TESTING THE UNIFICATION MODEL FOR ACTIVE GALACTIC NUCLEI IN THE INFRARED: ARE THE OBSCURING TORI OF TYPE 1 AND 2 SEYFERTS DIFFERENT?

In order to enlarge the sample of 18 Seyfert galaxies presented in RA09, which comprises 12 Seyfert 2 (Sy2), 2 Seyfert 1.9 (Sy1.9), 1 Seyfert 1.8 (Sy1.8), 2 Seyfert 1.5 (Sy1.5), and 1 Seyfert 1 galaxy (Sy1), we obtained new subarcsecond MIR observations of the Type-1 Seyferts NGC 6221, NGC 6814, and NGC 7469 (see Table 1).

The observations were performed with the MIR camera/spectrograph Thermal-Region Camera Spectrograph (T-ReCS; Telesco et al. 1998) on the Gemini-South telescope during the Summer of 2009. T-ReCS uses a Raytheon 320 × 240 pixel Si:As IBC array, providing a plate scale of 0089 pixel⁻¹, corresponding to a field of view of 285 × 214. The filters employed for the observations were the narrow Si-2 filter (λ_c = 8.74 μm, Δλ = 0.78 μm, 50% cut-on/off) and the broad Qa filter (λ_c = 18.3 μm, Δλ = 1.5 μm, 50% cut-on/off). The resolutions obtained were 03 at 8.7 μm and 05 at 18.3 μm, as measured from the observed point-spread function (PSF) stars. A summary of the observations is reported in Table 2.

Table 2. Summary of MIR Observations

Galaxy	Filters	Observation	On-source		PSF FWHM
		Epoch	Time (s)
			N Band	Q Band	N Band	Q Band
NGC 1097	Si-5, Qa	2005 Sep	456	912	041	052
NGC 1566	Si-2, Qa	2005 Sep	152	304	030	053
NGC 6221	Si-2, Qa	2009 Aug	145	203	032	055
NGC 6814	Si-2, Qa	2009 Aug	145	203	028	053
NGC 7469	Si-2, Qa	2009 Sep	145	203	031	055
NGC 3227	N'	2006 Apr	300	...	039	...
NGC 4151	N, IHW18	2001 May	360	480	053	058

Notes. Images were obtained in the 8.74 μm (Si-2, Δλ = 0.78 μm at 50% cut-on/off), 11.66 μm (Si-5, Δλ = 1.13 μm), and 18.30 μm (Qa, Δλ = 1.5 μm) T-ReCS filters. NGC 3227 was observed in the 11.29 μm (N', Δλ = 2.4 μm) Michelle/Gemini-North filter, and NGC 4151 in the 10.75 μm (N, Δλ = 5.2 μm) and 18.17 μm (IHW18, Δλ = 1.7 μm) OSCIR/Gemini-North filters.

Download table as:  ASCIITypeset image

The standard chopping–nodding technique was used to remove the time-variable sky background, the telescope thermal emission, and the so-called 1/f detector noise. The chopping throw was 15'', and the telescope was nodded 15'' in the direction of the chop every 45 s. The difference for each chopped pair for each given nodding set was calculated, and the nod sets were then differenced and combined until a single image was created. The data were reduced using in-house-developed IDL routines.

Observations of flux standard stars were made for the flux calibration of each galaxy through the same filters. The uncertainties in the flux calibration are typically ∼5%–10% at 8.7 μm and ∼15%–20% at 18.3 μm. PSF star observations were also made immediately prior to or after each galaxy observation to accurately sample the image quality. These images were employed to determine the unresolved (i.e., nuclear) component of each galaxy. The PSF star, scaled to the peak of the galaxy emission, represents the maximum contribution of the unresolved source (100%), where we integrate flux within an aperture of 2''. The residual of the total emission minus the scaled PSF represents the host galaxy contribution, which is analyzed in Section 3. We require a flat profile in the residual for a realistic galaxy profile over the central pixels. Therefore, we reduce the scale of the PSF from matching the peak of the galaxy emission (100%), when necessary, to obtain the unresolved fluxes reported in Table 3. They include corresponding aperture corrections to take into account possible flux losses when integrating the scaled PSF flux in the 2'' aperture.

Figure 1 shows an example of PSF subtraction at various levels (in contours) for NGC 7469 in the Si-2 T-ReCS filter, following Radomski et al. (2002, 2003). The residual profiles from the different scales demonstrate the best-fitting result. The uncertainty in the unresolved fluxes determination from PSF subtraction is ∼10%–15%. Thus, we estimated the errors in the flux densities reported in Table 3 by adding quadratically the flux calibration and PSF subtraction uncertainties, resulting in ∼15% at 8.7 μm and ∼25% at 18.3 μm.

