Fundamental Reference AGN Monitoring Experiment (FRAMEx). I. Jumping Out of the Plane with the VLBA

Our AGN sample was selected from the Swift BAT 105-month catalog (Oh et al. 2018). Because this catalog has a uniform flux limit (∼8 × 10⁻¹² erg cm⁻² s⁻¹) across the sky, and hard X-rays (14–195 keV) reliably pick out bona fide AGNs, even in the presence of heavy absorption (${N}_{{\rm{H}}}\lesssim {10}^{24}$ cm⁻²), constructing a volume-complete sample of AGNs above some luminosity threshold is straightforward. Additionally, AGNs from the BAT catalog have extensive multiwavelength coverage and have the benefit of intense recent study (e.g., Trakhtenbrot et al. 2017; Powell et al. 2018; Shimizu et al. 2018; Bär et al. 2019).

We choose AGNs above a hard X-ray luminosity of 10⁴² erg s⁻¹, which corresponds to a distance limit of ∼40 Mpc for a volume-complete sample using a flat ΛCDM cosmology with ${H}_{0}=70$ km s⁻¹ Mpc⁻¹ and ${{\rm{\Omega }}}_{{\rm{M}}}=0.3$. This luminosity threshold was chosen to sample both moderate- and high-luminosity AGNs, and to provide a sample of AGNs close enough to give exquisite physical resolution on the AGN. Additionally, heavy Compton-thick obscuration may cause lower luminosity AGNs beyond ∼40 Mpc to be undetected even by BAT (e.g., LaMassa et al. 2019), biasing a sample extending farther than 40 Mpc toward less obscured AGNs. At distances of less than 40 Mpc, the ∼1 mas beam of the VLBA at 6 cm (C band) samples physical scales of less than 0.2 pc within the dusty obscuring medium (i.e., the "torus") that is inferred to exist in most AGNs (e.g., Netzer 2015). Additionally, these physical scales correspond to light-crossing times of only a few months, enabling relatively short-term studies of both luminosity and morphological variability.

Out of the 1105 AGNs in the 105 -month BAT catalog, 43 are within the local volume defined here. For observability with the VLBA and other northern hemisphere facilities, we made an additional decl. cut of $-30^\circ \lt \delta \lt +60^\circ$, leaving 25 targets. We illustrate our selection method in Figure 1 and provide a list of targets in Table 1.

Zoom In Zoom Out Reset image size

Table 1.  FRAMEx Volume-complete Sample

Target	R.A. (ICRS)	Decl. (ICRS)	Type	Redshift	Distance	log(MBH)	log(${L}_{{\rm{H}}\alpha }$)	SFRmax	SNRmax
	(deg)	(deg)			(Mpc)	(M☉)	(erg s−1)	(M⊙ yr−1)	(yr−1)
NGC 1052	40.2699884	−8.25576190	Sy2	0.0050	21.5	8.67	40.15	0.11	2.2 × 10−3
NGC 1068	40.6696342	−0.01323785	Sy2	0.0038	16.3	6.93	41.34	1.74	3.5 × 10−2
NGC 1320	51.2028681	−3.04226840	Sy2	0.0089	38.4	7.96	40.49	0.24	4.9 × 10−3
NGC 2110	88.0473918	−7.45625094	Sy2	0.0078	33.6	9.38	40.68	0.38	7.5 × 10−3
NGC 2782	138.5212787	+40.11369022	Sy2	0.0085	36.6	6.07	41.51	2.62	5.2 × 10−2
IC 2461	139.9914308	+37.19100007	Sy2	0.0075	32.3	7.27	39.40	0.02	3.9 × 10−4
NGC 2992	146.4247756	−14.32626689	Sy1	0.0077	33.2	8.33	40.11	0.10	2.0 × 10−3
NGC 3081	149.8731005	−22.82631476	Sy2	0.0080	34.5	7.74	40.97	0.74	1.5 × 10−2
NGC 3089	149.9028701	−28.33129443	Sy2?	0.0090	38.8	6.55	⋯	⋯	⋯
NGC 3079	150.4908469	+55.67979744	Sy2	0.0037	15.9	6.38	40.04	0.09	1.7 × 10−3
NGC 3227	155.8774015	+19.86505766	Sy1	0.0039	16.8	6.77	40.78	0.48	9.6 × 10−3
NGC 3786	174.9271391	+31.90942732	Sy2	0.0089	38.4	7.48	40.73	0.42	8.4 × 10−3
NGC 4151	182.6357547	+39.40584860	Sy1	0.0033	14.2	7.55	41.21	1.27	2.5 × 10−2
NGC 4180	183.2626924	+7.03891255	LINER	0.0070	30.1	7.63	⋯	⋯	⋯
NGC 4235	184.2911678	+7.19157597	Sy1	0.0080	34.5	7.55	40.94	0.68	1.4 × 10−2
NGC 4388	186.4449188	+12.66215153	Sy2	0.0084	36.2	6.94	41.57	2.93	5.9 × 10−2
NGC 4593	189.9143400	−5.34417010	Sy1	0.0090	38.8	6.88	40.65	0.35	7.0 × 10−3
NGC 5290	206.3297085	+41.71241871	Sy2	0.0086	37.1	7.78	40.12	0.10	2.1 × 10−3
NGC 5506	213.3119888	−3.20768334	Sy1.9	0.0062	26.7	6.96	41.18	1.19	2.4 × 10−2
NGC 5899	228.7634964	+42.04991289	Sy2	0.0086	37.1	7.69	41.19	1.23	2.5 × 10−2
NGC 6814	295.6690092	−10.32345792	Sy1	0.0052	22.4	7.04	39.35	0.02	3.5 × 10−4
NGC 7314	338.9424567	−26.05043820	Sy1.9	0.0048	20.6	6.76	39.68	0.04	7.5 × 10−4
NGC 7378	341.9486864	−11.81658744	Sy2	0.0086	37.1	4.93	⋯	⋯	⋯
NGC 7465	345.5039963	+15.96477472	Sy2	0.0066	28.4	6.54	40.44	0.22	4.3 × 10−3
NGC 7479	346.2359605	+12.32295297	Sy2	0.0079	34.0	7.61	40.02	0.08	1.7 × 10−3

Download table as:  ASCIITypeset image

Table 1 also presents reddening-corrected nuclear narrow Hα luminosities, maximum expected star formation rates (SFR_max) and maximum expected supernova rate (SNR_max). The Hα luminosities were calculated using nuclear emission line fluxes from Koss et al. (2017), except for NGC 1320 (de Grijp et al. 1992), NGC 2782 (Ho et al. 1997), and NGC 7465 (Moustakas & Kennicutt 2006). These flux measurements were obtained with small apertures (2''–4'') and corrected for reddening using the Calzetti et al. (2000) extinction law. These Hα luminosities are used to determine the maximum expected SNR due to star formation, assuming that all of the narrow Hα is due to star formation, providing an upper limit on any possible contamination of the nuclear AGN radio emission by SNe. We calculated SFR_max for our sample using the calibration from Kennicutt (1998). These values were converted into SNR_max using starburst99 continuous star formation models (Leitherer et al. 1999), assuming a Salpeter (1955) initial mass function and an age higher than 30 Myr where the SNR reaches a plateau. We find SNR_max ranging from 3.5 × 10⁻⁴ to 5.9 × 10⁻² yr⁻¹, which allows us to conclude that the contribution from SNe to the nuclear radio emission is negligible.

We note that the distances we use are calculated from the sample redshifts, and the mean redshift of the sample is 0.007, so the effect of peculiar motions on the cosmological distance estimates may be significant. To estimate this, we retrieved redshift-independent distances for our sample using the NASA Extragalactic Database (NED),⁷ and calculated the amount of additional intrinsic uncertainty that is required for the luminosity using the cosmological distance estimate to be consistent with the luminosity using the redshift-independent distance, in a reduced χ² sense. We found 13 unique methods used to estimate the redshift-independent distances of our sample: ring diameter (Pedreros & Madore 1981), Faber-Jackson and "tertiary" methods (e.g., de Vaucouleurs & Olson 1984), Tully estimate distances (Tully & Fisher 1988), CO ring diameter (Sofue 1991), D-sigma and the Infrared Astronomical Satellite standard candle (Willick et al. 1997), the "fundamental plane" of early-type galaxies (unrelated to the fundamental plane of black hole activity; e.g., Blakeslee et al. 2001) look-alike ("sosie") galaxies (Terry et al. 2002), surface brightness fluctuations (e.g., Tully et al. 2013), AGN time lags (e.g., Hönig et al. 2014), Tully–Fisher (e.g., Tully et al. 2016), and type Ia supernovae (e.g., Koshida et al. 2017). We found the value the intrinsic uncertainty to be 0.31 dex. When the Merloni et al. (2003) FP relation and propagation of uncertainty is used, however, the scatter induced by the absolute luminosity uncertainty is only 0.19 dex. Moreover, this source of uncertainty also affected the sources studied by Merloni et al. (2003, see their Section 3.1), and makes only a minor contribution to the overall dispersion in the FP (see Section 6.1 in Merloni et al. 2003). We therefore use the cosmological distances for our sample, which are more generally available than redshift-independent distances.

