THE NuSTAR VIEW OF REFLECTION AND ABSORPTION IN NGC 7582

We reduced data from both NuSTAR focal plane modules, FPMA and FPMB (Harrison et al. 2013), using the standard pipeline in the NuSTAR Data Analysis Software (NUSTARDAS) version 1.4.1, distributed with HEASOFT version 6.16. Instrumental responses were taken from the NuSTAR calibration database (CALDB) version 20150316. The raw event files were cleaned and filtered for South Atlantic Anomaly (SAA) passages using the nupipeline script. The source was well detected in the entire NuSTAR bandpass. The cleaned events were further processed using the nuproducts script, producing light curves, spectra and response files. For both modules we extracted spectra and light curves from a circular extraction region with a $75^{\prime\prime}$ radius centered on the point source. This choice maximizes the signal-to-noise ratio above 25 keV and leaves ample detector area free for background sampling in short, ≃20 ks NuSTAR observations (Baloković et al. 2015, in preparation). Background was extracted from polygonal regions that cover the same detector as the source in each focal plane module. We do not use NuSTAR data below 3 keV or above 78 keV. We grouped all NuSTAR spectra with a minimum of 50 counts per bin before background subtraction.

The Swift-XRT observation was performed in the Photon Counting mode (Hill et al. 2004; Burrows et al. 2005). The data were reduced using the task xrtpipeline (version 0.12.6), which is a part of the XRT Data Analysis Software within HEASOFT. The spectrum was extracted from a circular region 20'' in radius centered on the point source. The background was extracted from a large annular source-free region encircling the source region. We used the response file swxpc0to12s6_20010101v014.rmf from the Swift-XRT CALDB, while the auxiliary response file was generated using the task xrtmkarf. Due to low count statistics, the Swift-XRT spectrum is binned to a minimum of 10 photons per bin.

An alternate explanation of the discrepancy is a hidden nucleus, where the Compton-thick torus actually intersects the line of sight to the nucleus, contributing obscuration as well as reflection. We can test this scenario using the MYTorus model (Murphy & Yaqoob 2009) in xspec. The model has three components (the absorbed zeroeth-order continuum, the Compton reflection hump, and the Fe Kα line) with the viewing angle frozen at 65°, an opening angle of the torus of 60°, and equatorial column density free to vary (although it is tied between the three components). Using the recommended fully coupled model plus soft extended emission as modeled before (power law plus emission lines) could not provide an acceptable fit (${\chi }^{2}/$dof $\;=\;300/185$ with ${\rm{\Gamma }}\sim 1.4$) due to extra emission in the 3–15 keV range. We therefore included the same absorbed power law which in the previous model was considered the primary continuum:

zphabs(zphabs × leaked power law + MYTZ(0th order power law) + MYTS(Compton reflection) + MYTL(Fe Kα)) + soft power law + zGauss[×5].

This provided a good fit to observation 1 with ${\chi }^{2}/$dof $=188/182.$ Data and data/model ratio plots for this model are shown in Figure 3. The relative strength of this power law compared to the inferred strength of the heavily absorbed/reflected zeroth order continuum is quite high to be scattered continuum, $\sim 20\%$ as compared to a more typical $\lt 10\%$ (e.g., Noguchi et al. 2010). It is more likely that it is leaked emission due to a patchy absorber with a covering fraction of $f\sim 80\%.$ This power law is still absorbed by a variable 10²³ cm⁻² column in the line of sight which may be part of the hidden BLR or the patchy torus.

Zoom In Zoom Out Reset image size

Including the data from observation 2, we obtain a good fit (${\chi }^{2}/$dof $\;=\;355/312$) with all parameters tied save for the leaked power law normalization and absorption column. Using this model we find that observation 2 has a covering fraction of $f\sim 90\%$ and an increase in the column density from 2.6 to 3.6 $\times {10}^{23}$ cm⁻² in the line of sight.

