Reclaiming WFC3 Pixels for Science

B. Sunnquist (bsunnquist[at]stsci.edu), M. Bourque (bourque[at]stsci.edu), S. Baggett (sbaggett[at]stsci.edu), and E. Sabbi (sabbi[at]stsci.edu)

Installed in May 2009, the Wide Field Camera 3 (WFC3) has spent nearly 10 years on-orbit. The high-energy radiation in this harsh low-Earth orbit environment damages detectors and changes pixel behavior. By analyzing pixel levels as a function of time in both the UVIS and IR detectors, we were able to quantify these changes and generate time-dependent bad pixel tables and dark calibration files that represent them—a process that saves many thousands of typically discarded pixels for use in scientific analysis.


WFC3/UVIS

Each day, the WFC3/UVIS detector acquires roughly 1000 new hot pixels (WFC3 ISR 2016-08). By warming the detector during the monthly WFC3 anneal procedures, ~10–20% of these hot pixels are fixed and return to their normal level; however, the remaining hundreds of thousands of hot pixels are flagged as bad in the WFC3/UVIS dark calibration reference files and are typically discarded from any analysis. During evaluation of the WFC3/UVIS pixel levels as a function of time, we developed a stability criterion to identify any pixels that vary significantly above and beyond the expected uncertainty. Surprisingly, ~95% of the WFC3/UVIS hot pixels were found to be stable and thus can be successfully calibrated during dark calibration. These hot-but-stable pixels are now identified as such in updated WFC3/UVIS bad pixel tables—and users have the opportunity to reclaim them for use in scientific analyses. Also, the new bad pixel tables flag a newly identified population of tens of thousands of cold and unstable pixels that were previously treated as good but should now be discarded by users. For more information on the new UVIS bad pixel tables see WFC3 ISR 2018-15.


WFC3/IR

In the WFC3/IR detector, roughly 3.5% of pixels have changed their behavior since launch (WFC3 ISR 2019-03); such variations include normal pixels becoming hot, unstable pixels regaining their stability, some pixels transitioning between many stable dark current levels, other pixels alternating between two stable levels, and more. In general, the WFC3/IR detector experiences an increase of ~200 hot pixels per year while the unstable pixel population can vary by up to ~4000 pixels per year (~0.02% and 0.4% of the detector, respectively). Past WFC3/IR reference files combined all of the calibration data together and were insensitive to these changes, simply flagging as bad any pixels which spent the majority of their time in a hot and/or unstable state. In our new approach, we use subsets of the calibration data to better monitor any time-variable pixel behaviors and capture those changes in a collection of time-dependent superdarks and bad pixel tables. As a result, the new superdarks provide an improved calibration for the changing hot and unstable pixels. Furthermore, with the new data quality file (DQF) flagging methodology, users have the opportunity to include stable pixels—even when hot—in their analyses; doing so can reclaim about 0.5% of the detector. Likewise, the WFC3/IR bad pixel tables now track the time-variable bad pixel populations which ensures that new bad pixels are flagged in a timely fashion and that the thousands of pixels which convert to good are not flagged. For more information on these new IR bad pixel tables and superdarks, see WFC3 ISR 2019-03 and 2019-04.

The trends in the various WFC3 pixel populations over time
Figure 1: The trends in the various WFC3/UVIS (left) and WFC3/IR (right) pixel populations over time (taken from WFC3 ISR 2018-15 and 2019-03, respectively). The IR detector has now joined UVIS in incorporating a time-dependence into the bad pixel tables and superdarks to more accurately represent these trends. In both plots, the yellow area represents those hot pixels that were found to be stable; while these are flagged with DQ value 16 in observations, they are now successfully calibrated and do not necessarily need to be discarded from analyses. The DQ value 32 is used to represent pixels that vary significantly more than expected between exposures (blue and red areas) and are not recommended for use.

References

Bourque, M., & Baggett, S. 2016, WFC3/UVIS Dark Calibration: Monitoring Results and Improvements to Dark Reference Files, WFC3 ISR 2016-08, STScI

Bourque, M., Borncamp, D., Baggett, S., Desjardins, T., & Grogin, N. 2018, Using Dark Images to Characterize Pixel Stability in the WFC3/UVIS Detector, WFC3 ISR 2018-15, STScI

Sunnquist, B., Brammer, G., & Baggett, S. 2019, Time-dependent WFC3/IR Bad Pixel Tables, WFC3 ISR 2019-03, STScI

Sunnquist, B., Mckay, M., & Baggett, S. 2019, Time-dependent WFC3/IR Superdarks, WFC3 ISR 2019-04, STScI