On type I cascaded quadratic soliton compression in lithium niobate: Compressing femtosecond pulses from high-power fiber lasers

TL;DR

This study numerically investigates cascaded quadratic soliton compression of femtosecond pulses in lithium niobate, revealing moderate compression capabilities and addressing challenges like nonlocal effects and spatial walk-off for high-quality pulse generation.

Contribution

It demonstrates the feasibility of pulse compression in LiNbO₃ using type I phase matching, analyzing the effects of nonlinearities, dispersion, and spatial walk-off on pulse quality and duration.

Findings

Moderate compression to sub-130 fs achievable in available crystal lengths.

Red-shifted long component in second harmonic can be filtered out for high-quality pulses.

Nonlocal effects limit compression below 100 fs in the nonstationary regime.

Abstract

The output pulses of a commercial high-power femtosecond fiber laser or amplifier are typically around 300-500 fs with a wavelength around 1030 nm and 10s of $\mu$J pulse energy. Here we present a numerical study of cascaded quadratic soliton compression of such pulses in LiNbO$_3$ using a type I phase matching configuration. We find that because of competing cubic material nonlinearities compression can only occur in the nonstationary regime, where group-velocity mismatch induced Raman-like nonlocal effects prevent compression to below 100 fs. However, the strong group velocity dispersion implies that the pulses can achieve moderate compression to sub-130 fs duration in available crystal lengths. Most of the pulse energy is conserved because the compression is moderate. The effects of diffraction and spatial walk-off is addressed, and in particular the latter could become an issue when compressing in such long crystals (around 10 cm long). We finally show that the second harmonic contains a short pulse locked to the pump and a long multi-ps red-shifted detrimental component. The latter is caused by the nonlocal effects in the nonstationary regime, but because it is strongly red-shifted to a position that can be predicted, we show that it can be removed using a bandpass filter, leaving a sub-100 fs visible component at $\lambda=515$ nm with excellent pulse quality.