Machine Learning Assisted Rapid Design of Terahertz Vortex Beam Metasurface
Generation and modulation of terahertz vortex beams is considered to be one of the key technologies in 6G communica-tion,radar detection and new sensors.The use of artificial metasurface to generate vortex beams has characteristics of planarization,integration and low cost compared with traditional method,but it is faced with complex parameter design and analysis,which requires a lot of time and computing resource.Therefore,we propose a machine learn-assisted metasurface design method for terahertz vortex beams.By using Long Short-Term Memory(LSTM)neural network,we can quickly obtain the meta-atoms that meet the phase require-ment and carry out an integrated array.Three terahertz vortex beams metasurfaces with modes l of 1,2,3 are designed by this method.Results show that mode purity of the designed terahertz vortex beam is more than 80％.Machine learning-assisted terahertz vortex beam metasurfaces design method has the advantages of high precision,fast,integrated design,etc.,and is expected to be used in terahertz amplitude,phase,polarization,orbital angular momentum and other complex beam control.
万方数据
维普
