A New Compton-thick AGN in Our Cosmic Backyard: Unveiling the Buried Nucleus in NGC 1448 with NuSTAR

NuSTAR (Harrison et al. 2013), launched in 2012 June, is the first orbiting X-ray observatory which focus light at high energies (E ≳ 10 keV). It consists of two co-aligned focal plane modules (FPMs), which are identical in design, and referred to as FPMA and FPMB. Each FPM covers the same 12' × 12' portion of the sky, and comprises four cadmium–zinc–telluride detectors placed in a 2 × 2 array. NuSTAR operates between 3 and 79 keV, and provides 100×  improvement in sensitivity as compared to previous hard X-ray orbiting observatories (E ≳ 10 keV; e.g., Swift-BAT and INTEGRAL). In addition, it has an angular resolution of 18'' full-width at half-maximum (FWHM), with a half power diameter of 58'', resulting in 10×  improvement over previous observatories operating at E ≳ 10 keV. These characteristics, particularly its high energy coverage and better sensitivity, have allowed us to perform accurate broadband X-ray spectral modelings of heavily obscured AGNs in the local universe (e.g., Baloković et al. 2014; Puccetti et al. 2014; Annuar et al. 2015; Bauer et al. 2015; Gandhi et al. 2016).

NGC 1448 was observed by NuSTAR in 2015 (2015 July 12; ObsID 60101101002) with an effective exposure time of 58.9 ks (60.3 ks on-source time) for each FPM. The source was observed as part of our program to form the most complete census of the CTAGN population and the N_H distribution of AGNs in the local universe, using a volume-limited (D < 15 Mpc) AGN sample from Goulding & Alexander (2009).²⁷

We processed the NuSTAR data of NGC 1448 with the NuSTAR Data Analysis Software (nustardas) v1.4.1 within heasoft v6.15.1 with CALDB v20150316. The nupipeline v0.4.3 script was used to produce calibrated and cleaned event files using standard filter flags. We extracted the spectra and response files using the nuproducts v0.2.5 task.

The AGN is detected in both of the NuSTAR FPMs. The combined images of the AGN from the two FPMs in the 3–8, 8–24, and 24–40 keV bands are shown in Figure 1. We extracted the NuSTAR spectrum of NGC 1448 from each FPM using a circular aperture region of 20'' radius (corresponding to ∼30% NuSTAR encircled energy fraction, ECF) centered on the Chandra position of the AGN (see Section 2.2). The aperture size was chosen to minimize contamination from off-nuclear sources observed in the Chandra data. Background photons were collected from an annulus region centered on the AGN with inner and outer radii of 40'' and 70'', respectively. The extracted spectrum from each FPM was then co-added using the addascaspec script to increase the overall S/N of the data. We detected significant counts up to ∼40 keV from this combined spectrum, and measured a net count rate of 2.79 × 10⁻³ counts s⁻¹ in the 3–40 keV band.

Zoom In Zoom Out Reset image size

We note that in the 24–40 keV band image, where we expect the AGN to completely dominate, the peak emission appears to be offset from the center of the extraction region. The offset is not observed in the 3–8 and 8–24 keV band images where NuSTAR is more sensitive. Performing our spectral fits up to only 24 keV gave consistent results with that obtained by the 0.5–40 keV spectral fits. We therefore attribute this offset due to NuSTAR statistical uncertainty due to lower S/N at this higher energy band. We also note that centering the NuSTAR extraction region according to this offset or enlarging the extraction region to account for this apparent offset do not significantly affect our final results on the analysis of the AGN in NGC 1448.

NGC 1448 was observed by Chandra in 2014 with an exposure time of 49.4 ks (50.1 ks on-source time) using the ACIS-S detector as part of the Chandra HRC-GTO program (2014 March 09; PI S. Murray; ObsID 15332). We reprocessed the data to create event files with updated calibration modifications, following standard procedures.

We determined the centroid position of the AGN in the Chandra hard energy band of 2–8 keV using the wavdetect tool within ciao with the threshold parameter set to 1 × 10⁻⁷. We detected three sources within the central 20'' radius of the galaxy in this energy band (see Figure 1). The brightest source was detected at a position of R.A. = 3:44:31.83, and decl. = −44:38:41.22, with errors of 017 and 011, respectively. This is consistent with the 2MASS and Gemini/T-ReCS (see Section 2.4) positions of the nucleus to within ∼1''. Therefore, we adopted this Chandra position as the AGN position.

