Millimeter‐Wave Orbital Angular Momentum Mode Conversion Based on Multilayer Quasi‐Optical Diffractive Neural Network
Jiaqi Cai, Jiexin Lai, Yuda Wu, Shufang LiABSTRACT
Orbital angular momentum (OAM) modes provide additional degrees of freedom for high‐capacity communication, advanced radar, and quantum information processing in millimeter‐wave and terahertz bands. This work investigates a multilayer quasi‐optical phase diffraction network operating at 150 GHz, in which each layer is represented by a 3D‐printable dielectric‐column array. The phase profiles are optimized with a differentiable angular‐spectrum model and Adam, and the resulting 3D dielectric structures are evaluated by CST full‐wave electromagnetic simulation. The inter‐layer spacing and higher‐order channel weights are examined separately. The fixed structure is also evaluated over frequency, and a second optimized design is used to test nonuniform OAM mappings. All CST results use the same definitions of Mode Purity and target‐normalized XT. The resulting passive three‐layer structure uses fabrication‐compatible dielectric columns and supports the four designed OAM conversion channels.