Toward a More Complete Optical Census of Active Galactic Nuclei via Spatially Resolved Spectroscopy

Eight of the off-nuclear Seyfert region galaxies were observed with Chandra/ACIS for the program GO9-20089X (PI: Comerford). We derived exposure times from the summed [O iii] λ5007 flux of the Seyfert spaxels in each system and the scaling relation between [O iii] λ5007 flux and hard X-ray (2–10 keV) flux for Type 2 AGNs, which has a scatter of 1.06 dex (Heckman et al. 2005). We selected exposure times that would ensure a firm detection of at least 3σ of each AGN and measurements of the extragalactic column density that are accurate to an order of magnitude or better. The galaxies were observed with exposure times of 20 ks to 35 ks (Table 2). The observations of SDSS J0743+4443 were split across two observing dates, and we registered each observation to SDSS broadband imaging (see below) and then used merge_obs to merge these two exposures. The merged observation was used for the image modeling described below.

For the remaining two off-nuclear Seyfert region galaxies in our sample, we analyzed archival Chandra observations. SDSS J1407+4428 (ObsID = 19990; PI: Secrest) was observed for 30 ks, while SDSS J1518+4244 (ObsID = 6858, PI: Komossa) was observed for 14 ks.

The galaxies were observed with the telescope aimpoint on the ACIS S3 chip in "timed exposure" mode and telemetered to the ground in "faint" mode. We reduced the data with the latest Chandra software (CIAO 4.13) in combination with the most recent set of calibration files (CALDB 4.9.4).

First, we registered the Chandra observations following the same approach as the MaNGA registration (Law et al. 2016), so that we can accurately compare the positions of the X-ray and optical sources. We registered against the SDSS broadband imaging in the g, r, i, and z bands, and used a biweight mean of the four bands as our final result.

For each galaxy, we used dmcopy to make a sky image of the field in the rest-frame soft (0.5–2 keV), hard (2–8 keV), and total (0.5–8 keV) energy ranges. Then, with the modeling facilities in Sherpa, we simultaneously modeled each X-ray source as a 2D Lorentzian function (beta2d: $f(r)=A(1+{[r/{r}_{0}]}^{2})-\alpha$) and the background as a fixed count rate estimated from an annulus of 25 width around the source region. We used the wavdetect source position that is closest to the galaxy centroid as the initial input position for each beta2d fit. Then, we allowed the model to fit a region of radius 25, which is more than double the radius of the Chandra PSF. To determine the best-fit model parameters, we used Sherpa's implementation of the "Simplex" minimization algorithm (Lagarias et al. 1998) and minimized the Cash statistic.

Our sample includes two merging galaxy systems with two stellar bulges each, SDSS J1407+4428 and SDSS J1518+4244, and we attempted a two-component beta2d model to test for additional X-ray sources in these systems. We set the initial positions of the two beta2d components to the stellar bulge positions and allowed them to wander within a circle of radius 25. In SDSS J1518+4244 we detected two X-ray sources with significances >3σ above the background, and these two sources are spatially coincident with the two stellar bulges. The extraction regions of 25 radius did not overlap for these two sources. For SDSS J1407+4428, we detected an X-ray source at the position of the eastern nucleus with a significance 32σ above the background. We found a western source with a significance of 2.6σ above the background, and since this is a <3σ result we do not classify it as a detection. In total, we find 11 X-ray sources with significances >3σ above the background: the eight nonmerging galaxies each have an X-ray source coincident with their centers, SDSS J1407+4428 has an X-ray source coincident with the eastern nucleus, and SDSS J1518+4244 has two X-ray sources (one coincident with each of the two stellar bulges).

Next, we used the Bayesian Estimation of Hardness Ratios (BEHR) code (Park et al. 2006) to measure the rest-frame soft, hard, and total counts in each X-ray source. We used calc_data_sum to determine the number of observed soft and hard counts from both the source region and a background region, and then BEHR uses a Bayesian approach to estimate the expected values and uncertainties of the rest-frame soft counts and rest-frame hard counts. Table 3 shows these values.

