Flexible hydrogel fiber-coupled SERS sensor for remote in situ food inspection

Technical Summary: Flexible Hydrogel Fiber-Coupled SERS Sensor for Remote In Situ Food Inspection

Problem Statement
Surface-enhanced Raman scattering (SERS) spectroscopy is a powerful tool for food safety monitoring due to its high sensitivity and non-destructive nature. However, conventional optical fiber SERS sensors face a critical limitation: they are typically constructed from rigid silica, which prevents conformal contact with irregular or curved food surfaces (such as fruit). This rigidity restricts their application primarily to liquid samples or requires complex sample pretreatment (e.g., juice extraction). Conversely, existing flexible SERS substrates can achieve conformal contact but generally lack remote signal transmission capabilities, necessitating offline transfer to a spectrometer. There is a distinct need for a sensor that integrates conformal surface sampling, remote signal readout, and high sensitivity into a single device for rapid, in situ detection of pesticide residues.

Methodology
The study proposes a flexible hydrogel fiber-coupled optical fiber (HF-COF) SERS sensor designed for remote in situ detection. The core methodology involves the following steps:

1. Sensor Fabrication: A flexible hydrogel fiber is assembled at the tip of a silica optical fiber using an in situ crosslinking method. Sodium alginate (SA) is mixed with silver nanoparticles (AgNPs) to form a precursor solution. The optical fiber tip is dipped into this precursor and then immersed in a calcium chloride (CaCl₂) solution. The Ca²⁺ ions crosslink with the SA to form a stable, porous calcium alginate hydrogel fiber.
2. Structural Design: The resulting hydrogel fiber (approximately 1 cm in length) maintains efficient optical coupling with the fiber core while providing a flexible, porous 3D network. This structure facilitates the dense loading of AgNPs and the enrichment of analyte molecules.
3. Optimization: Parameters such as hydrogel fiber length, SA content, and Ca²⁺ crosslinking time were optimized to balance signal transmission, mechanical stability, and SERS enhancement. The optimal configuration was determined to be a 15 cm optical fiber with a 1 cm hydrogel fiber, 1.2 g of SA, and a 15-second crosslinking time.
4. Detection Protocol: For practical application, a "spiking" method was used on apple surfaces. Ethanol solutions containing varying concentrations of thiram (a model fungicide) were applied to the apple, dried, and then re-extracted with ethanol. The HF-COF sensor was placed directly on the apple peel to perform conformal sampling. The Raman signal was transmitted remotely through the optical fiber to a spectrometer.
5. Intelligent Analysis: A one-dimensional convolutional neural network (1D-CNN) was employed to process the full-spectrum Raman data, establishing an intelligent recognition model for thiram concentration classification without relying solely on manual feature peak selection.

Key Results

• Performance Metrics: The optimized HF-COF sensor demonstrated a SERS enhancement factor (EF) of 2.65 × 10⁹ for Rhodamine 6G (R6G) and a limit of detection (LOD) of 8.63 × 10⁻¹² M.
• Stability and Reproducibility: The sensor exhibited excellent reproducibility with a relative standard deviation (RSD) of 5.41% across different batches. Stability tests showed that the sensor retained 88.27% of its initial signal intensity after 40 days of storage, attributed to the protective encapsulation of AgNPs within the hydrogel matrix.
• Thermal and Optical Characterization: Simulation and thermal imaging confirmed that the porous hydrogel structure promotes multidirectional light scattering. This scattering spreads the optical field energy across the entire fiber cross-section, effectively exciting "hotspots" between randomly distributed AgNPs and increasing the light-matter interaction path.
• Application to Food Safety: In the detection of thiram on apple surfaces, the sensor achieved an LOD of 1.6 × 10⁻³ mg/L. The 1D-CNN model successfully classified thiram concentrations with an accuracy of 98.04%. Grad-CAM analysis confirmed that the model focused on the same key spectral regions (around 1379 cm⁻¹) identified by manual analysis, while also integrating subtle variations across the full spectrum.

Significance and Claims
The authors claim that this work provides a new strategy for developing SERS sensors that overcome the trade-off between surface adaptability and remote signal readout. By integrating a flexible hydrogel fiber with an optical fiber, the sensor enables direct, conformal contact with curved food surfaces while maintaining the ability to transmit signals remotely.

The study highlights that the HF-COF sensor does not necessarily pursue the absolute lowest limit of detection compared to all existing methods but achieves a superior balance of performance metrics: conformal sampling, remote readout, high sensitivity, and good reproducibility. The integration of deep learning (1D-CNN) further enhances the robustness of the detection system against complex backgrounds and baseline drifts. The authors conclude that this approach offers a practical, efficient, and intelligent solution for the rapid on-site screening of pesticide residues, significantly improving the convenience and reliability of food safety inspections compared to traditional rigid fiber probes or offline flexible substrates.