Zoom In Zoom Out Reset image size

In addition to the new T-ReCS observations described in Section 2.1, we include here two more Sy1 galaxies with already published T-ReCS data: NGC 1097 and NGC 1566.⁹ We reduced the images presented in Mason et al. (2007) for NGC 1097 and obtained unresolved MIR fluxes by performing the same technique explained in Section 2.1 (see Table 3). For NGC 1566, we compiled the MIR nuclear fluxes reported RA09. Both galaxies were observed in 2005 September with T-ReCS: NGC 1566 in the Si-2 and Qa filters, and NGC 1097 in the Si-5 (λ_c = 11.66 μm, Δλ = 1.13 μm, 50% cut-on/off) and Qa filters (see Table 2). The resolutions of the images are 03 at 8.7 μm, ∼04 at 11.66 μm, and 05 at 18.3 μm.

NIR subarcsecond-resolution nuclear fluxes compiled from the literature are reported in Table 4. For NGC 1097, NGC 1566, and NGC 7469, Prieto et al. (2010) reported diffraction-limited and near-diffraction-limited adaptive optics NACO/VLT fluxes. The three galaxies are unresolved in the NIR down to the highest resolution achieved (FWHM ∼ 015 for NGC 1097 in the L' band, ∼012 for NGC 1566 in the K band, and ∼008 for NGC 7469 in the H band).

For NGC 7469, Prieto et al. (2010) reported NACO J-, H-, and K-band nuclear fluxes obtained from observations in 2002 November, as well as in the L' and NB-4.05 μm filters, observed in this case in 2005 December. First, we discarded the narrow-band filter, since it is designed to collect the Brα emission, which in the case of this galaxy is important (as inferred from the Brγ line detected with NIR spectroscopy; Genzel et al. 1995; Hicks & Malkan 2008). The L'-band filter also contains Brα, which is very likely compromising the flux measurement, since the L' and NB-4.05 fluxes do not match the SED shape of the remaining J, H, K, Si-2, and Qa data. Indeed, Prieto et al. (2010) also reported a NICMOS/HST flux measurement in the filter F187N, obtained in 2007. This data point lies in between the NACO H and K measurements in wavelength and flux, which gives us extra-confidence in the reliability of the NACO J, H, and K measurements. Considering all the previous, we finally decided to consider the NACO L'-band flux as an upper limit.¹⁰

For NGC 6221 and NGC 6814, which do not have any published subarcsecond-resolution NIR data, we retrieved broadband images from the Hubble Legacy Archive¹¹ obtained with NICMOS on HST. The two galaxies were observed in the F160W filter, using the NIC2 camera, as part of the program 7330 (PI: J. Mulchaey). The typical FWHM for an unresolved PSF is ∼013 using the F160W filter with NIC2. Details of the observations can be found in Scoville et al. (2000) and Regan & Mulchaey (1999). For the analysis, we first separated the nuclear emission from the underlying host galaxy emission. We then applied the two-dimensional image decomposition GALFIT program (Peng et al. 2002) to fit and subtract the unresolved component (PSF). PSF models were created using the TinyTim software package, which includes the optics of HST plus the specifics of the camera and filter system (Krist 1993). We checked that no prominent emission lines are included in the wavelength range covered by the filter F160W. The resulting unresolved NIR fluxes for NGC 6221 and NGC 6814 are reported in Table 4.

We finally include the Sy1.5 galaxies NGC 3227 and NGC 4151 in this study, which were also part of the RA09 sample. The MIR nuclear fluxes are the same reported in the latter work (see description of the observations in Section 2.1 in RA09), whereas the NIR fluxes have been updated (see Table 4).

Using the NIR nuclear fluxes reported in Table 4 in combination with our MIR unresolved measurements (Table 3) we construct AGN-dominated SEDs for the five Sy1 galaxies.

The most spectacular morphological feature of this Sy1 galaxy is a circumnuclear ring of powerful starbursts of ∼1.6 kpc diameter, which is deeply embedded in a large cloud of molecular gas and dust. This ring contains several super star clusters and regions of star formation and it accounts for two-thirds of the galaxy bolometric luminosity (Genzel et al. 1995). The star-forming ring has been the subject of study of several works in different wavelengths (see Díaz-Santos et al. 2007, hereafter DS07, and references therein), including the MIR (Miles et al. 1994; Soifer et al. 2003; Horst et al. 2009). Based on VISIR/VLT MIR observations at 12.3 μm, Horst et al. (2009) report the detection of distinct knots of star formation around the nucleus located at a distance of ∼13 (∼400 pc), although with low signal-to-noise ratio. Using NIR HST data, DS07 identified 30 clusters of star formation in the ring at 1.1 μm. By fitting the individual ultraviolet-to-NIR SEDs of the clusters, the authors reported the presence of a dominant intermediate age population (8–20 Myr) and a younger and more extinguished one (∼1–3 Myr). The latter does not coincide with the optical/NIR continuum-emitting regions, but seems to be traced by the MIR/radio emission, less affected by extinction than the optical/NIR.

