4.1. Null Drift

It is essential to know whether the changes in the background fluxes shown in Figs. 7 are characteristic of only certain classes of astronomical sources or whether they are symptoms of a more general problem affecting all SWP data. To address this question, we selected for study a set of 27 null images distributed in time. Null images were taken during the IUE mission for two reasons: first, to facilitate the study of degradation with time of camera response to illumination and background particle radiation and, second, to permit the construction of the Intensity Transfer function, or ITF. The ITF is a generally nonlinear conversion function between the natural Data Number units given by the camera's 8-bit Digital to Analog Converter and FN units proportional to incident flux.⁴ For the IUE cameras the ITF is unique to each pixel. The null images provide the input data for the determination of the first (zeropoint) of 14 exposure levels. For a given pixel, the mean of the raw fluxes (Data Numbers) from nulls of a common epoch define a zero-flux level which is the starting point (``toe") of the ITF. In practice, the fluxes of an extracted group of ``background pixels," even uncontaminated by interorder fluxes, may be nonzero for several reasons: (1) the geometric correction (shift) NEWSIPS computes is a complicated function of illumination (Linde & Dravins 1990), (2) both the pixel positions and ITF change with time and camera-head temperature, (3) the electron charge measured by the camera-read beam at a given exposure can vary slightly in time, and (4) the independent variable (Data Numbers) in the ITF construction is strongly quantized.

Typically, the nulls are used in low-dispersion analyses of the change in camera sensitivity with time. For this reason the NEWSIPS archiving-pipeline system processed them as low-dispersion images. Thus, special NEWSIPS were required to generate SIHI (high dispersion) images from the raw data. These computations were kindly carried out for us by Dr. Catherine Imhoff for this study for 27 null images listed in Table 3. These images were stacked into a data cube, permitting the computation of mean regression slopes and rms dispersion values along the time axis for each pixel in this cube. These statistics permitted us to search for trends in time of FN values for all pixels in unexposed images. For example, Figure 6 already suggests that the background fluxes diminish with time, particularly after 1991. Accordingly, we present the SWP background surfaces by three surface plots in Figure 8 representing the mean total change in FN units per year per pixel, first, prior to 1982.0 (a), then prior to 1989.0 (b), and finally after 1990.0 (c). A fourth panel shows the spatial dependence of the rms values for the entire mission lifetime of IUE (1978-1996). This is useful as a check for the relative changes over the surface during the mission lifetime. The first three panels show that even early in the mission lifetime the FN level corresponding to zero flux decreased most rapidly in the short-wavelength corner of the image, which appears in the left foreground of each surface. Further inspection of surfaces of individual null images shows that the zeropoint changes in the ``short-wavelength region" of the SWP were already discernible in 1979. As the camera aged, the progress of this decay halted in the short-wavelength corner (second panel) but then in the early-1990's spread to higher sample positions across the short-wavelength region (third panel). The overall drop from 1979 to 1996 in the background flux solutions amounts to 23.5 FN units at the position of Lyman $\alpha$ on the image. This figure is in good agreement with the net difference, $\approx$28 FN, in derived background fluxes found at the position of the Lyman $\alpha$ feature in Figs. 6 and 7.

SWP Image Year (1900+) SWP Image Year (1900+)

01216 78.2 33001 88.4

01792 78.5 34492 88.7

03929 79.0 36634 89.4

07396 80.0 38341 90.2

10565 80.9 40167 90.8

14283 81.5 41982 91.5

16362 82.1 44495 92.4

19941 83.3 46750 93.0

23044 84.4 50462 94.3

25012 85.1 51158 94.5

26947 85.8 54278 95.3

29280 86.7 56580 95.9

31026 87.3 57368 96.4

32656 88.1

Figure 8: Surface plots representing the spatial degradation of zero-fluxes of null SWP images. The short wavelength corner of the image is the left foreground. Vertical axis represents the mean change per year in flux (FN units) for a given pixel across the timeline represented in the plot (Panel a: 1978-1981, Panel b: 1978-1988, Panel c: 1990-1996). Panel d represents the mean rms of fluctuations of fluxes within the spatial bin sampled.

Ironically, the size of the drop in background fluxes from the long- to short-wavelength (spatial) end of the camera is largest in the late-epoch images. However, at no position along the Pass 1 swaths is the background flux gradient as large in late-epoch images as it is over the small distance between orders 113 and 125 in early-epoch images (cf.Fig 7). Thus, although the changes in the FN zeropoint across the entire image is large at late epochs, a 7th degree Chebyshev solution is able to accommodate the required slope. Moreover, the monotonic decrease of background flux with increasing echelle order makes it more likely that the BCKGRD solutions of late-epoch images will pass the pathology tests described in $\S$2.2., thereby permitting the Pass 1 Chebyshev fits to follow the background surface accurately. For these reasons background fits near the Lyman $\alpha$ line are actually comparatively good for late-epoch images.

Although Figure 8 shows a drop in the null-flux for short-wavelength areas of the SWP camera, but we have not yet discussed whether this anomaly is sensitive to wavelength alone or one which happens to occur in the area of the SWP camera image where short wavelength flux happens to fall. One can distinguish between these possibilities simply by inspecting difference images between early- and late-epoch spectra. Figure 9 exhibits the difference between SWP01216 and SWP57368 at epochs 1978 and 1996. To aid in orientation we have superposed on the difference surface the first of these images. One sees that the changes in null level are largest at the left lower edge (short-wavelength corner) and the right lower edge of the image. This image shows clearly that the change in the null surface arises from spatial and not wavelength characteristics of the camera detector surface. Comparison with other image pairs show this same property. Such difference images show that the changes in the zero level are more likely to be associated with changes in the camera electro-statics near the target edges. Our inspection of a number of SIHI and raw images from both the LWP and LWR cameras shows the same tendencies for latent static discharging from the edge of the camera image to its surrounding casing increased over the IUE mission lifetime. For example, we found a crescent-shaped enhancement in the background in LWR images dating to the commissioning period of the IUE mission. This enhancement subsequently increased in amplitude and in 1983 developed into the full-fledged ``flare" which curtailed use of the LWR camera.

Figure 9: Grayscale difference image between the early and late-epoch ``null" images SWP01216 and SWP57368. Dark tones indicate a rapid decrease in FN level with time; white indicates no degradation. An SWP echellogram of $\tau$ Sco is superimposed for reference. The horizontal arrow shows the position of the Lyman $\alpha$ absorption feature. The pair of horizontal bars shows where the SWP camera response stops. ``Uncontaminated" short-wavelength background fluxes may be sampled only below this boundary.