3}\mathrm{km}{{\rm{s}}}^{-1}}\right]}^{2.06}\,{M}_{\odot },\end{eqnarray} \tag{ 3 }$$

where L_Hα is the integrated luminosity of the broad component of the Hα line, determined from the best-fitting model. This prescription is therefore mostly unaffected by host light. Since the Hα-related prescription (Equation (3)) is based on a secondary calibration of a ${R}_{\mathrm{BLR}}$–$L({\rm{H}}\alpha )$ relation, it carries somewhat larger systematic uncertainties (compared with the Hβ-based one). However, it can be applied to Seyfert 1.9 AGNs without broad Hβ lines and it may perform better for sources that have high levels of stellar contamination and/or extinction. A summary of the results of the fitting is provided in Figure 4. We found that about a quarter of Seyfert 1.9 (27%, 31/116) have weak broad Hα lines (EW < 50 Å; see the general discussion of Seyfert sub-classes in Section 4.2). More details on these objects are given in the Appendix.

Zoom In Zoom Out Reset image size

For 19 high-redshift sources (z ∼ 1–3.3), the available optical spectra include either the Mg ii λ2798 or C iv $\lambda 1549$ broad emission lines. The emission complexes around these two lines were fitted using dedicated procedures, described in detail in Trakhtenbrot & Netzer (2012, see Appendices C2 and C3 therein). These take into account the (blended) emission features from iron and He ii $\lambda 1640$ transitions (in the case of Mg ii and C iv, respectively). For both broad lines, each of the doublet features is modeled with two broad Gaussians. We assume no narrow-line contribution to these transitions (see Trakhtenbrot & Netzer 2012, and references therein).

We used the best-fit models of the broad emission lines, together with the adjacent continuum luminosities, to estimate ${M}_{\mathrm{BH}}$ in these 19 high-redshift sources. For the Mg ii line, we used the prescription presented by Trakhtenbrot & Netzer (2012), which is calibrated against Hβ-based mass estimates using a larger sample of SDSS quasars for which both lines are available. An identical prescription was also independently derived by Shen et al. (2011). For the C iv line, we used the prescription presented by Vestergaard & Peterson (2006). We note that C iv-based estimates of ${M}_{\mathrm{BH}}$ are known to be considerably less reliable (e.g., Denney 2012) than those based on lower-ionization transitions, perhaps due to significant contributions from non-virialized BLR gas motion to the emission-line profile (see detailed discussion in Trakhtenbrot & Netzer 2012, and references therein). We therefore advise that the C iv-based determinations of ${M}_{\mathrm{BH}}$ provided here may carry large uncertainties and possibly systematic biases. At such high redshifts, however, they provide the only estimate of ${M}_{\mathrm{BH}}$, in lieu of NIR spectroscopy of the other emission lines.

The spectral and derived parameters of the 19 high-redshift sources are listed in Table 10. We finally note that the Swift-BAT sample of AGNs probably includes several other high-redshift (and therefore high-luminosity) sources, for which an optical spectrum is not available within the large optical surveys we use, and/or the effects of beaming are not yet well understood. We plan to address this population in a separate publication.

Table 10.  Properties Derived from Spectral Decomposition of High-redshift Sources