Figure 2 shows the high-resolution 8.74 and 18.3 μm T-ReCS images of NGC 7469. At both wavelengths we detect extended emission with high signal-to-noise coincident with the star-forming ring. Indeed, our flux maps appear very similar to the high-resolution 11.7 μm contours presented by Miles et al. (1994) and to the high-resolution (02) VLA 8.4 GHz radio map presented in Colina et al. (2001). We identified the brightest knots in our MIR images in Figure 2 using the same notation as in Miles et al. (1994), where the AGN is labeled A. The B and C knots in our images correspond to the brightest regions in radio, according to the 5 GHz (Wilson et al. 1991) and 8.4 GHz radio maps (Colina et al. 2001). We also identified the knots D, E, and F. In Table 5, we report the star clusters identified by DS07 which better match the positions of the A to F MIR compact regions. The distances between these knots and the AGN range from 14 to 18 (median distance of ∼480 pc). All the knots appear more compact in the 18.3 μm image than in the 8.74 μm, where the ring emission is more extended. This is expected since the 8.74 μm filter contains the 8.6 μm polycyclic aromatic hydrocarbon (PAH) feature, whereas the 18.3 μm filter is mostly probing hot dust emission. As found by Spitzer, the ∼8 μm PAH emission of nearby galaxies appears to be more extended than at ∼24 μm (e.g., Helou et al. 2004; Calzetti et al. 2005).

Zoom In Zoom Out Reset image size

Table 5. Positions and Flux Densities of NGC 7469 MIR Knots

Knot	ID	8.74 μm			18.3 μm
	DS07	(X)	(Y)	Flux (mJy)	(X)	(Y)	Flux (mJy)
A	...	0.00	0.00	229	0.00	0.00	1320
B	C20	−158	−089	11.9	−154	−093	50
C	C6	109	107	13.4	095	106	79
D	C10	−049	131	11.5	−058	138	107
E	C7, C15	−040	−136	15.8	−052	−149	104
F	C12	125	−089	3.7	147	−083	14.8

Notes. Columns 1 and 2 indicate the label assigned to each knot in Figure 2 and in DS07. B, C, and D are also coincident with the R3, R2, and R1 regions in the VLA 8.4 GHz radio maps in Colina et al. (2001). Columns 3, 4 and 6, 7 indicate the position of each knot relative to A at 8.7 and 18.3 μm, respectively. Columns 5 and 8 list the flux densities of the knots in a 055 aperture radius. Fluxes include aperture corrections of 19% at 8.7 μm and 37% at 18.3 μm. Errors in flux densities are dominated by uncertainties in the flux calibration (∼5%–10% at 8.7 μm and ∼15%–20% at 18.3 μm).

Download table as:  ASCIITypeset image

In Table 5, we report aperture fluxes for the six identified knots in both the 8.74 and 18.3 μm images calculated using IRAF.¹² The aperture radius was defined on the basis of the resolution element in the Qa band (055), and corresponding aperture corrections were applied, since the knots are only barely resolved in both bands. Positions relative to the nucleus (A) in arcseconds are also given.

The galaxy NGC 6221 is a prototypical example of the so-called X-ray–loud composite galaxies (see, e.g., Moran et al. 1996). This classification comes from the comparison between its optical spectrum, which is starburst-like, and its X-ray data, where the AGN is revealed (Levenson et al. 2001). According to the X-ray data, and in particular the width of the Fe Kα line, the orientation of the Seyfert nucleus is Type-1 (Levenson et al. 2001). However, as mentioned above, the optical spectrum of the galaxy resembles more that of a starburst galaxy, implying that there must be a big amount of dust (likely associated with the starburst) along the line of sight (LOS) hiding the BLR. A nuclear optical extinction of A_V = 3.0 mag was measured from the optical spectrum by Levenson et al. (2001). They also presented HST images of the central region of NGC 6221 in the optical (F606W/WFPC2) and in the NIR (F160W/NICMOS) and identified the bright central NIR source as the AGN. At optical wavelengths, the nucleus is diffuse and weaker than other bright knots, identified as star clusters.

Our 8.7 and 18.3 μm images of NGC 6221 are shown in Figure 3. They reveal spectacular extended emission with two bright knots. The AGN is centered in both images. The other bright knot, at ∼19 southwest from the nucleus, which is roughly coincident with the southwest starburst region shown in the F606W optical image in Figure 3 of Levenson et al. (2001), reaches practically the same intensity as the AGN at 8.7 μm and appears brighter at 18.3 μm. We measured aperture fluxes in a 07 radius for the southwest knot, selected to collect the bulk of its MIR emission, and obtain fluxes of 42 mJy at 8.7 μm and 422 mJy at 18.3 μm. Surrounding the AGN there is more diffuse emission extending toward the north, which in this case is more intense in the 8.7 μm image than in the 18.3 μm one. As already mentioned in Section 3.1, the ∼8 μm PAH emission of nearby galaxies appears more extended than the ∼24 μm emission (e.g., Helou et al. 2004; Calzetti et al. 2005).

