2. Observation and data reduction
The 21 cm absorption survey has been done toward 20 compact continuum sources, which have been selected from our ATCA snapshot survey at 1.4 GHz (Marx et al. 1997). The sources are mainly in directions near the 30 Doradus complex, the H supergiant shell LMC 4, and in the direction of the sharp H I edge at the eastern boundary of the LMC. These regions are illustrated on the H image of the LMC in Fig. 1, together with the location of the continuum sources. In Table 1 the positions of the sources are listed, Columns 2-5. The X and Y coordinates (Columns 4 and 5) are offsets in degrees toward the centre of the rectangular grid defined by Isserstedt (1975). The peak flux densities, which are between 21 and 80 mJy, are given in Column 6 (uncorrected for primary beam attenuation). The noise in the optical depth, , is shown in Column 7. The locations of the background sources with regard to the prominent structures are given in Column 8. "Others" are lines of sight, which we use as a reference sample, as they are far distant from the 30 Doradus region, LMC 4 and the eastern H I boundary. Source names (Column 9) are taken from our catalogue of compact continuum sources (Marx et al. 1997). Source No. 18 has not been selected from this snapshot survey, but was within the primary beam of one line of sight at the far east during the H I absorption study.
Table 1. Background sources
The spectral line observations were made in May 1994 using the single 6 km configuration (6D) of the ATCA. The resulting synthesized beamwidth is 7". A spectrometer configuration with 4 MHz bandwidth divided into 1024 channels was used. This gives a channel separation of 3.9 kHz = 0.825 km s-1. The central frequency was 1419 MHz, which translates to a centre velocity of = 283 km s-1. We have integrated two to seven hours on each source, depending on its continuum flux density, with individual scans of 10 min each, spread over various hour angles to give the best uv-coverage. This provides an optical depth sensitivity between 0.1 and 0.25. As primary flux and bandpass calibration source we used B1934-638, the secondary (phase and gain) calibrator was B0407-658. The calibration was done using the AT version of the Astronomical Image Processing System (AIPS). To reduce the noise in the spectra we averaged together every two channels using the AIPS task `AVSPC'. The final spectra have the same velocity resolution of 1.65 km s-1 as those of survey 2. The absorption spectra were computed using the task POSSM with vector averaging, shifting to the centre position of each source as computed from the continuum maps. A linear baseline was fitted to the velocity ranges 43-108 km s-1 and 439-504 km s-1, which are free of emission and absorption from either Galactic or Magellanic H I . Only baselines longer than 3 k (796 m) were used, so as to filter out emission features as far as possible. Since POSSM gives the resultant amplitude and phase of the vector averaged complex visibility function it is necessary to correct for the noise vector as described in Thompson et al. (1986, Eq. 9.46-9.52). For this correction the approximations of Dickey et al. (1994) have been used.
By rejecting data from baselines shorter than 3 k we are effectively filtering the image of the LMC to pass only the higher spatial frequencies. The equivalent largest angular size to which we are sensitive is 70", or 17 pc at a distance of 50 kpc. The background sources are much smaller than this, so the absorption spectra are not effected by the cutoff, except to slightly raise the noise because of the rejected data. However, including shorter baselines would pollute the spectra with "pseudo- absorption" (Radhakrishnan et al. 1972), which is a figment of small scale structure in the emission. We can estimate the level of such pollution from the study of the spatial power spectrum of the HI emission in the SMC by Stanimirovic et al. (1999). Although emission fluctuations are not detected at such long baselines, if we assume that the power law dependence of amplitude vs. baseline length seen for shorter baselines continues to such small angular sizes, then the maximum amplitude expected for the emission fluctuations would be well below 100 mJy. Since these fluctuations occur with random phases all over the primary beam, they integrate down toward a mean of zero. The absence of positive features in the absorption spectra shows that, with our cutoff, confusion by emission fluctuations is not a problem. Typically, the amplitude of the emission fluctuations is proportional to the overall amplitude of the emission (unless the optical depth is very high, which never occurs in the Magellanic Clouds except perhaps in 30 Dor itself). Thus regions of low emission brightness are even less subject to confusion in this way than the regions of higher HI line strength.
© European Southern Observatory (ESO) 2000
Online publication: February 25, 2000