Food safety monitoring requires highly sensitive and low-loss detection techniques for identifying hazardous food additives. In this study, a pentagonal hollow-core photonic crystal fiber (HC-PCF) sensor is introduced to detect two widely used but potentially harmful food additives, Sorbitol, and Butyl Acetate, in the terahertz (THz) regime. The objective of this work is to develop a PCF structure that ensures strong light analyte interaction, ultra-low loss, and high detection accuracy. The sensor was numerically modeled using COMSOL Multiphysics 6.1 based on the Finite Element Method (FEM) to evaluate key optical parameters such as relative sensitivity (RS), confinement loss (CL), effective material loss (EML), numerical aperture (NA), effective area (EA), and spot size. In addition, a Random Forest Regressor (RFR) machine learning model was employed to predict sensor performance and validate the simulation results. The optimized design, operating at 2.4 THz, achieved maximum RS values of 95.98% for Butyl Acetate and 95.07% for Sorbitol, along with ultra-low CL of 1.18 × 10⁻¹³ dB/m and 3.84 × 10⁻¹⁴ dB/m, respectively. The corresponding EML values were 0.0073 cm⁻¹ and 0.0083 cm⁻¹. The RFR model yielded a high R² score of 0.9935, confirming its predictive reliability and consistency with FEM results. These results demonstrate the effectiveness of the proposed sensor for accurate and reliable food additive detection. A pentagonal hollow-core PCF sensor is designed for rapid detection of Sorbitol and Butyl Acetate at 2.4 THz. The sensor achieves high sensitivity (95.98%) and ultra-low confinement loss, ensuring precise and efficient light guidance. Machine learning with a Random Forest model enhances accuracy, enabling smart and reliable food safety monitoring. Cross sectional view of our proposed sensor • Innovative pentagonal hollow-core PCF sensor designed and simulated using FEM in COMSOL Multiphysics for detecting harmful food additives. • Achieves high relative sensitivity of 95.98% for Butyl Acetate and 95.07% for Sorbitol at an operating frequency of 2.4 THz. • Demonstrates ultra-low confinement losses of 1.18 × 10⁻¹³ dB/m and 3.84 × 10⁻¹⁴ dB/m, ensuring excellent light guidance and minimal signal degradation. • Machine learning integration with a Random Forest Regressor yields a high R² score of 0.9935, improving detection accuracy and reliability. • Provides a real-time, accurate, and intelligent sensing platform for next-generation food safety monitoring and chemical detection systems.
Ferdous et al. (Wed,) studied this question.