Zoom In Zoom Out Reset image size

Using the MIR and NIR data reported in Tables 3 and 4, we construct subarcsecond-resolution nuclear SEDs in the wavelength range from ∼1 to 18 μm for the five Type-1 Seyferts analyzed here. Figure 4 shows a comparison between their spectral shapes and the average Sy2 SED from RA09. This mean template was constructed using individual Sy2 data of the same angular resolution as that achieved in this work (≲055). In the same way, we have constructed an average Type-1 Seyfert template using the IR nuclear SEDs of the seven galaxies studied here.¹³ The spectral shape of Sy1 and Sy1.5 galaxies is practically identical (as can be seen from Figure 4). Based on the previous, and on the fact that both types of nuclei present broad lines in their optical spectra, in this work we consider them as Type-1 Seyferts.

Zoom In Zoom Out Reset image size

The Sy2 template defines the wavelength grid, and we performed a quadratic interpolation of nearby measurements of the individual Sy1 galaxies onto its scale (1.265, 1.60, 2.18, 3.80, 4.80, 8.74, and 18.3 μm). We did not interpolate the sparse observations of NGC 6221 and NGC 6814, which only have NICMOS 1.6 μm data in addition to the MIR measurements. The interpolated fluxes were used solely for the purpose of deriving the average Sy1 template. The error bars correspond to the standard deviation of each averaged point, except for the 8.74 μm point (the wavelength chosen for the normalization). In this case, we assigned a 15% error, which is the nominal percentage considered for the N-band flux measurements (see Section 2).

We measured the 1.265–18.3 μm IR slope ($f_{\nu }\ \alpha \ \nu ^{-\alpha _{{\rm IR}}}$) of the Sy1 template, which is representative of the individual Sy1 SEDs: α_IR = 1.7 ± 0.3. We also calculated the NIR (α_NIR = 1.6 ± 0.2, from 1.265 to 8.74 μm) and MIR spectral indexes (α_MIR = 2.0 ± 0.2, using the 8.7 and 18.3 μm points). A flat NIR slope indicates an important contribution of hot dust emission (up to ∼1000–1200 K; Rieke & Lebofsky 1981; Barvainis 1987) that comes from the immediate vicinity of the AGN. α_IR, α_NIR, and α_MIR values for the individual Type-1 Seyfert galaxies and for the mean Sy1 and Sy2 SEDs are shown in Table 6. The shape of the Sy2 mean SED is very steep (α_IR = 3.1 ± 0.9, α_NIR = 3.6 ± 0.8, and α_MIR = 2.0 ± 0.2) compared with those of the Type-1 Seyferts. In general, Sy2 have steeper 1–10 μm SEDs than Sy1 (Rieke 1978; Edelson et al. 1987; Ward et al. 1987; Fadda et al. 1998; Alonso-Herrero et al. 2001, 2003). On the contrary, the MIR slope results to be the same (α_MIR = 2.0 ± 0.2) for both the Sy1 and Sy2 templates.

Alonso-Herrero et al. (2003) also reported IR spectral indices measured from 1 to 16 μm for Sy1 and Sy1.5 galaxies (α^Alonso_IR = 1.5–1.6). On the other hand, the NIR slopes of the Sy1.8 and Sy1.9 in RA09 have intermediate values between those of Sy2 and Sy1 (mean slope α_IR = 2.0 ± 0.4), also in agreement with the IR slopes reported in Alonso-Herrero et al. (2003) for Sy1.8 and Sy1.9 (α^Alonso_IR = 1.8–2.6). In summary, the slope of the IR nuclear SED is generally correlated with Seyfert type: Type-2 nuclei show steeper SEDs, whereas Type-1 and intermediate Seyferts are flatter. The NIR excess responsible for flattening the SED of the Type-1 nuclei would come from the contribution of hot dust in the directly illuminated faces of the clouds in the torus, as well as from the direct AGN emission (i.e., the tail of the optical power-law continuum).

A similar comparison between Type-1 and Type-2 spectral shapes can be done by using the H/N and N/Q flux ratios. The H/N ratio is larger for Type-1 Seyferts (0.07 ± 0.03) than for Sy2 (0.003 ± 0.002), as measured from the individual values of the Sy1 considered here and the Sy2 in RA09. The difference in the H/N ratio between Sy1 and Sy2 galaxies is significantly different at the 100% confidence level, according to the Kolmogorov–Smirnov (K-S) test. This ratio depends on both the torus inclination and covering factor, as we will discuss in Section 6.1. On the other hand, N/Q is very similar for Type-1 and Type-2 Seyferts (mean values of 0.27 ± 0.11 and 0.23 ± 0.14, respectively). Values of H/N and N/Q for the individual Sy1 galaxies and the average templates are reported in Table 6.

The clumpy dusty torus models of Nenkova et al. (2002) hold that the dust surrounding the central engine of an AGN is distributed in clumps, instead of homogeneously filling the torus volume. These clumps are distributed with a radial extent Y = R_o/R_d, where R_o and R_d are the outer and inner radius of the toroidal distribution, respectively (see Figure 5). The inner radius is defined by the dust sublimation temperature (T_sub ≈ 1500 K), with R_d = 0.4(1500KT⁻¹_sub)^2.6(L/10⁴⁵ erg s^-1)^0.5 pc. Within this geometry, each clump has the same optical depth (τ_V, defined at the V band). The average number of clouds along a radial equatorial ray is N₀. The radial density profile is a power law (∝r^−q). A width parameter, σ, characterizes the angular distribution of the clouds, which has a smooth edge. The number of clouds along the LOS at an inclination angle i is $N_{{\rm LOS}}(i) = N_0 e^{(-(i-90)^2/\sigma ^2)}$. Finally, the optical extinction produced by the torus along the LOS is computed as $A_{V}^{{\rm LOS}} = 1.086 N_0 \tau _{V} e^{(-(i-90)^{2}/\sigma ^{2})}$ mag. For a detailed description of the clumpy models, see Nenkova et al. (2002, 2008a, 2008b).

