SOFIA/FORCAST OBSERVATIONS OF WARM DUST IN S106: A FRAGMENTED ENVIRONMENT

S106 was observed with Spitzer/IRAC (Fazio et al. 2004; Werner et al. 2004) as part of the IRAC Guaranteed Time Observations (GTO) program (PID 6, AOR 3657472) and during the Cygnus-X Legacy Survey (PID 40184, AORs 27108352, 27107328, 27108608; Hora et al. 2009; Kraemer et al. 2010) at wavelengths of 3.6, 4.5, 5.8, and 8.0 μm. The images were obtained using the 12 s high dynamic range mode, which takes one short (0.6 s) and one long (12 s) frame at each pointing in the map. Part of the region was also observed in the GLIMPSE360 Exploration Science program during the Spitzer Warm Mission at 3.6 and 4.5 μm using 12 s frames (PID 61072, AOR 42053120; PI B. Whitney). We used the latest version of the Basic Calibrated Data (BCD) available in the archive¹⁸ , which was version 18.7.0 except for AOR 42053120 which was version 18.18.0. The BCD were used rather than the "corrected" or cBCD because the pipeline automatic column pulldown correction causes many artifacts in the data near regions of bright extended emission, as is present in S106. The IRAC images were individually processed using the routine imclean¹⁹ which is an IRAF²⁰ -based script for removing the bright source artifacts ("pulldown," "muxbleed," and "banding"; Hora et al. 2004; Pipher et al. 2004) from the BCD images. Saturated regions and the resulting artifacts in the 12 s frames were masked as well, allowing the unsaturated short frame data to be used in those locations. The artifact-corrected BCDs were mosaicked into larger images using the IRACproc package (Schuster et al. 2006). IRACproc is a PDL script based on the Spitzer Science Center's post-BCD processing software MOPEX (Makovoz & Khan 2005) which has been enhanced for better cosmic ray rejection. The final images have plate scales of 0863 per pixel, with intensity units in MJy/sr.

We include observations taken with the MIRLIN camera (Ressler et al. 1994) at the 3 m NASA Infrared Telescope Facility (IRTF) on 2002, June 13. These observations include images taken through the CVF at a wavelength of 11.300 μm and through the N5 filter at a center wavelength of 12.492 μm. The MIRLIN plate scale at the IRTF is 047 per pixel on the 128 × 128 Si:As BIB array. The observations were taken in chop/nod mode with cumulative integration times of 360 s and 112.5 s for 11.300 μm and 12.492 μm, respectively. The chopper throw was 30'' east/west with a similar north/south telescope nod that we varied slightly from one set of images to the next. After sky subtraction, the many individual frames were resampled to a smaller pixel scale (012 per pixel) and then shifted and added using the bright central star for spatial registration. The observations were flux-calibrated using observations of β Peg taken immediately after S106, adopting the zero-magnitude fluxes in the MIRLIN handbook.

We observed S106 with the FORCAST instrument (Adams et al. 2012a; Herter et al. 2012) on SOFIA (Young et al. 2012) at the wavelengths 19.7, 25.3, 31.5, and 37.1 μm, using FORCAST guaranteed observing time during Cycle 1 (Flight 108 on 2013, June 21) and Cycle 2 (Flight 170 on 2014, May 8). FORCAST is a dual channel imager and spectrometer that utilizes a 256 × 256 pixel format Si:As detector for 5–26 μm and a 256 × 256 pixel format Si:Sb detector for 26–40 μm, with a rectified plate scale of 0768 per pixel in each camera. The wavelengths 19.7, 25.3, and 31.5 μm were observed using the dichroic beamsplitter, while 37.1 μm was observed directly. The observations were performed with off-source chopping and nodding (C2NC2 mode) and dithering, whereby the off-source fields were chosen to be ones free of detectable emission in Midcourse Space Experiment (Price et al. 2001) 21 μm images. Typical dwell times in each nod beam/dither position were approximately 30–90 s, including chopping inefficiencies.

