Hang Sun , Zheng-Da Hu , Jingjing Wu , Jicheng Wang
{"title":"Design of robust topological edge states clock-shaped multiplexer in anisotropic photonic crystals enhanced by neural networks","authors":"Hang Sun , Zheng-Da Hu , Jingjing Wu , Jicheng Wang","doi":"10.1016/j.photonics.2025.101423","DOIUrl":null,"url":null,"abstract":"<div><div>Topological edge states have demonstrated significant potential in controlling light propagation. This paper presents a high-degree anisotropic photonic crystal based on a combination of convolutional and fully connected neural networks. The model accurately predicts the asymmetry parameters and band structure in both directions. We achieve topological edge state transmission over a broader frequency range compared to traditional valley photonic crystal array channels. We precisely predict accurate results for both wide-frequency boundary state outputs and topological bandgaps. We design a clock-shaped wavelength division multiplexer to enable precise optical channel selection. Our deep learning model demonstrates robust performance against input variations, and the structure remains highly resilient to disturbances. The deep learning-assisted anisotropic structure demonstrates precise selectivity, wideband high-intensity transmission and strong robustness in topological edge states significantly enhancing the application potential and development efficiency of topological edge states.</div></div>","PeriodicalId":49699,"journal":{"name":"Photonics and Nanostructures-Fundamentals and Applications","volume":"66 ","pages":"Article 101423"},"PeriodicalIF":2.5000,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Photonics and Nanostructures-Fundamentals and Applications","FirstCategoryId":"101","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1569441025000732","RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"MATERIALS SCIENCE, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 0
Abstract
Topological edge states have demonstrated significant potential in controlling light propagation. This paper presents a high-degree anisotropic photonic crystal based on a combination of convolutional and fully connected neural networks. The model accurately predicts the asymmetry parameters and band structure in both directions. We achieve topological edge state transmission over a broader frequency range compared to traditional valley photonic crystal array channels. We precisely predict accurate results for both wide-frequency boundary state outputs and topological bandgaps. We design a clock-shaped wavelength division multiplexer to enable precise optical channel selection. Our deep learning model demonstrates robust performance against input variations, and the structure remains highly resilient to disturbances. The deep learning-assisted anisotropic structure demonstrates precise selectivity, wideband high-intensity transmission and strong robustness in topological edge states significantly enhancing the application potential and development efficiency of topological edge states.
期刊介绍:
This journal establishes a dedicated channel for physicists, material scientists, chemists, engineers and computer scientists who are interested in photonics and nanostructures, and especially in research related to photonic crystals, photonic band gaps and metamaterials. The Journal sheds light on the latest developments in this growing field of science that will see the emergence of faster telecommunications and ultimately computers that use light instead of electrons to connect components.