Identifying Microplastics Using ATR-FT-IR Spectroscopy and Raman Spectroscopy

Recently, a research team comprised of scientists from several Chinese institutions developed and tested a novel method to better characterize, identify, and classify microplastics. This study, which was published in Microchemical Journal, showcases the integration of two spectroscopic techniques, Raman and attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy(1). Combining them with advanced artificial intelligence (AI), the researchers show how their method can improve and address an important challenge in environmental monitoring, and that is to reliably detect and differentiate microplastics across a wide range of samples (1).

What are microplastics?

Microplastics are tiny fragments of plastic that are less than 5 millimeters in size (2,3). Because of their small size, microplastics are often difficult to detect. As a result, they have been able, through anthropogenic activities, to infiltrate ecosystems all over the globe (1–3). The result is that microplastics find their way into our food and water supply.

Another challenge is that microplastics resist natural degradation. Therefore, monitoring their spread and understanding their health impacts is a growing global priority. However, accurate classification has remained a hurdle because of the limitations of using single spectroscopic techniques alone (1).

What did the researchers do in their study?

In their study, the researchers focused on eight common microplastics made from different polymer materials. The research team utilized Raman spectroscopy, prized for its ability to detect molecular vibrations with minimal sample preparation, and ATR-FTIR, a powerful tool for identifying chemical bonds in solid and liquid samples (1).

The researchers then built comprehensive databases for both the IR and Raman spectra of the microplastics, carefully adjusting instrument parameters such as laser wavelength, laser power, and acquisition time to ensure robust data collection (1). With these data sets in hand, they trained a one-dimensional convolutional neural network (1D-CNN), which is a deep learning model particularly adept at recognizing patterns in sequential data such as spectra (1,4).

When tested individually, the models based on ATR-FTIR and Raman data achieved recognition accuracies of 73% and 75%, respectively (1). This result indicates that both ATR-FTIR and Raman spectroscopy cannot acquire the spectral information needed by themselves.

Therefore, the researchers developed a novel three-level data fusion strategy that integrates both spectral data sets. This approach included low-level fusion, which combines raw data; mid-level fusion, which merges extracted features; and high-level fusion, which integrates final classification decisions (1).

The results demonstrated that this new strategy increased recognition accuracy. Low-level fusion reached an accuracy of 88%, mid-level fusion achieved 97%, and high-level fusion delivered a near-perfect 99% classification rate (1).

To test how their method can be applied in the real world, the researchers spiked milk, cola, and tap water with microplastics and used these as the external validation sets. Even under these complex and variable conditions, the high-level fusion model-maintained recognition accuracies exceeding 98% across all sample types, demonstrating both robustness and generalization capability (1).

By combining the complementary strengths of ATR-FTIR and Raman spectroscopy within an AI-driven fusion framework, the researchers showcased an improved method for microplastic classification (1).

What are the next steps in this work?

The implications of this work extend well beyond laboratory analysis. Reliable and highly accurate identification methods are essential for large-scale environmental surveys, regulatory enforcement, and understanding the ecological and human health impacts of microplastics. Furthermore, the fusion approach could be adapted to other areas of materials analysis where combining multiple detection methods can reveal a more complete picture of complex samples.

As global concern over microplastic pollution continues to rise, this research provides a promising path toward faster, more reliable, and more comprehensive monitoring strategies. By leveraging the synergy of two spectroscopic techniques and the power of deep learning, the team at the Chinese Academy of Sciences has set a new benchmark in microplastic detection—one that could help address one of the most persistent environmental challenges of our time.

References

1. Li, H.; Xu, S.; Teng, J.; et al. Deep Learning Assisted ATR-FTIR and Raman Spectroscopy Fusion Technology for Microplastic Identification. Microchem. J. 2025, 212, 113224. DOI: 10.1016/j.microc.2025.113224
2. Workman, Jr., J. Tracking Microplastics Across Air, Water, and Soil: What Spectroscopy Reveals About Global Pollution. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/tracking-microplastics-across-air-water-and-soil-what-spectroscopy-reveals-about-global-pollution (accessed 2025-08-14).
3. Wetzel, W. Microplastics Found in Deepest Reaches of Central Indian Ocean. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/microplastics-found-in-deepest-reaches-of-central-indian-ocean (accessed 2025-08-14).
4. Kiranyaz, S.; Avci, O.; Abdeljaber, O.; et al. 1D Convolutional Neural Networks and Applications: A Survey. Mechan. Sys. Sig. Proc. 2021, 151, 107398. DOI: 10.1016/j.ymssp.2020.107398