The raw data were reduced and calibrated using the pipeline described in Herter et al. (2013). This pipeline performs chop and nod subtraction and applies corrections for droop, detector nonlinearity, multiplexer crosstalk, and optical distortion. Calibration factors were derived from the average photometric instrument response to standard stars observed during each corresponding observing campaign (a series of four or more flights). The dither positions were aligned and averaged, producing images with effective on-source exposure times of 480, 371, 375, and 195 s for 19.7, 25.3, 31.5, and 37.1 μm, respectively. The rectified plate scale for all the FORCAST images was 0768 per pixel. The effective rms noise levels were approximately 8, 10, 10, and 13 mJy/pixel for 19.7, 25.3, 31.5, and 37.1 μm, respectively. We estimate the calibration error due to flat fielding to be ∼10%.

We performed image deconvolution on the FORCAST images using several iterations of a maximum likelihood algorithm (Richardson 1972; Lucy 1974) to produce images with an effective spatial resolution of ∼24. The application of this algorithm requires a point-spread function (PSF) to be used as input. We chose to use a synthetic PSF since there were no point sources in the field that were sufficiently bright to use as a reliable reference PSF. The synthetic PSF was a convolution of an Airy function for telescope diffraction and a two-dimensional Gaussian for pointing instability that was scaled so that the FWHMs of the final PSF matched those of the point sources in the images that were to be deconvolved (De Buizer et al. 2012).

Finally, we derived astrometric solutions (WCS) for the images using the location of three point sources in the field and a tangent plane projection. This method produces astrometry that is accurate to within ∼04 (Adams et al. 2012b). The astrometrically calibrated frames were then aligned with and sampled to that of the IRAC 3.6 μm image (0863 per pixel), whose astrometric solution is determined from the position of the field stars in the 2MASS catalog.

In this paper, we use Herschel/PACS (Poglitsch et al. 2010) 70 and 160 μm observations from the Herschel Open-Time program OT2_jhora_2 (obsIDs 1342257386 and 1342257387). The data were taken on 2012, December 18, simultaneously with SPIRE (Griffin et al. 2010) and PACS, using a scanning speed of 60'' s⁻¹ with one repetition in the nominal and orthogonal observing directions. The individual scan directions of the PACS data were first reduced from the raw data of level-0 to level-1 using the standard pipeline in HIPE 10.2751. This includes flat-field and nonlinearity corrections. The next step (from level-1 to level-2) was performed within Scanamorphos v.21 (Roussel 2013). The processing consists of subtracting long and short timescale drifts and masking glitches and brightness discontinuities. The two scan directions were merged and projected onto a spatial grid of 2''/pixel and 3''/pixel for 70 μm and 160 μm, respectively. The angular resolution of the data is 6'' × 12'' for 70 μm and 12'' × 16'' for 160 μm, respectively (PACS Manual v.2.5.1). The absolute calibration uncertainties for the integrated source flux densities are estimated to be ∼10% for 70 μm and ∼20% for 160 μm.

Figure 1 shows the Spitzer/IRAC images. Based on the spectrum of S106 presented by van den Ancker et al. (2000), these images trace dust continuum emission, recombination line emission, and PAH emission. In addition, the images contain stars with primarily photospheric emission and young stellar objects that exhibit infrared excess emission. The thermal emission is bipolar. A dark lane is seen in the region where a similar feature is seen in the near-IR (Oasa et al. 2006). There is also emission from a dusty clump (PL 1) to the southwest of the S106 IR as indicated in Figure 1.

