稳健性(进化)
制作
分路器
网络拓扑
反向
计算机科学
拓扑(电路)
电子工程
曲率
材料科学
工程类
光学
数学
物理
电气工程
医学
生物化学
替代医学
几何学
病理
基因
操作系统
化学
作者
Yuchen Chen,Jifang Qiu,Zhenli Dong,Xiaogang Wang,Yin Liu,Hongxiang Guo,Jian Wu
标识
DOI:10.1109/jlt.2023.3242472
摘要
In order to improve the fabrication robustness of the inverse-designed devices that have complex topology structures, we propose an optimization technique, named as structure transformation (ST), to transform structures of the inverse-designed devices into simple topologies that contain less irregularly tiny features, sharp corners and less boundary regions. The proposed ST technique, based on a new alternately-implementing strategy of optical performance optimization and modified curvature-constrained structure adjustment, effectively facilitates an overall simplification of the structure topologies of devices, enhances their robustness to fabrication uncertainties, and simultaneously maintains or even improves their optical performances. To verify the feasibility of the technique, three types of typical SOI-based devices (T-junction, 1×3 power splitter and 90° crossing) with and without ST were designed, among which the ones without ST were given as control groups. Thanks to the ST, the boundary regions and highly curved ranges in these devices are all significantly reduced, while their performances are well maintained or even improved. Next, the effect of the ST on fabrication tolerances were investigated, and the simulation results show that the devices with ST are more robust to over/under-etching errors than the control groups. At last, the T junction and 1×3 power splitter and their control groups were fabricated each with 6 samples. Measured results show that the performances of the devices with ST deviate from the simulated performances much less than their control groups, which agrees well with the simulated results in fabrication tolerances investigation. Summarily, we proposed and demonstrated a technique that provides the inverse-designed devices new possibilities to achieve simpler topologies with better fabrication robustness and improved performances at the same time.
科研通智能强力驱动
Strongly Powered by AbleSci AI