Thesis Format
Integrated Article
Degree
Doctor of Philosophy
Program
Mechanical and Materials Engineering
Supervisor
Johlin, Eric
Abstract
The strong influence of wavelength-scale geometry on electromagnetic fields as well as the unintuitive nature of these responses make the inverse design of nanophotonic structures promising to improve the efficiency of nanophotonic components. However, these design processes are accompanied by challenges, such as their high sensitivity to initial conditions, computational expense, time-intense, and complexity in integrating multiple design constraints. Machine learning and deep learning approaches, however, show strengths, addressing these limitations allowing huge sample sets to be generated nearly instantaneously, and with transfer learning, allowing modifications in design parameters to be integrated with limited retraining. Therefore, in this thesis, we explore a variety of machine learning and deep learning approaches to improve the inverse design of nanophotonic structures. That includes randomly generated simple absorbing 2D nanophotonic structures, high-quality adjoint optimized 3D nanolens structures, and color splitter nanoscale structures to replace color filters in standard cameras. The forward design and inverse design performances are investigated for both interpolation and extrapolation performances. We introduce a hybrid deep learning approach, leveraging the accuracy and performance of adjoint-based topology optimization to produce a high-quality training data set to enhance the potential and performance of deep learning techniques than conventional techniques alone. Further, the transfer learning approach is utilized to retain networks on new design parameters with very few new training samples. This process can be used for general nanophotonic design and is particularly beneficial when a range of design parameters and constraints need to be applied. Moreover, tandem neural networks have seen success in the inverse design of nanophotonic structures compared to other commonly used techniques but they suffer from a significant limitation as they are single-input single output networks with zero diversity. In this thesis, we further implement a novel single-input, multiple-output tandem network to address this limitation. Furthermore, our technique to use genetic algorithm optimization in inverse tandem network demonstrates that the extrapolation capability of the implemented network can be enhanced from the utilized methodology. Therefore, deep learning approaches are promising for the inverse design of nanophotonic structures in optoelectronic applications.
Summary for Lay Audience
Light is an incredible phenomenon that fills the world around us with brightness. It gives things their color, aids our vision, and even holds secrets about the universe. People have been fascinated by it for a very long time and are still learning about it to understand and push boundaries for a new era of technology. Exploring light at a very small scale (billionth of a meter) is known as nanophotonics and many studies have been performed to improve this area by introducing new conventional methods and mechanisms. Although those mechanisms could push the limits of the field, some major problems could not be resolved even after years of research such as sensitivity to the initial conditions, requirement of high computational power and time, and implementation complications when many design rules are involved. A huge attention has been gained in implementing concepts and methods of artificial intelligence (AI) in nanophotonics with the advancement of AI. The recent theoretical results show that AI techniques are promising to address most of the challenges in nanophotonics and it has the most important advantages to use over existing traditional methods. This study explored the utilization of AI techniques for the designing of unintuitive nanophotonic structures related to different optical performances. In this thesis, we demonstrate the use of AI techniques for two and three-dimensional nanophotonic structure designs, optimize and predict their optical performances for different optoelectronic applications. Training of the networks was first conducted using randomly generated structures, later, for more complex tasks, we combined AI techniques with conventional optimization techniques to produce high-quality training data and improve the network capabilities. This combined approach leveraged the performances and we further demonstrated a technique to transfer knowledge of one task to another newly defined application proving the promise of AI technique for nanophotonics. Moreover, we introduced a modified version of existing AI technique that addresses the limitation of the original network. Our work demonstrates that AI techniques can make designing nanophotonic structures more efficient and adaptable, which is beneficial for a wide range of applications.
Recommended Citation
Salwathura Acharige, Didulani A D, "Deep learning for inverse design of nanophotonic structures" (2024). Electronic Thesis and Dissertation Repository. 10306.
https://ir.lib.uwo.ca/etd/10306
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.