AN OFF-NUCLEUS NONSTELLAR BLACK HOLE IN THE SEYFERT GALAXY NGC 5252

NGC 5252 was first observed with the Chandra X-Ray Observatory (Weisskopf et al. 2002) on 2003 August 11 for 60.1 ks (observation identification number [ObsID] 4054; PI: M. Dadina), using the spectroscopic array of the Advanced CCD Imaging Spectrometer (ACIS) detector (Garmire et al. 2003) in the 1/4 subarray mode. The nucleus of NGC 5252 was placed near the aimpoint on the backside-illuminated S3 chip, and the data telemetry mode was set to "Very Faint." Three more ACIS-S observations with the same configuration were obtained on 2013 March 4, March 7, and May 9, with exposure times of 40.4, 67.8, and 62.5 ks, respectively (ObsIDs 15618, 15022, 15621; PI: J. Wang). The data were analyzed following the standard procedures using CIAO (Version 4.6) with CALDB (Version 4.5.9) provided by the Chandra X-ray Center (CXC). We reprocessed the event files using CIAO script chandra_repro to apply the latest version of relevant calibration. The light curve from regions free of bright sources on the S3 chip was used to identify time intervals of high background, and no further filtering was required for this data set. The total exposure time is 230.8 ks.

Although the Chandra observations were intended to image the diffuse X-ray emission in NGC 5252, a bright off-nuclear X-ray point source caught our attention besides the known X-ray-bright Seyfert nucleus. Using CIAO tool wavdetect, we have identified the source at a position of R.A. = 13:38:15.64 and decl. = +04:32:55.4. (Figure 1). The uncertainty (1σ) on the source position is 03.

Zoom In Zoom Out Reset image size

We obtained an optical long-slit spectrum using the Inamori-Magellan Areal Camera & Spectrograph (IMACS) on the Baade 6.5 m telescope at Las Campanas Observatory on 2013 May 29 UT. The data were taken with a 15 slit in grism mode (300 lines mm^–1). It covers 3900–8000 Å with a spectral resolution ranging from 650 to 1200. The sky was mostly clear but occasionally covered by thin cirrus clouds. The seeing was between 06 and 10. The observation was done in relatively low airmass (<1.35). The total integration time was 2 hr, divided into four exposures to avoid the saturation of the sky emission lines. The slit was aligned to cover the center of the galaxy and the ULX simultaneously so that we can gauge the contribution from the central active galactic nucleus (AGN) along the radius (Figure 2). The slit (∼30') is long enough to cover the entire galaxy and the sky. The spatial pixel size is 02 on the sky. Following the target observation, we took the arc spectrum and the flat frames for the calibration. The flux calibration was done with two standard stars (CD 32 and LTT 4816).

Zoom In Zoom Out Reset image size

We reduced the data using IRAF. Flat-field correction was done using two sets of images (dome flats and sky flats). We used He+Ne+Ar arc lamp spectra for wavelength calibration. Spatial rectification was done using negative features made by blocked gaps in every ∼100'' along the long slit in the slit mask (see Figure 3). We combined the two-dimensional spectra using cosmic-ray rejection and applied the flux calibration. In order to see a spatial variance of the spectra, one-dimensional spectra with a width of 2'' were extracted. Finally, the spectra are corrected for Galactic extinction using the extinction curve of Fitzpatrick (1999).

Zoom In Zoom Out Reset image size

At the position of the ULX, we extracted a spectrum with 2'' width and subtracted the local sky background in the vicinity of the source, so that we can remove the contribution from the galaxy. Throughout the paper, this spectrum is used for estimating the optical spectral properties of the ULX.

We assemble archival UV/optical imaging data taken with various telescopes (Galaxy Evolution Explorer (GALEX), Hubble Space Telescope (HST), Sloan Digital Sky Survey (SDSS), and Canada–France–Hawaii Telescope (CFHT)). Table 1 summarizes the archival data used in this paper. The data include 12 images obtained with far-UV (FUV), near-UV (NUV), F336W, u, FQUVN, g, F588N, F606W, r, F673N, i, and z filters. The basic data reduction steps of bias subtraction, flat-fielding, and alignment are carried out by the pipeline of each data set. For the data from HST, we use the lacos_im task (van Dokkum 2001) to remove cosmic rays. The flux calibration is also given by the pipeline, except the imaging data from the CFHT. We indirectly derive the zero point of the CFHT/u image by using the count rates of five bright stars with known magnitude from SDSS.

