Ag@C decorated GaN nanoflower enabled super-stable, single molecule level SERS substrate integrated with machine learning for multiple analytes identification

Abstract

Surface-enhanced Raman scattering (SERS) is the most promising noninvasive detection method. However, the long-term stability and multiple analytes identification are still challenging for the SERS sensor, especially for Ag materials, and inferior spectral interpretation in multispectral. Here, we developed a super-stable, single molecule-level SERS sensor via one-step synthesizing Ag@ organic carbon nanodots nanoparticles on the nanoflower gallium nitride surface. The SERS sensor not only achieves single molecule detection with the enhancement factor of 2.483 × 109 but also provides excellent long-term stability in air up to 8 months. The super-stability mechanism of the SERS substrate was proposed and discussed. In addition, integrating with machine learning technology-the back-propagation neural algorithm networks classifiers, three similar analytes were correctly recognized with an accuracy rate of 92.64%. This work could explore to the practical applications in the fields of the environmental pollution, food safety, and biomedicine.