3.1. NIR images
Figs. 1 and 2 show the results of our adaptive optics near-infrared imaging. The images presented in the figures were subject to filtering with the multi-scale maximum entropy method described by Pantin & Starck (1995). To enhance the details in the images as they are shown in the figures while at the same time preserving extended structures, we present an additive combination of the image with a deconvolved version. The deconvolution was achieved by applying 100 iterations of a maximum-likelihood algorithm as described in Lucy (1974). The combination was done by adding 0.3 times the deconvolved image and 0.7 times the original one. The reference position given in the figures denotes the centre of the approximately spherical shell visible in the radio contours. This position will be used extensively in forthcoming sections. Superimposed on both images are the VLA maps obtained by WC89.
In the lower left corner of the images, a bright star can be seen which served as wavefront reference for the AO system. A band of extended emission and unresolved sources stretches towards the north-east at a position angle of approximately 45o. Several point-like sources are embedded in the extended emission. Both extended and unresolved emission are stronger in than in H (note the different cut levels!). Apart from the bright star, the most prominent structure in follows the northern half of the radio shell. This is what we will later discuss as the visible half sphere of the dust shell surrounding a hot O6 ZAMS star. In H, only weak remains of these structures are visible, except two supposedly stellar sources at positions (+0:003,-1:005) and (+0:005,-2:005) in the southern half of the radio shell. Later, we will argue that these sources are probably foreground objects.
3.2. MIR images
The results of our broad band imaging in the mid-infrared are presented in Figs. 3 and 4. The Q-band image was subject to noise removal by the maximum entropy filter described above. Both images show the characteristic arc-like structure already visible in : A banana-shaped bow opening to the south and following the contours of the northern half of the radio shell. The contours of the L band image confirm that this general structure is present also at that wavelength. However, especially from Fig. 4 it becomes clear that extended emission exists beyond the radio shell. The extraordinary fact that the general shape of the object remains constant from the near- to the mid-infrared indicates that the optical depth does not change significantly over these wavelengths. On the other hand, it is noteworthy that no infrared radiation at all is detected in the southern part of the radio shell. This is extensively discussed in Sect. 4. We note that the maximum of the emission shifts to the west along the arc by 0:006 in N (11.7 µm) and by 1:004 in Q compared to .
3.3. 1.3 mm continuum map
Fig. 5a shows the result of our 1.3 mm continuum mapping. The 1 rms level measured in source-free regions in the map is 115 mJy/beam. Note that the continuum emission at 1.3 mm is the sum of thermal dust emission and free-free emission.
The contour map shows that the source appears slightly extended in the 23" SEST beam. Deconvolution yields an intrinsic FWHM of 16""". The total flux of the source is 22 Jy with 17 Jy in the compact core and 5 Jy in the envelope. The extended feature seen in the map might partially be the result of the beam side lobes. However, the signal denoted by the lowest contour line contains far more power than the error beam. Thus the extension is indeed real. This becomes especially obvious from part B of the figure, where profiles of the source and, for comparison, of Uranus are shown.
To access the distribution of the cold dust, we subtracted the free-free contribution from the 1.3 mm map. This was achieved by multiplying the 2 cm map with a factor of (assuming optically thin free-free emission) and convolving the resulting map with the SEST beam before subtracting it from the 1.3 mm map. The result of this operation is shown as contours in Fig. 6. Here it is obvious that the orientation and position of the thermal dust emission is consistent with that of a presumed foreground cloud (see Sect. 4.1.1) extincting the southern half of G5.89.
At the eastern edge of the mapped region (the scanning direction was north-south), a second source is visible with a peak flux of 1.5 Jy/beam (approximate peak position R.A.= and Dec.= -24o04´09"). However, this source is not completely covered by our map. This source is obviously the HII region G5.90-0.43, visible in the radio continuum maps of Zijlstra et al. (1990) and denoted as source B in their paper.
The results of our NIR narrow-band imaging are presented in Fig. 6. The image is colour-coded, red representing flux measured in the H2(1-0)S1 line, green the corresponding continuum filter flux and blue the flux measured in the Br filter. This image gives a large-scale overview of the region. In the lower right fewer stars are visible than in the upper left. This suggests that large clouds are obscuring our view towards that region. Close to the H2 source labelled "A", brownish, arc-shaped structures are visible which might be regarded as rims of clouds reflecting light from nearby stars (similar to the "fingers" in M16; see Pound 1998, Hester et al. 1996). The impression is that G5.89 (situated at the reference position) is sitting exactly at the rim of such a cloud with only its northern half visible.
