NuSTAR unveils a heavily obscured low-luminosity Active Galactic Nucleus in the Luminous Infrared Galaxy NGC 6286

We report the detection of a heavily obscured Active Galactic Nucleus (AGN) in the luminous infrared galaxy (LIRG) NGC 6286, identified in a 17.5 ks NuSTAR observation. The source is in an early merging stage, and was targeted as part of our ongoing NuSTAR campaign observing local luminous and ultra-luminous infrared galaxies in different merger stages. NGC 6286 is clearly detected above 10 keV and, by including the quasi-simultaneous Swift/XRT and archival XMM-Newton and Chandra data, we find that the source is heavily obscured [$N_{\rm\,H}\simeq (0.95-1.32)\times 10^{24}\rm\,cm^{-2}$], with a column density consistent with being Compton-thick [CT, $\log (N_{\rm\,H}/\rm cm^{-2})\geq 24$]. The AGN in NGC 6286 has a low absorption-corrected luminosity ($L_{2-10\rm\,keV}\sim 3-20\times 10^{41}\rm\,erg\,s^{-1}$) and contributes $\lesssim$1\% to the energetics of the system. Because of its low-luminosity, previous observations carried out in the soft X-ray band ($<10$ keV) and in the infrared did not notice the presence of a buried AGN. NGC 6286 has multi-wavelength characteristics typical of objects with the same infrared luminosity and in the same merger stage, which might imply that there is a significant population of obscured low-luminosity AGN in LIRGs that can only be detected by sensitive hard X-ray observations.


INTRODUCTION
Luminous [L IR (8 − 1000 µ m) = 10 11 − 10 12 L ⊙ ] and ultra-luminous (L IR ≥ 10 12 L ⊙ ) infrared galaxies (LIRGs and ULIRGs, respectively) were first discovered in the late sixties (Low & Kleinmann 1968;Kleinmann & Low 1970). With the advent of the Infrared Astronomical Satellite (IRAS, see Sanders & Mirabel 1996 for a review), which discovered a large number of U/LIRGs, the cosmological importance of these objects became evident. Although they are relatively rare at low redshift, their luminosity function is very steep (Le Floc'h et al. 2005), and they are the major contributor to the IR energy density at z ≃ 1 − 2 (e.g., Caputi et al. 2007, Goto et al. 2010. The discovery that most, if not all, U/LIRGs are triggered by galaxy mergers led to the development of an evolutionary scenario (Sanders et al. 1988) in which two gas rich disk galaxies collide, triggering an intense phase of star formation in which they are observed as U/LIRGs. This is then followed by a blowout phase, during which most of the material enshrouding the supermassive black hole (SMBH) is blown away and the system is observed as a luminous red quasar (e.g., Glikman et al. 2015 and references therein). When most of the dust is removed, the system is eventually observed as a blue quasar. This model is consistent with the observed increase of the fraction of obscured sources with redshift up to z ≃ 3 (Treister et al. 2010a;Ueda et al. 2014). The Wide-field Infrared Survey Explorer satellite (WISE) has recently found evidence of a new population of very luminous IR sources (L IR > 10 13 L ⊙ ), dubbed Hot Dust Obscured Galaxies (Hot DOGs, Wu et al. 2012), which might represent a short evolutionary phase in the evolution of galaxies, and be related to mergers (e.g., Eisenhardt et al. 2012;Stern et al. 2014;Assef et al. 2015). Numerical simulations (e.g., Springel et al. 2005) have also shown that tidal interactions can drive an inflow of material that triggers and feeds both accretion onto the SMBH and star formation. Therefore, mergers might play an important role in fuelling the SMBH, as it has been suggested by the discovery that the fraction of active galactic nuclei (AGN) in mergers increases with the AGN luminosity Schawinski et al. 2012), spanning from < 1% at a 2-10 keV luminosity of L 2−10 ∼ 10 41 erg s −1 to 70-80% in the most luminous quasars with L 2−10 ∼ 10 46 erg s −1 .
The contribution of AGN to the overall luminosity of U/LIRGs has been shown to increase with the IR luminosity of the system (e.g., Veilleux et al. 1995Veilleux et al. , 1999Imanishi 2009;Imanishi et al. 2010a,b;Nardini et al. 2010;Alonso-Herrero et al. 2012;Ichikawa et al. 2014). Due to the great opacity of the nuclear region, a clear identification of AGN in U/LIRGs is often complicated. Mid-IR (MIR) properties have been used to estimate the relative contribution of accretion onto the SMBH and star formation to the bolometric luminosity. This has been done by exploiting 5-8 µm spectroscopy (e.g., Nardini et al. 2010), and the characteristics of several features in the L (3-4 µm) and M (4-5 µm) bands (Imanishi & Dudley 2000;Risaliti et al. 2006;Sani et al. 2008;Risaliti et al. 2010): the 3.3 µm polycyclic aromatic hydrocarbon (PAH) emission feature, the bare carbonaceous 3.4 µm absorption feature, and the slope of the continuum. The 6.2µm (e.g., Stierwalt et al. 2013Stierwalt et al. , 2014 and 7.7µm PAH features (e.g., Veilleux et al. 2009), the presence of high-excitation MIR lines (e.g., [Ne V] 14.32µm, Veilleux et al. 2009), or the radio properties (e.g., Parra et al. 2010, Romero-Cañizales et al. 2012a, Vardoulaki et al. 2015 have also been used to infer the presence of a buried AGN. X-ray observations are a very powerful tool to detect accreting SMBHs and to disentangle the contributions of star formation and AGN emission to the total luminosity of U/LIRGs. Studies performed so far using XMM-Newton (e.g., Franceschini et al. 2003;Imanishi et al. 2003;Pereira-Santaella et al. 2011) and Chandra (e.g., Ptak et al. 2003;Teng et al. 2005;Iwasawa et al. 2011) have characterized the properties of a significant number of these systems. However, a significant fraction of U/LIRGs might be heavily obscured (e.g., Treister et al. 2010b;Bauer et al. 2010), and X-rays at energies 10 keV are strongly attenuated in Comptonthick (CT, N H ≥ 10 24 cm −2 ) AGN. Observations carried out in the hard X-ray band (≥ 10 keV) are less affected by absorption, and can be used to probe nuclear X-ray emission even in highly obscured systems (e.g., Baloković et al. 2014, Gandhi et al. 2014, Arévalo et al. 2014, Lansbury et al. 2015, Annuar et al. 2015, Puccetti et al. 2015, Ricci et al. 2015. Previous hard X-ray observations of U/LIRGs have been carried out with BeppoSAX (e.g., Vignati et al. 1999), Suzaku (e.g., Teng et al. 2009) and Swift/BAT (Koss et al. 2013).
The recent launch of the Nuclear Spectroscopic Telescope Array (NuSTAR, Harrison et al. 2013), the first focussing telescope in orbit operating at E ≥ 10 keV, has opened a new window in the study of U/LIRGs thanks to its unprecedented characteristics. The first studies of the hard X-ray emission of local ULIRGs carried out with NuSTAR have recently been reported by Teng et al. (2015) and Ptak et al. (2015), who show the importance of sensitive hard X-ray spectra to well constrain the lineof-sight column density.
We report here on the first results of a series of NuS-TAR observations awarded to our group during AO-1 as a part of a campaign aimed at observing ten local LIRGs in different merger stages (PI: F. E. Bauer). The sources were selected from the Great Observatories All-sky LIRG Survey (GOALS 15 , Armus et al. 2009). GOALS is a local (z < 0.088) sample which contains 181 LIRGs and 21 ULIRGs selected from the IRAS Revised Bright Galaxy Sample (Sanders et al. 2003).
This paper reports the detection of a heavily obscured AGN in NGC 6286 (also referred to as NGC 6286S), a LIRG (log L IR /L ⊙ =11.36, Howell et al. 2010) located at z=0.018349 (i.e., a luminosity distance of d L = 76.1 Mpc), which was not previously detected above 10 keV (Koss et al. 2013). The source has a starformation rate (SFR) of 41.3 M ⊙ yr −1 (Howell et al. 2010), is in an early merging stage (i.e., stage B or 2, following the classification of Stierwalt et al. 2013), and is interacting with the galaxy NGC 6285 (NGC 6286N), located at a distance of ∼ 1.5 arcmin (∼33 kpc, projected distance, see Fig. 1 and panel four of Fig. 2). The source is also known to host a OH megamaser (Baan et al. 1998). The only previous X-ray study of this source, carried out using XMM-Newton observations, did not find any evidence of an AGN (Brightman & Nandra 2011). The XMM-Newton can in fact be well represented by a model taking into account only a collisionally-ionized plasma and an unabsorbed power-law component, representing thermal emission from the starburst and X-ray radiation produced by X-ray binaries, respectively. Possible evidence of very faint AGN activity has been found studying the near-IR to radio spectral energy distribution (SED) , and could be inferred by the detection of [Ne V] lines, although the detection of these features has been questioned by Inami et al. (2013), and due to their weakness they might also be produced by a young starburst.
The paper is structured as follows. In § 2 and § 3 we present the X-ray and radio data used and describe the data reduction procedures, in § 4 we report on the X-ray spectral analysis of NGC 6286, in § 5 we discuss our results by taking into account the multi-wavelength properties of NGC 6286, and in § 6 we summarise the main results of our work. Throughout the paper we adopt standard cosmological parameters (H 0 = 70 km s −1 Mpc −1 , Ω m = 0.3, Ω Λ = 0.7).  The four regions shown in panels one and four represent the 3-8 keV core, the north, central, and south regions discussed in § 4.1. The 1.4 GHz VLA FIRST radio contour is illustrated in panel two, together with the two radio sources (R1 and R2) detected by our analysis of EVN observations at 5 GHz (see § 3), which show that the radio emission coincides with the hard X-ray core, suggesting the presence of a buried AGN. The black circle and the blue dashed annulus in panel three correspond to the NuSTAR source and background extraction regions, respectively. The image in panel four was smoothed with a Gaussian kernel of radius 5 pixels. The blue dashed circle in panel four represents the source region used for XMM-Newton EPIC/PN.

