Genetic algorithm optimization of broadband operation in a noise-like pulse fiber laser

Abstract

The noise-like pulse regime of optical fiber lasers is highly complex, and associated with multiscale emission of random sub-picosecond pulses underneath a much longer envelope. With the addition of highly nonlinear fiber in the cavity, noise-like pulse lasers can also exhibit supercontinuum broadening and the generation of output spectra spanning 100's of nm. Achieving these broadest bandwidths, however, requires careful optimization of the nonlinear polarization rotation based saturable absorber, which involves a very large potential parameter space. Here we study the spectral characteristics of a broadband noise-like pulse laser by scanning the laser operation over a random sample of 50,000 polarization settings, and we quantify that these broadest bandwidths are generated in only [Formula: see text] 0.5% of cases. We also show that a genetic algorithm can replace trial and error optimization to align the cavity for these broadband operating states.

Figure 1

Schematic of the NLP laser and our experimental setup. Labels A-H refer to different fiber segments. The saturable absorber segment is between F and G. The figure also shows how feedback from the optical spectrum analyser (OSA) and the computed power spectrum from the oscilloscope (bottom left panel) are used as inputs to the genetic algorithm for laser optimization. ISO isolator; ${\theta }_1$ and ${\theta }_4$ quarter-wave plate, ${\theta }_2$ and ${\theta }_3$ half-wave plate; PBS polarizing beamsplitter, EDF Erbium-doped fiber, HNLF highly nonlinear fiber, WDM wavelength-division multiplexer.

Figure 2

(a) False-color plot mapping how the − 20 dB spectral bandwidth of the laser output varies as a function of waveplate angles ${\theta }_3$ and ${\theta }_4$ with pump power at 195 mW. (b) An expanded view over the region indicated by the white square in (a) to illustrate the rarity of operating states of bandwidths exceeding $\sim$ 100 nm (corresponding to red regions in the plot).

Figure 3

2000 spectra measured with the Anritsu OSA, selected randomly from the full scan (gray curves). Specific examples of a narrowband spectrum (dashed black line) and a broadband spectrum (solid black line) are also shown. The inset plots a broadband spectrum measured using the NIRQuest spectrometer to show spectral extension above 1750 nm.

Figure 4

Histogram of the measured − 20 dB bandwidths from the full cavity scan of 50,000 polarization settings of the saturable absorber. The main plot uses a logarithmic scale for the vertical axis whereas the inset shows an exploded view on a linear scale to illustrate the long-tailed nature of the distribution.

Figure 5

(a) Evolution of the objective function over 20 generations of the genetic algorithm. At each generation, we show the mean of the population (red stars) and the particular individuals corresponding to the minimum fitness values (black stars). (b) For the data in (a) we show the spectrum of the best individual for the generations as indicated on the right hand axis.