Table 3. Chandra Source Properties in Different Energy Ranges

SDSS Name	Chandra Rest-frame	Counts	LX,abs	LX,unabs
	Energy Range (keV)		(1040 erg s−1)	(1040 erg s−1)
SDSS J0743+4443	0.5–2	${8.8}_{-3.7}^{+2.4}$	${0.3}_{-0.6}^{+0.2}$	${0.6}_{-0.7}^{+0.4}$
	2–8	${4.6}_{-2.9}^{+1.6}$	${0.4}_{-0.9}^{+0.3}$	${0.9}_{-1.0}^{+0.6}$
SDSS J0745+4604	0.5–2	${93.5}_{-10.4}^{+9.0}$	${4.2}_{-1.4}^{+1.0}$	${9.7}_{-2.1}^{+2.1}$
	2–8	${93.8}_{-10.4}^{+9.1}$	${8.7}_{-2.8}^{+2.2}$	${20.1}_{-4.3}^{+4.3}$
SDSS J0931+4904	0.5–2	${32.2}_{-6.3}^{+5.1}$	${2.4}_{-0.9}^{+0.7}$	${3.1}_{-1.0}^{+0.9}$
	2–8	${22.5}_{-5.4}^{+4.0}$	${3.0}_{-1.1}^{+0.9}$	${3.9}_{-1.2}^{+1.1}$
SDSS J1254+4409	0.5–2	${8.8}_{-3.7}^{+2.4}$	< 3.7	< 500
	2–8	${23.2}_{-5.5}^{+4.4}$	${14.6}_{-9.5}^{+7.0}$	${147.6}_{-66.1}^{+67.3}$
SDSS J1407+4428 E	0.5–2	${108.1}_{-11.0}^{+9.7}$	${239.3}_{-56.4}^{+51.0}$	${320.3}_{-66.9}^{+65.6}$
	2–8	${930.5}_{-30.4}^{+30.7}$	${2538.9}_{-598.5}^{+541.3}$	${3398.3}_{-709.3}^{+696.1}$
SDSS J1430+5242	0.5–2	${1036.0}_{-34.4}^{+29.8}$	${158.9}_{-19.6}^{+18.5}$	${216.6}_{-23.9}^{+21.1}$
	2–8	${1747.4}_{-42.7}^{+41.1}$	${575.0}_{-70.8}^{+67.0}$	${783.9}_{-86.5}^{+76.2}$
SDSS J1518+4244 NW	0.5–2	${70.8}_{-9.8}^{+8.4}$	${5.3}_{-3.9}^{+2.6}$	${10.9}_{-4.9}^{+4.9}$
	2–8	${23.5}_{-5.8}^{+4.6}$	${4.4}_{-3.3}^{+2.2}$	${9.1}_{-4.1}^{+4.1}$
SDSS J1518+4244 SE	0.5–2	${70.8}_{-9.8}^{+8.4}$	${0.9}_{-0.3}^{+0.3}$	${0.9}_{-0.3}^{+0.3}$
	2–8	${23.5}_{-5.8}^{+4.6}$	${1.0}_{-0.4}^{+0.4}$	${1.1}_{-0.4}^{+0.4}$
SDSS J1601+4521	0.5–2	${16.8}_{-4.7}^{+3.5}$	${3.4}_{-0.9}^{+1.0}$	${4.0}_{-1.0}^{+1.2}$
	2–8	${22.2}_{-5.4}^{+4.3}$	${4.9}_{-1.2}^{+1.4}$	${5.7}_{-1.4}^{+1.7}$
SDSS J1630+2612	0.5–2	${16.5}_{-4.6}^{+3.8}$	${9.3}_{-5.7}^{+3.9}$	${25.7}_{-12.2}^{+10.}$
	2–8	${540.9}_{-24.4}^{+22.3}$	${1563.5}_{-963.8}^{+657.1}$	${4318.7}_{-2048.1}^{+1677.7}$
SDSS J1633+3911	0.5–2	${12.5}_{-4.2}^{+2.8}$	${2.3}_{-8.4}^{+1.7}$	${4.9}_{-9.4}^{+3.4}$
	2–8	${6.8}_{-3.3}^{+2.0}$	${0.7}_{-2.5}^{+0.5}$	${1.5}_{-2.8}^{+1.0}$

Download table as:  ASCIITypeset image

We then used Sherpa to model the energy spectra of the extracted regions over the observed energy range 0.5–7 keV. We fit each unbinned spectrum with a redshifted power law, F ∼ E^−Γ, which represents the intrinsic AGN X-ray emission at the galaxy redshift. This spectrum is attenuated by passing through two absorbing column densities of neutral hydrogen. One of these is fixed to the Galactic value, which we determined using an all-sky interpolation of the H i in the Galaxy (Dickey & Lockman 1990), and the other (n_H) is assumed to be intrinsic to the source at the galaxy redshift.

