Integrated photonic device designs based on computational approaches

Nanophotonics is a field of science which investigates the behavior of light at the scale of wavelength or smaller dimensions. The design of photonic devices have been developed based on physical knowledge and intuition. Thus, many photonic devices have been introduced from past to present. In order to design a new photonic devices, a known physical phenomena is utilized which is followed by tuning small number of characteristic parameters on it. As a result, ranging from ultraviolet to mid-infrared, various photonic devices have been designed to operate at different wavelength regimes. The development of nanophotonics is being pursued by dense integration of photonic devices and increasing their bandwidth. However, as the complexity of photonic devices increase, problemshave been faced in design processes. For example, when an optical property depends on large number of characteristic parameters, it is required to optimize these parameters simultaneously. In this case, manual adjustment of parameters is not a decent approach to obtain desired optical properties in photonic devices. It can be concluded that these design approaches are not efficient enough for the future of nanophotonics. In order to overcome these design problems, in nanophotonics, inverse design approaches have been introduced. A photonic device with desired optical property can be designed in a defined design space by using inver desing approaches. Here, a known physical phenomena is not required; and large number of characteristic parameters can be optimized simultaneously. In these designs, various optimization algorithms and, recently, machine learning methods have been applied. The desired optical properties are defined as a cost function and algorithms adjust characteristic parameters by minimizing the value of cost function. In order to calculate cost function values, several numerical methods are utilized which simulate electromagnetic fields and waves. The computational approaches are obtained by combining numerical methods with algorithms. In the first part of this thesis, the histroical background of inverse design approaches in nanophotonics are summarized. From very first studies to recently introduced ones, the most important studies are briefly presented. In the second part of this thesis, the selected algorithms and numerical method is explained in details. In this dissertation, evolutionary algorithms and reinforcement learning algrotihms are applied to design integrated photonic devices. As evolutrionary algorithms, Differential Evolution and Genetic Algorithm are applied. Also, as reinforcement learning algorithms, additive reinforcement learning algorithm and attractor selection algorithm are utilized. In addition, as a numerical method, finite-difference time-domain method is utilized. In the third part of this thesis, beam shaping photonic devices such as an optical cloak, a lens, optical couplers, a mode order converter and a multilevel diffractive lens are presented. In the fourth part of this thesis, optimization algorithms are applied on photonic bandgap structures, namely photonic crystals (PCs). As a result, a mode order converter PC waveguide and an asymmetric light transmitter (ALT) PC are designed. In the fifth part of this thesis, an ALT, wavelength demultiplexers, a polarization beamsplitter, a waveguide crossing, an optical attenuator are introduced which can be classified as optical splitters. In the sixth part of this thesis, the studies in this thesis are summarized and concluding remarks on inverse design in nanophotonics are shared. To conclude, As a result of the studies in this dissertation, beam shaping photonic devices, photonic bandgap structures and integrated optical splitter photonic devices are designed. The obtained results are promising for the future of nanophotonics.