Revisiting the Complex Nuclear Region of NGC 6240 with Chandra

To obtain a fine alignment of the nuclear emission, we registered the observations using the centroids of the images in the energy band 6.0–6.4 keV, where the nuclear emission is dominated by the prominent 6.4 keV Fe i Kα line, which has a strong point-like component (see Wang et al. 2014); this line is observed at 6.2 keV, as expected given the redshift of NGC 6240, z = 0.02448. Visual inspection of the images led us to choose the N AGN as a reference, because the Fe i Kα image looks symmetrically round, as expected from point-like emission. To evaluate the centroid of the nuclear source for each observation, we used the following procedure. First, we produced 6.0–6.4 keV band images with subpixel scale of 1/8 of the native ACIS pixel; second, we smoothed these images with a two-dimensional Gaussian kernel with σ = 2 image pixels and selected the brightest pixel in the resulting images; finally, using the original 1/8 pixel image for each ObsID, we found the centroid of the count distribution within a circular region with a radius of 02 centered on the brightest pixel of the corresponding smoothed image.

We then used the wcs_update CIAO task to match the N AGN centroids, using the deepest observation (i.e., OBSID 12713) as a reference. The shifts needed to match the centroids range between 015 and 023, compatible with the 08 Chandra absolute astrometric uncertainty. ¹²

Figure 1 shows the results of the registration procedure. We reprojected the shifted observations to a common tangent point with the reproject_obs task and merged the reprojected observation with the flux_obs task.

Zoom In Zoom Out Reset image size

Figure 2 shows on the left the 6.0–6.4 keV merged images obtained from the observations performed in pure imaging mode (OBSIDs 12713 and 1590), and on the right the full data set adding to these the 0th order images of the HETG grating observations (OBSIDs 6908 and 6909). The latter adds to the signal to noise, without any change in quality. Both images suggest an extension of the surface brightness of the S AGN to the SE, which is confirmed by our following analysis. In what follows, we use the full available data set.

Zoom In Zoom Out Reset image size

The Chandra mirror PSF was well-calibrated before launch and characterized as a function of photon energy. ¹³ This PSF is well reproduced in flight by HRC observations, except for a small PSF artifact situated at a radial distance of ∼06–08 from the centroid of the PSF, in a well-determined angular position that depends on the roll angle of the observation. This artifact contains ∼6% of the counts in the PSF, and therefore it is important to consider near strong point sources (Juda & Karovska 2010 ¹⁴ ; see also the CIAO documentation page ¹⁵ ).

However, here we are using ACIS data in subpixel binning for our work, for which in-flight calibrations of the PSF have not yet been developed by the CXC. Given that we are attempting to investigate subarcsecond scale features, it is important to carefully model the PSF. Fortunately, our data include two prominent point-like sources from the two AGNs, which we could exploit as empirical observations of the PSF. With our observations of the N and S AGN, we can then effectively measure the PSF and compare it to simulations based on the preflight calibrations. We performed model-data comparisons using both 1/8 and 1/16 subpixel binning, with consistent results. We also separately compared the PSF models with the imaging observations (ObsID 1590 and 12713) and the 0th order grating observations (6908, 6909) data to exclude any possible unknown effect in the 0th order grating PSF. We obtained consistent results, as with the ACIS-S imaging observations.

The Chandra PSF was simulated using rays produced by the Chandra Ray Tracer (ChaRT ¹⁶ ) projected on the image plane by MARX. ¹⁷ For each observation, we generated the average from 1000 PSF simulations centered at the coordinates of the centroids shown in the lower panels of Figure 1. We then produced images of these PSFs in each energy band used in this study, which correspond to those in Wang et al. (2014): 5.5–5.9 keV, 6.0–6.4 keV, and 6.4–6.7 keV. These energy bands are representative of the hard continuum emission, and the redshifted Fe i, Kα I, and Fe xxv emission lines, respectively. The PSFs were then merged to be compared with the merged data set.

