Detection of Polystyrene Microplastics up to the Single Nanoparticle Limit Using SERS and Advanced ANN Design (KANformer)

Due to uncontrolled release, gradual accumulation, low degradation rate, and potential negative impact on human health, microplastics (MPs) pose a serious environmental and healthcare risk. Thus, the spread of MPs should be at least carefully monitored to identify and eliminate their main sources, as well as to provide a suitable alarm in the case of MP concentration increase. Among various detection methods, surface-enhanced Raman spectroscopy (SERS) poses a unique detection limit and the ability to perform outdoor measurements without preliminary sample treatment. However, the utilization of SERS for MPs detection is significantly limited for a few reasons. First, the maximal SERS enhancement occurs in the so-called hot spots, where the MPs cannot penetrate due to their size. In addition, the natural environment can produce a significant spectral background, which blocks the microplastic characteristic signal. To overcome these limitations, we propose a new alternative route for introduction of MPs into the plasmonic hot spots, using in situ MP annealing and an advanced artificial neural network (ANN) design, the Kolmogorov-Arnold transformer (KANformer, KANF). Polystyrene (PS) MPs were used as a model compound. We have also demonstrated the potential versatility of our approach using different microplastics, such as polyethylene, polypropylene, and polyethylene terephthalate. The proposed approach allows us to detect the presence of PS up to the single nanoparticle limit (in the mL of analyzed solution) with a probability of above 95%, even under mixing with groundwater model matrices.