We extracted the source spectrum using the specextract task in ciao from a circular region of 20'' radius centered on the detected position of the AGN to match the NuSTAR extraction region. A 50'' radius circular aperture was used to extract the background counts from an offset, source-free region. The total net count rate within the 20'' radius extraction region in the 0.5–8 keV band is 5.69 × 10⁻³ counts s⁻¹. The net count rate measured by wavdetect for the AGN is 1.39 × 10⁻³ counts s⁻¹ in the 0.5–8 keV band. The two other sources detected within the extraction region are located to the north east (NE) and north west (NW) of the AGN. They do not have counterparts at other wavelengths, and are likely to be X-ray binaries within NGC 1448 (see Section 3.2.1). These sources have 0.5–8 keV count rates of 6.67 × 10⁻⁴ counts s⁻¹ and 8.61 × 10⁻⁴ counts s⁻¹, for the NE and NW sources, respectively, as measured by wavdetect.

At optical wavelengths, we performed spectroscopy for NGC 1448 using the ESO NTT on 2015 December 07, using the Faint Object Spectrograph and Camera v.2 (EFOSC2) instrument (Program ID 096.B-0947(A); PI G. B. Lansbury). The source was observed for four 5 min exposures centered on the Chandra position (see Section 2.2). The slit adopted was 1'' (75 pc) in width, and the adopted grism yielded a spectral coverage of 4085–7520 Å, and spectral resolution of FWHM = 12.6 Å. The seeing at the time of the observation was ∼07. Standard iraf routines were followed to reduce the spectra, and spectrophotometric standard star observations from the same night were used for calibration. As shown in Figure 2, we extracted spectra from three different apertures along the slit, corresponding to: (1) the AGN (using the Chandra position as a reference); (2) the "optical peak"; and (3) the total galaxy. The optical peak corresponds to the position where the optical emission is the brightest in the imaging data and in the two-dimensional spectra, and is offset from the AGN position by ∼3'' (R.A. = 3:44:31.99 and decl. = −44:38:42.33). The spatial apertures adopted were 24 for the AGN and the optical peak, and 369 for the total galaxy (see Figure 2). The apertures for the AGN and the optical peak correspond to two spatially distinct, prominent line-emitting regions which are separated by a dust lane, from which we do not see any prominent line emission. The aperture extents for the AGN and the optical peak were chosen based on the light profile of the Hα and [O iii]λ5007 Å emission lines in the spatial direction, to ensure that we included the total emission of these lines.

Zoom In Zoom Out Reset image size

Before we can measure the fluxes of the emission lines, we first need to subtract stellar emission from the extracted spectra. This was done by fitting the emission line-free regions of each spectrum with a combination of a small subset of galaxy template spectra from the Bruzual & Charlot (2003) stellar library. This library consists of model templates for 39 stellar populations with ages between 5.0 × 10⁶ and 1.2 × 10¹⁰ years, and metallicities between 0.008 and 0.05, with a spectral sampling of 3 Å. This library has been used to fit the continua and measure the emission line fluxes of the Sloan Digital Sky Survey galaxy spectra (Tremonti et al. 2004). We performed our spectral synthesis for the three spectra in xspec v12.8.2,²⁸ assuming z = 0.00390, solar metallicity (Z_⊙ = 0.02), and the Cardelli et al. (1989) extinction law. The best-fitted stellar spectra were then subtracted from the observed spectra, which resulted in residual spectra with emission lines only. We then analyzed the detected emission lines using the splot tasks in iraf.

NGC 1448 was observed at MIR wavelengths in 2010 with high spatial resolution using the Thermal-Region Camera Spectrograph (T-ReCS; field of view 288 × 216; 0.09 arcsec pixel⁻¹; Telesco et al. 1998), mounted on the Gemini-South telescope. The observation was carried out on 2010 August 19 (Program ID GS-2010B-Q-3) for ≈319 s on-source time using the N-band filter (λ = 7.4–13.4 μm) in parallel chop and nod mode. The data were reduced using the midir pipeline in iraf provided by the Gemini Observatory, and the image analysis was performed using the idl package mirphot, following Asmus et al. (2014). The data were flux-calibrated with a standard star, HD22663, which was observed immediately before NGC 1448. The resolution of the observation is ∼04 as measured from the FWHM of the standard star.