Zoom In Zoom Out Reset image size

The MIRLIN 11.3 μm and 12.492 μm images are shown in Figures 2 and 3, respectively. Although the 11.3 μm image is not continuum subtracted, the emission is dominated by PAHs, while emission at 12.492 μm traces that of the dust thermal continuum. Emission can be seen from the bright, compact sources (IRS 1–8) listed in Gehrz et al. (1982), but also from the lobes. The dark lane seen in near-IR images is also dark at both these wavelengths. In Figure 2, we show the locations of the ionized region using Hα contours (Bally et al. 1998) and the UV shadow as the region devoid of 5 GHz free–free emission (Bally et al. 1983). The position PL 1 lies outside the MIRLIN field.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The deconvolved SOFIA/FORCAST images are shown in Figure 4. Emission was detected from both lobes and from several bright, compact sources. Several of these sources (IRS 1, IRS 3, IRS 5, IRS 6, IRS 7, IRS 8) were detected in ground-based images at ∼10 and ∼18 μm by Gehrz et al. (1982) and Smith et al. (2001) and are indicated in Figure 4. We did not detect the photosphere of S106 IR (IRS 4) at 19–37 μm. The dusty clump PL 1 was detected at all wavelengths by FORCAST.

Zoom In Zoom Out Reset image size

The Herschel/PACS images are shown in Figures 5 and 6. At 70 μm, the morphology follows that of the warm dust continuum that is seen at 19–37 μm. At 160 μm, the morphology changes dramatically due to detection of cold dust in the molecular cloud, and resembles the submillimeter continuum emission (Simon et al. 2012). A cavity is seen surrounding the western region around the lobes, which extends southward along the eastern and western sides of PL 1.

Zoom In Zoom Out Reset image size

The peak emission at the illuminated edge of PL 1 shifts deeper into the molecular cloud at 160 μm when compared with emission at 37 and 70 μm (Figure 7), suggesting that the clump is self-absorbing. The amount of this shift is nearly 0.05 pc in projection. In addition, the peak of the PAH emission (Smith et al. 2001) lies at the inner edge of the clump and starts to decrease where the 37 μm emission from larger grains peaks.

Zoom In Zoom Out Reset image size

In Table 1, we report the total detected dust continuum flux densities at 12–160 μm. We include the flux density at 350 μm (Simon et al. 2012). The aperture was chosen for each wavelength according to field of view and the extent of the detected dust emission. We limited the aperture at 70 and 160 μm in order to minimize the chance of the inclusion of dust that may be heated by external stars. We examined the 70/160 μm color temperature in this aperture and did not find a gradient that would indicate external heating. The apertures for the FORCAST and Herschel images are shown in Figures 4–6. The flux densities of the point sources IRS 2 and IRS 4 have been subtracted (where applicable). From these flux densities, we derive a lower limit to the total dust luminosity of L_dust ≳ (9.02 ± 1.01) × 10⁴ L_⊙. We state the luminosity as a lower limit because the some of the stellar radiation may be escaping through the bipolar geometry and there may be dust emission outside the 70 and 160 μm apertures.

Zoom In Zoom Out Reset image size

Table 1.  Total Dust Continuum Flux Densities in S106, Including 350 μm Data from Simon et al. (2012)

λ (μm)	Aperture	Fν (kJy)
12.492	101 × 111	0.0649 ± 0.0130
19.7	174 × 230	2.53 ± 0.25
25.3	174 × 230	4.32 ± 0.43
31.5	174 × 230	5.66 ± 0.57
37.1	174 × 230	6.86 ± 0.69
70	497 × 753 elliptical	19.1 ± 1.9
160	497 × 753 elliptical	15.8 ± 3.2
350	572 × 405	0.630 ± 0.126

Note. The values exclude emission from the point sources S106 IR and IRS 2 (Gehrz et al. 1982).

Download table as:  ASCIITypeset image

A modified blackbody color temperature map derived from the 19 and 37 μm flux ratios is shown in Figure 8 with Hα emission (Bally et al. 1998) overlaid with contours. For this map, the emissivity of the dust grains was assumed to be proportional to λ^−1.8 (Abergel et al. 2011). Overall, the dust temperatures are lower (∼70–90 K) in the equatorial region and tend to increase out into the lobes (∼107 K).

