Studies On Properties Of Microstructured Silicon Photonic Devices

Integrated circuits(ICs)are increasingly unable to meet the demands of modern society for high-speed data processing and information transmission due to quantum size and power consumption issues.People are paying more and more attention to the unique advantages of photons as information carriers,and hope to use photons to continue the development of Moore's Law.In this era,silicon photonics has been extensively studied and the improvement of the performance of silicon photonic systems has become an important issue.In addition to the design and optimization methods based on traditional optical theory,the promotion of theory in other fields to silicon photonic systems,combined with artificial intelligence and inverse design,will provide new motivation to the development of silicon photonic system.The parity-time symmetry theory is introduced and it is proposed that the lossy medium has an extra enhancement on the modal coupling loss in the two-mode waveguide with spontaneous broken PT symmetry.By designing the perturbation distribution of the real and imaginary parts of the effective refractive index on the two-mode SOI waveguide,three devices are constructed in the Hermitian phase,the broken PT symmetry phase and the pure absorption phase.Numerical simulations and experiments have effectively verified that when the system is in the broken PT symmetry phase,it will have a higher loss to the targeted output mode than the pure absorption phase.In this solution,the lossy medium introduces a phase transition from the Hermitian phase to the broken PT symmetry phase in addition to the normal light absorption.The former mechanism strongly prohibits the power coupling from the input mode to the output mode,which effectively enhances modal coupling loss.The experimental results demonstrate that the enhancement factor of modal coupling loss can reach 17dB and has a high bandwidth.This work may offer a different way to exploit the role of optical loss in boosting modal coupling loss that can directly translate to power extinction ratio.It thus may find potential applications in novel intensity modulator designs.The artificial neural network algorithm is introduced and a scheme is proposed for analyzing the modal power distribution in a multimode waveguide using convolutional neural networks.This solution does not rely on the spatial mode sorting commonly used in the past,thus eliminates the complexity on device level and shifts the burden to the data processing level.A specialized convolutional neural network and a general two-dimensional convolutional neural network are respectively designed for the thin SOI multimode waveguide and the SOI heavy multimode waveguide to predict the modal power distribution by taking the preprocessed far field intensity pattern as input.The performance of these trained convolutional neural networks has been proven to be great and the performance does not decay in the case of heavy multimode waveguide.In addition,it is verified that the trained convolutional neural network is highly robust to noise.It is believed that this new method of analyzing the modal power distribution in multimode waveguides using convolutional neural networks has high potential applications in the field of space-division multiplexing and structured light,and can be extended to multimode fiber systems.The principle of inverse design is introduced and a scheme is proposed for designing ultra-small and high-performance SOI mode converters using particle swarm optimization(PSO)algorithm.The one-sided width variations and total length of coupling region in the mode converter are parameterized and limited in a reasonable range to form a solution space.The average forward transmittance of the TE00 mode calculated by FDTD is taken as the fitness function,and the velocity and position of the particles searching in the solution space are updated iteratively so that they continuously approach the global optimization.According to the above inverse design principle,the global optimal parameters(or near optimal)of the mode converter are obtained.Multiple performance metrics such as device size,insertion loss,backscatter,mode conversion rate,mode purity and processing error tolerance have been analyzed.It turns out that our mode converter performs well compared to the previous work.