A compact nucleus is clearly detected in the MIR continuum at position R.A. = 3:44:31.75 and decl. = −44:38:41.8 (telescope astrometric uncertainty is on the order of 1'') with a S/N ∼ 5. The source appears to be embedded in ∼07 extended emission along the host major axis (position angle PA ∼ 44°). However, owing to the small extent and often unstable point-spread function (PSF) in ground-based MIR observations, a second epoch of high spatial resolution MIR images is required in order to confirm this extension. The source is best-fitted with a two-dimensional Gaussian model (see Figure 3), which measured a flux density of f_{12 μm} = 12.5 ± 2.3 mJy.

Zoom In Zoom Out Reset image size

In this section, we describe the broadband X-ray spectral analysis of the AGN in NGC 1448, performed using xspec. Given the non-negligible contribution of the background flux to the weak source flux, particularly in the NuSTAR data at high energies, we binned both the Chandra and NuSTAR spectra to a minimum of 1 count per bin and optimized the fitting parameters using the Poisson statistic, C-statistic (Cash 1979).²⁹ We included a fixed Galactic absorption component, ${N}_{{\rm{H}}}^{\mathrm{Gal}}=9.81\times {10}^{19}\,{\mathrm{cm}}^{-2}$ (Kalberla et al. 2005), using the xspec model " phabs" in all spectral fits, and assumed solar abundances for all models. The redshift was fixed at z = 0.00390 in all analyses. We quoted all errors at 90% confidence, unless stated otherwise. We summarize the main results of the analysis in Table 1, and present the X-ray data and best-fitting models in Figure 4.

Zoom In Zoom Out Reset image size

Table 1.  X-Ray Spectral Fitting Results for NGC 1448

Component	Parameter	Units	Model T	Model M
(1)	(2)	(3)	(4)	(5)
apec	kT	keV	${0.63}_{-0.33}^{+0.14}$	${0.63}_{-0.42}^{+0.14}$
Absorber/Reflector	NH(eq)	1024 cm−2	${4.2}_{-1.7}^{+u}$	${4.9}_{-2.0}^{+u}$
	NH(los)	1024 cm−2	${4.2}_{-1.7}^{+u}$	${4.5}_{-1.8}^{+u}$
	${\theta }_{\mathrm{inc}}$	deg	87f	${78.6}_{-11.8}^{+6.5}$
	${\theta }_{\mathrm{tor}}$	deg	${45.9}_{-18.9}^{+33.1}$	60f
AGN Continuum	fscatt	%	${0.2}_{-0.2}^{+0.3}$	${0.1}_{-0.1}^{+0.2}$
	Γ	⋯	1.9f	1.9f
	L0.5−2,obs	1039 erg s−1	0.2	0.2
	L${}_{2-10,\mathrm{obs}}$	1039 erg s−1	0.9	0.9
	L10−40,obs	1039 erg s−1	11.7	11.8
	L0.5−2,int	1040 erg s−1	2.6	5.6
	L2−10,int	1040 erg s−1	3.5	7.6
	L10−40,int	1040 erg s−1	3.5	7.6
C-stat/dof	⋯	⋯	431/445	429/440

Note. (1) Model component; (2) parameter associated with each component; (3) units of each parameter; (4) best-fitting parameters for Model T (torus model by Brightman & Nandra 2011); (5) best-fitting parameters for Model M (MYTorus model by Murphy & Yaqoob 2009). "f" is a fixed parameter and "u" is an unconstrained parameter. Details of each model are described in Section 3.

