Ultrafast true-green Ho:ZBLAN fiber laser inspired by the TD3 AI algorithm

Abstract

Ultrafast lasers in the true-green spectrum, which are scarce due to the “green gap” in semiconductor materials, are in high demand for the surging field of biomedical photonics. One ideal candidate for efficient green lasing is Ho:ZBLAN fiber, as ZBLAN-hosted fibers have already reached picosecond dissipative soliton resonance (DSR) in the yellow. When attempting to push the DSR mode locking further into the green, traditional manual cavity tuning is faced with extreme difficulty, as the emission regime for these fiber lasers is so deeply concealed. Breakthroughs in artificial intelligence (AI), however, provide the opportunity to fulfill the task in a fully automated manner. This work, inspired by the emerging twin delayed deep deterministic policy gradient (TD3) algorithm, represents the first application, to the best of our knowledge, of the TD3 AI algorithm to generate picosecond emissions at the unprecedented true-green wavelength of ∼545 nm. The study thus extends the ongoing AI technique further into the ultrafast photonics region.