The Origin of Double-peaked Narrow Lines in Active Galactic Nuclei. IV. Association with Galaxy Mergers

Of the 95 z > 0.1 galaxies in our sample, we previously observed 54 of them with the Kast Spectrograph at Lick Observatory (3 m telescope, pixel size 078; Miller & Stone 1993), the Double Spectrograph (DBSP) at Palomar Observatory (5 m telescope, pixel size 047 for the red detector and 039 for the blue detector; Oke & Gunn 1982), and the Blue Channel Spectrograph at MMT Observatory (6.5 m telescope, pixel size 029; Angel et al. 1979). We used 1200 lines mm⁻¹ gratings, and slit widths of 15, 15, and 1'' for the Lick, Palomar, and MMT observations, respectively.

In addition, we obtained optical longslit observations of 41 more z > 0.1 double-peaked AGNs with GMOS-North on the Gemini Observatory North 8 m telescope (pixel size 0073 for 1 × 1 pixel binning and 0145 for 1 × 2 pixel binning; Allington-Smith et al. 2002; Hook et al. 2004), GMOS-South on the Gemini Observatory South 8 m telescope (pixel size 0073 for 1 × 1 pixel binning and 0146 for 1 × 2 pixel binning), LRIS on the Keck I 10 m telescope (pixel size 0135; Oke et al. 1995), and DEIMOS on the Keck II 10 m telescope (pixel size 0119; Faber et al. 2003).

We observed 20 objects with Gemini/GMOS-N and 11 objects with Gemini/GMOS-S, and with both telescopes we used a 600 lines mm⁻¹ grating and 1'' slit width. We observed 9 objects with Keck/LRIS; 4 of them were observed with the red and blue cameras simultaneously, and 5 were observed with the blue camera only because the red camera was broken. We used a 1200 lines mm⁻¹ grating for the red camera and a 600 lines mm⁻¹ grating for the blue camera, and a 1'' slit width. We observed one object with Keck/DEIMOS, using a 1200 lines mm⁻¹ grating and 1'' slit width.

Table 1 summarizes the observations. We observed each target with two different slit position angles. Typically, we observed with the slit oriented along the isophotal position angle of the major axis of the object in SDSS r-band photometry and again with the slit at the corresponding orthogonal position angle. In some cases we deviated from this approach due to guide star requirements or mechanical constraints on the rotation of the telescope.

The data were reduced following standard procedures in IRAF and IDL (see Tody 1993; Comerford et al. 2012; Cooper et al. 2012).

[O iii] λ5007 is a strong emission line that is an excellent tracer of the extent and kinematics of ionized gas in the NLR, and has been used in many studies of AGN photoionization and outflows (e.g., Schmitt et al. 2003; Greene et al. 2014; Sun et al. 2017). Here, we measure kinematic parameters of the observed [O iii] λ5007 emission in each galaxy following the approach presented in Nevin et al. (2016), who analyzed a sample of 71 galaxies with double-peaked AGN emission lines at z < 0.1 that were selected from the same catalogs of double-peaked AGNs in SDSS that we use here (Section 2). The Nevin et al. (2016) sample does not overlap with the sample presented here. The full details of the kinematic analyses are given in Nevin et al. (2016), and we summarize the approach below.

First, we fit a Gaussian to the stellar continuum along the spatial position axis of each longslit spectrum, and measure the full width at half maximum of the spatial extent of the continuum (FWHM_cont). For each spatial row of the longslit spectrum that is within FWHM_cont, we determine how many Gaussians are required to best fit the [O iii] λ5007 emission. To determine the number of Gaussians required for the fit, we use the Akaike Information Criterion (AIC), which is a least-squares statistic that penalizes extra free parameters (Akaike 1974). The fraction of the spatial rows of the longslit spectrum within FWHM_cont that are best fit by more than two Gaussians is used in our kinematic classification of the galaxies (Section 4.1).