X-RAY OBSERVATIONS AND DATA REDUCTION
2.1. NuSTAR NuSTAR observed NGC 6286 on UT 2015 May 29 for 17.5 ks. We processed the data using the NuS-TAR Data Analysis Software nustardas v1.4.1 within HEASOFT v6.16, adopting the latest calibration files (Madsen et al. 2015). The source is clearly detected in the 3-24 keV image (panel three of Fig. 2). For both focal plane modules (FPMA and FPMB) we extracted source and background spectra and light-curves with the nuproducts task. A circular region of 45 arcsec was used for the source 16 , while the background was extracted from an annulus centred on the X-ray source, with inner and outer radii of 90 and 150 arcsec, respectively. The 3-10 and 10-50 keV light-curves of the sources do not show any evidence of flux variability. exposures of 20.8 and 8.9 ks. Both PN (Strüder et al. 2001) and MOS (Turner et al. 2001) data were analysed by reducing first the observation data files (ODFs) using the XMM-Newton Standard Analysis Software (SAS) version 12.0.1 (Gabriel et al. 2004), and then the raw PN and MOS data files using the epchain and emchain tasks, respectively. In order to filter the observations for periods of high background activity we analyzed the EPIC/PN and MOS background light curves in the 10-12 keV band and above 10 keV, respectively, and found that both observations show a significant background contamination. Observation 0203391201 was not used because the background flux dominates the whole observation (with an average count-rate of 6 ct s −1 and a minimum of 2 ct s −1 ). Observation 0203390701 showed less contamination, and we filtered the periods of high background activity using a threshold of 2 ct s −1 for both PN and MOS, which resulted in net exposure times of 2.3 and 4.7 ks, respectively. For both cameras we extracted the spectrum of the source using a circular region of 20 arcsec radius, while the background was extracted from a circular region of 40 arcsec radius, located on the same CCD of the source and in a zone devoid of other sources. No significant flux variability is found in the 0.3-10 keV band during the XMM-Newton observation. with an exposure of 14.2 ks. The data reduction was performed following the standard procedure, using CIAO v.4.6. The data were reprocessed using chandra repro, and then the spectra were extracted using the specextract tool.
The 0.2-2 keV Chandra image shows clear evidence of extended emission (panel one of Fig. 2). The 3-8 keV image (panel two) shows instead only a point-like source, which does not appear in the 0.2-2 keV image. This source is located at the center of the galaxy (see panel three) and could be associated with AGN emission. The spectra of these different regions are discussed in § 4.1.
In order to be consistent with the spectral extraction of XMM-Newton and Swift/XRT, which have a much lower spatial resolution than Chandra, the ACIS-S source spectrum used for the broad-band X-ray fitting was extracted from a circular region of 10.5 arcsec radius. The background spectrum was extracted from a circular region of the same size on the same CCD, where no other source was detected. Parra et al. (2010) reported VLA observations of NGC 6286 at 4.8 GHz, showing a compact morphology with a size of 0.25 × 0.21 arcsec and a flux density of 15.24 mJy. They also observed this galaxy with three of the most sensitive antennas (Effelsberg, Westerbork and Lovell) of the European very long baseline interferometry (VLBI) Network (EVN) at 5 GHz, and detected fringes in each one of the baselines with amplitudes between 5.36-6.11 mJy. We have extracted these observations from the archive and applied the pipelined calibration available. In Figure 3 we show the contour map obtained using the cleaning algorithm within the Caltech program difmap (Shepherd et al. 1995). No proper flux density could be obtained with such a small array since the measurements are still subject of instrumental amplitude errors. We can however rely on the source structure, as there is enough information to determine phase closure. We find two milli-arcsec sources with S/N > 5, one (R1) at RA= 16 h 58 m 31.7374 s , DEC = +58 • 56 ′ 14.705. ′′ , and the other (R2) at RA=16 h 58 m 31.6572 s , DEC= +58 • 56 ′ 14. ′′ 167. These two sources are consistent with the 3-8 keV core region (see panel two of Fig. 2 and § 4.1).

