Artificial Intelligence Enabled Spectral Reconfigurable Fiber Laser
The combinations of artificial intelligence and lasers provide powerful ways to form smart light sources with ground-breaking functions. Here, a Raman fiber laser (RFL) with reconfigurable and programmable spectra in an ultra-wide bandwidth is developed based on spectral-spatial manipulation of light in multimode fiber (MMF). The proposed fiber laser uses nonlinear gain from cascaded stimulated Raman scattering, random distributed feedback from Rayleigh scattering, and point feedback from an MMF-based smart spectral filter. Through wavefront shaping controlled by a genetic algorithm, light of selective wavelength(s) can be focused in the MMF, forming the filter that, together with the active part of the laser, actively shape the output spectrum with a high degree of freedom. We achieved arbitrary spectral shaping of the cascaded RFL (e.g., continuously tunable single-wavelength and multi-wavelength laser with customizable linewidth, mode separation, and power distribution) from the 1st- to the 3rd-order Stokes emission by adjusting the pump power and auto-optimization of the smart filter. Our research uses artificial-intelligence controlled light manipulation in a fiber platform with multi-eigenmodes and nonlinear gain, mapping the spatial control into the spectral domain as well as extending the linear control of light in MMF to active light emission, which is of great significance for applications in optical communication, sensing, and spectroscopy.