The ULX is visually detected in most images (Figure 4). But in FUV and z, it is unclear whether the counterpart is present at the position of the ULX. The position of the optical counterpart in the ground-based images is well matched with that derived in X-ray data within its positional uncertainty (03). But the absolute astrometry of the HST images, especially those obtained before Cycle 14, which have a typical astrometric error of 1''–3'' due to the uncertainty of the guide star positions (Koekemoer et al. 2007), is not as accurate as other data. Thus, we correct the astrometry of the HST images using the sources with known positions from the SDSS images. However, owing to the narrow field of view and shallowness of the HST images, we were only able to use three faint sources for the registration. Given the typical astrometric error of the SDSS images (≥01; Pier et al. 2003), the uncertainties of the astrometry in the HST images might be slightly larger than 01. We again find that the optical counterparts in the HST images are coincident with the X-ray position within 05. Note that the aperture radii (04–9'') for the aperture photometry are significantly larger than the astrometric uncertainties (∼01) of the optical images, so that the photometric error due to the image misalignment should be negligible.

Zoom In Zoom Out Reset image size

We extracted X-ray counts in the energy range of 0.3–8 keV from a circular region of 25 radius centered at the source. Background was taken locally from an annular region also centered at the source. The total number of background-subtracted source counts is 391 ± 21. Spectra and instrument responses were generated and combined using specextract. Spectra were grouped to have a minimum of 20 counts per energy bin to allow for χ² fitting. The spectral modeling was performed with XSPEC version 12.7 (Arnaud 1996). We fit the X-ray spectrum with an absorbed power-law model. The fit prefers negligible intrinsic absorption, so we fixed the absorption column at the Galactic neutral hydrogen column density in the sightline toward NGC 5252, N_H = 1.97 × 10²⁰ cm⁻² (given by Dickey & Lockman 1990). The best-fit power-law photon index Γ = 1.65 ± 0.11, with a reduced ${\chi }^{2}/$ dof = 19.8/17. Adding in an extra thermal emission or power-law component does not further improve the fit statistically. The absorption-corrected 0.5–8 keV flux is (1.33 ± 0.11) × 10⁻¹⁴ erg cm⁻² s¹. Adopting a distance of 98.4 Mpc to NGC 5252, this corresponds to L_X = 1.5 × 10⁴⁰ erg s⁻¹, placing the source into the ULX domain. We further examined the light curves for intra-observation and inter-observation variabilities and found no significant variations. Comparing to a constant light curve, the probability of variable signal is negligible in all cases. However, we found a long-term variability over 10 yr, evaluated using the count rates in 2003 and 2013. A χ² test against a constant light curve indicates that the probability for no variation is 0.0002, which implies that the long-term variation is statistically significant.

Strong narrow emission lines are found in the position of the ULX (Figure 5). The [O iii] and Hα luminosities corrected for the extinction using the observed Balmer decrement (Hα/Hβ ≈ 3.9) are as large as 10^39.7 and 10^39.2 erg s⁻¹, respectively. We note that these fluxes might be overestimated by 0.2 dex owing to the extended structures in the vicinity of the ULX (Section 3.3). Thus, if the lines are ionized by the source in the ULX, its energy is comparable to that of nearby low-luminosity AGNs (Ho 2008). However, it could also be ionized by the central AGN, which is ∼22'' away from the ULX.

