Chongfan Technology
News
25
2025
-
03
ACS Photonics | Intelligent Recognition of Light Beam Angular Momentum Spectrum
Author:
Introduction
Recently,the team of Shiyao Fu and Chunqing Gao from Beijing Institute of Technologyand collaborators proposed ametasurface-based intelligent sensing methodthat achieves precise measurement of the total angular momentum spectrum of photons. By using a periodic gradient metasurface, the two spin angular momentum modes of the beam to be measured are simultaneously controlled, and two diffracted light fields are obtained, which are in one-to-one mapping with the corresponding orbital angular momentum spectrum. The research group used a convolutional neural network to analyze the mapping relationship between the intensity characteristics of this set of diffracted light fields and the corresponding total angular momentum spectrum, and finally achieved high-precision measurement of the total angular momentum spectrum of beams containing up to 34 total angular momentum modes, with a mean square error of 10-6order of magnitude.
This research result is titled “Metasurface-Based Intelligent Identification of Total Angular Momentum Spectra for Beams” and published in the journalACS PhotonicsSchool of Optoelectronics, Beijing Institute of TechnologyPhD student Lang Li、Master's student Liliang GaoandDr. Yuxin Cheng from China Electric Power Research Instituteare the co-first authors of this paper, and ProfessorShiyao Fu from the School of Optoelectronics, Beijing Institute of Technologyis the corresponding author.
Design ideas and identification process
The team's scheme for measuring the total angular momentum of a light beam is shown in Figure 1. After the beam to be measured is modulated by a periodic gradient metasurface, two diffracted light fields corresponding to the two spin angular momentum eigenmodes are generated on the two diffraction orders of the metasurface. After being received by the camera, they are identified by the convolutional neural network SADT-Net designed by the team. After being trained with a large amount of data, the SADT-Net network can identify the corresponding total angular momentum spectrum from the intensity distribution characteristics of this set of diffracted light fields.
Figure 1: Total angular momentum spectrum identification process
SADT-Net network structure and mode distribution characteristics
During the training process, the research team investigated the influence of different network depths on the identification effect, and statistically analyzed the identification errors of SADT-Net under different numbers of total angular momentum modes and in each total angular momentum mode channel. Since the network depth is often related to the size of the image features to be identified, under the condition that the image resolution remains unchanged, the identification errors of different network depths on the same training set and validation set were tested, as shown in Figure 2(c), and the network depth was determined to beM. After a set of diffracted light intensities shown in Figure 2(a1-a2) are identified by the SADT-Net shown in Figure 2(b), the network output and the true value of the total angular momentum spectrum are shown in Figure 2(d1), and Figure 2(d2) shows the original intensity distribution of the beam to be measured. As shown in Figure 2(e-f), due to the mode distribution characteristics of the dataset, the mean square error of SADT-Net in identifying beams carrying different numbers of total angular momentum modes and in identifying different total angular momentum mode channels also fluctuates.
Figure 2: SADT-Net network structure, network depth parameter test, identification effect of a set of diffracted spots, and identification error of the network under different numbers of modes and in each mode channel
Robustness analysis of the identification system
In order to test the noise immunity of SATD-Net, the research team selected a set of diffracted spots from the test set and showed the comparison results between the network output and the true value of its total angular momentum spectrum. As shown in Figure 3 (a1-a2), Gaussian noise with a standard deviation of 0 to 0.3 was added to the images received by the camera, and then SATD-Net was used to identify these images with different intensities of Gaussian noise, as shown in Figure 3 (b1-b2). The test results show that SATD-Net can still accurately identify the total angular momentum spectrum even when there is weak Gaussian noise interference in the spot images. Since the experimental system involves the center alignment of the beam to be measured and the periodic gradient metasurface, various errors will inevitably be introduced during the experiment. The research team investigated the influence of two main alignment errors: tilt error and rotation error on the identification of SATD-Net, as shown in Figure 3 (c1-d2). When there are small tilt errors and rotation errors in the experimental system, the identification mean square error of SATD-Net is almost unaffected, but when the errors increase, the diffracted spots will be deformed, which will affect the identification of the network and lead to a rapid increase in the identification error.
Figure 3: Robustness analysis of SATD-Net
Summary and outlook
The research team proposed a metasurface-based intelligent sensing method, which achieves high-precision identification on beams carrying different numbers of angular momentum modes. This method can simultaneously identify the total angular momentum spectrum containing up to 34 angular momentum states, and maintains high-precision identification within a certain range when dealing with camera image acquisition noise and experimental system alignment errors. This method can lay the foundation for the development of fields based on precise measurement of total angular momentum spectrum, such as high-capacity optical communication, quantum information processing, and novel laser detection.
Paper Information
Intelligent Recognition of Light Beam Total Angular Momentum Spectrum Based on Metasurface, Li Lang, Gao Liliang, Cheng Yuxin, Zhou Shiyun, Wang Jiaqi, Yu Haoran, Gao Chunqing, Fu ShiyaoACS Photonics2025 12 （3）， 1418-1425
https://doi.org/10.1021/acsphotonics.4c01930
LATEST NEWS
Thank you for visiting the official website of Chongfan Technology. If you have cooperation intentions or suggestions, please contact us through the following methods, and we will reply as soon as possible, thank you!
Address: Room 403, Building 6, Phase III of R&D, No. 36 Xiyong Avenue, High tech Zone, Chongqing, China.
Telephone: +86-13658337211
E-mail: Sales@cfkeji.net
Website: www.cfkeji.net
Mobile Version
