Data processing and algorithm development for the WFIRST coronagraph: comparison of RDI and ADI strategies and impact of spatial sampling on post-processing

Abstract

Direct detection and characterization of mature giant or sub-Neptune exoplanets in the visible will require space- based instruments optimized for high-contrast imaging with contrasts of 10-9. In this context, the coronagraph instrument (CGI) on the Wide-Field Infrared Survey Telescope (WFIRST) will reach raw contrasts of about 10-8 or better using state-of-the-art starlight suppression and wavefront control techniques. A ten-fold contrast improvement is therefore required using post-processing techniques in order to detect 10-9 planets from speckles. Post-processing techniques that are successful on both ground-based and space-based instruments need to be validated at such high contrast levels. In this communication, we investigate speckle subtraction techniques for different observation strategies and hardware parameters on WFIRST-like simulated images in the presence of deformable mirrors and an hybrid lyot coronagraph (HLC). We compare the contrast gain after post-processing in both speckle-noise and photon-noise dominated regimes for two different observing scenarios: the reference star differential imaging (RDI) and the angular differential imaging (ADI). We find that the ADI observing strategy is more robust to speckle and photon noises than the RDI observing strategy, enabling up to a threefold gain with respect to the latter. Thus, we recommend that the telescope be able to roll by at least 13° off nominal. We investigated the impact of spatial sampling on post-processed sensitivity, in the context of design trade studies for the Integral Field Spectrograph (IFS) component of the instrument. Our preliminary results suggest that the spatial sampling can be halved from the baseline sampling rate (~4 lenslets per λ/D) without any degradation in final contrast, thereby reducing the integration time required for spectroscopic characterization. In the speckle-noise dominated regime, we also find that at Nyquist sampling or higher, sub-pixel reference- to-target offsets have a negligible impact on the level of residual speckles after post-processing.