Phase control method for generating orbital angular momentum beams using coherent beam combining based on deep learning

In recent years, orbital angular momentum (OAM) beams have shown great potential for applications in laser communication, laser processing, optical imaging, and detection. For free-space optical communication, high-power, high-quality vortex beams with a high signal-to-noise ratio are critical for long-distance communication. Coherent beam combining (CBC) of vortex beams enables the enhancement of power while maintaining high beam quality. Considering the orbital angular momentum spectrum as a new dimension of optical wave resources, achieving rapid phase locking of specific phases is crucial for increasing communication capacity. Traditional phase control methods based on wavefront intensity distribution face limitations in optimization, particularly for centrosymmetric laser phased arrays. To address this, we propose a deep learning-based method using spiral phase modulation. By designing a loss function that eliminates phase periodicity, we establish a nonlinear mapping between the sub-beam phases and the far-field image. To improve the phase prediction accuracy of the deep learning model, we introduce a power-in-the-bucket (PIB) metric for the vortex beam's main lobe, which mitigates dynamic phase errors caused by thermal and environmental disturbances. This method holds promise for application in high-power vortex beam optical systems with coherent combining of fiber laser phased arrays.