In order to estimate the required integration time of our sample, we used the FP given in Merloni et al. (2003). A 10⁶ M_⊙ black hole with a hard X-ray luminosity of 10⁴² erg s⁻¹ would result in a 6 cm (6 GHz) radio peak luminosity of approximately 10³⁶ erg s⁻¹. This yields a flux of ∼100 μJy beam⁻¹ at the maximum distance of ∼40 Mpc for our sample. To ensure that our targets are robustly detected, we designed our observations to achieve a signal-to-noise ratio (S/N) of 5, or a root mean square (rms) noise of ∼20 μJy beam⁻¹. Using the European VLBI Network online calculator,⁸ we determined that one hour of on-source integration time produces a 1σ image thermal noise of 23 μJy beam⁻¹.

To ensure accurate VLBA pointing for all of our targets, which is necessary due to the small effective field of view (a few arcsec) in the C band, we used the most accurate AGN positions of our targets. For 17 of our objects we determined the AGN positions from archival Very Large Array (VLA) A-array configuration observations in the X-band (∼9 GHz), which provided a pointing accuracy of roughly 50 mas. Two of our targets had archival C-band (∼5 GHz) VLA data in the A-array and B-array configurations for an a priori positional accuracy of 80 mas and 370 mas, respectively. The remaining 6 objects did not have any archival VLA data with the required positional accuracy for VLBA pointing, and we therefore used the Pan-STARRS y-band data for 5 galaxies, providing a pointing accuracy of ∼300 mas. For one galaxy we used the Chandra X-ray position, which has an a priori positional accuracy of 500 mas. These AGN positions are well within the desired pointing accuracy for the VLBA and provide high confidence that we imaged the AGN at our subparsec spatial scale, if it exists.

As one of the primary goals of these VLBA snapshot observations is to produce high-fidelity images, we devised an observing strategy to maximize uv sky coverage by dividing the 25 targets into bins of similar R.A. This produced six groups, each containing three to five sources. We observed each target in a given bin for 20 minutes before slewing to the next target in the bin. We continued in this fashion, cycling back through each of the targets three times to reach a full one-hour on-source total integration time. In this manner, we were able to track parallactic angles three to five times longer than if we had observed each source independently, thereby producing a better-sampled uv plane. We used phase referencing in order to accurately constrain the phases and ultimately the positions of each target. Table 2 lists the observing parameters, including the center frequency, restoring beam parameters as measured in the plane of the sky and then rotated by the beam position angle, rms noise of the cleaned image, and the phase calibrator name and position. For each observation, we used a data rate of 2048 Mbps with 2 bit sampling in a single right circular polarization with four intermediate frequency (IF) windows each with a width of 128 MHz and 512 channels. This recording setup permitted a bandwidth smearing limit of $\sim 22^{\prime\prime}$ and a time smearing limit of $\sim 5^{\prime\prime}$. The positions for sources are known to within the error of the positions of the phase reference calibrators, which have all come from the ICRF3 (P. Charlot et al. 2020, in preparation) and have a median formal error of ${\sigma }_{\alpha }=6.4$ μas, ${\sigma }_{\delta }=137.3$ μas.

Table 2.  VLBA Observations

Target	Frequency	Restoring Beam	Beam Angle	rms	Calibrator	R.A.	Decl.
	(GHz)	($\alpha \times \delta ;$ mas)	(deg)	(μJy bm−1)	IERS Name	(deg(μs))	(deg(μas))
NGC 1052	5.873692	3.83 × 1.49	8.4	4242.0	0240−060	40.801956142(4.78)	−5.8486932(137.3)
NGC 1068	5.858862	3.48 × 1.60	14.7	21.7	0237−027	39.939467802(2.09)	−2.57803183(31.9)
NGC 1320	5.867662	7.46 × 2.02	−15.9	39.1	0319−056	50.499459824(8.17)	−5.4367857(222.2)
NGC 2110	5.861240	3.37 × 1.51	5.3	162.0	0551−086	88.42454680(13.99)	−8.6671937(157.8)
NGC 2782	5.867662	2.82 × 2.23	18.1	33.0	0913+391	139.203769060(7.04)	38.9078184(115.6)
IC 2461	5.870108	2.92 × 2.20	17.0	35.3	0922+364	141.466047268(6.73)	36.2099096(118.2)
NGC 2992	5.869068	6.68 × 2.52	−13.7	105.1	0938−133	145.260622829(6.07)	−13.5974958(193.9)
NGC 3081	5.874081	6.80 × 3.23	−7.0	43.7	1004−217	151.693390348(4.80)	−21.9890028(109.5)
NGC 3089	5.873428	7.91 × 3.36	2.9	60.1	1008−285	152.772989020(7.74)	−28.7945604(246.8)
NGC 3079	5.887221	3.31 × 1.78	17.9	441.6	0954+556	149.40910265(63.41)	55.38271342(38.3)
NGC 3227	5.874904	5.13 × 1.77	−11.9	48.3	1022+194	156.186706652(2.61)	19.20567097(41.3)
NGC 3786	5.878159	3.25 × 1.70	21.6	47.9	1133+344	174.113933109(8.91)	34.1276345(181.3)
NGC 4151	5.855409	3.13 × 1.78	42.4	175.1	1204+399	181.654389055(8.75)	39.6843742(132.2)
NGC 4180	5.868327	6.92 × 1.55	−4.2	76.3	1212+087	183.749638201(8.64)	8.4895883(191.2)
NGC 4235	5.906512	3.36 × 1.48	−7.6	106.4	1212+087	183.749638201(8.64)	8.4895883(191.2)
NGC 4388	5.868400	2.76 × 1.60	13.0	107.9	1222+131	186.265597247(4.1)	12.8869831(138)
NGC 4593	5.869410	6.93 × 1.52	−6.7	145.3	1245−062	192.095731923(4.98)	−6.5360605(154.7)
NGC 5290	5.883406	3.69 × 2.02	−19.9	61.8	1357+404	209.908726091(7.73)	40.1939585(120.2)
NGC 5506	5.872330	5.36 × 1.78	−16.9	352.9	1402−012	211.191231047(5.89)	−1.50609643(93.6)
NGC 5899	5.869410	6.93 × 1.52	−6.7	42.1	1505+428	226.721007704(6.32)	42.65639874(79)
NGC 6814	5.873225	5.88 × 2.08	−11.6	50.0	1937−101	294.988569046(4.17)	−10.04486683(98.1)
NGC 7314	5.876574	7.68 × 3.73	3.9	81.5	2240−260	340.860036591(6.44)	−25.7418576(189.9)
NGC 7378	5.865883	6.82 × 2.38	−12.3	46.7	2243−123	341.575966543(2.11)	−12.11424378(33.5)
NGC 7465	5.876460	4.30 × 2.13	−15.9	47.4	2258+166	345.179129648(7.01)	16.9206644(148.4)
NGC 7479	5.872223	4.43 × 2.04	−17.1	43.5	2307+106	347.618823901(2.5)	10.92519353(41.8)

Download table as:  ASCIITypeset image

2.2.1. Data Calibration

2.2.2. Imaging

To image each target, we used the aips imaging task imagr. For each target we used similar input parameters, including a Briggs weighting with robust = 5.0 for a natural weighting in order to maximize sensitivity, and a pixel size of 0.8 mas, which Nyquist-samples the synthesized beam. We took extreme care in determining the rms of our VLBA data by initially determining the rms from a "dirty image," which is made when the number of iterations is set to zero for an initial iteration of clean. Next, we deconvolved the point-spread function using no more than 500 iterations of clean to derive the best estimate for the true rms noise of each image (see Table 2). We interactively placed clean boxes around all point-source emission and interactively cleaned until a thermal noise-limited image was made. For several of our observations, there were significant amounts of RFI, and some antennas required complete flagging altogether. For example, for all sources, the Kitt Peak VLBA antenna suffered a pointing error due to a software bug and thus required flagging altogether for the duration of our observations. In the observing session that included NGC 4151, NGC 4180, NGC 4235, NGC 4388, and NGC 4593, we had to flag an additional two antennas due to RFI and other issues in the data. For these reasons, the rms values for these targets presented in Table 2 are above the theoretical rms by factors of 4–5.

