Design & Analysis of Dual Peak-Double Plasmonic Layered PCF Sensors

Abstract

A number of researchers have established themselves in the area of PCF-based SPR sensors with innovative designs achieving exceptional sensing performance. The majority of designs, however, either show low sensitivity with greater loss or show high sensitivity with the disadvantage of higher confinement loss, which makes designing challenging. The designs may also get complex in order to attain high sensitivities, which is the most essential factor. In our thesis, we employ a novel plasmonic material configuration in which Silver (Ag) and Ga-doped ZnO (GZO) work together to detect analytes. The sensor's detecting capabilities are improved by the presence of two peaks with the same Refractive Index (RI). The first peak is caused by GZO, and the second peak is caused by Ag. In terms of y-polarization, a DPSS of 27,341.5 nm/RIU is found which is the highest number reported to date. The amplitude sensitivity (AS) of the sensor is 875.72 RIU-1, and the wavelength sensitivity (WS) is 27,360 nm/RIU. The sensor also exhibits a high amplitude resolution of 1.496×10-5 and a wavelength resolution of 6.032×10-6. A linearity of R2=0.9973 and a FOM of 243.4 RIU-1 can be seen from the sensor. After conducting a thorough fabrication tolerance analysis, we find that both the confinement loss and the resonant wavelength shift are unaffected by a ±10-tolerance limit. A wide range of RI up to 1.27 to 1.41 is investigated, which broadens its potential applicability to the detection of pharmaceuticals and other compounds. In addition, the sensing parameters in RIs that were not numerically analyzed have been sought through the use of machine learning regression techniques in this study. No relevant work was found which used regression algorithms to analyze the performance parameters were found to the best of our knowledge. As input, the RI of the analyte is given to the algorithm while the output target variables are the shift in the GZO and Ag peak resonant wavelengths in two consecutive RIs respectively. The accuracy of the Random Forest Regressor was determined to be 90.176%, whereas that of the K-Nearest Neighbors Regressor was 95.54%. This research presents a new approach to assessing sensor performance metrics, which not only facilitates quicker and more accurate predictions but also saves considerable amounts of time.