Astron. Astrophys. 327, 1123-1136 (1997)

2. Observations, data reduction and first results

The observations were conducted at the ESO 3.6 meter telescope in Chile (La Silla) during November 1994 using the ESO mid-infrared camera TIMMI (Thermal Multi Mode Instrument) built by SAp (Service d'Astrophysique, CE Saclay) (Lagage et al. 1993). Pictoris was observed during three nights with the smallest pixel scale of TIMMI, 0.33 arcsec, and with the 10.5-13.3µm band-pass filter. In order to benefit from the full pixel resolution, we have aligned the disk orientation with one of the axis of the detector, by rotating the telescope adaptor. The usual chopping and nodding techniques were used; the chopping and nodding frequencies were respectively 6 and 0.01 Hz. To avoid the saturation of the detector by the huge ambiant photon background, the image elementary integration time was set at 7.7 ms. The elementary images were coadded in real time during 20 chopping cycles. A total time of 300 min was spent on the source, which was always observed below 1.3 airmasses. The peak signal to noise ratioprior to any filtering is around 200. The nights were photometric; the number of counts recorded by the detectors varyied by less than 1 percent from night to night. The nearby reference star Car was frequently monitored. The photometry is in full agreement with previous measurements (1.1 Jy at 11.9µm for the disk plus star in a 4 arcsec beamin diameter). The images were flat fielded by using sky images with a sligtly different integration time. The array is quite homogeneous and the flat field corrections are always below 20 percent. The reference star Car is also used to remove the stellar contribution to the observed flux, as in LP94.

At 12µm, images are strongly degraded by diffraction, seeing and noise. To get the best angular resolution possible, we had to deconvolve (restore) the images. The Point Spread Function (PSF) was derived from observations of the reference star ( Car). There are various methods to restore images. We have selected the new Multi-Scale Maximum Entropy Method developed by Pantin and Starck, 1996, because it proved to be sligthly better in general than the other ones. In the case of the deconvolution of the Pictori s dust disk at 12µm, the method does not really bring determinant improvements and similar results were obtained with standard deconvolution methods, like the regularized Richardson-Lucy algorithm with noise suppression using a multiresolution support (Richardson, 1972, Lucy, 1974, Murtagh et al., 1994; Starck et al., 1994) or the popular Maximum Entropy restoration (Gull et Skilling, 1984). The deconvolved image obtained here has a resolution of about 0.33 arcsec (from deconvolution experiments of two PSFs) and is in agreement with the one shown in LP94. The error bars have been derived from the error estimated on the PSF ( 5 %) and the subsequent error on the stellar subtraction. Nevertheless, one should note that the resulting error in the central part of the disk (and also the later derived dust density) does not exceed 40 %. Note also that the profiles derived by LP94 and the new one are very close, when the error bars are taken into account (see Fig. 2).

Fig. 2. Flux profiles obtained with the new data compared with those of LP94 (diamonds).

The asymmetry S-W/N-E detected in LP94 is confirmed (Fig. 3), even if the new values are in the lower range of the uncertainties of the previous measurements (Pantin et al., 1995). One should note that the asymmetry is already seen in the "raw" images before deconvolution, but the very conservative errors we took concerning the deconvolution process lead to large error bars in Fig. 3. This may lead one to think that the asymmetry is only marginal, but this is not the case.

Fig. 3. S-W/N-E asymmetries as a function of the distance to the star

As the observations of November 94 are better than the earlier one in terms of angular resolution and signal to noise ratio, the disk is resolved in thickness (Fig. 4).

Fig. 4. The equivalent width (in AU) of the disk as a function of the distance x to the star (in AU). The thickness is defined as w(x)= , where z is the direction perpendicular to the disk plane.

© European Southern Observatory (ESO) 1997

Online publication: April 6, 1998
helpdesk.link@springer.de