3. Data reduction and analysis
3.1.1. N -band
At 10 µm, fields are not crowded: for each of our images we had only one object within a relatively small field of view of 40". This leaves no doubt about the identification of the correct counterpart in the TIMMI field at this wavelength, even in non-photometric weather conditions. The disadvantage is that there are no other points of reference to improve upon the IRAS coordinates other than the telescope pointing.
The positions and Full Width at Half Maximum (FWHM) of the sources were determined from the images after filtering. Because of the variable sky conditions, the background level around the sources was badly defined. Consequently the determination of the (standard) star's PSF was very uncertain. To remedy this, we subtracted the background in the following way: we made an image in which we replaced a circular area surrounding the source by a flat surface fitted to the background outside this circle and subtracted this image from the original. The resulting frame contained only the object against a virtual zero background. Next a two dimensional Gaussian was fitted to the source to determine the position and FWHM more accurately.
At regular intervals during the two nights, we pointed the telescope to a standard star and put it exactly on the cross hairs. Next, we took an image to determine exactly the corresponding position on the array. Similarly, for each object, we calibrated the telescope pointing using the closest bright SAO star, before doing an offset to the IRAS position. There was always only one source in the field, which left no doubt about the identification of the IRAS source in the N -band image. We assumed that the IRAS position corresponds to the reference position on the array as determined from the standard stars. We measured the offset from this array position to the position of the source in the TIMMI image, and adopted this as the improved position of the IRAS source. The difference between the IRAS and TIMMI positions were mostly less than 10" which is in agreement with what was found in positional difference between the radio detections and the IRAS positions (Van de Steene & Pottasch 1993, 1995). The uncertainty in the TIMMI positions was estimated to be less than 5" and is due to the fact that in these regions with high extinction, the nearest SAO star could be several arcminutes away and the pointing of the ESO 3.6-m was not very good.
The median seeing was FWHM during the first night and, even worse, FWHM, during the second. Fifteen objects remained unresolved. Five appear to be extended at 10 µm. The contour plots of the extended sources are presented in Fig. 8. The morphology of the resolved objects is similar to young PNe. The size of the extended sources is given in Table 3. One showed Br in absorption, two in emission, and in two we didn't detect any Br in absorption or emission.
Table 3. The angular sizes of the 6 extended sources. The first column gives the name of the object, the second the extent in arcseconds in RA and DEC in the N -band image, the third similarly the visible extent in the K -band image, and the last whether Br was seen in absorption (A), emission (E), or the object had a flat spectrum at Br (F). Unresolved objects are indicated by PS.
The J H K L -band images were reduced in IRAF using standard procedures as described in the CASPIR manual (McGregor 1994). The two images taken 24" apart in declination were combined. Because the objects were most prominent in the K -band, the resulting K -band images are presented in Fig. 9.
The positions of the objects determined from these images are listed in Table 4. We used the IMCOOR package in IRAF and the positions of reference stars from the United States Naval Observatory Catalogue (USNO release A1.0, available via the ESO SKYCAT tool) to determine an accurate position. If the object had a counterpart in the catalogue, we adopted the USNO catalogue position. The positional uncertainty is similar to the uncertainty in the USNO positions, about 1". The object for which we have obtained the Br spectrum is indicated with a box in Fig. 9. Note that in some fields more than one star was in the slit. The IRAS position is indicated with a cross in Fig. 9.
Table 4. Coordinate list of the IRAS sources. If the object was associated with a star in a catalogue, the catalogue position is given. Otherwise the position as determined from our imaging is given. The abbreviations in front of the catalogue numbers have the following meaning: U - USNO-A2.0 , G - GSC 1.1 , D - DENIS . The galactic longitude and latitude in degrees are given in Columns 2 & 3 respectively.
Although the K-band seeing was sub-arcsecond, most of the objects are unresolved at this level. The sizes of the extended objects are presented in Table 3. All objects which are extended in the K -band show Br in emission, but not vice versa. Obviously, comparison of morphology in the N - and K -band is not a good identification tool.