Zoom In Zoom Out Reset image size

The clumpy database now contains 1.2 × 10⁶ models, calculated for a fine grid of model parameters. The inherent degeneracy between these parameters has to be taken into account when fitting the observables. To this end, we recently developed a Bayesian inference tool (BayesClumpy), that extracts as much information as possible from the observations. Details on the interpolation methods and algorithms employed can be found in Asensio Ramos & Ramos Almeida (2009). Thus, for the following analysis of the Seyfert SEDs, we are not using the original set of models described in Nenkova et al. (2008a, 2008b), but an interpolated version of them (see Figures 3 and 4 in Asensio Ramos & Ramos Almeida 2009 for a comparison between the original and interpolated models). In this work, we use the most up-to-date version of the models that are corrected for a mistake in the torus emission calculations, which in principle only affects the AGN scaling factor (see erratum by Nenkova et al. 2010). The present version of BayesClumpy has also been updated to use the Multinest algorithm of Feroz & Hobson (2008) and Feroz et al. (2009) for sampling. This algorithm is very robust and efficient when sampling from complex posterior distributions.

The prior distributions for the model parameters are assumed to be truncated uniform distributions in the intervals reported in Table 7. Therefore, we give the same weight to all the values in each interval. Apart from the six parameters that characterize the models, there is an additional parameter that accounts for the vertical displacement required to match the fluxes of a chosen model to an observed SED, which we allow to vary freely. This vertical shift scales with the AGN bolometric luminosity (see Section 6). In order to compare with the observations, BayesClumpy simulates the effect of the employed filters on the simulated SED by integrating the product of the synthetic SED and the filter transmission curve. Observational errors are assumed to be Gaussian or upper/lower limit detections. A detailed description of the Bayesian inference applied to the clumpy models can be found in Asensio Ramos & Ramos Almeida (2009). In addition, to see an example of the use of clumpy model fitting to IR SEDs using BayesClumpy, see RA09.

5.2.1. Seyfert 1 Individual Fits

The results of the fitting process of the IR SEDs with the interpolated version of the clumpy models of Nenkova et al. (2008a, 2008b) are the posterior distributions for the six free parameters that describe the models and the vertical shift. These are indeed the probability distributions of each parameter, represented as histograms. When the observed data introduce sufficient information into the fit, the resulting posteriors will clearly differ from the input uniform priors, either showing trends or being centered at certain values within the intervals considered. For all the Sy1 fits, we considered uniform priors in the intervals shown in Table 7. The only exception is NGC 7469, for which we use a Gaussian prior of 85° ± 2° for the inclination angle of the torus, based on the value of the accretion disk viewing angle deduced from X-ray observations (Nandra et al. 2007), assuming that the disk and the torus are coplanar.

We fit the individual Sy1 SEDs with BayesClumpy, modeling the torus emission and the direct AGN contribution (the latter as a broken power law). We also consider the IR extinction curve of Chiar & Tielens (2006) to take into account any possible foreground extinction from the host galaxy. The AGN scales self-consistently with the torus flux, and the foreground extinction (separate from the clumpy torus) is another free parameter, which we set as a uniform prior ranging from A_V = 0 to 10 mag.¹⁴

Table 8. Parameters Derived from the Clumpy Model Fitting

Galaxy	σ (°)		Y		N0		q		i (°)		τV		ALOSV (mag)
	Median	Mode	Median	Mode	Median	Mode	Median	Mode	Median	Mode	Median	Mode	Median	Mode
NGC 1097	29±108	28	19±68	23	<2	1	2.7±0.20.3	2.9	36±1719	43	42±6910	36	<15	1
NGC 1566	46±1215	52	21 ± 4	20	2 ± 1	2	0.2 ± 0.2	0.1	79±715	86	104±2855	141	165±95100	160
NGC 6221	42±1512	36	21±56	24	8±43	7	0.8±0.70.5	0.4	59±1833	79	110±2332	128	<690	1
NGC 6814	23±135	20	18 ± 7	23	2 ± 1	2	1.1±0.60.5	1.1	43±2725	39	113±2230	139	<85	1
NGC 7469	41±176	38	22±56	27	3 ± 1	3	1.9 ± 0.4	1.9	84 ± 2 (fix)	84	141±615	148	500 ± 120	430
NGC 3227	36±1812	33	19 ± 6	19	5±43	2	0.6±0.50.4	0.5	49±2126	66	118±1827	129	<240	1
NGC 4151	24±176	19	16±87	8	2±21	2	1.8 ± 0.6	1.8	43±1826	57	110±2326	123	<60	1

Notes. Medians and modes of the Sy1 probability distributions. Those presenting single tails have been characterized with the mode and upper/lower limit at 68% confidence. The models include the intrinsic AGN continuum emission.