Download table as:  ASCIITypeset image

We began our spectral modeling by examining the NuSTAR and Chandra data separately. At this point we are not yet considering non-AGN contributions to the X-ray spectrum, such as from the two off-nuclear point sources within the extraction region. We modeled the two spectra in the 3–40 keV and 2–8 keV bands, respectively, using a simple power-law model with Galactic absorption. The best fitting photon indices are very flat: Γ = −0.29 ± 0.38 for NuSTAR (C-stat/dof = 229/237), and Γ = ${0.82}_{-0.77}^{+0.88}$ for Chandra (C-stat/dof = 125/126). A slight excess of counts just above 6 keV was observed in both the NuSTAR and Chandra spectra. This is likely an indication of a fluorescent Fe-Kα line emission, associated with a spectral component produced by AGN emission that is being reflected off a high column of gas. The fluxes of the two spectra within the common energy range of 3–8 keV measured by the model are f₃₋₈ = 2.00_−0.57^+0.23 × 10⁻¹⁴ erg s⁻¹ cm⁻² for NuSTAR, and f${}_{3-8}={3.50}_{-1.34}^{+0.56}\times {10}^{-14}$ erg s⁻¹ cm⁻² for Chandra. These fluxes are consistent with each other within the measurement uncertainties, indicating that there was no significant variability between the two observations.

We then fitted the NuSTAR spectrum simultaneously with the Chandra spectrum between 3 and 40 keV using a simple power-law model and a Gaussian component to model the possible Fe-Kα emission line, with the line energy and width fixed to E = 6.4 keV and σ = 10 eV, respectively. This model measured a photon index of Γ = ${0.02}_{-0.28}^{+0.27}$ and Fe-Kα line equivalent width of EW = ${2.1}_{-0.8}^{+1.0}$ keV (C-stat/dof = 315/331). The flat photon index and large EW measured for the Fe-Kα line (EW ≳ 1 keV), are characteristic signatures of CT absorption.

To test for this, we model both spectra simultaneously from 3–40 keV with the pexrav model (Magdziarz & Zdziarski 1995), which has historically been used to model reflection-dominated AGN spectra. This model produces an AGN continuum which is reflected from a slab-geometry torus with an infinite column density, which is highly unlikely to represent the actual AGN torus. However, it can provide a useful initial test to investigate the AGN spectrum for CT obscuration. Because it does not self-consistently model the fluorescence emission lines expected from a CTAGN, we included a Gaussian component in the model to simulate Fe-Kα narrow line emission, which is the most prominent line produced by CTAGNs. The centroid energy and width of the line were fixed to E = 6.4 keV and σ = 10 eV, respectively. Due to the limited number of counts, the inclination of the reflector was fixed to the default value set by the model; i.e., θ_inc ≈ 63°. However, we note that fixing θ_inc to other values (e.g., near the lower and upper limits of the model; θ_inc ≈ 26° and 84°, respectively) has insignificant effects on the parameters obtained. We also fixed the reflection scaling factor to R = −1, to simulate a pure reflection spectrum. In addition, we included the absorbed transmitted component of the AGN, simulating the Compton scattering and photoelectric absorption by the torus using cabs and zphabs models, respectively. This model provides a decent fit to the data (C-stat/dof = 318/330), and measured a column density of N_H(los) = ${2.7}_{-1.2}^{+u}\times {10}^{24}$ cm⁻², consistent with CT column.³⁰ The intrinsic photon index inferred from the best-fit model is Γ = 1.57 ± 0.35, in agreement with the typical intrinsic value for an AGN (e.g., Burlon et al. 2011; Corral et al. 2011). The reflection component of the model dominates the transmitted component at all spectral energies probed, indicating that the spectra are reflection-dominated, consistent with being heavily obscured.

We proceeded to model the X-ray broadband spectrum of the potentially heavily obscured AGN in NGC 1448 using more physically motivated obscuration models to better characterize the broadband spectrum. The two models used are the torus model by Brightman & Nandra (2011) (Model T), and MYTorus model by Murphy & Yaqoob (2009) (Model M). Details and results of the two models are described in Sections 3.2.2 and 3.2.3, respectively. In addition to these models, we added extra components required to account for non-AGN contributions to the X-ray spectrum, and to provide a good fit to the data as described below, and in Section 3.2.1.

