FUTURE | 远见 闵青云 文
近日，国科大杭州高等研究院杨陈楹团队和浙江大学沈伟东团队以「All-optical combinational logical units featuring fifth-order cascade」¹为题在Chip上发表研究论文，介绍了一种基于衍射神经网络的全光组合逻辑单元，展示了在全光时分复用中无延迟处理高速数据传输的能力。共同第一作者为高海淇和邵煜，通讯作者为杨陈楹、沈伟东。Chip是全球唯一聚焦芯片类研究的综合性国际期刊，是入选了国家高起点新刊计划的「三类高质量论文」期刊之一。
图1 | 全光组合逻辑单元的原理图和位置编码设计。a，全光组合逻辑单元的示意图。b，衍射过程的演示。c，分别通过传统电子方法和新型光学方法实现解码器的架构。d，编码器和解码器输入输出平面的位置编码设计。e，多路复用器的输入-输出设计。f，解复用器的输入-输出设计。
这种创新设计提供了更高的处理速度、更好的能效、强大的环境稳定性以及高容错率，使其非常适合于光学计算和通信领域的广泛应用。利用迁移学习，作者们还成功开发五阶级联组合逻辑单元。
图3 | 新型OTDM系统的结构和仿真结果。a，使用我们的组合逻辑单元进行信息传输的新型光时分复用OTDM框架。b，利用全光多路复用器和解复用器随时间变化的OTDM工作流程。c，在一致时钟信号下OTDM的部分仿真结果。这里仅展示了六个输入情况。d，复用器（Multiplexer，MUX）和解复用器（Demultiplexer，DEMUX）中使用的层数与串扰之间的关系。
All-optical combinational logical units featuring fifth-order cascade1
Fig. 1 | The schematic and location coding design of the proposed all-optical combinational logical unit. a, Schematic diagram of the proposed all-optical combinational logical unit. b, The demonstration of the diffraction process. c, Implementation architecture of decoders by traditional electronic and novel optical methods respectively. d, The location coding design of the input and output planes for encoder and decoder. e, The input-output design of multiplexer. f, The input-output design of demultiplexer.
This innovative design offers increased processing speed, improved energy efficiency, robust environmental stability, and high error tolerance, making it exceptionally well-suited for a broad spectrum of applications in optical computing and communications. By leveraging transfer learning, the authors successfully developed a fifth-order cascaded combinational logic circuit.
Fig. 2 | The implementation and results of the cascaded all-optical combinational logic circuit. a, The framework of the cascaded all-optical combinational logic circuit. b, Schematic diagram of training strategy applied during single training epoch of the cascaded all-optical combinational logic device. c, The accuracy and loss during fine-tuning the fifth-order cascade. d, The accuracy and loss of the fifth-order cascade during training from scratch. e, The output images and the output intensity distributions of the cascaded logical device.
Furthermore, the authors reveal a pioneering application of our device in optical time division multiplexing (OTDM), demonstrating its capability to manage high-speed data transfer seamlessly without the need for electronic conversion. Extensive simulations and experimental validations highlight the potential of this model as a foundational technology for future optical computing architectures, paving the way toward more sustainable and efficient optical data processing platforms.
Fig. 3 | The framework and simulated results of the novel OTDM system. a, Framework of novel OTDM for information transfer with our combinational logic units. b, The workflow of OTDM using the all-optical multiplexer and demultiplexer over time. c, Part of the simulated results of the OTDM under a consistent clock signal. d, The relationship between the number of layers used in MUX as well as DEMUX and crosstalk.
参考文献
1. Gao, H. et al. All-optical combinational logical units featuring fifth-order cascade. Chip 3, 100112 (2024).
2. Qian, C. et al. Performing optical logic operations by a diffractive neural network. Light Sci. Appl. 9, 59 (2020).
3. Zhao, Z. et al. Deep learning-enabled compact optical trigonometric operator with metasurface. PhotoniX 3, 15 (2022).
4. Işıl, Ç. et al. All-optical image denoising using a diffractive visual processor. Light Sci. Appl. 13, 43 (2024).
5. Luo, Y., Mengu, D. & Ozcan, A. Cascadable all-optical NAND gates using diffractive networks. Sci. Rep. 12, 7121 (2022).
6. Lin, X. et al. All-optical machine learning using diffractive deep neural networks. Science 361, 1004–1008 (2018).
论文链接：
https://www.sciencedirect.com/science/article/pii/S2709472324000303
作者简介
高海淇，国科大杭州高等研究院在读硕士生，研究方向主要为智能光计算的原理、技术及应用。
Haiqi Gao is a graduate student at the Hangzhou Institute for Advanced Study, UCAS. His primary research focuses on the principle, techniques and application of intelligent light computing.
邵煜，中国科学院大学杭州高等研究院在读博士，主要研究方向是智能光计算的原理、技术及应用，以及微纳器件的加工。
Yu Shao is a current PhD candidate at the Hangzhou Institute for Advanced Study, UCAS. His primary research focuses on the principles, techniques and applications of intelligent optical computing, as well as fabrication of micro/nano-devices.
杨陈楹，国科大杭州高等研究院副研究员。2016年博士毕业于浙江大学，研究方向包括光学结构功能器件、多维光学感知以及智能光计算，在重要学术期刊发表SCI论文30余篇。
Chenying Yang is an associate research fellow in Hangzhou Institute for Advanced Study, UCAS. He received his PHD degree from Zhejiang University in 2016. His research interests include nanostructured photonic devices, multi-dimensional optical perception and smart optical computing. He has published more than 30 articles in recent years.
沈伟东，浙江大学教授。2006年博士毕业于浙江大学，研究方向包括光学、光电子薄膜以及微纳光子器件，在重要学术期刊发表SCI论文130余篇。
Weidong Shen is a professor in Zhejiang University. He received his PHD degree from Zhejiang University in 2006. His research interests include optical coatings, optoelectronic films and nanostructured photonic devices. He has published more than 130 articles in recent years.
关于Chip
Chip（ISSN：2772-2724，CN：31-2189/O4）是全球唯一聚焦芯片类研究的综合性国际期刊，已入选由中国科协、教育部、科技部、中科院等单位联合实施的「中国科技期刊卓越行动计划高起点新刊项目」、「中国科技期刊卓越行动计划二期项目-英文梯队期刊」，为科技部鼓励发表「三类高质量论文」期刊之一。
爱思唯尔Chip官网：
往期鉴赏