For the fit to each spectrum, we allowed Γ and n_H to vary freely. For three sources, we found that the best-fit value of Γ was not within the typical range of observed power-law indices, i.e., 1 ≤ Γ ≤ 3 (Nandra & Pounds 1994; Reeves & Turner 2000; Piconcelli et al. 2005; Ishibashi & Courvoisier 2010), and for these we redid the fits with Γ fixed at a value of 1.8, which is a typical value for the continuum of Seyfert galaxies.

To determine the best-fit model parameters for each spectrum, we used Sherpa's implementation of the Levenberg–Marquardt optimization method (Bevington 1969) to minimize the Cash statistic. For SDSS J0743+4443 the observations were split across two observing dates, and so we extracted a spectrum from each observation and modeled them simultaneously. Table 4 shows the results of these spectral fits. For the three sources where we redid the fits with fixed Γ = 1.8, we use the results of the Γ = 1.8 fits for the analyses that follow.

Table 4. Chandra Spectral Fits

SDSS Name	nH	Γ	Reduced
	(1021 cm−2)		C-stat
SDSS J0743+4443	<3E-6	${2.3}_{-0.8}^{+0.6}$	0.09
SDSS J0745+4604	<6E-7	${2.5}_{-0.2}^{+0.1}$	0.53
SDSS J0931+4904	<7E-7	${2.1}_{-0.3}^{+0.2}$	0.28
SDSS J1254+4409	${10.8}_{-4.0}^{+4.5}$	${9.0}_{-1.6}^{+2.6}$	0.30
	${365.4}_{-103.1}^{+120.8}$	1.8a	0.25
SDSS J1407+4428 E	${10.4}_{-2.1}^{+2.2}$	${1.1}_{-0.1}^{+0.1}$	0.69
SDSS J1430+5242	${3.0}_{-0.6}^{+0.6}$	${1.6}_{-0.1}^{+0.1}$	0.76
SDSS J1518+4244 NW	${3.6}_{-1.6}^{+1.6}$	${2.8}_{-0.5}^{+0.5}$	0.30
SDSS J1518+4244 SE	${3.9}_{-2.7}^{+2.5}$	${3.8}_{-1.0}^{+0.9}$	0.15
	${0.0}_{-0.6}^{+0.5}$	1.8 a	0.16
SDSS J1601+4521	<4E-7	${0.5}_{-0.5}^{+0.4}$	0.25
	<6E-7	1.8a	0.25
SDSS J1630+2612	${89.8}_{-10.9}^{+6.8}$	${1.5}_{-0.2}^{+0.1}$	0.68
SDSS J1633+3911	<1E-6	${3.0}_{-1.7}^{+1.6}$	0.14

Note.

^a The best-fit spectrum had Γ outside of the usual AGN range 1 ≤ Γ ≤ 3, so we redid the fit by freezing Γ = 1.8. For these three sources, we use the results of the Γ = 1.8 spectral fits in the rest of our analyses.

Download table as:  ASCIITypeset image

All fluxes are k-corrected, and we calculated the observed flux values from the model sum (including the absorbing components) and the intrinsic flux values from the unabsorbed power-law component. Then, we used the redshift to determine the distance to each system and convert the X-ray fluxes to X-ray luminosities (Table 3). Finally, we converted the rest-frame 2–10 keV luminosities to bolometric luminosities by multiplying by a factor of 20, which is a typical bolometric correction for AGNs (e.g., Elvis et al. 1994; Marconi et al. 2004).

Our analyses of the MaNGA galaxies rely on galaxy properties measured in the Pipe3D Value Added Catalog (Sánchez et al. 2016, 2018), including galaxy stellar mass and star formation rate (SFR; derived from Hα measurements). For the two merging galaxy systems, we used the stellar mass of each bulge as measured from best-fit models of the nuclear spectra in Fu et al. (2018). For these two merging galaxy systems, we measured a separate SFR for an aperture of 25 diameter centered on each stellar bulge. We summed the Hα flux in each aperture and converted to SFR via the relation between Hα luminosity and SFR given in Kennicutt & Evans (2012), which is the same approach used to measure the SFRs in Pipe3D that we use for the remaining eight galaxies. These values are shown in Table 1, and we note that they may be overestimates of the SFR if an AGN is present and contributing to the Hα luminosity.

We also used the emission line fluxes measured in MaNGA's Data Analysis Pipeline (DAP; Westfall et al. 2019). We use the nonparametric summed fluxes of the emission lines, where these fluxes are measured after subtraction of the stellar continuum model, and Galactic reddening is also accounted for. Further, we use the DAP's nonparametric equivalent width measurements of Hα.