Then, we measure the spatial extent of the [O iii] λ5007 emission. For each spatial row of the longslit spectrum, we use the AIC to determine whether the [O iii] λ5007 spectrum is better fit by a line (representing the background; 2 free parameters) or by a Gaussian and a line (representing an emission line and the background; 5 free parameters in total). We define the spatial extent of the [O iii] λ5007 emission as the spatial extent along the slit for which each spatial row's spectrum is best fit by a Gaussian and a line. For each row of the longslit spectrum that is within this spatial extent, we fit the [O iii] λ5007 emission line first with one Gaussian, and then with two Gaussians. For the single-Gaussian fit, we measure the line-of-sight velocity difference between the central velocity of the Gaussian fit and the systemic velocity of the galaxy in each row, and we report the highest measured line-of-sight velocity difference V_r. We also measure the dispersion of each component of the double-Gaussian fit in each row, and we report the highest measured dispersions σ₁ and σ₂.

Furthermore, we determine the position angle of the maximum extent of [O iii] λ5007 on the sky by measuring the spatial positions of [O iii] λ5007 along each slit position angle as in Comerford et al. (2012). We compare this to PA_gal, the isophotal position angle of the major axis of the galaxy from SDSS r-band photometry. If PA_{[O iii]} = PA_gal to within 20° error (Nevin et al. 2016), then we classify the [O iii] λ5007 emission as aligned with the plane of the galaxy.

Finally, we measure the asymmetry A of the [O iii] λ5007 emission-line profile in each longslit spectrum, following the approach of Whittle (1985) and Liu et al. (2013a), and where positive asymmetry values indicate redshifted emission-line wings and negative asymmetry values indicate blueshifted emission-line wings. We consider the asymmetries measured from the two position angle observations, and we report the A that has the larger absolute value. Then, we define an emission-line profile as symmetric if the absolute value of its asymmetry falls below the 95% confidence interval around the mean asymmetry of the sample. Consequently, symmetric profiles are defined by $| A| \lt 0.19$, as in Nevin et al. (2016).

The kinematic parameters are given in Table 2, and these are the parameters that we will use to kinematically classify the source of the double-peaked emission lines in each galaxy. As described in Section 4.1, the main classifications are rotation-dominated, outflow, and ambiguous.

Table 2.  Kinematic Classifications of the 95 Galaxies

SDSS Name	PAobs	Num. Rows Fit	Single Gauss	Double Gauss	Double Gauss	PAgal	PA[O iii]	A	Kinematic
		by >2 Gaussians	Vr (km s−1)	σ1 (km s−1)	σ2 (km s−1)				Classification
J0006+1548	44.4	4/11	$-{231.5}_{-7.2}^{+27.1}$	${232.7}_{-18.8}^{+37.6}$	${173.0}_{-26.2}^{+59.5}$	44.4	172.6 ± 3.9	0.08	Ambiguous
	134.4	4/11	$-{241.9}_{-9.5}^{+25.1}$	${168.4}_{-0.0}^{+277.5}$	${172.7}_{-29.1}^{+77.2}$			0.11
J0107−0053	52.0	22/43	${305.6}_{-11.1}^{+23.1}$	${271.2}_{-43.0}^{+39.0}$	${222.4}_{-35.3}^{+13.9}$	52.0	176.6 ± 3.8	−0.35	Ambiguous
	142.0	16/41	${380.9}_{-71.0}^{+47.9}$	${353.2}_{-92.7}^{+224.6}$	${367.6}_{-105.7}^{+127.6}$			−0.55
J0116−1025	28.8	10/13	${33.7}_{-0.6}^{+6.0}$	${275.0}_{-0.6}^{+78.2}$	${106.6}_{-55.7}^{+78.2}$	118.8	117.0 ± 2.6	0.19	Rotation Dominated
	118.8	4/13	$-{185.2}_{-0.5}^{+71.4}$	${260.8}_{-0.2}^{+69.4}$	${258.5}_{-22.3}^{+14.8}$			−0.27	+ Disturbance
J0118−0826	42.3	9/15	$-{310.5}_{-8.0}^{+40.2}$	${218.2}_{-8.0}^{+108.4}$	${338.2}_{-65.3}^{+210.1}$	42.3	25.7 ± 3.4	0.18	Rotation Dominated
	132.3	4/13	$-{296.4}_{-15.2}^{+448.7}$	${220.3}_{-15.8}^{+208.1}$	${331.3}_{-67.5}^{+49.0}$			0.06	+ Obscuration