The photometry was done with the program QPHOT in IRAF . The J H K L -band magnitudes of the objects were determined for each of the two images separately, and their agreement checked before averaging. The average difference between the two measurements was 0.01 mag for both nights in J H K and 0.05 mag for the L -band. The near-infrared magnitudes are presented in Table 5. The uncertainties are estimated at 0.05 mag in J H K and 0.1 mag in the L -band, including measurements errors and the uncertainty in the correction for atmospheric extinction. Table 5 also contains estimates for the K magnitudes derived from the Br spectra (see also Sect. 3.2). Usually these magnitudes are a bit fainter than the ones determined from the CASPIR images. This is probably due to slit loss.
Table 5. Johnson J H K L magnitudes for the program stars. Columns 2, 3, 5 and 8 give the J H K L magnitudes determined from the CASPIR images. Column 4 gives the K -band magnitude estimated from the continuum flux at 2.166 µm measured by IRSPEC . Column 6 gives literature values for the K magnitude, if available, and column 7 the references. Column 9 gives the SED class defined by van der Veen et al. (1989).
Table 5 also presents near-infrared photometry from the literature, when available. The literature values are based on aperture photometry rather than imaging. In crowded regions they are often brighter, likely due to contamination by neighboring objects. In all cases where the literature value was very different from our magnitude, the two values were not associated with the same star, and we were able to identify the star measured by the other authors in our images.
Near-infrared imaging is preferred to aperture photometry to identify the object in the field. It can provide very accurate coordinates and finding charts for follow-up observations. Given photometric weather and good seeing conditions, the magnitudes won't be contaminated by neighboring stars. When several near-infrared bands are available, colors can be used as a secondary means of identification. The counterpart often is the reddest, but not necessarily the brightest object in the near infrared. Due to their thick circumstellar dust shells, post-AGB stars often are very reddened.
3.2. The Br spectra
The data reduction was done using the standard reduction macros contained in the MIDAS image processing system. After flatfielding, sky subtraction, and rectifying the spectrum the resultant image contained a positive and a negative spectrum of the source. The next steps were to extract the positive and negative beam, invert the negative beam, and average the two. The spectra were accurately wavelength calibrated using the sky-emission lines present in the non-subtracted images. The rest wavelength of these lines were taken from Oliva & Origlia (1992). Using the standard star spectra, a flux conversion factor was determined to calibrate all spectra.
The values of the continuum flux at Br (2.166 µm) have been converted to Johnson K -band magnitudes for all sources and are listed in Table 5. Since the central wavelength of the K -band filter (2.20 µm) nearly coincides with the central wavelength of our spectra, no attempt has been made to correct for the different slopes in the continuum. The error introduced by this assumption is well within the accuracy of the flux calibration. The K -band fluxes of four sources have been marked as a lower limit. For IRAS 16279-4757 and IRAS 16594-4656 the reason is that the absorption profile extends beyond the observed spectral range, hence the estimated continuum is a lower limit to the true level. IRAS 13428-6232 and IRAS 17009-4154 are larger than the slit at these wavelengths. Flux will have been missed for other sources as well, because the peak-up procedure to position the object in the slit was not very accurate. No attempt has been made to correct for slit loss, because it was impossible to optically verify how well the object was centered in the slit.
When a clear continuum was present, the spectra were normalized by fitting a second order polynomial to the continuum and dividing the spectrum by the fit. The spectra of IRAS 16279-4757 and IRAS 16594-4656 showed a very wide Br absorption and no or hardly any continuum. Therefore, the spectrum of IRAS 16279-4757 was divided by the maximum flux present in the spectrum. For IRAS 16594-4656 some continuum seemed present and a linear fit was made between both ends of the spectrum. This normalization should be considered uncertain. The continuum of the four detected PNe is very weak and therefore they were not normalized.
A Voigt profile was used to fit the Br absorption lines in the normalized spectrum. From the fits, the equivalent widths were determined using:
where is the residual flux normalized to 1 at the continuum. The emission lines were unresolved at our instrumental resolution, and could be fitted well with a Gaussian profile. Table 6 lists the equivalent widths of the absorption and emission profiles, along with the Doppler FWHM , and the Lorentz FWHM = (where is the effective damping constant) of the Voigt profile.