Finally, we also use the MaNGA emission line fluxes to measure the Balmer decrement, Hα/Hβ, which provides a measurement of the dust attenuation (Calzetti et al. 2000; Yuan et al. 2018). We then use the Balmer decrement to determine the color excess E(B − V) (Yuan et al. 2018), and in Table 1 we present our measurements of E(B − V) in the spaxel corresponding to the central 25 (∼1.5–6 kpc) of each galaxy's stellar bulge(s).

To determine whether the Chandra observations indicate the presence of an X-ray AGN, we first account for the X-ray binary (XRB) contribution to the hard X-rays. For each galaxy we calculate the predicted hard X-ray luminosity from XRBs L_{2–10keV,XRB}, as a function of galaxy stellar mass and SFR, using the relations in Lehmer et al. (2010). We use the galaxy stellar masses and SFRs described in Section 3.2, which includes separate stellar mass and SFR measurements for each bulge in the merging galaxy systems. We then determine the uncertainties on L_{2–10keV,XRB} from propagating the scatter (0.34 dex) in the Lehmer et al. (2010) relation, the coefficient uncertainties in the Lehmer et al. (2010) relation, the uncertainties on galaxy stellar mass, and the uncertainties on SFR. We define an X-ray AGN detection as those cases where L_{2–10keV,unabs} is more than 3σ greater than L_{2–10keV,XRB} (e.g., Barrows et al. 2019). By this definition, all 11 X-ray detections presented in this paper are X-ray AGNs (Table 5).

Table 5. Chandra Evidence for AGN

SDSS Name	L2−10keV,XRB	L2−10keV,unabs	L2−10keV,XRB/	σ	Chandra Evidence
	(1040 erg s−1)	(1040 erg s−1)	L2−10keV,unabs	Significance	for AGN?
SDSS J0743+4443	0.4 ± 0.2	${1.1}_{-1.3}^{+0.7}$	0.39	4	yes a
SDSS J0745+4604	1.2 ± 0.2	${24.4}_{-5.2}^{+5.2}$	0.05	>100	yes
SDSS J0931+4904	2.1 ± 0.3	${5.0}_{-1.6}^{+1.4}$	0.43	11	yes a
SDSS J1254+4409	2.3 ± 0.2	${293.1}_{-131.3}^{+133.6}$	<0.01	>100	yes
SDSS J1407+4428 E	0.3 ± 0.2	${5478.5}_{-1143.9}^{+1122.0}$	<0.01	>100	yes
SDSS J1430+5242	3.2 ± 0.3	${1100.9}_{-121.5}^{+107.1}$	<0.01	>100	yes
SDSS J1518+4244 NW	0.4 ± 0.2	${10.5}_{-4.7}^{+4.7}$	0.04	55	yes
SDSS J1518+4244 SE	0.1 ± 0.2	${1.5}_{-0.6}^{+0.5}$	0.06	9	yes
SDSS J1601+4521	1.2 ± 0.2	${7.6}_{-1.9}^{+2.2}$	0.15	33	yes
SDSS J1630+2612	4.7 ± 0.4	${7080.3}_{-3358.1}^{+2750.7}$	<0.01	>100	yes
SDSS J1633+3911	0.4 ± 0.2	${1.7}_{-3.2}^{+1.2}$	0.23	7	yes a

Note. Column 2 shows the estimated rest-frame 2–10 keV luminosity from XRBs (Section 4.1). Column 3 shows the rest-frame, unabsorbed 2–10 keV luminosity that we measure from the Chandra observations (Section 3.1). Column 4 is the ratio of the estimated rest-frame 2–10 keV luminosity from XRBs to the rest-frame, unabsorbed 2–10 keV luminosity. Column 5 indicates the σ significance with which L_{2−10keV,unabs} is greater than L_{2−10keV,XRB}. Column 6 indicates whether an AGN is present: if L_{2−10keV,unabs} is more than 3σ (Column 5) greater than L_{2−10keV,XRB}, then this is evidence for an AGN.

^a These galaxies have weaker AGN detections, since the XRBs contribute more than 20% to the total unabsorbed 2–10 keV luminosity and the significance of the X-ray luminosity in excess of the expected contribution from XRBs is ≲10σ.

Download table as:  ASCIITypeset image

However, for three galaxies (SDSS J0743+4443, SDSS J0931+4904, and SDSS J1633+3911), the XRBs contribute more than 20% to the total unabsorbed 2–10 keV luminosity and the significance of the X-ray luminosity in excess of the expected contribution from XRBs is ≲10σ. To be conservative, we label these three sources as weaker AGN detections (Table 5). Therefore, 7–10 (70%–100%) of the galaxies in our sample have at least one confirmed X-ray AGN, depending on how conservative our criteria are.

