Optical Sorting and Traceability: Making Recycled Polycarbonate Credible

Polycarbonate (PC) recovered from post-industrial and post-consumer streams occupies a narrow technical position: it is an engineering thermoplastic with high intrinsic toughness, but the recycled form is often priced as a lower-tier feedstock because batch-to-batch variation in melt flow rate, residual coating fragments, and non-PC contamination undermines confidence. The phrase “Optical Sorting and Traceability: Making Recycled Polycarbonate Credible” describes a two-part control strategy. Optical sorting supplies the physical separation required to reach a defined polymer purity. Traceability supplies the chain-of-custody and process-data record required to link that purity to final mechanical and rheological performance. Without both layers, recycled PC cannot be meaningfully qualified against ISO 1133-1:2022 melt flow measurements or ISO 527-2:2012 tensile tests, because the input stream itself is not stable enough to serve as a reference.

What Distinguishes Credible Recycled PC from Downcycled Feedstock?

To assign credibility to recycled PC, three variables must be controlled simultaneously: chemical identity of the principal phase, molecular weight retention, and additive or contaminant load. The first variable is measured by spectroscopic or thermal identification; the second by melt flow rate and, where warranted, size-exclusion chromatography; the third by digestion or combustion methods supplemented by sorting records. A compounder accepting baled PC sheet can receive material containing 7–12 wt% PMMA from extruded lighting covers, which is nearly transparent to density-based separation because the specific gravity ranges overlap substantially. PMMA degrades under PC processing conditions, generating methyl methacrylate volatiles and surface defects, and reduces elongation at break. Published data for a blend containing 10 wt% PMMA in PC indicate elongation at break can fall below 20% when measured according to ISO 527-2:2012; unmodified PC typically exceeds 80%. Such values are not linear in contaminant concentration and therefore require sorting before compounding, not after.

The molecular weight retention question is equally sensitive. Polycarbonate hydrolyses at elevated humidity and temperature. A recycled feedstock with moisture content beyond 0.02 wt% will show a measurable drop in molar mass during extrusion. The processing window in a co-rotating twin-screw extruder with L/D 40:1 and barrel temperatures between 280°C and 310°C is sufficiently narrow that pre-drying at 120°C for 4 h to a dew point of -40°C is mandatory. Recycled lots that cannot provide traceability records for drying and handling carry an unacceptable probability of viscosity breaks. A common gate check is melt volume-flow rate at 300°C / 1.2 kg per ISO 1133-1:2022; values higher than 25 cm³/10 min in a resin originally specified at 10 cm³/10 min indicate chain scission rather than benign flow improvement.

In a European post-industrial recycling line handling 12 t/h of shredded automotive glazing and solid polycarbonate sheet, two sequential optical units have been configured: a visible/NIR camera system and a dual-ejector chute sorter with 5 mm nozzle pitch. The first unit removes black PMMA and rubber particles; the second rejects pieces that still contain 4 wt% or more PC/ABS as detected by short-wave infrared absorption at 1,650 nm to 1,750 nm. The sorting threshold is set at 0.5 wt% PC/ABS in the accepted stream for optical clarity applications. Ejector latency in the second unit is 2.4 ms, and at belt speeds above 3.5 m/s valve overlap produces carryover losses of PC flakes. A yield audit on this configuration records 82–88% PC recovery at 99.2% target-polymer purity when feedstock comes from clear monolithic sheet. When the same line receives coated headlamp lenses, recovery drops to 74–79% because polyurethane hardcoat and silicone coatings mask the underlying PC infrared fingerprint. This field data illustrates the central limitation: optical sorting is a concentration-dependent, surface-sensitive unit operation. Traceability must therefore capture not only the accepted fraction but also the rejected fractions, because rejected material is often re-blended in secondary markets and later re-enters the supply chain as unlabelled regrind.

Sensor Configurations, Ejector Latency, and Polymer Identification Accuracy

Across the installed base of polycarbonate sorting lines, optical sorting is not a single technology but a stack of detectors whose resolution limits are defined by wavelength range, spot size, and ejection hardware. NIR hyperspectral imaging in the 1,000–2,500 nm region discriminates PC from PET, PMMA, and ABS via C–H, C=O, and aromatic overtone bands. However, black polycarbonate absorbs strongly in the visible and near-infrared; classification accuracy drops below 85% when carbon black pigment loading exceeds 0.1 wt%. Raman systems provide molecular structure from inelastic scattering and can identify black grades more reliably, but integration times of 10–50 ms per object constrain throughput to 1–2 t/h per sensor bar unless multiple channels are installed. X-ray fluorescence detects bromine and antimony in flame-retardant PC/ABS and in PCB-contaminated flake, but it cannot detect PMMA or low-tin silicone coatings. LIBS is used for rapid elemental screening but is less common in high-volume PC sorting due to spot size and surface roughness effects. Laser-induced fluorescence from deliberately added tracer molecules offers an alternative when the tracer has a quantum yield above 0.1 and an emission wavelength that does not overlap with polymer photo-oxidation fluorescence.