Note. Column 1: galaxy name. Column 2: observed position angle, in degrees east of north. Column 3: number of spatial rows of emission in the longslit spectrum that are best fit by >2 Gaussians within the FWHM of the continuum. Column 4: line-of-sight velocity difference between the velocity derived from a single-Gaussian fit to the emission-line profile and the systemic velocity of the galaxy. Columns 5 and 6: dispersion of each component of a double-Gaussian fit to the emission-line profile. Column 7: position angle of the major axis of the galaxy in the r-band, in degrees east of north. Column 8: position angle of the maximum extent of [O iii] λ5007 on the sky, in degrees east of north. Column 9: asymmetry of the emission-line profile, where positive asymmetry values indicate redshifted emission-line wings and negative asymmetry values indicate blueshifted emission-line wings. Column 10: Kinematic classification of the galaxy.

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

Download table as:  DataTypeset image

We searched the 95 galaxies for those with companion galaxies that have line-of-sight velocity separations $| {\rm{\Delta }}v| \,\lt 500$ km s⁻¹ and projected physical separations Δx < 30 kpc, where Sections 3.3.1 and 3.3.2 describe how Δv and Δx are measured, respectively. We selected the $| {\rm{\Delta }}v| \lt 500$ km s⁻¹ and Δx < 30 kpc criteria to match those of Ellison et al. (2008), for ease of later comparison.

We found that eight galaxies with double-peaked AGN emission lines have companion galaxies within $| {\rm{\Delta }}v| \lt 500$ km s⁻¹ and Δx < 30 kpc (Figure 2). Five of the galaxies have SDSS spectra that are classified as Type 1 AGNs (SDSS J0952+2552, SDSS J1157+0816, SDSS J1248−0257, SDSS J1541+2036, SDSS J1610+1308) and three have SDSS spectra that are classified as Type 2 AGNs (SDSS J1245+3723, SDSS J1301−0058, SDSS J1323+0308).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As a further test of whether two galaxies in a pair are indeed related, we can neglect the $| {\rm{\Delta }}v|$ cutoff and estimate the probability of the chance projection of two unassociated galaxies on the sky. We carry out this calculation for the galaxy pairs (SDSS J1157+0816, SDSS J1245+3723, SDSS J1323+0308) where both galaxies are detected in the SDSS photometric catalog, which has a limiting r-band magnitude of 22.0 (Alam et al. 2015). The larger separation pairs have greater probabilities of being chance projections of unrelated galaxies, and this subsample does include the largest separation pair (SDSS J1323+0308, with a separation of 28.5 kpc). The faintest galaxy in each pair has an r-band magnitude of (19.66, 20.41, 19.29) for (SDSS J1157+0816, SDSS J1245+3723, SDSS J1323+0308), respectively. For each galaxy pair, we calculate the surface density SDSS galaxies out to this r-band magnitude limit. For the example of the faintest galaxy in our pair sample, the r = 20.41 galaxy SDSS J1245+3723, the surface density out to r = 20.41 is 209 galaxies deg⁻². We calculate the probability of chance alignment for each galaxy pair using the surface density, redshift, and the solid angle subtended by the galaxy pair separation that we measured (Table 3). We find that the probability of chance alignment is (2 × 10⁻⁴, 1 × 10⁻⁴, 9 × 10⁻⁴) for the galaxy pair systems (SDSS J1157+0816, SDSS J1245+3723, SDSS J1323+0308), respectively. Consequently, the probability is 0.001 that there are one or more chance projections in this subsample. The likelier scenario is that the galaxies in these pairs are in fact associated with one another.

