Intelligent controllable ultrafast fiber laser via deep learning and adaptive optimization algorithm

Abstract

Ultrafast fiber lasers based on nonlinear polarization rotation can generate femtosecond pulses with different pulse durations and high peak powers, which are powerful tools for engineering applications and scientific research. However, achieving a precise and repeatable polarization state for generating the ultrashort pulses with the shortest pulse duration remains a significant challenge. In this paper, we extend the use of recurrent neural networks and adaptive optimization algorithms, specifically designed to optimize repetitive processes in optical systems, to facilitate intelligent search and control aimed at achieving the minimum pulse duration within a mode-locked fiber laser cavity. Our multi-algorithm-based intelligent system can fully simulate and optimize the processes involved in hands-on experiments. Our intelligent system identified a mode-locked fiber laser with the shortest pulse duration of 465 fs, which was experimentally verified. The proposed intelligent algorithm not only identifies the shortest pulse but also holds significant potential for selecting related laser characteristic parameters. We believe this work opens up a novel avenue for exploration and optimization in mode-locked lasers and the intelligent laser can find practical applications in engineering and scientific research.