MARX allows the application of an "aspect blur" to the simulated PSF to account for possible smearing of the image resulting from imperfect photon positioning. We found that the recommended 025 blur ¹⁸ created a PSF that did not match the central regions of the observed AGNs surface brightness. We investigated the amount of blur to be applied in two ways: first, we derived an encircled energy curve from the strong point-like Fe i Kα emission line (observed in the 6.0–6.4 keV band) and compared it with the prelaunch calibrations, within a radius of 04; we also derived a set of model PSFs for a range of blur parameters, and compared them with the radial profile of the Fe i Kα sources. In both cases, the radial profile of the observed emission is consistent with the Chandra PSF in the same energy band when no blur is applied. In particular, the comparison of the X-ray surface brightness with the PSF shows that a central region of positive residuals begins to appear for blur factor 0.04. (See, for example, the left and center panels of Figure 3. Here, the S AGN surface brightness excludes the SE excess, as in Figure 5.) We therefore did not apply a blur factor to the model PSF.

Zoom In Zoom Out Reset image size

We explored empirically how to best match the model PSF to each merged AGN observation, on the assumption that the observed emission of each AGN consists of an unresolved component that should match the PSF, plus possible residual extended emission. We explored matching the PSF to the data for radii from the data/PSF centroid, ranging from 02 to 05, taking into account the contribution of the PSF wings to each other's normalization circle. For sufficiently higher signal to noise ratios, it would be preferable to use as small a radius as possible to normalize the PSF to the data. However, the statistical uncertainties in the inner bins are large enough to bias the fit, when the nominal values are used, resulting in regions of cumulative negative residuals (i.e., over-subtracting the PSF). (See the right panel of Figure 3for the N nucleus; a similar behavior is observed for the S nucleus, excluding the "excess" region, discussed later in more detail.)

For the N AGN, normalizing the model PSF to the data within 04, we find an excellent agreement of the profiles with the PSF within 05. The cumulative residual counts within 05 are consistent with zero. At larger radii, excess extended emission is apparent. Figure 4 shows the radial profiles of the data and PSF for the N AGN for 1/16 binning for the data. In all our experiments, we obtain consistent results using 1/8 and 1/16 binning.

Zoom In Zoom Out Reset image size

In the S AGN, instead, the match between PSF and data was not straightforward, because of excess emission within 04 from the nuclear centroid in the SE quadrant. Inspection of the radial surface brightness profiles at different azimuthal angles showed that this excess emission is concentrated in the region at the position angle (PA) between 120° and 210° (counterclockwise from N). Excluding this azimuthal sector, we were able to match the PSF well to the remaining data within 05, normalizing the PSF to the data within a radius of 0''4. Figure 5 shows the azimuthal sector of the S AGN emission, which we used to match the PSF. Figure 6 shows the excess sector, also with 1/16 binning. We obtain consistent results using 1/8 binning.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We conclude that the merged PSFs with no blur we have produced are excellent models of the merged unresolved emission of the NGC 6240 AGNs within 05 of the centroids, when using 1/8 and 1/16 subpixel data.

In Section 2.2 we have shown that comparisons between the observed surface brightness of the two circumnuclear regions with the expected PSF reveal two types of discrepancies: (1) an excess surface brightness at larger radii than 05 and (2) a localized excess in the S AGN in the 120°–210° PA azimuthal region SE of the nuclear centroid. Here we explore these features and their spatial distribution in detail. We also address possible biases introduced by the PSF artifact in these images.

Figure 7 compares the residuals after PSF subtraction for the central ∼1 kpc (∼2'') radius region of NGC 6240 with 1/8 binning, in the three energy bands under study. Only positive residuals are shown. Some negative pixels occur within the central 04 circles, consistent with statistical noise. As shown in Figures 4 and 5 (Section 2), the cumulative residual counts are consistent with zero within the circles, with the exception of the "exclusion region" to the SE of the S AGN (Figure 6). To better visualize the extended emission, we have produced the adaptively smoothed residual maps shown in Figure 8 with the CIAO tool dmimgadapt, where we used 1/16 binned data smoothed with Gaussians, with 9 counts under the kernel and sizes ranging from 1/2 to 15 image pixels, in 30 logarithmic steps.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Although there is significant excess emission in all three energy bands, the excess is particularly significant in the Fe i Kα band. In both this image and in the Fe xxv image, there is a concentration of residual counts in the area between the two AGNs, which is not seen in the 5.5–5.9 keV continuum image. In the Fe i Kα (top) of Figure 7, we also show the regions affected by the PSF artifact for each observation in our data set (red: ObsID 1590; cyan: 6908; magenta: 6909; yellow: 12713). The overall artifact region shape is also reproduced in Figure 8. This artifact results in a spurious positioning of 6% of the counts of the nuclear AGN in each of these regions. Since the two nuclear sources are bright, it is important to assess the impact of this effect. We estimated the PSF artifact counts using the model PSFs for each nucleus.

