2. Observations and data reduction
A full scan observation of Tau has been obtained in the 2.38-45.2 µm wavelength range with the Short-Wavelength Spectrometer (SWS) (de Graauw et al. 1996) on board ISO. This has been done by use of the SWS observing mode AOT01 (= a single up-down scan for each aperture/order combination) with scanner speed 4, resulting in a resolving power of 1500. The observation lasted for 6650 s and was performed during revolution 636.
The data were processed to a calibrated spectrum sampled in all the pixels, the so-called Auto-Analysis product, using the procedures and calibration files of the ISO off-line pipeline version 7.0. For a description of the flux and wavelength calibration, we refer to Schaeidt et al. (1996) and Valentijn et al. (1996). This reduction resulted in an oversampled spectrum for each of the four AOT-bands, which each consist of 12 detectors.
Remaining instrumental effects, such as fringes in the 12.0-29.5 µm part of the spectrum, were removed using the SWS Interactive Analysis package (IA) provided by the SIDT. The band-2 data (4.08-12 µm) are severely affected by detector memory effects. Time and flux dependent memory effects occur for all four instruments (SWS, LWS, PHOT, CAM) on board ISO. So far, several attempts have been made to model the corrections of the transient effects (see e.g. Van Malderen et al. 1999), but currently no foolproof method exists to correct the SWS data for these transient effects. Memory effects appear to be less severe in the down-scan measurements, which are obtained after the up-scan data, suggesting a more stabilized response to the flux level for down-scan data. The SWS flux calibration of band 2 was corrected for a regime where the flux history of the detectors is comparable to the second scan in the observation, the so-called down scan. We therefore used the down-scan data of our observation as a reference to do a correction of the flux level of the first scan.
Several procedures were performed on each sub-band separately to combine the twelve detectors. A sigma-clipping procedure (with ) was used to reduce the noise by discarding the datapoints in every resolution element for which the difference to the mean was . After aligning the twelve different detector signals to their average level and rebinning to the expected resolution (see Table 1), the final result was obtained. Although the resolution changes in within one sub-band, the resolution was kept constant (see Table 1) and the value was taken to be the lowest value in Fig. 6 by Lorente (1998). The individual sub-band spectra, when combined into a single spectrum, can show jumps in flux levels at the band edges. This is due to imperfect flux calibration or wrong dark-current subtraction for low-flux observations. Using the overlap regions of the different sub-bands and looking at other SWS observations, several sub-bands were multiplied by a small factor (see Table 1) to construct a smooth spectrum. Due to the problems with memory effects in band 2, the factor of band 2B was determined by use of the template of Cohen et al. (1992) and the obtained synthetic spectra. Note that all shifts are well within the photometric absolute calibration uncertainties claimed by Schaeidt et al. (1996).
© European Southern Observatory (ESO) 2000
Online publication: December 15, 2000