2.4. Problems in the Short-Wavelength Region

  Many things can go wrong in an automated computation of a background solution. BCKGRD was designed to provide a viable solution in the most pathological cases, rather than permitting ringing in the solutions or a halt of the processing altogether. As indicated in Fig. 2, Pass 1 swaths which sample fluxes in the vicinity of Lyman $\alpha$ have fewer useable pixels than at other image locations because of the circular target ring, which constricts its length, and the large proportion of pixels which have severely contaminated fluxes. The problem is exacerbated by two considerations. First, the circular target ring cuts off the spectrum at the ends of short-wavelength orders. Pass 1 extraction swaths terminate at the target ring in the short-wavelength corner of the camera, so there is no unilluminated region to sample. Second, the echelle grating blazed is shifted to the short-wavelength edge (left in Fig. 2) within the target format. These geometric considerations are ultimately responsible for an unsolvable problem for background determinations in arbitrary high-dispersion SWP images: contamination-free pixels do not exist near Lyman $\alpha$ (the fifth vertical swath from the left in Fig. 2). Thus, no scheme can measure a truly uncontaminated background level in this region of the camera for a spectrum dominated by continuum flux. This conclusion pertains also to so-called customized re-extractions from the SIHI image which one might make in an interactive data-reduction environment. We remind the reader than IUE camera dark surfaces (which are nulls for short exposures) changed from exposure to exposure, either because of phosphorescence in the camera ultraviolet-converter (Coleman et al. 1977) or because of changes in particle radiation as the spacecraft moved through the Van Allen belt. In principle, a second readout immediately following a science exposure might have provided a handy instantaneous null surface for purposes of determining a better background. However, the overhead on satellite time would have made this procedure prohibitive, so it was almost never followed.

The absence of pixels containing reliable background fluxes in the Lyman $\alpha$ region of the SWP camera image often results in the failure of BCKGRD's first pathology test for the first swath and sometimes inaccurate Pass 1 solutions for the first three swaths. These particular extractions may result in a constant or even monotonically increasing background solution towards short wavelengths (spatial direction). In contrast, swaths taken through a more central part of the image and which are not intercepted by the target ring, show that the background fluxes should decrease at the short-wavelength (spatial) end of the swath because the detector response is low below $\sim$$\lambda$1200. Figure 5a shows the example of a final (Pass 2) background solution for the order containing Lyman $\alpha$ for the image SWP04262. Star symbols represent Pass 1 solutions either at, or constant-flux extensions to, the line position of this order. Notice that the leftmost Pass 1 value attains a comparatively high flux level. This is because no pixels uncontaminated by interorder flux can be sampled at the low-wavelength end of their Pass 1 swaths. BCKGRD attempts to correct for this problem (and only for the short-wavelength corner of SWP images) by interpolating to fluxes sampled by the fourth swath, which does reach uncontaminated pixels). Experience shows that this action tends to undercorrect the ``flux turnup" problem in this region of the camera for spectra of hot stars. We note that the BCKGRD assumes explicitly that maximum fluxes (within a tolerance criterion set by experimentation) should not be reached for the first few swaths in the Pass 1 operation. This expectation is valid for images at most epochs, but as we demonstrate below that it may not have been accurate for early-epoch images.

Figure 5: (Panel a): A comparison of the final background flux solution for order m = 113 in SWP04262, a spectrum of $\tau$ Sco taken early in the IUE mission lifetime. Star symbols represent the solutions from Pass 1 at the intersection of this order, while diamond figures are those intersections from the first few Pass 1 swath solutions which falsely rise near the short wavelength corner of the image. The Pass 1 solution is ordinarily not saved in NEWSIPS processings and was recovered for this figure from a special processing.

(Panel b): A comparison of the final background solutions with the Pass 1 for SWP35620, a spectrum taken of $\tau$ Sco midway in the IUE mission. Diamond points in this and the next panel fall off scale.

(Panel c): A comparison of the final background solutions with the Pass 1 for SWP55997, a spectrum taken of $\tau$ Sco late in the IUE mission.