IDa	Counterpart Name	Redshift	Lineb	$\mathrm{log}{L}_{\mathrm{cont}.}$ c	${F}_{\mathrm{line}}$ d	${\mathrm{EW}}_{\mathrm{line}}$	${\mathrm{FWHM}}_{\mathrm{line}}$	$\mathrm{log}{M}_{\mathrm{BH}}/{M}_{\odot }$	$\mathrm{log}L/{L}_{\mathrm{Edd}}$	Flage
				(erg s−1)	(erg cm−2 s−1)	(Å)	(km s−1)
120	[HB89]0212+735	2.367	C iv	46.26	6.02	21.13	8207	9.69	−0.89	2
188	4C+32.14	1.258	Mg ii	46.26	12.19	13.56	2790	9.04	−0.46	1
311	[HB89]0552+398	2.365	C iv	46.04	2.46	15.11	3134	8.74	−0.16	1
387	B2 0743+25	2.994	C iv	46.24	3.66	26.81	5689	9.36	−0.58	2
428	[HB89]0836+710	2.172	C iv	46.75	15.36	16.18	7754	9.90	−0.61	1
445	1RXS J090915.6+035453	3.288	C iv	46.30	3.36	25.64	6000	9.43	−0.60	1
545	PKS1127−14	1.184	Mg ii	45.65	3.74	18.26	3208	8.78	−0.79	1
601	B2 1210+33	2.504	C iv	46.30	9.44	38.44	4299	9.15	−0.31	2
645	3C279	0.536	C iv	45.11	20.74	27.27	6196	8.83	−1.19	2
693	LBQS 1344+0233	1.310	Mg ii	45.89	6.55	23.24	2326	8.65	−0.43	1
752	3C309.1	0.905	Mg ii	45.27	9.15	52.32	3374	8.59	−0.93	2

Notes.

^aSwift-BAT 70-month hard X-ray survey ID (http://swift.gsfc.nasa.gov/results/bs70mon/). ^bBroad emission line (C iv $\lambda 1549$ or Mg ii λ2798) fitted to measure spectral quantities listed in this table. ^cMonochromatic luminosity at rest-frame 1450 Å (C iv $\lambda 1549$) or 3000 Å (Mg ii λ2798). ^dEmission-line flux ($\times {10}^{-15}$). ^eSpectral fitting quality flag: 1—a good fit with small error, 2—acceptable fit.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

We estimated the bolometric luminosity of the AGNs in our sample (${L}_{\mathrm{bol}}$) from the observed X-ray luminosity measured by the Swift-BAT  survey in the energy range 14–195 keV. First, we divided the 14–195 keV luminosity by 2.67 to convert to the intrinsic 2–10 keV luminosity, following Rigby et al. (2009), which is based on scaling the Marconi et al. (2004) templates to higher X-ray energies. We then used the median bolometric correction from Vasudevan et al. (2009) of the BAT sample, which resulted in a factor of 8 difference between ${L}_{\mathrm{bol}}$ and ${L}_{14\mbox{--}195\mathrm{keV}}$. More advanced luminosity-dependent bolometric corrections will be examined in future studies.

We begin by showing a plot of the X-ray luminosities in the entire Swift-BAT AGN sample, using our new redshift measurements or those from NED when the spectra are not available (Figure 5). We also show several other, deep X-ray surveys, for comparison. The majority (≈90%) of BAT-detected AGNs are nearby ($z\lt 0.2$). Their X-ray luminosities are similar to AGNs found in deeper surveys because of the larger survey area. Our survey finds 46 new redshifts for sources without measurements in NED. This leads to a redshift completeness of 96% (803/836) for the 70-month BAT AGN catalog. The newly measured sources are at similar redshifts (median $z=0.049$) to the sources in the rest of the sample (median $z=0.038$). A summary of the spectroscopic coverage is in Figure 6.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Only three sources with redshifts measured in BASS show significant differences from those in NED. QSO B0347-121 is listed at $z=0.18$ in NED; however, our measurements of the 6DF spectrum clearly show emission lines at $z=0.032$, in agreement with the redshift tabulated in the 6DF catalog. ESO 509-IG 066 NED02 is listed in NED as $z=0.0446;$ however, our measurements of the 6DF spectrum clearly show emission lines at $z=0.034$, again in agreement with the 6DF catalog. 1RXS J090915.6+035453 is listed as $z=3.20$ in NED, but the lines are clearly redshifted to $z=3.28$ based on our measurements, and in agreement with the SDSS catalog measurement.

