Machine Learning-Enhanced Surface Plasmon Resonance Sensor with D-Shaped Dual-Core Photonic Crystal Fiber Design

J Fluoresc. 2025 Jun 4. doi: 10.1007/s10895-025-04384-x. Online ahead of print.

Abstract

A D-shaped dual-core surface plasmon resonance (SPR) sensor based on photonic crystal fibers (PCFs) has been created, and its sensing capabilities were evaluated through the finite element method (FEM). The design features a square lattice arrangement of air holes, with two central holes removed to form a D-shaped dual-core structure. The sensor's performance was assessed using both wavelength and amplitude interrogation approaches. It achieved a maximum wavelength sensitivity of 16,000 nm/RIU for y-polarized light at an analyte refractive index (RI) of 1.38, along with a peak amplitude sensitivity of 765.21 RIU-1 at the same RI, and a wavelength resolution of 2.5 × 10-6 RIU. Furthermore, machine learning (ML) techniques, particularly artificial neural networks (ANN), were employed to predict confinement loss (CL) with high accuracy, without the need for the imaginary component of the effective RI. For an RI of 1.32, the ANN model achieved a mean squared error (MSE) of 3.5363 × 10-6, showcasing the model's reliability in forecasting sensor performance.

Keywords: Confinement loss; D-shaped SPR sensor; Machine learning; Photonic crystal fibers; Wavelength sensitivity.