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

Slot waveguides provide high electric field amplitude and optical power in low-index materials that are not possible with conventional waveguides. This specific property of the slot waveguide provides interaction between active material and electric field, which led to many interesting applications, such as optical amplification, optical switching, and optical detection in integrated photonics. In the present work, we combine machine learning (ML) algorithms and finite element simulation to predict the power confinement ([Formula: see text]) and mode effective index ([Formula: see text]) of slot waveguides with respect to geometric parameters such as gap, slab width, and slab height. Three different ML techniques, such as artificial neural network (ANN), support vector regression (SVR), and random forest (RF), were tested to compute performance parameters for the slot waveguide. The RF method outperformed the other two with mean absolute error (MAE), root mean square error (RMSE), coefficient of determination ([Formula: see text]), and Nash-Sutcliffe efficiency (NSE) values corresponding [Formula: see text] and [Formula: see text] as 0.007, 0.054, 0.961, and 0.960, and 0.129, 0.185, 0.998, and 0.998, respectively. Thus, providing a useful ML methodology for efficient optimization of slot waveguide structures for future applications.