Deep learning assisted SERS for detection of nicotine and etomidate in E-cigarettes

With the rise of electronic cigarettes ( E -cigarettes), certain illegal additives have gradually emerged in these products, posing significant health risks to humans. Consequently, there is an urgent need for rapid and highly sensitive detection of illegal additives in E -cigarettes. This study combines deep learning algorithms with Surface-Enhanced Raman Spectroscopy (SERS) to achieve the sequential detection following phase separation of the illegal additive nicotine and the new psychoactive substance etomidate. First, by utilizing prepared gold nanoparticles as an enhancement substrate, we accomplished highly sensitive detection of nicotine and etomidate standards. Additionally, through density functional theory calculations, we established the attribution of the characteristic peaks of both substances, providing a basis for their qualitative differentiation. Additionally, by developing a liquid-liquid extraction method, we successfully isolated nicotine and etomidate from e-cigarettes, achieving highly sensitive SERS detection with lowest detectable concentrations of 0.18 ng/mL and 1 μg/mL, respectively. Finally, by integrating one-dimensional convolutional neural network (1D-CNN), we attained a classification accuracy of 94.0% for real samples. Therefore, this strategy may provide a reference for the detection of illegal substances in the future.