In this study, we employ a deep neural network(DNN) to achieve rapid prediction of dispersion and effective mode area in Pb-silicate photonic crystal fiber. Subsequently, this prediction model is used to drive a particle swarm optimization(PSO) algorithm to design tapered Pb-silicate photonic crystal fibers for pulse compression. The designed fibers exhibit an increasing nonlinear index profile, enabling self-similar pulse compression(SSPC) of pulses at a wavelength of 2μm. In compliance with fundamental soliton conditions, a 0.566 m tapered PCF achieves notable pulse compression. It reduces the initial pulse width of 1.76 ps to 52 fs, achieving a remarkable compression factor of 33.92, while maintaining a quality factor of 1.01 and less than 5.4% pedestal energy. In scenarios where fundamental soliton conditions are not met, a 1.74 m tapered PCF also achieves significant compression. It compresses a pulse with a width of 1.76 ps and a peak power of 1 kW to a pulse width of 13.7 fs and a peak power of 24.82 kW, resulting in a compression factor of 129. The proposed approach overcomes the empirical reliance inherent in manual design and significantly reduces computational demands, enabling the design of pulse compression fiber tapers tailored to input pulse shapes.
Read full abstract