Several factors first came to mind that might explain what was seen. We mention them here because each appeared to fall short of fully explaining all the observations.
Image Motion at the Spectrograph Aperture
Small motions of a star occur when it is observed with HST because of jitter and telescope "breathing." Jitter is usually very small in magnitude and random in effect, and is not expected to lead to systematic throughput changes. Telescope breathing, however, can lead to systematic changes. This would be especially apparent if the star were observed in the small aperture (SSA) of the GHRS, as was the case for U Gem in program 6075. However, the wavecals are immune to any effects of image motion, and so this hypothesis was rejected as the sole explanation for the drops in both the stellar spectra and wavecals.
Variations in Lamp Output
Wavelength calibration lamps can be very sensitive to the conditions under which they are run, and small changes in voltage or current can lead to changes in the lamp output. Thus if the calibration lamp alone were seen to vary slightly, these effects would probably be invoked to explain it. However, such changes cannot also explain the changes seen in the stellar spectra, and again this hypothesis was initially rejected.
Spectrum Drift within the Detector
A drift of the spectrum on the diode array could lead to a decrease in observed flux over time. Two possible source of drift are geomagnetic (GIMP) and thermal effects.
For these observations, G140L and the GHRS Side 1 detector were used. A spectrum with G140L goes across the Digicon diodes at a fairly steep angle, to the extent that one end of the spectrum differs from the other by about 45 deflection units (the diodes are 64 deflection units high) *1. That means that modest motions of the spectrum could lead to throughput changes as one end or the other of the spectrum gradually fall off the diodes. However, such a change should lead to an obvious decline in the strength of the comparison lines at one end of the spectrum relative to the other and not an overall decline in counts. No such effect is seen (the counts decline evenly from one end of the spectrum to the other).
The Earth's magnetic field can lead to motions of the image of the spectrum within the detector, and that would correlate with orbital position, leading to an effect which might look like what is seen. However, the geomagnetic effect on Side 1 is small, totalling no more than a couple of deflection steps over an orbit.
If the spectrum were drifting in y (perpendicular to the direction of dispersion), then the counts could drop, but when SPYBALs are done to recenter the spectrum, the observed counts should recover. The amount of correction needed to recenter the spectrum is recorded in the keyword ZSPYBLU.
In fact, for program 6075, SPYBALs were done fairly often. The progression of ZSPYBLU values is shown in Figure 2. The value changes steadily, but not by enough to account for the changes in counts seen. Also, the value levels off towards the end of the observations, but the declines in the data are present in every orbit. (See Figure 1.) As mentioned previously, a significant change should also lead to a drop in the spectrum at one end relative to the other, and that is not seen.
Side 1 is not as thermally stable as Side 2 because the Side 2 electronics are always on and consequently are in good thermal equilibrium. Again, if thermal effects in Side 1 as it approaches equilibrium were to explain the effect seen in the data, the declines should go away after a few orbits and not look alike in each and every orbit from beginning to end.
Lyman-alpha Background
These observations with G140L include the Earth's geocoronal Lyman-alphaemission at 1215 Å. That emission varies systematically over an orbit, being strong on the sunlit side and weak on the dark side. If geocoronal Lyman-alpha were adding to all the spectra, then an apparent drop in counts could result.
However, an examination of the observations shows that Lyman-alpha is not a significant contributor to the total count rate when the star is observed and cannot account for the effect seen. Moreover, the wavecals are 100 times stronger than the star (in counts), so subtracting a correction for Lyman-alpha cannot remove the effect from both kinds of data.
An examination was also made of detector background (due to particle radiation), but that cannot explain any change in the raw wavecal counts. Similarly, the phasing of observations relative to the South Atlantic Anomaly was checked and shown not to be relevant.
Bright Earth Background
Contaminating light from the bright Earth could contribute to this problem, in a manner similar to that for Lyman-alpha. However, the Earth is dim in ultraviolet light, especially compared to the calibration lamp, and falls far short of accounting for what is seen.
- Image Motion at the Spectrograph Aperture
- Variations in Lamp Output
- Spectrum Drift within the Detector
- Lyman-alpha Background
- Bright Earth Background