Reconciling kernel-phase and coronagraphy: new steps towards combining the performance of opposing techniques

Coronagraphs are powerful tools to probe the direct neighborhood of stars at very high contrasts but their vulnerability to wavefront errors however makes them less efficient for observations at angular separations smaller than two or three resolution elements. In this regime, observables robust to instrumental phase noise, like the closure- and kernel-phase extracted from non-coronagraphic images, have proven capable of effectively picking up moderate contrast features down to the formal diffraction limit. Direct kernel-phase analysis of coronagraphic images is unfortunately not possible in theory. The focal plane mask used in a coronagraph indeed destroys the shift-invariance properties that give meaning to the analysis of their Fourier transform. We nevertheless investigate how techniques initially developed in the context of coronagraphic observations can be applied with kernel-phase to boost the contrast detection limits. Firstly, inspired by Angular Differential Imaging, we devised a similar method called Angular Differential Kernel to remove static biases from our measurements which are a limiting factor for reaching high contrasts. We present a recent comparative on-sky analysis of its performance using the SCExAO instrument at the Subaru Telescope. Secondly, we show how pupil plane apodization masks can be used to locally decrease the photon noise in the images, and how their effect translates into the kernel-phase observables, therefore improving the capability of kernel-phase to detect faint companions around nearby stars.