Zoom In Zoom Out Reset image size

Figure 6 shows distributions of [O iii] flux and systematic velocity along the slit. As already known from previous studies, NGC 5252 harbors an extended narrow-line region (ENLR) up to a radius of 50'' (e.g., Tadhunter & Tsvetanov 1989). The ULX is located on the major axis of the ENLR. Acosta-Pulido et al. (1996) also reported an excess of line flux in [O iii] and Hα close to the ULX, while the slit in their study is slightly off from the ULX (NW2 in Acosta-Pulido et al. 1996). Using the flux ratios among optical emission lines (Hβ, [O iii] λ5007, Hα, [N ii] λ6583), one can determine the source of ionizing radiation (Baldwin et al. 1981). Figure 7 indicates that the ENLR is mainly ionized by the AGN rather than hot stars from the star-forming region, which is in good agreement with previous studies (e.g., Acosta-Pulido et al. 1996). Since the strong X-ray emission, optical emission lines, and radio continuum can be associated with shocks (e.g., Sulentic et al. 2001; Konstantopoulos et al. 2014), it is worth testing this possibility. But the flux ratios among oxygen lines (log ([O iii] λλ4363/5007) ≈ −1.9 and log ([O ii] λ3727/[O iii] λ5007) ≈ −1.6) lead us to rule out a significant contribution from the shocks (Dopita & Sutherland 1995; see also Morse et al. 1998).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Owing to the low instrumental spectral resolution (${\sigma }_{{\rm{ins}}}\sim 100-200$ km s⁻¹), the emission lines are barely spectrally resolved at the position of the ULX. The velocity distribution shows a rotating pattern within ∼5''–6'' of radius from the nucleus, which is in good agreement with previous studies (Tsvetanov et al. 1996; Morse et al. 1998). But above that radius, the dynamics of the ionizing gas appears to be complicated. Interestingly, we note that the systematic velocity offset of the ULX relative to the nucleus of the galaxy is fairly small (∼13 km s⁻¹). This indicates that the ULX might be gravitationally bound to NGC 5252.

Besides the strong emission lines, we find weak signs of underlying continuum (see Figure 3). The signal-to-noise ratio (S/N) of the continuum flux is ≈2.0 between ∼5000 Å and ∼8000 Å, while the S/N dramatically decreases ($\leqslant 1$) below 4500 Å. Thus, the absolute flux calibration below 4500 Å may be very uncertain. Whereas the source of the continuum is unclear, the spectral energy distribution (SED) based on the broadband photometry suggests that it is not entirely explained by the stellar continuum unless the stellar population is very young (∼10⁶⁻⁷ yr). No broad component is detected in Hβ or Hα. We carefully examine the spectrum of the ULX to test the hypothesis that X-ray continuum comes from a background source. But we fail to find any signature (e.g., emission line, featureless continuum) of its existence. We will discuss this further in Section 4.2.

The high-resolution images from HST reveal that the optical counterpart is compact and marginally resolved (e.g., F336W), while there is a sign of extended features associated with the ULX in the low-resolution images (e.g., NUV, u, and g) and the HST images with narrowband filters affected by strong emission lines (e.g., F673N). Using the F336W image, which has the highest S/N among the images, we estimate the size of the optical counterpart. Using GALFIT (Peng et al. 2002; Peng et al. 2010), we fit the source by employing a synthetic PSF generated by TinyTim (Krist et al. 1995). We model the source with a Sérsic (Sérsic 1968) profile or a Gaussian profile. The best fit yields that the effective radius (r_e) is ∼01 (46 pc). However, since the derived r_e is comparable to the size of the PSF, it thus shall be regarded as an upper limit. Even with this caveat, it is intriguing that this upper limit of the size is consistent with that of ultracompact dwarfs but significantly larger than that of globular clusters (e.g., Forbes et al. 2013).

Since the archival images are obtained with different detectors and in different observing conditions, it is important to measure the fluxes of the ULX in a consistent way. Moreover, it is crucial to remove the light of the host galaxy correctly because the gradient in the background can introduce an additional error on the photometric measurements. We model the host galaxy component with a Sérsic profile using GALFIT. We find that the host galaxy is well fit with two or three Sérsic components, but we do not ascribe physical significance to each individual component. After the host galaxy contribution is removed, we perform aperture photometry as a function of radius. The optimal aperture radius (red circles in Figure 4) is determined where the cumulative flux within the radius reaches a constant. If the source appears to be extended (e.g., u and F673N), we use the aperture radius determined from other images obtained with the same or a similar detector. Table 2 lists the flux measurements from the aperture photometry. While the flux in the FUV is very uncertain (∼0.3 mag uncertainty), we note that it is also contained in the GALEX archival catalog (Morrissey et al. 2007); the catalog lists an AB magnitude of 20.58 mag, which is slightly brighter than, but within the uncertainty of, our measurement of 20.92 mag. Thus, our measurement in FUV is reliable.

