ResNet18-SE assisted SPR/digital SERS dual-mode integrated fiber optic sensor for ultra-trace detection of mercury ions

Developed a deep learning-assisted dual-modal SPR/digital SERS optical fiber sensor. • Enabled highly selective measurement of mercury ions. • Adopted a silver (Ag)/Ti 3 C 2 T x MXene/gold nanoparticle (AuNP) hybrid composite structure. • Achieved Hg 2+ detection limits as low as 10 pM. A deep learning-assisted dual-mode optical fiber sensor integrating surface plasmon resonance (SPR) and digital surface-enhanced Raman scattering (SERS) is developed for ultra-trace Hg 2+ detection. The sensor employs a silver (Ag)/Ti 3 C 2 T x MXene/Gold nanoparticle (AuNP) hybrid interface, where synergistic LSPR–SPR coupling amplifies electromagnetic fields for sensitive sensing. Upon Hg 2+ exposure, thymine (T)-rich DNA immobilized on the sensing surface and 4-mercaptopyridine-functionalized gold nanoparticle (AuNP@4-MPY) can selectively bind Hg 2+ , forming stable T–Hg 2+ –4-MPY complexes. This process anchors AuNP@4-MPY onto the sensing surface to form AuNP dimers, inducing a distinct SPR redshift and a prominent digital SERS enhancement. The sensor achieves detection limits of 10 pM (SPR) and 1 pM (digital SERS). The 18-layer Residual Network-Squeeze-and-Excitation (ResNet18-SE) regression model is utilized in this study to precisely predict Hg 2+ concentrations, yielding coefficient of determination ( R 2 ) values of 0.9997 and 0.9991 for SPR and digital SERS modes, respectively. Spike-and-recovery experiments performed in tap water and river water produced recoveries in the range of 95.4%–106.4% for both detection modes. This work provides an effective strategy for ultra-trace Hg 2+ detection and offers new perspectives on dual-mode optical sensing. Download: Download high-res image (269KB) Download: Download full-size image