Download table as:  ASCIITypeset image

In addition to the Gemini MIR unresolved fluxes reported in Table 3, we consider MIR nuclear fluxes from VISIR compiled from the literature when available (see Table A1 in Appendix A). We find good agreement between the VISIR and T-ReCS unresolved fluxes. We did not include measurements in the VISIR PAH filters (8.59, 11.25, and 11.88 μm) for the galaxies NGC 7469 and NGC 1097 because of their intense star formation and their already well-sampled SED.

Although the solutions to the Bayesian inference problem are the probability distributions of each parameter, we can translate the results into corresponding spectra (Figure 6). The solid lines correspond to the model described by the combination of parameters that maximizes their probability distributions (maximum-a-posteriori, MAP). Dashed lines represent the model computed with the median value of the probability distribution of each parameter. Shaded regions indicate the range of models compatible with the 68% confidence interval for each parameter around the median. In Figure 7, we show the posteriors of the six torus parameters (the vertical shift and the foreground extinction have been marginalized) for the galaxy NGC 1097. Those for the rest of the Sy1 galaxies are presented in Appendix A (Figures A1–A6).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The more information the IR SEDs provide, the better the probability distributions are constrained. From the analysis performed here and in RA09 it appears important to sample the SED in the wavelength range around 3–4 μm and also at ∼18 μm to constrain the model parameters. A detailed study of the influence of different filters/wavelengths in the restriction of the clumpy models parameter space will be the subject of a forthcoming paper (A. Asensio Ramos et al. 2011, in preparation). The posterior information for the seven Sy1 galaxies is summarized in Table 8.

From the individual fits of the Sy1 galaxies in our sample with clumpy torus models, we obtain the following results.

• 1.  
  The average number of clouds along an equatorial ray is within the interval N₀ = [1, 8].
• 2.  
  Low values of σ are preferred: σ≃ [25°, 45°], and intermediate inclination angles of the torus are found: i≃ [35°, 85°].
• 3.  
  The radial extent of the torus (Y = R_o/R_d) is weakly constrained within the interval Y≃ [15,20].
• 4.  
  Values in the range from τ_V≃ [40,140] are found for the optical depth of each cloud for all the galaxies.
• 5.  
  The radial density profile appears constrained within the interval q = [0.2,1.9], with the only exception of NGC 1097, for which q = 2.7 ± ^0.2_0.3.
• 6.  
  The 10 μm silicate feature appears in shallow emission or absent in the fitted models with the exception of NGC 6221 MAP model. The weak silicate feature arises in the clumpy models because both illuminated and dark cloud sides contribute to the observed spectrum (see Section 5.4 in RA09 for a detailed discussion on the silicate feature modeling).
• 7.  
  The optical extinction produced by the torus along the LOS results in A^LOS_V < 690 mag for all the galaxies.
• 8.  
  The foreground extinction from the host galaxy, which obscures the AGN direct emission in our modeling, results in A_V < 5 mag (see Figure 6). The values derived are consistent with those published in the literature, e.g., the nuclear optical extinction of A_V = 3 mag measured from the optical spectrum of NGC 6221 (Levenson et al. 2001), A_V ∼ 1 mag determined from NACO/VLT color maps (Prieto et al. 2005) for NGC 1097, and A_V = 4.5–4.9 mag reported by Mundell et al. (1995) for NGC 3227.All the above intervals or limits of the parameters correspond to median values. We chose the medians instead of the MAPs because the former gives a less biased information about the result, since it takes into account degeneracies, while the MAP does not.

The clumpy models successfully reproduce the observed Sy1 SEDs studied here with compatible results among them. This is indicating that the NIR and MIR unresolved fluxes employed here are dominated by a combination of reprocessed emission from dust in the torus and direct AGN emission.

Our modeling results for NGC 1097 somehow contradict those presented in Mason et al. (2007). The latter authors unsuccessfully tried to reproduce the T-ReCS 11.66 and 18.3 μm aperture fluxes that they measured for this galaxy using the clumpy models of Nenkova et al. (2002). The difference with our result is probably due to (1) the fact that we obtained unresolved fluxes using PSF subtraction over the same images presented in Mason et al. (2007), reducing the potential contamination from star formation; (2) they did not consider NIR data, but only the MIR fluxes; and finally (3) they faced the degeneracy problem of the clumpy models without using any sophisticated tool as, e.g., BayesClumpy.

5.2.2. Sy2 and Intermediate-type Seyfert Results

As reported in Section 4.1, Type-1 Seyferts present characteristically higher H/N ratios and flatter SEDs than those of Sy2 nuclei. The enhancement on the NIR emission of Type-1 AGN is produced by the hot dust from the directly illuminated faces of the clumps in the torus which are close to the central engine, and also by direct AGN emission (i.e., the tail of the optical/ultraviolet power-law continuum), which strongly flattens their IR SEDs (Rieke & Lebofsky 1981; Barvainis 1987).