The F673N and F606W images show that there are faint, extended substructures to the southwest and northeast of the ULX, probably arising from Hα + [N ii] emission. This emission likely lies partly or entirely within the aperture size of our extracted spectrum, although its nature and its association with the ULX are unclear. From aperture photometry of the F673N and F606W images with and without the substructures masked, we estimate that its contribution to the ULX spectrum is at most 40%. Therefore, it is possible that the luminosities of emission lines (e.g., [O iii] and Hα) are overestimated by 0.2 dex owing to the substructures. Throughout this paper, we take this factor into account.

The optical counterpart is also detected by SDSS Data Release 7 (Abazajian et al. 2009). Their measurements are broadly in good agreement with ours in g, r, and i. But there is substantial discrepancy in u and z, which are much shallower than other bands. For the u band, we use a deeper image obtained with CFHT rather than that from SDSS. The flux from the z band is quite uncertain with an error of ∼0.77 mag.

We note that the flux density from the spectroscopy broadly agrees with the broadband photometry, except the blue part below 4000 Å (Figure 8). While it is unclear why the discrepancy is remarkable in the blue part, it indicates that there might be variability in the UV. This finding may be regarded as further evidence that the continuum source is nonstellar. Because we subtracted the local host galaxy light in the vicinity of the ULX when we extracted the spectrum (Section 2.2), the contribution from the host galaxy should be negligible.

Radio images of NGC 5252, obtained multiple times for ∼3 yr (from 1991 to 1993), show a source in coincidence with the position of the ULX (Wilson & Tsvetanov 1994; Becker et al. 1995; Kukula et al. 1995). The offset between the radio peak and the X-ray position of the ULX is smaller than 03. The radio flux densities of the ULX are on average 1.50 ± 0.12 mJy, 1.9 ± 0.2 mJy, and 3.14 ± 0.06 mJy at 3.6, 6, and 20 cm, respectively. While the beam sizes of the radio images range from 01 to 5'', the radio source at the position of the ULX is unresolved for all the radio data. The spectral index of the radio continuum between 3.6 and 20 cm of the counterpart of the ULX is ∼0.5 (α in ${f}_{\nu }\propto {\nu }^{-\alpha }$), which is consistent with that of nearby Seyferts (Ho & Ulvestad 2001). We find that the variability over 3 yr is less than 10% at 3.6 and 20 cm (see also Wilson & Tsvetanov 1994).

Strong emission lines are detected at the position of the ULX, and its velocity offset with respect to the center of NGC 5252 is remarkably small (∼13 km s⁻¹). This suggests that the ULX is associated with NGC 5252. However, it is still inconclusive mainly because the emission lines underneath the ULX could be contamination from the ENLR. To address this issue, we investigate the profile of [O iii] in the vicinity of the ULX. Intriguingly, we find a velocity gradient within a radius of 2''. As shown in Figure 10, the velocity inferred from the peak value of the [O iii] line is maximized at radii of ∼−2'' and 14 (916 and 640 pc), where the maximum velocities are −80 and 120 km s⁻¹, respectively. This strongly suggests that the line-emitting gas is affected by and therefore is physically associated with or in close proximity to the ULX. If so, then the ULX is not a background source but is physically located within NGC 5252.

Zoom In Zoom Out Reset image size

While the [O iii] line is barely resolved, it appears that the line dispersion slightly varies (${\rm{\Delta }}\sigma \leqslant 50$ km s⁻¹) within a radius of 2'', which indicates that the kinematics of the gas are affected by the ULX. This further suggests that the ULX is located within NGC 5252. The gas itself may be physically excited by the ULX, which appears to be powered by an active BH.

Although we do detect emission lines at the position of the ULX, it is still difficult to prove that the line emission comes from the ULX or is, instead, simply confused by the ENLR. Thus, it is worthwhile to consider whether the ULX might be an interloper. It is common that ULXs turned out to be foreground stars or background AGNs (Foschini et al. 2002; Masetti et al. 2003; Gutiérrez & López-Corredoira 2005; Wong et al. 2008; Gutiérrez 2013). Moreover, the incidence of background AGNs appears to be higher for ULXs found in early-type galaxies (e.g., Gutiérrez 2006; Heida et al. 2013).

