Raman Spectroscopy and Machine Learning Show Promise for PFAS Detection

In a recent study, Amit Kumar, a researcher at the University of Georgia in Athens, Georgia, and his team investigated a new method that combines Raman spectroscopy with machine learning (ML) algorithms to improve the detection of per- and polyfluoroalkyl substances (PFAS) in the environment. This study, which was published in the Journal of Hazardous Materials, explores the vibrational spectroscopy properties of nine PFAS compounds and how Raman spectroscopy and computational modeling can help differentiate them (1).

PFAS substances are ubiquitous in the environment. These chemicals are known for their properties that prevent them from degrading (2). They often accumulate in water, soil, and human tissue (2). This is a concern because PFAS have been linked to certain health illnesses, including immune dysfunction and several types of cancers (1). As a result, researchers have been exploring effective methods that can help detect PFAS in the environment so these substances can be removed. The most conventional approaches are good, but there are several issues with them, including the fact that they require complex sample preparation and expensive instrumentation (1).

Model of PFAS Structure Floating in Three-Dimensional Space Showcasing Molecular Interactions and Characteristics. Generated with AI. | Image Credit: © Anzhela - stock.adobe.com

In this study, the research team proposed that Raman spectroscopy can be an alternative for rapid, structure-based PFAS identification. Raman spectroscopy is a vibrational spectroscopy technique widely used in chemistry and materials science (1,3). Looking at nine PFAS compounds with varying functional groups and chain lengths, the research team collected the spectra of these compounds and then identified distinct vibrational peaks across low, medium, high, and ultra-high wavenumber regions (1). These spectral variations allowed the scientists to differentiate PFAS compounds based on structural features such as chain length and the presence of specific functional groups (1).

Then, the researchers employed a computational method, called density functional theory (DFT) calculations, to model the electronic structure of molecules. Using DFT allowed the team to associate vibrational modes to specific molecular motions and confirm the experimental Raman data (1). These tools ultimately helped researchers gain a better understanding of how PFAS structures dictate their spectral signatures.

Meanwhile, the team also used advanced data analysis technologies, including principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE), to classify and separate the Raman spectra. These unsupervised ML methods revealed both structural similarities and subtle differences between the PFAS compounds studied (1). The ability to cluster PFAS compounds based on spectral characteristics opens the door for more automated and data-driven environmental detection strategies.

Despite the positive results achieved, there were some limitations with this study. The research team acknowledged that some spectra displayed broad and weak peaks, which is a limitation they attributed to sample preparation methods and instrumentation (1). Moreover, PCA analysis did not consistently distinguish all PFAS compounds, pointing to the need for more rigorous experimental design protocols (1).

The method presented here can be used for other applications as well. A reliable Raman spectral database could serve as a foundation for multiple applications, including identifying thermal degradation products through in situ spectroscopy, supporting forensic analysis of contaminated sites, and advancing sensor technologies for water quality monitoring (1). By combining Raman spectroscopy with computational modeling and ML, this work sets the stage for next-generation detection platforms capable of tackling an important environmental challenge, one that is expected to remain ongoing over the next decade.

In conclusion, the findings from this study not only confirm theoretical predictions about PFAS vibrational properties. They also enrich the existing spectral databases critical for environmental forensics (1). Although challenges remain in achieving reproducibility and sensitivity, the study demonstrates how interdisciplinary methods can converge to address the PFAS crisis (1).

With further refinement and validation, the method presented by the research team proved to be an effective method for PFAS detection and differentiation, contributing toward the continued health of both ecosystems and people from these contaminants.

References

1. Kumar, A.; Rothstein, J. C.; Chen, Y.; et al. Experimental Raman Spectra Analysis of Selected PFAS Compounds: Comparison with DFT Predictions. J. Hazard. Mater. 2025, 494, 138704. DOI: 10.1016/j.jhazmat.2025.138704
2. Tackett, B. PFAS in Food Analysis. LCGC The Column 2025, 21 (2), 20–23. Available at: https://www.chromatographyonline.com/view/pfas-in-food-analysis (accessed 2025-08-20).
3. Horiba, What is Raman Spectroscopy? Horiba. Available at: https://www.horiba.com/usa/scientific/technologies/raman-imaging-and-spectroscopy/raman-spectroscopy/ (accessed 2025-08-20).