IR Photometric Calibration

The repeatability of photometric measurements stars over the course of one to a handful of orbits is found to be approximately +/- 1.0% (ISR 2024-06; ISR 2024-01), despite Poisson noise terms being quite a bit smaller than 1%.  This is due to self-persistence from previous images in the visit, and can be mitigated by dithering, by at least 10 pixels between exposures, to ensure that stars do not fall on the same pixels.  Using this dither strategy, stars with very small Poisson noise components achieve repeatability of approximately 0.5% (ISR 2021-05; ISR 2019-07).

Previous analyses have shown that WFC3-IR suffers a count-rate dependent non-linearity of about 1% per dex, an order of magnitude smaller than the prior HgCdTe detector, NICMOS, flown on HST, but large enough to potentially limit the accuracy of photometry. In Riess et al. 2019 (ISR 2019-01) we present new and more precise measurements of count-rate non-linearity (CRNL) through a combination of comparisons of cluster star photometry between WFC3-IR and WFC3-UVIS and by using observed and synthetic magnitudes of white dwarfs. We further extend the measured range of CRNL to higher count rates by comparing magnitudes between the ground and WFC3-IR for LMC and Milky Way Cepheids. Combining these results with all previous measurements and those from the WFC3 grism provides a consistent and improved characterization of the CRNL of WF3-IR, of 0.75% +/- 0.06% per dex, with no apparent wavelength dependence, measured across 16 astronomical magnitudes. To illustrate the value of the precision reached for the CRNL, we show it is sufficient to compare the photometry of sources along a distance ladder calibrated by Gaia parallaxes and produce a 1% determination of the Hubble constant.

Due to the difficulty of measuring CRNL for WFC3-IR on orbit, a broad set of approaches have been used. The initial, on-orbit calibration of CRNL was produced by Riess (2010) who compared the magnitudes of stars in clusters between two overlapping passbands, one from an instrument without CRNL (the CCDs of ACS WFC) or corrected for CRNL (NICMOS) and WFC3-IR.

In Riess et al. (2019) we present additional sets of measurements which provide greater precision in the on-orbit determination of the WFC3-IR CRNL. First we present new measurements comparing cluster star magnitudes in bands where CCD’s overlap with WFC3-IR, F850LP, and F098M using a combination of deep and shallow frames to extend the dynamic range. We also provide calibrations of the CRNL at brighter magnitudes (F160W=6-14) using comparisons between the ground and WFC3-IR of the photometry of Cepheids in the Large Magenallanic Cloud and the Milky Way. We then use an independent approach where we compare the near-infrared photometry of hot, DA White Dwarf (WD) to models constrained by optical measurements.

We find the CRNL in WFC3-IR to be 0.0077 +/- 0.0008 mag/dex (or 0.0075 +/- 0.006 mag/dex including the grism measurements), characterized over 16 magnitudes with no apparent wavelength dependence and independent corroboration. This result is consistent with the initial determination by Riess (2010) of 0.010 +/- 0.002 mag/dex but the present results are much more precise. This result may be used to correct IR photometry by using the difference in apparent flux (in dex) between where the WFC3-IR zeropoint is set (~12th mag) and the target source (fainter sources appear even fainter and thus are corrected to be brighter).