Table 1 lists the cumulative counts in the residual images within a 2'' radius (∼1 kpc). In parentheses, we give the cumulative counts expected from the PSF artifact within a 2'' radius. Given the total number of extended counts, this effect is small (∼20%) and does not change our conclusions on the presence of extended circumnuclear emission. However, there may be localized effects, since the artifact areas are concentrated in the regions between the two nuclei and to the NW of the northern nucleus. Extraction of counts from these regions indeed shows values consistent (within statistics) with those expected from the PSF artifacts. To investigate further the presence of diffuse emission in the larger region between the two AGNs, we extracted counts from the box highlighted with a dashed line in the top panel of Figure 8. Subtracting the counts expected from the PSF artifact within this box, we find that there is highly significant (5.6σ) diffuse emission in this region in the Fe i Kα band and a 3 σ excess in the Fe xxv band. No significant excess is found in the 5.0–5.5 keV hard continuum. The excess counts in the box are listed in Table 1.

As discussed in Section 2.2.1, the N AGN surface brightness is well fitted with the PSF, and correspondingly the residuals in the 04 N circle are consistent with zero, as shown in Table 1. In the S AGN, instead, we find significant count excess within the 04 circle in all the energy bands, but most significantly in the Fe i Kα band (Table 1). This excess originates from the 120° to 210° PA sector SE of the nucleus. Outside this sector, the surface brightness distribution of the S AGN is well fitted by the PSF (see Figure 5). These circumnuclear circles of 04 radius (∼200 pc) roughly correspond to the spheres of influence of each nuclear black hole (BH)—that is, the region within which the BH contributes half of the total mass (Medling et al. 2015).

The excess to the SE of the S AGN is free of contamination from the PSF artifact. This excess is most prominent in the Fe i Kα emission line. It extends outward in the SE direction to a ∼15 radius (Figure 8). Figure 8 suggests that this feature may fan out to a wider angle outside the 04 radius circle, where it occupies the 120°–210° sector. Using the image for guidance, we extracted the counts contributed by this feature, from 110° to 210° PA angular sectors of 15 radius centered on the Fe i Kα centroid of the southern nucleus. This region is shown in red in the 6.0–6.4 keV panel of Figure 8. Table 1 lists the total counts estimated for this SE feature in each energy band.

Our analysis of the residual images after the subtraction of the nuclear AGN PSF models has demonstrated the presence of extranuclear X-ray emission in the central few hundred parsecs of NGC 6240 and has suggested energy-dependent differences in this emission. To further probe these differences, we have produced reconstructed images of the central region, using the image restoration algorithm Expectation through MCMC (EMC2; Esch et al. 2004; Karovska et al. 2005, 2007, 2010; Wang et al. 2014; Fabbiano et al. 2018a, 2018b; sometimes referred to as PSF deconvolution) with the PSF models described in Section 2.2.

EMC2 (Expectation through MCMC) is a statistical deconvolution technique allowing reconstruction/restoration of an astronomical image affected by the blurring effects of the PSF (including telescope and detector effects). This technique uses a wavelet-like multiscale representation of the true image (unblurred by the PSF) in order to achieve smoothing at all scales of resolution simultaneously. Wavelet decomposition of the image allows the Poisson likelihood to be factored into separate parts, corresponding to the wavelet basis, and each of these can be reparametrized as a split of the intensity from the previous, coarser factor. A prior is assigned to these splits (which can be viewed as smoothing parameters), and a model is fit using MCMC methods. This allows one to capture small- and large-scale structures in the image simultaneously, including diffuse emission as well as point-like sources (and sharp features in the image). This Bayesian model-based analysis is specifically applicable to low-counts Poisson data and allows assessment of the uncertainties in the reconstructed (deconvolved, restored) image. Details of this technique and examples are described in Esch et al. (2004).

