6.7 Sink Pixels

With the advent of post-flashing in 2012, a new type of image defect called sink pixels was identified (see WFC3 ISR 2014-04 for the initial discovery and WFC3 ISR 2014-19 for additional analysis). Sink defects are caused by pixels that contain a modest number of charge traps (typically 20–100 electrons). During exposure integration, the pixels fill with charge, and some of that charge gets stuck in traps. When the parallel shifting begins, the trapped charge does not shuffle down with the other electrons associated with the pixel, resulting in less charge recorded in that pixel than expected. 

Sink pixels stand out in post-flashed bias images, as when they are read out, they contain fewer counts than adjacent “normal” pixels (see Figure 4 of WFC3 ISR 2014-22). This phenomenon is distinct from normal pixel-to-pixel sensitivity variations, in that photons that interact with sink pixels do generate electrons, but some of those electrons do not shuffle out of the pixel during the readout process, and are thus not recorded with that pixel.  Sensitivity variations apply to all impacting photons irrespective of how bright the pixel is, whereas sink-pixel trapping generally has a smaller fractional effect on brighter pixels. 

Only a very small number of sink pixels were found in data taken before launch.  Investigations suggest that most of the sink pixels are likely a consequence of on-orbit radiation damage (WFC3 ISR 2014-19, WFC3 ISR 2014-22), in fact that sink pixel traps are likely the same traps that cause CTE losses.  The distinction is that CTE losses correspond to charge getting shuffled through traps.  Sink pixels represent trapping before any shuffling starts. As such, sink pixel effects are localized losses, while CTE losses are more generalized.  At present, no sink pixel has been found to heal or recover, consistent with the monotonically increasing CTE losses. 

The impact of sink pixels depends both on their locations in an image and the pre-readout pixel value. For images with high backgrounds (~85 electrons/pixel) and for sink pixels near the readout register, the sinks are single low pixels and have little effect on upstream or downstream pixels in the same column. However, for images with low backgrounds or for pixels far from the readout register, the interplay between CTE losses and sink pixels can extend the sink pixel profile to more than 10 pixels (see Figure 3 of WFC3 ISR 2014-19). So although sink pixels are rare (at last count, ~0.25% of the detector), in low-background imaging they can corrupt as much as ~1% of the detector. The impact of the sink pixels is also scene-dependent. For example, if a source lands on a sink pixel (or the streak of a sink pixel), then the electrons in the source will affect the trail behind the sink in the same way that a higher background would. Thus, the WFC3 team has adopted a conservative approach to flag all pixels that are likely to be affected in a given image; observers can choose whether to use or ignore the sink pixel flags (DQF value of 1024, see Table 3.2) in the science image DQF files.

The calibration pipeline calwf3 (version 3.3+) uses the sink pixel reference file (SNKCFILE) to populate the data quality array of a science image with 1024 for all flagged sink pixels. The sinks are identified in the SNKCFILE with the modified-Julian date (MJD) of the appearance of the sink pixel; any upstream/downstream adjacent pixels affected by the sink pixel are flagged as well (see Section 6 in WFC3 ISR 2014-22). More details on the flagging process are given in Section 3.2.7, as well as in WFC3 ISR 2014-22. Note that the sink pixels flagging is performed regardless of whether the CTE-correction is performed, i.e. both flt and flc will have sink pixels flagged in the DQ extensions; the science data pixels are not changed in any way.