Zoom In Zoom Out Reset image size

We resolved several cool lanes in the S106 IR UV shadow, with temperatures of ∼70 K, and three compact regions in the UV shadow with temperatures of ∼90 K. One of these warmer regions was identified as a 10 μm compact source (IRS 7) by Gehrz et al. (1982).

Figure 8 also depicts the locations used for dust modeling (Section 4). The positions and fluxes at these locations are listed in Table 2.

Using the temperature map shown in Figure 8 and the 37 μm image, we computed the optical depth τ along the line of sight. We show this optical depth map in Figure 9. The highest column densities (0.08 ≲ τ ≲ 0.12) are located near S106 IR, in the UV shadow, with the lobes becoming increasingly optically thin (τ ≲ 0.08) at larger distances in the lobes. The FORCAST images have low signal-to-noise ratios between IRS 6 and PL 1, indicating a low dust column density or absence of dust in this area.

Zoom In Zoom Out Reset image size

At a location within the nebula, the incident radiation field from the star is ${({R}_{*}/D)}^{2}{F}_{\lambda ,*}{e}^{-{\tau }_{\lambda }},$ where R_* is the radius of the star, D is the distance from the stellar surface to the dust, F_λ,* is the flux density of the star at the surface of the star, and τ_λ is the optical depth of the dust between the star and the location. The distance from the star to the dust was set to the projected distance (D_proj) between them, scaled to a distance of 1.4 kpc (Schneider et al. 2007). We adopt ${\tau }_{\lambda }={\tau }_{\mathrm{UV}}{(121.5\;{\rm{nm}}/\lambda )}^{\beta },$ where β = 1.85 (Landini et al. 1984) and τ_UV is a free parameter. For the stellar flux density, we used a stellar atmosphere with a spectral type of O7V and an effective temperature of 37,000 K from Castelli & Kurucz (2004).

For the dust component, we consider contributions to the emission from amorphous silicates (Draine & Li 2007), amorphous carbon, neutral PAHs, and ionized PAHs (Compiègne et al. 2010). In cases of relatively low extinction, we consider emission from both very small (1.2 × 10⁻⁷ ≤ a ≤ 1.5 × 10⁻⁶ cm, where a is the grain radius) grains (VSGs), which can be transiently heated in a UV radiation field (Ryter et al. 1987), and big (1.5 × 10⁻⁶ ≤ a ≤ 1.1 × 10⁻⁵ cm) grains (BGs), which are thought to be composed of silicates (Désert et al. 1990). The grain size distribution for both PAH species is modeled as

$$\begin{eqnarray}&&\displaystyle \frac{{dn}}{d{\rm{log}}a}\propto \displaystyle \frac{{e}^{-{\rm{log}}{(\displaystyle \frac{a}{{a}_{0}})}^{2}}}{\sigma }\end{eqnarray} \tag{ 1 }$$

where 3.1 × 10⁻⁸ cm ≤ a ≤ 1.2 × 10⁻⁷ cm, a₀ = 3.1 × 10⁻⁸ cm, and σ = 0.4. The grain size distributions for the amorphous silicates, amorphous carbon, VSGs, and BGs were modeled as dn/da ∝ a^α with α as a free parameter for each species. For the amorphous grains, 3.1 × 10⁻⁸ cm ≤ a ≤ 2.0 × 10⁻⁴ cm, for VSGs, 1.2 × 10⁻⁷ cm ≤ a ≤ 1.5 × 10⁻⁶ cm, and for BGs, 1.5 × 10⁻⁶ cm ≤ a ≤ 1.1 × 10⁻⁵ cm.

In Figure 8, we specify locations in the nebula for which we show the SED and perform dust SED modeling. These locations are cool lane positions CL 1–7, warm lane positions WL 1–3, southern lobe positions SL 1–3, southwestern clump position PL 1, and the positions of the bright, compact sources IRS 1, IRS 3, IRS 5, IRS 6, IRS 7, and IRS 8 identified in Gehrz et al. (1982). In Figures 10 and 11, we show the SEDs for these positions, including 3.29 μm IRTF/NSFCAM data from Smith et al. (2001) and 350 μm CSO/SHARC-II data from Simon et al. (2012).