To demonstrate how the diffuse emission recovered by the application of the EMC2 reconstruction to the data is not likely to be an artifact of the algorithm, we have simulated an image consisting of only two point-like sources, based on the two 6.0–6.4 keV model merged PSFs for the nuclear sources (Section 2). We have applied the same EMC2 reconstruction to this simulated image and to the real data. The results are shown in Figure 9. The EMC2 reconstructed image from the two PSF simulated images (left panel), shows only two point-like sources. Conversely, the EMC2 reconstruction of the observed merged 6.0–6.4 keV image (right panel) shows clear extended emission, demonstrating that the extended emission cannot be an artifact of the reconstruction method. Similar simulations and thorough testing of the EMC2 techniques were performed in Esch et al. (2004; see, e.g., their Figure 7).

Zoom In Zoom Out Reset image size

Figure 10 shows the results of the 200-iteration EMC2 reconstruction in the three energy bands, which includes the larger-spatial scales, for the 1/8 pixel images, with a further Gaussian smoothing with a 2 pixel kernel. These images show diffuse emission extending out to ∼5'' (∼2.5 kpc) from the center (see also Wang et al. 2014). The 100-iteration reconstructions for the 1/16 pixel binned images of the inner region are shown in Figure 11 to emphasize the small-scale features in the inner ∼1 kpc region.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 11 shows that there is extended X-ray emission both in the two nuclear regions and in between the nuclei. The differences between the images in the various energy bands suggest that these extended features are unlikely to arise from the PSF artifact, even in the regions where the artifact would contribute. The artifact is not included in the PSF model generated with the tools provided by the Chandra Data Center, so it would generate localized count excesses (up to ∼6% of the PSF) in the areas within its footprint in Figure 11.

Figure 11 shows extended emission to the north of the southern AGN in both neutral and ionized Iron, in an area where the PSF anomaly is not expected to interfere. In the areas potentially affected by anomaly, there are spatial differences between the energy bands that suggest that some features may be real (Figure 12). These differences are also—albeit less strongly—reflected in the adaptively smoothed residuals in Figure 8. In particular, Figure 11 shows an apparent filament connecting the nuclei to the NE of the S AGN, in both hard continuum and Fe i Kɑ. Instead, a wider "bridge" connects the two AGNs in the Fe xxv emission to the W of the Fe i Kα and continuum filaments (see the lower-left panel of Figure 11). This bridge is statistically significant. Extracting the counts from this region from the raw Fe xxv image, we find an excess of 33 counts (4.2σ) over the diffuse emission in adjacent regions. The bridge is evident also in the adaptively smoothed image (lower-right panel of Figure 11). While energy-dependent variability may in principle affect the images, it seems unlikely in our case. A point source, prominent in the 6.4–6.7 keV energy band, present in only a subset of the data, should appear more localized than the "bridge" emission. For example, see what the EMC2 reconstruction of point-like emission shows in Figure 9 (left). No corresponding hard continuum feature is observed (Figure 12).

Zoom In Zoom Out Reset image size

The Fe i Kα image (top right of Figure 11) may suggest a possible third point-like emission region to the SE of the southern AGN, which is spatially consistent with the position of the third possible nucleus of NGC 6240 reported by Kollatschny et al. (2020). The Fe xxv (bottom-left) image does not show a source at the position of the third nucleus. There is an elongation in the SE direction in the hard continuum image (top left). We have shown that there is a prominent SE elongation in the residual images after PSF subtraction in all the energy bands, but especially in the continuum and Fe i Kα (Figure 8; Section 3.1).

