4.2.2. Subtraction of Dark Backgrounds
We have developed a new method for dark subtraction because the pipeline subtraction ignores the fact that the signal which accumulates on the WF chips while the shutter is closed varies between 0.5-1.1 DN per 1800 sec exposures for the WF3 and WF4 chips (more for WF1, less for WF2). This variability is not caused by thermal dark current, but rather by variations in the dark "glow" contributed by the MgF2 field-flattening lenses, positioned immediately above each of the WFPC2 chips. When struck by even low-energy (200-2000 eV) ions and electrons (which are abundant in the upper atmosphere in the form of N2+ and cosmic rays), MgF2 will produce broad-band fluorescence on a time-scale << 1 sec (Qi et al. 1991). We have found that the cumulative flux in cosmic ray hits recorded in a dark (shutter-closed) exposure correlates strongly with the mean dark level in that frame. This correlation is shown Figure 5 in which we plot the cosmic ray flux and dark glow for each WFPC2 CCD in each of 15 dark exposures. The scatter around the linear fit for each chip is ~ 0.06 DN. This plot also shows clearly that the dark background in a 1700 sec exposure can vary by as much as ± 0.35 DN. As the total background sky signal through the F300W is only 0.3 DN/1700sec, accurate subtraction of the dark glow is essential for this measurement.
In collaboration with H. Ferguson (STScI), we developed a method which uses the correlation between the MgF2 glow and the flux in cosmic rays to isolate the thermal dark component from the dark glow. We obtained a final dark solution using 80 individual dark images taken between October and December 1995, and have tested the accuracy of this method by test-reducing the 15 dark exposures taken between the science exposures of our program. These 15 darks were not used in determining the dark solution itself. After ADC correction, overscan subtraction, bias subtraction, and dark subtraction by the prescription described above, the average mean level of the 15 darks is consistent with zero, with a scatter of < 0.05 DN per 1700 sec exposure. The statistical error in the pipeline dark-subtraction, by comparison, is roughly 0.15-0.25 DN for the WF3.
Figure 5. Correlation between the cumulative flux in cosmic ray events and the mean dark level (thermal dark plus MgF2 glow) for each WFPC2 chip in 15 exposures: × 's for PC1 fluxes, open boxes for WF2, circles for WF3, and filled circles for WF4. Statistical errors in both quantities plotted are negligible; scatter in the correlation probably reflects statistical variations in the energy level of the cosmic rays hitting the CCD and the MgF2 plates.