Neural network approach for faster optical properties predictions for different PCF designs

What is it about?

Photonic Crystal Fibres (PCFs) are emerging as an alternative to standard fibres for applications in many disciplines like fibre lasers & amplifiers, imaging, spectroscopy and telecommunications. They have superior light guiding properties compared to ordinary Optical Fibres (OFs). This paper illustrates the potential of neural networks to efficiently and accurately compute the optical properties of PCFs including solid-core, hollow-core and multi-core designs. The proposed method takes a range of design parameters and wavelengths as input to predict PCF optical properties like effective index, effective mode area, confinement loss and dispersion desired for optimal specifications. The neural network approach is significantly better in terms of the low computational runtimes (~5 milli-sec) required for predicting the properties against the longer runtimes (~18 sec) required for similar calculations by traditional numerical methods.

Featured Image

Photo by JJ Ying on Unsplash

Photo by JJ Ying on Unsplash