Table 6. Parameters for the fits to the Br lines. The 2nd and 3rd columns list the central wavelength and the equivalent width of the absorption feature. The 4th and 5th column give the same information for the emission feature. The 6th and 7th column give the Doppler and Lorentz FWHM of the Voigt profile (accuracy 5 km s-1, except entries with a colon). The 8th column gives the total flux of the H I , He I , He II blend (the helium lines only being included where indicated), and the 9th column the flux of the He I blend. Fluxes of sources larger than the slit are marked as lower limits. The last column gives the heliocentric radial velocities (accuracy 10 km s-1). If two entries are given, they are the radial velocities from the absorption and emission line, respectively.
For IRAS 14488-5405 and IRAS 16594-4656, both absorption and emission components were present. A combination of a Voigt and a Gaussian profile, each with its own central wavelength, was used in the fit for these objects. For IRAS 16594-4656, the central part of the absorption and the emission were fitted well, but the outer wings of the absorption were not. For IRAS 16086-5255, a weak absorption line at the central wavelength of Br was observed. It could not be fitted well. For IRAS 11159-5954, and IRAS 15553-5230, no Br was seen in emission or absorption. For IRAS 16279-4757 the resulting `best fit' was so bad that no parameters are listed. The fact that the absorption lines in IRAS 16279-4757 and IRAS 16594-4656 cannot be fitted properly with a Voigt profile, indicates the presence of an additional broadening mechanism. One likely candidate is the linear Stark effect. Further investigations are needed to verify this, but if confirmed, this would indicate a high surface gravity of the central star.
For the spectra containing emission features, the absolute line fluxes were determined. In most spectra only Br was detected, but some also showed the presence of He I emission. In these spectra the the He I and multiplets would be blended with the Br line, while the He I and blend could be seen separately. It is unlikely that He II line was also blended with Br, except maybe for high excitation PNe ( K), such as IRAS 18401-1109. In the fitting procedure, it was assumed that each line had a Gaussian profile with the same (unresolved) width and that the wavelength interval between the lines was fixed. The continuum was assumed to be flat except for IRAS 14488-5405 and IRAS 16594-4656 where the underlying Br absorption was assumed to have a Voigt profile. The helium lines were only fitted when there were indications for their presence upon visual inspection. The separate flux values in the decomposition of the Br complex could not be determined accurately and therefore only the total flux of the blend is listed. The He I lines never contributed more than 8.5% to the total flux of this blend. The results of these fits are listed in Table 6. No attempt has been made to correct any of the fluxes for slit loss.
The central wavelengths of the hydrogen absorption and emission features can be used to calculate the heliocentric velocities of these objects, using the routine RVCORRECT in IRAF . The results can be found in Table 6. The literature values for the radial velocities of the standard stars (Hirshfeld et al. 1991) were compared with their observed radial velocities, yielding an average accuracy of 11 km s-1.
3.3. Radio observations
The data were reduced using the package MIRIAD following standard reduction steps as described in the reference guide by Bob Sault and Neil Killeen (http://www.atnf.csiro.au/Software/Analysis/miriad ). Images were made using the multi-frequency synthesis technique and robust weighting with a robustness parameter of 0.5. Any confusing sources were CLEAN -ed before determining the upper limits and noise in the map.
We didn't detect any of the six emission line sources above a detection limit of 0.70 mJy/beam at 3 cm and 0.55 mJy/beam at 6 cm. Our individual upper limits to the flux for the sources are given in Table 7. The weakest confusing source which was clearly detected was 1 mJy and the strongest 3.8 mJy. The map of IRAS 15544-5332 showed large scale structure, but still no source was present after deleting the two shortest baselines.
Table 7. Results of the radio continuum observations. The total on source integration time was 96 min. Columns 2 and 3 give the upper limit and rms at 6 cm, Columns 4 and 5 the upper limit and rms at 3 cm, and the last column the radio flux predicted on basis of the observed Br flux, assuming Case B recombination. The last four objects are known PNe. For these objects only the observed and predicted 6 cm flux are given.
Using the Br flux and assuming Case B recombination we calculated the expected optically thin radio flux. These values are given in the last column of Table 7. The predicted values appear to be a factor ten or more higher than the upper limits on the observed flux. Even the flux per beam of the two extended objects would have been well above the detection limit at 6 cm. For the PNe, however, the predicted values are somewhat lower than what was observed in the radio. The Br flux is probably underestimated due to extinction and slit loss. The radio flux has an uncertainty of 10% to 20%.
© European Southern Observatory (ESO) 2000
Online publication: October 30, 2000