Zoom In Zoom Out Reset image size

For CL 1–7, the dust composition is assumed to be dominated by a mixture of amorphous silicates, amorphous carbon, and PAHs (Figure 10). Positions CL 1–3 contain contamination from the S106 IR photosphere at 3.29–8.0 μm, and those data are not shown in the SEDs. For CL 1, some PAH emission may be present at 11.3 μm, but these components are not modeled as the ionization fraction is unconstrained. For CL 2, we model the PAH components and note that the ionized PAH component is an upper limit based on emission at 3.29 and 11.3 μm. At CL 3, there is emission at 3.29 μm but not at 11.3 μm. In this case, the PAH components are not modeled as the ionization fraction is unconstrained. For CL 4, we show 3σ upper limits at 3.29 and 12.492 μm. At this position, the IRAC 3.6 and 4.5 μm points help constrain the size distribution for amorphous carbon. Emission at 11.3 μm is used to place an upper limit on the PAH emission, which underpredicts the emission at 5.8 μm. There is a possibility that the IRAC photometry at this position is affected by emission from IRS 6. At CL 5, the 3σ upper limit for 3.29 μm and 4σ upper limit for 12.492 μm are shown. Again, the IRAC fluxes at 3.6 and 4.5 μm help constrain the size distribution for amorphous carbon. For CL 6, a 3σ upper limit at 3.29 μm is shown. At position CL 7, the IRAC data points are contaminated by emission from IRS3 and are shown as upper limits. A 3σ upper limit is shown for 12.492 μm.

The modeling parameters and results for the CL positions are given in Table 3. The modeling results yield values of τ_UV in the range 8–40, with the highest extinction at position CL 1. We note that τ_UV does not necessarily increase with increasing distance from S106 IR. The amorphous silicates contribute ∼50%–94% to the dust mass, with the remainder contributed by amorphous carbon (∼6%–50%) and, where detected, PAHs (<1%).

For WL 1–3, a mixture of PAHs, VSGs, and BGs was modeled (Figure 10). The ionized PAH model component was averaged over the IRAC instrument response for the 5.8 μm bandpass and then normalized to match the emission at 5.8 μm, which is dominated by the 6.3 μm PAH feature and contains little recombination line emission (van den Ancker et al. 2000). The IRAC 8.0 μm band contains [Ar ii] and [Ar iii] lines at 6.99 and 8.99 μm, respectively, and was not used to constrain the model components. Also, since the IRAC 3.6 and 4.5 μm bands conain hydrogen recombination lines, such as Pf δ (3.2970 μm), Pf γ (3.7406 μm), Br α (4.0523 μm), and Pf β (4.6539 μm; van den Ancker et al. 2000), they were not used for model fitting. Residual emission at 3.29 μm above the emission from ionized PAHs was used to normalize the neutral PAH component. We attribute excess emission at 70–350 μm to contamination from IRS 6 and emission from cooler dust along the line of sight that can be seen at both 160 μm (Figure 5) and 350 μm (Simon et al. 2012). From Table 4, values of τ_UV for the warm lanes are in the range 2–6. The dust mass is dominated by BGs (∼91%–98%), with contributions from VSGs (≲1%) and PAHs (≲2%). The fraction of PAHs that are ionized spans the range 82%–97%.

SEDs and models for IRS sources, SL 1–3, and PL 1 are shown in Figure 11. We reiterate that PL 1 lies outside the MIRLIN field. The modeling approach for these sources was similar to that for WL 1–3, but the results are slightly different (Table 4). In these cases, values for τ_UV are in the range ∼0–5, with the lowest values of extinction found at the southernmost positions, SL 3 and PL 1. Again, the model composition is dominated by BGs (∼50%–98%), with contributions from VSGs (∼1%–43%) and PAHs (∼2%–7%). The ionized PAH fractions are 88%–97%, with the higher ionization fraction values correlated with the lower values of τ_UV.

