Banner
This article is submitted by the authors of the paper
Introduction
Recently, the team of Aydogan Ozcan from the University of California, Los Angeles, designed an innovative deep diffraction neural network structure, called Pyramid Deep Diffraction Neural Network (P–D2NN).
This structure can perform unidirectional image magnification or demagnification while significantly reducing the required number of diffraction features, meaning it can enlarge or shrink images during forward propagation while preventing any image generation during backward propagation. This design can operate over a wide range of illumination wavelengths and can simultaneously perform unidirectional image magnification and demagnification at different wavelengths. The researchers experimentally validated the P–D2NN framework under terahertz illumination. This new structure shows great potential applications in optical communication, surveillance, and photonic device isolation.
The work was published under the title Pyramid diffractive optical networks for unidirectional image magnification and demagnification in Light: Science & Applications.
Figure 1: Artistic depiction of the pyramid diffraction optical network for unidirectional image magnification and demagnification
The deep diffraction neural network (P–D2NN) is an optical system composed of multiple continuous diffraction surfaces in free space, optimized by deep learning, capable of performing various computational tasks in an all-optical manner. The research team led by Professor Aydogan Ozcan at UCLA developed a pyramid-structured diffractive optical network, where the size of the diffraction surfaces increases or decreases in a pyramid shape, consistent with the trend of image magnification or demagnification. This design uses fewer diffraction neurons to achieve unidirectional imaging, ensuring high fidelity in forward imaging while suppressing image generation in the reverse direction. The researchers also demonstrated higher magnification by cascading multiple P-D2NN modules, showcasing the modularity and scalability of the system.
The P-D2NN framework was experimentally validated under terahertz (THz) illumination. The diffraction layers were manufactured using 3D printing and tested under continuous-wave terahertz illumination. The experimental results for three different designs (containing different magnification and demagnification ratios) were highly consistent with numerical simulations. During the forward imaging process, the output plane accurately reflected the magnified or demagnified version of the input image, while a low-intensity light field with no useful information was generated during the backward transmission process, fulfilling the requirements of unidirectional imaging tasks.
P-D2NN framework can reduce the energy transmitted in the reverse direction and disperse its original signal into imperceptible noise, making it perform well in various application scenarios. These applications include optical isolation for photonic devices, decoupling transmitters and receivers in telecommunications, and privacy-preserving optical communication and surveillance.Additionally, the system is insensitive to polarization and can focus high-power structured beams onto target objects while simultaneously protecting the source from counterattacks, highlighting its potential in various defense-related applications.
Paper Information
Bai, B., Yang, X., Gan, T. et al. Pyramid diffractive optical networks for unidirectional image magnification and demagnification. Light Sci Appl 13, 178 (2024).
https://doi.org/10.1038/s41377-024-01543-w
Banner
Highly Cited Article Statistics
The following data comes from Web of Science, with the number of highly cited articles in Light: Science & Applications consistently leading among similar domestic journals. As of now:
Articles with over 2000 citations: 1 article
https://doi.org/10.1038/lsa.2014.99
Articles with over 1000 citations: 3 articles
https://doi.org/10.1038/lsa.2014.30
https://doi.org/10.1038/s41377-019-0194-2
Articles with over 900 citations: 4 articles
https://doi.org/10.1038/lsa.2016.133
Articles with over 700 citations: 8 articles
https://doi.org/10.1038/lsa.2014.48https://doi.org/10.1038/lsa.2017.141https://doi.org/10.1038/lsa.2017.168https://doi.org/10.1038/s41377-020-0341-9
Articles with over 600 citations: 9 articles
https://doi.org/10.1038/lsa.2013.28
Articles with over 500 citations: 14 articles
https://doi.org/10.1038/lsa.2015.67https://doi.org/10.1038/lsa.2014.60https://doi.org/10.1038/lsa.2013.26https://doi.org/10.1038/lsa.2014.46
https://doi.org/10.1038/s41377-018-0078-x
Articles with over 400 citations: 30 articleshttps://doi.org/10.1038/lsa.2014.42https://doi.org/10.1038/lsa.2016.17https://doi.org/10.1038/lsa.2015.30https://doi.org/10.1038/lsa.2015.97https://doi.org/10.1038/s41377-020-0326-8https://doi.org/10.1038/lsa.2015.131https://doi.org/10.1038/s41377-021-00658-8https://doi.org/10.1038/s41377-018-0060-7https://doi.org/10.1038/lsa.2017.39https://doi.org/10.1038/lsa.2016.76https://doi.org/10.1038/lsa.2012.1https://doi.org/10.1038/s41377-020-0264-5https://doi.org/10.1038/lsa.2017.146https://doi.org/10.1038/s41377-020-0268-1https://doi.org/10.1038/lsa.2014.94https://doi.org/10.1038/s41377-019-0148-8
Articles with over 300 citations: 55 articles
Articles with over 200 citations: 129 articlesArticles with over 100 citations: 331 articles
Articles with over 50 citations: 639 articles
Welcome to submit promotional articles for the research group
Please scan the code to contact the duty editor
👇 Follow me 👇
Light up the “like” and “looking” to not miss article updates