Chandra Reveals Heavy Obscuration and Circumnuclear Star Formation in Seyfert 2 Galaxy NGC 4968
Stephanie M. LaMassa, Tahir Yaqoob, N. A. Levenson, Peter Boorman, Timothy M. Heckman, Poshak Gandhi, Jane R. Rigby, C. Megan Urry, Andrew F. Ptak
DDraft version November 13, 2018
Preprint typeset using L A TEX style emulateapj v. 01/23/15
CHANDRA
REVEALS HEAVY OBSCURATION AND CIRCUMNUCLEAR STAR FORMATION IN SEYFERT 2GALAXY NGC 4968
Stephanie M. LaMassa , , Tahir Yaqoob , N. A. Levenson , Peter Boorman , Timothy M. Heckman , PoshakGandhi , Jane R. Rigby , C. Megan Urry , Andrew F. Ptak NASA Postdoctoral Program Fellow; NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA; Department of Physics,University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250, USA; Gemini Observatory, Casilla 603, La Serena,Chile; Department of Physics & Astronomy, University of Southampton, Southampton, SO17 1BJ, UK; Department of Physics &Astronomy, The Johns Hopkins University, 3400 North Chales Street, Baltimore, MD 21218, USA; Yale Center for Astronomy &Astrophysics, Physics Department, P.O. Box 208120, New Haven, CT 06520, USA;
Draft version November 13, 2018
ABSTRACTWe present the
Chandra imaging and spectral analysis of NGC 4968, a nearby ( z = 0 . ∼ ∼ (cid:12) yr − ). The soft emission at circumnuclear scales(inner ∼
400 pc) originates from hot gas, with kT ∼ ∼ α equivalent widthin this source (EW = 2.5 +2 . − . keV), which suggests the central engine is completely embedded withinCompton-thick levels of obscuration. Using physically motivated models fit to the Chandra spectrum,we derive a Compton-thick column density ( N H > . × cm − ) and an intrinsic hard (2-10 keV)X-ray luminosity of ∼ × erg s − (depending on the presumed geometry of the obscurer), whichis over two orders of magnitude larger than that observed. The large Fe K α EW suggests a sphericalcovering geometry, which could be confirmed with X-ray measurements above 10 keV. NGC 4968 issimilar to other active galaxies that exhibit extreme Fe K α EWs (i.e., > INTRODUCTION
Active galactic nuclei (AGN) are powered by accretingsupermassive black holes in galactic centers. A signifi-cant portion of this black hole growth is obscured by gasand dust (e.g., Risaliti et al. 1999; Fabian 1999; Treisteret al. 2004, 2009), which according to the AGN unifi-cation model, resides on ∼ N H ) of the obscuring material becomes optically thickto Compton scattering ( ≥ × cm − ), appearto represent a significant fraction of cosmic black holegrowth, though the percentage of this population is amatter of some debate (e.g., Treister & Urry 2005; Gilliet al. 2007; Treister et al. 2009; Ueda et al. 2014; Buchneret al. 2015; Ricci et al. 2015; Aird et al. 2015; Akylas et al.2016). Hence, to constrain population synthesis modelsand accurately determine the accretion history of cosmicblack hole growth, it is important to have a proper censusof Compton-thick AGN. These heavily obscured sourcescan be identified by signatures in their X-ray spectra,with the fluorescent Fe K α line emission at 6.4 keV be- Accounting for the contribution of electrons from Hydrogenand Helium, the mean number of electrons per Hydrogen atom is1.2 so that the Compton optical depth is N H = (1 . × σ T ) − , where σ T is the Thomson cross-section (see, e.g., Murphy & Yaqoob 2009) ing perhaps the most notable feature (Krolik & Kallman1987; Ghisellini et al. 1994; Matt et al. 1996). Detailedstudy of individual Compton-thick AGN then providesan opportunity to probe AGN physics. Furthermore, thesuppression of optical AGN emission allows us to studythe host galaxy in greater detail than is possible with un-obscured AGN since galaxy features are not swamped byAGN light. Circumnuclear star formation may even playa role in obscuring some AGN (e.g., Ballantyne 2008).NGC 4968 is a nearby ( z = 0 . IRAS µ m sample (Spinoglio& Malkan 1989), hosted in an Sa galaxy (Malkan etal. 1998) with a WISE measured 12 µ m luminosity of1.7 × erg s − (Wright et al. 2010). Our X-ray anal-ysis of this source with XMM-Newton revealed a strongFe K α line with equivalent width (EW) of 3.10 +1 . − . keV(LaMassa et al. 2011; see also Guainazzi et al. 2005).This extreme value indicates that the AGN in NGC 4968is embedded within Compton-thick material: the totalcontinuum from the transmitted and reflected compo-nents is suppressed relative to the fluorescent line pho-tons which can be created anywhere throughout the ob-scuring medium. However, using a phenomenologicalmodel of an absorbed power law plus a Gaussian compo-nent to fit the AGN continuum and Fe K α line, respec-tively, we did not find an absorbing column larger thanthat from our Galaxy. Excess emission at soft energies(0.5-2 keV) prevented the column density from being ac-curately measured with this overly simplistic model. Isthis soft X-ray emission due to the host galaxy, scatteredAGN emission, or both? Levenson et al. (2002) noted a r X i v : . [ a s t r o - ph . H E ] D ec that Sy2s with large Fe K α EW values ( > XMM-Newton data, we were unable to fit self-consistent, physically mo-tivated models (e.g., MYTorus, Murphy & Yaqoob 2009;spherical and toroidal absorption models of Brightman& Nandra 2011) to the spectra, nor could we attempt todisentangle nuclear from unresolved galaxy light. There-fore, we obtained a deep (50 ks)