We first classify the BAT AGNs depending on the presence and strength of broad emission lines (e.g., Osterbrock 1981). A Seyfert 1.9 classification is a source with a narrow Hβ line and broad Hα line. We use the quantitative classifications for Seyfert sub-classes (1, 1.2, 1.5, and 1.8) based on Winkler (1992), using the total flux of [O iii] and Hβ. A summary of the results of this classification can be found in Figure 7. About half of the sources are Seyfert 2 or Seyfert 1.9, and about a quarter are Seyfert 1.2 and 1.5. A small fraction are true Seyfert 1 (7%) and only two are Seyfert 1.8 sources.

Zoom In Zoom Out Reset image size

We compare our Seyfert types (Sy 1, Sy 1.2, Sy 1.5, Sy 1.8, Sy 1.9, and Sy 2) to the most recent 13th edition Veron-Cetty catalog of AGNs (Veron-Cetty & Veron 2010). Only a minority (36%, 230/642) of the BASS sample is classified in this catalog by Seyfert type. Using this subsample of 230 we find that the majority, 89% (206/230), agree with the Seyfert type classification with the Veron-Cetty catalog. These include 77% (177/230) that show exact type agreement or are listed as a unspecified Sy 1 in the Veron-Cetty catalog and found to be Sy 1, Sy 1.2, Sy 1.5, or Sy 1.9 in BASS. Another 13% (29/230) are Sy 1, but are listed as Sy 1.2 or Sy 1.5 and vice versa in BASS.

The remaining 10% (24/230) show disagreement among Seyfert type between BASS and Veron-Cetty. The majority of these are listed as Sy 1.9 in our sample but are found to be Sy 2 in Veron-Cetty 67% (16/24), most likely because our of higher-quality spectra. There are no examples of Sy 1.9 in the Veron-Cetty catalog that are found to be Sy 2 in our catalog.

Another six sources are Sy 1 unspecified in Veron-Cetty, but Sy 2 in our catalog. For PKS 0326-288, the Veron-Cetty reference (Mahony et al. 2011) lists the source as a narrow-line source in agreement with our classification. MCG +02-21-013 has no references in the Veron-Cetty catalog for the Seyfert type, but NED lists both a Sy 1 and Sy 2 in agreement with our classification. NGC 4992 has no references in the Veron-Cetty catalog for the Seyfert type, but is classified as a Sy 2 in a recent paper by Smith et al. (2014) in agreement with our classification. A recent paper on MCG+04-48-002 (Koss et al. 2016b) lists it as a Sy 2, in agreement with our classification, while it has no references in the Veron-Cetty catalog for the Seyfert type. NGC 5231 is listed as a Sy 2 to match our classification in recent work (Parisi et al. 2012). Finally, 2MASX J10084862–0954510 is shown to be an Sy 1 in Bauer et al. (2000), with broad Balmer lines, which is very different from our spectra and may be a case of variability.

Finally, two sources are listed as Sy 1.0 in Veron-Cetty, but are listed as Sy 1.9 in the BASS catalog. The reference for ESO 198-024 as a Sy 1.0 in Veron-Cetty is based on imaging variability (Winkler et al. 1992) in the optical rather than spectroscopy, and could be consistent with our Sy 1.9 classification. Finally, the reference in Veron-Cetty for a Sy 1.0 classification (Guainazzi et al. 2000) lists the source as a Sy 1.9, which is the same as our classification, though the authors mention the source may have changed from a narrow-line AGN to a Sy 1.9.

In summary, we find broad agreement (89%) with past studies of Seyfert type classification from the Veron-Cetty catalog. The major difference is that 16 sources in the BASS catalog are now classified as Sy 1.9 because of broad lines detected in Hα for the first time. Finally, eight sources show a different Seyfert type, possibly due to variability, which will be further studied in future publications.

In Figure 8 we present the fraction of sources with broad Hα and/or Hβ lines, plotted against X-ray luminosity. For both the 14–195 keV and 2–10 keV luminosities we find a general increase in type 1 fraction with increasing luminosity, as has been found in past studies using broad Hβ (see, e.g., Merloni et al. 2014, and references therein). We find that AGNs with broad Hα lines are consistently more common, by about 10%–20%, than AGNs with Hβ lines, across a wide range of X-ray luminosity.

