Ultra-sensitive and photocatalytic SERS platforms for machine learning assisted multiplex antibiotic detection using...

Dispersion and accumulation of antibiotics in the environment have harmed the ecosystem and cause serious antibiotic resistance, which has a significant impact on human health. This study presents the construction of a highly sensitive and photocatalytic surface enhanced Raman spectroscopy (SERS) platform with the multiplex antibiotic detection capability by the machine learning technique. The developed SERS platform utilizes the Ag/NiFe₂O₄ microcomposite composed of highly active NiFe₂O₄ microroses with oxygen vacancies and closely adjacent ultra-thin petals on which photoreduced silver nanoparticles are densely distributed. As a binary transition metal oxide semiconductor, NiFe₂O₄ has the conduction band edge near the middle point between the Fermi level of silver and the lowest unoccupied molecular orbital (LUMO) of nitrofurantoin to strengthen photoinduced charge transfer for Raman signal enhancement. Its medium bandgap facilitates optical excitation to increase the amount of electrons accumulated in the LUMO of nitrofurantoin. The Ag/NiFe₂O₄ microcomposite owns superior detection performance, including an ultra-low limit of detection of 3.18 × 10^-12 M, a substantial enhancement factor of 2.08 × 10¹⁰, high uniformity, high reproducibility, and satisfactory storage stability. Its ultra-high sensitivity can be attributed to the action of numerous electromagnetic hotpots between Ag NPs, the effective charge transfer from the microcomposite to the analyte, and the synergistic action of electromagnetic and charge transfer mechanisms. Multiple antibiotic discrimination for nitrofurantoin, nitrofurazone, and furazolidone, is performed by the machine learning algorithms, including decision tree, support vector machine, adaptive boosting, k-nearest neighbors, and random forest, with the best accuracy of 99.42 %. The photocatalytic degradation mechanism of Ag/NiFe₂O₄ microcomposite is verified by the radical scavenger studies. The proposed SERS sensing platform provides an effective and promising strategy for the trace-level detection of multiple antibiotics in food and environmental samples.