Chandra observation inCycle 16 in order to physically diagnose the conditions atthe center of this galaxy. We present our results in thispaper as follows: in Section 2, we detail the reduction ofthe
Chandra data. We discuss the imaging and spectralanalysis of NGC 4968 in Sections 3 and 4, respectively.From these results, we comment on the nature of the X-ray obscuring medium, estimate the star formation rateof the host galaxy, and propose that the gas responsiblefor circumnuclear star formation increases the coveringfraction of the obscuration towards unity, such that theAGN is almost completely enshrouded within Compton-thick material (Section 5). Throughout, we use a cosmol-ogy where H =67.8 km s − Mpc − , Ω M =0.31, Ω Λ =0.69(Planck Collaboration et al. 2015). DATA ANALYSIS
NGC 4968 was observed for 50 ks on 2015 March 9 withACIS-S (ObsID: 17126; PI: LaMassa). The data werereduced with
CIAO v4.8 , with
CALDB v. 4.7.1 . Weused
CIAO task chandra repro to produce a filteredlevel 2 events file with the latest calibration files.From this events file, we extracted images of NGC 4968in the soft (0.5-2 keV) and hard (2-10 keV) bands, asshown in Figure 1 (top), where the location of the nar-row line region (Schmitt et al. 2003), and slightly elon-gated radio emission (Schmitt et al. 2001) are noted bythe solid blue and dashed magenta lines, respectively. As can be seen via visual inspection of the images, thesource is more spatially extended in the soft band com-pared with the hard band (see Section 3 for a full treat-ment of the spatial analysis). For reference, we showthe
Hubble Space Telescope
WFPC2 image of NGC 4968(Malkan et al. 1998) in the F606W filter in the bottompanel of Figure 1, overplotted with X-ray contours fromthe soft (red) and hard (cyan) bands. While the emissionat softer energies has a larger spatial extent, it is confinedto the bulge of the host galaxy (at least in projection).We used
CIAO routine specextract to extract thespectrum from the full X-ray emitting region as well asthe nuclear and extended components separately. Forthe global modelling, the spectrum was extracted froma circular aperture with radius 6 (cid:48)(cid:48) to ensure all extendedemission was captured. To isolate the nuclear emission,we ran wavdetect on the hard energy band, extract-ing the spectrum from the output region file centered onthe AGN (an ellipse with major axis 1.86 (cid:48)(cid:48) , minor axis1.6 (cid:48)(cid:48) and angle 137.4 ◦ , measured counter-clockwise fromWest to East in WCS coordinates). This nuclear regionwas excised from the global region to produce a spec-trum of the extended-only emission (as noted in Section The lengths of the lines are larger than the size of the emittingregions to ease visualization.
Figure 1.
Top : Chandra images of NGC 4968 in the soft (0.5-2keV, left ) and hard (2-10 keV, right ) bands. The blue line denotesthe position angle of the narrow line region, as traced by the [OIII]emission (Schmitt et al. 2003), and the dashed magenta line indi-cates the position angle of the radio emission (Schmitt et al. 2001);both emitting regions are smaller than indicated by the lines, ex-tending 2.3 (cid:48)(cid:48) and 2.2 (cid:48)(cid:48) , respectively.
Bottom: Hubble
WFPC2 im-age of NGC 4968 (Malkan et al. 1998) in the F606W filter with theX-ray contours in the soft (red) and hard (cyan) bands overplotted;the contours correspond to count levels of 2.8, 16.1, and 90.9 (softband) and 1.8, 10.6, and 59.9 (hard band). The X-ray emission isspatially extended in the soft band (see Figure 2), though confinedto the bulge of the galaxy (in projection). (cid:48)(cid:48) and outer radius of 30 (cid:48)(cid:48) . Thespectra were grouped by 5 counts per bin; visual inspec-tion of the spectra demonstrates that the chosen binningdoes not over sample the resolution and thus we lose noinformation with this binning. CHANDRA IMAGING ANALYSIS
From Figure 1, the X-ray emission appears more ex-tended in the soft band than in the hard band, and liesin a direction not coincident with the narrow line region(Schmitt et al. 2003) nor radio structure (Schmitt et al.2001). To test whether the X-ray emission is extendedand to quantify the size of this region, we simulated apoint spread function (PSF) at the position where thesource was detected. Since the
Chandra
PSF varies as afunction of energy as well as position, we fit the
Chan-dra spectrum with an ad hoc model that reproduces theshape of the observed spectrum so that the energy de-pendence of the source can be accounted for by
ChaRT ,the
Chandra ray tracing software (see Section 4 for afull treatment of the spectral modeling). The outputray-traced data from the simulated PSF are then inputinto
MARX which projects the photons onto the ACIS-Sdetector-plane to create a pseudo-events file. We simu-lated pseudo-events files for a point source separately forthe soft and hard band.
