A recycling facility receives a truckload of post-consumer plastic. The bales are labeled HDPE. Some of it is. Some of it is LDPE. Some contains PP contamination. A small fraction has flame retardants that will cause problems downstream. The labels say one thing; the material says another.
This is the reality of circular economy ambitions: the sorting challenge is harder than the rhetoric suggests.
Why recycling is harder than it looks
Polymer recycling works when material streams are clean and consistent. Virgin material comes with a data sheet, a supplier specification, and predictable properties. Recyclate comes with uncertainty.
The same resin code can cover materials with very different molecular weights, additive packages, and processing histories. A PET bottle and a PET tray have different crystallization behavior. An HDPE milk jug and an HDPE industrial drum container have different stress crack resistance. Labels and resin codes do not capture these differences.
Contamination makes it worse. A small percentage of PVC in a PET stream will degrade the entire batch. Brominated flame retardants in WEEE plastics create compliance issues. Cross-contamination between polymer families causes processing failures and poor mechanical properties.
The result: conservative acceptance limits, repeated testing, and recycled content targets that stay out of reach.
What certificates and labels miss
Supplier certificates of analysis for recyclate provide part of the picture: MFR range, density, maybe ash content. These scalar values help, but they do not capture the full material fingerprint.
Two batches with identical MFR can behave differently on the extruder. One might have more low-molecular-weight tails. One might contain trace contamination that affects color stability. One might have residual additives from its previous life that interact unpredictably with your stabilizer package.
The information exists in the material itself. The question is whether you are measuring it and whether those measurements are structured for decision-making.
Spectroscopic methods for sorting
Three techniques dominate polymer identification and sorting:
NIR (Near-Infrared) spectroscopy works at production line speeds. It is the workhorse of automated sorting systems. NIR identifies polymer families well and can detect some contaminants. Limitations: struggles with black plastics (carbon black absorbs NIR), limited discrimination within polymer families, and surface-sensitive.
FTIR (Fourier Transform Infrared) provides more detailed chemical information. It can distinguish polymer grades, identify additives, and detect degradation markers. More time-consuming than NIR, but higher information content. Often used for incoming QC rather than high-speed sorting.
Raman spectroscopy works through packaging, identifies some additives FTIR misses, and handles black plastics better than NIR. Fluorescence interference can be a problem with some materials.
Each technique has strengths. The practical question is: how do you turn spectra into accept/reject/blend decisions?
Application: plastic recyclate screening
Incoming recyclate QC is where sorting decisions get real. A compounder receives a shipment. The question is not just "is this HDPE?" but "is this batch consistent with what we have successfully processed before?"
The approach:
- Fingerprint incoming batches with FTIR (and optionally NIR, density, MFR)
- Compare to a reference library of previously processed materials with known outcomes
- Flag anomalies before they reach the production line
- Document the decision rationale for traceability
This is not about replacing existing tests. It is about adding a rapid screening layer that catches problems upstream. A 30-second FTIR scan costs far less than a failed extrusion run.
One practical example: a compounder built a library of FTIR spectra from recyclate batches that processed well vs. those that caused problems (gels, color drift, poor mechanicals). New incoming batches get scanned and compared. Batches that fall outside the "good processing" envelope get flagged for additional testing or blending adjustments.
Application: contamination detection
Contamination detection is critical and often overlooked until something goes wrong.
Cross-polymer contamination: Small amounts of PVC in PET cause degradation. PP in PE can affect crystallization. These contaminants may be below visual detection thresholds but spectroscopically visible.
Additive contamination: Flame retardants carry through recycling streams. Some are regulated (brominated compounds, certain phthalates). Spectroscopy can flag batches with suspicious peaks for further analysis.
Black plastic identification: Carbon black-pigmented plastics defeat NIR sorting. Raman and mid-IR techniques can still identify the base polymer. This matters because black plastics are a significant fraction of post-consumer streams, especially from automotive and electronics.
The key is building reference libraries that include contaminants, not just clean materials. A system that only recognizes pure HDPE will not help when you need to detect 2% PVC contamination.
Application: textile sorting
Textile recycling is having a moment. EU regulations are pushing for separate collection and recycling targets. The challenge: textiles are rarely single-fiber.
A garment labeled "100% cotton" might contain polyester stitching. A "polyester" fleece might be a blend. Performance fabrics combine multiple fiber types intentionally. And the labels? Often missing, faded, or inaccurate.
Manual sorting by trained workers is slow and inconsistent. Automated spectroscopic sorting is emerging as the scalable solution.
NIR can distinguish major fiber families (cellulosic, synthetic, protein). Identifying specific fiber types within families (cotton vs. viscose vs. lyocell within cellulosics) requires higher-resolution techniques or combined approaches.
The textile sorting problem is fundamentally a classification problem: take a spectrum, assign it to a category, route the material accordingly. The technical capability exists. The challenge is building robust classification models that handle the variability of real-world textile waste streams.
Building a material fingerprint library
Effective sorting requires reference data. A classifier is only as good as its training set.
What makes a useful fingerprint library:
- Representative samples of materials you actually encounter (not just pure lab standards)
- Known outcomes linked to spectra (did this batch process well? did it meet specs?)
- Context preserved (instrument settings, sample preparation, conditions)
- Continuous updating as new materials and contaminants appear
The library becomes an asset. Every batch you characterize adds to your ability to screen future batches. Historical outcomes inform current decisions.
From waste stream to value stream
When sorting and screening work:
- Higher recycled content becomes achievable with controlled risk
- Fewer processing failures from undetected contamination or variability
- Faster accept/reject decisions at intake
- Documented rationale for compliance and customer requirements
The circular economy will not run on good intentions. It needs data infrastructure that handles the messiness of real-world recycling streams.
How PolyCore helps
PolyCore structures spectral data alongside QC results, processing outcomes, and batch metadata. The platform builds fingerprint libraries from your actual materials, not generic databases. Classification models learn from your outcomes, and decisions link back to traceable evidence.
Whether you are screening incoming recyclate, detecting contamination, or building capability for textile sorting, the foundation is the same: structured data that turns spectra into decisions.
Interested in exploring this?
If material variability or contamination is limiting your recycled content targets, or if you are looking at textile recycling infrastructure, we can design a pilot that uses your actual material streams to build practical sorting and screening workflows.