Machine learning assisted design of photonic crystal fiber-based plasmonic sensor

A plasmonic refractive index sensor based on Photonic Crystal Fiber, capable of detecting analytes within a refractive index range of 1.33–1.365, is proposed in this work. This design features a simple structure with external sensing, utilizing a circular gold layer coating as a plasmonic material over the PCF. The design underwent numerical exploration employing the finite element method. However, to overcome FEM’s long simulation times and trial-and-error-based optimization process, machine learning techniques are employed to forecast the sensor’s operational characteristics. The FEM generated dataset was utilized for training and validation of the machine learning model which then provided a low mean bias error of less than 2% and was substantially 82.88 times faster than those obtained through simulations using FEM. The optimized design achieved a maximum amplitude sensitivity of 618 RIU^− 1, a maximum wavelength sensitivity of 20,000 nm/RIU and a high figure of merit of 386.14 RIU^− 1. The sensor’s wavelength resolution was determined to be 5 × 10^− 6 RIU. The proposed sensor can demonstrate remarkable sensitivity in detecting various compounds such as different types of milk (camel, cow, and buffalo), carbonated beverages, polymers, salinity at different levels, methanol, and various other bio-originated analytes. Due to the enhanced sensing performance and utilization of the machine learning approach, the proposed sensor can serve as an accurate and efficient alternative to existing sensors, while its design approach can reduce the need for resource-intensive and laborious simulations.