At low energies, typically at E ≲ 2 keV, where the direct emission from a heavily obscured AGN is completely absorbed, other processes can dominate. These processes include X-ray emission radiated by unresolved off-nuclear X-ray sources, thermal emission from a hot interstellar medium, gas photoionized by the AGN, and scattered emission from the AGN. Our data are not of sufficient quality to accurately distinguish between these different physical processes. Therefore, we simply parametrized the low energy part of the spectrum covered by Chandra using the thermal model "apec" (Smith et al. 2001),³¹ and a power-law component to simulate the scattered emission from the AGN.

A constant parameter C is often included when analyzing data from multiple X-ray observatories simultaneously to account for cross-calibration uncertainties between the various instruments, and to account for any significant variability of the targets between the observations. As detailed earlier, we do not find any significant differences between the Chandra and NuSTAR spectra, indicating that there has not been significant variability between the two observations. Based on the NuSTAR calibration paper (Madsen et al. 2015), the cross-calibration uncertainty of Chandra with respect to NuSTAR is ≈1.1. We therefore decided to fix this parameter to the value found by Madsen et al. (2015) as a conservative approach. However, we note that allowing this parameter to vary in all models returns results that are consistent with this value within statistical uncertainties.

Our two models are described in xspec as follows:

$$\begin{eqnarray}\mathrm{Model\; T} & = & {\rm{constant}}\ast {\rm{phabs}}\ast ({\rm{apec}}\\ & & +{\rm{cons}}\ast {\rm{zpow}}+2\times {\rm{tbabs}}\ast {\rm{zpow}}\\ & & +{\rm{torus}}),\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}\mathrm{Model}\ {\rm{M}} & = & {\rm{constant}}\ast {\rm{phabs}}\ast ({\rm{apec}}\\ & & +{\rm{cons}}\ast {\rm{zpow}}+2\times {\rm{tbabs}}\ast {\rm{zpow}}\\ & & +{\rm{zpow}}\ast {\rm{mytz}}+{\rm{myts}}+{\rm{mytl}}).\end{eqnarray} \tag{ 2 }$$

The two additional absorbed power-law components (tbabs ∗  zpow) are included in the models to take into account the two off-nuclear sources detected within the extraction region in the Chandra data. We discuss these sources in the following section.

3.2.1. Off-nuclear X-Ray Sources

3.2.2. Model T

3.2.3. Model M

As described in Section 2.3, we extracted optical spectra from three different regions along the slit; i.e., the AGN, optical peak, and across the whole galaxy. The extracted spectra, along with the fits obtained from the spectral synthesis modeling (see Section 2.3), and the fit residuals, are shown in Figure 5. As shown in this figure, no significant residuals were left after fitting the observed spectra with stellar population templates, without the need of additional AGN continuum components. This suggests that the AGN optical continuum is highly obscured. Spectral analysis revealed that the stellar emission from all three regions is due to both young (5 Myr) and old (5 Gyr) stellar populations, dominated by the latter with ∼75%–84% contributions measured between the three spectra. Although the optical peak is where the total optical emission is the brightest, the emission lines are the strongest at the Chandra position, providing independent evidence of the AGN location within the galaxy.

Zoom In Zoom Out Reset image size

The interstellar extinction measured toward the stellar population at the AGN position from the spectral synthesis modeling is much higher than at the optical peak position, ${A}_{V}^{\mathrm{AGN}}={2.15}_{-0.06}^{+0.16}$ mag and ${A}_{V}^{\mathrm{peak}}={0.59}_{-0.05}^{+0.04}$ mag, respectively, further suggesting that the optical emission from the AGN is obscured along our line of sight. The total extinction measured from the whole galaxy spectra, ${A}_{V}^{\mathrm{gal}}=$ 0.73 ± 0.06 mag, is ∼3×  lower than that measured at the AGN position, demonstrating that the extinction across the galaxy is non-uniform and inhomogeneous.

