Abstract
Metasurfaces are materials with a nanostructured surface and sub-wavelength thickness. Recent advances in the design and fabrication technology have enabled unprecedented control of light through the precise determination of the size and arrangement of the nanostructures, generating exotic applications, such as flat lenses. This thesis explores the potential of designing and fabricating metasurfaces with different levels of designer’s control. The thesis starts with an investigation to push the limits of precision in nanofabrication, and subsequently explores alternative views of metasurface design with a lower degree of control by the designer. The developed metasurfaces can be applied for a wide range of optical functionalities, e.g., beam steering, extraordinary optical transmission (EOT), and structural color generation. The investigation of precision nanofabrication focuses on the development of a largescale Si waveguide surface grating structure to vertically couple near-infrared (NIR) light at ~1.55 µm wavelength. The design approach is based on the fine-tuning of periodicity and gap size to optimize the local emission angles and the out-coupling intensities. The study presents the challenges to attain high nanofabrication fidelity over a large area, and subsequently provides a strategy for the lithographic and pattern transfer processes. The fabricated device is characterized, while the performance is evaluated with simulation and experiment, showing a focal spot with a 1/e2 width of 3.82 µm located close to the desired focal position, with ~33% source to free-space focusing efficiency. The study then investigates an approach to create near-infrared resonant apertures in Au film by using a bottom-up process based on dewetting. This process enables the formation of sub-10 nm apertures via a guiding template, but initially apertures larger than 100 nm could not be formed. Different templating strategies, e.g., fin-like templates with radial symmetry, are explored to expand the fabrication capability to form larger apertures exhibiting EOT. Compared to existing methods, such as milling and lift-off, the process improves the throughput of fabrication and the aspect ratio of the nanostructures. Interestingly, the method also shows evidence of grain boundary 4 engineering through template pinning effect. The 3D phase-field simulation of the aperture forming and grain control phenomenon matched well with the experimental results. The results show the potential of the template assisted metal dewetting to form a wide range of motifs, e.g. suspended nanodisks or trefoil apertures, crucial for applications that require high metal purity and low grain boundary scattering. Finally, the thesis investigates a data-driven approach to design dielectric nanoantennas that exhibit desired optical response. A deep neural network (DNN) algorithm is used to teach a machine to learn about the interaction between light and dielectric nano-blobs, i.e., irregular geometries with rounded corners, and the geometrical dependency on the source polarization angle. Leveraging on the pattern recognition and predictive strengths of neural networks, the DNN is designed to provide a rapid prediction of optical responses and to solve the intractable inverse design problem of complex and arbitrary nanostructures with desired optical response. The results show that the data-driven approach can provide a comprehensive solution for the optics of nanostructures.