For the mid-infrared data, no subtraction of continuum flux from the narrow-band images was attempted. The main contribution to the flux density in the image taken at 12.8 µm comes from the [NeII ] line. Thus, bright regions in that image should be the highly ionized regions in G5.89. From Fig. 7, we learn that indeed these regions correlate very well with the highest radio contours. As no [NeII ] emission is seen in the southern half of the radio shell, we infer that the optical depth there is still high at 12.8 µm .
The image taken at 10.6 µm could contain flux from the [SIV ] line. From Faison et al. (1998) we learn that this line is virtually not present in G5.89. Thus, we use the resulting image to determine the optical depth of the silicate feature by comparing its flux to that in our 11.7 µm image. The result of this procedure is presented in Fig. 8. The natural logarithm of the ratio of the two fluxes is shown as gray scale, superimposed are contours denoting the 2 cm radio contours. We note that the logarithm of the flux ratio can only serve as a measure for the optical depth (The flux is not even measured at 9.7 µm after all!), but does not yield its absolute value. From Fig. 8, we learn that the silicate optical depth is highest at the locations of the largest radio flux. This behaviour is somewhat peculiar and not fully understood. We will come back to this feature in Sect. 4.2.1 when discussing the properties of our model of G5.89.
3.4.2. Line fluxes
Line flux densities were measured using the calibration procedure described in Appendix A and then performing aperture photometry on the calibrated images. The total line fluxes were then obtained via the filter widths and are presented in Table 2. We note that our measured Br flux is about 3.8 times higher than that measured by Moorwood & Salinari (1983). Also, the flux density in the [NeII ] filter measured by us differs from that measured by Faison et al. (1998) by around 80 Jy. The latter authors measured a peak flux density of 225 Jy. Of course different spectral resolutions and responses might account for this difference, because of the averaging over adjacent continuum flux. On the other hand, our measured flux of 24.6 Jy at 10.6 µm seems to agree with their data as seen in their Fig. 1.
The neon flux is particularly interesting because it allows to determine the ratio of Ne+/H+. Unfortunately, we have no chance of subtracting the underlying continuum flux density from the total flux density to get the line flux. However, we can estimate the Ne+ abundance by judging the line contribution to be roughly 20% from Faison et al. (1998). If we convert the measured total flux density of 307 Jy to a flux of 5.4 W cm-2 using the filter width of 0.4 µm and the 20% mentioned above, we can use Eq. (3) of Watarai et al. (1998) to determine the ratio of the number abundances. Inserting the appropriate values for temperature (104 K) and electron number density (1.1 cm-3, see Sect. 3.5) as well as the size of the source (1.7 sr) and the emission measure (64 pc cm-6, see Sect. 3.5), we derive . This value seems reasonable compared to the 9.6 found by Watarai et al. (1998) for G29.96-0.2 and the 8.1 found in the Orion Nebula (Rubin et al., 1991). The value for Orion gives the ratio of the neutral atom number abundances. Since we cannot tell anything about the Ne abundance, except that it should be high because of the early type central star of G5.89, the difference might be easy to understand in this case.
To derive the NIR-extinction towards an ionized source, one can compare the radio flux to the flux in recombination lines such as Br . In this work we used essentially the same procedure as in Paper I which in turn followed that described by Watson et al. (1998). One main difference to Paper I should be noted: This time, we have two radio images, taken at 2 and 6 cm. However, the spectral energy distribution of G5.89 (see, e.g. WC89) shows that the emission at 6 cm is already optically thick. Thus, it is possible to use the beam temperature of the 6 cm map as electron temperature directly. The temperatures then range from 0.8 K to 1.6 K with a mean value of 1.1 K. The temperature distribution is of course the same as the that of the 6 cm emission. With this knowledge, the emission measure and the expected Br flux were computed from the optically thin 2 cm emission. Missing flux in the VLA maps should not be a problem because large-scale maps by Gomez et al. (1991) taken in VLA-D configuration show that no large-scale emission exists, except in their lowest contours. The computed emission measure peaks at 360 pc cm-6 with an average value of 64.2 pc cm-6. This deviates from the results of WC89 and those of Zijlstra et al. (1990), because they used uniform electron temperatures of 104 K and 8000 K, respectively. The mean electron densities implied by these values and the mean path length of 0.05 pc is 1.1 cm-3. The resulting Br -extinction map is shown in Fig. 9. The Br -extinction rises constantly from northeast to southwest. We will later argue (Sect. 4.1.1) that this is also due to the foreground cloud, not due to internal extinction in the dust shell.
