Astron. Astrophys. 318, 472-484 (1997)

3. Speckle observations and data reduction

The data discussed in this paper were obtained over a period of about ten years, using different instruments at several telescopes in both hemispheres, as listed in Table 2. Here we only provide a basic description of the observing techniques and of the data reduction methods. More details can be found in previous dedicated papers (Haas 1990, Christou 1991, Haas et al.  1993).

Table 2. Log of observations

The instruments that we used recorded either 1-D or 2-D speckle data. To the first class belongs the Calar Alto specklegraph (Leinert and Haas 1989), based on a single InSb photodiode, with a scanning slit that can be oriented at any chosen angle on the sky. The results are obviously one-dimensional, however two orientations are in general sufficient to obtain the true position angle and separation of a binary, while the interpretation can be more difficult in the case of an extended source. The field of view of the 1-D measurements is characterised by the scan length and by the height of the slit, both of which are about 10 arcsec.

To the 2-D instruments belong the Calar Alto InSb (Lenzen et al.  1990) and Nicmos (Herbst et al.  1993) cameras (the latter dubbed MAGIC), the Kitt Peak InSb camera (Beckers et al.  1988) and the ESO NTT Nicmos camera (dubbed SHARP, Hofmann et al.  1993). In this case the 2-D structure of the source is preserved, making the observations more efficient, but with the possible disadvantage that the data reduction is burdened by the need for flat-fielding and bad pixel correction and by the increased computational needs due to the two-dimensional Fourier treatment. The format of the cameras ranged from 58 62 pixels for the InSb arrays to 256 256 for MAGIC and SHARP, which however are limited to the 1-2.5µm range. With MAGIC and SHARP data taking was often restricted to one 128x128 pixel quadrant or to a format of 128X256 pixels, combining simultaneous object and sky measurements. The field of view for the InSb cameras usually was x or x , while a quadrant size of the NICMOS cameras was x for SHARP and x for MAGIC.

The technique employed was that of speckle interferometry, by which many short exposures images (or scans in the 1-D case) are obtained. The integration time was chosen as a compromise between the coherence time of atmospheric turbulence at the given wavelength (typically 0.05s at 2µm) and the minimum time necessary to achieve a satisfactory SNR for the source on a single exposure. For the fainter sources (or in case of bad seeing) the integration time was taken a few times longer (typically 0.2-0.5s), which improved the signal but resulted in some loss in the high spatial frequency resolution. The typical observing sequence consisted of a total of frames for a given source, broken into blocks of frames each, which were interspersed with a comparable number of frames on one or more nearby reference stars, i.e. stars which are presumably point-like and which, because of their position in the neighborhood of the object, are aberrated by similar atmospheric turbulence. Sky exposures (also useful for the flat-fielding necessary for the 2-D instruments) were usually obtained next to the object and reference measurements.

The data reduction typically consisted in the reconstruction of the source visibility according to the classical power spectrum analysis (Labeyrie 1970), and additionally of the phase by means of the Knox-Thompson (1974) method or the bispectrum formalism (Lohmann et al.  1984). The number of "subplanes" used in the bispectrum reconstruction was typically several tens. From the resulting complex visibility it is possible to reconstruct a (nearly) diffraction limited image of the source, which could be analysed for objects in the circumstellar environment. However, it is more convenient to fit directly the complex visibility by the Fourier transform of model images (see for instance the example given in Fig. 4). Our fitting procedures allow to choose a number of point sources and/or sources with a gaussian profile (these latter are useful in the presence of extended emission but were not needed for the evaluation of the binaries in our sample). The parameters resulting from the fits are separation, position angle and brightness ratio of the components. For some sources we have this information at several wavelengths and thus obtain the spectral energy distributions (SEDs) for each component separately. Following usual practice we denote the component brighter at K as 'A' and the fainter one as 'B'.

For the purposes of this study we consider two infrared sources to form a binary if their projected linear separation is less than 3600 AU. A value of 1800 AU appeared appropiate for the less massive T Tauri stars (Leinert et al.  1993) because at this separation confusion with foreground (or also background) sources started to become noticeable. The Herbig Ae/Be stars of our list are on the average four times as distant as the Taurus sources. We therefore expect roughly the same number of foreground sources if we double the allowed projected separation. This increased maximum separation is also an approximation to the effect that the more massive Herbig Ae/Be stars are gravitationally dominating a larger volume than the less massive T Tauri stars.

© European Southern Observatory (ESO) 1997

Online publication: July 8, 1998
helpdesk.link@springer.de