Hybrid Ag/Au nanocavity arrays for ultrasensitive SERS discrimination of molecular isomers via machine learning

Surface-enhanced Raman scattering (SERS) has been one of the efficient strategies for trace molecular analysis, but the traditional substrates with sharp tips cannot effectively detecting the molecular isomers. Here, we develop a facile way to fabricate large-area, reproducible Ag/Au nanocavity arrays by the combination of microsphere-templated soft lithography and self-assembly. Compared to the flat and convex substrate, the Ag/Au units arranged on the concave surface exhibit a high-density of hotspots via spatial light confinement and multi-reflection-enhanced electromagnetic coupling, which enables highly sensitive detection of probe molecules down to10^-12 M. Moreover, by integrating machine learning algorithms, the test set shows a coefficient of determination (R²) of 0.9896, achieving advanced training performance for classifying four structurally analogous quinolone antibiotics (ciprofloxacin, enrofloxacin, danofloxacin, and enoxacin) with a quantitative prediction of 10^-10 M. This work demonstrates a robust pathway for the controllable fabrication of high-performance SERS substrates for the development of intelligent, quantitative, and on-site sensing platforms in complex practical scenarios.