While BPT diagrams mainly focus on SF and AGNs as the ionization sources producing emission lines, here we broaden the analysis of these MaNGA galaxies by also considering ionization by post-AGB stars and shocks. Post-AGB stars are capable of producing hard ionized spectra (e.g., Binette et al. 1994; Yan & Blanton 2012), but they are not thought to produce Hα equivalent widths greater than ∼3 Å (e.g., Belfiore et al. 2016). Consequently, a cutoff of Hα equivalent width <3 Å can identify regions of a galaxy that are predominantly ionized by post-AGB stars (Cid Fernandes et al. 2010). We present such individual spaxel-by-spaxel maps in Figure 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The [O i] λ6300 emission line is an indicator of shocks (e.g., Dopita 1976; Allen et al. 2008; Rich et al. 2011), and line flux ratios of [O i] λ6300/Hα > 0.1 indicate that shocks with velocities of 160–300 km s⁻¹ are the main excitation source of [O i] λ6300 (Riffel et al. 2021). Further, the line flux ratio cutoff [S ii] λλ6717, 6731/Hα > 0.4 has also been established empirically to be an identifier of supernova remnant (SNR) shocks (Dodorico 1978; Dodorico et al. 1980). Figure 1 shows individual spaxel-by-spaxel maps of these emission line ratio cutoffs for the MaNGA galaxies in our sample, and we use them for the analyses of the individual galaxies that follow (Section 6).

We selected our galaxies from the Wylezalek et al. (2018) sample of AGN candidates in MaNGA, and we found that all of them have AGNs based on Chandra observations. Now, we explore whether these systems have AGN detections in other wavelength regimes and using other approaches. Table 6 summarizes our results.

First, we crossmatch our sample to the MaNGA AGN catalog (Comerford et al. 2020), which is a catalog of MaNGA galaxies where an AGN was detected via mid-infrared colors of the Wide-field Infrared Survey Explorer (WISE) (using color cuts from Assef et al. 2018), Swift/BAT ultrahard X-ray detections (using the 105 month BAT catalog of AGNs; Oh et al. 2018), NRAO Very Large Array Sky Survey (NVSS) and Faint Images of the Radio Sky at Twenty Centimeters (FIRST) radio observations (using the catalog of radio sources corresponding to AGNs; Best & Heckman 2012), and broad emission lines detected in single-fiber SDSS spectra (Oh et al. 2015). The radio sources are further subdivided into high-excitation radio galaxies (HERGs) and low-excitation radio galaxies (LERGs). Each data set in the MaNGA AGN catalog covers the full sample of MaNGA galaxies considered here. While we find that four of the galaxies in our sample have AGNs identified by WISE colors, none of them are identified as AGNs by Swift/BAT, radio HERG or LERG classifications, or broad emission lines (which would only identify Type I AGNs).

Next, we crossmatch our sample to MaNGA catalogs of AGN candidates that were selected spectroscopically. Sánchez et al. (2018) selected AGN candidates based on the integrated spectra of the central 3'' × 3'' of MaNGA galaxies. To classify an object as an AGN candidate, they require that this central integrated spectrum have emission line ratios that lie above the theoretical maximum for starbursts in the BPT diagram (Kewley et al. 2001, 2006) and an Hα equivalent width that is >1.5 Å (Cid Fernandes et al. 2010). Out of 2755 galaxies in MaNGA MPL-5, they identify 98 AGN candidates. We find that five of our galaxies are AGN candidates based on these criteria.

In addition, Rembold et al. (2017) searched for AGNs in MaNGA using both BPT and WHAN (where Hα equivalent widths >3 Å identify ionization by AGNs, and not ionization by post-AGB stars; Cid Fernandes et al. 2010) diagnostics simultaneously. Using the SDSS-III integrated nuclear spectra of the 2778 galaxies observed in MaNGA MPL-5, they identify 62 "true" AGNs whose line flux ratios and Hα equivalent widths lie in the Seyfert or LINER regions of both the BPT and WHAN diagrams. Five of our galaxies are included in this sample of AGNs.

