3.1. Time Degradation of Background Fluxes Near Lyman $\alpha$

  Our strategy for determining the cause of time-dependence of Lyman $\alpha$fluxes in $\tau$ Sco was to examine the the Pass 1 solutions for a subsample of 23 large-aperture images distributed in time from Massa's larger sample (Massa, priv. commun.). Table 2 lists these images and their observation epochs. To minimize the possible variables each of these images was taken through the Large Aperture.²

 

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

04262 79.1 39192 90.5

08683 80.3 39400 90.7

10255 80.8 41319 91.3

14425 81.5 42009 91.5

19860 83.2 44007 92.2

20500 83.6 47088 93.2

23379 84.5 50499 94.3

26512 85.6 51380 94.5

31369 87.5 54438 95.3

33884 88.5 55238 95.5

35620 89.2 55997 95.7

38467 90.2

To obtain the critical Pass 1 solutions of the images in Table Table 2, we processed them with a prototype version of NEWSIPS which stores intermediate data files. The IUE Project extensively tested the archiving-``pipeline" NEWSIPS software by byte-to-byte comparisons of output files with corresponding files produced by a prototype version of NEWSIPS. ³ In order to confirm that the solutions in the MXHI files are consistent with their Pass 1 solutions, we compared the final background fluxes along the echelle order containing Lyman $\alpha$, m = 113, for a number of images with the intermediate (Pass 1) results. The three panels of Figure 5 show three such comparisons for images of $\tau$ Sco taken early, midway, and late in the IUE mission. This figure brings out two important points. First, the close consistency of the Pass 1 and 2 solutions demonstrates that aberrations in the determined background fluxes are likely to be caused either by circumstances which largely predetermine Pass 1 results or by Pass 1 itself, and not by any post-Pass 1 operation. Second, although the early- and mid-epoch solutions decrease monotonically towards longer wavelengths (sample positions), the late-epoch solution shows a different behavior. Both the Pass 1 and 2 solutions for SWP55997 show a steep rise at the short-wavelength end of order m = 113. This feature is replicated for several adjacent orders in this image and also for the same orders in other late-epoch images. This fact suggests that the underlying background surface changes over time in this area of the camera and that BCKGRD has been able to follow this change faithfully.

To verify this conclusion, we stacked the Chebyshev fits for all 26 Pass 1 extractions for the 23 SWP images listed in Table 2. We then determined the evolution in background flux during the mission lifetime. The background fluxes at three positions along order m = 113 are plotted in Figure 6a. The positions chosen were samples 100, 150, and 250 in m = 113, which roughly correspond, respectively, to the short-wavelength edge of the order (near the target ring), to the Lyman core wavelength, and to $\sim$3 Angstroms to the red of the core where the background continuum in the SIHI image is at a maximum. In Figure 6b we show the degradations of change of derived background fluxes for the sample position value, 157, corresponding to the Lyman $\alpha$ core but at the shortest-wavelength order, m = 125. The primary result from these determinations is the secular decrease in background fluxes. The correspondence between these solutions in both plots is quite good. The agreement between the two passes (Fig. 6b) again tells us that the final backgrounds are essentially those determined from Pass 1 (Fig. 6a). The morphological similarity between these two plots, as well as with Fig. 1, suggests that the accelerated decrease starting at time 1990-91 is robustly determined. Similar plots for neighboring orders show the same result. The fact that qualitatively similar results obtain for both a well-exposed and much-underexposed order (m = 125) in Figure  Figure 6b suggests that neither exposure levels nor instrumental conditions cause the temporal degradation of the far-UV background fluxes. We will return to this point in $\S$4.

Figure 6: (Panel a): The decrease of derived background fluxes with time from ``Pass 1" NEWSIPS (BCKGRD) processing for a sample of 23 SWP high-dispersion images of $\tau$ Sco for the echelle order containing Lyman $\alpha$. Solid, dotted, and dashed lines represent solutions obtained near sample positions 100, 150, and 250 pixels, respectively.

(Panel b): The decrease in determined background fluxes for a point in the shortest-wavelength echelle order (m = 125) following Passes 1 and 2 of BCKGRD.