Astron. Astrophys. 339, 759-772 (1998)

2. Observations and data reduction

2.1. Adaptive optics imaging

G45 was observed in August 1995. ESO's adaptive optics system ADONIS (Beuzit et al. 1994) was used at the 3.6 m telescope on La Silla/Chile to obtain high-resolution images in the H and bands. In , a mosaic of two frames was made that resulted in a total integration time of 400 s in the image centre. During the observations, the seeing was , the high-order adaptive optics correction improved the full-width half-maximum (FWHM) of the point spread function (PSF) to 0:004. The integration time for the H image was 300 s. All frames were subject to standard bad-pixel removal, flat fielding, and dark-frame subtraction processes before being combined in the resulting images. For better identification of the point sources, part of the image was deconvolved with the point source located at position (,) using 200 iterations of a slightly modified maximum-likelihood deconvolution algorithm. The modification concerns the last step of each iteration, where the re-convolution is done with a narrower Gaussian instead of the original point-spread-function as described in Lucy (1974). This results in a better resolution of close point sources. Photometric calibration for the NIR images was obtained by calibrating the fluxes in the images using separate images of the UKIRT standard Y 4338.

2.2. MIR imaging

Three mid-infrared images were obtained. The first one was taken using SpectroCam-10 (Hayward et al. 1993) at the 200-inch Hale Telescope of the Palomar Observatory ². The filter effective wavelength was 11.7 µm with µm. A 5 frame mosaic was combined into the image, the average on-source time at each pixel is 20 s. The star Lyr served as a standard for flux calibration. This image will be referred to as the 12 µm image in the discussion. The second image was taken using MANIAC (Böker et al. 1997) at ESO's 2.2 m telescope on La Silla/Chile. MANIAC's N-band filter was used with µm and µm. The total integration time sums up to 1135 s for the frames that were combined into the image. Photometric calibration was obtained by observing the standard stars Aqu, Ser, and Gru. This will be our 10 µm or N image.

A third MIR image was obtained in the L-band ( µm, µm). The new MIR-Camera TC-MIRC (Robberto et al. 1994) was used at the 1.5 m TIRGO telescope on Gornergrat/Switzerland. The total integration time for this image was 600 s. Photometric calibration was achieved by observing the standard star HD203856. This image will be referred to as 3.8 µm or L image.

2.3. Narrow band imaging

The Br image was taken using BLUE-MAGIC (Herbst et al. 1993) at the 2.2 m telescope on Calar Alto/Spain which belongs to the German-Spanish Astronomical Centre. One image was taken using the Br ( µm, µm), another one with the adjacent continuum filter ( µm, = 0.022 µm). The total integration time for each of these images was 900 s. Photometric calibration was achieved by using the standard star GL748. Its magnitudes at the wavelengths of the continuum and the Br filter were derived by interpolating its known spectral energy distribution, assuming a featureless spectrum. After subtraction of the continuum, it turned out that the residual total fluxes in the stellar PSFs in the image were below the sky sigma level, so no further corrections were applied.

2.4. The astrometric reference frame

When drawing conclusions on the nature of a certain object, a great deal of the information is taken from the comparison of its structural appearances at different wavelengths. Apart from other issues like the different beam sizes for each observation, it is crucial to establish a common astrometric reference frame for all images. This is especially difficult when one tries to match images of 0:002 resolution like our deconvolved image and the 6 cm VLA maps from WC89. We adopted a two-way strategy to build our reference frame: First, the position of the wavefront calibrator which is visible in the image at position (,) (see Fig. 1) was taken from the digitized sky survey. The image was then calibrated using the known scale of 50 milliseconds of arc per pixel. This tied the NIR reference frame to the digitized sky survey. Secondly, the astrometry for the Br image taken with MAGIC was obtained by optimizing the cross-correlation with the 6 cm VLA image, the latter one being smoothed to the same resolution before the procedure. As the Br image contains the same stars as the image (before continuum subtraction, of course), we were able to check the two astrometric frames against each other. It turned out that the deviations between them were below 0:002. This indicates that all larger offsets are indeed real. The method of optimizing cross-correlation was also applied to tie the 10µm image to the reference frame. Before optimizing the cross-correlation between the images, a mask was applied that blanked everything except the ionization front. The reference frames of the 12 and 3.5 µm images were chosen to make the MIR point-source at position (,+10:005) appear at identical positions in all MIR images.

Fig. 1. image of G45.

2.5. Photometry

Photometry of the point sources detected in the AO images was performed in an aperture of 0:005 diameter because of the crowdedness of the field. Due to the non-Gaussian shape of the AO-corrected PSF, an aperture correction of 0.95 mag in and 1.23 mag in H was applied to account for flux outside the aperture. This correction was determined using sources at the reference position as well as a and b (in H, a is not visible). Background subtraction was achieved by subtracting the mean value of six sky measurements around the corresponding source using the same aperture to account for the varying background. The detection limit in the -image is mag, in the H-image it is mag, both derived from the sky noise. Photometric accuracy is generally better than mag. Photometry of the MIR images was done using an aperture of (2:0075 at 3.5 µm) diameter. Background subtraction was achieved in the same way as in and H. The accuracy of the photometric measurements in these images amounts to approximately 7% of the flux level or 0.1 magnitudes at most. The total source flux was determined in these images as well as in the image by applying an aperture of diameter.

2.6. A word on image filtering

All images presented as results from our imaging campaign were subject to filtering with the multi-scale maximum entropy method described by Pantin & Starck (1995). This method uses a wavelet decomposition to detect and remove the noise from an image. The band image was also subject to this kind of filtering before the deconvolution was applied. However, all this filtering was done solely for illustrative purposes. All photometric and mathematical operations described throughout the paper were performed on the "raw" data, before filtering or deconvolution. It should be stressed that all features discussed in the text can be found in the raw data. Therefore, we are certain not to describe any filtering artifacts whatsoever!

© European Southern Observatory (ESO) 1998

Online publication: October 22, 1998
helpdesk.link@springer.de