Optical fibers transmit information at the speed of light over thousands of kilometers. These properties have enabled their massive deployment to carry global computer, audiovisual, and telephone data. They have also revolutionized medical imaging, sensors, and indoor lighting. Yet a limitation has persisted since the technology was developed in the 1970s: modulation instability.
This phenomenon corresponds to an amplification of small disturbances (noise) arising from the interaction between the material's nonlinearity and the natural dispersion of light. As signal intensity increases, the noise strengthens and triggers a chaotic cascade. Until now, scientists had only managed to marginally mitigate it.
Illustration image.
© Benjamin Wetzel
In July 2025, this paradigm was overturned. The team of Benjamin Wetzel harnessed the power of artificial neural networks to decode, predict, and control these chaotic dynamics.
The scientists adopted an approach of injecting weak coherent optical pulses to alter the properties of the chaos linked to modulation instability. Paradoxical? Perhaps, but these injected signals are deliberately controlled, unlike the random noise of modulation instability. And the results show they prove effective at interfering with the chaos and restructuring it.
To develop this strategy, the team first simulated nearly 100,000 scenarios using artificial intelligence and tested 500 noise configurations for each of them. The network learns to decode hidden information, camouflaged by layers of nonlinear processes and noise. In parallel, an experimental platform tested ultra-short optical pulses traveling through several hundred meters (roughly 1,000 feet) of optical fiber.
At the output, two measurement approaches—a classical optical analyzer and a real-time detection architecture—demonstrate the method's effectiveness. The hidden parameters of the chaos are revealed, and spectral fluctuations are predicted with unprecedented accuracy.
These results are promising on several fronts: secure communications, real-time imaging, and quantum optics. Better-understood nonlinearities can be exploited to encrypt information within the optical chaos. The team thus envisions more sophisticated network architectures to transform optical fibers into more efficient and secure systems.