Identification of Degraded Plastics Cross-checking IR Spectra using Machine Learning Classification and Library Search

February 2, 2024


The shape of the infrared (IR) spectra of plastics depends on the degree of degradation (Figure 1). For example, oxidative degradation generates bands attributed to OH and C=O stretching vibrations. If an attempt to identify such spectra is made using a general database based only on the degree of agreement of the spectral shape, it may provide erroneous qualitative results. However, since multiple factors such as ultraviolet radiation, heat, moisture, and mechanical stress can affect the degradation of plastics, it is not practical to create a database that covers all of these factors in a combined manner, as it would require a great deal of effort.

Focusing on the fact that key IR bands associated with plastics are not lost even when the plastics are degraded,*1 JASCO recommends two different approaches*2 to identify degraded plastics: Classification based on key bands and Search based on libraries.

*1: This is based on the idea that if the molecular structure has been completely destroyed so that the material can be no longer be

called a plastic, the key bands associated with the original plastic will not be detected.

*2: The ADSS-4X [Advanced Spectral Search] program performs both Classification and Search for sample spectra.

Fig. 1   Spectra of undegraded (blue) and degraded (red) polystyrene

Sample spectra can be analyzed by Classification in the [Advanced Spectral Search] program based on the key bands associated with plastics. Classification is based on machine-learning results for approximately 10,000 spectra, and it classifies sample spectra into 35 categories using key bands exhibited by common plastic components such as hydrocarbons, styrenes and polyesters. Specific key bands associated with molecular bonding for a substance are displayed in the spectrum in the classification results, allowing for visual spectral determination.

To further verify the results of the Classification, Search narrows down the type of plastic using the LIB-PLA-ADSS IR plastics library, which contains 153 spectra grouped into 39 categories, in the [Advanced Spectral Search] program.

This report presents the results of a qualitative analysis of degraded polystyrene using a combination of Classification and an IR plastics library Search in the [Advanced Spectral Search] program.


Microplastics, Degraded plastics, Machine-learning, Advanced Spectral Search, IR plastics library


Figure 2 shows the Classification results for a spectrum of degraded polystyrene. The scores and specific key bands in the spectrum indicate that styrenes are highly probable. Assuming that the sample is a microplastic, the styrenes are considered to be polystyrene (including polystyrene foam).

Fig. 2   Classification results

To further validate the Classification results, Search was performed. Styrofoam and polystyrene were the top two rankings in the Search results (Figure 3). By cross-checking using the Classification and Search results, it was determined that the substance was polystyrene.

The same analysis method was attempted for the spectra of 11 other types of degraded plastics, including polyethylene and polypropylene, and correct results were again obtained. The validation results also showed that “Euclidean distance after differentiation” is the best search method.*3

*3 Depending on the degree of plastics degradation, it may be necessary to select another search method.

Fig. 3   Search results



The above results for qualitative analysis of degraded plastics whose spectral shape has changed indicate the effectiveness of the [Advanced Spectral Search] program equipped with Classification and Search.

About the Author

Spectroscopy Group