CIAO tool srcextent compares the PSF with the ac-tual data to measure the size of the PSF at the source lo-cation on the detector and to quanity whether the sourceis extended at the 90% confidence level. From this rou-tine, we find that in both the soft and hard bands, thesource is extended at the 90% confidence level. The ob-served size of the PSF at the location of the source is0 . (cid:48)(cid:48)
69, where the PSF is measured as (cid:112) (1 / a + b )with a and b as the semi-major and semi-minor axes ofan elliptical Gaussian.To measure the extent of the X-ray emission, we com-pared the surface brightness of the source with whatwould be expected from a point source: if the surfacebrightness of the source exceeds that of a point sourceover a range of distances, then we can confirm that thesource is extended and measure this spatial extent. Weperform this exercise separately for the soft and hardband. In each band, we extracted radial surface bright-ness profiles from the soft and hard bands by calculatingthe background-subtracted net counts per pixel squaredwithin a series of annuli around the nucleus. The radiusof the inner annulus was set to 1 (cid:48)(cid:48) and the subsequentannuli were incremented in steps of 0 . (cid:48)(cid:48) ∼ (cid:48)(cid:48) while thehard X-ray emission is more compact, yet extends upto ∼ . (cid:48)(cid:48) . These sizes correspond to physical scales of900 pc and 500 pc, respectively. As discussed in Section5.3, the size scale, and direction of the extended hardemission is consistent with the narrow line region mea-sured via the [OIII] ionization cone (Schmitt et al. 2003),while the soft emission appears to be independent of thenarrow line region. SPECTRAL ANALYSIS
The
Chandra spectrum was fitted with
XSpec version12.9.0 (Arnaud 1996). We used the Cash-statistic for fit-ting, which is more appropriate in the low count regimethan χ since it only requires that the count distribu-tion be Poissonian (Cash 1979). The energy range wasrestricted to be between 0.5 and 8 keV. Errors on the fitparameters represent the 90% confidence interval for oneinteresting parameter. Global Fit
Phenomenological Fit
We begin by fitting the
Chandra spectrum with a phe-nomenological model to measure the equivalent width ofthe Fe K α feature as this provides insight into the geome-try of the obscurer. A simple absorbed power law fails tofit the spectrum, requiring a more complex model. Here,we adopt a double absorbed power law model which de-scribes a partial covering geometry where a portion ofthe intrinsic AGN continuum is scattered into our lineof sight, or leaks through the circumnuclear obscuration,while the rest is absorbed. To this model, we add a Figure 2.
Radial surface brightness profiles of NGC 4968 (blackdiamonds) in the soft ( top ) and hard ( bottom ) bands compared withthe radial surface brightness profiles for a simulated point source(red asterisks) at the source position. The source is significantlyextended to ∼ (cid:48)(cid:48) (900 pc) in soft band and up to ∼ (cid:48)(cid:48) (500 pc)in the hard band. −4 −3 no r m a li z ed c oun t s s − k e V − ( da t a − m ode l ) / e rr o r Energy (keV)
Figure 3.
Chandra spectrum fitted with a phenomenologicalmodel of a double absorbed power law with a Gaussian componentto accommodate the Fe K α line and an apec component to bestmodel emission at soft energies. From measuring the Fe K α EWagainst the total continuum, we find a value of 2.9 keV.
Gaussian component, at the redshift of the source, toaccomodate the Fe K α emission and a thermal emissionmodel ( apec ) to best fit the soft emission (see below).One absorption component ( N H ) is frozen to the Galac-tic value of 9 × cm − and an additional component isleft free to attenuate the the transmitted AGN emission.This fit to the Chandra spectrum is shown in Figure 3,where we find N H =1.8 +5 . − . × cm − and a power lawslope (Γ) of 1.69 +0 . − . .In Figure 4 (top left), we show a close-up of the Fe Kregion and note that this model can not completely fitthe line, despite the line being spectrally unresolved (i.e.,the best-fit σ pegged at our imposed lower limit of 0.01keV, which is several times below the approximate Chan-dra resolution of ∼ α EW against the total continum,finding a value of 2.9 keV in the rest-frame.We also perform a local fit (3.5 keV - 7.8 keV) fora more detailed investigation of the Fe K α EW. Here,the spectrum is modeled with a power law and Gaus-sian component. Taking into account errors on the localpower law slope and normalization, and width and nor-malization of the Gaussian model (the line center waswell measured at 6.38 keV and thus was frozen), we cal-culated ∆ χ as a function of Fe K α EW in Figure 5; thehorizontal line denotes the 90% confidence level (∆ χ =7.779 for four free parameters). From this exercise, wefind the (rest-frame) Fe K α EW = 2.5 +2 . − . keV.This large EW is quite telling. To obtain such largevalues, the transmitted emission must be attenuatedmore than the fluorescent line emission produced withinthe medium, which is possible when the column densityreaches Compton-thick levels of obscuration. The sameeffect could be induced by variability, where the AGNcontinuum weakens and the travel time to the Fe K α emitting region causes a delay in the response to theEW. As we discuss further in Section 5.1, archival X-ray observations of this source indicate the flux and EWhave remained constant over time (20 and 15 years, re-spectively), such that this scenario seems unlikely. Themeasured EW thus implies a higher N H than what ismeasured, indicating that another continuum componentmust dominate at energies below the Fe K α line. Addi-tionally, as discussed in Murphy & Yaqoob (2009), theFe K α EW is sensitive to the inclination angle of theobscuring medium when the line-of-sight intersects thetorus, with the most extreme EW values (i.e., > ◦ for the geometry assumed in the MYTorus model, i.e., atorus with a fixed 60 ◦ opening angle. Toroidal Obscuration
To obtain a physically motivated fit to the NGC 4968spectrum, we use the
MYTorus model which self-consistently treats the attenuated transmitted contin-uum (modeled as a powerlaw), reflection spectrum, andFe K α and K β fluorescence emission (Murphy & Yaqoob2009). To preserve the self-consistency of the model, thenormalization and Γ of the powerlaw, the inclination an-gle of the torus, and the column density ( N H ) are tiedtogether among the MYTorus model components. Mo-tivated by our phenomenological modeling of the spec-trum and resulting Fe K α EW measurement, we freezethe inclination angle to 90 ◦ .The data require an additional powerlaw componentto properly model the soft emission (i.e., ∆C-Stat=703for a change of 1 in degrees of freedom (DOF) betweenthe model lacking this component and the one where itis included). We attribute this component to be due toa portion of the intrinsic AGN continuum that is scat-tered into or directly enters our line of sight after leakingthrough the torus: the powerlaw normalization and Γare tied to the MYTorus components, with a constantfactor left free to account for the scattered/leaked frac- tion of AGN light. In Figure 6 (top left), we show thisfit to the
Chandra spectrum, with the model parameterssummarized in the first column of Table 1. Residuals atsofter energies are apparent.We therefore added an apec component to this modelwhich accounts for thermal emission, putatively associ-ated with star formation; the abundances are kept frozenat solar. The lower energy spectrum is much better mod-eled (∆C-Stat = 60.3 for ∆DOF = 2, corresponding to
P < .
