|
| 基于串联神经网络的多焦点超透镜逆向设计研究 |
| Research on inverse design of multifocal metalens based on tandem neural network |
| 投稿时间：2025-02-18 |
| DOI：10.3969/j.issn.1005-5630.202502180026 |
| 中文关键词: 超透镜 多焦点 逆向设计 串联神经网络 |
| 英文关键词:metalens multifocal inverse design tandem neural network |
| 基金项目:国家自然科学基金青年科学基金（62205209） |
|
| 摘要点击次数: 1044 |
| 全文下载次数: 591 |
| 中文摘要: |
| 超透镜作为一种典型的衍射光学器件，在光学成像、光场调控等领域展现出显著潜力。然而，传统设计方法依赖于电磁波与材料的时域有限差分矢量三维麦克斯韦方程求解，通常伴随着高昂的计算资源和时间成本。近年来，利用深度学习算法实现逆向设计方法，提高了衍射光学器件的设计自由度与设计效率，因此受到广泛的关注。提出了一种基于深度学习串联神经网络实现多焦点超透镜逆向设计方法，可实现高效的超透镜相位前向预测和结构参数设计。基于该网络，实现了目标焦距为9 μm的超透镜设计，设计误差低于2.8%，聚焦效率达到60.6%。该方法还可用于实现多焦点超透镜设计。该研究对于集成衍射光学器件与信息处理等领域的研究具有重要价值。 |
| 英文摘要: |
| As a typical diffractive optical device, metalenses have shown significant potential in the field of optical imaging, light field modulation, etc. However, traditional design methods rely on the time-domain finite difference vector three-dimensional Maxwell equation solution of electromagnetic waves and materials, which usually require high computational resources and time costs. In recent years, the inverse design methods based on deep learning improve the design freedom and efficiency of diffractive optical devices. Therefore, they have received widespread attention. An inverse design method for multifocal metalenses based on deep-learning tandem neural networks is proposed, which achieves efficient forward prediction of the phases and the design of structural parameters. Based on this network, a metalens designed with the target focal length of 9 μm has been achieved with the design error controlled below 2.8%, and the focusing efficiency of 60.6%. This method can also be used to realize the design of the multifocal metalens. The study has significant value for the research in the field of integrated diffractive optical devices, information processing, etc. |
| HTML 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