The subsections that follow focus on the longslit spectra and the images of these eight merging galaxy systems. We note that many other galaxies in our sample of 95 galaxies have companions within 30 kpc that were resolved with near-infrared imaging (Rosario et al. 2011; Fu et al. 2012; McGurk et al. 2015), but since the companions do not have spectroscopic redshifts they cannot yet be confirmed as associated with the primary galaxies.

3.3.1. Longslit Spectral Analysis

3.3.2. Imaging Analysis

Here, we use the [O iii] λ5007 emission measurements from Section 3.2 to kinematically classify each galaxy using the classification scheme of Nevin et al. (2016). The three main classifications are as follows.

Rotation Dominated. We classify a system as rotation-dominated if it has Keplerian rotation in the plane of the galaxy, exhibited by V_r < 400 km s⁻¹, σ₁ < 500 km s⁻¹, σ₂ < 500 km s⁻¹, and [O iii] λ5007 emission that is aligned with the plane of the galaxy (e.g., Osterbrock & Ferland 2006). If the emission-line profile is symmetric ($| A| \lt 0.19$), then we classify the system as "Rotation Dominated + Obscuration" since dust obscuring a rotating disk could explain a symmetric double-peaked emission-line profile (e.g., Smith et al. 2012). If the emission-line profile is asymmetric ($| A| \gt 0.19$), then we classify the system as "Rotation Dominated + Disturbance" since nuclear bars, spiral arms, dual AGNs, or other dynamical disturbances could produce an asymmetric double-peaked emission-line profile (e.g., Schoenmakers et al. 1997; Davies et al. 2009; Blecha et al. 2013).

Outflow. For a system to be classified an outflow, we apply the conservative criteria of V_r > 400 km s⁻¹, σ₁ > 500 km s⁻¹, or σ₂ > 500 km s⁻¹ (e.g., Das et al. 2006; Müller-Sánchez et al. 2011; Fischer et al. 2013). Then, if more than half of the spatial rows of the longslit spectra within FWHM_cont have [O iii] λ5007 emission that is best fit by more than two Gaussians, we classify the system as "Outflow Composite." Otherwise, the system is "Outflow." The "Outflow Composite" class can be explained by an outflow that includes many different gas clouds with their own distinct velocities.

Ambiguous. If a galaxy fits neither the rotation-dominated nor the outflow classification, then it is ambiguous. These are the systems with V_r < 400 km s⁻¹, σ₁ < 500 km s⁻¹, σ₂ < 500 km s⁻¹, and [O iii] λ5007 emission that is not aligned with the plane of the galaxy. Possible explanations for the emission-line profiles of these galaxies include a counter-rotating disk, inflowing gas, weak outflows, and dual AGNs.

Table 2 shows the results of our kinematic classifications. We find that the galaxies are classified as follows: 6% (6/95) are Rotation Dominated + Obscuration, 9% (9/95) are Rotation Dominated + Disturbance, 21% (20/95) are Outflows, 33% (31/95) are Outflow Composite, and 30% (29/95) are Ambiguous.

First, we consider the galaxies classified as Rotation Dominated + Obscuration or Rotation Dominated + Disturbance. Simulations of the emission-line profiles of AGNs in galaxy mergers have shown that a Rotation Dominated + Disturbance profile can be associated with dual AGNs, where a rotating disk around one AGN is disturbed by the second AGN (Blecha et al. 2013). In contrast, dual AGNs are unlikely to produce the symmetry of the emission lines in the Rotation Dominated + Obscuration classification; instead, these symmetric emission-line profiles are most likely associated with obscured rotating disks (e.g., Smith et al. 2012; Blecha et al. 2013).

Next, we examine the galaxies classified as Outflow or Outflow Composite. Outflow galaxies are the archetypal outflows that have a single redshifted component and a single blueshifted component, while Outflow Composite galaxies have evidence of three or more emission knots that are moving at distinct velocities. This complex structure can be due to, e.g., outflowing gas shocking as it encounters the interstellar medium, or multiple different outflows (e.g., Cecil et al. 2002; Crenshaw & Kraemer 2005; Comerford et al. 2017). We note that the number of galaxies classified as Outflow Composite is a lower limit, since higher signal-to-noise follow-up spectra may reveal additional Gaussian components that transform an Outflow classification into an Outflow Composite classification. Therefore, we do not draw a significant distinction between Outflow and Outflow Composite classifications here.