Zoom In Zoom Out Reset image size

Finally, in Figure 9 we plot the FWHM of Hα as a function of the column density derived from the X-rays (${N}_{{\rm{H}}}$). We note that roughly half (57%, 128/223) of Seyfert 1–1.8 broad-line AGNs with column density measurements have only upper limits on ${N}_{{\rm{H}}}$ at 10²⁰ cm⁻², corresponding to being unobscured.

Zoom In Zoom Out Reset image size

The FWHMs of the emission lines show broad agreement with the X-ray obscuration (∼94%), such that Seyfert sub-types 1, 1.2, 1.5, and 1.8 have ${N}_{{\rm{H}}}\lt {10}^{21.9}$ cm⁻², and Seyfert 2 have ${N}_{{\rm{H}}}\gt {10}^{21.9}$ cm⁻². Seyfert 1.9, however, show a range of column densities.

Additionally, a small fraction of Seyfert 2 sources (6%, 14/221) have X-ray obscuration below ${N}_{{\rm{H}}}={10}^{21.9}$ cm⁻². We note, however, that Seyfert 1.9 sources, which have evidence of a broad-line in Hα, but not Hβ, span the full range of column densities from unobscured to Compton-thick (i.e., ${N}_{{\rm{H}}}\lt {10}^{24}$ cm⁻²).

We use the emission-line diagnostics of Veilleux & Osterbrock (1987), revised by Kewley et al. (2006). We classify each AGN using the [O iii] λ5007/Hβ versus [N ii] λ6583/Hα, [S ii] λ6717,6731/Hα, and [O i]$\,\lambda 6300$/Hα diagnostics (Figure 10). For the [N ii] λ6583 diagnostic, we further separate the star-forming (H ii) galaxies and composite galaxies, and separate AGNs into LINERs and Seyferts (following Schawinski et al. 2007b). Finally, we also apply the [O iii] λ5007/[O ii] $\lambda 3727$ and He ii $\,\lambda 4686$/Hβ diagnostics, defined by Shirazi & Brinchmann (2012).

Zoom In Zoom Out Reset image size

We find that roughly half of BAT AGNs are found in the Seyfert region of the [N ii] diagram (53%, 338/642). The next largest sub-group is sources without an Hβ detection (4%, 338/642), though the detection limits imply either a Seyfert or LINER AGN classification. About 15% of sources have weak lines, and despite high-S/N optical spectra, lack enough emission-line measurements for line diagnostic diagrams. The remaining categories of LINERs, composite galaxies, and H ii classifications are rare, with only a few percent of BAT AGNs found in each. A few percent of sources also have complex emission-line profiles where a good fit to the emission lines was not obtained. Finally, about 10% of sources lack sufficient wavelength coverage, because of the instrumental setup or their high redshifts.

The [S ii] λ6717,6731 diagnostic shows a very similar distribution, though there is a slightly lower fraction of Seyferts (50%, 317/642), due to the weaker [S ii] line (and limited S/N of the data), and a larger fraction of H ii regions. For the [O i] diagnostic, the line is markedly weaker than the [N ii] and [S ii] line, and therefore identifies somewhat fewer Seyferts (38%, 242/642), and has about double the number of sources that lack emission-line detection. For sources with line detections in all three diagnostics (35%, 225/642), we find good agreement in the AGN classification across the diagnostics (81%, 182/225).

We also classify the sample using the [O iii] λ5007/[O ii] $\lambda 3727$ versus [O i]$\,\lambda 6300$/Hα and He ii $\,\lambda 4686$/Hβ versus& [N ii] λ6583/Hα diagnostic diagrams (Figure 11). Compared to the more commonly used diagnostics (i.e., [N ii], [S ii], and [O i]), these two diagnostics are not efficient for classifying the majority of AGNs in our sample, because of the difficulty in detecting the He ii line and the lack of blue coverage in most spectra for the [O ii] line.

