From Massa et al. 1998, ESA SP-413, "Ultraviolet Astrophysics Beyond the IUE Final Archive", eds. Wamsteker and Riestra, (ESA Pubs. Div., Noordwijk, The Netherlands), p723
To date, our investigations have concentrated on extracted, untrailed, large aperture data from all cameras in low dispersion and the SWP and LWP in high dispersion. Section 2 describes the low dispersion results, § 3 presents the high dispersion analysis, and § 4 is a summary.
The analysis is performed on the IUE standard stars listed in Table 1 (Pérez et al. 1990) using standard IUEDAC software. The Table lists the star name, spectral type, photometry, and the number of well exposed NEWSIPS spectra used in the analyses of both low and high dispersion data.
Low dispersion standards
High dispersion standards
Figure 1: Time dependence of 237 LWR NEWSIPS spectra of the 3 low dispersion standards in Table 1. Each standard is normalized by its mean and then interleaved in time. The data were placed on a linear time grid using nearest neighbor interpolation, with dark gaps appearing whenever 30 days past between spectra.
Figure 2 shows this time dependence integrated over the wavelength band 2400< lambda < 2800Å. Each point is the mean flux across the band normalized by the mean value for that star. It is clear that flux values for the same source change by roughly 10% between 1978 and 1984.
Figure 2:Plot of LWR fluxes across the band 2400< lambda< 2800Å. Each point is the mean flux across the band for a star normalized by the mean value for that star.
We also examined the time dependence of 930 well exposed, untrailed large aperture SWP NEWSIPS spectra of the 3 low dispersion calibration standards listed in Table 1. We found that even the most recent NEWSIPS spectra (version 2.5.2 for low dispersion) still contain some residual time dependence shortward of Ly alpha (up to 10% near the end of the mission). Grey scale plots of the SWP data also showed that the wavelength alignment is inconsistent, changing with time at the shortest wavelengths. However, wavelength calibration has always been problematic for lambda < 1250Å, due to a paucity of calibration lines.
We next turned our attention to the veracity of the modeled standard deviations provided in the MXLO files. For each of the low dispersion standards listed in Table 1, we determined the actual standard deviations and compared these to the square root of the quadratic mean of the MXLO errors. The results are shown in Figure 3. This figure shows that; the ratios agree from one standard to the next and, the MXLO error model systematically underestimates actual standard deviations by up to 40%. However, this is not considered a very serious effect, since the relative weights agree fairly well, and the distinction between 1 and 1.4 sigma is marginal in most practical situations.
Figure 3:: Ratios of observed to MXLO modeled standard deviations for the low dispersion standards listed in Table 1. All 3 standards are overplotted, with line styles given by the key.
We should also point out that Fitzpatrick & Massa (1998) have compared the NEWSIPS and HST/FOS (Bohlin 1996) absolute flux calibrations and found that they differ by as much as 15%. We emphasize that their transformation was derived from hundreds of spectra of several blue stars observed by both satellites, and that it is not known whether it applies to objects with very different energy distributions.
Figure 4: Time dependence of 71 SWP high dispersion NEWSIPS MXHI spectra of tau Sco obtained over the lifetime of IUE -- only the short wavelength region is shown. The spectra are normalized by their mean and the temporal ordinate is simply the accumulated number of spectra.
Figure 4 shows the time dependence of 71 large aperture high dispersion NEWSIPS MXHI spectra of tau Sco. Several aspects of the figure are worth noting:
Figure 5 examines the temporal dependence more quantitatively, by comparing the total observed changes in eta UMa and zeta Cas. This plot demonstrates the following points:
Figure 5: Time dependence of SWP NEWSIPS large aperture extractions of eta UMa (solid) and zeta Cas (dotted). The curves are ratios of 10 spectra means from 1993-94 divided by 10 spectra means from 1981-82 for each star (all binned to 0.5 Å/point).
Figure 6: Time dependence of SWP NEWSIPS (solid) and IUESIPS (dashed) extractions of large aperture spectra of eta UMa. Each curve is a ratio of a 10 spectra mean obtained in 1993-94 divided by a 10 spectra mean from 1981-82 (all binned to 0.5 Å/point).
In the process of analyzing the data, we also noticed the following general anomalies in MXHI extracted spectra:
Figure 7: Time dependence of 110 LWP NEWSIPS large aperture high dispersion extractions of eta UMa obtained over the lifetime of IUE. Only a portion of the wavelength coverage is shown, and the display is the same as for Fig. 4.
Figure 8 displays the LWP time dependence of zeta Cas (solid) and eta UMa (dotted) spectra and is similar to Figure 5 for the SWP spectra. As in Figure 5, the ripple pattern is present, but the effect is quantitatively smaller. Additionally, as in Fig. 6, the systematic effects can dominate the random noise.
Figure 8: Time dependence of LWP NEWSIPS large aperture extractions of zeta Cas (solid) and eta UMa (dotted). The curves are ratios of 10 spectra means from 1993-94 divided by 10 spectra means from 1981-82 for each star (all binned to 0.5 Å/point).
The high dispersion anomalies probably result from compromises needed to make the difficult process of echelle spectral extraction automated. Nevertheless, they underscore the necessity of making the unextracted data products easily accessible, so that users can apply customized extractions as warranted.
Further details and updates can be obtained from
|ADF IUE Homepage||Questions/Comments|
Web Curator and Responsible NASA Organization/Contacts