Finally, the Ambiguous galaxies have the most complicated kinematic structures. They likely contain some combination of less energetic outflows, inflows, rotation, and possibly dual AGNs, although with the longslit data alone we cannot distinguish these individual contributions to the emission-line profiles.

When we consider only the galaxies with unambiguous classifications, we find that ${77}_{-12}^{+10} \%$ of SDSS galaxies at z > 0.1 with double-peaked narrow AGN emission lines are explained by outflows and ${23}_{-10}^{+12} \%$ are explained by rotation, where the error bars present the 95% binomial confidence intervals. These results are consistent, within the confidence intervals, with the results of Nevin et al. (2016) for their sample of 71 SDSS galaxies at z < 0.1 with double-peaked narrow AGN emission lines. Other studies, which do not use the same kinematic classifications, also find that the majority of double-peaked emission lines are produced by NLR gas kinematics (e.g., Shen et al. 2011; Fu et al. 2012). We conclude that most double-peaked narrow AGN emission lines are produced by outflows.

Seven of the galaxies with double-peaked AGN emission lines have companion galaxies that are classified as Seyferts, and one has a companion that is classified as composite. Although the Seyfert-like line flux ratios in each galaxy pair suggest the presence of dual AGNs, there may be a single ionizing source (a single AGN) producing the emission features in both galaxies (e.g., Moran et al. 1992). As a result, these eight merging galaxy systems are candidate dual AGNs. Follow-up high-resolution X-ray or radio observations are necessary to resolve whether two AGNs are present and whether the systems indeed host dual AGNs.

Of these eight dual AGN candidates, four (SDSS J0952+2552, SDSS J1245+3723, SDSS J1248−0257, and SDSS J1301−0058) have a companion galaxy located within the 15 radius SDSS fiber. These systems are kinematically classified as outflow (SDSS J0952+2552) or ambiguous (SDSS J1245+3723, SDSS J1248−0257, and SDSS J1301−0058). However, because both galaxies are located within the SDSS fiber, the motion of dual AGNs could contribute to the double-peaked emission lines in these systems. The case of dual AGNs where one AGN is driving an outflow and the other AGN is relatively faint could be classified as an outflow, while dual AGNs combined with rotation, inflows, weak outflows, and other kinematic disturbances could be classified as ambiguous (Nevin et al. 2016). There is no clean kinematic classification category for dual AGNs, since the kinematics alone are insufficient to confirm dual AGNs.

For the other four dual AGN candidates, the galaxy pair separations are too large for dual AGNs (where there is one central AGN in each galaxy) to contribute to the double peaks in the SDSS spectra.

Here, we comment on each dual AGN candidate individually and compare to other analyses published in the literature. We note that these published analyses sometimes define dual AGNs as those that are responsible for the double-peaked emission in the SDSS spectrum, whereas our definition of dual AGNs includes AGN pairs with separations <30 kpc, regardless of whether they contribute to the double peaks.

SDSS J0952+2552 (candidate dual AGNs). The two stellar bulges seen in the HST/F160W image are also seen in near-infrared laser guide star adaptive optics H-band imaging with the NIRC2 near-infrared camera on the Keck II telescope (Rosario et al. 2011) and in H-band imaging with the OH-Suppressing Infrared Imaging Spectrograph (OSIRIS) on the Keck II telescope (Fu et al. 2012). While previous studies have classified this system as confirmed dual AGNs (McGurk et al. 2011; Fu et al. 2012; McGurk et al. 2015), we classify it as candidate dual AGNs because of the need for X-ray or radio confirmation of two AGNs with such a small separation (100).

