Neural Networks in Optimizing the Performance of the Elliptical-Plasmonic Sensor

Abstract

In this work, we report the capability of a PCF-SPR sensor with an elliptical core, which has high sensitivity, and it is explained using a machine learning approach. The sensor component consists of fused silica as the background material, TiO2 as the adhesive material between the dielectric material and the plasmonic material, and Au was chosen as plasmonic material with optimal thicknesses of 35 nm for TiO2and 45 nm for Au. Numerical results show that the sensor component has a high sensitivity of 24,000 nm/RIU for four modes that have consistent shifts, including x-polarized, x-odd, y-polarized, and y-odd. Meanwhile, AS maximums were found of -91.82 1/RIU for x-polarized, -91.88 1/RIU for y-polarized, -90.98 1/RIU for x-odd, and -89.276 1/RIU for y-odd respectively, on the refractive index of the analyte of 1,365 RIU. The ML algorithm was used to optimize the sensor parameters, and it was found that the algorithm had a very low MSE of 0.00083; this result is better than the previous report work. Doi: 10.28991/ESJ-2024-08-05-07 Full Text: PDF