RADIO OBSERVATIONS AND DATA REDUCTION
We have also extracted and analysed the very long baseline array (VLBA) experiment BC196 observed at 8 GHz on UT 2012 January 12. We used the NRAO Astronomical Image Processing System (aips) to reduce the data, following standard procedures. We note that the source chosen as phase reference (J1651+5805) is not detected in this experiment, and constraints for it are also not available in the VLBA Calibrator search engine at NRAO. We have resorted to the use of another nearby calibrator (J1656+6012, at 2.22 • ) which was observed 2 min before the NGC 6286 scan. We found that there are no sources detected above ∼ 0.8 mJy/beam (3×r.m.s.) in the VLBA observations convolved with a 3.16 × 0.94 arcsec at 29.31 • beam. If any of the sources detected with the EVN is the AGN core, we would expect a similar peak intensity measured in a baseline with comparable length as that from Ef-Jb or Ef-Wb baselines. The fact that we do not detect any source in the VLBA observations leaves two possible explanations: i) the sources are variable and the VLBA observations are not sensitive enough; ii) the sources are resolved at resolutions better than ∼ 3 arcsec. Although the VLBA array includes three times as much antennas as the EVN small array, it also observed the target for only 1/3 of the time with respect to the EVN, and using antennas ∼3-100 times smaller than those in the small EVN array. We made the exercise of producing an image with similar uv-range for both EVN and VLBA observations. The obtained uv-coverages result in the VLBA being sensitive to emission close to perpendicular to the structure we detected with the EVN (at an inclination ∼ 50 • ), and since there is no emission in such orientation, the VLBA cannot detect any structure, unlike the EVN. Threfore, in order to better constrain the radio emission of NGC 6286, further VLBI observations covering proper hour angles at high sensitivity are needed.

X-RAY SPECTRAL ANALYSIS
The X-ray spectral analysis was performed within xspec v.12.8.2 (Arnaud 1996). Galactic absorption in the direction of the source (N G H = 1.8 × 10 20 cm −2 , Kalberla et al. 2005) was taken into account by adding photoelectric absorption (TBabs in xspec, Wilms et al. 2000). Abundances were set to the solar value. Spectra were rebinned to have at least 20 counts per bin in order to use χ 2 statistics, unless reported otherwise.
In the following we first present the X-ray spectral analysis of the extended and nuclear emission revealed by Chandra ( § 4.1) and then discuss the spatially integrated broad-band X-ray emission ( § 4.2) considering all observations available.