005 that the statistical improvement to the fit isdue to chance). However, the Fe K α complex is worse fitwhen compared with the phenomenological model, withlarger fit residuals around the emission feature (Figure4). We note that in this modeling, we used MYTorustables which have a power law termination energy of 200keV. However when we fit the spectra with terminationenergies of 100 and 400 keV, which span the range of thehighest and lowest values available with MYTorus , weobtain essentially identical results.
Spherical absorption
To produce such an extreme Fe K α EW, the centralengine may be completely enshrouded, because differ-ential extinction of continuum and fluorescent line pho-tons is greater for a closed geometry such as a uniformspherical distribution, compared to an open geometrysuch as a torus (e.g., compare left panel of Figure 2from Brightman & Nandra 2011 with Figure 8 from Mur-phy & Yaqoob 2009). We therefore fit the spectra withthe spherical absorption model of Brightman & Nandra(2011).Similar to the MYTorus model, the spherical absorp-tion model treats the transmitted, reflected and fluores-cent line emission self-consistently. In this model, thecentral continuum is fully covered over 4 π steradian,though we do include a scattered powerlaw component toaccount for light leaking through this obscuration, sim-ilar to the toroidal model; the percentage of escapinglight is small enough to not significantly affect the self-consistency of the model. Again, we include an apec component to best fit the soft emission (∆C-Stat = 48.2for ∆DOF = 2, corresponding to P < . Summary of the Chandra Spectral Modeling to theGlobal Emission
The spectral modeling points to the AGN in NGC 4968being heavily obscured to Compton-thick. The impliedcolumn density of the obscurer is inconsistent betweenthe
MYTorus and spherical absorption models, rangingfrom N H = 1 . +0 . − . × cm − (marginally Compton-thick) to 7 . +8 . − . × cm − (extending to extremeCompton-thick levels). Furthermore, circumnuclear starformation is on-going in the host galaxy, evidenced bythe thermal nature of the soft X-ray spectrum.We note that both model fits return a shallow spec- Figure 4.
Close-up of the Fe K α complex for different model fits, with best-fit model in upper panels and fit residuals (i.e., the ratio ofdata to the model) in bottom panels. The phenomenological model ( top left ) fits the extreme Fe K α emission feature, though it is unableto accommodate all the data points; the line is unresolved (i.e., the best fit width of the line is pegged at the lower limit which is set toseveral times below the instrumental resolution). The physically motivated MYTorus model ( top right ) does a poorer job of fitting theemission line than the phenomenological model and the Brightman & Nandra (2011) spherical absorption model ( bottom ). The scenariowhere the AGN continuum is nearly completely cocooned within Compton-thick material appears the most likely description of the data. Table 1
Spectral Fit to Global EmissionParameter
MYTorus MYTorus + apec
Sphere + apec
Γ 1.57 +0 . − . < < N H (10 cm − ) 1.36 +0 . − . +0 . − . +8 . − . Power law normalization (10 − ) +0 . − . +0 . − . +0 . − . f scatt (10 − ) +0 . − . +0 . − . +0 . − . kT (keV) · · · +0 . − . +0 . − . apec normalization (10 − ) · · · +1 . − . +1 . − . C-Stat (DOF) 217.2 (106) 156.9 (104) 128.5 (104) Power law normalization is units of photons keV − cm − s − at 1 keV. Fraction of intrinsic AGN light that leaks through the circumnuclear obscuration or isscattered into the observer line of sight. The apec normalization is given as − π [ D A (1+ z )] (cid:82) n e n h dV , where D A is the angulardiameter distance to the source in cm and n e and n H are the electron and Hydrogendensities, respectively, in cm − . EW(FeKα)/keV ∆ χ EW(FeKα) = 2.47 +2.55−1.02 keV90%confidenceEW(FeKα)
Figure 5.