For the imaging procedure, we started with a field of view of 512 pixels per side, which we increased systematically to search for point-source emission, while being mindful not to introduce effects from bandwidth or time smearing. For sources where we did not detect a point source, we imaged out to a radius of ∼15 from the phase center of the source positions. One of our detected targets, NGC 5290, had a position offset of +035, −02 in R.A. and decl., respectively, from the a priori position as determined from the Pan-STARRS y-band data. Thus, we applied this R.A. and decl. offset in the imaging process in order to center the source in the field of view.

Once we had produced initial images, we then proceeded with self-calibration for the nine objects that we detected. Our self-calibration procedure used the AIPS task calib, and we started by self-calibrating on phase using the initial input image. Self-calibration is only viable for high S/N detections, and thus, we are only able to apply this technique to the nine objects for which we achieved an S/N of >15. This is an iterative process, running CALIB and IMAGR to create new images from the preceding calibrated uv data files. We repeated this process for several cycles, usually between three or four, until there was little to no improvement in the rms noise. When this was achieved, we then continued self-calibration by calibrating both the amplitudes and phases. After a few more iterations, our images converged and the rms noise in the final resulting image no longer improved. Because we used phase referencing, our absolute source positions were preserved throughout the self-calibration process. For NGC 1068, however, self-calibration was not applied as this source suffers from severe short-spacing artifacts and requires a multiscale imaging technique. We used four spatial scales for our multiscale imaging algorithm at 0, 2, 4, and 8 times the restoring beam, and we imaged 2048 × 2048 pixels. The final images for our nine detected nuclear AGN sources are presented in Figure 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We also include archival images of our sample from the NRAO VLA Archive Survey (NVAS),⁹ when available, at ∼4.89 GHz. Largely taken in A-configuration, we include these data sets in our analysis to compare how peak fluxes of ostensibly core radio AGN emission may in fact differ depending on different resolutions. Of the 25 targets in our sample, 19 possessed relevant observations in the archive, which are presented in Figure 3.

The NRAO Common Astronomical Software Applications (casa; McMullin et al. 2007)¹⁰ was used to analyze each image in our VLBA and VLA observations. Using the viewer command in casa, we used the two-dimensional fitting application to determine both peak-flux and integrated flux densities for the 9 detected sources in our VLBA observations and the entirety of the 19 sources observed by the VLA. Peak-flux values were measured directly from the image, and integrated values were measured by enclosing all nuclear flux greater than 3σ or 5σ for VLBA and VLA measurements, respectively. The VLA flux distribution in NGC 1068 contains several discrete regions that are not connected to the nuclear emitting region that were also included in its integrated flux density measurement. Table 3 shows the flux parameters for all 9 detections, and Table 4 shows the peak-flux and integrated flux densities for the VLA observations.

Table 3.  VLBA Measurements

Name	R.A.	Decl.	Fpeak	Log Fpeaka	Log Lpeak	Sinta
	(deg)	(deg)	(mJy bm−1)	(×10−16 erg s−1 cm−2)	(erg ${{\rm{s}}}^{-1})$	(mJy)
NGC 1052	40.2699939	−8.2557642	745 ± 15	437 ± 8.8	39.38	1269 ± 14
NGC 1068	40.6696212	−0.0133181	0.198 ± 0.016	0.116 ± 0.0094	35.57	2.37 ± 0.19
NGC 2110	88.0474013	−7.4562554	6.89 ± 0.07	4.04 ± 0.041	37.73	11.15 ± 0.17
NGC 2992	146.4247680	−14.3262771	1.01 ± 0.04	0.592 ± 0.023	36.89	1.29 ± 0.10
NGC 3079	150.4908485	55.6797891	9.95 ± 0.29	5.84 ± 0.17	37.24	27.02 ± 0.96
NGC 4151	182.6357579	39.4058501	2.94 ± 0.14	1.73 ± 0.082	36.62	11.61 ± 0.66
NGC 4235	184.2911738	7.1915751	2.56 ± 0.023	1.50 ± 0.013	37.33	3.11 ± 0.049
NGC 5290	206.3298364	41.7123508	2.32 ± 0.051	1.36 ± 0.029	37.35	6.49 ± 0.19
NGC 5506	213.3119889	−3.2076829	21.90 ± 0.14	12.9 ± 0.0082	38.04	29.67 ± 0.32

Note.

^aPeak-flux values, corresponding to the observations in Table 2, derived from the CASA 2D Gaussian model-fitting algorithm, and integrated flux densities are determined from flux measured within the 3σ outer contour shown in Figure 2.

Download table as:  ASCIITypeset image

Table 4.  VLA Archival Observations and Measurements

Name	Frequency	VLA	Restoring Beam	Beam Angle	rms	Fpeak	Fpeak	Log Lpeak	Sinta	Fpeak/
	(GHz)	Config.	($\alpha \times \delta ;$ '')	(deg)	( μJy bm−1)	(mJy bm−1)	(erg s−1 cm−2)	(erg ${{\rm{s}}}^{-1})$	(mJy)	VLBArms
VLBA Detected Sources
NGC 1052	4.86	A	0.50 × 0.38	−32	177.0	1808	8.79 × 10−14	39.67	1889	142.1
NGC 1068	4.86	A	0.43 × 0.39	−671	80.6	252	1.22 × 10−14	38.59	980	3871.0
NGC 2110	4.99	A	0.45 × 0.36	−26	78.9	53.5	2.67 × 10−15	38.55	108	110.1
NGC 2992	4.86	A	0.53 × 0.39	−52	18.3	8.0	3.88 × 10−16	37.71	57.7	25.3
NGC 3079	4.99	A	0.48 × 0.33	−635	162.0	58.4	2.91 × 10−15	37.95	89.0	44.1
NGC 4151	4.99	A	0.40 × 0.35	889	90.4	38.2	1.91 × 10−15	37.66	102.0	72.8
NGC 4235	4.86	A	0.43 × 0.40	04	59.7	4.6	2.25 × 10−16	37.50	5.17	14.5
NGC 5290	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
NGC 5506	4.86	A	0.43 × 0.37	122	103.0	102.	4.96 × 10−15	38.63	154	96.3
VLBA Nondetections
NGC 1320	4.86	B	1.56 × 1.39	−218	22.9	1.65	8.04 × 10−17	37.15	1.92	20.9
NGC 2782	4.86	B	1.34 × 1.16	−544	60.3	1.50	7.29 × 10−17	37.07	2.58	17.3
IC 2461	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
NGC 3081	4.86	A	0.66 × 0.35	−82	54.5	0.500	2.43 × 10−17	36.54	0.502	4.5
NGC 3089	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
NGC 3227	4.86	A	0.37 × 0.34	182	91.9	10.8	5.24 × 10−16	37.25	22.2	88.0
NGC 3786	4.86	A	0.37 × 0.33	148	87.6	1.70	8.26 × 10−17.	37.16	2.51	13.8
NGC 4180	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
NGC 4388	4.86	A	0.38 × 0.36	−285	44.8	3.17	1.54 × 10−16	37.38	9.26	19.7
NGC 4593	4.86	A	0.46 × 0.38	−20	94.0	1.15	5.59 × 10−17	37.00	1.23	10.3
NGC 5899	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
NGC 6814	4.86	A	0.52 × 0.4	−06	135.0	1.16	5.64 × 10−17	36.52	1.40	8.9
NGC 7314	4.86	A	0.82 × 0.35	220	65.9	1.72	8.36 × 10−17	36.53	2.00	8.3
NGC 7378	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
NGC 7465	4.86	A	0.51 × 0.37	666	48.6	1.21	5.88 × 10−17	36.76	1.61	10.8
NGC 7479	4.86	A	0.49 × 0.41	−726	51.4	2.43	1.18 × 10−16	37.22	2.46	18.7

Note.

^a5σ contour.

Download table as:  ASCIITypeset image

We obtained Target of Opportunity (ToO) observations (Table 5; PI: N. Secrest) of 20 out of 25 of our sample with the Neil Gehrels Swift Observatory X-ray Telescope (XRT; Burrows et al. 2003; Hill et al. 2004; Burrows et al. 2005), which has a point-spread function (PSF) with a half-power diameter of $18^{\prime\prime}$ and a full width at half maximum (FWHM) of $7^{\prime\prime}$ at 1.5 keV, and a positional accuracy of $3^{\prime\prime}$. These observations were simultaneous with our VLBA observations to obtain a simultaneous snapshot of the radio and X-ray properties of our volume-complete sample of AGNs, as is generally warranted when comparing the radio and X-ray properties of AGNs (e.g., Merloni et al. 2003; Gültekin et al. 2009). Three of the five objects we did not acquire observations of were visibility-constrained, and the ToOs for the remaining objects, IC 2451 and NGC 2992, were approved but not executed. We requested 1.8 ks integrations for all ToOs, with one ToO set up in Windowed Timing (WT) mode to avoid possible pileup, and the remaining ToOs in Photon Counting (PC) mode following the recommendation of the Science Operations Team. Finally, although our ToO observation for NGC 2992 was not executed, a separate XRT observation of NGC 2992 was taken during the requested time in PC mode, which we include with our data.

