As the decline in charge transfer efficiency (CTE) became a very serious issue for ACS/WFC, there was increased interest in developing a method to correct its effects pixel-by-pixel. Anderson & Bedin (2010) present a model for how charge is trapped and released during readout of the ACS/WFC CCDs over the full range of background intensity and source flux. This model was developed by studying the charge trails of warm and hot pixels in a set of calibration dark images, which determined the distribution of charge traps in the detector and the release profile of trapped charge. In the spring of 2012, pixel-based CTE correction software using the Anderson & Bedin model was included in the data processing pipeline CALACS, which produces one standard (CRJ or FLT) and one CTE-corrected (CRC or FLC) data product for each WFC observation. An in-depth description of the pixel-based CTE correction software is provided in the ACS Data Handbook.
In the summer of 2017, a new version of the pixel-based CTE correction software was implemented in
CALACS. As described in ACS ISR 2018-04, the updated model includes a technique for mitigating amplification of readnoise during the CTE correction process, updated charge trap distribution and trapped charge release profile, and the ability to run the software over multiple CPUs. The new correction provides a photometric accuracy of better than 5% and an astrometric accuracy of better than 0.05 pixels for moderately faint sources (>200e-) on moderate backgrounds (>20e-).
In fall of 2018, the capability to simulate the effects of CTE losses during readout on input images, i.e., adding the charge trails to images rather than correcting them, was added to
CALACS. Called the ACS CTE forward model, this software is intended for users who wish to more fully understand the effects of the pixel-based CTE correction on their measurements by forward-modeling synthetic data (or other data unaffected by CTE losses) and then CTE-correcting it. A Jupyter notebook is available to guide users in the proper usage of the CTE forward model (See ACS Analysis Tools).
Figure 1 shows a region of 47 Tuc from a March 2016 observation, which was a 339 second exposure in the F775W filter. The left panel shows the FLT image, which has been bias-subtracted, dark-corrected, and flat-fielded, but not CTE-corrected. The middle panel shows the FLC image, which the same as the FLT, except that it has been CTE-corrected. The right panel shows the result of CTE forward-modeling the FLC image according to the guidelines in the Jupyter notebook referenced above. The significant charge trails from stars in the FLT have been removed and added back to the stars which generated them in the FLC. The forward-modeled FLC appears nearly identical to the FLT.
The figure below shows the photometric accuracy as a function of instrumental magnitude of Milky Way bulge stars that are located at least 1500 pixels from the readout amplifiers. The observed magnitudes are measured from short exposures with varying levels of post-flash, and the "true" magnitudes are measured from a stack of deep CTE-corrected images of the same field scaled down to the short exposure time. Uncorrected data (black curves) and data corrected with the new version of the CTE-correction software (blue curves) are shown. Background level decreases from the top (70e-) to bottom (10e-) panels. Even the brightest stars on the highest backgrounds lose about 5-10% of their flux when the data are not CTE-corrected. After applying the CTE correction, the photometric accuracy remains near zero until stars of -6.75 mag for background levels above 20e-. For fainter stars and backgrounds, the photometric accuracy suffers. Note that this figure shows stars that experience the largest CTE losses due to the number of parallel transfers endured during readout.
The figure below shows the profiles of Milky Way bulge stars relative to the star's center along a column. Again, the stars are located at least 1500 pixels from the amplifiers, and the "truth" profiles (black curves) are produced by scaling down profiles from the deep stack of CTE-corrected images. Uncorrected data (red curves) and data corrected with the new version of the CTE-correction software (blue curves) are shown. Background is higher in the top row of panels than in the bottom, and the instrumental magnitude of stars decreases from the left to right panels. For both background levels, the CTE-corrected data reproduces the true stellar profiles fairly well down to about -6 mag (~250 e-), but is much worse for fainter objects.