Inverse design of figure eight fiber laser by artificial neural network

Abstract

Fiber lasers have become indispensable tools in modern photonics, offering unparalleled efficiency, stability, and versatility. Among them, the figure-eight fiber laser (F8FL) has gained prominence for its ability to generate ultra-short pulses with high peak power, making it highly suitable for applications in ultrafast spectroscopy, nonlinear microscopy, and optical frequency comb generation. However, designing and optimizing F8FLs remains a significant challenge due to the intricate interplay of nonlinear effects, dispersion management, and gain dynamics. Traditional design approaches rely on numerical simulations and iterative experimental tuning, which are computationally expensive and often yield suboptimal results. To address these challenges, we introduce a machine learning-based inverse design framework for optimizing F8FL parameters. Using a dataset generated from numerical simulations, an artificial neural network (ANN) is trained to establish a direct mapping between pulse characteristics and the key amplifier parameters, including small-signal gain and saturation energy. This approach enables rapid and accurate prediction of laser settings required to achieve a target pulse profile, significantly reducing the computational burden compared to conventional numerical methods. Our results demonstrate that the trained ANN model achieves excellent agreement with numerical simulations, effectively predicting the optimal parameters for producing high-energy rectangular pulses in the dissipative soliton resonance (DSR) regime. To validate the effectiveness of the predicted parameters, the ANN outputs were independently verified using OptiSystem simulations, confirming strong agreement with the desired pulse profiles. This study highlights the potential of machine learning in photonics, paving the way for the development of self-optimizing, adaptive laser systems with enhanced precision and efficiency. The proposed methodology can be extended to other nonlinear optical systems, offering a powerful tool for accelerating the design and optimization of advanced fiber lasers. • Neural networks optimize soliton properties, replacing grid search methods. • The approach applies to fiber lasers with diverse pulse generation regimes. • ANNs predict amplifier parameters, reducing complex numerical computations. • Simulations validate ANN-based soliton resonance optimization in fiber lasers. • OptiSystem simulations confirm the physical validity of ANN-predicted results.