We also crossmatch our sample to a study of coronal line-emitting galaxies in MaNGA, where coronal lines are emission lines with high ionization potentials (≳100 eV) that are suggestive of AGN activity. Negus et al. (2021) use the full MaNGA footprint of each galaxy to search for emission from one or more coronal lines ([Ne v] λ3347, [Ne v] λ3427, [Fe vii] λ3586, [Fe vii] λ3760, or [Fe vii] λ6086) detected at ≥5σ above the background continuum in at least 10 spaxels. Using these criteria, they find 10 coronal line-emitting galaxies out of 6263 galaxies observed in MaNGA MPL-8. None of these 10 are galaxies in our sample.

Finally, we note that two galaxies (SDSS J0743+4443 and SDSS J1430+5242) host AGNs that are identified by Chandra but not by any other approach, and in Section 6 we suggest that these AGNs are flickering, where the X-rays are the signature of current AGN activity.

Since the X-ray and optical classifications of these 10 galaxies are different, we compare the hard X-ray luminosity and optical emission line luminosity (in this case, [O iii] λ5007 luminosity) to explore whether this galaxy sample is unusual in its X-ray or optical luminosities. We use the [O iii] λ5007 observed fluxes of the central 3'' spectra from SDSS DR7, as measured by the OSSY catalog (Oh et al. 2011), and the absorbed 2–10 keV luminosities measured in Section 3.1.

The fiducial relation between these two luminosities for Type 2 AGNs is log(L_X,2–10keV/L_{[O III]λ5007}) = 0.57 ± 1.06, based on a sample of 29 Type 2 AGNs at z < 0.2 (Heckman et al. 2005). However, Type 2 dual AGNs fall systematically below this relation, suggesting that mergers drive excess gas onto the AGNs, which increases their [O iii] λ5007 luminosities, and/or that mergers induce higher nuclear gas columns that suppress the hard X-ray luminosities of the AGNs (Liu et al. 2013; Comerford et al. 2015; Barrows et al. 2017). This trend for dual AGNs illustrates that the location of an AGN on this plot can reveal unusual circumstances.

As Figure 2 shows, we find that our sample lies within the rms scatter of the fiducial relation for Type 2 AGNs, with the exceptions of two galaxies: SDSS J1630+2612 and SDSS J1430+5242. For these two galaxies, the relatively high hard X-ray luminosities and low [O iii] luminosities could be due to AGN flickering. In this scenario, the spatially extended AGN emission (traced by [O iii]) may be left over from past AGN activity (e.g., Eracleous et al. 1995; Lintott et al. 2009) while the central AGN emission (traced by hard X-rays) may be a sign of current AGN activity.

Zoom In Zoom Out Reset image size

We note that dust reddening in the centers of host galaxies may also contribute, since the optical [O iii] luminosity would be more affected by dust than the X-ray luminosity (since the X-ray sources are Compton thin; Table 4).

To understand whether the stellar masses or AGN bolometric luminosities of the off-nuclear Seyfert region galaxies are unusual, we compare them to the MaNGA AGN catalog, which is a catalog of MaNGA AGNs detected via WISE, Swift/BAT, radio, and/or broad emission lines (Comerford et al. 2020). The stellar mass ranges of the off-nuclear Seyfert region galaxies and the MaNGA AGN catalog galaxies are the same; the median log stellar mass is 11.1 for both samples. According to a Kolmogorov–Smirnov test, there is an 82% probability that the stellar masses of these two samples were drawn from the same distribution.

However, as Figure 3 shows, on average the AGN bolometric luminosities of the off-nuclear Seyfert region galaxies are significantly lower than the AGN bolometric luminosities of the MaNGA AGN catalog galaxies. We conclude that the off-nuclear Seyfert regions are signatures of relatively low-luminosity AGNs, where the AGN radiation in the central 3'' spectrum is not the dominant radiation (Section 6).

Zoom In Zoom Out Reset image size

Two of the AGNs (SDSS J1407+4428 and SDSS J1630+2612) have ${\rm{log}}({L}_{{\rm{bol}}}/{\rm{erg}}\,{{\rm{s}}}^{-1})\gt 45$ and they fall in the same general parameter space for bolometric luminosity and stellar mass as the MaNGA AGN catalog (Figure 3); they are also WISE-detected AGNs and we note that their WISE-derived bolometric luminosities (taken from Comerford et al. 2020) are consistent with their Chandra-derived bolometric luminosities (Table 7).