SDSS J1157+0816 (candidate dual AGNs). The galaxy with the double-peaked AGN emission lines has two companions visible in the SDSS imaging: one to the northeast, and one to the southwest. These three sources are also seen in NIRC2 K'-band imaging (Fu et al. 2012). The galaxy with the double-peaked AGN emission lines and the southwest companion have similar redshifts (Table 3), but the northeast companion's continuum was too faint in our longslit spectrum to extract a redshift. We cannot confirm whether it is associated with the other two galaxies, although in the SDSS image all three galaxies appear to be related (Figure 2). McGurk et al. (2015) classified the [O iii] λ5007 peaks in SDSS J1157+0816 as unresolved structure with a range of possible explanations, including dual AGNs, NLR kinematics, and small-scale outflows.

SDSS J1245+3723 (candidate dual AGNs). The longslit spectra show continuous emission connecting the two galaxies, and the SDSS image also shows that the companion galaxy has an elongated morphology (Figure 2). This suggests that the two galaxies may already be interacting. No other studies have published follow-up observations for this object, so it has no previous classifications.

SDSS J1248−0257 (candidate dual AGNs). NIRC2 H-band and K'-band imaging of this system reveal the companion galaxy (Rosario et al. 2011; Fu et al. 2012). Fu et al. (2012) classified this system as having an extended NLR.

SDSS J1301−0058 (candidate dual AGNs). A companion galaxy is seen in HST/ACS/F550M imaging (Fu et al. 2012), and the companion's elongated tidal tail suggests that the two galaxies are already interacting. Fu et al. (2012) classified this system as having an unresolved NLR.

SDSS J1323+0308 (candidate dual AGNs). The SDSS image shows tidal debris in the system, and the longslit spectrum shows ionized gas forming a bridge between the two galaxies (Figure 2). No other studies have published follow-up observations for this object, so it has no previous classifications.

SDSS J1541+2036 (candidate dual AGNs). The companion galaxy is seen in NIRC2 H-band and K'-band imaging (Rosario et al. 2011; Fu et al. 2012). Fu et al. (2012) classified this system as having an unresolved NLR.

SDSS J1610+1308 (candidate dual AGNs). The galaxy with the double-peaked AGN emission lines has a companion that is found in NIRC2 H-band and K'-band imaging (Rosario et al. 2011; Fu et al. 2012), and a stream of ionized gas connects the galaxy with the double-peaked AGN emission lines to the companion (Figure 2). McGurk et al. (2015) classified the double peaks in [O iii] λ5007 as being produced by outflows, and this complicated system may include both outflows and dual AGNs.

Many previous studies have explored the relative AGN fractions in merging and isolated galaxies, as a means of constraining the importance of different AGN triggering mechanisms (e.g., Ellison et al. 2011; Silverman et al. 2011; Koss et al. 2012; Liu et al. 2012; Satyapal et al. 2014; Fu et al. 2018). Here, we also examine the fraction of AGNs in galaxy mergers, but specifically for AGNs that exhibit double-peaked emission lines in their spectra.

Of the eight merging galaxy systems that we find, three of them have both merging galaxies resolved in SDSS imaging (SDSS J1157+0816, SDSS J1245+3723, SDSS J1323+0308). Consequently, we find that at least ${3.2}_{-1.7}^{+3.0} \%$ (3/95) of SDSS galaxies with double-peaked AGN emission lines are in galaxy mergers (defined by $| {\rm{\Delta }}v| \lt 500$ km s⁻¹ and Δx < 30 kpc) that are resolvable in SDSS imaging. This is a lower limit, given that the necessary spectroscopic redshifts do no exist for all of the galaxies that have another galaxy located within 30 kpc. Our lower limit is ∼2 times higher than the ${1.7}_{-0.2}^{+0.2} \%$ of all SDSS AGNs that are in galaxy mergers, defined with the same $| {\rm{\Delta }}v|$ and Δx cutoffs, and with the same range of galaxy stellar masses (the Kolmogorov–Smirnov probability is 72% that the masses were derived from the same distribution; Ellison et al. 2008).