Based on the IR SEDs presented here, the relative NIR contribution to the SED is generally correlated with the Seyfert type. Sy2 galaxies show lower NIR to MIR ratios (H/N = 0.003 ± 0.002) than Sy1 (0.06 ± 0.03).

In the context of the clumpy models, the presence of a cloud along the LOS, which may occur from any viewing angle, results in a Type-2 classification. Cloud encounters are more probable at large inclination angles (i), but there is always a finite probability of having an unobscured view of the AGN. In fact, the likelihood of intercepting a cloud along a LOS depends on the combination of i, N₀, and σ. Thus, the preference for lower values of these parameters (especially N₀ and σ) in Type-1 Seyferts increases the likelihood of unimpeded views of some directly illuminated cloud faces (i.e., those on the "back" side of the torus) and direct AGN emission, resulting in an increase of the NIR flux. The latter can be represented in terms of the escape probability P_esc (see Equation (4) in Nenkova et al. 2008a). For a total number of clouds N_LOS along a path, P_esc ≃ exp(−N_LOS) when the clouds are optically thick (τ_λ>1).

In Figure 12, we show the dependence of the H/N ratio on the escape probability. All the Sy2 are in the bottom left corner of the diagram, whereas the Sy1 have higher values of the H/N ratio and P_esc = [1%, 92%]. Sy1.8 and Sy1.9 galaxies present intermediate values between those of Sy1 and Sy2. The derived joint posterior distributions of the escape probabilities for Sy1 and Sy2 (KLD = 29 between them) are shown in Figure 11, and the median values of the histograms are P_esc(Sy1) = 18% ± 3% and P_esc(Sy2) = 0.05% ± ^0.08_0.03%.

Zoom In Zoom Out Reset image size

Thus, while for tori with high values of i, N₀, and σ the probability of having a direct view of the AGN is very little, which increases for objects with narrower and less inclined tori, and containing less clumps. In this work, we show for the first time that, in the clumpy torus scenario, the classification as a Type-1 or Type-2 Seyfert depends more on the intrinsic properties of the torus, rather than on the inclination angle itself. The Sy1 galaxies in our sample have larger P_esc than the Type-2 Seyferts, and as a consequence of that, we detect the broad lines in their spectra.

The clumpy model fits yield the intrinsic bolometric luminosity of AGN (L^AGN_bol) by means of the vertical shift applied to match the observational data points. Combining this value with the torus luminosity (L^tor_bol), obtained by integrating the corresponding model torus emission (without the AGN contribution), we derive the reprocessing efficiency (RE) of the torus (L^tor_bol/L^AGN_bol). The previous values are calculated on the Bayesian framework by combining the posterior distributions of the model parameters. Median values and 1σ intervals for Sy1 galaxies are reported in Table 11, and those for Sy2 and intermediate-type Seyferts are shown in Table B3 (Appendix B).

Zoom In Zoom Out Reset image size

Sy2 tori in our sample are more efficient reprocessors than Sy1, absorbing and re-emitting the majority of the intrinsic AGN luminosity in the IR: RE(Sy2) = [0.4, 1.0], with a median value of 0.8 and RE(Sy1) = [0.2,0.7], with median of 0.5. It makes no sense to compare the derived joint Sy1 and Sy2 posterior distributions of RE, because we normalized the SEDs to perform the fits, and consequently the derived L^AGN_bol are meaningless.

We considered a possible dependency of the RE (or alternatively the covering factor, C_T) on L^AGN_bol, since the amount of incoming radiation from the AGN could possibly have some influence on the reprocessed energy or even in the torus properties (e.g., receding torus scenario; Lawrence 1991). However, we find no relationship between the two quantities in the luminosity range considered. This means that the RE depends primarily on the total number of clouds available to absorb the incident radiation, i.e., on the torus covering factor. However, the possible dependence of the torus properties on the AGN luminosity considering a broader luminosity range is further investigated in Alonso-Herrero et al. (2011).

Figure 13 shows that there is a correlation between RE and the torus covering factor for the galaxies in our sample. The larger C_T the more efficient reprocessor. Type-2 tori have in general larger RE than those of Type-1, with the exceptions of Centaurus A and Mrk 573. Despite the relatively small sample, Figure 13 shows a segregation between Sy1 and Sy2 galaxies in terms of RE and covering factor.

If all Seyfert nuclei are identical, as the unified model predicts, only the viewing angle should determine the classification, not the properties of the torus itself. While our results are limited by the small sample analyzed, they suggest instead that the Type-1/Type-2 classification depends on the torus intrinsic properties rather than in the mere torus inclination.

The bolometric luminosity of the intrinsic AGN derived from the fits (L^AGN_bol; column 2 in Table 11) can be directly compared with those from the absorption-corrected 2–10 keV luminosities compiled from the literature (L^AGN_Xbol; Column 6 in Table 11), which is an effective proxy for the AGN bolometric luminosity (Elvis et al. 1994). To obtain L^AGN_Xbol from the intrinsic 2–10 keV values, we applied a bolometric correction factor of 20 (Elvis et al. 1994).