In place of the MYTorus model with the extra leaked power law component, we also tried a decoupled MYTorus model. This models clumpiness of the obscuring and reflecting medium, allowing the material in our line of sight to have a different density than the systemic average. To do this we used the recommended full model as above, but untied the column density and the normalization between the MYTorus direct and scattered components (and between the observations), although degeneracy led us to tie ${N}_{{\rm{H}}}$ for the line component to the scattered component. This model provided an acceptable fit (${\chi }^{2}/$dof $\;=\;351/300$), not quite as good as with the leaked power law model. The best fit ${N}_{{\rm{H}}}$ values were 5 $\times \;{10}^{23}$ cm⁻² in the line of sight and 3 $\times \;{10}^{24}$ cm⁻² out of the line of sight. Additionally, this model led to a scattered component that was six times higher than expected and a line component that was three times higher than expected, even with the higher column. This could potentially be explained by variation in the source power, except that we know our 20–50 keV flux is comparable to the average (see Figure 1, lower panel). Simulating Compton scattering from clumpy material we also tried using two scattering components with inclination angles of 0° and 90°. This led to a good fit but with the same issue of unexpectedly high reflecting fractions.

We also applied the alternative bntorus model from Brightman & Nandra (2011) including an absorbed leaked power law component as with the MYTorus model. We found very similar absorption and continuum parameters to the MYTorus model with an equatorial torus column density of $2.4\times {10}^{24}$ cm⁻² and an incident power law normalization of 1.3 $\times \;{10}^{-2}$ counts s⁻¹ keV⁻¹ at 1 keV. We measured a viewing angle of $65^\circ \pm 5^\circ$ and an opening angle of $60^\circ \pm 5^\circ .$ Note that for both toroidal models, the equatorial column density and therefore the reflection spectrum strength, is heavily influenced by the absorption column which may be variable in this source given the variability of the lower column density absorber.

It has been 16 years since the bright BeppoSAX observation in 1998 (T00) and the reflection signatures have not shown the substantial reduction predicted by P07 (see Figure 1 and Table 3). With no X-ray data between 1998 and 2000 this means a lower limit of 12 years until the 2012 NuSTAR observations, implying a minimum inner reflecting radius of the material at 6 lt-yr (2 pc) from the source. We note that HST data put an upper limit on the size of the nuclear region of 8 pc (Bianchi et al. 2007).

Table 3.  Historical Absorption/Reflection Parameters

Date	Observatory	Observed	${N}_{{\rm{H}}}$	Fe Kα	pexrav
		F2–10a	(1022 cm−2)	Normb	Normc
1978 Dec	Einstein	64	8${}_{-1}^{+2}$	⋯	⋯
1979 May	Einstein	36	10${}_{-0.3}^{+0.6}$	⋯	⋯
1979 Jun	Einstein	24	13${}_{-3}^{+7}$	⋯	⋯
1979 Nov	Einstein	52	20${}_{-9}^{+5}$	⋯	⋯
1980 May	Einstein	34	5${}_{-3}^{+4}$	⋯	⋯
1984 Jun	EXOSAT	16	15${}_{-9}^{+17}$	⋯	⋯
1988 Oct	Ginga	27	48${}_{-15}^{+25}$	⋯	⋯
1994 Nov	ASCA	15 ± 4	8 ± 1	⋯	⋯
1996 Nov	ASCA	16 ± 2	12 ± 1	⋯	⋯
1998 Nov	BeppoSAX	19.7 ± 0.3	14 ± 1	7.0 ± 3.5	⋯
2001 May	XMM-Newton	2.4 ± 0.3	55${}_{-2}^{+7}$	2.2 ± 0.3	≈10
2003 Jun	RXTE	6.5 ± 0.6	-	3.6 ± 0.8	36 ± 10
2003 Oct	RXTE	13.2 ± 0.2	15 ± 2	3.6*	36*
2004 Feb	RXTE	6.5 ± 0.2	-	3.6*	36*
2004 Aug	RXTE	8.8 ± 0.2	16 ± 5	3.6*	36*
2005 Apr	XMM-Newton	4.0 ± 0.4	130${}_{-7}^{+6}$	2.3 ± 0.1	9.7${}_{-1.8}^{+0.8}$
2007 Apr	XMM-Newton	7.6 ± 0.4	33${}_{-5}^{+4}$	2.4 ± 0.9	9.3 ± 2.1
2007 May	Suzaku	5.3 ± 0.3	44${}_{-2}^{+3}$	2.5 ± 0.5	9.3*
2007 May	Suzaku	4.1 ± 0.3	68${}_{-7}^{+6}$	2.2 ± 0.4	9.3*
2007 Nov	Suzaku	3.2 ± 0.2	110${}_{-14}^{+11}$	2.2 ± 0.4	9.3*
2007 Nov	Suzaku	2.6 ± 0.5	120 ± 20	2.4 ± 0.3	9.3*
2012 Aug	NuSTAR	4.8 ± 0.1	24${}_{-7}^{+3}$	2.1 ± 0.4	10 ± 3
2012 Sep	NuSTAR	3.2 ± 0.1	56${}_{-20}^{+10}$	2.1*	10*

