Deep Learning of Coherent Laser Arrays in Angular Domain for Orbital Angular Momentum Beams Customization

Abstract

Artificial intelligence methods for the phase control of laser arrays have been extensively developed in the recent years. To tailor high-power and fast switchable orbital angular momentum (OAM) beams by the coherent combination of laser arrays, complex phase control difficulties arise, and the introduction of deep-learning (DL) methods to learn the mapping from the intensity information to the relative phases of array elements offers a promising solution. In this work, we theoretically propose a DL-based phase control scheme that can customize OAM beams with desirable topological charges from laser array systems, which relies on the acquisition of the optical field information in the view of angular domain. When the system in closed loops, the optical field and phase errors information of laser array are extracted and decoupled in the angular domain perspective by using an OAM demultiplexer. The processed optical field of the coherently combined beam serves the DL-based phase control servo to realize the efficient phase stabilization, thus ensuring the controllable customization of OAM beams. This work could provide a useful reference on the intelligent control of laser array systems for structured light beams customization.