Feruglio et al. (2013) and Wang et al. (2014) reported convincing evidence of thermal emission in the X-rays, thought to be caused by strong shocks in the ISM of NGC 6240. These thermal processes produce both hard continuum and Fe xxv line emission. The presence of extended Fe i Kα emission, however, cannot be explained in this framework. While Fe xxv and the thermal hard continuum arise from the shocked ISM (see Feruglio et al. 2013; Wang et al. 2014), Fe i Kα is thought to arise from fluorescence, excited by hard nuclear X-ray photons interacting with dense clouds. Hard nonthermal continuum emission is also produced by the scattering of nuclear photons by these dense clouds (e.g., Baloković et al. 2018).

The presence of strong Fe i Kα emission in CT AGNs suggested that this interaction occurred in the obscuring torus, a region of sub-parsec radius surrounding the AGN (Suganuma et al. 2006). However, more recent Chandra observations of several CT AGNs have detected extended Fe i Kα emission and associated it with regions ranging from several tens of parsecs up to several kiloparsecs (Marinucci et al. 2012, 2017; Bauer et al. 2015; Koss et al. 2015; Fabbiano et al. 2017, 2018a, 2018b; Jones et al. 2020; Ma et al. 2020a). These observations suggest that fluorescence can also arise from the interaction of the AGN photons with dense molecular clouds in the host galaxy (Fabbiano et al. 2018a) or in a substructure such as a circumnuclear disk (Marinucci et al. 2012, 2017; Fabbiano et al. 2019b). Even in the Milky Way, there is strong evidence of Fe i Kα emitting clouds, first discovered with ASCA (Koyama et al. 1996) and more recently studied with Chandra and XMM-Newton (Ponti et al. 2015; Churazov et al. 2017a, 2017b). In the CT AGNs studied so far, wherever extended Fe i Kα is found, the hard continuum also presents similar extended emission components, consistent with scattering by dense non-nuclear clouds, not just the nuclear torus.

In ESO 428-G014, the first well-studied case of a CT AGN with hard continuum and Fe i Kα emission extended out to a few kiloparsecs (Fabbiano et al. 2017, 2018a, 2018b) and the extended Fe i Kα line emission has a large equivalent width (EW ∼2 keV, Fabbiano et al. 2017), typical of the large EW measured from the overall spectra of CT AGNs (Levenson et al. 2002). This large EW is consistent with the scattering/fluorescence picture that explains the CT AGN spectra (e.g., Baloković et al. 2018).

In NGC 6240, from the values in Table 1, we measure a smaller EW ∼330 ± 0.1 eV for the Fe i Kα line of the extended component in the inner 2'' (∼1 kpc) circle. This EW is consistent with the value found by Wang et al. 2014, for a slightly larger emission area. This lower Fe i Kα EW than typical of CT AGNs (∼1 keV or more) could be explained by a mix of thermal and nonthermal continuum emission. We measure the nonthermal Fe i Kα directly, as well as the continuum, which we can assume to be composed of nonthermal emission (from the scattering of hard nuclear photons from the same dense clouds responsible for the Fe i Kα line) and thermal emission from shock-heated gas (see, e.g., Wang et al. 2014). If we assume that the EW of the extended component of the nonthermal Fe i Kα line is a nominal EW∼1 keV (e.g., Levenson et al. 2002), from the definition of EW we can estimate the amount of extra continuum needed for decreasing the EW to the measured value. Under these assumptions, we derive that the nonthermal component could account for ∼1/3 of the observed extended hard continuum emission in the central 2'' radius circumnuclear region of NGC 6240. In the region between the two AGNs (the box in Figure 8), the nonthermal continuum emission may be less diluted. In this box, we find that the Fe i Kα EW∼1.6 ± 0.9 keV, consistent—within the large uncertainty—with a CT AGN Fe i Kα EW, using the "box excess" values in Table 1, which are corrected for the PSF artifact (Section 3.1).

Both Feruglio et al. (2013) and Wang et al. (2014) discussed the presence of hard shock-ionized ISM emission from the central regions of NGC 6240. In Feruglio et al.'s paper, this shock ionization was related to the strong nuclear outflows (v ∼400–800 km s⁻¹) suggested by the CO kinematics. Wang et al. instead suggested stellar winds from the active star-forming region in the center of this galaxy as a cause. Their star formation ionization picture was supported by the similarity of the Fe xxv spatial distribution with that of the near-IR H₂(1−0) 2.12 μm emission from Max et al. (2005). However, a more recent paper by Cicone et al. (2018), based on ALMA observations, reports a large wide-spread outflow from NGC 6240, whose energetics challenge the star formation-only-powered hypothesis. In a study based on HST and VLT/SINFONI observations, Müller-Sánchez et al. (2018) conclude that both nuclear and starburst outflows are present in NGC 6240.

