Pixel-Based Non-Linearity Correction for the WFC3 IR Detector

Abstract

The current non-linearity correction for the Wide Field Camera 3 Infrared (WFC3/IR) channel is based on ground-based data acquired during WFC3's Thermal Vacuum 3 (TV3) testing campaign. In the current reference file, the correction coefficients derived for each pixel are averaged over each of the four detector quadrants. In this work, we compute a new pixel-based non-linearity correction using in-flight calibration observations with the internal tungsten lamp flats acquired between 2011 and 2013. We derive the new correction coefficients by fitting a third-order polynomial to the accumulated signal ``up-the-ramp" for each pixel. Approximately 2\% of IR detector pixels are flagged as bad, and a solution cannot be computed. For these, we use the quadrant averages of the new correction coefficients. An accompanying report (\cite{huynh2025}) provides detailed testing results using both internal flats and external science targets acquired in a variety of observing modes. The report highlights improvements in photometry derived from ``\_flt.fits" data products calibrated using the new reference file, with the largest improvement for pixels with fluence levels approaching the full well limit of $\sim$80,000 $e^- $. A new NLINFILE reference file was delivered in October 2025 and will be used to reprocess all WFC3/IR imaging and grism observations in the MAST archive.

Figures (6)

• Figure 1: Example of bias-subtracted data. The first column shows the flat SCI array, the middle column is the bias frame (zeroth read of a RAW dark file), and the last column is the bias-subtracted flat. The second row is a histogram of pixel values for each respective frame. Images in the top row are displayed in a linear scale with zscale interval.
• Figure 2: The top row displays the cumulative read of the SCI and ERR arrays from the master flat. The red squares are early saturated pixels, dead pixels, and hard saturated pixels. The bottom row is the histogram of pixel values in the SCI and ERR arrays, respectively. The images in the top row are displayed in a linear scale with zscale interval.
• Figure 3: WFC3/IR non-linear response of the mean signal (DN) in the master flat as a function of sample time (sec). Top Panel: Shows the mean signal in the master flat for a sample pixel (indices 465, 987). The green diamonds represent the mean signal values. The red dashed line is the linear fit derived from the first four reads. The horizontal black dotted line indicates the pixel's saturation limit. Blue open circles highlight mean signal values exceeding this saturation limit. Bottom Panel: Displays the fractional difference between the mean signal and the linear fit shown in the top panel. The dotted and dashed lines indicate the 5% and 25% difference thresholds, respectively.
• Figure 4: Maps of the four fit coefficients derived using the model in Equation \ref{['eq:polyfit']}. The coefficients are labeled at the top of each image and highlight both pixel-to-pixel and large-scale variations across the detector. In each image, the data is displayed with a linear scale and zscale interval.
• Figure 5: Histogram of the four non-linearity correction coefficient values derived for each quadrant of the WFC3 IR detector using the model in Equation \ref{['eq:polyfit']}.The four coefficients are labeled in the top-left of each plot, and the quadrant IDs are labeled in the top-right box. The spikes observed at the peak of each histogram originate from bad pixels where the NaN values have been replaced with a quadrant-based average of the correction coefficients.