A new publication from Opto-Electronic Sciences; DOI 10.29026/oes.2024.240014 , discusses how AI and physics unite for meta-antennas design.
Ka-band metasurface antennas, with their low-cost, low-profile design and superior beam-steering capabilities, show significant potential in the field of satellite communications. However, the constraints of limited satellite resources and significant atmospheric losses at Ka-band frequencies require these antennas to achieve wide-angle beam scanning capabilities and high antenna gain, adding considerable complexity to their design. In order to achieve the design of a multifunctional and highly efficient meta-antenna, the design optimization will involve numerous parameters, greatly increasing the use of computational resources and optimization time. Addressing the critical issue of balancing multiple optimization objectives, such as gain and scanning angle, while improving optimization speed, remains a key challenge in the design process.
To address these challenges of meta-antenna design, researchers from the University of Electronic Science and Technology of China, Tongji University, and City University of Hong Kong have joined forces in an extensive collaboration. Leveraging their long-term expertise in the field of meta-optics, they proposed a Ka-band meta-antenna design method based on a Physics-Assisted Particle Swarm Optimization (PA-PSO) algorithm. Using this method, they designed and fabricated a Ka-band meta-antenna.
The antenna proposed in the paper is designed using the PA-PSO algorithm. Compared to the traditional PSO algorithm, the optimization direction of particles in the PA-PSO algorithm is guided by extremum conditions derived from the variational method. This not only reduces computation time but also decreases the likelihood of finding suboptimal designs, as shown in Figs. 1c−1d. The final optimized results indicate that the relative strength achieved by the PA-PSO algorithm is 94.62806, which is comparable to the relative strength of 94.62786 achieved by the traditional PSO algorithm. However, the computational cost of the PA-PSO algorithm is significantly lower; it reaches the optimal state after only 650 iterations, whereas the traditional PSO algorithm requires 4100 iterations. This means the computation time of the PA-PSO algorithm is less than one-sixth of that for the PSO algorithm. Therefore, the PA-PSO method can guide particle swarms more efficiently, reducing computation time, making it an important tool for addressing complex multivariate and multi-objective optimization challenges.
The purple line shows the calculation errors. The four hexagons from bottom to top represent phase distributions at different stages: initial phase distribution, PSO algorithm iteration 650 times, PSO algorithm iteration 1500 times, and PSO algorithm iteration 4100 times (PA-PSO algorithm iteration 650 times). (b) Comparison of FOVs and F/D for planar lens antennas. The colors of the points indicate the fluctuation of gains when scanning within the field of view range.
Based on the phase distribution optimized by the PA-PSO algorithm, the team designed and fabricated a hexagonal meta-antenna sample with a focal length of 22 mm, diagonal length of 110 mm, and a thickness of only 1.524 mm. As shown in Figure 3, the antenna has an f-number of only 0.2, a beam scanning angle of ±55°, a maximum gain of 21.7 dBi, and a gain flatness of within 4 dB. This innovative hexagonal meta-antenna, with its wide scanning angle, compact design, and high transmission gain, exhibits enormous potential for applications in satellite communication, radar systems, 5G networks, and the Internet of Things, among many other fields.
Keywords: multiple-feed lens antennas / PA-PSO algorithm / metalens / metasurfaces / Ka-band antenna.