Fe K α EW as a function of ∆ χ , calculated over a fourdimensional grid of power law slope and normalization and Gaus-sian width and normalization. The horizontal black line indicatesthe 90% confidence level for four free parameters (∆ χ = 7.779).The large EW indicates that the inclination angle of the obscureris completely edge-on, and that the opening angle of the torus mayapproach 0 ◦ , such that the AGN is completely enshrouded. Table 2
X-ray Luminosities Luminosity MYTorus Sphere L . − (observed) 40.11 +0 . − . +0 . − . L − (observed) 40.91 +0 . − . +0 . − . L thermal +0 . − . +0 . − . L − , int +0 . − . +0 . − . Luminosities are in log space and units of ergs − and are rest-frame values. The quoted errorson the luminosities refer to the statistical error ofthe fit. tral slope, inconsistent with the canonical Γ=1.8-1.9 forAGN (e.g., Kim et al. 2004; Tozzi et al. 2006). However,if we freeze the photon index to 1.9, we obtain statisti-cally worse fits (∆C-Stat = 53.5, 40.7 for ∆DOF = 1, forthe MYTorus and spherical absorption models, respec-tively), with larger residuals around the Fe K complexthan when the photon index is a free parameter. Fur-thermore, a hard spectrum facilitates the production ofextreme Fe K α EW values: with more photons in thecontinuum above the Fe K edge, more Fe K α photonscan be produced (see Murphy & Yaqoob 2009).From these fits, we calculated the total observed softand hard X-ray luminosities ( L . − and L − ,respectively), the apec contribution to the soft X-rayluminosity ( L thermal ) from which we estimate the starformation rate in Section 5.4, and the estimated instrinsic2-10 keV X-ray luminosity ( L − , int ) from the powerlaw fit parameters. These luminosities, and associatederrors (which reflect the statistical errors on the powerlawand/or apec normalizations) are listed in Table 2. Bothmodels return similar observed X-ray luminosities andpredict intrinsic 2-10 keV luminosities that are over twoorders of magnitude larger than that observed. Clues From the Mid-Infrared
The scenario where the source is completely obscuredgeometrically, with a small amount of leakage permitted,appears to be the most plausible for explaining the X-rayspectrum of NGC 4968: the residuals around the Fe Kcomplex are the lowest among the self-consistent models,as would be expected if complete coverage facilitates con-ditions to produce an extreme Fe K α EW. This modelalso predicts the highest column density.Further clues of extreme obscuration are evidencedby the mid-infrared (MIR) emission of this source.The 12 µ m luminosity of NGC 4968 as measured by WISE (Wright et al. 2010) is 1.7 × erg s − .Based on the correlation found between the nuclear12 µ m luminosity and intrinsic 2-10 keV X-ray lu-minosity (log( L µ m ergs − ) = (0 . ± .
04) + (0 . ± . L − , int ergs − ); Asmus et al. 2015; Gandhi et al.2009), the predicted X-ray luminosity is ∼ × ergs − , fully consistent with the value we derived via fittingthe X-ray spectrum with the spherical absorption model.Though the MIR - X-ray correlation can also be drivenby the X-ray spectral slope and covering factor of theobscurer (Yaqoob & Murphy 2011), the MIR W − W consistent with the MIR emissionbeing dominated by AGN heated dust ( W − W > . W ≤
15; Stern et al. 2012). A high quality X-rayspectrum above 10 keV from e.g.,
NuSTAR , would deter-mine whether a spherical or toroidal geometry is correct,since the predicted spectral shapes above 10 keV differbetween these models.
Nuclear Region
Here we capitalize on
Chandra ’s arcsecond resolutionto hone in on nuclear emission at scales of ∼ (cid:48)(cid:48) ( < MYTorus and summarize those results in Ta-ble 3 for completeness.For both models, where we include a scattered pow-erlaw model to accommodate leaked AGN emission, weagain find residuals at softer energies, such that addingin an apec component significantly improves the fit (thechange in C-Stat is significant at the
P < .
005 level;Table 3). Thus it appears that thermal X-ray emissionis not confined to the extended emission, but also arisesat nuclear scales. The spherical absorption plus apec fitis shown in Figure 7.
Extended Emission
The spectrum from the extended only emission, ex-tracted from a 6 (cid:48)(cid:48) ( ∼ apec plus powerlaw model (see Table 4), where bothcomponents are required to fit the spectrum. The pow-erlaw emission can arise from unresolved X-ray binaries We use the
WISE magnitudes measured with profile-fittedphotometry.
Figure 6.
Chandra spectrum (black) of the global X-ray emission fitted with
MYTorus only ( top left ) and with
MYTorus plus apec ( top right ). In the bottom panel , the spherical absorption model of Brightman & Nandra (2011) is used, again with an apec componentadded to best accommodate emission at softer energies. All models include a component where a fraction of the AGN continuum emissionis scattered into the observer line of sight. The apec component best fits the features at lower energies, indicating that a portion of thesoft emission is thermal, putatively linked to circumnuclear star formation. Table 3
Spectral Fit to Nuclear EmissionParameter
MYTorus MYTorus + apec
Sphere Sphere + apec Γ < < +0 . − . < N H (10 cm − ) 1.36 +0 . − . +0 . − . +1 . − . +13 . − . Power law normalization (10 − ) 1.47 +0 . − . +0 . − . +6 . − . +2 . − . f scatt (10 − ) 1.4 +0 . − . +0 . − . +1 . − . +1 . − . kT (keV) · · · +0 . − . · · · +0 . − . apec normalization (10 − ) · · · +1 . − . · · · +1 . − . C-Stat (DOF) 173.3 (101) 141.0 (99) 146.9 (101) 112.7 (99) in the host galaxy and/or AGN continuum photons thatscatter into our line of sight.In this extended zone, the temperature of the gas issignificantly lower than that observed on nuclear scaleswith kT = 0.33 +0 . − . keV versus 0.76 +0 . − . keV. In fact,from the global fit to the X-ray emission, we find agas temperature (kT ∼ +1 . − . × erg s − and 1.77 +1 . − . × erg s − , respectively), suggestingthat circumnuclear star formation dominates the totalthermal emission in this galaxy. Table 4
Spectral Fit to Extended EmissionParameter apec + powerlaw kT (keV) 0.33 +0 . − . apec normalization (10 − ) 4.40 +2 . − . Γ 0.69 +0 . − . Power law normalization (10 − ) 9.39 +9 . − . C-Stat (DOF) 19.7 (14) DISCUSSION
Figure 7.
Chandra spectrum of nuclear emission, fit with theBrightman & Nandra (2011) spherical absorption model with ascattered AGN component (parameterized by a powerlaw model)and apec component. Thermal emission appears to be present onnuclear, as well as global, scales (see Figure 8).
Figure 8.
Chandra spectrum of extended emission, fit with an apec plus powerlaw model. The spectrum is soft, consistent withbeing dominated by thermal emission likely arising from star for-mation.