Zoom In Zoom Out Reset image size

5.1.1. Luminosity

5.1.2. The Circumstellar Environment

It is likely that the star illuminates dust in the lobes through an opening in the equatorial geometry of the dust. This would explain the overall lower temperatures observed near S106 IR and relatively high temperatures in the southern lobe. Smith et al. (2001) presented a color temperature map that exhibited a more uniform structure at a temperature of ∼135 K, using a λ⁻¹ emissivity law. The SOFIA observations contain sensitivity that extends farther into the lobes, where the temperature gradient is more apparent, which may explain the difference in temperature gradient between the two maps.

The agreement between the observed dust color temperatures and the equilibrium temperature computed by radiative transfer in the southern lobe and PL 1, where stellar extinction is low, suggest that the dust heating is dominated by stellar radiation, rather than by grain–electron collisions or trapped Lyα heating. The following calculations support this assertion. We computed the radiative heating rate on a single grain assuming, conservatively, a moderate distance (D = 0.15 pc) from the star with some extinction of the UV field (τ_UV = 3), and a grain aborption efficiency Q_λ = 1 for λ < 0.4 μm and Q_λ ∝ λ^−1.6 for λ ≥ 0.4 μm (Draine & Lee 1984). The radiative heating rate, Γ_rad is given by

$$\begin{eqnarray}&&{{\rm{\Gamma }}}_{\mathrm{rad}}=\displaystyle \frac{{R}_{*}^{2}}{{D}^{2}}\pi {a}^{2}\int {F}_{\lambda ,*}{e}^{-{\tau }_{\lambda }}{Q}_{\lambda }d\lambda .\end{eqnarray} \tag{ 2 }$$

The heating rate Γ_coll for grain–electron collisions is given by Dwek (1987):

$$\begin{eqnarray}&&{{\rm{\Gamma }}}_{\mathrm{coll}}={\left(\displaystyle \frac{32}{\pi {m}_{{\rm{e}}}}\right)}^{1/2}\pi {a}^{2}{n}_{{\rm{e}}}{({{kT}}_{{\rm{e}}})}^{3/2},\end{eqnarray} \tag{ 3 }$$

where m_e is the mass of an electron, n_e is the number density of free electrons, and T_e is the electron temperature. We assumed a typical value for the electron temperatures in a gas-cooled nebula (T_e ≈ 10,000; Tielens 2005) and used an estimated value for the electron number density (n_e ≈ 3 × 10⁴ cm⁻³) from Bally et al. (1983). We find the heating rate for grain–electron collisions is ≲15% that of the radiative heating. Such a small component of the heating will have little effect on the grain equilibrium temperature.

We also expect a small contribution to the dust heating from Lyα radiation due to the presence of warm gas (e.g., Schneider et al. 2002, 2003; Simon et al. 2012). The rate of this heating Γ_α is given by Tielens (2005):

$$\begin{eqnarray}&&{{\rm{\Gamma }}}_{\alpha }=\displaystyle \frac{{n}_{{\rm{e}}}^{2}{\beta }_{B}h{\nu }_{\alpha }}{{n}_{{\rm{d}}}},\end{eqnarray} \tag{ 4 }$$

where β_B is the hydrogen recombination coefficient to all levels with n ≥ 2, ν_α is the frequency of a Lyα photon, and n_d is the number density of dust particles. We assumed a high degree of ionization (x ≈ 1), a charge neutral gas (n_e = n_p, where n_p is the number density of protons), and a standard gas-to-dust mass ratio of 100. For a grain with radius 0.1 μm and mass density ρ = 3 g cm⁻³, we find Γ_α ≈ 3.2 × 10⁻⁶ Γ_rad, where Γ_rad is the heating rate on the grain from the star at a distance of 0.15 pc with τ_UV = 3. Therefore, Lyα radiation will not appreciably affect the dust grain equilibrium temperatures.

