Machine Learning Optimized Graphene and MXene-Based Surface Plasmon Resonance Biosensor Design for Cyanide Detection

Abstract

Cyanide, a highly toxic chemical compound, presents severe risks to both human health and the environment. Its presence is particularly concerning in various industrial sectors, including mining, electroplating and chemical manufacturing, as well as in natural water bodies due to industrial discharge. This study introduces a graphene-based metasurface sensor designed for highly sensitive cyanide detection within the terahertz frequency range. The sensor’s design was refined through comprehensive electromagnetic modelling and analysis. Performance characterization demonstrates optimal sensitivity of 929 GHz RIU⁻¹, coupled with a figure of merit of 14.286 RIU⁻¹ between 0.806 and 0.856 THz frequencies. The detection limit achieved is 0.053 RIU. Adjustments to graphene’s chemical potential and structural dimensions demonstrated the device’s adaptability. Additionally, the application of machine learning techniques, specifically 1D-CNN regression, proved effective in optimizing sensor performance. The predictive model demonstrated remarkable accuracy, with an optimal R² score exceeding 95%, indicating that over 94.9% of the variance in the data was accounted for. This high precision enables accurate estimation of absorption values for wavelengths between measured points, underscoring the model’s reliability in spectroscopic analysis. This work highlights a versatile platform for rapid, label-free cyanide detection, with significant potential for applications in environmental monitoring, industrial safety and public health protection.