There are two other AGNs (SDSS J0931+4904 and SDSS J1518+4244) that have WISE-derived bolometric luminosities that are two orders of magnitude higher than their Chandra-derived bolometric luminosities (Table 7). Since the WISE-derived bolometric luminosities are based on a mid-infrared to X-ray conversion that is not well constrained at the lower X-ray luminosities of these two AGNs (Stern 2015), we suggest that the Chandra-derived bolometric luminosities are the more accurate luminosities for these two systems. Dust heated by star formation or by shocks (e.g., merger-induced shocks) can contribute to the mid-infrared emission (e.g., Padovani et al. 2017; Barrows et al. 2021), which may also partially explain the high WISE-derived bolometric luminosities for the star-forming galaxy SDSS J0931+4904 (SFR = 8 M_⊙ yr⁻¹; Table 1) and the merger SDSS J1518+4244.

Our parent sample is the 2727 galaxies that have been observed in MaNGA MPL-5. From the single-fiber, central 3'' spectrum alone, 130/2727 = 5% of these galaxies are classified as hosting AGNs (Section 2). However, another 173 galaxies were identified as AGN candidates based on spatially resolved BPT and Hα analyses of their full maps of spaxels (for details see Section 2; Wylezalek et al. 2018), even though their single-fiber central 3'' spectrum did not show AGN signatures. Here, we have studied a sample of 10 of these galaxies, and found that 70%–100% of them indeed host AGNs (Section 4.1).

If we assume that these results are representative of the full population of 173 AGN candidates, then another 121–173 galaxies (70%–100% of 173) in this sample may host AGNs. In that case, the full AGN fraction would be (251–303)/2727 ∼ 10%. With the caveat that our sample is small, our results suggest that single-fiber BPT diagnostics are potentially missing a significant number (up to a factor of 2) of AGNs in galaxies.

We find that there are multiple, often overlapping, reasons for single-fiber BPT diagnostics to fail to identify a complete population of AGNs. First, the BPT classification is highly sensitive to the size and placement of the fiber: a fiber centered on a galaxy's nucleus may miss AGNs that are located outside the fiber diameter, and even in the case of a central AGN the BPT diagnostic can be dominated by emission from SF, post-AGB stars, shocks, and other ionization sources, depending on the fiber size and placement. This domination of other (non-Seyfert) emission sources is heightened if the AGN has intrinsically low luminosity or is obscured by dust. We also find evidence for AGN flickering in some of these galaxies, which is another avenue for other emission sources to dominate the BPT classification. While the shutdown is a brief phase in an AGN's lifetime (the light echo is visible for ∼10⁴–10⁵ yr; Schawinski et al. 2015), it may be that selecting AGN candidates by their off-nuclear Seyfert signatures (as we have here) introduces a bias toward flickering AGNs.

This galaxy hosts a weak AGN (Table 5), and it also was not classified as an AGN by any other set of observations (Table 6). We interpret these results as evidence for a flickering AGN. This explains why the X-ray source is weak, and why the central 3'' spectrum shows evidence of SF but not AGN activity. Further, the BPT map shows off-nuclear Seyfert emission outside the plane of the host galaxy (Figure 1), and this optical emission could be a light echo of past AGN activity.

The LINER-like emission line ratios in this galaxy's central region are likely due to a combination of post-AGB stars and shocks, as the diagnostics in Figure 1 show.

This galaxy hosts a weak AGN (Table 5) and has a relatively high SFR (8 M_⊙ yr⁻¹; Table 1), which implies that the SF may be dominating over the AGN emission in the central region of the galaxy. This would then explain the SF-like emission line ratios observed in the central 3'' spectrum.

This galaxy has evidence of shocks in its center (Figure 1), which can explain why the central 3'' spectrum is classified as LINER.

SDSS J1407+4428 is a major merger with two stellar bulges that have a projected separation of 8 kpc and a line-of-sight velocity difference of ∼50 km s⁻¹ (Ellison et al. 2017 measured ∼30 km s⁻¹, while Fu et al. (2018) measured 76.8 km s⁻¹).

Using the MaNGA data, Fu et al. (2018) extracted a spectrum from a circular aperture of 2.6 kpc diameter centered on each stellar bulge. They classified the spectra using the [N ii]-BPT diagnostic in combination with the WHAN diagram that relates Hα equivalent width and [N ii]/Hα (Cid Fernandes et al. 2010). They found that the western nuclear spectrum is Seyfert, while the eastern nuclear spectrum is LINER. We note that this result differs from the Portsmouth [S ii]-BPT classification that we use, which classifies the western nuclear spectrum as LINER. Fu et al. (2018) classify this system as "apparent" binary AGNs, since the nuclear spectra could be explained by merger-driven shocks, a single AGN cross-ionizing its companion galaxy, or two separate AGNs.

