Parametric optimization of the slot waveguide characteristics using a machine-learning approach.

Area of Science:

• Integrated photonics
• Computational electromagnetics
• Materials science

Background:

• Slot waveguides enable high electric field confinement in low-index materials, crucial for integrated photonic devices.
• Their unique properties facilitate interactions with active materials, enabling applications like optical amplification and switching.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting the power confinement and mode effective index of slot waveguides.
• To compare the performance of Artificial Neural Network (ANN), Support Vector Regression (SVR), and Random Forest (RF) algorithms for this prediction task.

Main Methods:

• Finite element simulations were used to generate training data for slot waveguide parameters.
• Three ML algorithms (ANN, SVR, RF) were trained and tested to predict power confinement and mode effective index.
• Performance was evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), R-squared, and Nash-Sutcliffe Efficiency (NSE).

Main Results:

• The Random Forest (RF) model demonstrated superior performance over ANN and SVR.
• RF achieved excellent accuracy with MAE of 0.007 and 0.129, RMSE of 0.054 and 0.185, R-squared of 0.961 and 0.998, and NSE of 0.960 and 0.998 for the respective parameters.
• The study validates ML as an efficient methodology for slot waveguide design optimization.

Conclusions:

• Machine learning, particularly the Random Forest algorithm, provides an effective and accurate method for predicting slot waveguide performance metrics.
• This ML-driven approach facilitates efficient optimization of geometric parameters for advanced integrated photonic applications.
• The developed methodology can accelerate the design and development of novel slot waveguide structures.