Mapping 2D Spatial Orbital Angular Momentum Information of Light to 1D Spatial and Temporal Domain

Orbital Angular Momentum (OAM) beams, characterized by their 2D spatial twisted phase fronts and doughnut intensity profiles allow them to carry additional information. Due to their 2D spatial phase and intensity profiles, real-time OAM recognition is challenging in high-speed OAM applications. Capturing 2D OAM information requires complex, resource-intensive systems, hindering real-time processing. To overcome this, we developed a speckle-learned technique that maps 2D spatial OAM information to 1D spatial / temporal domains. The unique speckle patterns, corresponding to the azimuthal and radial indices, enable high-fidelity OAM recognition using 1D data in both domains. The mapped OAM features onto 1D speckle information are 1/n^th of the corresponding 2D information. We have employed a custom-designed lightweight 1D CNN and a standard SVM model to classify OAM using its 1D spatial or temporal speckle information with accuracies 98 / 99.5% respectively. The spatial and temporal 1D speckle information accelerates the OAM training and recognition process with high speed and accuracy with minimal computational overhead. This approach not only simplifies hardware requirements but also enables real-time detection in resource-constrained environments, with potential applications in high-speed communication, medical diagnostics, and on-chip photonic processors.