Towards pulse compressing in high power fiber CPA systems: a spectral-dispersion pre-compensation method based on genetic algorithm

Le résumé fourni par la source

Fiber chirped pulse amplification (CPA) has been recognized as an efficient scheme for power and energy scaling of ultrafast lasers. However, limited by the gain narrowing effect and high-order dispersion mismatch, the duration of the amplified laser pulse would broaden with obvious quality degradation compared to that of the seed laser. To address this issue, previous practice is to empirically pre-shape the pulse spectrum or pre-compensate the high-order dispersion of the seed pulse. In this work, we propose a collaborative optimization strategy for the pulse width suppression of the fiber CPA system based on the genetic algorithm-driven spectral pre-shaping and high-order dispersion compensation. A weighted flatness metric combining global-local spectral features is exploited to pre-shape the pulse spectrum, while a stepwise dispersion optimization framework is utilized to compensate the dispersion. Specifically, the stepwise optimization strategy (spectrum pre-shaping followed by dispersion compensation) accelerates computational speed by 58% compared to concurrent spectral reshaping and dispersion adjustments, achieving equivalent pulse compression with 40% fewer algorithmic costs. This method allows for one to attain near-optimal ultrafast laser performance within minutes – a task conventionally requiring hours of experimental optimization. Our approach realizes effective broadening of the pulse spectrum from 5.573 nm to 9.941 nm, and correspondingly suppressing of the de-chirped pulse duration from 454 fs to 323 fs at an operation power of 50 W. Remaining challenges in suppressing nonlinear pedestal growth and speeding up the algorithm are being addressed through utilizing real-time multi-objective Kolmogorov-Arnold network algorithms.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.