We generated X-ray spectra for our data using the online XRT product generator (Evans et al. 2009),¹¹ allowing the routine to center the X-ray sources within the default $1^{\prime}$ search radius. For objects where centroiding failed due to faintness, such as NGC 4180, we turned off centroiding and set the extraction coordinates to the target coordinates used for the ToO, which were the same as the VLBA targeting coordinates.

Similar to the procedure outlined in Ricci et al. (2017), we carried out a joint X-ray spectral analysis by combining the contemporaneous XRT data with the time-averaged 105-month BAT data, and fitting models of increasing complexity. We fit the X-ray spectra using xspec (Arnaud 1996), version 12.10.1f, using phabs to model both intrinsic and Galactic absorption for consistency with the MYTorus model (Murphy & Yaqoob 2009) that we use for Compton-thick sources, which requires the Anders & Grevesse (1989) abundances and the Verner et al. (1996) photoionization cross-sections. For Galactic absorption, we use the Swift online Galactic N_H tool,¹² which uses the method of Willingale et al. (2013).

We began by fitting a single power-law component pow, allowing the normalizations to vary between the XRT and BAT data to account for variability. If soft X-ray absorption is clearly present, we appended a phabs component to account for intrinsic absorption, again allowing the normalizations to vary. For some sources, a soft excess is present. This we fit with either a blackbody component (bb) or an additional, nonabsorbed power-law component. For sources in which the best-fit N_H approached Compton-thick levels ($\gt {10}^{24}$ cm⁻²), we used MYTorus. We note that because of the ∼10 keV coverage gap between the XRT and BAT data, for some sources we relied on the spectral curvature in the BAT data to determine whether a source is Compton-thick (e.g., Koss et al. 2016). In all cases, the 2–10 keV flux we calculate is rest-frame, corrected for absorption, and from the intrinsic power-law component only. For Compton-thick sources, the scattered component may dominate the high-energy continuum (e.g., Murphy & Yaqoob 2009), necessitating the use of a physical model such as MYTorus to infer the strength of the intrinsic power-law continuum. In order to constrain the intrinsic flux from the power-law component at the epoch of the XRT data, we freeze the best-fit spectral parameters corresponding to components existing at large physical scales and unlikely to exhibit significant short-term variability, such as a blackbody or plasma emission component, or optically thin scattered emission. We then fit the XRT data separately from the fit that included the BAT data, determining the 90% upper limit on the power-law normalization allowable by the XRT data. We note that when using the MYTorus model the normalization of the scattered continuum is usually held fixed to that of the intrinsic power-law continuum. Because there are likely time lags of unknown duration between the intrinsic power-law emission and the scattered continuum, it is possible that the intrinsic power-law emission at the epoch of the XRT data may be considerably stronger than the scattered continuum. Fixing the scattered continuum means that a much larger range of power-law emission is allowed by the XRT data, yielding a more conservative upper limit.

As a check to ensure we are performing a core X-ray luminosity to core radio luminosity comparison, we additionally fit the Swift BAT data alone, which removes model assumptions and potential contamination from diffuse X-ray emission at softer energies. We fit the BAT spectra with the simplest model required to adequately fit the data (${\chi }^{2}/\mathrm{dof}\lt 2$), where dof is the degrees of freedom. For 15 out of 25 objects a simple power law was sufficient; for 5 objects a power law with a high-energy cutoff was required; for the remainder, either a power law or cutoff power law with an absorption component was warranted. We do not physically interpret these models or attempt to deconvolve an intrinsic power-law continuum from the fit as the 105-month BAT catalog data are limited to only eight spectral bins. We appended the cflux convolution component in front of the best-fit spectral model and fit for the total observed 14–195 keV flux after freezing the power-law normalization. We found that we were able to more robustly estimate the uncertainty of the 14–195 keV flux using a Markov chain Monte Carlo (MCMC) algorithm with the default Goodman-Weare algorithm with 10 walkers. For every object we ran chains of length 10⁵ after determining the appropriate burn-in period. For NGC 4180, the MCMC chain produced a bimodality in Γ with one mode at unphysically low values $\lt 1$. We alleviated this by setting a Gaussian prior on Γ with a mean of 2.00 and standard deviation of 0.38, taken from the distribution of Γ for all AGN in the 105-month catalog, using the bayes command.

Finally, as we are comparing the XRT data to data from other facilities, we include in the X-ray fluxes a systematic uncertainty floor of 10% (Burrows et al. 2005), which we add in quadrature to the formal errors.

An immediate point of interest from our initial snapshot program was the lack of VLBA detections for a majority of our sample. As our VLBA integration time calculations were derived from the FP for AGNs, we expected detections for the entirety of our sample. The left plot of Figure 4 places our sample on the FP, in comparison to data from Merloni et al. (2003), using peak radio and X-ray luminosities calculated from VLBA and Swift XRT measurements in Tables 3 and 6, respectively, and black hole mass measurements from Table 1. Targets with VLBA detections are represented by red and blue points that use simultaneous and archival XRT observations, respectively. Green points with arrows mark radio luminosity upper limits for targets with VLBA nondetections, which we define as 3σ over the rms in each observation. From this comparison, we find that although several radio-luminous targets fall within the Merloni et al. (2003) distribution, our sample lies largely below the defined plane. As the FP was derived using VLA observations, we replace our peak radio luminosities measured from our VLBA observations with peak radio luminosities from archival VLA observations in Table 4. Using these values, as shown in the right plot of Figure 4, we find our targets are 100% detected, versus 36% with VLBA, and become well aligned with the FP and the Merloni et al. (2003) data distribution.

Zoom In Zoom Out Reset image size

Table 6.  X-Ray Spectral Fitting Results

Object	Model	Total stat/dof	NH	Γ	${\mathrm{log}}_{10}({F}_{2-10\mathrm{keV}})$
			(cm−2)		(erg cm−2 s−1)
NGC 2110	phabs∗pow + const∗pow	383.25/473	${6.8}_{-1.7}^{+2.1}\times {10}^{22}$	${2.5}_{-0.5}^{+0.6}$	$-{10.12}_{-0.04}^{+0.05}$
NGC 2782	apec + zpow + const∗pow	29.31/30	${1.4}_{-0.1}^{+0.3}\times {10}^{23}$	${1.8}_{-0.5}^{+0.5}$	$-{11.9}_{-0.4}^{+0.5}$
NGC 2992	phabs∗pow	358.22/416	${7.5}_{-1.7}^{+1.9}\times {10}^{21}$	${1.6}_{-0.2}^{+0.2}$	$-{10.00}_{-0.02}^{+0.02}$
NGC 3079	vapec + MYTorus	120.30/95	${8.5}_{-1.0}^{+1.2}\times {10}^{24}$	${2.2}_{-0.1}^{+0.1}$	−9.9
NGC 3081	apec + MYTorus	41.05/34	${6.1}_{-0.3}^{+0.4}\times {10}^{23}$	${1.96}_{-0.01}^{+0.01}$	$-{10.5}_{-0.2}^{+0.1}$
NGC 3089	phabs∗pow	28.55/27	${7.1}_{-3.5}^{+5.4}\times {10}^{22}$	${2.7}_{-0.8}^{+1.1}$	$-{11.7}_{-0.1}^{+0.2}$
NGC 3227	pow	290.09/292	$\lt 6.1\times {10}^{20}$	${1.7}_{-0.1}^{+0.1}$	$-{10.10}_{-0.03}^{+0.03}$
NGC 4151	zbbody + zxipcf∗zcutoffpl	258.15/328	${7.8}_{-2.0}^{+2.2}\times {10}^{22}$	${1.51}_{-0.03}^{+0.03}$	$-{9.60}_{-0.08}^{+0.06}$
NGC 4180	MYTorus	5.40/2	${7.3}_{-1.5}^{+1.7}\times {10}^{24}$	${1.5}_{-0.2}^{+0.3}$	−11.1
NGC 4235	phabs∗pow	150.28/181	${2.7}_{-1.4}^{+1.6}\times {10}^{21}$	${1.7}_{-0.3}^{+0.3}$	$-{11.13}_{-0.06}^{+0.06}$
NGC 4388	phabs∗pow + const∗pow	248.06/265	${4.6}_{-0.9}^{+1.0}\times {10}^{23}$	${3.2}_{-0.9}^{+1.0}$	$-{9.5}_{-0.1}^{+0.2}$
NGC 4593	pow	304.40/353	$\lt 2.7\times {10}^{20}$	${1.6}_{-0.09}^{+0.09}$	$-{10.45}_{-0.03}^{+0.03}$
NGC 5290	phabs∗pow	114.27/145	${9.5}_{-5.3}^{+6.5}\times {10}^{21}$	${1.1}_{-0.5}^{+0.5}$	$-{10.83}_{-0.05}^{+0.05}$
NGC 5506	zxipcf∗pexrav	257.79/304	${1.2}_{-0.1}^{+0.6}\times {10}^{22}$	${1.7}_{-0.2}^{+0.3}$	$-{9.97}_{-0.03}^{+0.10}$
NGC 5899	phabs∗pow	51.39/49	${10.9}_{-3.8}^{+5.1}\times {10}^{22}$	${2.0}_{-0.2}^{+0.2}$	$-{11.1}_{-0.1}^{+0.1}$
NGC 6814	pow	324.07/350	$\lt 5.4\times {10}^{20}$	${1.6}_{-0.1}^{+0.1}$	$-{10.68}_{-0.03}^{+0.03}$
NGC 7314	phabs∗pow	247.97/293	${8.6}_{-1.3}^{+1.4}\times {10}^{21}$	${1.9}_{-0.1}^{+0.1}$	$-{10.69}_{-0.03}^{+0.03}$
NGC 7378	phabs∗pow	8.94/9	${5.8}_{-2.1}^{+2.2}\times {10}^{22}$	${1.7}_{-0.1}^{+0.1}$	$-{12.1}_{-0.2}^{+1.1}$
NGC 7465	phabs∗pow	166.46/215	${8.2}_{-2.6}^{+3.0}\times {10}^{21}$	${1.6}_{-0.2}^{+0.3}$	$-{10.95}_{-0.05}^{+0.05}$
NGC 7479	apec + MYTorus	15.52/9	${5.7}_{-0.5}^{+0.5}\times {10}^{24}$	${2.5}_{-0.1}^{+0.1}$	−8.8