Zoom In Zoom Out Reset image size

We perform a comparison of the demographics of the BASS X-ray-selected AGNs to optically selected Seyferts in the SDSS, based on the OSSY catalog (Oh et al. 2011, 2015). The results are shown in Figure 12. We find that the BASS X-ray-selected AGNs show a relatively constant type 1 to type 2 fraction of ∼55% over the redshift range of $z\lt 0.1$, while the fraction in SDSS AGNs is much lower and furthermore shows a strong dependence on redshift (2%–30%).²³ The [O iii] luminosities of BAT AGNs are higher, on average, than those of SDSS AGNs, for both Seyfert 1 and Seyfert 2 AGNs. The BASS narrow-line AGNs show a larger number of sources with high Balmer decrements (Hα/Hβ > 5), compared to SDSS AGNs. This can be clearly understood by the requirement to have robust detections of all relevant emission lines for SDSS AGNs to be classified as such, which the hard X-ray selection of BASS AGNs overcomes. Finally, the average stellar velocity dispersions of the BAT narrow-line AGNs ($191\pm 17\,\mathrm{km}\,{{\rm{s}}}^{-1}$) are significantly higher than those of narrow-line SDSS AGNs ($101\pm 10\,\mathrm{km}\,{{\rm{s}}}^{-1}$), and show a stronger redshift dependence.

Zoom In Zoom Out Reset image size

Using the black hole mass estimated from velocity dispersion and broad Balmer lines, we find that the black hole masses of our BASS AGNs range between ${10}^{5.4}\,\mathrm{and}\,{10}^{10}\,{M}_{\odot }$. Figure 13 shows the distribution of ${M}_{\mathrm{BH}}$ in different redshift ranges and for Seyfert 1–1.9, Seyfert 2, LINERs, and beamed AGNs.

Zoom In Zoom Out Reset image size

We use the median and median absolute deviation (MAD) to compare the populations because of the spread over several orders of magnitude. The median and MAD are ${M}_{\mathrm{BH}}\,=(5.8\pm 8.5)\times {10}^{7}\,{M}_{\odot }$ for $z\lt 0.01$, ${M}_{\mathrm{BH}}=(6.4\pm 8.1)\,\times {10}^{7}\,{M}_{\odot }$ for $0.01\lt z\lt 0.04$, and ${M}_{\mathrm{BH}}=(1.2\pm 1.5)\,\times {10}^{8}\,{M}_{\odot }$ for $0.04\lt z\lt 0.85$, and ${M}_{\mathrm{BH}}=(7.4\pm 6.4)\,\times {10}^{8}\,{M}_{\odot }$ for $z\gt 0.85$. An Anderson–Darling test indicates the distributions of black hole masses at $z\lt 0.01$ and $0.01\lt z\lt 0.04$ consistent with being drawn from the same population, but those at $0.04\lt z\lt 0.85$ and $z\gt 0.85$ are drawn from the same population at less than the 1% level, consistent with their higher median black hole masses.

The higher median black hole masses found for high-redshift AGNs are likely a selection effect driven by the fixed survey flux limit.

The median and MAD values are ${M}_{\mathrm{BH}}=(1.9\pm 2.6)\,\times {10}^{8}\,{M}_{\odot }$ for Seyfert 1–1.9; ${M}_{\mathrm{BH}}=(1.9\pm 2.1)\times {10}^{8}\,{M}_{\odot }$ for Seyfert 2; ${M}_{\mathrm{BH}}=(1.7\pm 2.1)\times {10}^{8}\,{M}_{\odot }$ for LINERs; and ${M}_{\mathrm{BH}}=(6.0\pm 6.0)\times {10}^{8}\,{M}_{\odot }$ for beamed AGNs. An Anderson–Darling test indicates the distributions of black hole masses of Seyfert 1–1.9, Seyfert 2, and LINERS are consistent with being drawn from the same population, but the likelihood that beamed AGNs are drawn from the same population is less than the 1% level.