The presence of extended Fe i Kα emission provides another view of the AGN–ISM interaction—that of hard X-ray photons interacting with dense molecular clouds. As already noted for the Milky Way and ESO 428-G014 (Koyama et al. 1996; Ponti et al. 2015; Churazov et al. 2017a, 2017b; Fabbiano et al. 2018a, 2019b), high-spatial-resolution observations of the Fe i Kα emission may provide a complementary way to direct CO observations, for mapping giant molecular clouds complexes in galaxies.

We have established that there is a highly significant excess of Fe i Kα emission in the region between the two AGNs of NGC 6240 (Table 1), although we cannot map this region in detail because of the intervening Chandra PSF artifact (see Section 3.1). The EMC2 reconstructed maps (Section 3.2) tentatively suggest possible filaments connecting the two AGNs, both in Fe i Kα and hard continuum, which appear displaced from the Fe xxv bridge that joins the two nuclei (Figure 12). While the Fe xxv emission is thermal and believed to originate from shock-ionized hot ISM regions (Wang et al. 2014), the inter-nuclear continuum and Fe i Kα emission are likely to be the result of the scattering and fluorescence of the AGN photons interacting with the molecular clouds.

Manohar & Scoville (2017) imaged the central region of NGC 6240 in the 3 mm band transitions of CO, HCN, HCO⁺, HNC, and CS, finding that the line ratios are not consistent with the chemistry of "X-ray Dominated Regions" (AGN excitation). The concentration of CO (2−1) emission in regions between the nuclei (Treister et al. 2020) appears to be consistent with this conclusion. Comparison of Chandra and ALMA observations have shown an anticorrelation near the active nuclei of ESO 428-G014 (Feruglio et al. 2020) and NGC 2110 (Fabbiano et al. 2019a), suggestive of X-ray excitation of the molecular clouds—a localized lack of CO (2−1) emission in regions of prominent X-ray emission near the AGN—which is not present here. The situation of NGC 6240 is instead suggestive of fluorescent emission of the dense molecular clouds in the inter-nuclei regions, stimulated by the interaction with nuclear X-ray photons, as reported for the extended components of CT AGNs imaged with Chandra. A similar example of near-nucleus interaction is found in NGC 5643, where a Fe i Kα spatially linear emission feature has been associated with the outer regions (Fabbiano et al. 2018c) of the rotating nuclear molecular disk, detected with ALMA (Alonso-Herrero et al. 2018).

However, some localized X-ray suppression of the CO (2−1) emission may be present, particularly if we take the EMC2 maps (Figures 11, 12) at face value and compare them with the high angular resolution CO (2−1) ALMA images in the recent paper by Treister et al. (2020). These authors resolved a system of clouds in the region between the two AGNs and studied their kinematics and velocity dispersion, concluding that there is a high-velocity turbulent stream of molecular clouds in between the two nuclei, resulting from the interaction of the two circumnuclear disks. Figure 13 shows the EMC2 Fe xxv contours (as in Figure 11) superimposed on the CO (2−1) intensity map of Treister et al. (2020, their Figure 1). The Fe xxv line emission appears displaced from the peak of the CO (2−1) emission, suggesting a possible local X-ray excitation and suppression of the CO (2−1) emission. We cannot, however, exclude that the hot shock-ionized gas, responsible for the Fe xxv emission, is physically displaced by the densest molecular clouds detected with ALMA.

Zoom In Zoom Out Reset image size

The region to the SE of the southern AGN (6240S in Figure 12; Treister et al. 2020, their Figure 1), where we find evidence of extended emission, particularly prominent in the Fe i Kα line and hard continuum (Section 3.1, Figure 8), is also rich in CO emitting clouds, which may belong to a rotating disk associated with the S AGN (Treister et al. 2020). In this region, we find a fairly low EW∼340 eV for the Fe i Kα line, suggesting a mix of thermal and scattered emission for the hard continuum. The presence of correlated Fe xxv emission is consistent with shock excitation in this region (Wang et al. 2014).