Note. All (formal) confidence intervals and upper limits are 90% except for the X-ray fluxes, which are $\pm 1\sigma$. X-ray fluxes without a confidence interval are from fits not using contemporaneous Swift XRT data. Note that Swift XRT was in an anomalous state for NGC 3786, so we could not use its data.

Download table as:  ASCIITypeset image

Our findings are consistent with those of Panessa & Giroletti (2013), who studied a more nearby (${D}_{L}\lt 22$ Mpc) sample of Seyfert galaxies with the European VLBI Network (EVN), which has a similar maximum baseline to the VLBA (∼9000 km). They find that out of 21 Seyfert galaxies observed with the EVN, 13 (62%) are detected at 5 GHz. The 90% confidence interval from binomial statistics for this fraction is 42%–79%, whereas for our targets, of which 36% were detected, the 90% confidence interval is 20%–54%, with a p-value between our study and Panessa & Giroletti (2013) of p = 0.16. Nonetheless, there are differences between these samples that are worth noting. First, the Panessa & Giroletti (2013) sample is again considerably closer, with a mean distance of 14.7 Mpc,¹³ compared to the mean distance of 30.0 Mpc for our sample. Given the 3σ luminosity upper limits listed in their Table 1, the mean rms of their EVN data is ∼57 μJy, about half as sensitive as our VLBA observations. The difference in mean distance and sensitivity implies that the 5 GHz observations used by Panessa & Giroletti (2013) are comparable in terms of probing the target luminosities, with upper luminosity limits about ∼0.6 times that of our observations. Of the 20 objects in Panessa & Giroletti (2013) with 2–10 keV X-ray luminosity measurements, the mean X-ray luminosity is $8.1\times {10}^{41}$ erg s⁻¹, while for our sample the mean luminosity is $1.7\times {10}^{43}$ erg s⁻¹, indicating that higher X-ray luminosity is not necessarily correlated with greater prevalence of radio core detection at mas scales, given the statistical consistency of the latter between our study and that of Panessa & Giroletti (2013). As in this work, Panessa & Giroletti (2013) also find that the X-ray emission is more correlated with larger-scale radio emission as captured by the archival VLA observations they use, and they suggest that this implies that the hard 2–10 keV luminosity in an AGN is associated with larger-scale structure than the radio core. With both 5 GHz and 1.7 GHz VLBI observations, they are able to estimate spectral index α, and find that the disparity between the VLA and VLBI fluxes is smaller in flatter-spectrum objects, consistent with the emission being truly core dominated.

It is interesting to note that other studies aimed at detecting radio-quiet quasars on spatial scales of the VLBA using ancillary catalogs such as those from the VLA as a priori source lists find similar detection rates to those we find in this work. For example, the recent work by Herrera Ruiz et al. (2017) mosaicked the COSMOS field with the VLBA at 1.4 GHz and find a detection rate of ∼20% for all VLA detected sources at 3 GHz. Herrera Ruiz et al. (2018) followed up this work by combining the Robert C. Byrd Green Bank Telescope together with the VLBA (VLBA+GBT) for increased sensitivity at 1.4 GHz and detected an additional 10 sources that were not previously seen with the VLBA alone. The overall detection rate for this follow-up study, however, remained at ∼20%. Similarly, Middelberg et al. (2013) mosaicked the Lockman Hole/XMM field with the VLBA at 1.4 GHz and out of their parent sample of 217 sources, only 65 sources (∼30%) were detected at their 1σ sensitivity threshold of ∼20 μJy beam⁻¹. Although these studies present results at a different frequency than our study (1.4 GHz versus 6 GHz), the radio emission across both of these frequencies is expected to be dominated by synchrotron, nonthermal emission processes (see e.g., Condon 1992, for a discussion of radio emission in galaxies). There are clear differences in the observing strategies, sensitivity limits, and populations of radio quasars studied by Panessa & Giroletti (2013), Herrera Ruiz et al. (2017, 2018), Middelberg et al. (2013), and our volume-complete sample, but the detection rates from all of these studies are within the ∼20%–40% range, which may indicate that a similar process is responsible for the core radio emission seen in AGNs at these VLBA spatial scales.

To explore this idea further, we compare the VLBA and VLA peak and integrated flux density measurements in our targets. We find that the difference in peak-flux measurements between VLBA and VLA observations appears to be greater in targets such as NGC 1068 and NGC 2992, with VLBA measurements that place them farther from the FP, as shown in Figure 5. We also find that targets farther from the FP have more extended 6 cm radio morphologies in our VLA observations. We quantify this by measuring the ratio between the VLA flux peak (F_peak) and the VLA integrated flux density greater than 5× rms (F_int) using the values listed in Table 4, where more extended targets have a greater portion of their flux outside the nucleus and a ratio closer to zero. Furthermore, we find a relationship between these two ratios, where AGN with greater VLA/VLBA flux ratios are typically more extended as they exhibit higher VLA F_peak/F_int ratios. Together, these relationships suggest that the emission observed in VLA observations is largely extended, where AGN that deviate farther from the FP have larger differences in VLA and VLBA peak-flux measurements and exhibit more extended radio emission than the nucleus. Additionally, we can show that emission measured in VLA observations is extended in targets with VLBA nondetections by comparing our VLA peak-flux measurements to the VLBA background rms as listed in Table 3. Assuming that the VLA emission originated from a point source, one would expect to detect this source in the VLBA observations with an S/N ${\rm{S}}/{\rm{N}}={F}_{\mathrm{peak}}(\mathrm{VLA})/{\mathrm{VLBA}}_{\mathrm{rms}}$. We measure this hypothetical S/N to be >3 in all targets with VLA observations, with S/N > 5 in 10 of 11 targets, suggesting a point source would have been detected for each AGN if present.

Zoom In Zoom Out Reset image size

Following Zakamska & Greene (2014), we suggest that the extended radio emission observed in the VLA observations for this sample is largely a byproduct of relativistic particles accelerated in shocks caused by AGN-driven winds, similar to processes in supernova remnants. Assuming a scenario where radio emission in each of these targets is attributed to a jet emanating from the nucleus, one would expect larger discrepancies between VLA and VLBA observations in compact sources, where the jet is pointed along our line of sight. Instead of being spread across the plane of the sky, the bulk of the extended contaminating emission in these targets would be contained near the nucleus along our line of sight, captured in VLA observations but resolved out in VLBA observations. As we note, we find higher ratios to exist between VLA and VLBA peak-flux measurements in targets with more extended morphologies, thus it is likely that we are not observing an effect dependent on the orientation of the AGN with respect to our line of sight. Instead, in a scenario where emission is formed by shocks via AGN feedback, the observed ratio between VLA and VLBA fluxes would be dependent on the orientation of the AGN with respect to its host galaxy, where targets in our sample with extended VLA morphologies are exhibiting more interactions between the AGN and its host medium than targets with compact VLA morphologies.