As discussed in Section 3.2, we do not see any signs of emission lines, narrow or broad, from a background AGN in the spectrum. How strong should any hypothetical line emission be? We use the empirical correlation between X-ray and [O iii] luminosity to estimate the strength of the expected optical line emission, if the source were a line-emitting background AGN. We adopt a conversion factor of ${\rm{log}}\ ({L}_{[{\rm{O}}\;{\rm{III}}]}/$ ${10}^{42}\ \mathrm{erg}\ {{\rm{s}}}^{-1})=0.91\ {\rm{log}}$ $({L}_{{\rm{2}}\;{\rm{keV}}}/{10}^{42}\;\mathrm{erg}\;{{\rm{s}}}^{-1})-1.21$ based on type 1 AGNs (Stern & Laor 2012a, 2012b) to calculate the expected [O iii] luminosity, given a rest-frame 2 keV luminosity estimated from the observed Γ = 1.65. [O iii] can be detectable up to z ∼ 0.8 owing to the wavelength limit of the spectrum. We expect an [O iii] luminosity ranging from 10⁴⁰ to 1.5 × 10⁴² erg s⁻¹ at 0.1 < z < 0.8. Given the S/N of the continuum of the optical spectrum, we can detect [O iii], if it is unresolved at the instrumental resolution of our spectrum, up to z ∼ 0.8 at a significance of 6σ. This estimate assumes negligible internal extinction, which is not unreasonable given the low observed X-ray column.

Having ruled out the possibility that the ULX is a low-redshift AGN, we now evaluate whether it can be an AGN at z > 0.8. With ${L}_{{\rm{2keV}}}\gt {10}^{43.5}$ erg s⁻¹, such a hypothetical source would be luminous enough to qualify as a type 2 quasar (e.g., Ptak et al. 2006), but the low observed column density is incompatible with the heavy obscuration typical of this class (Jia et al. 2013). At the same time, the featureless optical spectrum, which would correspond to rest-frame UV, is also inconsistent with that of a type 1 quasar. To estimate the expected strength of broad UV emission lines, we adopt the empirical scaling between the strength of the [O iii] and Mg ii and C iv, as well as the median line width of FWHM = 4500 km s⁻¹, as derived from the large SDSS quasar catalog of Shen et al. (2011). Given the S/N of our spectrum, we can rule out the existence of broad UV emission lines at the level of 10σ.

From the above analysis, we reject the possibility that the ULX is a line-emitting background AGN at any reasonable redshift. What about a weak-line AGN, namely, a BL Lac object? We argue against this possibility using three lines of evidence. The compactness of the radio source (Kukula et al. 1995) is consistent with the radio morphology of BL Lacs. First, the SED of the ULX does not look like that of typical BL Lacs (Fossati et al. 1998; Donato et al. 2001). The left panel of Figure 10 shows that there is substantial discrepancy, especially in the UV, between the SED of the ULX and those of BL Lacs. We varied the redshift but could not find any satisfactory solution that can simultaneously match the observed SED of the ULX in the UV and X-ray bands. One obvious weakness in this conclusion, of course, is that the SED measurements were not simultaneous, whereas BL Lac objects are known to be highly variable.

Zoom In Zoom Out Reset image size

Second, the optical properties of the ULX do not agree with the expected properties of a BL Lac host. BL Lac objects are commonly hosted by massive (e.g., Falomo 1996; Urry et al. 1999; Kotilainen et al. 2005), luminous ellipticals with M_R = −22.9 ± 0.5 mag (e.g., Sbarufatti et al. 2005). Thus, if the ULX is a BL Lac object, one can expect that the host galaxy would be detected in the HST images. By using the distribution of the apparent R-band magnitude of BL Lac host galaxies (Sbarufatti et al. 2005), we find that the host galaxy would have been detected if the ULX is a low-redshift BL Lac with z < 0.4. However, a high-z BL Lac cannot be ruled out as an interloper because the expected r-band host brightness for BL Lacs with $z\geqslant 0.4$ would be lower than our observational limits.

