Shining new light on plastic traceability: Fluorescence fingerprint analysis

Daisuke Yagi

Research & Development Group
Hitachi, Ltd.

The need for traceability in plastic resource circulation

In the global push towards a circular economy, improving the quality, traceability, and availability of recycled plastics is becoming vital in industry and society. Traceability systems that indicate "how much recycled material is used" play a key role in delivering accurate product information and provide value for both manufacturers and consumers. By offering clear and transparent data on the amount of recycled content, manufacturers demonstrate a concrete commitment to environmentally responsible manufacturing, as well as meeting regulatory requirements. This, in turn, builds customer trust and enhances the competitiveness of their products. For consumers, transparent information such as "X% recycled material used" empowers them to make environmentally conscious informed choices when selecting products.

Achieving effective traceability is, however, not without its challenges. Conventional analytical methods, such as chemical composition analysis, often rely on destructive testing or are time intensive. This makes it difficult to accurately assess recycled content during the manufacturing process, posing a significant barrier to implementing practical traceability systems.

Fluorescence fingerprint analysis: A non-destructive approach to traceability

To overcome this issue, we focused our attention on fluorescence fingerprint analysis which is a non-destructive spectroscopic technique[1]. We felt that this could be a promising solution because it does not require destroying the sample, and fluorescence fingerprint analysis can be seamlessly integrated into manufacturing lines. In addition, the advantage of this technique is its ability to analyze some black plastics, which conventional visible-light-based spectroscopic techniques find challenging due to the high light absorption properties of black pigments. This capability is particularly valuable because recycled plastics often come from mixing various colored waste, which tends to result in darker shades such as black. As a result, the ability to accurately analyze some black plastics significantly enhances traceability in recycled materials.

So, how does it work? Let me first introduce the principles and data analysis methods behind this technology. Firstly, plastics often contain additives such as antioxidants, which are often fluorescent substances. When these fluorescent substances are exposed to specific wavelengths of light, they exhibit unique fluorescence wavelengths and intensities. We can use this characteristic to create an Excitation-Emission Matrix (EEM) (Fig.1) that maps the fluorescence intensity across various excitation and emission wavelengths by exposing the plastics to these different wavelengths. The EEM can then be used as a "digital fingerprint" that indicates the different characteristics, additives, and concentrations within the plastics, thus enabling the non-destructive identification.

Advancing traceability with AI-driven fluorescence analysis

Using this characteristic of EEM, we developed a technique to predict the recycled material content by analyzing the EEM of molded products made from a mixture of virgin and recycled materials. I would like to introduce a use-case on polypropylene (PP), a plastic widely used in manufacturing.

Test samples were molded with varying ratios of virgin PP (vPP) and recycled PP (rPP). Using the Hitachi High-Tech Fluorescence Spectrophotometer F-7100 [2], we captured the Excitation-Emission Matrix (EEM) for each sample, allowing us to identify fluorescence spectra directly correlated to rPP content (Fig. 2).

Recognizing the non-linear relationship between fluorescence intensity and the concentration of fluorescent substances, we applied a Neural Network (NN) to develop a prediction model. This AI-driven approach achieved high predictive accuracy, with an R² value of 0.96 (Fig. 3). In Fig. 3, the data points closely align with the diagonal line, indicating that the predicted values closely match the actual measurements. This alignment underscores the model’s high predictive accuracy for recycled material content.

This research was presented at the Seikei-kakou Autumnal Meeting of the Japan Society of Polymer Processing held in Okinawa, Japan, in November 2024, and recognized with a Poster Presentation Award [3].

Future directions: Scaling the technology for a sustainable future

In our research, we analyzed the EEM of molded products made from a mixture of virgin and recycled materials to develop a technique for non-destructively estimating recycled material content. The technique can be integrated into manufacturing processes, where molded products are non-destructively measured to determine their recycled material content, and the content data is then added as traceability information (Fig.4). This approach enables reliable management of recycled material use in manufacturing.

Moving forward, our aim is to extend this technique to a broader range of plastics to enable more comprehensive traceability across a wider range of recycled plastic products. Furthermore, by refining and adapting the technology for practical use in manufacturing environments, we hope to bridge the gap between environmental sustainability goals and practical manufacturing. By enhancing trust through transparent labeling of recycled content, we aim to empower consumers to make more informed and responsible choices. When products clearly display how much recycled material they contain, consumers can better factor environmental considerations into their purchasing decisions.