These interactions translate into parsec-scale levels, where VLA peak-flux measurements of our AGN are likely contaminated by unresolved extranuclear host interactions that resolve out of VLBA peak-flux observations with the AGN becoming much less luminous than hypothesized from the FP relation. We see this explicitly in VLBA observations of NGC 1068 in Figure 6, which exhibit several knots of extranuclear emission in addition to the AGN, whose position was previously identified by Gallimore et al. (2004) and references therein. Placing the synthesized beam and flux peak location obtained from the archival VLA observations onto our VLBA observations, we see that the peak flux measured by the VLA is derived largely from extranuclear emission north of the nucleus and likely does not contain emission from the AGN itself. While the source at the center of the VLA beam and the central AGN exhibit similar peak fluxes in our VLBA measurements, MERLIN observations from Gallimore et al. (2004) reveal the non-AGN source to be much more luminous at scales slightly larger (MERLIN restoring beam ∼60 mas) than what is resolvable by VLBA (maximum resolvable size ∼47 mas). We note that NGC 1068 is the only source in our sample that exhibits such a misalignment between VLA peak and VLBA peak, which likely accounts for the extreme ratio between VLA and VLBA fluxes for NGC 1068 compared to the seven other sources shown in Figure 5. However, even without a misalignment where the VLA peak flux would have been centered over the AGN, Figure 6 also exhibits several knots of extranuclear emission adjacent to the central engine that would contaminate VLA peak-flux measurements for NGC 1068.

Zoom In Zoom Out Reset image size

We have shown that when probed at the exquisite angular resolution of the VLBA, AGN are underluminous or altogether undetected, at odds with their properties in lower angular resolution VLA data, but potentially more in line with the scaling relation for Galactic X-ray binaries. We can remove the dependence of black hole mass in this relation for a pure comparison between X-ray and radio luminosities and produce a relationship that may better encapsulate the core behavior of AGNs. Figure 7 compares ${L}_{6\mathrm{cm}}$ and ${L}_{2-10\mathrm{keV}}$ in our sample, with detections as red points and radio luminosity upper limits derived from rms values for nondetections as green points, revealing VLBA observed AGN largely fall in line with measurements for Galactic black holes (GBHs). While the majority of our sample is undetected, these nondetections are significant, and so in order to preserve the statistical power of our sample, we fold in the information these nondetections carry by performing regression with censored data using the asurv package (Lavalley et al. 1992).

Zoom In Zoom Out Reset image size

${L}_{5\mathrm{GHz}}$ and ${L}_{2-10\mathrm{keV}}$ measurements from Table 1 of Panessa & Giroletti (2013) are also included as blue points. Initially, there appears to be a discrepancy between how the two data sets compare to measurements from GBHs. However, separating radio-quiet and radio-loud AGN according to the X-ray radio-loudness parameter (${R}_{X}\equiv {L}_{6\mathrm{cm}}/{L}_{2-10\mathrm{keV}}$ limit (${R}_{X}\,=-4.5$) from Terashima & Wilson (2003), we find the eight radio-loud AGN in the Panessa & Giroletti (2013) sample are the targets that do not strongly agree with the general trend. Additionally, NGC 1052 from our sample is also radio-loud by this definition, likewise separating itself from the general trend, which concurs with its use as an ICRF target and the documented presence of a relativistic radio jet (Lister et al. 2019) often present in radio-loud AGN. We note that while GBHs may themselves exhibit bimodality in L_R/L_X (e.g., Gallo et al. 2012), these modes do not appear to be reflected in the AGN population. When we use the two best-fit relations from Gallo et al. (2012), the GBHs in this work are consistent with one or the other, but both relations predict radio luminosities that are three to four orders of magnitude below the radio-loud relation at the X-ray luminosities of AGNs. Of the two GBH L_R/L_X "tracks" presented in Gallo et al. (2012), the lower track, corresponding to GBHs that are less radio-luminous, is more consistent with the AGNs studied in this work.

As even completely uncorrelated fluxes can become highly correlated when multiplied by random distances spanning several orders of magnitude, we fit the FP in terms of flux as well as luminosity normalized by BH mass. The results are shown in Figure 8. They show that the correlation between the 2–10 keV and 6 cm fluxes and BH mass-normalized luminosities is weak, and that the values derived using the high physical resolution VLBA data overlap with the values seen in XRBs, which are systematically lower than AGNs observed in lower-resolution studies.

Zoom In Zoom Out Reset image size

Finally, we compared the radio luminosities to the observed 14–195 keV luminosities from the 105-month BAT spectra alone, using the fluxes calculated in Section 2.6 and provided in Table 7. For the X-ray data from the Merloni et al. (2003) sample, which was mostly taken from published literature using a wide variety of X-ray observatories, we multiply the 2–10 keV luminosities by a factor of 1.6 to estimate the 14–195 keV luminosities, a factor derived from a power law with a spectral index of 2. The results are shown in the right panel of Figure 7. As can be seen, the 2–10 keV fluxes calculated using our simultaneous Swift XRT observations are consistent with those inferred from the BAT data, indicating that apparent underluminosity seen in the radio between VLA and VLBA data is not in fact due to an overluminosity at softer X-rays due to contamination from extranuclear sources.

Table 7.  Best-fit xspec Models for the Swift BAT Data from the 105 months catalog (Oh et al. 2018), with Observed 14–195 keV Fluxes

Object	Model	${\chi }^{2}/\mathrm{dof}$	${\mathrm{log}}_{10}\left({F}_{14-195\mathrm{keV}}\right)$
			(erg cm−2 s−1)
NGC 1052	pow	4.87/5	$-{10.505}_{-0.046}^{+0.036}$
NGC 1068	phabs∗pow	2.51/4	$-{10.515}_{-0.047}^{+0.031}$
NGC 1320	pow	3.31/5	$-{10.872}_{-0.121}^{+0.067}$
NGC 2110	cutoffpl	4.16/4	$-{9.496}_{-0.006}^{+0.004}$
NGC 2782	pow	1.45/5	$-{10.921}_{-0.136}^{+0.068}$
IC 2461	pow	3.26/5	$-{10.722}_{-0.025}^{+0.011}$
NGC 2992	pow	1.43/5	$-{10.487}_{-0.046}^{+0.035}$
NGC 3079	phabs∗cutoffpl	2.78/3	$-{10.550}_{-0.008}^{+0.053}$
NGC 3081	phabs∗cutoffpl	1.53/3	$-{10.125}_{-0.026}^{+0.012}$
NGC 3089	pow	0.76/5	$-{11.156}_{-0.248}^{+0.072}$
NGC 3227	phabs∗cutoffpl	4.34/3	$-{9.990}_{-0.010}^{+0.015}$
NGC 3786	pow	2.1/5	$-{10.836}_{-0.095}^{+0.058}$
NGC 4151	cutoffpl	7.11/4	$-{9.249}_{-0.002}^{+0.003}$
NGC 4180	pow	9.14/5	$-{10.847}_{-0.139}^{+0.074}$
NGC 4235	pow	4.29/5	$-{10.414}_{-0.034}^{+0.028}$
NGC 4388	pow	2.18/4	$-{9.569}_{-0.005}^{+0.005}$
NGC 4593	cutoffpl	0.87/4	$-{10.112}_{-0.013}^{+0.034}$
NGC 5290	pow	2.54/5	$-{10.835}_{-0.076}^{+0.054}$
NGC 5506	phabs∗cutoffpl	2.93/3	$-{9.700}_{-0.004}^{+0.013}$
NGC 5899	pow	2.03/5	$-{10.688}_{-0.053}^{+0.037}$
NGC 6814	cutoffpl	2.18/4	$-{10.258}_{-0.012}^{+0.050}$
NGC 7314	cutoffpl	0.44/4	$-{10.335}_{-0.018}^{+0.046}$
NGC 7378	pow	2.21/5	$-{10.863}_{-0.170}^{+0.079}$
NGC 7465	pow	3.5/5	$-{10.722}_{-0.083}^{+0.055}$
NGC 7479	pow	5.31/5	$-{10.774}_{-0.096}^{+0.058}$

Download table as:  ASCIITypeset image

With the results detailed in Sections 3.1 and 3.2, it is clear that there is a resolution dependence of the perceived radio luminosity of nearby AGNs, that at subparsec scales the radio emission is considerably less luminous, and that the FP does not appear to capture the relationship between the black hole mass, X-ray luminosity, and radio emission at these scales. It is therefore worthwhile to explore what physical mechanisms may cause the radio emission that we do detect on mas scales.