Notes. Comparing observed flux, absorption, and reflection measurements over time with the traditional pexrav model. A star (*) indicates a tied or frozen parameter. A dash (-) indicates an unconstrained parameter. Early models did not include the reflection hump, which may affect measured ${N}_{{\rm{H}}}$ and Fe line flux values. References: Risaliti et al. (2002 and references therein), T00, P07, B09, and this work. Note that analysis of the four epochs of RXTE data are presented in Appendix.

^aFlux is in units of 10⁻¹² erg cm⁻² s⁻¹. ^bFe line flux is in units of 10⁻⁵ photons cm⁻² s⁻¹. ^c pexrav Norm is in units of 10⁻³ photons keV⁻¹ cm⁻² s⁻¹.

Download table as:  ASCIITypeset image

From the unabsorbed X-ray luminosity (measured in 1998) of $\sim 7\times {10}^{41}$ erg s⁻¹ we can estimate a bolometric luminosity of $\sim 2\times {10}^{43}$ erg s⁻¹ (see, e.g., Vasudevan et al. 2010) corresponding to a dust sublimation radius of ∼0.06 pc. This is consistent with mid-infrared interferometry results for Seyfert tori that predict a torus size of 0.1–0.5 pc for this luminosity (Burtscher et al. 2013, their Figure 36), with the hot inner wall predicted from near-infrared dust reverberation at 0.01–0.1 pc (e.g., Suganuma et al. 2006). An estimation of the torus size from fitting a clumpy torus model to infrared data gives an approximation of the median torus size of 1.5 pc (Alonso-Herrero et al. 2011). This is in conflict with the shut-off theory as it makes it unlikely that the torus is larger than the necessary 2 pc. It is possible that the bulk of the reflection signal is originating outside the torus, for instance in the extended dust lanes of the galaxy, however the covering factor of the dust lane would likely be far too low to explain the extreme strength of the reflection and is optically thin in the line of sight.

Other observations between the mid 1990's and today also detract from the shut-off interpretation. HST observed the source twice, in 1995 June and 2001 July, and found that the optical source flux had increased by a factor of 60% (Bianchi et al. 2007), inconsistent with the source shutting off during this time. Additionally, RXTE and Swift-BAT have monitored the source from 2003 to 2011 and found a highly variable observed flux, the brightest of which was nearly as bright as the BeppoSAX observation, though relatively short lived. Soldi et al. (2014) performed variability analysis on the Swift-BAT light curves of 110 Seyferts, including NGC 7582. Their results show that above 35 keV, the variability estimator of NGC 7582 is consistent with that of Seyferts in general, which rules out a dominant contribution to the high energy flux from reflection.

With the advent of physical torus models we are able to investigate fully an alternative to this scenario: that the source has remained at a relatively stable average X-ray luminosity (though with high intrinsic source variability) while increased obscuration by the torus has weakened the observed signal in the past decade. Our obscuring torus model shows that this scenario is entirely plausible with a patchy absorber (${N}_{{\rm{H}}}$ $\;\sim \;{10}^{24}$ cm⁻²) providing a strong reflection signal and heavily absorbed primary continuum while allowing for leaked emission of the primary continuum. Another layer of absorption (${N}_{{\rm{H}}}$ $\;\sim \;{10}^{23}$ cm⁻²) fully covers this leaked emission. This scenario is illustrated in Figure 5.

Zoom In Zoom Out Reset image size

B09 noted that the full-covering absorber accounted for nearly all the source variability observed by XMM-Newton and Suzaku. They theorized that this layer might be associated with hidden BLR clouds close to the central source due to the rapidity of the variations. We confirm this variability with a 50% change in the absorption column in 15 days. We also see some variability in the normalization of this leaked power law, a drop of 30% which could be explained by an increase in the covering fraction of the torus clouds in the line of sight.

