The 7.45-7.9 µm ISO-SWS spectra of W 33A and C 7538 : IRS9 were published in Boogert et al. (1996). Due to a scanner problem in the region short-ward of 7.45 µm, the spectrum of W 33A was re-observed in revolution 332. The old and new spectra agree very well, although the individual detector up and down scans of the new observation overlap better. Perhaps the cosmic ray flux was lower in revolution 332. Nevertheless, some of the weak gas phase lines reported here, can be recognized in the old spectrum (notably the Q-branch and R(3) lines; Boogert et al. 1996; Sect. 3). The spectra of C 7538 : IRS9 and W 33A were (re-)reduced with the SWS pipeline of June 24 1997. The standard wavelength calibration was applied (Valentijn et al. 1996) and corrections for spacecraft and source velocities were made, using and -60 km s-1 for W 33A and C 7538 : IRS9 respectively (Mitchell et al. 1990). Each of the 24 detector up and down scans of both spectra was inspected on detector jumps and excessive noise levels. Within the wavelength range considered here (7.35-7.85 µm) no deviating scans were found. Systematic differences in the flux scale and slope of the scans were corrected for by fitting third order polynomials to each scan and using these fits to shift the data points per scan to the mean of all data points (`flat-fielding'). These systematic differences are most likely caused by dark current variations due to detector memory effects in this wavelength range. Thereafter, data points deviating more than 2.7 sigma from the mean per resolution element were removed. These points are mainly caused by (minor) cosmic ray hits, that were not recognized in the SWS pipeline. Finally, the up and down scans were separately convolved with a Gaussian to , and rebinned to two points per resolution element. Although the instrumental resolving power of the SWS grating spectrometer is at this wavelength, we found that smoothing to removes the high frequency fringes seen in many spectra (see e.g. Fig. 1 of Boogert et al. 1996). The frequency of these fringes is the same as the fringes in the detector responsivity, although the amplitude is a factor larger in the interstellar spectra. This is explained by the low resolution and under-sampling of the current SWS responsivity tables. In order to check our reduction method, we have reduced a spectrum of the standard star Lyrae in exactly the same way. We find that none of the absorption features reported in this paper correlates with features in this standard spectrum. The final spectra, i.e. the average of the up and down scans, show point-to-point variations at the level of 0.3% of the continuum. However, on a larger scale the separate up and down scans deviate up to 1.5% from each other. These systematic differences are probably caused by (residuals of) cosmic ray impacts and subsequent dark current variations. Since the continuum is uncertain at this level, we adopt the difference of up and down scans divided by 2 as the 1- uncertainty for the line depths (Fig. 1).
© European Southern Observatory (ESO) 1998
Online publication: July 7, 1998