In the cool lane positions, the abundance ratio of amorphous carbon to amorphous silicate grains is variable, but brackets the values found in the interstellar diffuse high galactic latitude fields of Compiègne et al. (2011; 83% amorphous silicates, 17% amorphous carbon). In the regions with a stronger UV field, the relative abundances of the BGs, VSGs, and PAHs are similar to those found in the ISM (88% BGs, 6.4% VSGs, 5.9% PAHs; Désert et al. 1990) and in the M16 PDR locations (Flagey et al. 2011). The exception is the relative VSG abundance in the southern lobe location SL 3 and the southwestern clump (PL 1) is higher than at locations closer to the star. The increase in VSGs is accompanied by a relatively high proportion of ionized PAHs when compared with the other locations. The increase in relative VSG abundance is similar to that in the "reverse shell" in M16 (Flagey et al. 2011); however the "reverse shell" shows an absence of PAHs. Flagey et al. (2011) provide wind-driven grain–grain collisions as one possible explanation for the enhancement of small grains in the M16 shell. Such a process may be at work at PL 1, although another explanation may be that there is less photo-evaporation of VSGs as a consequence of the weaker radiation field at PL 1. Photo-evaporation of VSGs has been proposed to explain the abundances of VSGs and PAHs in other regions such as Ced 201 (Cesarsky et al. 2000; Berné et al. 2007), NGC 2023 north (Compiègne et al. 2008), ρ Oph-SR3 and NGC 7023-NW (Rapacioli et al. 2005), and NGC 7023-E and ρ Oph-filament (Berné et al. 2007).

Hα images show that the size of the H ii region extends southwest from S106 IR to an ionization front near location PL 1 (Bally et al. 1998). Theory states that the Strömgren radius of an H ii region expands with time according to the equation (Ward-Thompson & Whitworth 2011):

$$\begin{eqnarray}&&R(t)\simeq 5{\left(\displaystyle \frac{{\dot{N}}_{{\rm{H}}\;{\rm{I}}}}{{10}^{50}\;{{\rm{s}}}^{-1}}\right)}^{1/7}{\left(\displaystyle \frac{{n}_{0}}{{10}^{3}{{\rm{cm}}}^{-3}}\right)}^{-2/7}{\left(\displaystyle \frac{t}{{\rm{Myr}}}\right)}^{4/7}{\rm{pc}},\end{eqnarray} \tag{ 5 }$$

where ${\dot{N}}_{{\rm{H}}\;{\rm{I}}}$ is the number of hydrogen-ionizing photons per second emitted by the star, n₀ is the number density of gas molecules in the surrounding molecular cloud, and t is time. If we assume ${\dot{N}}_{{\rm{H}}\;{\rm{I}}}=4.1\times {10}^{48}\;{{\rm{s}}}^{-1}$ (Sternberg et al. 2003) and n₀ ≈ 1.4 × 10³ cm⁻³ (Schneider et al. 2002), then a Strömgren radius of 0.6 pc (the distance from S106 IR to the ionization front assuming an inclination angle of 30°; Gehrz et al. 1982) is reached in ∼5 × 10⁴ years. This timescale is consistent with the existence of a photo-evaporating disk, but we caution that it is only a rough estimate as n₀ may vary in the surrounding cloud.

Bipolar continuum emission extends beyond the ionization front. The size of the bipolar region depends on the accretion and outflow histories of S106 IR, and the clump at PL 1 may be a remnant, higher density clump that is still eroding from the formation of the lobes through outflow. Alternatively, it may be a clump of swept-up material at the ionization front. Either case indicates that there are complex dynamics in the bipolar region. Further study of the dynamics of the region will require multi-wavelength, spectrally resolved observations.