We find similar values of L^AGN_Xbol and L^AGN_bol for all the Sy1 (see Table 11). In the case of NGC 1097, we expect to have some contamination from the nuclear starburst (see Section 5.2.1 and Mason et al. 2007). Thus, if the L^AGN_Xbol value represents the AGN contribution only, L^AGN_bol may be overestimated because of the starburst contribution, thus resulting in a low ratio between the two measurements. The good agreement with the observational X-ray measurements reinforces the results from our SED modeling. The same comparison with X-ray data for Sy2 and intermediate-type Seyferts is shown in Table B3 in Appendix B.

In general, the IR SED fitting does not constrain the size of the torus (Y) as well as other model parameters (see Section 5.3 in RA09). The NIR and MIR observations are sensitive to the warm dust (located within ∼10 pc of the nucleus), which depends on the combination of model parameters N₀, q, and Y (Thompson et al. 2009). FIR observations would be more sensitive to the torus extent independently. However, the fits performed here for Type-1 and Type-2 Seyferts are consistent with a small torus size, confined to scales of less than 6 pc (see below).

Uniform density models require the dusty torus to extend over large dimensions, to provide cool dust that produces the IR emission (e.g., Granato & Danese 1994). In contrast, in a clumpy distribution, different dust temperatures can coexist at the same distance, including cool dust at small radii (Nenkova et al. 2002), so large tori, which are inconsistent with imaging and interferometry results, are not necessary. Indeed, for the Seyfert galaxies considered here, we found Y ranging from 10 to 25 (see Table 8), showing that small tori can account for the observed IR nuclear emission.

The outer size of the torus scales with the AGN bolometric luminosity: R_o = YR_d, so assuming a dust sublimation temperature of 1500 K, R_o = 0.4Y(L^AGN_bol/10⁴⁵)^0.5 pc. We derived R_o posterior distributions from those of L^AGN_bol and Y and find that all tori in our sample have outer radii smaller than 6 pc (Tables 11 and B3), in agreement with MIR direct imaging of nearby Seyferts (Packham et al. 2005; Radomski et al. 2008) and also interferometric observations (Jaffe et al. 2004; Tristram et al. 2007; Meisenheimer et al. 2007; Raban et al. 2009).

For example, the estimated outer radius for NGC 1097 (R_o = 0.4 ± ^0.1_0.2 pc) is in agreement with the value derived from NIR NACO/VLT observations (Prieto et al. 2005), which placed the radius of the central compact source in r < 5 pc. Mason et al. (2007) also derived an upper limit of 19 pc for the size of the unresolved component from the MIR images employed in this work for NGC 1097. For the case of NGC 7469, for which we derive R_o = 5 ± ¹₂ pc, Tristram et al. (2009) reported an estimation of the size of the dust distribution of 10 pc, based on MIDI/VLT interferometric observations, although compromised by the high level of noise and the lack of fringes.

Here we include the posterior distributions of the Sy1 galaxies NGC 1566, NGC 6221, NGC 6814, NGC 7469, NGC 3227, and NGC 4151 (Figures A1–A6).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Here, we include the results of the fits of the intermediate-type Seyferts and the Sy2 galaxies in RA09 (Figures B1 and B2) using the data reported in Table B1 and the most up-to-date version of the Nenkova et al. (2008a, 2008b) models (see erratum Nenkova et al. 2010). The resulting model parameters for the individual galaxies are shown in Tables B2 and B3. In Figure B3, we show the posterior distributions from the fit of Centaurus A (see the electronic edition of the journal for the rest of the Sy2 and intermediate-type Seyfert posteriors).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View extended figure (14 panels)

As mentioned in Section 5.2.2, the results presented in this appendix are compatible in general with those presented in RA09 at the 1σ level. For the majority of the galaxies, we have included IR data from recent publications. In particular, the use of VISIR MIR data have been key in predicting the silicate feature in absorption in the case of Centaurus A²⁰ and NGC 3281, and in emission in NGC 3227.

In the case of Centaurus A, the fit presented in RA09 predicted the silicate feature in emission, in clear contradiction with MIR spectroscopic data (Siebenmorgen et al. 2004; Meisenheimer et al. 2007). By including the VISIR data for this galaxy, the fitted models predict a broad silicate feature in agreement with the observations. In addition, we have also introduced the inclination angle of the torus as a Gaussian prior centered in 70° as inferred from the inclination of the inner radio jet of Centaurus A (Tingay et al. 1998). This value is also coincident with the 66° inclination angle of the obscuring structure derived from MIR interferometric observations (Burtscher et al. 2010).

For the Sy2 galaxy NGC 3281, we also have considered a uniform prior in the range [60°, 90°] for the inclination angle of the torus, based on the inclination of the ionization cone axis derived in Storchi-Bergmann et al. (1992) from optical imaging and spectroscopic observations, which implies a nearly edge-on orientation of the obscuring structure.