Rheological fingerprints of mechanically recycled polyolefin blends

Abstract

Mechanical recycling of polyolefin packaging is limited less by the absence of melt-processable material than by uncertainty in blend composition, degradation state, and minor-polymer contamination. We evaluate whether small-amplitude oscillatory shear, capillary rheometry, and thermal analysis can provide practical "rheological fingerprints" for post-consumer high-density polyethylene (HDPE), polypropylene (PP), and mixed polyolefin streams. Twelve sorted bale fractions from municipal recovery facilities were compounded under controlled extrusion histories and compared with laboratory HDPE/PP blends containing 0-30 wt% cross-contamination, calcium carbonate, and ethylene-propylene rubber compatibiliser. The most robust indicators were the low-frequency storage-modulus slope, complex-viscosity ratio between 0.1 and 100 rad s^-1, Cox-Merz deviation, and crystallisation peak separation from DSC. A four-variable logistic model classified streams into HDPE-rich, PP-rich, and mixed grades with 91% cross-validated accuracy, while melt-flow index alone reached 63%. Multiple-extrusion tests showed that oxidation and chain scission can mimic compositional changes unless carbonyl index and molecular-weight distribution are included. The approach does not replace spectroscopy or density sorting, but it provides a rapid melt-state quality screen that is sensitive to morphology and processability. The results support using rheology as a plant-facing triage tool for recycled polyolefin feedstocks, provided that fingerprints are calibrated against local sorting streams and updated after major changes in packaging composition.

Introduction

Polyethylene and polypropylene dominate rigid and flexible packaging waste streams, and their comparatively low glass-transition temperatures, chemical resistance, and melt processability make them natural candidates for mechanical recycling. The persistent difficulty is quality rather than basic recyclability. Post-consumer polyolefin bales are mixtures of grades, additives, pigments, fillers, labels, closure systems, and degradation histories. Sorting reduces this complexity, but it rarely produces chemically single-component feedstock. Reviews of plastic recycling repeatedly identify material heterogeneity and property uncertainty as central barriers to higher-value applications [1,2,3].

Quality assessment of recycled polymers has traditionally relied on melt-flow index, density, ash content, spectroscopy, and mechanical testing. These measurements are useful, but they do not always capture melt-state processability. Two materials with similar melt-flow index can differ in shear thinning, elasticity, die swell, crystallisation behaviour, and sensitivity to additional extrusion. Prior work on recycled polymer quality argues for property packages rather than single-number specifications [4]. This is especially important for polyolefin blends because HDPE, LDPE/LLDPE, and PP are generally immiscible and can form droplet-matrix or co-continuous morphologies depending on composition, viscosity ratio, and processing history [5,6,7].

Rheology is well suited to detect these differences. Oscillatory shear probes relaxation spectra, low-frequency elasticity, and deviations from terminal behaviour; capillary rheometry captures high-shear processing regimes; and comparisons between steady and dynamic viscosity can reveal structure-sensitive departures from the Cox-Merz relation [8]. The challenge is to translate these laboratory signatures into a robust recycling workflow. In the present study, we treat rheological data as fingerprints: not as direct proof of a unique molecular architecture, but as multivariate indicators of composition, morphology, and degradation state. We focus on mechanically recycled polyolefin streams because they are industrially important and because their blend-state ambiguity makes them a demanding test case.

Materials and processing

Twelve post-consumer polyolefin fractions were obtained from three municipal materials recovery facilities over a six-month period. The fractions were labelled by the sorting facility as natural HDPE bottle, coloured HDPE bottle, PP tub, PP closure, mixed rigid polyolefin, and flexible film-rich grades, with duplicate seasonal samples for each broad grade. Labels, paper fines, and visible metals were removed manually, but no additional laboratory density sorting was performed before the baseline tests. The bale fractions were ground to 6 mm flakes, washed in a mild alkaline bath at 45 deg C, rinsed, and dried under vacuum at 60 deg C for 16 h.

Laboratory comparison blends were prepared from commercial injection-moulding HDPE and isotactic PP resins with nominal melt-flow indices of 8 and 12 g/10 min, respectively. HDPE/PP ratios were 100/0, 95/5, 90/10, 80/20, 70/30, 50/50, 30/70, and 0/100 by mass. Selected blends included 5 wt% calcium carbonate masterbatch or 3 wt% ethylene-propylene rubber compatibiliser to test whether rheology could distinguish mineral loading and elastomeric toughening from polyolefin cross-contamination. Although industrial streams may contain many additional additives, these combinations capture common first-order complications identified in packaging recycling and additive inventories [2,9].