Finally, it is noteworthy that the ULX source does not vary in the radio by more than 10%. BL Lacs, by contrast, are highly variable (>10%) in the radio (Aller et al. 1992). However, it is possible that we do not have enough data to correctly judge the variability because the radio fluxes were obtained in only three different epochs within 3 yr. In order to test this possibility, we perform a Monte Carlo simulation by randomly choosing the data points at three different epochs from monitoring radio data of BL Lacs (Richards et al. 2011). We find that ∼40% of BL Lacs show variability less than 10%. This indicates that additional multiepoch data are needed to use radio variability to test whether the ULX is a background BL Lac. Interestingly, the X-ray data measured at three different epochs for 5 days also do not exhibit short-timescale variability at a level greater than 17%.

It is very intriguing that the SED of the ULX is roughly consistent with that of radio-loud quasars (Shang et al. 2011; see Figure 10, right panel). However, the lack of emission lines, as discussed above, is strongly at odds with this interpretation.

As we discuss above, the probability that the ULX is a background AGN appears to be relatively low. Under the assumption that the X-rays, optical emission lines, and radio continuum are associated with the ULX, it is natural to suppose that it is an off-nucleus stellar-mass BH. For stellar-mass BH binaries, the X-ray flux is correlated closely with the radio flux (e.g., Corbel et al. 2003; Gallo et al. 2003). Using the empirical correlation from Gallo et al. (2003), we find that the expected radio flux density (∼0.003–0.022 mJy) is 2 orders of magnitude smaller than the observed flux value (∼1.5–3.1 mJy) if the ULX is a "low/hard"-state BH binary. For an X-ray binary to be as luminous in the radio as observed relative to the X-rays, it would have to be in the "very high" state (Fender et al. 2004). However, this is incompatible with the lack of X-ray and radio variability, which is typically observed in very high state X-ray binaries (Miyamoto et al. 1991; Kubota & Done 2004). In addition, the X-ray spectrum is expected to be very steep in the very high state (${\rm{\Gamma }}\geqslant 2.5;$Miyamoto et al. 1991; Hori et al. 2014), which is inconsistent with the value of Γ = 1.65 observed for the ULX. We thus argue that our source is unlikely to be a stellar-mass BH.

If the ULX is an off-nucleus nonstellar BH, for example, from the accretion of a tidally stripped galaxy nucleus, one can expect that the line emission should correlate with the ionizing continuum from the BH. To test this scenario, we use the well-known correlation between [O iii] luminosity and X-ray luminosity in AGNs (e.g., Bassani et al. 1999; Heckman et al. 2005). Figure 11 shows that the observed ${L}_{{\rm{X}}}/{L}_{[{\rm{O}}\;{\rm{III}}]}$ of the ULX is comparable to that of local Seyferts. The X-ray luminosity from 2 to 10 keV is corrected for Galactic and intrinsic absorption. The [O iii] luminosity is corrected for extinction using the observed Balmer decrement (Hα/Hβ ≈ 3.9) and the extinction curve of Fitzpatrick (1999). As discussed in Section 3.3, the [O iii] luminosity may have been overestimated by 0.2 dex. Thus, a lower limit of the [O iii] luminosity is also displayed in Figure 11; however, this is not significant enough to change any of our conclusions. That the ULX follows the L_X−L_{[O iii]} relation of Seyferts suggests that the optical line emission is powered by the local X-ray continuum rather than the central AGN in the host galaxy. As expected, the nucleus of NGC 5252 also falls on the relation; the X-ray and [O iii] luminosities are estimated within the nucleus region (Cruz-Gonzalez et al. 1994; Dadina et al. 2010). Since the X-ray and [O iii] emissions are broadly extended, the luminosities might be slightly underestimated. But the amount of underestimation should be negligible because the flux contribution from the extended region is less than 1% (Dadina et al. 2010).

Zoom In Zoom Out Reset image size

A correlation between radio power and line ([O iii]) luminosity in bright AGNs has been reported by various studies (e.g., Rawlings et al. 1989; Miller et al. 1993). This relation appears to hold also for Seyferts (Ho & Peng 2001). We find that the observed luminosity ratio between [O iii] ($\approx {10}^{39.5-39.7}$ erg s⁻¹) and radio continuum at 5 GHz (≈10^21.3 W Hz⁻¹) is in broad agreement with Seyferts (Figure 11). Indeed, the ULX is bright enough in the radio to be classified as a radio-loud AGN. These findings suggest that the radio, X-ray continuum, and [O iii] emission originate from the same source, namely, an off-nucleus nonstellar, presumably fairly massive BH.