Extended and nuclear emission
The diffuse soft X-ray emission detected by Chandra has an angular size of ∼ 12 arcsec, which at the distance of the source corresponds to ∼ 4.4 kpc. This diffuse emission might either be related to thermal plasma in a star-forming region, to X-ray binaries, or to shocks created by the interaction between outflows from the AGN and the galactic medium. To analyse the diffuse and nuclear emission we extracted the spectra of the four regions shown in panel one of Fig. 2. Besides the 3-8 keV core, in order to study how the extended emission varies, we arbitrarily selected three regions (A, B and C) where most of the 0.2-2 keV photons were detected. Due to the low number of counts we rebinned the spectra to have at least 1 count per bin, and used Cash statistics (Cash 1979) to fit the data. In the following we discuss the spectral properties of the Core, and the regions A, B and C.
Core. The spectrum of the core was extracted from a circular region of radius 1.5 arcsec centred on the peak of the 3-8 keV emission. Ignoring the data below 1.2 keV to avoid contamination from the diffuse soft X-ray emission, and fitting with a power-law model (tbabs Gal ×zpowerlaw in xspec) we obtain a photon index of Γ = −0.17 +1.01 −1.03 . This low value is indicative of heavy absorption in the nuclear region. Fitting the X-ray spectrum using the whole energy range with a model that includes also a collisionally ionized plasma model (tbabs Gal ×zpowerlaw+apec) we obtain C-stat/DOF=20.9/21, Γ ≤ −0.03 and a plasma temperature of kT = 0.99 +0.28 −0.35 keV. The 3-8 keV core coincides with the 1.4 GHz radio emission measured by the VLA FIRST survey (Becker et al. 1995).
Region A. Fitting the spectrum with a collisionally ionized thermal plasma model (tbabs Gal ×apec) results in a good fit (C-stat/DOF=31.0/37), with kT = 0.91 +0.10 −0.18 keV. Applying a spectral model which reproduces a non-equilibrium plasma created in a shock (pshock in xspec) yields C-stat/DOF=31.4/36, a plasma temperature of kT S = 0.88 +0.11 −0.13 keV and an upper limit on the ionization timescale of τ u ≥ 1.6 × 10 12 s cm −3 .
Region B. Using the thermal plasma model yields a rather poor fit (C-stat/DOF=48.0/34). This can be improved by adding photoelectric absorption (tbabs Gal ×ztbabs×apec, C-stat/DOF=43.7/33), and would be consistent with the presence of larger absorption in the central part of the edge-on galaxy with respect to other regions. The shock plasma model fails to reproduce well the spectrum both without (Cstat/DOF=48.7/33) and with (C-stat/DOF=43.6/32) an absorption component. Region C. A thermal plasma model with kT = 1.23 +0.37 −0.35 keV yields a good fit (C-stat/DOF=34/34), while a shock plasma model cannot reproduce well the data (C-stat/DOF=45.1/33), and results in kT s = 1.09 +0.73 −0.38 keV and τ u ≤ 1.4 × 10 8 s cm −3 .

4.2.
Spatially integrated X-ray emission The XMM-Newton EPIC spectrum of NGC 6286 was analysed by Brightman & Nandra (2011), who found that it could be well represented by an unabsorbed power-law continuum plus a thermal plasma 17 , with the photon index fixed to Γ = 1.9. This is in disagreement with the hard (Γ = −0.17 +1.01 −1.03 , see § 4.1) 1-8 keV spectrum of the 3-8 keV core. Fitting the NuSTAR FPMA/FPMB data with a simple power-law model we also find a very flat continuum (Γ = 0.49 +0.46 −0.41 ). The low values of the photon index obtained in the 3-8 keV and 3-30 keV bands could indicate that the X-ray emission is highly absorbed.
While the model used by Brightman & Nandra (2011) can reproduce well the XMM-Newton and the spatially integrated Chandra spectra, it severely under predicts the hard X-ray flux inferred by NuSTAR, as illustrated in Fig. 4. This might be related either to heavy obscuration of the X-ray source or to flux variability between XMM-Newton and NuSTAR observations, although variability would not be able to explain the very flat hard X-ray spectrum. The Swift/XRT observation allows us to constrain the flux level below 10 keV band at the time of the NuSTAR observations. We find that the Swift/XRT 0.3-2 and 2-5 keV fluxes are consistent with that inferred by Chandra and XMM-Newton EPIC/PN observations (see Table 1), which implies the lack of significant variability 17 zpowerlaw+apec in xspec  Table 1 Observed X-ray fluxes between the different observations. To further test the variability scenario we fitted NuSTAR and the spatially integrated Chandra spectra with a model that consists of a power-law plus a thermal plasma [tbabs Gal ×(apec + power law)], allowing for different normalisations of the power-law continuum to vary (fixing Γ = 1.9). We found that the model cannot reproduce well the spectra (χ 2 /DOF=28.9/20), with the fit 18 showing clear residuals between 10 and 30 keV. This rules out variability as a likely explanation for the large ratio between the 10-50 keV and 2-10 keV fluxes. In the following we report the results obtained by adopting several different X-ray spectral models to infer the properties of the AGN in NGC 6286. In order to reduce the possible degeneracies in the models, we constrained the average properties of the diffuse soft Xray emission. To do this we first extracted the Chandra X-ray spectrum of the diffuse emission by excluding from the circular region of 10.5 arcsec a circle of 1.5 arcsec centred on the 3-8 keV core. We then fitted the spectrum with a model that includes i) a thermal plasma and ii) a power-law component (Γ = 1.9) to take into account the scattered emission. We obtained a normalization of the power law n scatt po = (1.03 ± 0.37) × 10 −5 ph keV −1 cm −2 s −1 , and a temperature and normalization of the thermal plasma of kT = 0.77 +0.07 −0.08 keV and n apec = (2.02 ± 0.33) × 10 −5 ph keV −1 cm −2 s −1 , respectively. In all the spectral models reported below we set n scatt po , kT and n apec to the values obtained for the dif-18 the ratio of the power-law normalisations is ≃ 4.
fuse emission, and allow them to vary only within their 90% uncertainties.

PEXRAV
To infer the value of the line-of-sight column density (N H ) we fitted the joint Swift/XRT, Chandra ACIS-S, XMM-Newton EPIC/PN and MOS, and NuSTAR FPMA and FPMB data with a model that consists of: a) an absorbed power-law with a photon index fixed to Γ = 1.9, consistent with the average value of AGN (e.g., Nandra & Pounds 1994a;Piconcelli et al. 2005;Ricci et al. 2011), b) unabsorbed reprocessed X-ray emission from a slab, c) a Gaussian to reproduce the fluorescent Fe Kα emission line (with the rest-frame energy fixed to E Kα = 6.4 keV), d) a second power-law to reproduce the scattered component, and e) emission from a collisionally ionized plasma. To reproduce the effect of obscuration we included both Compton scattering and photoelectric absorption. Reprocessed X-ray emission (excluding fluorescent lines) was taken into account using the pexrav model (Magdziarz & Zdziarski 1995). The fraction of scattered flux (f scatt ) is calculated as the ratio between the normalization at 1 keV of the primary power law (n po ) and n scatt po . The width of the Gaussian line was fixed to σ = 40 eV, consistent with the results obtained by Chandra/HETG studies (e.g., Shu et al. 2010). An Fe Kα line at 6.4 keV is usually found in the X-ray spectrum of AGN (e.g., Nandra & Pounds 1994b, Shu et al. 2010, Ricci et al. 2014b, and is believed to originate in the material surrounding the SMBH (e.g., Ricci et al. 2014a, Gandhi et al. 2015 andreferences therein). In xspec our model is: tbabs Gal (ztbabs×cabs×zpowerlaw + pexrav + zgauss + apec + zpowerlaw). The model yields a good fit (χ 2 /DOF=47.2/44) and results in a column density consistent with borderline Compton-thick obscuration (N H = 1.32 +0.82 −0.54 × 10 24 cm −2 ). Due to the low signal-to-noise ratio, the Fe Kα is not spectrally resolved, and only an upper limit of its equivalent width was obtained (EW ≤ 2318 eV), which is consistent with heavy obscuration.

TORUS
To further study the absorbing material we used the torus model developed by Brightman & Nandra (2011), Figure 6. Value of ∆χ 2 = χ 2 −χ 2 best (where χ 2 best is the minimum value of the χ 2 ) versus the column density for the different Xray spectral models discussed in § 4.2. The horizontal dashed line represents ∆χ 2 = 2.7. The plot shows that NGC 6286 is heavily obscured, with N H consistent with the source being CT for the five models considered.
which considers reprocessed and absorbed X-ray emission from a spherical-toroidal structure. In this model the line-of-sight column density is independent of the inclination angle, which we fixed to the maximum value permitted (θ i = 87.1 • ). Similarly to what was done for pexrav, we added to the model a power law, to take into account the scattered emission, and a collisionally ionized plasma model. In xspec, the model is tbabs Gal (atable{torus1006.fits} + apec + zpowerlaw). We fixed Γ = 1.9 and tested several values of the halfopening angle of the torus (θ OA = 40 • , 60 • , 80 • ). The three models are statistically indistinguishable, and in all cases we obtained good fits. For the three values of θ OA the column densities are consistent within the uncertainties with CT absorption.

SPHERE
To test the scenario in which the X-ray source is fully covered by the obscuring material we applied the sphere model (Brightman & Nandra 2011), using the same setting as for the torus model: tbabs Gal (atable{sphere0708.fits} + apec + zpowerlaw). This model provides a good fit (Fig. 5), and confirms the presence of heavy obscuration (N H = 9.8 +4.6 −3.5 × 10 23 cm −2 ).

Next
we applied the MYTorus model (Murphy & Yaqoob 2009), which considers absorbed and reprocessed X-ray emission from a smooth torus with θ OA = 60 • , and can be used for spectral fitting as a combination of three additive and exponential table models: the zeroth-order continuum (mytorusZ), the scattered continuum (mytorusS) and a component containing the fluorescent emission lines (mytorusL). We used the decoupled version of MYTorus  Table 3, while in Figure 6 we show the values of ∆χ 2 versus N H for the models described above. Depending on the X-ray spectral model adopted, the intrinsic (i.e. absorption and k-corrected) 2-10 keV luminosity of NGC 6286 is 3 − 20 × 10 41 erg s −1 .

DISCUSSION
The X-ray spectral analysis of NGC 6286 reported above clearly shows that the accreting SMBH is heavily obscured, possibly by CT material (see Fig. 6). The very flat continuum found by both Chandra (for the hard X-ray core) and NuSTAR, together with the fact that the 1.4 GHz emission coincides with the 3-8 keV Chandra point-source (Panel 2 of Fig. 2) confirms the presence of a heavily obscured AGN. While the buried AGN in NGC 6286 could be easily identified at hard X-rays, several other diagnostics failed to detect it because of its low-luminosity. In § 5.1 we illustrate the most commonly adopted techniques to detect AGN in U/LIRGs, and discuss the case of NGC 6286 by exploiting the wealth of multi-wavelength data available for the GOALS sample. In § 5.2 we estimate the contribution of the AGN to the luminosity of NGC 6286, while in § 5.3 we discuss the optical and radio properties of the galaxy, comparing them to those of other similar LIRGs. Finally, in § 5.4, we discuss the presence of heavily obscured low-luminosity AGN in LIRGs.

IR and X-ray tracers of AGN activity in U/LIRGs
AGN in U/LIRGs can be identified in the IR by several means: i) with the detection of high-excitation MIR emission lines (e.g., Sturm et al. 2002), and in particular of [Ne V] 14.32µm and [Ne V] 24.32µm (e.g., Weedman et al. 2005, Goulding & Alexander 2009 ii) using the ratios of high-to-low ionization fine-structure emission lines (e.g., [ Lutz et al. 1999, Petric et al. 2011; iii) with the EW of the PAH features, which tend to be lower in the presence of a  (6)  bright AGN, since it can destroy PAH molecules (e.g., Imanishi et al. 2010b); iv) studying the slope of the 2.5-5µm continuum (Γ 2.5−5 , e.g., Imanishi et al. 2010b) or the continuum 30µm/15µm flux density ratio (e.g., Stierwalt et al. 2013), which tend to be red in the presence of an AGN; v) using the depth of absorption features (e.g., Imanishi & Dudley 2000;Risaliti et al. 2006;Georgantopoulos et al. 2011b), with large depths pointing towards AGN obscured by dust; and/or vi) from deviations of the well known correlation between the far-IR (FIR) and the radio luminosity (Helou et al. 1985, Condon et al. 1991, Condon 1992, using the radio-FIR flux ratio q (e.g., Yun et al. 2001). We find that all these proxies (  (Inami, private communication). Spitzer/IRAC selection provides another important tool for identifying AGN (e.g., Lacy et al. 2004, Stern et al. 2005. Using the AGN selection criteria proposed by Donley et al. (2012) (see Eq.1 and 2 in their paper), and considering the fluxes reported by U et al. (2012), we find that NGC 6286 does not satisfy the conditions for the presence of an AGN. The fact that the IR proxies fail to identify the AGN emission in NGC 6286 is due to the problematic identification of low-luminosity AGN with IR spectra dominated by the host. For example, in a low-luminosity AGN the silicate absorption feature would be diluted by the strong IR continuum of the host galaxy. Iwasawa et al. (2011) studied 44 LIRGs from the GOALS sample with Chandra, and assessed the presence of an AGN using the hardness ratio HR ≡ (H − S)/(H + S), where H and S are the background-corrected counts in the 2-8 and 0.5-2 keV ranges, respectively. Sources with HR > −0.3 are reported as candidate AGN. This value was chosen because ULIRGs which are known to host AGNs, such as Mrk 231, Mrk 273, and UGC 5101, tend to cluster just above this limit (Iwasawa et al. 2009). Considering the spatially-integrated X-ray flux NGC 6286 has a hardness ratio HR = −0.85 ± 0.07, which would not allow to infer the presence of an AGN. However, as discussed by Iwasawa et al. (2011) this threshold could become less reliable for some CT AGN, since mostly reprocessed radiation is observed in the hard X-ray band. Another criteria commonly used to identify AGN is the observed 2-10 keV X-ray luminosity. Using log(L 2−10 / erg s −1 ) > 42 as a threshold (e.g., Szokoly et al. 2004, Kartaltepe et al. 2010, one would also miss identifying NGC 6286 as a buried AGN [log(L 2−10 / erg s −1 ) = 40.80].
Spectral decomposition (e.g., Nardini et al. 2008, Alonso-Herrero et al. 2012) is another powerful method to constrain the contribution of AGN to the multiwavelength SED. Vega et al. (2008) found that a pure starburst model fails to reproduce well the near-IR to radio SED of NGC 6286, and a buried AGN accounting for 5% of the IR luminosity is required by the data. An useful diagnostic of the presence of a heavily obscured AGN is the ratio between the MIR and the 2-10 keV luminosities (e.g., Alexander et al. 2008;Rovilos et al. 2014;Georgantopoulos et al. 2011a). It has been shown indeed that for AGN the absorptioncorrected 2-10 keV and the 6 and 12 µm luminosities are well correlated (e.g., Gandhi et al. 2009;Stern 2015;Asmus et al. 2015), so that deviations from the correlation might imply the presence of heavy obscuration. Vega et al. (2008) report that at 6µm about 58% of the flux is produced by the AGN. This would imply that the ratio between the IR and observed X-ray AGN luminosity is very low: L 2−10 /L 6µm ≃ 2.4 × 10 −3 . This value is consistent with undetected DOGs in the CDF-N (Georgakakis et al. 2010) and with other U/LIRGs (Georgantopoulos et al. 2011a), which is related to the fact that in U/LIRGs the IR emission is enhanced by strong star formation, leading to very low values of L 2−10 /L 6µm . Using the largest 2-10 keV X-ray luminosity obtained in § 4 (L 2−10 ∼ 2 × 10 42 erg s −1 ) one would still find that L 2−10 /L 6µm ∼ 0.1, a value lower than that expected from the L 2−10 − L 6µm correlation. This might imply that the AGN contribution to the IR flux is significantly lower than that reported by Vega et al. (2008) (see §5.2 and Fig. 7).

AGN contribution to the IR luminosity
The IR luminosity of NGC 6286 is 8.8 × 10 44 erg s −1 , which would imply that, depending on the X-ray spectral model used, we obtain a ratio L 2−10 /L IR ≃ 4 × 10 −4 − 2.3 × 10 −3 , significantly lower than the value expected from pure AGN (e.g., Mullaney et al. 2011). Considering the observed 2-10 keV luminosity, the ratio is log(L obs 2−10 /L IR ) ≃ −4.14, which is consistent with the average value found for the GOALS sample Comparing this to the IR luminosity of the system [log(L IR / erg s −1 ) = 44.96] we find that the IR luminosity of the AGN is between 0.1 and 0.6% of the total IR luminosity. This value is in disagreement with that obtained by Vega et al. (2008) using spectral decomposition, who found that the contribution of the AGN to the total IR luminosity is about one order of magnitude larger. A 5% contribution to the total IR luminosity would imply that log(L AGN IR / erg s −1 ) = 43.66 and the intrinsic 2-10 keV luminosity of the AGN would be log(L 2−10 / erg s −1 ) = 43.12, also an order of magnitude larger than predicted by our X-ray spectral analysis. To have such a luminosity, the AGN should be obscured by log(N H /cm −2 ) > 25, which is inconsistent with the results obtained here. An alternative explanation for this discrepancy is that the AGN is intrinsically weak at Xray wavelengths, as recently found by NuSTAR for the AGN in Mrk 231 (Teng et al. 2014, see also Teng et al. 2015).
Assuming a 2-10 keV bolometric correction of κ x = 20 (e.g., Vasudevan & Fabian 2007), the bolometric output of the AGN would be 7 − 40 × 10 42 erg s −1 . This implies that the ratio between the IR luminosity and the total output of the AGN is L Bol AGN /L IR ≃ 0.8 − 4.5%. The AGN bolometric output can also be inferred from the [Ne V] 14.32µm luminosity, following the relation obtained by Satyapal et al. (2007): and is log(L AGN Bol / erg s −1 ) ∼ 42.7, consistent with the estimate obtained using the X-ray luminosity. The 2-10 keV bolometric correction obtained using this value is κ x ≃ 3 − 17. The black hole mass of NGC 6286 has been estimated to be M BH ∼ 2.7 × 10 8 M ⊙ by Caramete & Biermann (2010) using the black hole massspheroid correlation (e.g., Magorrian et al. 1998). The Eddington ratio of the source would then be λ Edd ≃ (0.2 − 1.2) × 10 −3 , consistent with a low accretion rate AGN.
The lack of a significant AGN contribution to the total IR flux is also confirmed considering the [Ne V]/[Ne II] ratio versus the EW of the 6.2µm PAH feature [see Fig. 1 and 2 of Petric et al. (2011)], which shows that the ratio between L AGN IR and L IR is below 1% for this object. This, together with the 2-10 keV bolometric correction obtained using [Ne V] 14.32µm, clearly disfavours the intrinsically X-ray weak AGN scenario. We can therefore conclude that the energetics of NGC 6286 are clearly dominated by the host galaxy, with the low-luminosity AGN providing only a minor contribution to the total flux. The contribution of the AGN to the IR flux of the system is shown in Fig. 7. 5.3. Optical and radio emission NGC 6286 has been classified as a lowionization nuclear emission-line region (LINER) by Veilleux & Osterbrock (1987) using a classification scheme based on the diagram first proposed by Baldwin, Phillips & Terlevich (1981). While most LINERs appear to be driven by old stellar populations (e.g., Sarzi et al. 2010) and by shocks in ULIRGs (e.g., Soto & Martin 2010, 2012, in some cases they can be associated to low-luminosity AGN (e.g., Ho 2008). Yuan et al.   Kewley et al. (2006) to separate starburst galaxies, starburst/AGN composite galaxies, Seyfert 2s, and LINERs. In the scheme of Kewley et al. (2006) objects that were classified as LINERs according to Veilleux & Osterbrock (1987) would be either true LINERs, Seyfert 2s, composite HII-AGN galaxies, or high metallicity star-forming galaxies. Yuan et al. (2010) found that true LINERs are rare in IR-selected samples (< 5%), and most of the objects would be either classified as star-forming galaxies or starburst/AGN composites. Yuan et al. (2010) classified NGC 6286 as a composite using [NII], a HII region using [SII] and a LINER using [OI]. Therefore they adopted a composite classification for the source, which might imply the presence of an AGN. Yuan et al. (2010) found that in the IR luminosity bin L IR = 10 11 − 10 12 L ⊙ about 37% of the objects in the IRAS Bright Galaxy Sample (BGS, Sanders et al. 1995, Veilleux et al. 1995 are classified as composites. To characterise the relative AGN contribution to the extreme ultraviolet (EUV) radiation field, Yuan et al. (2010) use D AGN , which is the normalised distance from the outer boundary of the star-forming sequence. While this quantity does not provide information on the fraction of emission due to the AGN, it can be used to compare the amount of EUV radiation due to the AGN in different objects. For NGC 6286 they found D AGN = 0.5 using both the [OI]/H α and the [NII]/H α diagram. Yuan et al. (2010) found a statistically significant increase of D AGN with L IR , consistent with the idea that the fraction of AGN increases for increasing values of the 8-100µm luminosity (e.g., Veilleux et al. 1995). The value obtained for NGC 6286 is marginally larger than the average value obtained by Yuan et al. (2010) for the BGS sample for 11 < log(L IR /L ⊙ ) < 12 (D AGN ≃ 0.35).
The dense molecular gas tracer HCN has been found to be enhanced (relative to HCO+ and CO) in systems with dominant AGN (e.g., Imanishi et al. 2007). Privon et al. (2015) have shown that some pure starburst and composite sources show similarly enhanced HCN emission. The origin of this enhancement is uncertain, but might be due to mid-infrared pumping associated with a compact obscured nucleus (CON; e.g., Aalto et al. 2015). However, the HCN/HCO+ ratio of NGC 6286 is consistent with that of normal starbursts, rather than CONs. From this we can conclude that the starburst does not appear to be compact.
A radio core is rather common in low-luminosity AGN, as shown by the work of Nagar et al. (2005), who found evidence of radio emission in ≥ 50% of the low-luminosity AGN of the Palomar Spectroscopic sample (see also Ho 2008). The flux of NGC 6286 at 1.4 GHz is f 1.4 GHz = 157.4 ± 5.6 mJy (Condon et al. 1998), which implies that the radio loudness is log R X = log(f 1.4 GHz /f 2−10 ) = −2.6 to −3.1, depending on the X-ray spectral model assumed. These values were obtained taking into account only the nuclear emission in the computation of the 2-10 keV flux. Considering the threshold suggested by La Franca et al. (2010) (see also Panessa et al. 2007, Terashima & Wilson 2003, log R X = −4.3, NGC 6286 would be classified as radio-loud AGN. Murphy (2013) report that the radio spectral index 19 of NGC 6286 is α low = −0.73 ± 0.03, α mid = −0.89 ± 0.03 and α high = −1.02 ± 0.12 for ν < 5 GHz, 1 < ν/GHz < 10 and ν > 10 GHz, respectively. This would point towards a significant contribution of synchrotron emission, possibly from a jet. The two radio sources detected by EVN and coincident with the 3-8 keV core could be in fact associated to a jet and counter jet, consistent with the radio-loud classification of NGC 6286.

