Source-linked AI summary
Towards 'smart lasers': self-optimisation of an ultrafast pulse source using a genetic algorithm
R. I. Woodward, E. J. R. Kelleher
TL;DR
Ultrafast single-pulse CW mode-locked sources enable manufacturing, research, and medical imaging, where pulse-to-pulse stability and system reliability are critical. A genetic algorithm searches laser cavity settings using a compound fitness function that combines temporal and spectral diagnostics. Across ten initially ‘off’ realisations, the laser consistently converged to stable fundamentally mode-locked output in an average of ∼20 generations, typically in less than 30 minutes.
Problem
Ultrafast single-pulse CW mode-locked sources enable manufacturing, research, and medical imaging, where pulse-to-pulse stability and system reliability are critical.
Source: Choice of Fitness Function · p. 4Method
A genetic algorithm searches laser cavity settings using a compound fitness function that combines temporal and spectral diagnostics.
Source: Genetic Algorithm Development · p. 3, Choice of Fitness Function · p. 4Results
Across ten initially ‘off’ realisations, the laser consistently converged to stable fundamentally mode-locked output in an average of ∼20 generations, typically in less than 30 minutes.
Source: Results and Discussion · p. 6Takeaways & Limitations
The approach could benefit pulsed lasers with real or artificial saturable absorbers, although real absorbers restrict the available operating-regime space.
Source: Conclusion · p. 7The study focuses on single-pulse mode-locking; targeting other regimes requires defining a corresponding fitness function.
Source: Conclusion · p. 7
Abstract
from arXiv · showhide
Short-pulse fibre lasers are a complex dynamical system possessing a broad space of operating states that can be accessed through control of cavity parameters. Determination of target regimes is a multi-parameter global optimisation problem. Here, we report the implementation of a genetic algorithm to intelligently locate optimum parameters for stable single-pulse mode-locking in a Figure-8 fibre laser, and fully automate the system turn-on procedure. Stable ultrashort pulses are repeatably achieved by employing a compound fitness function that monitors both temporal and spectral output properties of the laser. Our method of encoding photonics expertise into an algorithm and applying machine-learning principles paves the way to self-optimising `smart' optical technologies.
Self-Optimising Laser Design
The Figure-8 laser’s operating regime depends on cavity polarization and power settings, while stable single-pulse states occupy sparse regions of the parameter space. A genetic algorithm searches these settings using a compound fitness score that combines temporal and spectral diagnostics to distinguish target operation from other regimes.
Operating States in Fibre Lasers
- Stable single-pulse continuous-wave mode-locking occurs in sparse, localized regions of the waveplate parameter space and operates at the 7.4 MHz fundamental cavity frequency.The parameter sweep holds pump power constant while two of four waveplates are varied; intermediate scores include Q-switching and unstable or partial multiple-pulse regimes.Source: Operating States in Fibre Lasers · p. 2, Figure 2 · p. 3
Genetic Algorithm Development
- The genetic algorithm scores each parameter-setting individual, then breeds the next generation through score-weighted parent selection, gene mixing, mutation, and elitist preservation.Individuals are evaluated by electronically setting cavity parameters and measuring the resulting laser output.Source: Genetic Algorithm Development · p. 3, Figure 3 · p. 4
- The implemented population contains 30 individuals with five genes: four waveplate angles and pump current, allowing the algorithm to automate the full turn-on cycle.Pump-current bounds keep the diode above threshold and prevent damage; mutation starts at 25% and is damped as the search progresses, while the best four individuals are cloned into each generation.Source: Genetic Algorithm Development · p. 3
Choice of Fitness Function
- The compound score equally weights an oscilloscope score for a single dominant pulse, an electrical-spectrum score favoring the fundamental while penalizing harmonics’ amplitude fluctuations, and an optical-spectrum score favoring broad bandwidth.An autocorrelator could add temporal information, but the authors state its polarization sensitivity prevents use with non-stationary output polarization.Source: Choice of Fitness Function · pp. 4–5
- The compound fitness function provides the highest contrast between stable and unstable pulsating regimes, whereas individual diagnostics miss distinctions among mode-locking states.Optical bandwidth distinguishes Q-switched from mode-locked regimes but not stable single-pulse operation; electrical harmonics highlight mode-locking, while the oscilloscope helps distinguish single- from multiple-pulse operation.Source: Choice of Fitness Function · p. 5, Figure 4 · p. 5
- Without resetting the laser between parameter adjustments, hysteresis extends high-fitness regions, but the overall stable-operation pattern remains largely unchanged.Some high-scoring hysteretic states may not self-start reliably; the algorithm’s dependence on prior system state is expected to favor states that repeat over time.Source: Choice of Fitness Function · p. 5, Figure 4 · p. 5
Results and Discussion
The compound-fitness genetic algorithm converges from random, initially off states to stable mode-locking, with repeatable results and recovery after a mechanical perturbation. Multiple diagnostics support stable sub-picosecond pulses, while diagnostic bandwidth limits pulse-duration optimization.