While we do not find any significant difference between the black hole mass distributions of Seyfert 1–1.9 and Seyfert 2, we note that our survey has systematic biases against the smaller black holes (${M}_{\mathrm{BH}}\lt {10}^{7}\,{M}_{\odot }$) in Seyfert 2 AGNs. Specifically, the velocity dispersion measurements for Seyfert 2 are limited by the instrumental broadening in lower spectral resolution setups. Typical instrumental resolutions are between 2 and 7 Å FWHM, (corresponding to limiting black hole masses of ${M}_{\mathrm{BH}}={10}^{6}\mbox{--}{10}^{8}\,{M}_{\odot }$). We have recently been granted two filler programs with VLT/XSHOOTER (K. Oh et al. 2017, in preparation) that will further address this issue, as the spectral resolution would be sufficient to measure limiting black hole masses of ${M}_{\mathrm{BH}}={10}^{5}\,{M}_{\odot }$. A final issue is that for galaxies with a significant rotation component, ${\sigma }_{* }$ measured from a single aperture spectrum can vary by up to ∼20%, depending on the size of the adopted extraction aperture (Kang et al. 2013).

We combined the ${M}_{\mathrm{BH}}$ estimates with the estimates of ${L}_{\mathrm{bol}}$, derived from the BAT X-ray luminosity, to calculate the Eddington ratios of the BASS AGNs, $L/{L}_{\mathrm{Edd}}$ erg s⁻¹(where ${L}_{\mathrm{Edd}}\equiv 1.3\times {10}^{38}\,({M}_{\mathrm{BH}}/{M}_{\odot })$). The maximum value of the bolometric luminosity of our sample is ${L}_{\mathrm{bol}}={10}^{48.5}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$. The AGNs with higher ${L}_{\mathrm{bol}}$ have, in general, higher $L/{L}_{\mathrm{Edd}}$, but there are also some AGNs with relatively high bolometric luminosity (${L}_{\mathrm{bol}}\gt {10}^{45}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$) and low Eddington ratio ($L/{L}_{\mathrm{Edd}}\lt 0.01$). The sources with the highest bolometric luminosity (${L}_{\mathrm{bol}}\gt {10}^{47}\,\mathrm{erg}\,{{\rm{s}}}^{-1}$), however, do have high accretion rates ($L/{L}_{\mathrm{Edd}}\gt 0.1$). Conversely, several of the most massive BHs in unbeamed AGN in our sample (${M}_{\mathrm{BH}}\gt {10}^{9.7}\,{M}_{\odot }$) also have the lowest accretion rates. Regarding the redshift distributions, the median and MAD Eddington ratios are $L/{L}_{\mathrm{Edd}}=0.005\pm 0.008$ for $z\lt 0.01$, $L/{L}_{\mathrm{Edd}}=0.026\pm 0.036$ for $0.01\lt z\lt 0.04$, $L/{L}_{\mathrm{Edd}}\,=0.047\pm 0.062$ for $0.04\lt z\lt 0.85$, and $L/{L}_{\mathrm{Edd}}=1.46\,\pm 2.16$ for $z\gt 0.85$. An Anderson–Darling test indicates each of the redshift distributions of Eddington ratios are each drawn from the same population at the less than 1% level, consistent with their steadily increasing medians. These properties of our sample are not surprising, given the flux limited (and low-redshift) nature of our sample.

Regarding the Eddington ratios among different Seyfert types, we do find that Seyfert 2 AGNs have, in general, lower Eddington ratios because Seyfert 1 AGNs have higher bolometric luminosities. The peak of the distribution for Seyfert 2 is at $L/{L}_{\mathrm{Edd}}\simeq 0.01$. For the Seyfert 1 AGNs, the peak of the distribution is between $L/{L}_{\mathrm{Edd}}=0.01$ and 0.1, with a small number (13) of unbeamed sources above the Eddington limit, $L/{L}_{\mathrm{Edd}}\gtrsim 1$. The median and MAD are $L/{L}_{\mathrm{Edd}}=0.10\pm 0.09$ for Seyfert 1–1.9, $L/{L}_{\mathrm{Edd}}=0.014\,\pm 0.014$ for Seyfert 2, and $L/{L}_{\mathrm{Edd}}=0.009\pm 0.012$ for LINERs, and $L/{L}_{\mathrm{Edd}}=1.46\pm 2.16$ for beamed AGNs.

