DUST EMISSION FROM A PARSEC-SCALE STRUCTURE IN THE SEYFERT 1 NUCLEUS OF NGC 4151

The resulting single-dish spectrum with an effective aperture of ∼300 mas is shown in Figure 1 together with a Spitzer IRS spectrum for this source (observed 2004 April 8; aperture ∼35; Weedman et al. 2005). The higher Spitzer flux most likely indicates emission from the narrow-line region (see also Radomski et al. 2003).

Zoom In Zoom Out Reset image size

The color temperature of the spectrum is (285⁺²⁵₋₅₀) K. Additionally, we detect the [Ne ii] 12.81 μm line commonly seen in star-forming regions. The [S iv] (10.51 μm) line, clearly seen in the Spitzer spectrum, is not significant in the MIDI spectrum. The errors in the MIDI spectrum arise from incomplete thermal background subtraction and hence rise steeply with increasing wavelength.

The two correlated flux spectra are shown in Figure 2.

Zoom In Zoom Out Reset image size

The correlated flux observed with BL = 61m rises from (0.13 ± 0.01) Jy at 8.5 μm to (0.43 ± 0.05) Jy at 12.5 μm. The second spectrum (BL = 89 m) rises from (0.12 ± 0.02) Jy at 8.5 μm to (0.30 ± 0.05) Jy at 12.5 μm. Correlated flux uncertainties arise primarily from background photon noise and increase with wavelength but are smaller than the single-dish errors.

The correlated flux on the shorter baseline (the one that has a higher flux) shows a broad "bump" between 9 μm and 12 μm that we interpret as a silicate emission feature (see Section 3.3). No [Ne ii] emission is seen, indicating that this arises on a scale that is fully resolved by the interferometer (i.e., ≳20 mas ∼1.3 pc). The fact that the correlated flux is lower on the longer baseline is a clear sign that the source is resolved by the interferometer.

Since the resolution of the interferometer θ_min ∼ λ/3BL changes with wavelength, the correlated flux reflects both the source spectrum and the source structure. It is not possible to draw any conclusions from the spectral slope of a correlated flux spectrum without assuming a source geometry.

Although the silicate absorption feature has often been detected in type 2 nuclei, the emission feature, predicted for type 1 nuclei by torus models (e.g., Pier & Krolik 1992; Schartmann et al. 2005), has not been detected except in a handful of objects, most of them quasars (Hao et al. 2005; Weedman et al. 2005; Buchanan et al. 2006).

As noted by Weedman et al. (2005) and Buchanan et al. (2006), Spitzer spectra (with an aperture of ∼35) of NGC 4151 show weak excess emission at 10 μm and 18 μm that is most easily seen when plotted as ν²F_ν with a sufficiently large wavelength range. In Figure 3, we plotted our spectra (which are limited to the atmospheric N band) in such a way together with a Spitzer spectrum.

Zoom In Zoom Out Reset image size

The hypothesis of silicate emission is not inconsistent with our single-dish spectrum (aperture ∼03) but since this spectrum suffers from incomplete background subtraction, especially at longer wavelengths, it is probably hidden in the resulting uncertainties. The emission feature seems to be most prominent in our observations on the 61 m baseline observation and is clearly not detected on the 89 m baseline.

This suggests that silicate emission is distributed over ∼1 pc as derived from the 10.5 μm Gaussian model (see Section 3.4).

With the limited baselines available from Paranal, it is not possible to reconstruct an image of NGC 4151 from our data. Instead we consider simple model distributions of the emission on the sky and compare the predicted interferometric and single-dish spectra with our measurements in order to fix parameters in such a model. We chose a model containing an unresolved point source (flux F_p) and an extended Gaussian distribution (flux normalization F_g, FWHM θ). Although we might expect the mid-IR brightness distribution in Sy 1 galaxies to have a hole in the middle, neither observed nor modeled fluxes change as long as the hole, whose radius is determined by the sublimation radius of the dust, is unresolved. This is certainly the case in NGC 4151 (see Section 4). An upper limit to the size of our point source is given by the effective resolution of the interferometer; this corresponds to a diameter of ∼λ/3BL ∼ 7 mas, i.e., a radius of ∼0.2 pc, at 10.3 μm.

Because of the east–west baseline orientation, the north–south distribution of emission is undetermined. Equivalently, we assume the source to be circularly symmetric on the sky.

The correlated flux density of the model at wavelength λ is then given by F_ν(λ) =  F_p(λ) +  F_g(λ) ·  exp(−(BL/λ · π/2 · θ)²/ln(2)). With the given data points, this model is uniquely determined. We calculated the parameters of such a model separately at 8.5, 10.3, and 12.5 μm where we are safely away from the regions of very low S/N (at the edges of the N band) and the ozone feature. At 10.3 μm, the parameter values may be affected by the silicate feature. The modeled visibilities are shown in Figure 4 and the resulting parameters are given in Table 1.

Zoom In Zoom Out Reset image size

The errors of these parameters were estimated from a Monte Carlo simulation in which we randomly placed 10,000 measurements in a Gaussian distribution around the measured value with the σ as determined from our data reduction. The parameter errors are then given by the standard deviations of the resulting model solutions to the simulated data.

