Design of computational imaging systems using wavefront-coded dielectric metasurfaces

Abstract

High-quality cameras are widely accessible today owing to the ubiquity of smartphones and the miniaturization of sensors driven by improvements in manufacturing. These cameras, however, still rely on assemblies of aberration-correcting, refractive elements that are fundamentally the same technology in use for optical systems for centuries. While for smartphones these assemblies can be made sufficiently small, these lenses present a bulky form factor for the most stringent size and weight-constrained applications. One candidate for enabling further miniaturization is metasurfaces, which are ultrathin surfaces comprising arrays of subwavelength-spaced, spatially varying scatterers. By locally tailoring the response of each scatterer, metasurfaces can manipulate the phase, amplitude, and polarization of wavefronts at subwavelength resolution with only a wavelength-scale thickness. While promising for next-generation miniaturized optics, in an imaging context, metasurfaces exhibit significant aberrations. Most metasurface research has also focused on producing static elements, which pose a challenge for systems that require varifocal control. While there has been significant work towards circumventing these challenges through innovations in scatterer design, such approaches often entail tradeoffs in terms of system complexity, polarization dependence, efficiency, and limitations on scaling to large area apertures. In this dissertation, we instead examine the utility of computational imaging in conjunction with metasurfaces so that we can simultaneously enhance performance while maintaining the size benefits offered by metasurfaces. In a computational imaging system, software is treated as a component in the image formation process. Here, we examine this approach by exploring several different imaging modalities supported by metasurfaces combined with computation so that we can simultaneously deliver a compact form factor and high-quality images. These modalities include imaging in full color, varifocal zoom capability, and acquiring depth information from a scene. Through a combination of wavefront coding, deconvolution, and Alvarez lens-inspired conjugate metasurfaces, we demonstrate a set of separate metasurface systems that image over the full visible spectrum, can achieve more than 200% change in focal length with a 1 cm aperture, and can discriminate depths with a fractional ranging error of 1.7%. The demonstrated approach may find applications in microscopy, planar cameras, machine vision, and augmented reality.