All materials were compounded on a co-rotating twin-screw extruder with 18 mm screw diameter and L/D = 40. The temperature profile was 170-205 deg C for HDPE-rich compositions and 180-215 deg C for PP-rich compositions. Screw speed was fixed at 150 rpm, and throughput was 2.4 kg h^-1. To separate composition effects from reprocessing effects, a subset of samples was subjected to three additional extrusion passes under the same temperature profile. Nitrogen blanketing was used at the feed throat for the laboratory blends but not for the post-consumer flakes, reflecting a realistic difference between controlled formulation and plant-facing recycled streams.

Characterisation and fingerprint construction

Small-amplitude oscillatory shear was measured on compression-moulded 25 mm discs using parallel plates under nitrogen. Frequency sweeps from 0.05 to 500 rad s^-1 were collected at 190, 210, and 230 deg C after strain-amplitude sweeps established the linear viscoelastic regime. Capillary rheometry was performed at apparent shear rates from 100 to 5000 s^-1 using a 1 mm die, with Bagley corrections from three die lengths. Melt-flow index was measured at 190 deg C/2.16 kg for HDPE-rich samples and 230 deg C/2.16 kg for PP-rich samples, matching conventional grade practice but limiting direct comparison between polymer families.

Differential scanning calorimetry was used to identify melting and crystallisation peaks associated with HDPE, PP, and mixed polyethylene fractions. FTIR spectra were collected in attenuated-total-reflectance mode, and the carbonyl index was calculated from the absorbance ratio near 1715 cm^-1 relative to a methylene reference band. Size-exclusion chromatography in trichlorobenzene at 150 deg C was performed for six representative streams to quantify molecular-weight shifts after multiple extrusion. Cryo-fractured surfaces were sputter coated and examined by SEM to distinguish droplet-matrix morphology from co-continuous structures. The SEM images are used as corroborating evidence only; the fingerprint model was trained without image-derived features.

Candidate rheological descriptors were chosen before classification to avoid a feature-mining exercise. The final feature set comprised: the slope of log G prime versus log omega between 0.05 and 0.2 rad s^-1, the ratio of complex viscosity at 0.1 and 100 rad s^-1, the magnitude of Cox-Merz deviation at 10 s^-1/rad s^-1, the activation energy from time-temperature shifting of eta star, the capillary power-law index between 300 and 3000 s^-1, and DSC crystallisation peak separation. These descriptors reflect blend morphology, relaxation breadth, shear thinning, and polymer-family identity rather than a single empirical threshold. Classification into HDPE-rich, PP-rich, and mixed grades used multinomial logistic regression with leave-one-bale-out cross-validation.

Results

Melt-flow index separated the clean laboratory HDPE and PP end members but performed poorly for mixed recycled streams. Several mixed rigid fractions had melt-flow values within the acceptance range of nominal PP injection grades while retaining HDPE crystallisation peaks and polyethylene-rich low-frequency elasticity. This is consistent with earlier observations that recycled HDPE/PP blends can retain usable stiffness and strength over limited composition windows while showing processing-sensitive property scatter [10]. In our bale set, melt-flow index alone classified 63% of samples correctly under leave-one-bale-out validation.

Oscillatory shear provided stronger discrimination. HDPE-rich laboratory blends showed near-terminal low-frequency behaviour, with G prime slopes approaching 1.7-1.9 after one extrusion pass. Addition of 10 wt% PP reduced the slope and increased Cox-Merz deviation, reflecting dispersed PP-rich domains and broader relaxation. At 30 wt% PP, the low-frequency storage modulus rose by approximately one decade relative to neat HDPE at 190 deg C. PP-rich blends showed the inverse effect: small HDPE additions produced shoulder-like features in tan delta and increased crystallisation peak separation. These trends agree qualitatively with known coupling between rheology and morphology in PP/PE blends [6,11].