Ellison et al. (2017) also analyzed the MaNGA observations of this system and found, via [N ii]-BPT diagnostics (and supported by [O i]- and [S ii]-BPT diagnostics), that the emission coincident with both of the stellar bulges is consistent with AGN photoionization. This differs from the Portsmouth [S ii]-BPT classification of the western nuclear spectrum as LINER, perhaps because the Portsmouth classification uses a 3'' fiber whereas the Ellison et al. (2017) classification does not use an aperture.

In their analysis of the Chandra/ACIS observations, Ellison et al. (2017) found two hard X-ray sources associated with the two stellar bulges. For the eastern source, they measured L_{2–10keV,unabs} = (3.5 ± 0.4) × 10⁴³ erg s⁻¹ and concluded that it is an AGN. This is similar to our measurement of L_{2–10keV,unabs} = (5.5 ± 1.1) × 10⁴³ erg s⁻¹, which we also used to conclude that the eastern X-ray source is an AGN.

For the western source, Ellison et al. (2017) placed a 3'' aperture on the position of the western galaxy bulge and detected two soft (0.3–8 keV) and 10 hard (2–8 keV) X-rays. From this, they estimated a hardness ratio of 0.58 and L_{2–10keV,unabs} = (5.5 ± 1.1) × 10⁴³ erg s⁻¹ (assuming a power-law spectrum with Γ = 1.8). Based on the optical emission line ratio diagnostics, hardness ratio, and X-ray luminosity, they concluded that the western X-ray source is likely an AGN. Our search for a western X-ray source did not return a source detected at >3σ above the background (Section 3.1), so we cannot claim a detection of a western X-ray source.

In this system, the off-nuclear Seyfert region observed in the MaNGA data is explained by an AGN (the eastern X-ray source) in the nucleus of a companion galaxy, so that the Seyfert emission does indeed spatially coincide with an AGN. The LINER-like emission in the western source is then explained by post-AGB stars and shocks (Figure 1).

We interpret this galaxy as hosting a flickering AGN. Since the source was not classified as an AGN by any other set of observations (Table 6), and the BPT map reveals spatially extended Seyfert emission with a hole at the middle of the galaxy (Figure 1), the spatially extended emission could be an extended light echo of past AGN activity. The X-ray luminosity is also higher than the canonical scaling relation between L_X,2–10keV and L_{[O iii] λ5007} for AGNs (Section 5.2), which is evidence of new, current AGN activity. The central bulge is obscured (E(B − V) = 0.78; the second highest obscuration in our sample), and the SFR is relatively high (12 M_⊙ yr⁻¹; Table 1), which would then explain why the central 3'' spectrum is dominated by SF.

SDSS J1518+4244 is a major merger with two stellar bulges that have a projected separation of 5.5 kpc and a line-of-sight velocity difference of 91.9 km s⁻¹ (Fu et al. 2018). Using the MaNGA data, Fu et al. (2018) extracted a spectrum from a circular aperture of 2.6 kpc diameter centered on each stellar bulge and classified the spectra using the [N ii]-BPT diagnostic along with the WHAN diagram (Cid Fernandes et al. 2010). They classified both nuclear spectra as starburst-AGN composite, which is different from the Portsmouth [S ii]-BPT classification of the northwestern nuclear spectrum as star-forming. Fu et al. (2018) concluded that this system hosts "apparent" binary AGNs, where the nuclear spectra could be explained by merger-driven shocks, a single AGN cross-ionizing its companion galaxy, or two separate AGNs.

Here, we use Chandra observations to confirm that there are indeed dual AGNs in this galaxy merger (Table 5). The 3'' SDSS fiber is centered on the northwest AGN, and that stellar bulge is obscured (E(B − V) = 0.81; the highest obscuration in our sample), which helps explain why that fiber spectrum is dominated by SF-like emission line ratios.

This galaxy has signs of shocks in its center (Figure 1), which explain why the central 3'' spectrum is classified as LINER.

This galaxy's BPT map shows off-nuclear Seyfert emission outside the plane of the host galaxy (Figure 1), and the AGN has an X-ray luminosity that is higher than the canonical scaling relation between L_X,2–10keV and L_{[O iii] λ5007} for AGNs (Section 5.2). Consequently, this galaxy could host a flickering AGN that shut off in the past and left a light echo behind, and then turned on again and produced the X-ray emission. The SFR is relatively high (17 M_⊙ yr⁻¹; Table 1), which would explain the SF classification of the central 3'' spectrum.

This galaxy hosts a weak AGN (Table 5) and its central region has signatures of shock emission (Figure 1), which explains why the central 3'' spectrum is classified as LINER.