If the ULX is an off-nucleus BH, what is its mass? In order to address this question, we investigate the physical properties of the ULX. In the hypothesis that the bolometric luminosity does not exceed the Eddington luminosity, we can estimate a lower limit of the BH mass from the observed luminosity (but see Abramowicz et al. 1988). The bolometric conversion factors for the [O iii] and X-ray luminosity strongly depend on luminosity, in the sense that they (${L}_{{\rm{bol}}}/{L}_{[{\rm{O}}\;{\rm{III}}]}$ and ${L}_{{\rm{bol}}}/{L}_{{2\mbox{--}10}\;\mathrm{keV}}$) tend to be smaller for lower-luminosity AGNs (e.g., Ho 2008; Stern & Laor 2012b). Adopting the conversion equation from Stern & Laor (2012b), ${L}_{{\rm{bol}}}/{L}_{[{\rm{O}}\;{\rm{III}}]}\sim 162,$ which leads to ${{\bf{M}}}_{{\rm{BH}}}\gt {10}^{3.5}\;$ ${M}_{\odot }$. We can carry out the same experiment using the X-ray luminosity. By adopting the bolometric conversion factor of 16–28 for low- and moderate-luminosity AGNs (Ho 2008), we estimate ${M}_{{\rm{BH}}}\;\geqslant$ ${10}^{3.0-3.6}\;$ ${M}_{\odot }$, consistent with the estimate based on [O iii] luminosity.

The BH mass can also be independently estimated using the fundamental plane of BH accretion (Merloni et al. 2003), which is described as

$$\begin{eqnarray}&&{\rm{log}}\ {L}_{R}=0.6{\rm{log}}\ {L}_{{\rm{X}}}+0.78{\rm{log}}\ {M}_{{\rm{BH}}},\end{eqnarray} \tag{ 1 }$$

where L_R is the luminosity at 5 GHz in units of erg s⁻¹, L_X is the luminosity at 2–10 keV in units of erg s⁻¹, and M_BH is the BH mass in units of ${M}_{\odot }$. By assuming that the ULX follows this relation, we find that ${M}_{{\rm{BH}}}$ is as large as ∼10⁸ ${M}_{\odot }$ (Figure 12). While the fundamental plane has been recalibrated with various types of samples (e.g., Gültekin et al. 2009; Plotkin et al. 2012), our estimate does not change significantly if we adopt the revised versions. A caveat is that this empirical relation is only applicable for BHs accreting in the low state (i.e., ${L}_{{\rm{bol}}}/{L}_{{\rm{Edd}}}\leqslant 0.1$). If the BH mass is indeed as large as suggested by the fundamental plane relation, we are faced with a serious challenge. First, the Eddington ratio would be very low (∼2 × 10⁻⁵). Such a low Eddington ratio should produce a low-ionization parameter for the optical emission lines (Ho 2008, 2009), similar to those seen in low-ionization nuclear emission-line regions, which have [O iii]/Hβ < 3 (Ho et al. 1997). Instead, the ULX is observed to have Seyfert-like line ratios, with [O iii]/Hβ ∼ 9.

Zoom In Zoom Out Reset image size

If the ULX is really associated with a nonstellar mass BH, it raises the question about its nature. Morse et al. (1998) showed that the kinematics of the ionized gas in NGC 5252 can be modeled with at least three different components, which suggests an external origin of the gas. Although NGC 5252 is an early-type galaxy, it has abundant H i gas (M_HI ∼ 10⁹${M}_{\odot }$), and its structure is highly complicated (Prieto & Freudling 1996). Thus, it seems that NGC 5252 may have undergone a gas-rich merger. In this case, we might speculate that the off-nucleus BH powering the ULX is the remnant of an already-merged companion galaxy. If our estimation for the BH mass (∼10⁴⁻⁸ ${M}_{\odot }$) is correct, the merging galaxy should have had a bulge with stellar mass of ∼10^7.0−10.4 ${M}_{\odot }$, derived from the correlation between BH mass and bulge stellar mass (Kormendy & Ho 2013).