We found in our application of the pexrav model that the Fe Kα line strength is unexpectedly weak. The Fe line strength is about half what we would have expected were it to arise in the same material as the Compton reflection hump. This could be explained by a low Fe abundance (∼0.6) or obscuration of the Fe line region by Compton thin material. The MYTorus model does not find this mismatch, the Fe line strength matching well with the modeled torus.

The torus model implies a much higher intrinsic luminosity for the mostly obscured X-ray emitting corona. The unabsorbed 2–10 keV fluxes from Table 2 correspond to intrinsic luminosities of ${L}_{2-10}=5.3$ (4.2) $\times \;{10}^{41}$ erg s⁻¹ for observations 1(2) from the pexrav model and ${L}_{2-10}=3.5\;$ $\times \;{10}^{41}$ erg s⁻¹ for the MYTorus model. The infrared 6 μm flux of the source is 0.183 Jy (Lutz et al. 2004) corresponding to an infrared luminosity of 5 $\times \;{10}^{42}$ erg s⁻¹. Using empirical relations we predict an intrinsic 2–10 keV X-ray flux of ∼(5–10) $\times \;{10}^{-11}$ erg cm⁻² s⁻¹ (e.g., Gandhi et al. 2009; Stern 2015), which is consistent with the unabsorbed source flux in our hidden nucleus model, 6 $\times \;{10}^{-11}$ erg cm⁻² s⁻¹.

If we accept the torus model as the most plausible physical interpretation of the data, we find that this source includes three distinct absorbers: the slowly varying patchy torus ($3.6\times {10}^{24}$ cm⁻²), the rapidly varying full covering absorber (∼3–12 × ${10}^{23}$ cm⁻²), and the dust lane in the host galaxy ($\sim {10}^{22}$ cm⁻²), seen by Chandra and included in our models, though not directly measured due to its degeneracy with the full-covering absorber. The Compton-thick absorbed continuum is degenerate with the reflection hump in spectral modeling, however it should not affect the measurements of the 10²³ cm⁻² absorption that have been performed by, e.g., B09 and others. We might expect similar variability in Compton-thick absorption from the patchy torus on longer timescales.

From our spectroscopic analysis of NGC 7582 we find that it is plausible that this source could indeed contain a hidden nucleus absorbed by a patchy torus intersecting the line of sight with a covering fraction of ∼80%–90%. The more traditional phenomenological model that has been used to describe this source in many recent papers fits the data equally well, however, the NuSTAR data require a higher than expected reflection fraction. One suggested physical explanation, that the source has switched off, is not viable since it requires a torus at a distance of ≳2 pc, contrary to predictions from infrared reverberation and interferometry measurements of AGN.

Much of the variability of the source is also consistent with patchy absorption. Continuum flux changes could be due to a combination of a changing covering factor of Compton-thick material in the line of sight and intrinsic luminosity changes. Short term spectral variation observed in this source over the past decade has been attributed to changing column density in the full covering Compton-thin ($\sim {10}^{23}$ cm⁻²) BLR clouds as seen by B09. Additional full covering absorption ($\sim {10}^{22}$ cm⁻²) from the galactic scale dust lane has also been observed. Detection of the extended soft X-ray emitting gas is consistent with previous observations by Chandra and XMM-Newton.

RXTE observed NGC 7582 from 2003 to 2004 with four clusters of 3–10 exposures each. The light curve and stacked spectrum for these observations were published in Rivers et al. (2011, 2013), with the caveat that the source was highly variable over these observations on timescales as short as a few days, both in flux and hardness. In order to investigate this variability we have grouped the spectra by month: 2003 June (10.3 ks total net exposure), 2003 October (83.6 ks), 2004 February (45.4 ks) and 2004 August (45.7 ks). The PCA data were extracted using HEASOFT version 6.7 software (no updates to RXTE extraction procedures have been included since). We extracted PCA STANDARD-2 data from PCU 2 using only events from the top Xe layer in order to maximize signal-to-noise. Standard screening was applied with time since SAA passage >20 minutes and background model file "pca_bkgd_cmfaintl7_eMv20111129.mdl."

We performed spectral fitting on all four epochs of RXTE data between 3.5 and ∼20 keV. Naive modeling of these data is difficult owing to the low spectral quality. We therefore applied both the pexrav and MYTorus models to the four RXTE epochs simultaneously, omitting the soft power law and lines and tying parameters in order to reduce the number of free parameters.