In higher-frequency studies (e.g., 22 GHz Baek et al. 2019; Smith et al. 2020), there may be some mixture of synchrotron and thermal bremsstrahlung (free–free) emission, which begins to dominate for star-forming galaxies above ∼30 GHz (Condon 1992). However, for our observations, which were taken at 5 GHz, it is more likely that the emission is synchrotron dominated. This is not necessarily true, however, if the radio emission is "young," which induces a spectral curvature, or turnover at lower frequencies. One means of distinguishing these emission mechanisms is by calculating the brightness temperature,

$$\begin{eqnarray}&&{T}_{b}=\displaystyle \frac{4\mathrm{ln}\left(2\right){c}^{2}{S}_{\nu }}{2\pi {k}_{B}{\theta }_{A}{\theta }_{B}},\end{eqnarray} \tag{ 1 }$$

where S_ν is the specific intensity over the solid angle, and ${\theta }_{A}$ and ${\theta }_{B}$ are the full widths at half-power of the major and minor axes of an elliptical Gaussian beam. We find that the brightness temperatures calculated from our VLBA detections range from ${10}^{6.1}$ to ${10}^{9.7}$ K, indicating a likely dominant synchrotron source. One potential source of compact emission, thermal free–free emission from accretion disk winds, may be significantly present in NGC 1068 and NGC 2992, which have brightness temperatures of ${10}^{6.1}$ K and ${10}^{6.3}$ K, respectively. NGC 1068 exhibits extreme line-of-sight absorption ${N}_{{\rm{H}}}\sim {10}^{25}$ cm⁻² (Ricci et al. 2017), which may scatter or absorb direct nonthermal AGN emission (Gallimore et al. 1997, 2004). For the remaining objects, ${T}_{b}\geqslant {10}^{7}$, suggesting either jet-associated or corona-associated synchrotron. While our observations sample down to the parsec or subparsec scale, it is possible that subparsec jets may exist, as was recently found for a small fraction (4%) of BAT-selected AGNs using other VLBI observations (Baek et al. 2019). However, the AGNs in that study are more distant, with radio luminosities several orders of magnitude higher than the sample studied here. Indeed, Baek et al. (2019) find that the L_R/L_X relation of their sample spans a range of ${10}^{-3.5}\mbox{--}{10}^{-1.5}$, indicating jet-dominated radio emission. By contrast, our sample has L_R/L_X between $\sim {10}^{-7}\mbox{--}{10}^{-4}$, more in line with expectations from coronal emission, which scales as approximately ${L}_{R}/{L}_{X}\sim {10}^{-5}$ (Laor & Behar 2008; Panessa et al. 2019).

If the radio emission is coronal in origin, then one might expect that the relationship between the X-ray luminosity, radio luminosity, and black hole mass would be more pronounced (i.e., a tighter FP), on account not only of the X-ray and radio emission originating in the same physical structure, but also that structure existing only a few gravitational radii from the black hole (e.g., Reis & Miller 2013). However, we have found that the correlation between the X-ray and parsec-scale radio emission in hard X-ray selected AGNs is rather poor, and is not improved by the introduction of the black hole mass. Indeed, the majority of our sample is not even detected in the radio on parsec scales, in stark contrast with lower-resolution studies (e.g., Smith et al. 2020), with L_R/L_X well below the fiducial 10⁻⁵ scaling for pure coronal emission (Laor & Behar 2008; Panessa et al. 2019). We suggest that there may be "radio-silent" AGNs: AGNs that do not even produce significant coronal synchrotron radio emission. Deeper follow-up studies of these undetected sources will prove critical to confirming this, but in either case, our results underscore the need to better understand and characterize the physical properties of the corona.

This object was suggested to be Compton-thick by Zhang et al. (2006), who fit Chandra X-ray Observatory (CXO) data with a plasma emission component plus a high-energy reflection model. Tzanavaris & Georgantopoulos (2007), using the same data, noted an Fe Kα line (6.4 keV) at high significance with an equivalent width (EW) of ∼1.5 keV. We use an APEC model to account for the plasma component, setting the temperature and abundance equal to the values in Zhang et al. (2006), and fit the spectrum using MYTorus. We obtain a column density N_H value of ${1.3}_{-1.3}^{+4.7}\times {10}^{23}$ cm⁻², suggesting that the source is heavily absorbed if not Compton-thick. However, the relative normalization between the scattered ("reflection") continuum and the intrinsic power-law component is over four orders of magnitude different between time-averaged BAT spectrum and the XRT data, despite their intrinsic power-law fits varying in normalization only marginally, which is unphysical. Conversely, if we assume that the relative normalization of the scattered continuum is constant between epochs, the best-fit N_H is unchanged but the intrinsic power-law normalization corresponding to the XRT data is ∼25% that of the BAT data, indicating considerable variability. Setting the strength of the scattered continuum to be fixed between the BAT and XRT epochs, which allows for the possibility of a time delay between a change in the intrinsic power-law continuum and a change in the scattered continuum, leaves this result unchanged, likely due to the weakness of the scattered continuum at low N_H values. However, the complex nature of the soft X-ray spectrum allows for a wide range of possible N_H values, as we have seen. To mitigate this issue, we downloaded and reprocessed the archival CXO data discussed above (Obs ID 3014; 2002 May 17; PI: Stevens) using ciao, version 4.10 with CALDB 4.7.8. Using a r = 15 aperture aperture centered on the AGN, and defining an $1\buildrel{\prime\prime}\over{.} 4\lt r\lt 3\buildrel{\prime\prime}\over{.} 5$ annular background region immediately surrounding the AGN to sample the diffuse soft X-ray emission that contaminates the low angular resolution XRT data (Figure 9), we extracted the isolated AGN spectrum using specextract. With the AGN isolated, we were able to delete the diffuse plasma APEC component, fitting only the MYTorus model. We find a hydrogen column density of ${N}_{{\rm{H}}}={1.4}_{-0.1}^{+0.3}\times {10}^{23}$ cm⁻², consistent with the best-fit value using the XRT data, but with significantly tighter confidence bounds. We therefore classify NGC 2782 as not being Compton-thick, contrary to the findings of Zhang et al. (2006). With N_H frozen to this value, we re-fit the XRT+BAT data, finding that the power-law normalization at the XRT epoch is 26% that of the BAT data, in agreement with our initial finding of variability between the epochs. Finally, we note that in addition to diffuse soft X-ray emission, there are five unresolved X-ray sources within the XRT aperture, consistent with being X-ray binaries (XRBs) or background AGNs (Figure 9). We extracted the background-isolated summed spectrum for all five, finding it to be well fit (${\chi }^{2}/\mathrm{dof}=1.03$) with a power-law component with Γ = 1.7 and ${N}_{{\rm{H}}}=6.5\times {10}^{20}$ cm⁻². The total 2–10 keV flux is 3.5 × 10⁻¹⁴ erg cm⁻² s⁻¹, $\sim 3 \%$ of the flux from the absorbed power-law component fit to the XRT data. We therefore do not consider these sources to be a significant contaminant in our results.

Zoom In Zoom Out Reset image size

This is a known Compton-thick object (e.g., Ricci et al. 2017), and has an extensive nuclear superbubble (e.g., Li et al. 2019) that contributes significantly to the X-ray spectrum at soft (<2 keV) energies. As with NGC 2782, we isolate the nuclear X-ray emission using CXO data (Obs ID 19307; 2018 January 30; PI: Li), using the immediate vicinity of the AGN for the background spectrum. We also found an archival NuSTAR observation (Obs ID 60061097002; 2013 November 12; PI: Harrison) taken as part of the NuSTAR extragalactic surveys (Harrison et al. 2016). Using MYTorus, we found a good fit (${\chi }^{2}/\mathrm{dof}=197.95/155$) with ${N}_{{\rm{H}}}={8.5}_{-1.0}^{+1.2}\times {10}^{24}$ cm⁻², although we have to fit the CXO and NuSTAR data above 4 and 7 keV, respectively, to avoid residual contamination from the diffuse X-ray emission seen in the CXO image (Figure 10). We show this spectrum in Figure 11. We note that we find evidence of variability in the intrinsic power-law spectral index Γ between the NuSTAR and BAT data, with the former having ${\rm{\Gamma }}={1.7}_{-0.2}^{+0.2}$ and the latter having ${\rm{\Gamma }}={2.2}_{-0.1}^{+0.1}$.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We then fit the XRT and BAT data separately, holding the MYTorus component parameters fixed to the best-fit values found above, and adding before the MYTorus component an Astrophysical Plasma Emission Code (APEC; Smith et al. 2001)¹⁴ model, using the default angr abundances (Anders & Grevesse 1989). With the exception of the normalization, we keep all APEC parameters frozen to those found in Li et al. (2019) by using the vapec variant. Although the intrinsic power-law continuum is of course not detectable in the XRT data for this bone fide Compton-thick object, precluding a measurement of the unabsorbed 2–10 keV flux contemporaneous with our VLBA observations, the XRT data do reveal the presence of additional X-ray emission above ∼1 keV that is not attributable to the plasma emission component. We find that this additional emission is consistent with an optically thin scattered continuum component, which can be interpreted as intrinsic power-law emission originating along an unobscured direction being scattered off free electrons back into our line of sight. We model this component as a power law with parameters held fixed to the intrinsic power-law component fit to the BAT data, with a coefficient to account for the scattering fraction. We find this value to be ∼0.1%. We show this spectrum in Figure 12.

