Machine Learning Methods for Control of Fibre Lasers with Double Gain Nonlinear Loop Mirror

Area of Science:

• Nonlinear optics
• Laser physics
• Machine learning applications

Background:

• Modern lasers, especially fiber lasers, exhibit complex nonlinear properties.
• High-power fiber laser systems often display multiple equilibrium states, complicating control and reproducibility.
• Machine learning offers potential solutions for managing complex laser dynamics.

Purpose of the Study:

• To experimentally demonstrate machine learning for controlling pulsed regimes in a figure-eight fiber laser.
• To investigate the use of self-tuning machine learning for on-demand pulse generation.
• To evaluate objective functions for controlling pulse duration, energy, and temporal coherence.

Main Methods:

• Utilized an all-normal dispersion, figure-eight fiber laser with two independent amplifying loops.
• Employed machine learning algorithms to electronically control two drive currents, adjusting gain levels.
• Defined and assessed objective functions for pulse characteristics selection.

Main Results:

• Successfully demonstrated experimental application of machine learning for pulsed regime control.
• Achieved generation-on-demand pulses with controllable duration, energy, spectral characteristics, and time coherence.
• Validated the effectiveness of self-tuning machine learning in managing laser operation states.

Conclusions:

• Machine learning provides a robust method for controlling pulsed fiber lasers.
• This approach enables precise, on-demand tailoring of laser pulse properties.
• Opens possibilities for new designs of electronically-managed pulsed fiber lasers.