Spectral data are commonly used to study nonlinear instabilities in optical fibers. However, these provide very limited information on the properties of the wave field in the time domain. A new approach based on artificial intelligence (AI) applied by researchers at Tampere University (Tampere, Finland) and the University of Franche-Comté (Besançon, France) can provide insight into the time-domain characteristics of optical-fiber modulation instabilities, enabling the prediction of high-intensity peaks from the spectral intensity only.
In the experiment, pulses from a Ti:sapphire mode-locked laser were injected into a nonlinear photonic-crystal fiber in which a random wave field developed as a result of modulation instability. The output spectrum of thousands of pulses was measured in real time with very high dynamic range by a moving mirror and grating system. Subsequent analysis of the recorded data using AI from a neural network trained from numerical simulations yielded information on the temporal properties of the wave field associated with those spectra, and in particular the probability distribution associated with the emergence of ultrahigh-intensity spectral peaks—known as rogue waves—that are generated in many natural environments. Modulation instability is a central process in physics that underpins the occurrence of extreme events, and in particular on the surface of the oceans. By showing that AI can predict the extreme intensity of an unstable wave field from only partial information, the work could represent a major step towards the predictability of rogue waves in general. Reference: M. Närhi et al., Nat. Commun., 9, 4923 (Nov. 22, 2018).
