Inverse design of photonic structures with particle swarm optimization, objective-first algorithm and deep learning

Photonics is a term that refers to a group of emerging technologies that focus on the storage, transmission, and processing of data using light beams. Photonic devices are components used to create, manipulate, or perceive light. By increasing the functionality of photonic devices, more effective applications can be obtained in many areas of use such as light detection, telecommunications, information processing, photonic computing, lighting, metrology, spectroscopy, holography, medicine, biophotonics, and military technology. Therefore, it is important to examine the interaction of photons with micrometer and nanometer-scale structures. The design of the photonic environment is the first step to be handled in this direction. The traditional design methods suffer from a long design cycle and do not take full advantage of the design space. The lack of a strategy for broadband applications, the inability to overcome complex design problems, and the design of structures in restricted parameter space, limit the performance of devices designed with traditional methods. The inverse design method is proposed to overcome these problems. It is a method based on the question of whether it is possible to design a photonic device whose electromagnetic response is predetermined. Inverse design methods are used to determine the material distributions required for the structure to provide the desired electromagnetic response. In other words, the best structure that gives the predetermined electromagnetic response is generated. In inverse design, a large parameter space is scanned with various optimization algorithms and high-performance devices can be obtained. Convenient optimization methods should be selected for near-optimum designs according to the design requirements of the device and the defined problems. In this thesis, the inverse design of various structures with Particle Swarm Optimization, Objective-first Algorithm, and Artificial Neural Networks is presented. In this context, these methods are given in detail, in the second part of the thesis. In the third part of the thesis, anti-reflective coatings, optical beam splitters, and broadband achromatic lens designs are optimized with the specified methods. The figure of merits for the designed structures are given in detail and compared with the structures in the literature. Besides, using PLA material, a low refractive index lens structure is fabricated with 3D printing and tested at microwave frequencies. The experimental results are agreed with the numerical results. The limits of the used algorithms are founded and expanded.