X-ray History of NGC 4968
NGC 4968 was detected by the
HEAO
A1 survey (Pol-letta et al. 1996), though was undetected by
GINGA andthe
ROSAT
All-Sky Survey (Bianchi et al. 2005). Turneret al. (1997) found NGC 4968 to be about five times dim-mer in an
ASCA observation compared with the previous
HEAO
A1 detection. It was thus targeted as a potentialX-ray “changing-look” AGN with
XMM-Newton , whereit was thought that perhaps a change in column den-sity drove the flux variation. However, as Bianchi et al.(2005) report, two epochs of
XMM-Newton observationsfrom 2001 and 2004 revealed no variability. Indeed, theobserved 2-10 keV flux with
XMM-Newton ( ∼ × − erg cm − s − ) is consistent with the reported ASCA flux ( ∼ × − erg cm − s − ) from a 1994 observation.Bianchi et al. (2005) concluded that the HEAO
A1 flux,extrapolated from the observed count rate between 1-20keV due to loss of spectral resolution resulting from ahardware failure, was likely erroneous.From our
Chandra spectrum, we find a 2-10 keV flux(3.5 +3 . − . × − erg cm − s − ) consistent with our pre- viously reported XMM-Newton flux (2.1 ± . × − ergcm − s − ; LaMassa et al. 2011), as well as similar Fe K α EWs ( 2.5 +2 . − . keV and 3.1 +1 . − . ). Fukazawa et al. (2011)report an Fe K α EW of 1.8 ± Suzaku spectrum obtained in 2006, which is consistentwith our past
XMM-Newton and current
Chandra mea-surements. Hence it appears that the X-ray flux andspectrum of NGC 4968 has remained relatively constantthroughout at least two decades of observations, indicat-ing that the extreme Fe K α EW is due to obscurationrather than time lags between the continuum and the FeK α emitting region. Comparison of Torus Properties with Past X-rayand Infrared Studies
Brightman & Nandra (2011) fitted the
XMM-Newton spectrum of NGC 4968 with their spherical absorptionmodel, obtaining a column density of log ( N H [cm − ]) =24.5 +0 . − . dex. We obtain similar column densities whenfitting the Chandra spectrum from both the global andnuclear emission regions with the same model. However,we demonstrate that a scattered powerlaw is required inorder to adequately model the data.Lira et al. (2013) fitted the MIR spectrum with aclumpy torus model (Nenkova et al. 2008a,b), findingthe viewing angle of the torus to be between 75 ◦ -90 ◦ atthe 67% confidence level. Consistent with our resultsfrom the X-ray modeling, the line-of-sight intersects theobscurer edge-on. They determine a dust line-of-sightextinction of ∼
100 mag, which with their assumed gas-to-dust ratio of N H /A V = 1 . × cm − mag − , cor-responds to N H ∼ × cm − . The finding of a lowercolumn density from MIR observations is not surprising.First, this technique probes only dust, and dust-free gaswithin the sublimation radius may dominate the X-raymeasured column density. Secondly, the MIR emissionis due to the ensemble emission of all the central cloudsand is not sensitive to the conditions along this specificline of sight.Relatedly, it is entirely consistent that the 9.7 µ m sili-cate strength is relatively weak, despite the large columndensity. Wu et al. (2009) measured S sil = -0.21 from the Spitzer spectrum, where S sil < -0.5 in the cases of strongabsorption. Silicate strength does not directly measureline-of-sight optical depth but is instead sensitive to thegeometric and temperature distribution of the emittingand absorbing material (Levenson et al. 2007). Indeed,only about 45% of the Compton-thick AGN studied byGoulding et al. (2012) showed strong silicate absorption,which they attribute to dust in the host galxay, not im-mediately associated with AGN. NGC 4968 provides fur-ther evidence that MIR-only diagnostics do not alwaysreveal the extreme nature of Compton-thick AGN. Extended X-ray Emission is Not Coincident withFeatures Discovered at Other Wavelengths NGC 4968 was not detected at energies above 10 keV with
Suzaku , precluding us from using higher energy spectra to constrainour spectral fitting. S sil = ln( f obs (9.7 µ m)/ f cont (9.7 µ m)), where f obs is the fluxdensity of the silicate feature and f cont is the flux density of thelocal mid-infrared continuum (Spoon et al. 2007). NGC 4968 was observed with the
Hubble Space Tele-scope
WFPC2 camera to image the [OIII] emission whichtraces the narrow line region (Schmitt et al. 2003). The[OIII] emission, and by proxy the narrow line region, wasfound to extend 2.2 (cid:48)(cid:48) along position angle 40 ◦ , with an ex-tent of 1.3 (cid:48)(cid:48) in the perpendicular direction. Additionally,high-resolution Very Large Array (VLA) observations at8.46 GHz revealed a radio structure at a position angleof 85 (cid:48)(cid:48) , extending 2.3 (cid:48)(cid:48) . The locations of these regions arenoted in the top panel of Figure 1.While the narrow line region seems coincident with theslightly extended X-ray emission in the hard and softbands along the northeast-southwest direction, the ex-panse of soft X-ray emission East of the nucleus is not co-incident with the narrow line region nor the radio struc-ture. Additionally, this soft X-ray emission is about twicethe size of both the narrow line region and radio emis-sion ( ∼ (cid:48)(cid:48) compared with 2.2-2.3 (cid:48)(cid:48) ). The size and locationof this expanse, plus its apparent thermal nature, lendscredence to the interpretation that it is associated withstar formation rather than AGN activity that permeatesbeyond nuclear scales. The X-ray View of Star Formation
