Maxwell equations are behind an incredible number of physical phenomena, explaining the behavior of light, electricity and magnetism, from Gamma rays to ultra-low frequency radio-waves. Since their inception in 1861, many approximations have been derived, many devices have been modelled and fabricated to manipulate the electromagnetic fields, and more recently many computational techniques have been developed to model the behavior of light-matter interactions numerically. However, until recently only forward techniques have been developed, to simulate the behavior of the electromagnetic field inside devices and materials coming from the imagination of a human being. ', '\tIn this 21st century, artificial intelligence replaces the human for more and more complex tasks. Electromagnetics will not miss this opportunity to bring tremendous computing power and powerful algorithms to automate the device design task, called inverse-design. In this work, we develop an algorithm to do inverse-design of dielectric devices with sizes in the order of magnitude of the wavelength. The algorithm is used to design and fabricate multiple types of electromagnetic devices for the manipulation of light on-chip or in free-space, with low index or high-index dielectrics, from the microwaves to the near-infrared and maybe the visible in the future.', '\tThis dissertation is organized as follows. Chapter 1 presents the wave equations, various approximations, various computational techniques to simulate electromagnetic fields and various types of electromagnetic devices based on the inputs, outputs and types of materials inside. Chapter 2 explains the theory behind inverse-design, the comparison between the algorithm in this work and some other algorithms in the literature, regularization methods and a few insights for the implementation. Chapter 3 demonstrates our first inverse-designed device: a two-dimensional optical diode made of Silicon and air. This chapter explores some details of the optimization process as a function of the number of iterations, and the tuning to select the best hyperparameters of the algorithm to optimize the performance of the final optical diode. Chapter 4 explores the design and fabrication of polymer meta-devices to transform free-space incoming electromagnetic radiation. Meta-gratings including polarization splitters and bends are demonstrated, as well as meta-lenses, one with short focal distance, one with long focal distance and one with tunable focal distance. The algorithm is adapted to new types of inputs and outputs such as plane-waves and cylindrical waves. The physical behavior of polymer-based devices, which have a low index, does not rely on strong resonances such as those observed in Silicon-based devices. This allows to design devices with a very large bandwidth. Furthermore, polymers can be 3D-printed, which allows to create devices with much more complex geometries and much larger aspect ratio than devices made with traditional lithography methods. In this chapter, we explore millimeter-scale 3D-printing to make microwave and millimeter-wave meta-devices, as well as nanometer-scale 2-photon lithography 3D-printing for the fabrication of a Near InfraRed polarization splitter. Finally, chapter 5 concludes this dissertation and provides some perspectives on the future opportunities for the improvement of the inverse-design computational method, its extension to more types of electromagnetic devices and its combination with advanced fabrication methods.

Last modified

• 10/21/2019

Creator

• Francois Callewaert

DOI

• https://doi.org/10.21985/n2-xghj-hk80

Subject

• Electrical Engineering

Keyword

• Electromagnetics

Date created

• 2018-01-01

Resource type

• Dissertation

Rights statement

• In Copyright