Title:
Wavefront Sensing and Subdiffraction Target Discrimination Using Spatial Mode Sorting
Abstract:
Spatial mode sorting is a generalization of traditional imaging, where instead of measuring (or filtering) an optical field’s Fourier components, we instead measure (or filter) photons residing in some arbitrary set of orthogonal spatial modes. Processing the light in this manner has been shown to be “quantum-optimal” for a number of imaging and sensing tasks, including exoplanet detection around unresolved stars and estimating the subdiffraction separation between two incoherent point sources. Despite these potential sensitivity advantages, however, contemporary mode sorting hardware often suffers from large crosstalk, high loss, or sort modes that are poorly adapted to circular apertures.
In this dissertation, I propose a spatial mode sorting architecture that can, in principle, realize lossless and crosstalk-free spatial demultiplexing of both the real- and complex-valued Zernike polynomials. My design relies on iterative application of vortex-phase Fourier filters and Mach-Zehnder interferometers, both of which can be assembled using commercially available optics hardware. This mode sorting paradigm is then applied to two optical sensing tasks: astronomical wavefront sensing and subdiffraction target discrimination.
For wavefront sensing, I propose a hardware platform termed the piston-adapted wavefront sensor (PAWS), which is shown to saturate the Holevo Cramér-Rao bound to simultaneous phase and log-amplitude estimation in the high-Strehl regime. The same hardware architecture can also be used to effect a perfect second-order coronagraph (as well as coronagraphs of higher order).
For target discrimination, I present a simulation study examining the use of my coronagraph for classifying subdiffraction objects. While direct imaging fails to outperform a random classifier at deep subdiffraction (one-twentieth the width of the point spread function) target sizes, the coronagraph approach excels, offering classification accuracies greater than 60% on the ten-class MNIST dataset. These simulation results are supplemented by a discussion on experimental design, wherein I present an alternative layout for the coronagraph based on polarization control.
Committee:
Dr. Amit Ashok (Chair)
Dr. Ron Driggers
Dr. Daniel Soh
