JWST NIRCam Defocused Imaging: Photometric Stability Performance and How it Can Sense Mirror Tilts

We use JWST NIRCam short wavelength photometry to capture a transit lightcurve of the exoplanet HAT-P-14 b to assess performance as part of instrument commissioning. The short wavelength precision is 152 ppm per 27 second integration as measured over the full time series compared to a theoretical limit of 107 ppm, after corrections to spatially correlated 1/f noise. Persistence effects from charge trapping are well fit by an exponential function with short characteristic timescales, settling on the order of 5-15 minutes. The short wavelength defocused photometry is also uniquely well suited to measure the realtime wavefront error of JWST. Analysis of the images and reconstructed wavefront maps indicate that two different hexagonal primary mirror segments exhibited "tilt events" where they changed orientation rapidly in less than ~1.4 seconds. In some cases, the magnitude and timing of the flux jumps caused by tilt events can be accurately predicted with a telescope model. These tilt events can be sensed by simultaneous longer-wavelength NIRCam grism spectral images alone in the form of changes to the point spread function, diagnosed from the FWHM. They can also be sensed with the FGS instrument from difference images. Tilt events possibly from sudden releases of stress in the backplane structure behind the mirrors were expected during the commissioning period because they were found in ground-based testing. Tilt events have shown signs of decreasing in frequency but have not disappeared completely. The detectors exhibit some minor (less than 1%) deviations from linear behavior in the first few groups of each integration, potentially impacting absolute fluxes and transit depths on bright targets where only a handful of groups are possible. Overall, the noise is within 50% of the theoretical photon noise and read noise. This bodes well for high precision time series measurements.