We note that the Eddington ratios of beamed AGNs are derived from the observed luminosities (and masses), and have not been corrected for beaming angle. An Anderson–Darling test indicates that the distributions of Eddington ratios for Seyfert 2 AGNs and LINERs are consistent with being drawn from the same population. However, the likelihood that the Eddington ratios of type 1 and beamed AGNs are drawn from the same population is less than 1%, consistent with their much larger medians. It is interesting to note that using the observed Eddington ratios is highly efficient at separating beamed AGNs from unbeamed sources because typical beamed AGNs are above the Eddington limit.

We also found sources with extremely low accretion rates ($L/{L}_{\mathrm{Edd}}\lt 0.001$). There are 15 type 2 AGNs and 13 type 1 AGNs with Eddington ratios $L/{L}_{\mathrm{Edd}}\lt 0.001$. In general, type 1 AGNs are found to have high accretion rates, $L/{L}_{\mathrm{Edd}}\gt 0.01$ (e.g., Nicastro 2000; Yuan & Narayan 2004; Trump et al. 2011; Elitzur et al. 2014), and therefore it is quite surprising to find type 1 sources with such a low value of the Eddington ratio.

As we studied the properties of our sample in the optical and X-rays, a number of objects showed interesting and unusual characteristics. Some examples of these objects are presented in Figure 14. We discuss here four different types: AGNs with very low X-ray column density, but lacking broad emission lines, or vice versa; double-broad-line AGNs; and weak-line AGNs. We note that because the X-ray and optical spectroscopy are not simultaneous, the contradictory optical and X-ray classification could be caused by variability.

Zoom In Zoom Out Reset image size

4.6.1. AGN with Contradictory Optical and X-Ray Classification

4.6.2. Double-broad-line AGNs

4.6.3. Weak-line AGNs

Examples of the velocity dispersion fits from different telescopes can be found in Figures 15 and 16. We first compared our results for the spectra from SDSS with the values given from the SDSS 12th data release, measured using a direct fitting method rather than pPxf (Figure 17). The values calculated using pPxf are in good agreement, with maximal differences always less than 50 $\mathrm{km}\,{{\rm{s}}}^{-1}$.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We also compared velocity dispersion ${\sigma }_{* }$, using pPxf, to literature values obtained with other telescopes. We used the data from the Hyperleda catalog (Paturel et al. 2003) and literature references (Cid Fernandes et al. 2004; Garcia-Rissmann et al. 2005; Ho et al. 2009; van den Bosch et al. 2015). For 41 galaxies in our sample we found a value of ${\sigma }_{* }$ in the literature. Figure 18 shows the comparison plot. Most (31/41, 76%) of the obtained values are in agreement with the literature values, with the differences between our values and the literature at <50 $\mathrm{km}\,{{\rm{s}}}^{-1}$. For 10/41 (24%) of sources, the difference is between 50 and 100 $\mathrm{km}\,{{\rm{s}}}^{-1}$ and none are above this value.

Zoom In Zoom Out Reset image size

We used Monte Carlo (MC) simulations to check the velocity dispersion ${\sigma }_{* }$ measured by pPxf and any possible systematics with larger noise. We considered five spectra from different telescopes (SDSS, KPNO, 6dF) with different values for the pPxferrors (10, 20, 30 $\mathrm{km}\,{{\rm{s}}}^{-1}$). Then, we added random noise to the spectrum to increase the values of the pPxferror until an established level (20, 30, 40, 50 $\mathrm{km}\,{{\rm{s}}}^{-1}$). For every level of noise, we ran 100 MC simulations and we calculated the mean of the results obtained for ${\sigma }_{* }$. The difference between the original value ${\sigma }_{* ,\mathrm{ppxf}}$ and the mean value from the MC simulations ${\sigma }_{* ,\mathrm{MC}}$ is always smaller than 20 $\mathrm{km}\,{{\rm{s}}}^{-1}$ (Figure 19).