We measured the emission line fluxes and ratios for the AGN, optical peak, and the whole galaxy from the extinction-corrected residual spectra (see Table 2). First, we calculated the extinctions toward the AGN narrow line region (NLR) from the Balmer decrements at the three positions. We assumed an intrinsic Balmer decrement of (Hα/Hβ)_int = 2.86, which corresponds to a temperature of T = 10⁴ K and an electron density n_e = 10² cm⁻³ for Case B recombination (Osterbrock 1989). Based on these, we found ${A}_{V}^{\mathrm{AGN}}=$ 1.89 ± 0.03 mag, ${A}_{V}^{\mathrm{peak}}=$ 0.74 ± 0.09 mag, and ${A}_{V}^{\mathrm{gal}}=$ 1.03 ± 0.04 mag. The high extinction measured at the AGN position provides strong evidence that the AGN is heavily obscured at optical wavelengths by the host galaxy.

Using the emission line ratios tabulated in Table 2, we constructed Baldwin–Philips–Terlevich (BPT) diagnostic diagrams (Baldwin et al. 1981) for NGC 1448. This is shown in Figure 6. Based on these diagrams, all three spectra fall within the region of Seyfert galaxies, with the AGN having slightly higher emission line ratios. These results provide the first identification of the AGN in NGC 1448 at optical wavelengths, and also confirm the position of the AGN in the galaxy. The total galaxy spectrum is located close to the H ii/low ionization emission line regions (LINERs) of the BPT diagrams. This might explain why the AGN was misidentified as an H ii region in Veron-Cetty & Veron (1986); i.e., due to host galaxy contaminations as a result of the larger slit size used to extract the spectrum.

Zoom In Zoom Out Reset image size

The separation between the AGN and the optical peak, ∼225 pc, is consistent with the NLR scale observed in AGNs with comparable luminosities as NGC 1448 (Bennert et al. 2006). The apparent separation observed between the two can be attributed to the presence of a dust lane which obscures part of the system along our line of sight.

The intrinsic [O iii]λ5007 Å flux of the AGN corrected for the Balmer decrement is about an order of magnitude higher than the observed luminosity (see Table 2). The [O iii] line emission in AGN is mostly produced in the NLR due to photoionization from the central source. Since this region extends beyond the torus, it does not suffer from nuclear obscuration like the X-ray emission. However as we demonstrated earlier, it can suffer from significant optical extinction from the host galaxy. Indeed, in extreme cases, the host galaxy obscuration can be so high that the optical Balmer decrement only provides a lower limit on the extinction (Goulding & Alexander 2009).

Using our intrinsic [O iii] luminosity measured for the AGN, L${}_{[\mathrm{OIII}]}=(6.89\pm 0.08)\times {10}^{38}$ erg s⁻¹, we proceeded to compare the intrinsic X-ray luminosity determined from our X-ray spectral fitting with that predicted from the X-ray:[O iii] intrinsic luminosity relationship of Panessa et al. (2006). Based on this correlation, we infer an intrinsic X-ray luminosity of L_2−10,int = (0.2–4.5) × 10⁴⁰ erg s⁻¹.³³ This is consistent with that determined from our X-ray spectral fittings, providing confidence that our analysis is reliable.

The high spatial resolution MIR observation by Gemini/T-ReCS measured a 12 μm luminosity of λL_{12 μm} = (4.90 ± 0.93) × 10⁴⁰ erg s⁻¹ for NGC 1448 (see Section 2.4). MIR luminosity is predicted to provide an accurate estimate for the intrinsic luminosity of the AGN. This is because the absorbed X-ray radiation from the central engine is mostly re-emitted in the MIR by the torus. To further test the absorption-corrected X-ray luminosity measured from our spectral analysis, we also compared the luminosity measured from our analysis with that predicted from the X-ray:MIR correlation constructed based upon high angular resolution MIR observations of local Seyferts, which has been shown to trace the intrinsic X-ray luminosity of AGNs very well (e.g., Horst et al. 2008; Gandhi et al. 2009; Asmus et al. 2015). The correlation with respect to the recent Asmus et al. (2015) relation is shown in Figure 7. The figure shows that the observed 2–10 keV luminosity of the AGN in NGC 1448 is about two orders of magnitude lower than this relation, similar to other local bona-fide CTAGNs (Gandhi et al. 2014), indicating extreme obscuration.³⁴ However, after correcting for absorption measured from our X-ray spectral modeling, the luminosity agrees very well with the intrinsic relationship of Asmus et al. (2015), further supporting our analysis.

Zoom In Zoom Out Reset image size