Title:  A New Paradigm for Knowledge Discovery and Design in Naophotonics Based on Artificial Intelligence

Committee: 

Dr. Adibi, Advisor

Dr. Chang, Chair

Dr. Wang

Abstract: The objective of the proposed research is to develop a new paradigm based on leveraging the intelligent aspect of artificial intelligence (AI) to design nanostructure and understand the underlying physics of light-matter interactions.  Considering the large number of design parameters and the complex and non-unique nature of the input-output relations in nanophotonic structures, conventional approaches cannot be used for their design and analysis. The dimensionality reduction (DR) techniques in this research considerably reduce the computing requirements. This proposal also focuses on developing a reliable inverse design approach by overcoming the non-uniqueness challenge. I have developed a double-step DR technique to reduce the complexity of the inverse design problem while preserving the necessary information for finding the optimum nanostructure for the desired functionality. I established an approach based on defining physics-driven metrics to explore the low-dimensional manifold of design-response space and provide a sweet region in the reduced design space for the desired functionality. In the remaining work, I propose using a mixture density network to map the low-dimensional instances to the actual design parameters to complete the inverse design approach for non-unique problems.