Zoom In Zoom Out Reset image size

We provide several examples of our emission-line fits using different telescopes. A separate BASS study found that the physical slit size, telescope sample, or X-ray obscuration level was not a significant contributor to the scatter in emission-line measurements compared to the X-ray emission (Berney et al. 2015) for sources with $0.01\lt z\lt 0.4$, where the size of the slit is typically kiloparsec scales. With some setups we were able to cover the blue [Ne v]$\,\lambda 3426$, [O ii] $\lambda 3727$, and [Ne iii]$\,\lambda 3968$ region (e.g., Figure 20) in addition to the Hβ and Hα regions (e.g., Figure 21). We tested our emission-line measurements with a number of literature values to confirm their accuracy. Figure 22 shows the comparison between values from the OSSY catalog (Oh et al. 2011) and our measured values for the [O iii] emission-line flux derived from the same SDSS spectra. We obtain a standard deviation of σ = 0.051 dex and a median offset of 0.014 dex for flux values over three orders of magnitude.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We also tested how the inclusion of the effect of fitting empirical models with absorption lines that we applied to all narrow-line sources changes the emission-line measurements. Due to the absorption features on Hα and Hβ, the obtained values for the two Balmer lines are often underevaluated. The effect is more pronounced for the Hβ emission line. The result on the line diagnostic diagram is a strong shift to the bottom of the diagram and a relatively small shift to the left, which is toward the H ii and Composite regions (See Figure 23). We note that in this study this fitting correction for absorption has only been done for narrow-line sources, where the empirical model fits are not biased by AGN light.

Zoom In Zoom Out Reset image size

We compared our black hole mass measurements (Figure 24) using broad-line measurements of the Hβ or Hα regions in single-epoch observations. We also compare the mass measurements to higher-quality reverberation mapping measurements for the small sample where these are available. We find that the slope is consistent with unity, with fairly large scatter (≈0.4–0.5 dex).

Zoom In Zoom Out Reset image size

We also made a comparison of Seyfert 1.9 black hole masses, measured with very weak broad Hα lines, to those with no broad Hβ detected (Figure 25). We find that below equivalent widths (EWs) of 50 Å in Hα, the velocity dispersion differs significantly, suggesting these broad features could be spurious; thus we have used this limit for flag 2 sources in our sample. The value of this limit is consistent with many past quasar studies (e.g., EW < 45 Å; Shen et al. 2011), though some studies of nearby AGNs have used much weaker broad lines (e.g., EW < 15 Å; Greene & Ho 2007).

Zoom In Zoom Out Reset image size

The offset in black hole mass measurements between Seyfert 1.9 using velocity dispersion measurements and broad-line ${\rm{H}}\alpha$ is concerning, though offsets have been found in AGN samples (e.g., Shankar et al. 2017). We note, however, that these two methods are tied to reproduce similar masses for systems where both are applicable (e.g., Graham et al. 2011; Woo et al. 2013), so it is likely that the subsample of Seyfert 1.9 explored here is not representative. One particular concern is that the weak broad-line ${\rm{H}}\alpha$ may suffer high extinction and is underestimated, which we are exploring in a current VLT/XSHOOTER program using the NIR broad Paschen emission lines (K. Oh et al. 2017, in preparation).

For consistency, we compare our values of the bolometric luminosity L_bol, derived from the X-ray luminosity, with the values of ${L}_{\mathrm{bol}}$ obtained from the optical luminosity at 5100 Å (e.g., Wandel et al. 1999) for Seyfert 1 AGNs. In Figure 26, comparisons between the L_bol obtained from the X-ray bolometric corrections and the L_bol derived from the optical luminosity are shown. The two methods to infer L_bol show fairly larger scatter (0.46 dex), with more scatter at high luminosities.

Zoom In Zoom Out Reset image size