Experimental Demonstration of Arbitrary Wave-Shaping with High Q Metasurfaces—A Route to Ultra-Efficient, High-Resolution Spatial-Light-Modulators

Phase gradient metasurfaces are revolutionizing optical system design, providing a way to shrink the thickness of any component down to the nanoscale, from lenses and prisms to polarizers. Tailoring diffracted light using non-uniform arrays of nanoscale antennas, the performance of these metasurfaces often matches if not surpasses their refracting counterparts. For example, meta-lenses with comparable resolution to high-end infinity corrected objectives 1000 times their size and bright, artifact-free metasurface holograms have both been realized. With most metasurface elements being resonant, it has been suggested that they are also ideal for dynamic light wave manipulation, potentially unlocking exciting application areas, including LiDAR and AR/VR. However, the very weak refractive index modulation found in most materials leads current metasurface designs to require impractically large electrical tuning voltages. Exciting demonstrations have been given using special materials, such as liquid crystal or epsilon-near-zero materials, but these come with limitations, including slow switching speeds, poor resolution, and often extremely strong absorption.<br/>Here, we reveal a route to dynamic metasurface wave shaping free from these limitations by amplifying weak index modulation. Specifically, we theoretically and experimentally introduce a new high quality (Q) factor metasurface platform, capable of generating arbitrary reflected phase profiles over an extremely narrow bandwidth in the infrared. These metasurfaces are composed of silicon nanoantennas sitting above a metallic ground-plane. The nanoantennas are patterned to support high-Q dipolar guided mode resonances (DGMRs). Due to the lack of absorption and the formation of an image dipole in the ground plane, the reflected phase from each element varies from 0-2π across the resonance, while maintaining unity reflectance. Importantly, the DGMRs or localized to individual nanoantennas. So, by independently tuning the resonant frequency of each element, the reflected phase of light at the center wavelength can be defined as a function of position across the array. As the element spacing can be subwavelength, very precise phase profiles can be realized, resulting in a high numerical aperture and minimizing diffractive artifacts. By simulating the reflection from an electrically biased Si-LNO metasurface, we show that beamsteering can be achieved with just a few volts applied to each element, which is a direct result of the high Q of the DGMRs.<br/>As experimental proof of the dynamic high-Q meta-reflect-array concept, we have realized fixed beam steering metasurfaces with a structurally tuned steering direction. While a dynamically tunable device requires resonant shifts to be realized via a bias dependent index change, the same physics applies to a system with resonance shifts resulting from slight structural variations of the antennas. Our nanoantenna design involves columns of silicon block pairs sitting on a sapphire substrate, with the slight difference in length between neighboring blocks determining the Q. After e-beam patterning, the chip is coated with PMMA and silver to produce the reflect-array. Phase tuning of each nanoantenna is achieved by altering the width of the smaller block. We vary this parameter across the array to generate linear phase gradients near the operating wavelength of 1485nm. Measuring the change in resonant deflection angle for devices with different numbers of elements per supercell, we confirm that arbitrary phase slopes can be realized. The relatively high-Q of the modes, Q≈500, allows the very small structural variation, just a 14nm (or 6%) difference between largest and smallest blocks, to produce strong diffraction. This small structural change is a proxy for needing only a very weak bias in a dynamic system. The Q here was also limited mostly by e-beam resolution, so we expect even better performance with a dynamic system.