We use the thermal emission from the apec componentas a starting point to estimate the star formation rate.As pointed out in LaMassa et al. (2012), using solely the apec luminosity neglects the contribution from X-ray bi-naries which emit as powerlaws. However, the unresolvedX-ray binary population can not be easily disentangledfrom the AGN powerlaw emission. In LaMassa et al.(2012), we presented a template to adjust L thermal torecover X-ray binary emission and more accurately es-timate the star formation rate (SFR). Using a sampleof 22 star forming galaxies from the XSINGS sample (a Chandra survey of the SINGS sample, Kennicutt et al.2003), we calculated the mean ratio of the X-ray emis-sion from non-nuclear point sources to the total 0.5-2 keVemission ( R ). The 0.5-2 keV luminosity attributable tostar-formation ( L . − , SF ) is then: L . − , SF = L thermal − R , (1)where R = 0 . ± . L thermal = 6.0 × erg/s (Table 2), we estimate that the X-ray luminosityassociated with star-formation is 1.2 ± . × erg s − (taking into account errors on the fitted apec luminosityand R value). Using the X-ray luminosity to SFR conver-sion presented in Pereira-Santaella et al. (2011), wherethey combine ultraviolet and infrared data to accountfor both the unobscured and obscured star formation ina local sample of luminous infrared galaxies: SF R (M (cid:12) yr − ) = 3 . × − L . − , SF (erg s − ) , (2)we find a SFR of 4.1 ± (cid:12) yr − . We note that whenusing the Ranalli et al. (2003) L . − /SFR relation,which is calibrated on infrared data only and thus ne-glects unobscured star formation and also assumes a dif-ferent initial mass function than Pereira-Santaella et al.(2011), we find a SFR of 2.6 ± (cid:12) yr − .Additionally, we estimate the far-infrared SFR fromthe IRAS flux densities at 60 µ m and 100 µ m (2.35 Jy and 3.75 Jy, respectively; Spinoglio & Malkan 1989). Usingthe far-infrared derived SFR calibration from Kennicutt(1998): SF R (M (cid:12) yr − ) = 4 . × − L FIR (erg s − ) , (3)we obtain a SFR of 2.4 M (cid:12) yr, − which is essentiallyidentical to our X-ray derived SFR using the Ranalliet al. (2003) relation; both works assume the same ini-tial mass function. We note that the Kennicutt (1998)SFR calibration includes mid-infrared wavelengths whichwe did not include here to minimize contamination fromAGN-heated dust to our star formation estimates.The mid-infrared spectroscopy provides an indepen-dent probe of the AGN/star-formation connection inNGC 4968. In LaMassa et al. (2010), we reported thepolycyclic aromatic hydrocarbons (PAHs) EW measure-ments at 11.3 µ m and 17 µ m for the 12 µ m Sy2 sampleusing Spitzer data. PAHs are associated with star for-mation though the presence of AGN heated dust dilutesthe strength of the PAH EW: the stronger the AGN con-tribution to the MIR continuum, the weaker the PAHEW. We found the PAH EWs to be weak. We can in-fer that circumnuclear star formation is on-going by thedetection of the PAH features and presence of thermalX-ray emission, though AGN heated dust dominates themid-infrared emission, evidenced by both the MIR spec-troscopy and
WISE photometry.
Extreme Fe K α EW: An Indication of More thanJust Obscuration?
In addition to providing insight into the geometry ofthe X-ray reprocessor, an extreme Fe K α EW ( > α EWs thatthe sources with the most extreme EWs had evidence ofon-going star formation within the centers of their hostgalaxies. Recently, Boorman et al. (2016) noted that IC3639 has an Fe K α EW above 2 keV and an X-ray spec-trum that requires the apec component to fit the soft X-ray emission, with estimated X-ray and far-infrared SFRsabove 10 M (cid:12) yr − , consistent with starburst galaxies.Our analysis on NGC 4968 further supports the ideathat extreme Fe K α EWs can be an indicator that cir-cumnuclear star formation may be present. As shownby our imagining analysis, the soft emission, associatedwith star formation, extends to physical scales of ∼ (cid:12) yr − kpc − .Though this SFR is not prodigious, it seems clear weare observing X-ray emission from recent star formationprocesses within the central kpc of the galaxy. This re-sult is consistent with a symbiotic relationship betweenstarbursts and the obscuring torus: feedback from su-pernovae and stellar winds can puff up the torus (e.g.,Wada & Norman 2002; Wada et al. 2009; Schartmann etal. 2009). If the torus is sufficiently inflated such that thecovering factor approaches unity, this obscuring mediumwill preferentially attenuate the total continum with re-spect to the line photons, which has the effect of boost-ing the EW. The gas associated with star formation canperhaps also act as this obscuring material (e.g., Thomp-son et al. 2005; Ballantyne 2008) and ultimately fuel theblack hole (Hobbs et al. 2011), though the connection0between nuclear star formation and AGN obscuration orblack hole fueling depend on a number of unknown vari-ables (e.g., gas fraction, outer radius of accretion disk,Mach number of accretion disk).However, starburst/AGN composites may not alwaysbe systems that harbor extreme Fe K α EWs. For in-stance, Arp 220 is a well known starburst galaxy hostingan AGN (e.g., Ptak et al. 2003), yet it lacks a strongneutral Fe K α line (Teng et al. 2009). Though AGNin starburst galaxies may not always be completely co-cooned inside gas with a high column density, providingthe requisite conditions to produce extreme neutral FeK α EWs, AGN with these prominent features are primecandidates to search for signatures of on-going star for-mation. CONCLUSIONS
We presented the
Chandra imaging and spectral anal-ysis of the Seyfert 2 galaxy NGC 4968. Capitalizing on
Chandra ’s arcsecond resolution, we investigated the spa-tially resolved X-ray emission, finding: • The hard (2-10 keV) X-ray emission is extendedon scales of ∼
500 pc, coincident with the narrowline region as mapped by [OIII] 5007 ˚A emissiondetected by the