Heavily obscured low-luminosity AGN in U/LIRGs
As discussed above for the case of NGC 6286, the identification of heavily obscured AGN in LIRGs can be rather difficult if the AGN has a low-luminosity. The EW of PAH features would not be significantly affected by the AGN if it is highly obscured, since the gas and dust would shield the PAH-emitting molecules, or if it is not very luminous. A low-luminosity AGN would also be difficult to find by studying the 2.5-5µm slope, since the IR emission would be dominated by the starburst, and the AGN emission can still be self-absorbed. Absorption features also would not be able to help if the AGN is not very luminous. A more reliable tracer is [Ne V], but while its detection might indicate the presence of an AGN, its non-detection does not exclude it. Moreover, [Ne V] could be created in young starbursts, and for low-luminosity AGN it could be too faint to be detected (see Eq. 2). Radio studies can also give important insights, but since not all AGN are very strong at these wavelengths, results are not always conclusive. Hard Xray studies are possibly the best way to unveil obscured AGN in U/LIRGs, although they can also be limited by absorption for log(N H /cm −2 ) ≫ 24. By using multi-wavelength indicators of AGN for a subsample of 53 U/LIRGs within the GOALS sample, U et al. (2012) found that ∼ 60% and ∼ 25% of ULIRGs and LIRGs host AGN. Studying the whole GOALS sample, Petric et al. (2011) found that 18% of the LIRGs show evidence of [Ne V] 14.32µm, and hence might host an AGN. By means of optical spectroscopy, Yuan et al. (2010) found that 59% of the 51 single nuclei galaxies with 11 < log(L IR /L ⊙ ) < 12 in the BGS sample host an AGN 20 . The fraction of AGN is larger (77%) if one considers only two of the three diagrams for the spectral classification. A significant fraction of the composite systems might hide buried low-luminosity AGN, as in the case of NGC 6286, although an important contribution to the line emission in these objects might be due to shocks (e.g., . Treister et al. (2010b) have shown, by stacking Chandra spectra of LIRGs in the Chandra Deep Field-South, that 15% of the objects with L IR > 10 11 L ⊙ contain heavily obscured AGN. By stacking X-ray spectra in different bins of stellar mass, they found a significant excess at E = 6 − 7 keV in the stacked spectrum of sources with mass M > 10 11 M ⊙ , very likely related to a prominent Fe Kα line, while no clear evidence of AGN activity was found in less-massive galaxies. Treister et al. (2010b) concluded that there might be a large population of heavily obscured AGN in high mass galaxies. NGC 6286, with a stellar mass of 1.26×10 11 M ⊙ (Howell et al. 2010), fits extremely well into this scenario in the local Universe.
We have shown in § 5.1 and 5.3 that NGC 6286 has optical and IR characteristics quite typical of LIRGs, and consistent with other galaxies of the GOALS sample for the same merger stage. It is interesting to notice that also the hardness ratio and the observed 2-10 keV luminosity inferred by Chandra are consistent with a large fraction of the objects of the sample of Iwasawa et al. (2011) (see Fig. 5 and 6 of their paper, respectively), which might indicate that several more heavily obscured low-luminosity AGN are present in LIRGs of the GOALS sample. Moreover, we have shown that in the low-count regime it is possible to miss obscured AGN by adopting a simple phenomenological model to reproduce their X-ray spectra. Therefore there might be a significant population of lowluminosity heavily obscured AGN in LIRGs that we are missing due to the lack of sensitive hard X-ray observations. Numerical simulations have shown that accretion onto SMBHs might be happen at some level even after the first encounter (e.g., Di , although the expected accretion rate varies depending on the galaxy mergers code adopted (e.g., Gabor et al. 2015). Our on-going campaign of NuSTAR observations of ten LIRGs will allow us to study the AGN fraction in merging galaxies in the hard X-ray band across the whole merger sequence.
Another object showing similar characteristics to NGC 6286 is IC 883, a LIRG in a late merger stage that was found to host a low-luminosity AGN from radio observations (Romero-Cañizales et al. 2012b;Romero-Cañizales et al., in prep.). As for NGC 6286, the IR emission of IC 883 is dominated by star-formation and the AGN does not contribute significantly to the ener-getics of the system. Interestingly, similar to NGC 6268, IC 883 is also reported as a composite AGN/starburst system by Yuan et al. (2010), along with more than one third of LIRGs from the BGS sample.

SUMMARY AND CONCLUSIONS
We have reported here the first results of a NuSTAR campaign aimed at observing ten LIRGs in different merger stages, focussing on the first detection of a heavily obscured AGN in NGC 6286. The Chandra/ACIS-S 0.3-2 keV image of the source shows extended emission that covers ∼ 4.4 kpc (Fig. 2), and which might be due to collisionally ionized plasma. In the 3-8 keV band we found a compact source, with a flat 1.2-8 keV spectrum (Γ ∼ −0.2), which coincides with the radio emission detected by FIRST. The NuSTAR spectrum also shows a flat X-ray continuum (Γ ∼ 0.5). By analysing the broadband X-ray spectrum of the source, combining archival XMM-Newton, Chandra and quasi-simultaneous NuS-TAR and Swift/XRT observations, we have found that the source is consistent with being obscured by mildly CT material (N H = 1.08 +0.63 −0.38 × 10 24 cm −2 , Fig. 6). The presence of a heavily obscured AGN is confirmed by the possible detection of weak [Ne V] 14.32µm and [Ne V] 24.32µm lines (Dudik et al. 2009), by near-IR to radio spectral decomposition ) and by the optical classification of the galaxy as an AGN/starburst composite (Yuan et al. 2010).
The buried AGN has an intrinsically low luminosity (L 2−10 ∼ 3 − 20 × 10 41 erg s −1 ), a low value of the Eddington ratio [λ Edd ≃ (0.2 − 1.2) × 10 −3 ] and seems to contribute less than 1% to the energetics of the system (Fig. 7). Because of its low luminosity, previous observations carried out below 10 keV and in the infrared did not notice the presence of a buried AGN. By exploiting the rich multi-wavelength coverage of U/LIRGs in the GOALS sample we have discussed the radio to Xray characteristics of NGC 6286, showing that they are consistent with those of objects with similar luminosities and in a similar merger stage. We speculated that there might be a significant fraction of low-luminosity AGN in LIRGs that we are missing due to their low contribution to the total IR flux of the system. In particular, more than one third of LIRGs from the BGS sample are classified as composite AGN/starburst system by Yuan et al. (2010), and might hide low-luminosity highly obscured AGN.
The case of NGC 6286 clearly shows the importance of hard X-ray coverage in order to detect low-luminosity heavily obscured AGN in LIRGs. Our ongoing NuSTAR campaign of LIRGs will put better constraints on the fraction of these objects and the relation between obscured accretion and merger stage.
We thank the anonymous referee for his/her comments, that helped us to improve the quality of our manuscript, and the NuSTAR Cycle 1 TAC for the NuSTAR data on which this paper is based. CR acknowledges C.S. Chang, H. Inami, P. Gandhi and S. Satyapal for useful discussion. We thank Adam Block (Mount Lemmon SkyCenter/University of Arizona) for allowing us to publish his optical image of NGC 6286/NGC 6285. This research has made use of the NuSTAR Data Analysis Soft- Table 3 Summary of the X-ray spectral analysis for the spatially integrated X-ray spectrum of NGC 6286. Notes. A parameter left free to vary within the uncertainties of the value obtained fitting the extended emission (see § 4.2 for details). B value calculated from the ratio of n po and n scatt po . C the photon index in MyTORUS is calculated only in the range Γ = 1.4 − 2.6.