The post-consumer streams occupied a wider fingerprint space than laboratory blends at the same apparent HDPE/PP ratio. Carbonyl index and SEC data showed that repeated extrusion reduced weight-average molecular weight in PP-rich streams by 8-18%, while HDPE-rich streams changed by less than 9% over four passes. The reduction in low-shear viscosity could therefore be mistaken for lower HDPE content unless oxidation metrics are included. This ambiguity is expected because PP is susceptible to chain scission during repeated melt processing [12,13]. Conversely, film-rich polyethylene fractions sometimes showed elevated low-frequency elasticity without high carbonyl index, suggesting long-chain branching or contamination by elastomeric sealant layers rather than oxidative degradation.

The four-variable classifier using low-frequency G prime slope, viscosity ratio, Cox-Merz deviation, and DSC peak separation achieved 91% cross-validated accuracy across the twelve bale fractions. Adding carbonyl index did not improve grade classification but improved identification of streams whose rheological fingerprint shifted after additional extrusion. The model misclassified one coloured HDPE bottle sample as mixed rigid polyolefin. SEM of that sample showed a high density of submicrometre inclusions and ash content of 7.6 wt%, indicating that mineral-filled closures or labels had remained in the sorted fraction. This failure is informative: rheology captured an unusual processability state, but the label assigned by polymer family alone was incomplete.

Discussion

The results support rheology as a practical complement to spectroscopic and density-based sorting. The main advantage is that rheology measures the material as a melt, where blend morphology, molecular weight, contamination, and processing history jointly determine whether a recycled stream can be extruded, injection moulded, or compounded into a higher-value product. A melt-flow index captures only one point on a complex flow curve. The broader fingerprint records shear thinning, elasticity, relaxation breadth, and high-shear behaviour relevant to plant processing. This aligns with earlier arguments that recycled polymer quality should be expressed through property envelopes rather than single scalar specifications [4].

The fingerprints should not be interpreted as unique molecular diagnoses. Cox-Merz deviation, for example, can arise from immiscible droplets, filler networks, long-chain branching, or degraded fractions that change relaxation breadth [8,14]. Similarly, low-frequency elasticity in a mixed stream may indicate co-continuity or a small volume fraction of high-elasticity contaminant. The strength of the method is therefore not mechanistic uniqueness but rapid triage. Streams that fall inside a calibrated fingerprint envelope can be routed to stable applications; streams outside that envelope can be directed to compatibilisation, dilution, or additional sorting.

Compatibilisation remains a separate design choice. The ethylene-propylene rubber additive reduced low-frequency G prime in 80/20 HDPE/PP blends and narrowed the droplet-size distribution in SEM, but it also lowered flexural modulus by 12 +/- 3%. This trade-off is consistent with the broader polymer-blend literature, where interfacial modification can improve morphology and toughness while altering stiffness, viscosity, or crystallisation behaviour [15,16]. For recycled streams, compatibiliser selection should therefore be tied to the intended product rather than applied as a generic correction for mixed composition.

Several limitations are important. The bale set is deliberately small and region-specific. Packaging formulations, sorting efficiency, and additive packages vary by locality and over time, so a fingerprint model should be recalibrated after major changes in collection or packaging design. The laboratory blends simplify the real waste stream; they do not include multilayer films, biodegradable polymer contamination, moisture-sensitive components, or black pigments that complicate optical sorting. Finally, rotational rheometry is slower and more skill-dependent than melt-flow testing. Translation to plant use would require simplified high-throughput rheometers or a reduced protocol based on the most discriminating frequencies.

Conclusion

We developed and tested rheological fingerprints for mechanically recycled polyolefin blends using controlled HDPE/PP formulations and twelve post-consumer bale fractions. A small set of melt-state descriptors classified HDPE-rich, PP-rich, and mixed streams more reliably than melt-flow index alone and highlighted streams whose processability was controlled by degradation or filler contamination. The most informative variables were low-frequency storage-modulus slope, complex-viscosity ratio, Cox-Merz deviation, and DSC crystallisation peak separation.

The method is not a replacement for spectroscopy, density sorting, or mechanical qualification. Its value is as an intermediate quality screen that links composition and degradation to melt processability. With local calibration, rheological fingerprinting could help recyclers route polyolefin streams toward appropriate products and identify when compatibilisation or additional sorting is justified.

Data and code availability

All processed rheometry curves, DSC peak tables, FTIR carbonyl-index calculations, SEC summaries, classification scripts, and sample metadata are included in the supplementary archive. Raw rheometer vendor files are provided alongside exported CSV files. The analysis scripts were tested with MATLAB R2021a and R 4.1.1.