For comparison, we estimate the total stellar mass of NGC 5252 to be ${M}_{*}\approx 1.4\times {10}^{11}\;$ ${M}_{\odot }$, using its K-band luminosity (M_K = −24.55 mag; Peng et al. 2006) and stellar velocity dispersion (${\sigma }_{*}=209$ km s⁻¹; Cid Fernandes et al. 2004) following the conversions given in Kormendy & Ho (2013). If the ULX has a BH mass as large as ∼10⁸ ${M}_{\odot }$, its progenitor bulge of ${M}_{*}\sim {10}^{10.4}\;$ ${M}_{\odot }$ implies that NGC 5252 experienced a fairly significant merger with a mass ratio of at least 1:6. This seems highly implausible, in view of the fact that NGC 5252 shows no evidence for any perturbation (tidal features, lopsidedness, etc.) in its current stellar distribution. On the other hand, it is not unreasonable that NGC 5252 could have accreted a small dwarf galaxy with ${M}_{*}\sim {10}^{7}\;$ ${M}_{\odot }$ that would have accompanied a ∼10⁴ ${M}_{\odot }$ BH. Such a minor merger event should have little impact on the global structure of a galaxy 10⁴ times more massive.

Unlike most ULXs studied to date, this particular ULX is unusual in that it has a prominent counterpart in the optical continuum. The physical nature of the optical continuum is unclear. We do not detect any obvious stellar features, but it is difficult to rule out that the continuum comes mainly from starlight, presumably the nucleus of a larger galaxy whose outer parts have already been stripped. In the following, we assume that the optical continuum is stellar and use it to place an upper limit on the stellar mass associated with the ULX. To minimize the possible contribution from the featureless continuum of the BH accretion disk, we use i-band photometry, which gives M_i = −13.3 mag. While mass-to-light ratio strongly depends on the initial mass function, metallicity, and age of the stellar population, we approximately estimate an upper limit of stellar mass of ∼10⁸ ${M}_{\odot }$ using the stellar population synthesis models from Vazdekis et al. (2012) and Ricciardelli et al. (2012).

Interestingly, the properties of the ULX are comparable to those of an ultracompact dwarf galaxy (M60-UCD1) around NGC 4649 (Strader et al. 2013). M60-UCD1 is very compact (r_e ≈ 24 pc) and one of the most luminous UCDs known (M_g ≈ −13.7). There appears to be an X-ray source at the center of M60-UCD1, although its luminosity (∼10³⁸ erg s⁻¹) is significantly lower than that of the ULX studied here. While the origin of the X-ray emission is somewhat uncertain, it is very likely to originate from a massive BH at the center, rather than a stellar BH (Sivakoff et al. 2007; Strader et al. 2013; Seth et al. 2014). Strader et al. (2013) and Seth et al. (2014) argue that M60-UCD1 is a tidally stripped remnant of a merging galaxy, which is consistent with our expectation for the ULX in NGC 5252.

Interestingly, the properties of the ULX in this study are somewhat similar to those of HLX-1 in ESO 243–39. Both of them seem to have BHs significantly larger than stellar-mass BHs, reside in early-type galaxies at similar distances (z ∼ 0.023), and have optical counterparts. In addition, their origin is thought to be merging galaxies. However, at the same time, it is intriguing to note that there are significant differences between the two sources. The ULX in NGC 5252 shows much higher flux ratio of optical emission to X-ray continuum. The luminosity of Hα emission of the ULX in NGC 5252 is $\sim 40-90$ times greater than that of HLX-1 (Wiersema et al. 2010), whereas the X-ray luminosity of the ULX in NGC 5252 is only 1%–65% of that of HLX-1 (Servillat et al. 2011). This implies that the difference in ${L}_{{\rm{X}}}/{L}_{{\rm{H}}\alpha }$ between the two sources is a factor of 90–9000. Moreover, the radio flux density of the ULX in NGC 5252 is at least ∼40 times larger than that of HLX-1 (Webb et al. 2012). This might suggest that they have distinctive physical properties (e.g., BH mass and/or accretion rate).

NGC 3341 is another interesting case of an off-nucleus source exhibiting AGN signatures in both the optical and X-rays (Barth et al. 2008; Bianchi et al. 2013). Optical images of NGC 3341 clearly show evidence of ongoing merging with multiple companions (Barth et al. 2008). The off-nucleus AGN appears to be associated with a dwarf galaxy, indicating that this AGN might be triggered during ongoing minor mergers.