Deep Learning Approach for Predicting Optical Properties of Chalcogenide Planar Waveguide

Abstract

In this paper, deep learning-based Fully Connected Feedforward Neural Network (FFNN) model is applied to predict and analyze the optical properties comprising mode effective area, nonlinearity, effective index, and dispersion for a Ge <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">11.5</inf> As <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">24</inf> Se <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">64.5</inf> chalcogenide (ChG) rib planar optical waveguide. This deep neural network algorithm gives exact predictions of optical phenomena mentioned above for common parameter settings of wavelength in the range 0.5 – 15 μm, waveguide core width of 1 – 8 μm and waveguide core thickness of 0.5 – 2.5 μm. The computational time required with deep neural network (for training) and finite-element method (FEM) solutions is also compared. This simple and fast-training FFNN, the deep learning approach employed here, predict the output for unfamiliar parameter setting of the optical waveguide faster than traditional numerical simulation techniques. This FFNN is further compared with state-of-the-art Random Forest (RF) algorithm and it is found that RF performs comparably to the FFNN.