Results and Discussion
- Starting from random polarisation parameters and sub-threshold pump power, the algorithm evolves from mostly non-lasing or continuous-wave states through Q-switched operation to stable single-pulse mode-locking.Elitism preserves strong individuals, while crossover and mutation enable improved regimes to be found beyond an early Q-switched local maximum.Source: Results and Discussion · p. 6
- Across ten initially off realisations, the laser converges to stable fundamental mode-locking in about 20 generations and typically under 30 minutes.Remote interfacing with controllers and diagnostics, plus cavity settling, dominates convergence time.Source: Results and Discussion · p. 6, Figure 5 · p. 6
- Compound fitness using multiple diagnostics repeatedly produces stable pulses with sub-picosecond durations, unlike the unstable multi-pulsing seen with a singular fitness function.Pulse widths vary by about 150 fs between realisations because the fitness diagnostics lack bandwidth for real-time ultrafast characterisation and pulse-duration optimisation.Source: Results and Discussion · p. 6, Figure 6 · p. 7
- After a mechanical cavity perturbation alters fibre birefringence and NALM phase bias, the algorithm recovers optimum mode-locking by increasing mutation and exploring new parameter settings.The fitness-score drop raises mutation after its rate had been damped during stable operation.Source: Results and Discussion · pp. 6–7, Figure 6 · p. 7
Conclusion
The demonstrated Figure-8 fibre laser uses a genetic algorithm and a compound fitness function to target stable mode-locked ultrashort pulses. The approach is best suited to electronically controlled artificial saturable absorbers, while faster convergence and other target regimes remain areas for improvement.
Conclusion
- The compound fitness function assesses temporal and spectral laser outputs, and the genetic algorithm maximizes its score to target stable mode-locked ultrashort pulses.The demonstrated system is a self-optimising ultrafast Figure-8 fibre laser.Source: Conclusion · p. 7
- The approach is ideally suited to mode-locked lasers with artificial saturable absorbers whose nonlinear transfer function can be controlled electronically.It could also benefit lasers with real saturable absorbers, although their fixed transfer function restricts the operating-regime space.Source: Conclusion · p. 7
- The reported target is single-pulse mode-locking; convergence may be reduced through algorithm tuning or electronic integration, and other operating regimes would require a suitable fitness function.All-normal-dispersion and long-cavity lasers are proposed as potential applications; without soliton pulse shaping, all-normal-dispersion dynamics make stable pulsation more challenging.Source: Conclusion · p. 7
Methods
The Figure-8 cavity combines a passive unidirectional ring with an active bidirectional NALM loop and electronic polarization control. Its output is split between fitness diagnostics and the stabilized laser output.
Methods
- The Figure-8 design pairs a passive unidirectional ring with an active bidirectional nonlinear amplifying loop mirror (NALM).The passive ring contains an isolator and 10% output coupler; the NALM contains 2.3 m of ytterbium-erbium co-doped fibre.Source: Methods · p. 8
- Four stepper motor-controlled fibre-loop quarter waveplates form the electronic polarisation controller, with 0.18-degree rotation resolution and full Poincaré-sphere traversal.Traversal is enabled by stress-induced birefringence.Source: Methods · p. 8
- The approximately 28 m cavity has approximately -0.6 ps^2 group-delay dispersion, and its output is split between fitness diagnostics and the GA-stabilised laser output.The dispersion indicates that soliton pulse shaping is expected.Source: Methods · p. 8
Author contributions statement
Both authors contributed to experimental design, genetic algorithm development, and manuscript writing.
Author contributions statement
- Both authors contributed to experimental design, genetic algorithm development, and writing the manuscript.Source: Author contributions statement · p. 9