Deep-Learning-Empowered Gold Nanoparticle Dimer-Assisted Hyperbolic Metamaterial Surface Plasmon Resonance Fiber-Optic Sensor for Ultrasensitive Mercury Ion Detection

Abstract Mercury ions (Hg2+) are highly toxic heavy-metal contaminants that pose serious environmental and health hazards, necessitating the development of highly sensitive and accurate detection methods. Herein, a fiber-optic surface plasmon resonance (SPR) sensor is developed for the ultrasensitive detection of Hg2+. The sensing probe is constructed on a hyperbolic metamaterial (HMM) consisting of three repeated Ag/ITO bilayers, decorated with thymine (T)-rich DNA-functionalized gold nanoparticles (AuNPs). Upon Hg2+ addition, specific T–Hg2+–T coordination captures a second population of DNA-functionalized AuNPs, forming AuNP dimers on the HMM surface. The nanometer interparticle gap of the AuNP dimer induces strong plasmonic coupling, which amplifies the local refractive index (RI) changes at the sensing interface and produces a pronounced SPR wavelength shift, enabling highly sensitive detection of Hg2+. Experimental results demonstrate that the proposed sensor exhibits a peak RI sensitivity of 8600 nm/RIU and achieves a limit of detection as low as 2.63 pM for Hg2+. Moreover, a DenseNet-based deep-learning model is employed to extract subtle spectral features from SPR spectra acquired at seven Hg2+ concentration levels, enabling quantitative concentration prediction with an R2 of 0.998505. Therefore, the proposed fiber-optic SPR sensor provides a promising strategy for sensitive and accurate Hg2+ detection.