Using AI To Control Properties Of Light | Supercontinuum Generation

  • Researchers used a photonic chip and an AI algorithm to configure the properties of broadband light sources. 
  • The technology will help in the development of various smart optical systems through self-optimization methods. 

In our everyday life, we use several complex systems that rely on a large number of parameters based on chaotic dynamics. In photonics, many systems fall under this category, including advanced optical sources that are used in metrology, laser science, and biomedical imaging.

To make such techniques better and control the properties of light effectively, it’s necessary to keep pushing the limits of photonic methodologies. Over the last few couples of years, scientists all over the world have been trying to generate Supercontinuum – a broadband spectrum created by an optical pulse propagating under a combined effect of scattering, dispersion, and nonlinearities.

The development of ultrashort and intense laser pulses – which led to the 2018 Physics Nobel Prize – along with techniques of spatially confining and guiding propagation of light gave rise to extremely powerful optical architectures.

Recently, a research team at the Institut National de la Recherche Scientifique, Canada, successfully generated and manipulated intense ultrashort pulse patterns to generate Supercontinuum. They used integrated photonic structures to create reconfigurable bunches of femtosecond optical pulses.

What Exactly Did They Use To Generate Supercontinuum?

In this work, researchers have demonstrated diverse patterns of ultrashort pulses that can be manipulated in a controlled manner. They harnessed the stability, compactness and sub-nanometer resolution provided by an integrated photonic system to create femtoseconds optical pulses.

Reference: Nature Communications | doi:10.1038/s41467-018-07141-w | INRS 

They scaled the parameter space exponentially, which yielded more than 1036 different combinations of possible pulse patterns. For such a large number of combinations — larger than the total number of planets in the universe — the team used a machine learning method to analyze the results of light manipulation.

Pulses separated by one picosecond | Courtesy of Benjamin Wetzel

With an appropriate AI algorithm, researchers were able to optimize different patterns of pulses and achieve the desired Supercontinuum results. They measured spectral output and applied a genetic algorithm to alter the integrated pulse-splitter configurations in order to enhance the dynamics of nonlinear fiber propagation towards a particular Supercontinuum criterion, for example, increasing the spectral intensity at certain wavelengths.


The technique allowed researchers to experimentally obtain 7 times more Supercontinuum spectral density than a single pulse excitation with the same power. It has the potential to provide a complete temporal control of Supercontinuum generation. The fabulous outcomes will affect applied research in various fields.

Read: Existing Laser Technology Is Strong Enough To Attract Aliens 20,000 Light Years Away

Especially, it will aid in developing other smart optical structures through self-optimization methods, including pulse amplification, self-adjusting lasers, optical frequency combs, and fundamental AI approaches like photonic neural networks.

Written by
Varun Kumar

Varun Kumar is an experienced science and technology journalist interested in machines, AI, and space exploration. He received degree in computer science from Indraprastha University. To find out what his latest project is, feel free to directly email him at [email protected] 

View all articles
Leave a reply