With Chandra, Komossa et al. (2003) discovered the double CT AGN system of NGC 6240. Given the complexity of the hard circumnuclear emission, the spectra extracted (r < 1'' for each AGN) by Komossa et al. for their spectral analysis are contaminated by the diffuse emission in the area (see Nardini 2017). Using the 1/16 pixel binned data, we can estimate directly from our spectral imaging the EW of the Fe i Kα line within 02 (∼100 pc) of the X-ray centroid of each AGN. For the N AGN, we find EW = 0.9 ± 0.2 keV, consistent with the typical values for CT AGNs (Levenson et al. 2002). For the S AGN, instead, we derive a lower EW = 0.47 ± 0.1 keV, suggesting that the continuum may be augmented by non-nuclear thermal emission. Komossa et al. (2003) do not quote EW values; their plotted spectra appear consistent with the values found here. Even at these small radii, however, the SE excess may affect the data (Figures 6 and 8).

It is worth pointing out that with Chandra, we can explore down to these small radii, which are within the sphere of influence of each nuclear black hole, ∼235–212 pc (Medling et al. 2015; see Figure 7)—that is, the regions within which the black hole contributes >50% of the total mass (although these figures have been somewhat diminished by more recent measurements of molecular gas in the areas; Medling et al. 2019). Treister et al. (2020) report that the S nucleus has peculiar kinematics within its sphere of influence (∼04), the same radius within which we find a discrepancy between the Chandra PSF and the data (SE excess). Medling et al. (2019) also report a larger contribution of molecular gas mass in the S AGN area. The SE excess is also in the direction of the third nucleus reported by Kollatschny et al. (2020), but the distribution of residuals (Figure 8) suggests that this excess is not due to a single point source.

Kollatschny et al. (2020) find that the N AGN, the S AGN, and the possible third AGN all lie close to the LINER/AGN boundary in the BPT excitation diagram (Veilleux & Osterbrock 1987). This suggests that all three are AGNs. However, BPT mapping of other nearby AGNs shows that regions of LINER emission are often found surrounding extended regions of AGN emission on a similar scale of hundreds of parsecs, with boundary values found in the interface between them (Maksym et al. 2016; Ma et al. 2020b). Hence the excitation diagnostic alone is not enough to require a third nucleus.

The EMC2 image reconstruction of the Fe i Kα emission (top-right panel of Figure 11) suggests a possible source at the third nucleus position. If we conservatively use the residual map (Section 3.1) and extract the counts from a circle of 02 radius at the position of the reported third nucleus, we obtain ∼30 ± 5.9 counts. For the same circular region, we estimate an Fe i Kα EW = 400 ± 200 eV. Given the possible dilution of the continuum by thermal emission, Fe i Kα is a more reliable measure of the AGN luminosity. The third nucleus has ∼1/4 of the counts detected from a similar circle at the centroid of the N AGN from the raw data in the same energy band. Assuming the N AGN to be a typical CT AGN benchmark, given its large Fe i Kα EW (discussed previously) and scaling by the X-ray luminosity of the N AGN (Komossa et al. 2003) by the same factor, we estimate an X-ray luminosity of ∼2 × 10⁴¹ erg s⁻¹ for the third nucleus. Although this is not a detection, it is interesting to note that the mass reported for the third nucleus (Kollatschny et al. 2020) is ∼1/4 of that of the N AGN. Alternatively, we can use the EMC2 Fe i Kα image (Figure 11), which has nine counts from a 02 circle encompassing the region coincident with the third nucleus position. This reconstruction was run with 100 iterations to be more sensitive to small-scale features, so it is likely to underestimate the total counts in the observed image, which also would include larger-spatial-scale extended emission. This would result in a 3 s upper limits of 18 counts on the source counts, along with an 1.4 × 10⁴¹ erg s⁻¹ upper limit on the luminosity.