Title: Is Flat Fielding Safe for Precision CCD Astronomy?

Abstract: The ambitious goals of precision cosmology with wide-field optical surveys such as the Dark Energy Survey (DES) and the Large Synoptic Survey Telescope (LSST) demand, as their foundation, precision CCD astronomy. This in turn requires an understanding of previously uncharacterized sources of systematic error in CCD sensors, many of which manifest themselves as static effective variations in pixel area. Such variation renders a critical assumption behind the traditional procedure of flat fielding--that a sensor's pixels comprise a uniform grid--invalid. In this work, we present a method to infer a curl-free model of a sensor's underlying pixel grid from flat field images, incorporating the superposition of all electrostatic sensor effects--both known and unknown--present in flat field data. We use these pixel grid models to estimate the overall impact of sensor systematics on photometry, astrometry, and PSF shape measurements in a representative sensor from the Dark Energy Camera (DECam) and a prototype LSST sensor. Applying the method to DECam data recovers known significant sensor effects for which corrections are currently being developed within DES. For an LSST prototype CCD with pixel-response non-uniformity (PRNU) of 0.4%, we find the impact of "improper" flat-fielding on these observables is negligible in nominal .7" seeing conditions. These errors scale linearly with the PRNU, so for future LSST production sensors, which may have larger PRNU, our method provides a way to assess whether pixel-level calibration beyond flat fielding will be required.