Photonic Crystal Nanobeam Cavity With a High Experimental Q Factor Exceeding Two Million Based on Machine Learning

Q-Factor Optimization of Modes in Ordered and Disordered Photonic Systems Using Non-Hermitian Perturbation Theory

“…32,33 However, while they allow a pathway for iterative optimization, these approaches are supervised, and their extension to the case of random modes is not trivial. In parallel, rapid growth of computational resources has helped the development of both gradient-free and gradient-based automated optimization methods such as nature-inspired search algorithms, 34,35 machine learning, 27,36,37 and densitybased topology optimization. 38 In particular, gradient-based inverse design, which is transforming the paradigm of highefficiency component design in nanophotonics, 39 uses adjoint sensitivity analysis to efficiently compute gradients of a wide variety of objective functions.…”

“…This boils down to modifying the momentum-space representation of the resonant modes via either first-principles group symmetry arguments, the direct observation of the smoothness of the field envelope, real-space analysis of the leaky components, or semianalytic formalisms that tackle the problem as a reverse design one. , However, while they allow a pathway for iterative optimization, these approaches are supervised, and their extension to the case of random modes is not trivial. In parallel, rapid growth of computational resources has helped the development of both gradient-free and gradient-based automated optimization methods such as nature-inspired search algorithms, , machine learning, ,, and density-based topology optimization . In particular, gradient-based inverse design, which is transforming the paradigm of high-efficiency component design in nanophotonics, uses adjoint sensitivity analysis to efficiently compute gradients of a wide variety of objective functions.…”