Zoom In Zoom Out Reset image size

One archival CXO observation exists for this object (Obs ID 20622; 2018 January 24; PI: Maksym). The high angular resolution X-ray imaging enabled by these data reveal a compact source with little diffuse emission and few nearby XRBs (Figure 13). A joint analysis of the CXO and BAT spectrum gives an excellent fit (${\chi }^{2}/\mathrm{dof}=141.93/141$) with an APEC thermal plasma model appended to MYTorus, consistent with the findings of Eguchi et al. (2011), who performed a joint fit with Suzaku and Swift BAT 22-month catalog data (Tueller et al. 2010). However, by allowing the spectral index corresponding to the CXO data to vary, we find that this fit produces an unusually hard spectral index of ${{\rm{\Gamma }}}_{\mathrm{CXO}}={0.9}_{-0.4}^{+0.5}$, which given the uncertainties, may be due to the relative insensitivity of the CXO data to the intrinsic power-law continuum in the presence of heavy absorption. Tying Γ to the best-fit value of the BAT data group Γ = 1.96 is disfavored by an F-test, which gives $p=1.5\times {10}^{-3}$. Setting ${{\rm{\Gamma }}}_{\mathrm{CXO}}$ to the 90% upper confidence limit of 1.4 gives ${\chi }^{2}/\mathrm{dof}=144.71/142$ and ${N}_{{\rm{H}}}={6.1}_{-0.3}^{+0.4}\times {10}^{23}$ cm⁻², within 0.2 and 0.1 dex of the values reported in Eguchi et al. (2011) and Ricci et al. (2017), respectively. We show the best-fit spectrum in Figure 14.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The X-ray spectrum of NGC 4151 is known to be complex, containing a rich array of emission lines at soft ($\lt 1$ keV) energies (e.g., Schurch et al. 2004), as well as significant spectral variation (e.g., Zoghbi et al. 2019). The BAT spectrum exhibits downward curvature at the highest energies suggestive of reflection, and we initially fit the BAT spectrum with the exponentially cutoff power-law reflection model pexrav (Magdziarz & Zdziarski 1995). Holding the best-fit parameters fixed, we added the XRT data and appended the partial covering, partially ionized absorber component zxipcf as was used for NGC 4151 in Lubiński et al. (2016). To account for unresolved line emission below $\lt 1\,\,\mathrm{keV}$, we appended a simple blackbody component, but we do not attribute to its best-fit parameters any physical significance. We found that the fit is insensitive to the reflection scaling factor, with the best-fit value being statistically consistent with zero. We therefore replaced the pexrav component with a power law with a high-energy cutoff (zcutoffpl). The best-fit parameters yield ${N}_{{\rm{H}}}={1.9}_{-1.3}^{+0.4}\times {10}^{23}$ cm⁻² and ${\rm{\Gamma }}={1.54}_{-0.04}^{+0.04}$.

One CXO observation exists for this object (Obs ID 9438; 2008 November 16; PI: Walter), but the object is exceptionally faint, with only eight counts within a $1\buildrel{\prime\prime}\over{.} 5$ aperture centered on the AGN. Using an archival NuSTAR observation (Obs ID 60201038002), we performed a joint fit with the BAT 105-month spectrum using MYTorus. We find an excellent fit (${\chi }^{2}/\mathrm{dof}=107.31/107$ with ${N}_{{\rm{H}}}={7.3}_{-1.5}^{+1.7}\times {10}^{24}$ cm⁻², indicating that NGC 4180 is a bona fide Compton-thick AGN. We note that this result somewhat contradicts Ricci et al. (2017), who find ${N}_{{\rm{H}}}={1.4}_{-0.6}^{+0.6}\times {10}^{24}$ cm⁻² using a more phenomenological model and BAT 70-month catalog data (Baumgartner et al. 2013). We show the best-fit spectrum in Figure 15.

Zoom In Zoom Out Reset image size

As with NGC 4151, a downward spectral curvature at high energies is present in the X-ray spectrum of NGC 5506, suggesting reflection or a high-energy cutoff, and a partially ionized X-ray absorber is more consistent with the data than a simple photoelectric absorption model (i.e., phabs). However, unlike for NGC 4151, we find that the BAT spectrum is significantly better fit with a pexrav reflection model than a zcutoffpl cutoff power-law model (${\chi }_{\nu }^{2}=0.44$ versus 3.5), and the latter yields an unusually hard spectral index of Γ = 0.8, while the former yields Γ = 1.8.

The contemporaneous XRT data for this object, ObsID 00088600003, is exceptionally faint, with only five spectral data counts. Consequently, the XRT and BAT data do not preclude the possibility that this source is Compton-thick, and a naive fit using the MYTorus model gives ${N}_{{\rm{H}}}\sim 3\times {10}^{24}$ cm⁻², with a 90% confidence lower limit of 2 × 10²³ cm⁻² and the upper limit pegged at 10²⁵ cm⁻². To resolve this ambiguity, we used archival NuSTAR (Harrison et al. 2013) FPMA, FPMB data of NGC 7378, ObsID 60464202002 (2018 December 22; 20.3 ks; PI: Harrison), taken as part of the NuSTAR extragalactic surveys (Harrison et al. 2016), for which we extracted data products using the interactive analysis feature on the Space Science Data Center website.¹⁵

After including the NuSTAR data, we find that the X-ray spectrum of NGC 7378 is well described by an absorbed power law with ${N}_{{\rm{H}}}={5.8}_{-2.1}^{+2.2}\times {10}^{22}$ cm⁻² and ${\rm{\Gamma }}={1.7}_{-0.1}^{+0.1}$. The model normalizations of the power-law fits to the BAT and NuSTAR data are consistent with a $\sim 10 \%$ instrumental calibration uncertainty; however, the power-law normalization of the XRT data is considerably lower, being ∼26% of the value of the NuSTAR data normalization (Figure 16), with a 90% upper limit of 54%. This indicates potential variability between the NuSTAR and XRT epochs, which are separated by 4.5 months, and so we calculate the 2–10 keV flux corresponding to the XRT data normalization while holding the other spectral parameters to the values determined from the fit to all four data sets.

Zoom In Zoom Out Reset image size

This object is listed as borderline Compton-thick (${N}_{{\rm{H}}}\,=1.4\times {10}^{24}$ cm⁻²) by Ricci et al. (2017). We reprocessed an archival NuSTAR observation (Obs ID 60201037002; 2016 May 12; PI: Ricci) of NGC 7479 and jointly fit its BAT 105-month data using MYTorus above 3 keV, finding an excellent fit (${\chi }^{2}/\mathrm{dof}=76.99/77$) with ${\rm{\Gamma }}={2.47}_{-0.11}^{+0.12}$ and ${N}_{{\rm{H}}}\,={5.7}_{-0.5}^{+0.5}\times {10}^{24}$ cm⁻² for a torus inclination angle of θ = 79°, where 0^◦ is face-on. NGC 7479 is therefore fully Compton-thick, and we do not expect our XRT observations to have any sensitivity to the intrinsic power-law continuum. However, we briefly explore the soft X-ray properties of this object by downloading and reprocessing an archival CXO observation (Obs ID 11230; 2009 August 11; PI: Pooley). The high-resolution CXO observation reveals extensive diffuse emission near the AGN, as shown in Figure 17. We extract the soft X-ray spectrum using an aperture matched to that of NuSTAR. We find that the joint CXO, NuSTAR, and BAT spectrum is well fit (${\chi }^{2}/\mathrm{dof}=151.04/134$ with the addition of a single APEC plasma component with kT = 0.7 keV. The intrinsic power-law spectral index and column density are not significantly affected by the addition of the CXO data, although curiously, the relative normalization of the Fe Kα line in the CXO data is required to be about four times higher than in the other data. We show the fit to the full X-ray spectrum in Figure 18.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size