In the strictest version of unified models, we expect for both a Sy 1 and a Sy 2 galaxy an extended dust structure with the same size, morphology, and temperature distribution (at a given UV luminosity, L_UV). In less strict versions, this is only true statistically (e.g., Elitzur & Shlosman 2006). Additionally, Sy 1 galaxies should have an unobscured point source (the unresolved accretion disk and inner rim of the torus)—but the relative strength of these two components in the mid-IR is model-dependent.

To test this, we can compare our observations with other MIDI observations: in the Circinus galaxy (L_UV ∼ 4 × 10³⁶ W, L_torus ∼ 5 × 10³⁵ W), Tristram et al. (2007) found a warm (T ∼ 330 K) disk with major axis FWHM ∼0.4 pc and a larger, similarly warm, component of ∼2 pc FWHM.

In NGC 1068 (L_UV ∼ 3 × 10³⁷ W, L_torus ∼ 10³⁷ W), Raban et al. (2009) found a hot (T ∼ 800 K) disk of 1.35 × 0.45 pc and a warm component 3 × 4 pc in FWHM. They identified the disks with the densest parts of the torus of the unified model.

For NGC 4151 (L_UV ∼ 1.5 × 10³⁶ W—variable (from NED), L_torus ∼ 4πD²νF_g ∼ 6 × 10³⁵ W), we determined a torus size (FWHM) of ∼(2.0 ± 0.4) pc and a dust temperature of (285⁺²⁵₋₅₀) K.

When scaled to the accretion disk luminosity L_UV, these values agree well with the torus sizes (r ∼ L_UV^1/2) and torus luminosities (L_torus ∼ L_UV) of the two Seyfert 2 galaxies and the temperatures are also very similar. Note, however, that the torus luminosity used here is calculated from the 12 μm flux density, not taking into account emission at longer wavelengths.

In addition to the single-wavelength Gaussian models discussed above, we can also connect the various wavelengths together by constructing graybody models where we assume a smooth power-law dependence of temperature with radius: T(r) ∝ r^−α, and an essentially constant emissivity with radius and wavelength. With such models, we get acceptable fits (reduced χ² ∼ 1) to both the total and the correlated flux spectra with T(1 pc) ∼ 250 K and 0.35 ≲ α ≲ 0.6. These values of α are consistent with dust receiving direct radiation from a central source (e.g., Barvainis 1987), i.e., an optically thin medium with optically thick clumps in it.

These models can be extrapolated to shorter wavelengths to check their consistency with the K-band measurements. All plausible extrapolations of the MIDI data yield K-band fluxes of <10 mJy, much less than observed by Swain et al. (2003) and Riffel et al. (2009) (see below). On the basis of these graybody models, we therefore conclude that the K-band emission arises from structures which can probably not be extrapolated from the larger structure seen in the N band.

From the graybody models one can further infer an emissivity ∼ 10⁻¹—similar to what has been seen in NGC 1068 and Circinus.

To conclude, the resolved nuclear mid-IR structure in NGC 4151 has a size, temperature, and emissivity that is comparable to those in type 2 objects where the existence of clumpy tori is established. Apart from the similarities with type 2 tori and the temperature profile of the graybody models, there is further evidence for clumpiness in the NGC 4151 torus from radio observations: Mundell et al. (2003) measured H i absorption against the radio jet (P.A. ∼ 77°) and found a structure of ∼3 pc in size. From the velocities they further suggest that the gas is distributed in clumps. Warm dust possibly traces the H i gas in the mid-IR.

It therefore seems reasonable to identify the warm dust structure resolved now in NGC 4151 with the clumpy tori seen in Sy 2 galaxies. Due to the limited observing geometry and the limited amount of observations, we cannot reconstruct its apparent shape nor can we constrain a model with more than the two components discussed above.

K-band interferometry measurements by Swain et al. (2003) and, with higher significance, more recent Keck observations (J.-U. Pott et al. 2009, in preparation) revealed a marginally resolved source, compatible with hot dust at the sublimation radius of ∼0.05 pc. Reverberation measurements by Minezaki et al. (2004) and Koshida et al. (2009) find lag times Δt between the UV/optical continuum and the K-band corresponding to a radius of ∼0.04 pc (variable) which they interpret as the sublimation radius of dust.

Riffel et al. (2009) measured a flux of ∼65 mJy at 2.2 μm and from spectral fitting they found that, at 2.2 μm, this flux is entirely dominated by a blackbody with a temperature of (1285 ±  50) K—again consistent with hot dust at the sublimation radius and the K-band interferometry, taking into account that L_UV is variable by at least a factor of 10 (Ulrich 2000). To account for that variability when comparing our observations with these K-band observations of different epochs, we looked at the X-ray flux as a proxy for the UV–optical radiation⁷ and find that, at the date of our observations, the source was probably in a higher state than at the time of Riffel et al.'s (2009) observations and emitted ∼0.2 Jy in the K band.

Since the flux density F_ν of a ∼1285 K blackbody is roughly the same at 2.2 μm as at 8.5 μm, we can compare the flux in the K band (∼0.2 Jy at the time of our measurement) with our point source flux at 8.5 μm (∼0.119 ± 0.016 Jy). From this it seems likely that hot dust is contributing to our point source at 8.5 μm. The spectrum of the point source rises by more than a factor of 2 from 8.5 μm to 12.5 μm, however, and thus requires emission from an additional small, "red" component. This could be core synchrotron emission, although such sources usually have flat spectra, or emission from a small, cool, optically thick central dust structure, possibly shadowed from direct accretion disk radiation.