Statistical Methods for Determining Turbulence in Supercontinuum Generation

Abstract

Distinguishing coherent, turbulent, and chaotic operating regimes in supercontinuum generation is important for understanding nonlinear optical dynamics and optimizing broadband light sources. Experimentally identifying the onset of turbulence remains challenging because the most common metric, first-order coherence, requires access to the complex optical field and cannot be directly obtained from intensity-only measurements. In this work, we investigate whether experimentally accessible statistical observables can identify turbulence in supercontinuum generation. We compare wavelength-integrated variance and kurtosis with simulation-based first-order coherence over a chirp-controlled pulse-duration sweep implemented through additional $\beta_2$ dispersion. The study combines generalized nonlinear Schr\"odinger equation simulations with shot-to-shot dispersive Fourier transform measurements validated against optical spectrum analyzer spectra. Statistical intensity distributions were analyzed using histograms, complementary cumulative distribution functions, and kurtosis measurements across the generated supercontinuum bandwidth. Simulations and experiments both revealed heavy-tailed intensity statistics in the intermediate pulse-duration regime associated with reduced spectral coherence. The integrated kurtosis reached a maximum near 600 fs in simulations and near 700 fs in experiments, while the integrated variance within the first 20 dB spectral range decreased with increasing pulse duration. The agreement between simulations and experiments demonstrates that variance- and kurtosis-based observables can serve as experimentally accessible indicators of turbulence in supercontinuum generation. These results show that intensity-only statistical measurements can distinguish coherent and incoherent operating regimes without requiring direct field-resolved coherence measurements.