Torus geometry difference between polarized and non-polarized BLR AGN

# On the difference of torus geometry between hidden and non-hidden broad line active galactic nuclei

Kohei Ichikawa11affiliationmark: 1414affiliation: JSPS research fellow (DC1), Japan , Christopher Packham22affiliationmark: , Cristina Ramos Almeida33affiliationmark: 44affiliationmark: 1515affiliation: Marie Curie fellow , Andrés Asensio Ramos33affiliationmark: 44affiliationmark: , Almudena Alonso-Herrero55affiliationmark: 1616affiliation: Visiting Professor, Department of Physics and Astronomy, University of Texas at San Antonio , Omaira González-Martín33affiliationmark: 44affiliationmark: , Enrique Lopez-Rodriguez22affiliationmark: , Yoshihiro Ueda11affiliationmark: , Tanio Díaz-Santos66affiliationmark: 77affiliationmark: , Moshe Elitzur88affiliationmark: , Sebastian F. Hönig99affiliationmark: 1515affiliation: Marie Curie fellow , Masatoshi Imanishi1010affiliationmark: , Nancy A. Levenson1111affiliationmark: , Rachel E. Mason1212affiliationmark: , Eric S. Perlman1313affiliationmark: , and Crystal D. Alsip22affiliationmark: 11affiliationmark: Department of Astronomy, Kyoto University, Kitashirakawa-Oiwake-cho, Sakyo-ku, Kyoto 606-8502, Japan 22affiliationmark: Department of Physics and Astronomy, University of Texas at San Antonio, One UTSA Circle, San Antonio, TX 78249, USA 33affiliationmark: Instituto de Astrofísica de Canarias, C/Vía Láctea, s/n, E-38205 La Laguna, Tenerife, Spain 44affiliationmark: Departamento de Astrofísica, Universidad de La Laguna, E-38205 La Laguna, Tenerife, Spain 55affiliationmark: Instituto de Física de Cantabria, CSIC-Universidad de Cantabria, 39005 Santander, Spain 66affiliationmark: Spitzer Science Center, California Institute of Technology, MS 220-6, Pasadena, CA 91125, USA 77affiliationmark: Nucleo de Astronomia de la Facultad de Ingenieria, Universidad Diego Portales, Av. Ejercito Libertador 441, Santiago, Chile 88affiliationmark: Department of Physics and Astronomy, University of Kentucky, Lexington, KY 40506-0055, USA 99affiliationmark: School of Physics and Astronomy, University of Southampton, Southampton SO17 1BJ, UK 1010affiliationmark: Subaru Telescope, 650 North A’ohoku Place, Hilo, HI 96720, USA 1111affiliationmark: Gemini Observatory, Southern Operations Center, c/o AURA, Casilla 603, La Serena, Chile 1212affiliationmark: Gemini Observatory, Northern Operations Center, 670 N. A’ohoku Place, Hilo, HI 96720, USA 1313affiliationmark: Department of Physics and Space Sciences, 150 W. University Blvd., Florida Institute of Technology, Melbourne, FL 32901, USA
###### Abstract

We present results from the fitting of infrared (IR) spectral energy distributions of 21 active galactic nuclei (AGN) with clumpy torus models. We compiled high spatial resolution ( arcsec) mid-IR -band spectroscopy, -band imaging and nuclear near- and mid-IR photometry from the literature. Combining these nuclear near- and mid-IR observations, far-IR photometry and clumpy torus models, enables us to put constraints on the torus properties and geometry. We divide the sample into three types according to the broad line region (BLR) properties; type-1s, type-2s with scattered or hidden broad line region (HBLR) previously observed, and type-2s without any published HBLR signature (NHBLR). Comparing the torus model parameters gives us the first quantitative torus geometrical view for each subgroup. We find that NHBLR AGN have smaller torus opening angles and larger covering factors than those of HBLR AGN. This suggests that the chance to observe scattered (polarized) flux from the BLR in NHBLR could be reduced by the dual effects of (a) less scattering medium due to the reduced scattering volume given the small torus opening angle and (b) the increased torus obscuration between the observer and the scattering region. These effects give a reasonable explanation for the lack of observed HBLR in some type-2 AGN.

galaxies: active — galaxies: nuclei — infrared: galaxies

## 1. Introduction

While active galactic nuclei (AGN) present a variety of observational characteristics, the unified model for AGN proposes the ubiquitous presence of an obscuring torus around their central engines, and that all AGN are fundamentally the same (Antonucci, 1993). This optically and geometrically thick torus produces the effect of a line of sight viewing angle dependency. Type-1 AGN are observed with the direct view of fast moving material close to the supermassive black hole (SMBH), resulting in broad emission lines in their spectra, while type-2 AGN are observed from an edge-on view and the torus blocks the broad emission line region (BLR) component from our line of sight. The most compelling evidence of the unified model was the detection of polarized broad emission lines (PBLs) in type-2 AGN (e.g., Antonucci & Miller, 1985). Further evidence supporting the unified model comes from infrared (IR) observations of several type-2 AGN showing the existence of obscured/hidden broad line regions (HBLRs) detectable only with dust penetrating infrared observations (e.g., Blanco et al., 1990; Nagar et al., 2002; Reunanen et al., 2003; Ramos Almeida et al., 2008).

Against the fact that the observations generally support the unified model, there is the question of why some, but not all, type-2 AGN do not show any observational signs of PBLs. Tran (2001, 2003) and Moran et al. (2001) found that only 30–50% of type-2 AGN show PBLs. Some studies have advocated that the non-detection of a PBL is due to genuine lack of a BLR (e.g., Tran et al., 2011). Others have suggested that the non-detection is due to obscuration effects, rendering the detection of PBLs as difficult or impossible, even with deep near-IR (NIR) spectro-polarimetric observations (Alexander, 2001). Using a statistically complete IRAS 60 m selected type-2 AGN catalog, Heisler et al. (1997) investigated the relationship between the detectability of PBLs and IR color as an indicator of the torus inclination angle. They showed that only AGN with a low torus inclination angle have high detection rate of PBLs compared to those with high inclinations. This result strongly suggests that PBLs could be obscured when there is an edge-on view through the torus and/or nuclear obscuration in the host galaxies. In addition to the optical spectro-polarimetry, X-ray observations suggest that there is a weak evidence showing different absorption in two types of type-2 AGN. Gu et al. (2001) found that the AGN with PBL have slightly lower column density () than those without PBL. Similarly, Lumsden et al. (2004) showed that the detection rate of PBL decreases as a function of , suggesting that the absorption effect by dusty torus could play a role on the detectability of PBL in AGN.

To understand the role of the obscuration by the torus in type-2 AGN and the detectability of PBLs, knowing the torus geometry and properties is crucial. In recent years much progress has been made toward understanding the torus geometrical structure. Thanks to the improvement of computing power, more physically realistic torus models assuming “clumpy” distributions (called clumpy torus models) have been coded by several authors (Nenkova et al., 2002, 2008a, 2008b; Hönig et al., 2006; Schartmann et al., 2008; Hönig & Kishimoto, 2010; Stalevski et al., 2012). These models readily reproduce high spatial resolution nuclear NIR to mid-IR (MIR) spectral energy distributions (SEDs) and spectra of AGN with a compact torus of  pc radius (e.g., Ramos Almeida et al., 2009, 2011a; Nikutta et al., 2009; Alonso-Herrero et al., 2011; Hönig et al., 2011; Lira et al., 2013). On the other hand, traditional smooth torus models (Pier & Krolik, 1992, 1993; Efstathiou & Rowan-Robinson, 1995) had difficulties to describe the variety of nuclear SEDs of nearby AGN (e.g., Alonso-Herrero et al., 2003; Gandhi et al., 2009; Asmus et al., 2011; Ichikawa et al., 2012a). Still, the true torus morphology remains far from being conclusively determined until future observations resolve the torus. Both smooth and clumpy torus models have degeneracies, and from SED-fitting alone is not possible to choose one or the other (see Feltre et al., 2012, for a comparison between smooth and clumpy torus models).

In this paper, under the assumption that the torus follows a clumpy distribution of dust, which we consider more realistic in principle , we discuss how the precise modeled torus morphology plays a key role in the probability of the detection of PBL by fitting clumpy torus models to our series of IR SEDs. We use 21 high spatial resolution MIR spectra in combination with NIR to far-IR (FIR) photometry, constituting one of the largest compilations of nuclear IR SEDs of AGN in the local universe. These SEDs afforded by 8 m class telescopes minimize contamination of the MIR torus spectra from surrounded diffuse MIR emission from warm dust and/or starformation region emission, and hence we are able to construct the highest fidelity torus SED. Therefore we can probe the detectability of the PBL with the least amount of host galaxy contaminations than ever before. The luminosity dependence of the torus morphology will be discussed in a forthcoming paper (Ichikawa et al. in prep.).

## 2. Observations

### 2.1. The Sample

Our principal motivation in this study is to investigate whether the torus model morphology plays a major role in the chance of PBL detections. To achieve this goal, we compiled the nearby AGN sources from the MIR samples of González-Martín et al. (2013) (21 sources) and Alonso-Herrero et al. (2011) (13 sources) as both samples already compiled the currently available data set of ground-based band spectroscopy. Of the 34 sources, five are in common in both samples, therefore the total number is 29 sources. We further set the criteria for survey inclusion that the objects must have at least one high spatial resolution NIR (1–5 m) measurement, as the NIR bands significantly help to constrain the torus parameters (Ramos Almeida et al., 2014). 22 out of 29 sources fulfilled this criterion. We also removed NGC 1808 from this study as controversy remains to whether it hosts AGN or ultra-luminous X-ray sources in the galactic center due to the low X-ray luminosity  erg s (Scarrott et al., 1993; Jiménez-Bailón et al., 2005). We summarize the properties of the 21 sources in Table 1.

Our sample spans AGN bolometric luminosities taken from the literature () in the range 42.7–45.1 erg s (see Table 1), with a mean value of 44.0 erg s. This value is fairly consistent with that of magnitude-limited Seyfert catalogs (Maiolino & Rieke, 1995; Ho et al., 1997). This suggests that our sample could be representative of AGN and their tori in the local universe, although the sample is not complete.

### 2.2. New Observations

We obtained (Si2 filter; the central wavelength with  m and 50% cutoff range of  m) and (Qa filter;  m and  m) band imaging data of NGC 5135 and NGC 5643, observed by T-ReCS (Program ID GS-2012A-Q-43, PI: Nancy Levenson). The standard MIR chop-nod technique was performed for the observations. The data reduction was made by using RedCan (González-Martín et al., 2013).

### 2.3. Published Data from the Literature

We collected the estimated values of the nuclear NIR to MIR emission when available. We compiled nuclear NIR data from both ground- and space-based telescopes such as VLT/NACO and HST/NICMOS. The only exception is the galaxy NGC 5728, whose only NIR flux is from 2MASS (Peng et al., 2006), which we use as an upper limit. Information about the NIR fluxes used here is compiled in Table 2 (columns 2-6).

In order to further reduce the parameters space at longer wavelengths, we used the Spitzer/IRS 30 m continuum fluxes reported by Deo et al. (2009) as upper limits in our fits. We do that because the star formation component can be important at 20–30 m, and beyond 30 m completely overwhelms the AGN torus emission in most cases (Netzer et al., 2007) at the spatial resolutions afforded by Spitzer. We only consider the Spitzer 30 m photometry as data points, and also included the Spitzer/IRS spectroscopy when the AGN spectral turnover at 20–30 m is clearly seen in the IRS spectra. This feature suggests that the torus emission is dominant even in the large aperture data from Spitzer (Alonso-Herrero et al., 2012b; Hönig et al., 2014). Only two sources fulfilled the criterion (IC 4329A and MCG -5-23-16). For those galaxies, we also collected available Herschel/PACS data. All the FIR flux information is tabulated in column 9 to 11 in Table 2.

The errors were estimated using the prescription given by Alonso-Herrero et al. (2012a). For the NACO AO observations, we used 20% in band and 15% in the band. For the other ground-based observation data, we applied 30% for band, 25% for and band, and 20% for the band. These errors include the photometric error, the background subtraction uncertainty, and the uncertainty from estimating the unresolved flux. band fluxes were always used as upper limits due to the difficulties of estimating the unresolved component. For the NICMOS observations, we used 20% for the band, 20% for the and band. For the and band,we use 15% and 25% errors, respectively.

### 2.4. Subsample

To examine the torus model properties of different AGN populations, we divide the sample into subgroups based on whether or not the source has HBLR signs in previously published observations. We first divide the sample into type-1 and type-2 AGN. Although Seyfert 1.8/1.9 are very ambiguous objects (e.g., see Elitzur et al., 2014, for the details), here we define type-1 as AGN which have at least one broad emission lines in their optical spectra. Therefore, we consider Seyfert 1 to 1.9 as type-1 AGN and Seyfert 2 as type-2 AGN. Next, we divide the type-2 AGN into those with any published polarized BLR detections in the optical and/or in the NIR (HBLR) and those without (non-HBLR; hereafter NHBLR). We use Marin (2014), who compiled almost all the previously published polarization information of nearby AGN. These spectro-polarimetric data are taken from several large surveys including the infrared-selected sample of Heisler et al. (1997), the FIR flux limited sample of Lumsden et al. (2001), the distance-limited sample of Moran et al. (2000, 2002), and the heterogeneous optical- and MIR selected sample of Tran (2001, 2003). Mainly the spectro-polarimetric observations were conducted with small or medium size telescopes (up to 4 m-class), while only NGC 3081 has been confirmed to have HBLR features with the Keck 10 m telescope (Moran et al., 2000). Therefore, we should note that some HBLR AGN could contaminate the subgroup of NHBLR in the cases where the BLR is below the signal-to-noise afforded by the 4 m class telescope observations (see Ramos Almeida et al., in prep.). Some sources have currently no published spectro-polarimetric data, but have clear broad emission lines in NIR wavelengths. These sources are MCG-5-23-16, NGC 2110, and NGC 7582 (Alonso-Herrero et al., 2011). All the references used for dividing the sample into each subgroup are indicated in column 12 in Table 1.

Finally, the sources in this study are categorized into three groups (type-1, HBLR, and NHBLR; see column 6 in Table 1). The sample contains 4 type-1, 10 HBLR, and 7 NHBLR AGN.

## 3. Application of Torus Model

### 3.1. Clumpy Torus Model

We fit the clumpy torus models of Nenkova et al. (2008a), known as CLUMPY, to the data using a Bayesian approach (BayesClumpy; Asensio Ramos & Ramos Almeida 2009). Here we describe the six free CLUMPY model parameters used for the SED fitting and the model set-up, which are listed in Table 3. The torus clumps are distributed in a smooth, rather than sharp, toroidal-shaped boundary of angular width . The inner radius () of the torus is set by the location of the dust at the sublimation temperature ( K). This is computed using the AGN bolometric luminosity

 rin=0.4(Lbol(AGN)1045 erg s−1)0.5 pc. (1)

The torus has a radial extent () defined by , where is the outer radius of the torus. The average number of clouds along the line of sight () at a viewing angle is set as

 NLOS=N0exp[−(90−i)2σ2], (2)

where is the average number of clouds along the radial equatorial ray. allows us to derive the escape probability of photons from the AGN central engines (). In the CLUMPY dust distribution, the classification of type-1 or type-2 AGN depends on whether or not there is a clump along the line of sight, which is a function of the viewing angle of the torus, the number of clumps and the torus width. This is different from smooth torus models, for which the classification of an AGN as type-1 or type-2 is solely determined by the viewing angle. The escape probability of photons passing through the torus at a given viewing angle () can be calculated as

 Pesc∼e−NLOS. (3)

In the CLUMPY model, the radiative transfer equations are solved for each clump and thus the calculations depend on the clump distribution within the torus, the optical depth of each clump, and also its dust composition. Here we assume each clump has the same optical depth (), which is defined at the optical band. The CLUMPY model applies a standard cold oxygen-rich interstellar medium dust, which is called OHMc dust (Ossenkopf et al., 1992). The torus clumps are distributed as a power law with index as a function of radius, .

In addition to these six physical parameters, we add two additional parameters to be fitted or fixed. The first parameter is the foreground extinction (), unrelated to the torus. Some authors demonstrated that some AGN have an extremely deep 9.7 m silicate absorption feature which cannot be reproduced solely by the torus obscuration (Alonso-Herrero et al., 2003; Polletta et al., 2008; Alonso-Herrero et al., 2011; Goulding et al., 2012; González-Martín et al., 2013). They suggested that dust in inclined host galaxies can contribute significantly to the observed SED and silicate feature absorption. 10 out of 21 sources are inclined galaxies with low minor-to-major axis ratios (; see Table 1). Therefore, some portion of the observed SED is accounted from by cool foreground dust extinction. Alonso-Herrero et al. (2011) discussed this issue and concluded that for , the effects of foreground extinction cannot be ignored for reproducing the silicate 9.7 m feature. We gathered available values of foreground from the literature and compiled them in column 10 in Table 1. The other additional parameter accounts for the multiplicative factor that has to be applied to match the fluxes of a given model to an observed SED. Deriving this factor enables us to calculate the model AGN bolometric luminosity (Nenkova et al., 2008b). As shown by Alonso-Herrero et al. (2011), reproduces well the values of for Seyfert galaxies, and therefore, in the following we will refer to as the bolometric luminosities of the sample studied here.

By combining the derived output parameters of the CLUMPY model, we can derive other important torus morphological parameters as the torus outer radius , defined as:

 rout=rinY  pc. (4)

We can also calculate the torus scale height as:

 H=routsinσ  pc. (5)

Finally, we define the “geometrical” torus covering factor, which is unaffected by the viewing angle, and it is defined by integrating the AGN escape probability over all angles (Nenkova et al., 2008a). This can be written as

 CT=1−∫π/20Pesc(β)cos(β)dβ, (6)

where . Considering that our motivation is to characterize the intrinsic torus morphology, the “geometrical” torus covering factor is more relevant here than the apparent covering factor.

### 3.2. BayesClumpy and Modeling Details

The CLUMPY database currently contains more than models. Therefore, when fitting the models to the observations, inherent degeneracies have to be taken into account. We then use the BayesClumpy fitting tool (Asensio Ramos & Ramos Almeida, 2009), as it does a fast synthesis of the CLUMPY SEDs. In the last version of BayesClumpy  the inference over the model parameters can be done either using neural network interpolation or multilinear interpolation in the full database. After running several tests, Ramos Almeida et al. (2014) concluded that the latter interpolation produces more robust results. Therefore, here we use linear interpolation, which results in slight differences in the fitted parameters (within 1 for the majority of the fits) for the 13 galaxies that were modeled by Alonso-Herrero et al. (2011) using the neural network interpolation and subsequently re-fitted in this paper.

BayesClumpy can be used to fit photometry and/or spectra in a Bayesian scheme, carrying out inference over the model parameters for observed SEDs. This way we can specify a-priori information about the model parameters. Here we consider uniform prior distribution in the range of each parameter, as summarized in Table 3. The prior distribution of inclination angle () is fixed from previous observations if available in the literature, following the same approach as in Alonso-Herrero et al. (2011). From the objects in our sample taken from González-Martín et al. (2013), NGC 1386 has two possible inclination angles 65 and 85 (Levenson et al., 2006). Thus, we use a uniform prior in the range 60–90 for this source. For the galaxies taken from Alonso-Herrero et al. (2011), we use the same inclination angle constraints they employed, which are compiled in column 11 in Table 1.

We finally include the AGN direct emission (i.e. a broken power-law) which is defined in Eq. (13) of Nenkova et al. (2008a) to the SED for type-1 AGN, in order to reproduce the flatter slope of the NIR band (see Ramos Almeida et al., 2014, for further details).

## 4. RESULTS and DISCUSSIONS

### 4.1. Infrared SEDs with BayesClumpy Fitting

The results of the fitting process to the IR SEDs are the posterior distributions for the six parameters that describe the model (defined in Table 3), the foreground extinction and the multiplicative factor needed to match the SED fluxes. However, we can also translate the results into corresponding spectra, as shown in Figure 1.

Figure 1 shows the observed SEDs and nuclear MIR spectra (black filled dots) of the galaxies with the best fit results overlaid (blue solid lines), based on the inference done with BayesClumpy. The fitted models correspond to those described by the median of the posterior distribution of each parameter. All the derived torus parameters obtained from BayesClumpy are presented in Table 4.

Some SEDs show lower -band fluxes than those predicted by the fitted model. This effect is prominent when silicate 9.7 m feature is observed in deep absorption. Although this may suggest that the model spectra still have a difficulty to reproduce the 18 m silicate feature, the difference is only within a factor of 3 in the worst case (NGC 3281). See Ramos Almeida et al. (2009) for further discussion on the -band excess.

The SEDs of NGC 5506 and NGC 7172 show NIR excesses compared to the model spectra as shown in Figure 1. NIR interferometric observations (Kishimoto et al., 2009, 2011) and NIR reverberation mapping (Kishimoto et al., 2007; Koshida et al., 2009) of Seyfert galaxies suggest that the AGN torus has much smaller sublimation radius than expected from Eq. (1). Kawaguchi & Mori (2011) also showed in their model that if they add the rim darkening effect of the accretion disk, the torus inner radius naturally connects to the outer disk. Thus, the NIR excess is readily accounted for by this connection. This is also shown in Stalevski et al. (2012) when they apply the rim darkening effect to produce the model SEDs. However, in this study, we did not include the SED of a hot dust component or apply such torus geometry including rim darkening effect for the fitting as this hot dust remains rather unconstrained. Further discussion on the possible origins of NIR excesses can be found in Alonso-Herrero et al. (2011).

### 4.2. General Torus Properties for Whole Sample

In this section, we describe how we obtained the global distribution of the torus model parameters for the whole sample and the three subgroups considered here. To take full advantage of the data employed here, we apply a hierarchical Bayesian approach to derive information about the global distribution of the CLUMPY parameters for a given subgroup. We use a generalized beta distribution as the prior for each parameter (given that they are defined in closed intervals) and learn the hyperparameters of the prior using importance sampling (e.g. Brewer & Elliott 2014). This allows us to derive the posterior distribution for each parameter taking into account all the observed data that belongs to an AGN subgroup.

In Table 5 we report the median parameters of the global posterior distributions for the whole sample, as well as for each subgroup. Note that the global posterior distribution of each parameter can be derived only in the circumstance that the prior distribution is the same for all the sources. This is not the case for the inclination angle because we use the constrained prior distributions for some sources as described in Section 3. For the inclination angle, we derive the median parameters of the individual galaxy fits within the each subgroup.

Based on interferometric observations, Kishimoto et al. (2011) reported a typical torus half-light radius of  pc for local AGN at MIR wavelengths. Our derived torus outer radius  pc is consistent with the interferometry results, although a little larger. This is understood because here we are considering colder dust within the torus than that traced by the MIR interferometry, as we include data at  m in our fits. These colder clumps will be generally located at larger radii, which explain the value of obtained from the global posterior distribution. The torus outer radius including cooler dust would be larger still than the value of  pc. Further studies including Herschel, SOFIA, and/or ALMA observations will help constraining the extent of the FIR-emitting dust. Indeed, some pilot studies already showed the importance of FIR data to trace cooler dust (Ramos Almeida et al., 2011b; García-Burillo et al., 2014).

### 4.3. Distribution of Torus Model Parameters

The key science direction of this paper is to investigate the torus morphology quantitatively by deriving torus parameters for each subgroup and comparing them. The median values of the model parameters fitted to our nuclear IR SEDs are reported in Table 4. The median values of the global posterior distributions of the free parameters , , , , and for each subgroup are compiled in Table 5. As discussed in Section 4.2, for the inclination angle we derive the median values from the individual galaxy fits for each subgroup.

Figure 2 shows the global posterior distributions of each physical parameter for the different subgroups. Black, blue, red, and green histograms show the parameter distribution of all, type-1, HBLR, and NHBLR sources, respectively.

As shown in Figure 2, for some of the parameters, the global distributions are similar for the three subgroups, but others are clearly different. In order to quantify these differences, we follow the same approach as in Ramos Almeida et al. (2011a). They used the Kullback-Leibler divergence (KLD; Kullback & Leibler, 1951) to show that the joint posterior distributions of type-1 and type-2 AGN were quantitatively different. The KLD takes into account the full shape of two posterior distributions to compare them. When the two distributions are identical, the value is and, the larger the KLD value, the more different the two distributions. Ramos Almeida et al. (2011a) concluded that if , the two posteriors can be considered significantly different. We calculate the KLD values for the global distributions of each torus parameter among the three groups. The values are reported in Table 6.

We find significant differences for the parameters , , and between type-1 AGN and HBLR AGN. The differences in these parameters between type-1 and type-2 AGN were already reported in Ramos Almeida et al. (2011a) with larger significance, but based on fits to NIR and MIR photometry only, where spectroscopic data were not included. Besides, they did not consider information from spectropolarimetry data, as we are doing here. Therefore, we confirm the results of Ramos Almeida et al. (2011a) after including  band spectroscopy to the IR photometry, which is crucial to constrain the six torus parameters (Alonso-Herrero et al., 2011; Ramos Almeida et al., 2014). We also find that the parameters , , and of HBLR and NHBLR AGN are significantly different. Considering the average values in Table 5, the tori of NHBLR AGN have larger , larger , and larger than those of HBLR AGN.

There are various possible interpretations for the difference in among the subgroups. One possibility is that the smaller could be due to larger AGN luminosities (receding torus model; Lawrence, 1991; Ricci et al., 2013). However, as shown in Table 5, the differences of median bolometric AGN luminosities among the subgroups are within the uncertainties. Therefore, we consider that the effect of the AGN luminosity is negligible in this study. Another possible interpretation is a selection bias in the optical type-1/type-2 AGN selection. Ramos Almeida et al. (2011a) and Elitzur (2012) discussed that AGN classification would depend on the distribution of the obscuring material; type-1 AGN would be preferentially selected from lower-obscuration AGN, while type-2 AGN (HBLR and NHBLR) from higher-obscuration AGN. This could be partly producing the differences in that we found for type-1 and type-2 AGN, but correcting this effect quantitatively is extremely difficult and beyond the scope of this paper.

### 4.4. Distribution of Covering Factor

As shown in Section 3.1., we can derive physical parameters of the torus model by combining the model parameters. The individual values of these physical parameters are reported in Table 4, and those obtained from the global posterior distribution for each subgroup are shown in Figure 3 and Table 5.

An interesting comparison can be made between the geometrical covering factor of the torus model (; described in equation 6) of the different subgroups and the average column densities derived from X-ray data (). NHBLRs have the largest column densities, with an average value of  cm (i.e. Compton-thick), followed by HBLRs, with  cm, and type-1s, with  cm. Based on hard X-ray (50–200 keV) observations of nearby AGN obtained with INTEGRAL, Ricci et al. (2011) reported differences between the X-ray reflection component of type-1 and type-2 AGN. Type-1 and “lightly obscured” AGN with  cm have the same X-ray reflection component, with reflection amplitude . On the other hand, “mildly obscured” AGN ( cm) show a clearly stronger X-ray reflection component with , suggesting that the central engine of “mildly obscured” AGN would be covered by an X-ray reflection wall. Our results are in good agreement with Ricci et al. (2011) if we consider the and values for each subgroup. The type-1 AGN in our sample fall in the “lightly obscured” AGN in their study, and indeed they show small covering factors (), suggesting small torus X-ray reflection solid angle. The HBLR and NHBLR AGN subgroups would fall in the “mildly obscured” AGN category, and we found large covering factors for them ( and 0.96 respectively), suggesting a larger X-ray reflection component (see also Ricci et al., 2014). Figure 4 shows a schematic illustration of the torus geometrical differences among type-1 (top), HBLR (middle), and NHBLR (bottom).

### 4.5. Torus Model Morphological Differences between HBLR and NHBLR

In this section we focus on the differences between the modeled tori of HBLR and NHBLR. In the case of HBLR, we obtain smaller values than for NHBLR, which is equivalent to larger torus opening angles (). This implies that HBLR objects can have a larger scattering region. (see middle panel of Figure 4). The scattering region (shown schematically as a filled green bar) can be larger due to the larger opening angle of the torus, allowing more photons to be scattered, and hence polarized, from the BLR. We note that the larger opening of the ionization cone, if unresolved, will produce a slightly lower degree of polarization due to (partial) cancellation of those polarization vectors at the edges of the scattering region. However, in our case, the increased amount of scattered photons will significantly increase the polarized flux, with only a very small reduction in the degree of polarization. To confirm this effect, we produced a toy polarization model assuming that the scattering region is a two dimensional biconical structure centered to the central engine. Then, we measure the degree of polarization and polarized flux, and found that the measured polarized flux is larger for HBLR than those for NHBLR, supporting our results. Miller & Goodrich (1990) also find the same results using a 3 dimensional cone and more sophisticated modeling, and more complex assumptions.

In the case of the NHBLR, we obtain larger values than for HBLR. This means that the probability that scattered radiation from the BLR can be blocked is higher than for HBLR objects. This is also in agreement with the larger value of  cm estimated from X-ray observations of the NHBLR objects in our sample.

To summarize, as shown in the bottom panel of Figure 4, the chance to observe scattered (polarized) flux from the BLR is reduced by the double effect of (a) less scattering of the flux from the BLR (due to the reduced scattering area) and (b) more obscuration between the observer and the scattering region. Therefore, the classification of an AGN as HBLR or NHBLR is probabilistic, and it would depend on the intrinsic properties of the torus, in particular of . This could be a reasonable explanation for the lack of a hidden (polarized) BLR in some type-2 objects111 A similar explanation for the lack of HBLR detection in 40% of type-2 objects, based on the distribution of dust within the torus and its inclination being not as simple as predicted by the unified model, was shown in the talk by C. Ramos Almeida at the Polarization & active galactic nuclei workshop held 2012 October 16-17 at the Royal Observatory of Belgium..

However, we note that the classification of the galaxies as HBLR and NHBLR is mainly based on spectropolarimetric observations from 3–4 m telescopes, and some of the NHBLR could be then misclassified. Therefore, further higher sensitivity spectropolarimetry observations of NHBLR AGN with 8 m class telescopes such as Subaru/FOCAS and/or VLT/FORS2 are highly encouraged to search for the HBLR in those AGN (Ramos Almeida et al. in prep.).

### 4.6. Inclination Angle Effect on Detectability of HBLR in Type-2 AGN

The dependence of HBLR detection in type-2 AGN on torus inclination angle is still a matter of debate. This idea arises from the observational trend that the IRAS 60 m to 25 m color ratio () of HBLR AGN is on average while that of NHBLR AGN is (Heisler et al., 1997; Lumsden et al., 2001). Several authors have suggested that this trend is due to the inclination angle of the torus: the cooler and outer dust within the torus blocks the warm and inner hot dust for edge-on views, producing the high , while the warm inner dust can be seen from the face-on views, reducing the value of . Following this idea, type-2 AGN with low would tend to be detected as HBLR AGN due to the more face-on view of the torus and vice versa.

Here we can take advantage of the torus models fitted to our SEDs, which are available for each AGN and shown in Figure 1. The wavelength range covered by the models allows us to calculate color ratios for each source. We have also compiled color ratios from IRAS for comparison, which probe much larger scales than our torus SEDs. Out of 21 sources, we obtained 17 IRAS colors with good quality of fluxes (, which is the highest quality)222 See Beichman et al. (1988) for the definition of FQUAL in the IRAS catalogs. False detections may be included when ..

Figure 5 shows the relationship between the flux ratios obtained from IRAS and those from the torus model SEDs. The averages IRAS flux ratio for type-1, HBLR, and NHBLR AGN are , and , respectively, showing the previously mentioned correlation for HBLR and NHBLR AGN, although with large error bars. However, when we compare the values obtained from the torus model SEDs, which exclude contamination from the host galaxy, we find that they are very similar for the three groups and smaller than the IRAS colors ( = , , and for type-1, HBLR, and NHBLR AGN respectively).

These results show that the differences of IRAS among the three subgroups are not produced from the torus dust. A similar result was reported by Alexander (2001), but using X-ray observations. Although the standard deviations of the average values of the IRAS colors are large for the three subgroups, one possible explanation for the cooler IRAS colors of NHBLR in comparison with those of HBLR AGN could be dust emission from stronger starbursts in their host galaxies. This is in good agreement with previous results showing that highly obscured AGN tend to have higher star formation activity in their host galaxies (Goulding et al., 2012; Ichikawa et al., 2012a, b, 2014; Castro et al., 2014). Therefore, larger obscuration from the torus in NHBLR AGN (as shown in Figure 4) and higher star formation activity in the host galaxy could be somehow coupled. For example, using three-dimensional hydrodynamical simulations, Wada & Norman (2002) showed that starbursts and supernovae within the central 100 pc of host galaxies help lifting up the torus, suggesting that high star formation activity could influence the scale height of the torus. Comparing the nuclear and overall star formation activity of AGN with the torus obscuration is crucial to find out if they are coupled (e.g., Imanishi et al., 2011; Esquej et al., 2014; Ichikawa et al., 2014).

## 5. Conclusions

We constructed 21 infrared torus-dominated SEDs including high spatial resolution NIR and MIR photometry, MIR spectroscopy, and Spitzer and Herschel FIR fluxes. By performing SED fitting using clumpy torus models and a Bayesian approach we derived torus parameters such as the torus covering factor (), the torus inner and outer radius ( and ), and the torus scale height (). We divided the sample into subgroups based on whether or not they are optically type-1, type-2 with observational hidden broad line regions sign (HBLR), and type-2 without any observational broad line region signs (NHBLR). Our results are summarized as follows:

1. Under the assumption of a clumpy distribution of the dust, we obtained a quantitative description of the torus geometry and intrinsic properties. We found that the median torus outer radius for the whole sample  pc is consistent with the results from MIR interferometry observations.

2. We found that the tori of Type-1 AGN have smaller , , , and than those of HBLR and NHBLR. Moreover, the tori of NHBLR are thicker and therefore have higher than those of HBLR. These differences in the torus properties of HBLR and NHBLR AGN would make it more difficult to detect hidden BLR in NHBLR.

3. Combining colors obtained from IRAS photometry and from torus model SEDs, we showed that the low measured fof HBLR using IRAS data are not due to a more face-on inclination of torus, but rather to star formation activity in their host galaxies.

We are grateful for useful comments from the anonymous referee. We also thank C. Ricci, M. Stalevski for useful comments and discussions. K.I. thanks the Department of Physics and Astronomy at University of Texas at San Antonio, where most of the research was conducted. This paper is based on observations obtained at the Gemini Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of the Gemini partnership. This work was partly supported by the Grant-in-Aid for JSPS Fellows for young researchers (K.I.). C.P. acknowledges support from UTSA and NSF (grant number 0904421). C.R.A. is supported by a Marie Curie Intra European Fellowship within the 7th European Community Framework Programme (PIEF-GA-2012-327934). A.A.-H. acknowledges support from the Spanish Plan Nacional de Astronomia y Astrofisica under grant AYA2012-31447.

## References

• Alexander et al. (1999) Alexander, D. M., Hough, J. H., Young, S., et al. 1999, MNRAS, 303, L17
• Alexander (2001) Alexander, D. M. 2001, MNRAS, 320, L15
• Alonso-Herrero et al. (2003) Alonso-Herrero, A., Quillen, A. C., Rieke, G. H., Ivanov, V. D., & Efstathiou, A. 2003, AJ, 126, 81
• Alonso-Herrero et al. (2011) Alonso-Herrero, A., Ramos Almeida, C., Mason, R., et al. 2011, ApJ, 736, 82
• Alonso-Herrero et al. (2012a) Alonso-Herrero, A., Sánchez-Portal, M., Ramos Almeida, C., et al. 2012, MNRAS, 425, 311
• Alonso-Herrero et al. (2012b) Alonso-Herrero, A., Pereira-Santaella, M., Rieke, G. H., & Rigopoulou, D. 2012, ApJ, 744, 2
• Antonucci & Miller (1985) Antonucci, R. R. J., & Miller, J. S. 1985, ApJ, 297, 621
• Antonucci (1993) Antonucci, R. 1993, ARA&A, 31, 473
• Asensio Ramos & Ramos Almeida (2009) Asensio Ramos, A., & Ramos Almeida, C. 2009, ApJ, 696, 2075
• Asmus et al. (2011) Asmus, D., Gandhi, P., Smette, A., Hönig, S. F., & Duschl, W. J. 2011, A&A, 536, A36
• Baumgartner et al. (2013) Baumgartner, W. H., Tueller, J., Markwardt, C. B., et al. 2013, ApJS, 207, 19
• Beichman et al. (1988) Beichman, C. A., Neugebauer, G., Habing, H. J., Clegg, P. E., & Chester, T. J. 1988, Infrared astronomical satellite (IRAS) catalogs and atlases. Volume 1: Explanatory supplement, 1,
• Blanco et al. (1990) Blanco, P. R., Ward, M. J., & Wright, G. S. 1990, MNRAS, 242, 4P
• Brightman & Nandra (2011) Brightman, M., & Nandra, K. 2011, MNRAS, 413, 1206
• Brewer & Elliott (2014) Brewer, B. J., Elliott, T. M. 2014, MNRAS, 439, L31
• Carollo et al. (2002) Carollo, C. M., Stiavelli, M., Seigar, M., de Zeeuw, P. T., & Dejonghe, H. 2002, AJ, 123, 159
• Castro et al. (2014) Castro, A., Miyaji, T., Shirahata, M., et al. 2014, arXiv:1408.3172
• Deo et al. (2009) Deo, R. P., Richards, G. T., Crenshaw, D. M., & Kraemer, S. B. 2009, ApJ, 705, 14
• Efstathiou & Rowan-Robinson (1995) Efstathiou, A., & Rowan-Robinson, M. 1995, MNRAS, 273, 649
• Elitzur (2012) Elitzur, M. 2012, ApJ, 747, L33
• Elitzur et al. (2014) Elitzur, M., Ho, L. C., & Trump, J. R. 2014, MNRAS, 438, 3340
• Elvis et al. (1994) Elvis, M., Wilkes, B. J., McDowell, J. C., et al. 1994, ApJS, 95, 1
• Esquej et al. (2014) Esquej, P., Alonso-Herrero, A., González-Martín, O., et al. 2014, ApJ, 780, 86
• Feltre et al. (2012) Feltre, A., Hatziminaoglou, E., Fritz, J., & Franceschini, A. 2012, MNRAS, 426, 120
• Galliano & Alloin (2008) Galliano, E., & Alloin, D. 2008, A&A, 487, 519
• Gandhi et al. (2009) Gandhi, P., Horst, H., Smette, A., et al. 2009, A&A, 502, 457
• García-Burillo et al. (2014) García-Burillo, S., Combes, F., Usero, A., et al. 2014, A&A, 567, A125
• González-Martín et al. (2013) González-Martín, O., Rodríguez-Espinosa, J. M., Díaz-Santos, T., et al. 2013, A&A, 553, A35
• Goulding et al. (2012) Goulding, A. D., Alexander, D. M., Bauer, F. E., et al. 2012, ApJ, 755, 5
• Gu et al. (2001) Gu, Q., Maiolino, R., & Dultzin-Hacyan, D. 2001, A&A, 366, 765
• Guainazzi et al. (2004) Guainazzi, M., Rodriguez-Pascual, P., Fabian, A. C., Iwasawa, K., & Matt, G. 2004, MNRAS, 355, 297
• Heisler et al. (1997) Heisler, C. A., Lumsden, S. L., & Bailey, J. A. 1997, Nature, 385, 700
• Ho et al. (1997) Ho, L. C., Filippenko, A. V., & Sargent, W. L. W. 1997, ApJS, 112, 315
• Hönig et al. (2006) Hönig, S. F., Beckert, T., Ohnaka, K., & Weigelt, G. 2006, A&A, 452, 459
• Hönig & Kishimoto (2010) Hönig, S. F., & Kishimoto, M. 2010, A&A, 523, A27
• Hönig et al. (2011) Hönig, S. F., Leipski, C., Antonucci, R., & Haas, M. 2011, ApJ, 736, 26
• Hönig et al. (2014) Hönig, S. F., Gandhi, P., Asmus, D., et al. 2014, MNRAS, 438, 647
• Ichikawa et al. (2012a) Ichikawa, K., Ueda, Y., Terashima, Y., et al. 2012, ApJ, 754, 45
• Ichikawa et al. (2012b) Ichikawa, K., Ueda, Y., Terashima, Y., et al. 2012, Torus Workshop, 2012, 109
• Ichikawa et al. (2014) Ichikawa, K., Imanishi, M., Ueda, Y., et al. 2014, ApJ, 794, 139
• Imanishi (2002) Imanishi, M. 2002, ApJ, 569, 44
• Imanishi et al. (2011) Imanishi, M., Ichikawa, K., Takeuchi, T., et al. 2011, PASJ, 63, 447
• Itoh et al. (2008) Itoh, T., Done, C., Makishima, K., et al. 2008, PASJ, 60, 251
• Jiménez-Bailón et al. (2005) Jiménez-Bailón, E., Santos-Lleó, M., Dahlem, M., et al. 2005, A&A, 442, 861
• Kawaguchi & Mori (2010) Kawaguchi, T., & Mori, M. 2010, ApJ, 724, L183
• Kawaguchi & Mori (2011) Kawaguchi, T., & Mori, M. 2011, ApJ, 737, 105
• Kishimoto et al. (2007) Kishimoto, M., Hönig, S. F., Beckert, T., & Weigelt, G. 2007, A&A, 476, 713
• Kishimoto et al. (2009) Kishimoto, M., Hönig, S. F., Antonucci, R., et al. 2009, A&A, 507, L57
• Kishimoto et al. (2011) Kishimoto, M., Hönig, S. F., Antonucci, R., et al. 2011, A&A, 536, 78
• Koshida et al. (2009) Koshida, S., Yoshii, Y., Kobayashi, Y., et al. 2009, ApJ, 700, L109
• Kullback & Leibler (1951) Kullback, S., & Leibler, A. 1951, Ann. Math. Stat., 22, 79
• Lawrence (1991) Lawrence, A. 1991, MNRAS, 252, 586
• Levenson et al. (2006) Levenson, N. A., Heckman, T. M., Krolik, J. H., Weaver, K. A., &Zdotycki, P. T. 2006, ApJ, 648, 111
• Lira et al. (2013) Lira, P., Videla, L., Wu, Y., et al. 2013, ApJ, 764, 159
• Lumsden et al. (2001) Lumsden, S. L., Heisler, C. A., Bailey, J. A., Hough, J. H., & Young, S. 2001, MNRAS, 327, 459
• Lumsden et al. (2004) Lumsden, S. L., Alexander, D. M., & Hough, J. H. 2004, MNRAS, 348, 1451
• Maiolino & Rieke (1995) Maiolino, R., & Rieke, G. H. 1995, ApJ, 454, 95
• Marin (2014) Marin, F. 2014, MNRAS, 441, 551
• Marinucci et al. (2012) Marinucci, A., Bianchi, S., Nicastro, F., Matt, G., & Goulding, A. D. 2012, ApJ, 748, 130
• Meisenheimer et al. (2007) Meisenheimer, K., Tristram, K. R. W., Jaffe, W., et al. 2007, A&A, 471, 453
• Meléndez et al. (2014) Meléndez, M., Mushotzky, R. F., Shimizu, T. T., Barger, A. J., & Cowie, L. L. 2014, ApJ, 794, 152
• Miller & Goodrich (1990) Miller, J. S., & Goodrich, R. W. 1990, ApJ, 355, 456
• Mor & Trakhtenbrot (2011) Mor, R., & Trakhtenbrot, B. 2011, ApJ, 737, L36
• Mor & Netzer (2012) Mor, R., & Netzer, H. 2012, MNRAS, 420, 526
• Moran et al. (2000) Moran, E. C., Barth, A. J., Kay, L. E., & Filippenko, A. V. 2000, ApJ, 540, L73
• Moran et al. (2001) Moran, E. C., Kay, L. E., Davis, M., Filippenko, A. V., & Barth, A. J. 2001, ApJ, 556, L75
• Moran et al. (2002) Moran, E. C., Filippenko, A. V., & Chornock, R. 2002, ApJ, 579, L71
• Nagar et al. (2002) Nagar, N. M., Oliva, E., Marconi, A., Maiolino, R. 2002, A&A, 391, L21
• Nenkova et al. (2002) Nenkova, M., Ivezić, Ž., & Elitzur, M. 2002, ApJ, 570, L9
• Nenkova et al. (2008a) Nenkova, M., Sirocky, M. M., Ivezić, Ž., & Elitzur, M. 2008, ApJ, 685, 147
• Nenkova et al. (2008b) Nenkova, M., Sirocky, M. M., Nikutta, R., Ivezić, Ž., & Elitzur, M. 2008, ApJ, 685, 160
• Nenkova et al. (2010) Nenkova, M., Sirocky, M. M., Nikutta, R., Ivezić, Ž., & Elitzur, M. 2010, ApJ, 723, 1827
• Netzer et al. (2007) Netzer, H., Lira, P., Trakhtenbrot, B., Shemmer, O., & Cury, I. 2007, ApJ, 671, 1256
• Nicastro et al. (2003) Nicastro, F., Martocchia, A., & Matt, G. 2003, ApJ, 589, L13
• Nikutta et al. (2009) Nikutta, R., Elitzur, M., & Lacy, M. 2009, ApJ, 707, 1550
• Ossenkopf et al. (1992) Ossenkopf, V., Henning, T., & Mathis, J. S. 1992, A&A, 261, 567
• Peng et al. (2006) Peng, Z., Gu, Q., Melnick, J., & Zhao, Y. 2006, A&A, 453, 863
• Pier & Krolik (1992) Pier, E. A., & Krolik, J. H. 1992, ApJ, 401, 99
• Pier & Krolik (1993) Pier, E. A., & Krolik, J. H. 1993, ApJ, 418, 673
• Polletta et al. (2008) Polletta, M., Weedman, D., Hönig, S., et al. 2008, ApJ, 675, 960
• Quillen et al. (2001) Quillen, A. C., McDonald, C., Alonso-Herrero, A., et al. 2001, ApJ, 547, 129
• Ramos Almeida et al. (2008) Ramos Almeida, C., Pérez García, A. M., Acosta-Pulido, J. A., González-Martín,O. 2008, ApJ, 680, L17
• Ramos Almeida et al. (2009) Ramos Almeida, C., Levenson, N. A., Rodríguez Espinosa, J. M., et al. 2009, ApJ, 702, 1127
• Ramos Almeida et al. (2011a) Ramos Almeida, C., Levenson, N. A., Alonso-Herrero, A., et al. 2011, ApJ, 731, 92
• Ramos Almeida et al. (2011b) Ramos Almeida, C., Sánchez-Portal, M., Pérez García, A. M., et al. 2011, MNRAS, 417, L46
• Ramos Almeida et al. (2014) Ramos Almeida, C., Alonso-Herrero, A., Levenson, N. A., et al. 2014, MNRAS, 439, 3847
• Reunanen et al. (2003) Reunanen, J., Kotilainen, J. K., & Prieto, M. A. 2003, MNRAS, 343, 192
• Ricci et al. (2011) Ricci, C., Walter, R., Courvoisier, T. J.-L., & Paltani, S. 2011, A&A, 532, A102
• Ricci et al. (2013) Ricci, C., Paltani, S., Awaki, H., et al. 2013, A&A, 553, A29
• Ricci et al. (2014) Ricci, C., Ueda, Y., Ichikawa, K., et al. 2014, A&A, 567, A142
• Scarrott et al. (1993) Scarrott, S. M., Draper, P. W., Stockdale, D. P., & Wolstencroft, R. D. 1993, MNRAS, 264, L7
• Schartmann et al. (2008) Schartmann, M., Meisenheimer, K., Camenzind, M., et al. 2008, A&A, 482, 67
• Schmitt (2001) Schmitt, H. R. 2001, AJ, 122, 2243
• Simpson (1998) Simpson, C. 1998, ApJ, 509, 653
• Stalevski et al. (2012) Stalevski, M., Fritz, J., Baes, M., Nakos, T., Popovic, L.  2012, MNRAS, 240,2756
• Tran (2001) Tran, H. D. 2001, ApJ, 554, L19
• Tran (2003) Tran, H. D. 2003, ApJ, 583, 632
• Tran (2010) Tran, H. D. 2010, ApJ, 711, 1174
• Tran et al. (2011) Tran, H. D., Lyke, J. E., & Mader, J. A. 2011, ApJ, 726, L21
• Tueller et al. (2008) Tueller, J., Mushotzky, R. F., Barthelmy, S., et al. 2008, ApJ, 681, 113
• Véron-Cetty & Véron (2006) Véron-Cetty, M.-P., & Véron, P. 2006, A&A, 455, 773
• Wada & Norman (2002) Wada, K., & Norman, C. A. 2002, ApJ, 566, L21
• Wang & Zhang (2007) Wang, J.-M., & Zhang, E.-P. 2007, ApJ, 660, 1072
You are adding the first comment!
How to quickly get a good reply:
• Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
• Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
• Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters