Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 360, 562-574 (2000)

Previous Section Next Section Title Page Table of Contents

3. ISO observations and data reduction

The data discussed in this paper is the SWS full resolution scan (AOT06) of EP Aqr in the wavelength range 12-16.5 µm (AOT band 3a) observed on 4 May 1997. The data were reduced with the SWS Interactive Analysis (IA) data reduction software package using standard ISO pipeline products (version 7.0) at the ISO datacenter in Groningen. The observed wavelength range is scanned twice in different directions (the up- and down-scan). Variations in dark currents and memory effects in the detectors generally produce differences in the fluxes seen in the up and down scan. In this AOT band however, memory effects are small and dark currents are negligible compared to the high fluxes. From overlaying the up- and down-scan for each of the 12 detectors in this AOT band, it is found that differences between the up- and down-scan due to cosmic-ray glitches or signal jumps are very scarce and only present at the 2 % level. The main uncertainties in this wavelength range arise from instrumental fringing. In principle a plain division of the data with the Relative Spectral Response Function (RSRF) should correct for this, but this is not the case as the RSRF is determined from pre-flight measurements on extended calibration sources. As a consequence, the fringe pattern is not sampled at the resolution obtained for point sources, resulting in fringe amplitude differences between data and RSRF. Moreover, pointing offsets cause the observed fringe pattern to be slightly shifted with respect to the RSRF. Although there exist several tools in IA to `defringe' the spectrum after the responsivity calibration, we found that for this kind of high-resolution observations these tools are confused by the overwhelming amount of (relatively regular) strong CO2 emission lines in the spectrum. None of the present defringing tools is able to get rid of the fringes without also changing the CO2 spectrum. Therefore we chose another strategy and performed the responsivity correction interactively using the dollar; circ; lcub; lcub; hyphen;8 rcub; rcub; dollar; routine. First, data and RSRF were divided in a high number (15-20) of bins of the same width. For each of the bins that do not contain strong CO2 emission bands, the RSRF was shifted in wavelength direction until the fringes in the RSRF were coincident with those in the data. The shifts thus found are virtually the same for all bins. Next, we enhanced the RSRF in each of these bins in order to minimize the residuals after division. The enhancement factors show a slight decrease with increasing wavelength, as is expected when the resolution increases with increasing wavelength. The shifts and enhancement factors for the bins that do show strong emission bands are an interpolation between the values found for the adjacent bins. This procedure removes most of the fringes - although residuals are still present in some parts of the spectrum, especially in the 12 - 13 µm region where the RSRF drops below 30 % of the peak value. The observed emission lines however are present in both the up and the down-scan at the same flux level and not in the RSRF, confirming that the observed lines are real (see Fig. 1). The main uncertainties left are possibly broad features that influence the underlying continuum. Next, the absolute flux levels of the different detectors are aligned and finally the spectrum is rebinned. Narrow lines that are clearly present in both the up- and down-scan (see Fig. 1) allowed us to determine the actual resolution for this particular observation from the measured FWHM of these lines. We found a typical resolution of about 2800 at 15.3 µm, 40 % higher than generally assumed. We finally rebinned the spectrum with a constant resolution of 2500 specifying an oversampling rate of 1 and applied a radial velocity correction to the whole spectrum. The resulting spectrum is shown in Fig. 2.

[FIGURE] Fig. 1. A representative part of the spectrum of EP Aqr before rebinning. Closed circles are the data for all detectors in the down-scan, open squares are for the up-scan and the solid line is the RSRF for one of the detectors, scaled and shifted for the sake of clarity. The nearly perfect match between both scans illustrates the excellent quality of the data. All lines are present in both scans, and not in the RSRF, confirming that the fringes have been well removed and the lines are real. Note also the clear presence of a CO2 line at 15.32 µm (indicated by the arrow), suggesting that the SWS resolution is higher than previously thought.

[FIGURE] Fig. 2. The full AOT band 3a spectrum of EP Aqr after rebinning. Between 12 and 13 µm the RSRF drops to very low values and moreover there are no obvious CO2 bands in this wavelength range. We therefore consider the rapid variations in this part to be mainly due to fringe resiudals superposed on the continuum. The solid circles define the continuum used for the analysis of the CO2 bands. Assuming that the 13 µm feature is the only dust feature in this wavelength region, we chose only a few points longward of 13 µm in order to obtain a smooth continuum. The grey line is a spline fit through these points and is used as a continuum when comparing data and models.

The IA package provides an estimate of the uncertainties on the datapoints in the rebinned spectrum which are determined by calculating the variance of the flux values of the different detectors for each bin. However, when reducing both the up- and down-scan separately, it is found that the differences between these scans are often larger than these uncertainties. For those points where this is the case, we adopted the difference between up- and down-scan as the observational uncertainty.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 2000

Online publication: August 17, 2000