The actual fraction of galaxies with double-peaked AGN emission lines that are associated with galaxy mergers is likely much higher. Studies using 01 resolution imaging have found that ${29}_{-4}^{+5} \%$ of SDSS galaxies with double-peaked AGN emission lines have another galaxy located within the 3'' SDSS fiber (Fu et al. 2012).

Next, we examine whether our eight merging galaxy systems are in major or minor mergers. If we define major mergers as mass ratios less than 4:1 and minor mergers as mass ratios greater than 4:1, then we find six major mergers and two minor mergers. Since the SDSS photometric catalog has a limiting r-band magnitude of 22.0, we cannot detect the faintest companion galaxies in our sample. Consequently, we are underestimating the number of minor mergers.

A previous study of galaxy pairs in SDSS, which used a similar pair definition ($| {\rm{\Delta }}v| \lt 350$ km s⁻¹ and Δx < 35 kpc), explored the relative numbers of major and minor merger galaxy pairs that are detectable in SDSS (Lambas et al. 2012). Using the same mass ratio cutoff as we use between major and minor mergers (4:1), they find that major merger galaxy pairs are twice as common as minor merger galaxy pairs in SDSS. While this is not indicative of the true mass ratio trend for galaxy mergers (where the minor merger rate is ∼3 times larger than the major merger rate; Lotz et al. 2011), it accounts for bias in the SDSS selection and offers a more accurate comparison for our sample. For the subset of our galaxy pairs where both galaxies are resolved in SDSS imaging, our results (two major mergers and one minor merger) are consistent with those of the general galaxy pair population in SDSS (Lambas et al. 2012).

In six of the dual AGN candidates the more luminous AGN is in the more massive stellar bulge, and in two of the dual AGN candidates the more luminous AGN is in the less massive stellar bulge. This fits with phenomenological models of dual AGNs that find that 60%–70% of dual AGNs have the more luminous AGN in the more massive stellar bulge (Yu et al. 2011) and simulations that find that the distinction of more luminous AGN switches between stellar bulges during the course of the galaxy merger (Van Wassenhove et al. 2012).

Of the two AGNs in each dual AGN candidate system, we determine which AGN is accreting at a higher rate using the Eddington ratio f_Edd ≡ L_bol/L_Edd. The bolometric luminosity L_bol is determined from the [O iii] λ5007 luminosity (Heckman et al. 2004; with a scatter of 0.38 dex), while the Eddington luminosity is L_Edd = 4πcGM_BHm_p/σ_T, where M_BH is the black hole mass, m_p is the mass of a proton, and σ_T is the Thomson scattering cross section.

If the more luminous stellar bulge in the merging system is "1" and the less luminous stellar bulge is "2," then L_bol,1/L_bol,2 = (f_Edd,1/f_Edd,2) (M_BH,1/M_BH,2). Assuming that the black hole mass traces the host stellar bulge luminosity (Marconi & Hunt 2003; McLure & Dunlop 2004; Graham 2007), then M_BH,1/M_BH,2 = L_*,1/L_*,2 and f_Edd,1/f_Edd,2 = (L_*,2/L_*,1)(L_bol,1/L_bol,2). Using this approach, we estimate f_Edd,1/f_Edd,2 for each dual AGN candidate (Table 4).

We find that three of the dual AGN candidates have f_Edd,1/f_Edd,2 < 1, and five have f_Edd,1/f_Edd,2 > 1. This implies that the more massive supermassive black hole in a dual AGN system typically accretes with a higher Eddington ratio than the less massive supermassive black holes in a dual AGN system. Simulations of galaxy mergers find that the less massive black hole accretes with a higher Eddington ratio until the less massive black hole's host galaxy is stripped of its gas (Van Wassenhove et al. 2012; Steinborn et al. 2016), which implies that in the five systems with f_Edd,1/f_Edd,2 > 1 (SDSS J1245+3723, SDSS J1301–0058, SDSS J1323+0308, SDSS J1541+2036, SDSS J1610+1308), the less massive galaxy in the merger may have depleted its gas reservoir. Follow-up observations of the gas contents of the galaxies would confirm whether this is the case.