Because the sorter is surface-sensitive, flake thickness and surface roughness alter apparent spectral intensity. A PC flake with an acrylic film of 0.1 mm will read as PMMA even though the core remains PC. This is why the first sorting pass must be preceded by granulation to a defined particle size, typically 8–16 mm, and by surface cleaning using a friction washer. The accepted fraction is then sampled at intervals of 1 t or less. Without that sampling interval, the optical sorter’s frame-level purity has no relationship to the extruder’s melt quality.

If Fluorescent Tracers Are Added to Polycarbonate Resin, How Must Sorting Calibration Change?

In tracer-based sorting, the detection target shifts from intrinsic molecular absorption to an intentionally encoded signal. A PC resin producer can incorporate 50–200 ppm of a thermally stable fluorescent tracer that survives extrusion at 310°C and is excited by a 365 nm or 395 nm LED array. The sorter then uses a bandpass filter in the emission region, typically 450–500 nm. Calibration requires a negative control of untraced PC and a positive control of 50 ppm traced PC. The main conflict is signal quenching by pigments, particularly carbon black, and by UV absorbers such as benzotriazole at loadings above 1 wt%. If a recycled lot contains mixed traced and untraced PC, the tracer intensity correlates with the traced fraction only when flake thickness is controlled between 2 mm and 4 mm; thickness variation outside this range changes path length and apparent tracer concentration. This is why tracer-based optical sorting is best deployed in controlled post-industrial loops rather than in mixed post-consumer bales. The chemical tracer itself must not migrate or alter the base polymer’s impact performance. A tracer added at 100 ppm is unlikely to shift Charpy notched impact strength by more than ±2%, but no claim is valid without testing per ISO 179-1:2010 on the specific compound because matrix interactions differ.

Traceability Records Provide the Missing Link Between Optical Purity and Mechanical Property Retention

Because optical sorting is a surface-sensitive unit operation, the accepted fraction must be linked to melt-phase property data. A credible recycled PC lot requires records that connect the input bale, the sorting run, the percentage reject rate, the contaminant thresholds, the granulation temperature, and the final melt flow rate. The relevant chain-of-custody framework is ISO 22095:2020, which defines differentiated and mass balance approaches. For recycled polycarbonate intended for engineering applications, mass balance is operationally superior to physical segregation because PC flakes from multiple optical-sort passes are often homogenized in a silo before compounding. Under mass balance, the claim “contains 30% recycled PC” requires documentation that the certified input mass was not exceeded by the output mass over a defined period, not that each pellet contains exactly 30% recycled material. This differs from identity-preserved systems, which are technically feasible only in closed-loop post-industrial streams such as compact disc scrap or water bottle reject.

Traceability methods include alphanumeric lot codes, barcoded super-sacks, RFID-tagged pallets, and digital watermarks printed on packaging. For polycarbonate, the more difficult challenge is linking optical-sorter ejector data to lot-level quality. A typical sorter logs frame-level classification, pneumatically ejected counts, and belt speed every 5 ms. These data can be aggregated to estimate the accepted fraction’s contaminant distribution. However, this estimate is not a chemical certificate. It must be validated by solvent extraction followed by Fourier-transform infrared spectroscopy on composite samples taken at intervals of 1 t or less. The sampling frequency is the limiting factor for traceability credibility; a 1 t grab sample from a 25 t lot carries an uncertainty that exceeds the typical optical sorting resolution unless the sample is ground to 0.5 mm and divided by a rotary riffle splitter following ISO 565:1990 sieve principles.

Compounding of optically sorted recycled PC into a 30% glass-filled grade requires melt filtration through a 200 µm screen pack and addition of a phosphite antioxidant at 0.2 wt%. On a 70 mm twin-screw co-rotating extruder at screw speed 450 rpm and melt temperature 305°C, a shift in melt pressure from 48 bar to 61 bar over a 4 h run indicates screen blockage from residual silicone coatings or crosslinked elastomer particles that optical sorting did not reject. This is a production-scale bottleneck observed on equipment processing mixed clear and opaque polycarbonate sheet. The sorting system must therefore be paired with a pressure-rise monitoring protocol; an acceptance rule that rejects the extruder batch if melt pressure increases by more than 25% over baseline prevents strand breakage and die freeze-off. In such lines, traceability records are extended to include differential pressure values, screen pack mesh size, and granulator knife gap, because the final pellet certificate needs to account for process-generated particulates as well as input contaminants.

Use in food-contact applications is governed by FDA 21 CFR 177.1580 or European Regulation EU No 10/2011. Recycled PC used in such applications requires decontamination validation, which optical sorting alone cannot provide. The sorter may deliver 99% PC purity, but migration of residues from sorted contaminants or degraded surface layers may still exceed the overall migration limit of 10 mg/dm². Therefore, food-contact recycled polycarbonate requires additional super-clean processes, such as high-vacuum devolatilization at 280–300°C and surface renewal via twin-screw elements with 0.5 mm clearances, plus migration testing per EU No 10/2011. Published data for combined optical sorting and vacuum devolatilization in this specific application configuration remain limited, so lot-by-lot functional barrier verification is necessary.