Hubble Space Telescope (Schmittet al. 2003). • The soft (0.5-2 keV) X-ray emission is significantlyextended to scales of about a kiloparsec, and ismore extended than, and spatially distinct from,both the narrow line region and a previously de-tected elongated radio structure (Schmitt et al.2001).The most striking feature of NGC 4968 comes fromits X-ray spectrum, which reveals a prominent Fe K α EW ( > Chandra spectrum,we measure an EW of 2.5 +2 . − . keV. This extreme valuecan only be achieved if the continuum is suppressed byCompton-thick levels of obscuration, and is more likelyto occur in a geometry where the central engine is com-pletely enshrouded. In that case, the differential extinc-tion of the total continuum (emanating from the cen-ter of the obscuring medium, and including the reflectedand scattered continua) compared with the fluorescentline emission (produced throughout the medium) is thegreatest, boosting the EW.We use the MYTorus (Murphy & Yaqoob 2009) andthe spherical absorption models of Brightman & Nan-dra (2011), which are physically motivated models thatself-consistently treat the transmitted, reflected, and flu-orescent line emission, to fit the X-ray spectra of NGC4968. From the spectral analysis, we learn: • The obscuration is at least near Compton-thick lev-els and is likely more extreme: – The MYTorus fit to the spectrum returns N H = 1.35 +0 . − . × cm − , while the spher-ical absorption model measures an N H of7 . +8 . − . × cm − . Though an ionized Fe line is detected at ∼ – The instrinsic X-ray luminosity from thespherical absorption fit ( L − , int =7 . +0 . − . × erg s − ), about 2 orders of mag-nitude higher than that which is observed,is consistent with what would be predictedbased on the MIR - X-ray correlation (Asmuset al. 2015; Gandhi et al. 2009).A future significant high energy (i.e., >
10 keV) ob-servation from e.g.,
NuSTAR should differentiatebetween a toroidal and spherical absorption model.These data would provide the most accurate con-straints on the column density and intrinsic X-rayluminosity. • A thermal model ( apec ) is needed to adequately fitthe soft X-ray spectrum, indicating the presence ofhot gas with temperature kT ∼ • We separately investigated the X-ray spectrumfrom the nuclear ( <
400 pc) and extended (400 pc- 1.3 kpc) regions: – Nuclear spectrum : Thermal X-ray emission ispresent at the smallest scales we are able toresolve, with a temperature of kT ∼ – Extended emission spectrum : Within the ex-tended region, both thermal and non-thermalemission are present and are ascribed tostar formation and unresolved X-ray binariesand/or scattered AGN photons, respectively.The temperature of the gas (kT ∼ L thermal ∼ × erg s − ) and the templatepresented in LaMassa et al. (2012), finding a SFR of ∼ (cid:12) yr − , consistent with the far-infrared SFR(based on the 60 µ m and 100 µ m fluxes) of 2.4 M (cid:12) yr − .Though this rate is not as prodigious as most starburstgalaxies, we can conclude that star formation is on-goingin the central ∼ kpc of this galaxy.It is interesting that AGN with extreme Fe K α EWsappear to reside in galaxies with circumnuclear star-formation (e.g., Levenson et al. 2002; Boorman et al.2016). This association may be a natural consequenceof nuclear star formation, where gas completely envelopsthe AGN (perhaps due to stellar feedback inflating theobscuring medium and increasing its covering factor tounity; Wada & Norman 2002; Wada et al. 2009; Schart-mann et al. 2009), providing the conditions necessary toproduce large EW values. Though a global trend be-tween star formation activity and X-ray obscuration isnot always apparent (see, e.g., LaMassa et al. 2011), gasfrom star formation may contribute to the column den-sity in these systems (Thompson et al. 2005; Ballantyne2008). We emphasize that the converse is not necessar-ily true: not all starburst/AGN composities have promi-nent Fe K α EWs. However, AGN with large Fe K α EWvalues are promising candidates to look for clues of on-going, circumnuclear star formation. Such a search maybe especially relevant at higher redshift where the star1formation rates and central gas masses are much higherthan in the present day Universe.We close by noting that though the Fe K α emissionline is the most prominent feature in the Chandra spec-trum, the line is spectrally unresolved. Follow up highresolution grating spectroscopy, or microcalorimeter ob-servations, would spectrally resolve resolve this featureand determine whether it originates within the standardbroad line region or putative torus (e.g., Yaqoob et al.2001, 2003; Shu et al. 2011).We thank the referee for a thorough reading of thismanuscript and for providing constructive comments.Support for this work was provided by the NationalAeronautics and Space Administration through ChandraAward Number GO5-16112X issued by the SmithsonianAstrophysical Observatory for and on behalf of the Na-tional Aeronautics Space Administration under contractNAS8-03060. SML is supported by an appointment tothe NASA Postdoctoral Program at the NASA GoddardSpace Flight Center, administered by Universities SpaceResearch Association under contract with NASA. NALis supported by the Gemini Observatory, which is oper-ated by the Association of Universities for Research inAstronomy, Inc., on behalf of the international Geminipartnership of Argentina, Brazil, Canada, Chile, and theUnited States of America. PB and PG thank STFC forsupport (trant reference ST/J003697/2).The scientific results reported in this article are basedto a significant degree on observations made by the
Chan-dra
X-ray Observatory. This research has made use ofsoftware provided by the
Chandra
X-ray Center (CXC)in the application packages CIAO, ChIPS, and Sherpa.Chandra ObsId 17126
Facilities: