Gemini Mid-IR Polarimetry of NGC1068: Polarized Structures Around the Nucleus

TL;DR

This study uses high-resolution mid-infrared polarimetry to reveal the structure and polarization mechanisms around the nucleus of NGC1068, providing new insights into the AGN's dusty environment.

Contribution

First mid-IR polarimetric imaging of NGC1068 with an 8-m class telescope, identifying distinct polarized regions and linking the torus to the host galaxy environment.

Findings

Identified three polarized regions around the nucleus.

Detected a compact torus consistent with polarization minimum.

Provided the first mid-IR polarimetry data from an 8-m class telescope.

Abstract

We present diffraction limited, 10um imaging polarimetry data for the central regions of the archetypal Seyfert AGN, NGC1068. The position angle of polarization is consistent with three dominant polarizing mechanisms. We identify three distinct regions of polarization: (a) north of the nucleus, arising from aligned dust in the NLR, (b) south, east and west of the nucleus, consistent with dust being channeled toward the central engine and (c) a central minimum of polarization consistent with a compact (<22pc) torus. These observations provide continuity between the geometrically and optically thick torus and the host galaxy's nuclear environments. These images represent the first published mid-IR polarimetry from an 8-m class telescope and illustrate the potential of such observations.

Figures

Tables

Table 1. Table 1: Polarization component summary

Locale, Aperture, $\lambda$	Degree of Polarization	PA of Polarization	Emitting Component
Nucleus, 1.7$″$x1.2$″$	2.48$\pm$0.57%	26.7$\pm$15.3$°$	Several
9.7$\mathrm{\mu m}$
North region	$\sim$2%	$\sim$8$°$	Dichroic emission from dust
9.7$\mathrm{\mu m}$			aligned through jet interaction
East, West, South	$\sim$3.5%	$\sim$35$°$	Dichroic emission from
regions, 9.7$\mathrm{\mu m}$			galactic dust or torus outer edge
Innermost region	$\le$1%	-	Dichroic absorption or
9.7$\mathrm{\mu m}$			unpolarized source
Nucleus, 2$″$	1.3$\pm$0.05%	49$\pm$3$°$	Several, including dichroic
10$\mathrm{\mu m}$			emission
Nucleus, 2$″$	4.11$\pm$0.46%	120.6$\pm$2.38$°$	Several, including dichroic
2.2$\mathrm{\mu m}$			absorption

Full text

galaxies: nuclei — galaxies: Seyfert — galaxies: structure galaxies: individual(NGC 1068) — infrared: galaxies — polarization: galaxies

1 Introduction

The unified theory of Seyfert (Sy) type active galactic nuclei (AGN) holds that all types of Sy AGN are essentially the same object, viewed from different lines of sight (LOS). Surrounding the central engine is a geometrically and optically thick, dusty, molecular torus, obscuring the broad emission line region from some LOS. In this scheme, the Sy classification depends solely on the LOS and exact torus properties. Such theories received a major boost through the detection of scattered, and hence polarized, broad emission lines in the spectrum of NGC1068 (Antonucci & Miller, 1985), revealing an obscured Sy 1 central engine in the previously classified Sy 2 AGN, entirely consistent with unified theories.

Whilst fundamental to unified theories, the torus remains difficult to image directly at optical/IR wavelengths, with perhaps the most direct observation of the torus made by speckle interferometry in the near-IR (Weigelt et al., 2004). Strong evidence for significant amounts of obscuring material in the central 100pc-scale nuclear regions of NGC1068, possibly in the form of a torus, is provided by observations of CO and HCN emission (Planesas et al., 1991; Jackson et al., 1993; Schinnerer et al., 2000), and recent Chandra X-ray observations (e.g. Ogle et al., 2003). Mid-IR spectroscopy reveals a moderately deep ($\tau_{9.7}\approx 0.4$) silicate absorption feature at the nucleus (Roche et al., 1984; Siebenmorgen et al., 2004), whose strength is approximately constant up to $\sim$1″ south of the brightest mid-IR point (Mason et al., 2006; Rhee & Larkin, 2006). Applying the Nenkova et al. (2002) clumpy torus model, Mason et al. (2006) suggested the torus is compact ($\leq 15$pc), in good agreement with mid-IR interferometric observations (Jaffe et al., 2004). Further evidence for a compact torus was found through AO fed H2 1-0S(1) observations (Davies et al., 2006), finding a 15pc clump of H2 extending from the nucleus at the same PA as the line of masers. The observed extent of the torus, or nuclear obscuring material in NGC1068, is partly dependent on the wavelength and/or observational technique. Young et al. (1996) used imaging polarimetry to observe the silhouette of obscuring material against the southern ionization cone, which they attributed to the torus, with a derived diameter of $\sim$200pc in the H-band.

The close proximity (1$\arcsec$ $\equiv$ 72pc) and high brightness of NGC1068 makes it an ideal target for polarimetry, a traditionally photon-starved application. Near-IR studies of NGC1068 by Packham et al. (1997), Lumsden et al. (1999) and Simpson et al. (2002) clearly detected the bi-conical ionization structure in scattered light. In the nuclear regions, there is a trend to a position angle (PA) of polarization being perpendicular to the radio jet with increasing wavelength. Modeling of the nuclear regions requires both an extended area of scattering particles as well as dichroic absorption of nuclear emission, possibly by dust in, or associated with, the torus (Young et al. (1995), Watanabe et al. (2003)).

Bailey et al. (1988) found that the PA of polarization rotates by $\sim$70$\arcdeg$ between 4 and 5$\mu$m. The 10$\mu$m spectropolarimetry of Aitken et al. (1984) showed a similar PA of polarization to that at 5$\mu$m, and a constant degree of polarization through the silicate absorption feature. These results are entirely consistent with the predicted 90$\arcdeg$ change from dichroic absorption to dichroic emission from aligned dust grains. That the PA change was only $\sim$70$\arcdeg$ is attributable to dilution of the dichroic emission component by polarized flux in the extended scattering cones, most likely from dichroic emission from dust in the narrow emission line region (NLR) (Bailey et al., 1988).

To investigate the contributions of the various polarizing mechanisms and structures in the nucleus of NGC1068, Lumsden et al. (1999) performed imaging polarimetry using a broad-band 8-13 $\mu$m filter. These data represented the first and only published mid-IR imaging polarimetry of an AGN, but their interpretation was complicated by the $\sim$0.7″ resolution of the data. To take advantage of the improved spatial resolution attainable from an 8 m-class telescope, we obtained new mid-IR imaging polarimetry of NGC1068 during commissioning of this mode at the Gemini North telescope.

2 Observations

We obtained imaging polarimetry of NGC1068 during commissioning of the polarimetry unit of Michelle (Glasse et al., 1997) on UT 2005 December 19 on the Gemini North 8.1m telescope. These observations were primarily aimed at measuring the degree and position angle (PA) of polarization with NGC 1068 as a test object, and hence used a limited on-source time of 148 seconds. Michelle uses a Raytheon 320 x 240 pixel Si:As IBC array, providing a plate scale of 0.1$\arcsec$ per pixel in imaging mode. Images were obtained in the 9.7$\mu$m (\delta$$\lambda = 1.0$\mu$m, 50$\%$ cut-on/off) filter only, using the standard chop-nod technique to remove time-variable sky background, telescope thermal emission and so-called “1/f” detector noise. The chop throw was 15$\arcsec$ and the telescope was nodded every $\sim$90 s. The chop throw was fixed at 0$\arcdeg$ (N-S). Conditions were photometric and the observations were diffraction limited ($\sim$0.30″FWHM).

Michelle employs a warm, rotatable half wave retarder (or half wave plate, HWP) to modulate the polarization signal, located upstream of the entrance window of the dewar. A cold wire grid polarizer is used as the polarimetric analyzer, located in a collimated beam. Images were taken at four HWP PAs in the following sequence: 0$\arcdeg$, 45$\arcdeg$, 45$\arcdeg$, 0$\arcdeg$, 22.5$\arcdeg$, 67.5$\arcdeg$, 67.5$\arcdeg$, 22.5$\arcdeg$ in the first nod position, and the sequence repeated in the second nod position. In this manner, the Stokes parameters can be computed as close in time as possible, reducing the effects of variations in sky transmission and emission. This sequence, however, requires many motions of the HWP, and is therefore under evaluation with a view to reducing the number of HWP motions to increase observing efficiency. Data were reduced using the Gemini IRAF package in conjunction with Starlink POLPACK software (Berry & Gledhill, 2003). The difference for each chop pair in a given nod position and HWP PA was calculated, and then differenced with the second nod position at the same HWP PA. Images were aligned through shifting by fractional pixel values to account for slight image drift between frames, and then the Stokes parameters I, Q and U computed using POLPACK. A total of 20 nod positions were recorded, and residual array/electronic noise was removed through use of a median-filter noise mask. The data were reduced through creation of four individual I, Q and U maps and also through coadding all frames at their respective HWP PA first and then producing a single I, Q and U map. The S/N in the latter method is slightly higher in the individual Q and U maps, presumably due to a ’smoothing’ of the Q and U during the co-addition; these are the data used in this paper.

The efficiency and zero angle calibration of the polarimeter were measured through observations of two polarized sources and comparison with measurements published by Smith et al. (2000). The instrumental polarization was estimated to be $\leq$0.3% through observations of two stars that fulfilled the criteria of (a) high proper motion (hence nearby), (b) high galactic latitude (to minimize the presence of Galactic dust) and (c) an intermediate spectral type star (to minimize intrinsic stellar nebulosity).

3 Results

Figure 1 shows the total flux image (color-scale and contours) with the polarization vectors overlaid. The polarization vectors are plotted where the S$/$N is $\geq$54 in total flux, and contours are linearly spaced in intensity, starting at a S$/$N of 27. Figure 2 shows the polarized intensity map, produced by multiplying the degree of polarization by the total intensity image. As in Figure 1, only where the S/N in the total intensity image is $\geq$54 are data plotted. The resultant polarization vectors are contained within an approximate N-S oriented ellipse, major/minor axes 1.7$\arcsec$/1.2$\arcsec$ respectively. The integrated degree of polarization within that ellipse is 2.48$\pm$0.57% at a PA of 26.7$\pm$15.3$\arcdeg$. The errors in the degree and PA of polarization are estimated through independent measurements of the four individual polarization maps and computing the standard deviation. It should be noted that the exposure time in those four individual maps was very short, where systematic effects could dominate, and hence the quoted errors may be an overestimation. The degree of polarization is higher than the 1.30% measured by Lumsden et al. (1999) in a 2″ aperture, consistent with an increased observed polarization as often arises with improved spatial resolution. The PA of polarization is significantly different from Lumsden’s measurement of 49$\arcdeg$ in a 2$\arcsec$ aperture. However, our data shows a PA rotation of 94$\arcdeg$ from the Kn band data of Packham et al. (1997) and Lumsden et al. (1999), entirely as expected if the dominant polarizing mechanism changes from dichroic absorption to emission between the two wavebands, as described in §1. We speculate the Lumsden et al. (1999) $\sim$0.68$\arcsec$ results suffered significantly greater contamination in their beam, possibly from surrounding extended polarization, as compared to our $\sim$0.30$\arcsec$ results. Additionally, the wider bandwidth of Lumsden et al. (1999) would have been more affected by the different and competing polarizing components.

The degree and PA of polarization suggests contributions from three components. The first extends $\sim$1$\arcsec$ north of the mid-IR peak and is coincident with the inner regions of the radio jet, with a PA of polarization approximately N-S. The second region extends south, east and west of the nucleus, with a PA of polarization of $\sim$35$\arcdeg$. Finally, the degree of polarization drops to a minimum very close to the mid-IR total flux peak, which we believe is most likely to arise from an unresolved polarization contribution with a PA of polarization approximately orthogonal to that of the more extended emission, leading to a reduction in the measured polarization. The polarized intensity image reveals polarized emission extending north of the mid-IR peak, and two areas of enhanced flux east and west of the mid-IR peak, and a minimum close to the mid-IR total flux peak. Table 1 summarizes the locations and polarization components.

4 Discussion

Polarization at mid-IR wavelengths most likely arises from either dichroic absorption or emission, both due to dust grains with a preferred alignment. The integrated PA of polarization in these data confirms and enhances the interpretation of the near-90$\arcdeg$ PA flip from near- to mid-IR wavelengths, with the factor $\sim$2.5 increase in spatial resolution providing a more accurate and consistent result. Galliano et al. (2003, 2005) suggested, based on spatial coincidence, the [OIII] clouds in the ionization cone are the dominant mid-IR sources away from the compact torus. The polarized flux image shows a similar spatial correspondence with the [OIII], and the PA of polarization north of the nucleus is consistent with the interpretation of Bailey et al. (1988) of dichroic emission in the NLR, possibly through dust alignment via jet streaming or a helical magnetic field associated with the jet. We discount directly observed synchrotron radiation from the radio jet accounting for the polarization, as an extrapolation of the radio emission to the mid-IR provides too little flux. Hence, this data confirms the extended mid-IR polarized emission north of the nucleus is dominantly from dust in the ionization cone.

South, east and west of the nucleus, as the PA of polarization is perpendicular to that in the near-IR where the polarization is thought to be produced by dichroic absorption, we suggest the dominant polarizing mechanism is dichroic emission by grains aligned to the same field direction as the absorbing grains, in agreement with other authors (i.e. Bailey et al. (1988)). Due to the Barnett effect (Lazarian (2003), and references therein), grains align with their short axes parallel to the local magnetic field, and the PA of polarization is parallel to the direction of the magnetic field for dichroic absorption and orthogonal for dichroic emission. The location of the polarized emission areas and PA of polarization is suggestive of warm aligned dust grains being channeled from the host galaxy toward the torus. Indeed, the PA of the polarized flux is coincident with the H2 material that Davies et al. (2006) associated with molecular material in a compact torus.

Dichroic absorption in an unresolved optically thick central region could account for the minimum in polarization close to the peak of mid-IR flux, with a PA of polarization approximately orthogonal to the more extended dichroic emission to the east, west and south. Alternatively the mid-IR flux in the innermost regions could arise from a strong mid-IR, intrinsically unpolarized, source. However, there is tentative evidence of the central regions showing a twist in the PA of polarization, tending toward a similar PA found in the dichroic absorption at near-IR wavelengths (e.g. Packham et al. (1997)), which supports the dichroic absorption interpretation. In both possibilities, a compact ($\leq$0.3$\arcsec$ ($\leq$22pc) diameter) torus could account for this result. If correct, the polarization minimum indicates the true position of the central engine, which is not coincident with the mid-IR total flux peak, but displaced by $\sim$0.2$\arcsec$ to the west.

CO (Schinnerer et al., 2000) and optical HST (Catchpole & Boksenberg, 1997) observations are interpreted as evidence of a warped molecular disk on 100pc scales, partially obscuring the nuclear regions of the host galaxy and ionization cone pointing away from Earth. Indeed, Schinnerer et al. (2000) speculate this material, rather than a compact torus, is responsible for obscuring the AGN. Elitzur & Shlosman (2006) suggest the 100pc molecular structure is the extension of the pc scale disk of masers (Greenhill & Gwinn (1997), Gallimore et al. (2001), Galliano et al. (2003)). We suggest that our data provide continuity between the geometrically thick torus (height/radius $\sim$1) to the flatter, larger galactic disk (height/radius $\sim$0.15).

The western polarized feature is considerably larger than the compact ($\leq$ few pc) torus suggested by several authors (e.g. Jaffe et al., 2004; Mason et al., 2006; Packham et al., 2005; Radomski et al., 2006) on the basis of mid-IR imaging and modeling, but much smaller than the suggested torus detected by Young et al. (1996). However, the feature is detected in polarized flux, a technique that increases contrast by removing the dominant, unpolarized, emission. Distinct from total flux, polarimetric observations are therefore potentially much more sensitive to emission from the putative faint, outer regions of the torus where the interaction with the inner regions of the host galaxy must occur. We suggest that a way to reconcile the evidence for a compact torus with these observations and others, such as extended silicate absorption (Roche et al. (2006), Roche et al. (2007)) and 100 pc-scale CO discs, is that the compact, geometrically and optically thick torus is often surrounded by a larger, more diffuse structure, associated with the dusty central regions of the host galaxy. Where the torus ends and the host galaxy dust structure starts may be more of a question of semantics rather than a true physical boundary. These observations examine the interaction between the host galaxy and possible entrainment into the outer torus regions. Further multiple-wavelength polarimetric observations of both NGC1068 and other AGN are required to test this hypothesis.

We are grateful to the Gemini, UKIRT and ATC science and engineering staff for their outstanding work on Michelle and the Gemini telescope, and wish to note especially Chris Carter. 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: the National Science Foundation (United States), the Particle Physics and Astronomy Research Council (United Kingdom), the National Research Council (Canada), CONICYT (Chile), the Australian Research Council (Australia), CNPq (Brazil) and CONICET (Argentina).