Updates to the WFC3/UVIS Saturation Map

Abstract

The calwf3 software for WFC3/UVIS utilizes a reference file to flag pixels that are saturated beyond their full-well depth. Previously, this was accomplished using a constant threshold of 65,500 e$^{-}$ across the entire detector. In this study, we retrieved $\sim$1 million stars from the Mikulski Archive for Space Telescopes (MAST) to determine the flux level at which the Point Spread Function begins to flatten, which occurs as the central pixel saturates. We quantified the saturation limit as a function of position on the detector in 1,024 discrete regions, and interpolated to a pixel-by-pixel saturation map to construct a spatially-variable saturation map reference file that is now implemented in the calwf3 calibration pipeline. We find the saturation varies by 13% across the UVIS detectors, from 63,465 e$^{-}$ to 72,356 e$^{-}$. These values agree well with earlier studies using sparser datasets, with the current analysis leading to improved characterization on small scales. Critically, the revised saturation values are larger than the previous constant threshold over 87% of the UVIS detector, leading to the recovery of usable science pixels near bright sources. This update greatly improves the robustness of saturation flags in the Data Quality arrays of observations obtained with WFC3/UVIS, and users are encouraged to redownload their data from MAST to benefit from the improved flags.

Figures (10)

• Figure 1: A demonstration of our procedure for determining how the saturation limit varies across the WFC3/UVIS detector. In the left panel, we show the peak flux (ordinate) versus 3$\times$3 aperture flux (abscissa) for thousands of stars retrieved from MAST, which were observed at three discrete locations on the detector (A, B, C). As the central pixel saturates, the rate at which it accumulates flux slows dramatically, leading to a breakpoint that we fit using a piecewise linear function, with sigma-clipped outliers shown by gray points. In the right panel, we show the 2D saturation map determined by Gilliland2010 using a related technique. The regions labeled A, B, and C correspond to the 512$\times$512 pixel detector regions for the stars that are fit in the left panel. The best-fit breakpoints are very similar to the medians of each corresponding region in the Gilliland2010 map. This serves as a proof of concept, and the small differences are expected as discussed in the Appendix.
• Figure 2: Properties of the 924,667 stars used in the saturation map analysis. The histograms show, from left to right, the distributions for quality of fit (qfit), central pixel flux (pixc), background sky flux (sky), and the number of saturated pixels (nsat). These are displayed for stars in the RAW images and so fluxes are in Data Numbers (DN). The gap in pixc is a feature of how HST1PASS flags saturated stars, and which is not utilized in this analysis.
• Figure 3: The effect of pixel phase on the peak to aperture flux ratio. Stars landing near the center of a pixel have the highest ratio, with most of the flux contained within the central pixel. Stars landing near the edges and corners lose more flux to the surrounding pixels. The main panel shows this effect as a function of flux color-coded by pixel phase, while the inset shows the effect as a function of pixel phase color-coded by flux ratio. These panels use stars drawn from a 400$\times$400 box at (x, y) = (1024, 512) on UVIS1. Stars at $r >$ 0.5 pixels (white dashed circle) bias the best-fit turnover by $>$ 100 e$^{-}$ and are excluded from the analysis.
• Figure 4: The results of our fitting procedure with the saturation map shown on the left, and a map for the number of stars used in each fit on the right. Fitting was performed on the RAW images and so are in units of Data Numbers (DN). The high density of stars near the center of the detector is due to preferentially centering targets on each chip over 15 years of observations, while the subtle enhancement in the lower-left is due to the use of subarrays near the detector readout. There are $>$ 250 stars at all locations as required for robust fits.
• Figure 5: The post-processing steps applied to the saturation map from Figure \ref{['fig:results']}. We start with the original map (left panel) and apply a minimal Gaussian smoothing with a FWHM of 2 pixels (center panel) to ensure monotonic changes across adjacent boxes while still preserving large-scale variations. This ensures that interpolation to the native WFC3/UVIS plate scale on a pixel-by-pixel basis (right panel) is free of small-scale interpolation artifacts.