For the pexrav model we initially tied the photon index and Fe line energy (width frozen) between the spectra. The Fe line and pexrav normalizations were consistent across all four epochs and we tied those parameters as well. Best fit parameters are listed in Table 4 and data and data/model ratio plots are shown in Figure 6. The first epoch was highly reflection dominated with a weak continuum which did not allow us to place constraints on the absorption. Epochs two and four showed very similar levels of absorption (${N}_{{\rm{H}}}$ $\sim 2\times {10}^{23}$ cm⁻²) while epoch three appeared to be heavily absorbed (${N}_{{\rm{H}}}$ $=1.4\times {10}^{24}$ cm⁻²) and reflection dominated. There is strong degeneracy between the pexrav strength and photon index in these data. If we freeze the photon index to 1.9 (1.8) to bring it more in line with other measurements, we find that the pexrav strength drops to ∼25 (15) $\times {10}^{-3}$ photons keV⁻¹ cm⁻² s⁻¹.

Zoom In Zoom Out Reset image size

Table 4.  Broadband Model Parameters

pexrav	Observed	Unabsorbed	Photon	PL	Column	Fe Kα	pexrav	${\chi }^{2}/$dof
Model	${F}_{2-10}$ a	Power Law	Index	Normb	Densityc	Line Normd	Norme
		${F}_{2-10}$ a	(Γ)		(NH)
RXTE Jun 03	6.5 ± 0.6	10 ± 8	2.1 ± 0.1	0.4 ± 0.4	unconstrained	3.6 ± 0.8	36 ± 10	157/181
RXTE Oct 03	13.2 ± 0.2	17 ± 2	⋯	7.2 ± 1.0	15 ± 2	⋯	⋯	⋯
RXTE Feb 04	6.5 ± 0.2	11 ± 7	⋯	0.5 ± 0.3	unconstrained	⋯	⋯	⋯
RXTE Aug 04	8.8 ± 0.2	8 ± 2	⋯	3.1 ± 0.6	16 ± 5	⋯	⋯	⋯
MYTorus	Observed	Unabsorbed	Photon	Absorbed PL	Leaked PL	Full Covering	Torus Column	${\chi }^{2}/$dof
Model	${F}_{2-10}$ a	Power Law	Index	Normb	Normb	Column Densityc	Densityc
		${F}_{2-10}$ a	(Γ)			(NH)	(NH)
RXTE Jun 03	6.5 ± 0.6	100 ± 20	$2.1\;\pm \;0.1$	44 ± 10	3.0 ± 0.7	7 ± 5	250 ± 30	166/181
RXTE Oct 03	13.1 ± 0.3	⋯	⋯	⋯	13 ± 2	15 ± 1	⋯	⋯
RXTE Feb 04	6.5 ± 0.2	⋯	⋯	⋯	3.5 ± 0.6	10 ± 4	⋯	⋯
RXTE Aug 04	8.7 ± 0.2	⋯	⋯	⋯	7.1 ± 0.9	15 ± 2	⋯	⋯

Notes. Model parameters to simultaneous fitting of the four epochs of RXTE spectral data with the pexrav and MYTorus models. Note we have fixed the inclination angle to 65° for both models.

^aFlux is in units of 10⁻¹² erg cm⁻² s⁻¹. ^bPower law norm is in units of 10⁻³ counts s⁻¹ keV⁻¹ at 1 keV. ^cColumn densities are in units of 10²² cm⁻². ^dFe line flux is in units of 10⁻⁵ photons cm⁻² s⁻¹. ^e pexrav Norm is in units of 10⁻³ photons keV⁻¹ cm⁻² s⁻¹.

Download table as:  ASCIITypeset image

Applying the MYTorus model to these data, we found strong consistency between the epochs with a torus column density of ${N}_{{\rm{H}}}$ $=\;(2.5\pm 0.3)\times {10}^{24}$ cm⁻², ${\rm{\Gamma }}=2.1,$ and ${\chi }^{2}/$dof = 156/181. The only parameters required to be left free were the full covering absorption and leaked power law normalization. The leaked power law portions were 6 ± 1%, 20 ± 3%, 7 ± 1%, and 15 ± 2% for each epoch, respectively. Allowing the Fe line energy to be free also led to an improvement in the fit (${\chi }^{2}/$dof = 143/177) and a visual improvement in the residuals around 5–6 keV, however is likely not real but calibration related.