From our 11.7 µm and Q-band images, we can derive the dust distribution inside G5.89 as well as the temperature of that distribution starting from a simple model. Assuming the emission to be optically thin at both wavelengths, the ratio of the two flux densities yields the temperature, while the total flux density in combination with the derived temperature yields the mass of dust.
In this expression, the dust mass absorption coefficient is taken from Ossenkopf & Henning (1994) for the appropriate wavelengths assuming a gas density of cm-3 and a size distribution after MRN (Mathis et al., 1977) without ice mantles. The quantity T is the dust temperature, h the Planck constant, k the Boltzmann constant, and d the distance towards the source. After adjusting the PSFs of both images by convolving the 11.7 µm image with an appropriate Gaussian, we were able to derive both the temperature and the mass of the dust. This procedure was applied in image areas where both signals were above their corresponding 3 levels.
The result of this procedure can be seen in Fig. 10. The grey scale represents the mass distribution of the dust. Superimposed are contours denoting the temperature distribution. The lowest contour line is at 100 K, the spacing is 5 K.
We note that our result concerning the temperature qualitatively agrees with that of Ball et al. (1992). However, we are measuring a much lower absolute temperature with a peak value of only 123 K (Ball et al. derived a peak value of roughly 500 K) and an average temperature of 112 K. Several explanations apply. Ball et al. (1992) discuss in depth the requirements for deriving a physical temperature from an MIR flux ratio and the drawbacks suffered by deriving a simple colour temperature. However, we believe that our estimates are considerably closer to the actual physical temperature because of the following reasons:
The total mass we derive by this procedure is 1.7 . Simply correcting for optical depths from Sect. 4.2.1, we get 2.5 . When assuming the geometry of a half-sphere with radius 2:005, we derive a dust density of g cm-3 (g cm-3 in the optical depth corrected case.) The total flux at 11.7 µm used is 127 Jy, the total Q-Band flux 509 Jy.
Of course we are aware that taking the from Ossenkopf & Henning (1994) introduces some arbitrariness. The model of Ossenkopf & Henning (1994) assumes a gas density of 105 cm-3 and coagulation for 105 years. To estimate this uncertainty, we calculated the dust mass for the initial MRN-distribution of dust, where no coagulation has taken place. Such a scenario might apply if the dust aggregates have been destroyed by the heat and the opacity is similar to that of the interstellar medium. Using the appropriate opacities leads to a total mass of 2.1 (3.1 when correcting for the optical depth). An additional complication is the possible impact of very small grains, which might be subject to "quantum heating". Such grains can be heated stochastically by the impact of single photons and are not in thermodynamic equilibrium. However, we have no way to tell whether such grains are present in G5.89 and while modelling the source with radiative transfer (see Sect. 4.2.1), it turned out that such an effect has very little impact on the fluxes at 11.7 and 21 µm in the case of G5.89. In conclusion, we can say that the total hot dust mass of 1.7 represents a lower limit with an uncertainty of about a factor of 2 to 3.
Twenty-one areas were selected in the high-resolution H and images for aperture photometry. Some of them contain point sources. All areas show remarkable deviations from their surroundings in either H or . The locations and sizes of the photometric apertures are denoted in Fig. 11, while the results are presented in a colour-magnitude diagram shown in Fig. 12. The plotted magnitudes are derived from the integrated flux densities inside the apertures, no sky subtraction was done. Where the measured sources were inside our extinction map (Fig. 9), de-reddening was applied for a reddening vector after Rieke & Lebofsky (1985).
At first glance, most of the sources scatter along a band from the lower left to the upper right of the diagram. The lower border of this band is given by the detection limits. No apparent systematics can be recognised and we hesitate to assign the label "star" to most of the measured sources for reasons given in the next section.
© European Southern Observatory (ESO) 1999
Online publication: May 6, 1999