Density and Copolymer Content in Polyethylene Samples by FT-NIR Spectroscopy

Importance of the topic

Polyethylene is among the highest-volume polymers globally and a major target for quality control and recycling. Rapid, non-destructive techniques that can distinguish polyethylene types and quantify composition (density, copolymer content) are valuable for production control, grade separation and sorting of recyclable streams. FT‑NIR combined with multivariate calibration offers a fast route to both qualitative classification and quantitative determination without chemical sample preparation.

Objectives and overview of the study

This application note reports two related studies using a Thermo Scientific Antaris FT‑NIR analyzer equipped with a Method Development Sampling (MDS) integrating sphere and spinning sample cup. The goals were:

• Study 1: classify polyethylene samples by density class (LLDPE, MDPE, HDPE) and develop a PLS model to predict density within MDPE samples.
• Study 2: quantify ethylene content in ethylene‑propylene copolymer films (2–16% ethylene) using PLS calibration.

The work demonstrates feasibility of rapid FT‑NIR assays for polymer sorting and composition monitoring using simple spectral pretreatment and chemometric models implemented in TQ Analyst software.

Methodology and instrumentation

Samples and scope:

• Study 1: 34 polyethylene specimens spanning three density classes: LLDPE (≈0.9170–0.9200 g·cm⁻³), MDPE (≈0.9340–0.9395 g·cm⁻³) and HDPE (>0.941 g·cm⁻³).
• Study 2: 28 random and impact ethylene‑polypropylene copolymer films containing 2–16% ethylene.

Spectral acquisition and preprocessing:

• Instrument: Thermo Scientific Antaris II FT‑NIR with MDS integrating sphere and spinning sample cup.
• Scan ranges used: general scans 10,000–4,000 cm⁻¹; model windows varied by application.
• Study 1 classification: first derivative spectra analyzed between 6,000 and 5,700 cm⁻¹; a Norris derivative smoothing filter was applied prior to discriminant analysis.
• Study 1 MDPE density PLS: unprocessed spectra from 10,000 to 6,200 cm⁻¹ with a one‑point baseline correction at 8,840 cm⁻¹.
• Study 2 copolymer PLS: unsmoothed unprocessed spectra from 9,000 to 4,500 cm⁻¹ with one‑point baseline at 9,029 cm⁻¹.

Chemometrics:

• Discriminant analysis (principal component scores) was used to separate density classes.
• Partial least squares (PLS) regression models were built and validated in TQ Analyst for quantitative predictions (density and ethylene content).

Main results and discussion

Classification (Study 1):

• Principal component analysis and discriminant modelling produced clear spectral separation among LLDPE, MDPE and HDPE classes using the selected NIR region and derivative preprocessing. All samples in the test set were correctly assigned to their density class.

MDPE density prediction (Study 1 quantitative):

• PLS model for 11 MDPE samples (density 0.9340–0.9395 g·cm⁻³) provided accurate density prediction with RMSEP = 0.0005 g·cm⁻³ and correlation coefficient r ≈ 0.977.

Ethylene content in polypropylene (Study 2):

• PLS calibration for ethylene content in copolymer films (2–16% ethylene) produced a strong fit with RMSEP ≈ 0.386% (reported <0.4%) and r ≈ 0.998, indicating excellent quantitative performance for low‑level copolymer content.

Implications of results:

• The discriminant model shows that specific NIR spectral features (notably in the 6,000–5,700 cm⁻¹ region) reliably reflect structural/density differences in polyethylene.
• The low RMSEP values for both density and ethylene content demonstrate that well‑designed PLS calibrations enable high precision composition measurements suitable for QC and sorting tasks.

Instrumentation used

The study employed Thermo Scientific Antaris II FT‑NIR analyzer with the Method Development Sampling (MDS) system, integrating sphere module and a spinning sample cup accessory. Data treatment and chemometric modelling were performed with TQ Analyst software.

Benefits and practical applications

Key practical advantages demonstrated:

• Rapid, non‑destructive measurement with no sample preparation, enabling high throughput screening.
• Robust class assignment for polyethylene grades supports automated sorting in recycling streams and incoming material verification.
• Accurate quantitative predictions (density and copolymer fraction) support in‑process quality control, blend verification and property prediction (e.g., melting point related to ethylene content).
• Compact FT‑NIR instrumentation and standardized chemometric workflows facilitate deployment at-line or in laboratory QA/QC environments.

Future trends and possibilities for use

Opportunities to extend and enhance the approach include:

• Calibration transfer and model robustness: expanding sample sets to cover broader suppliers, additives and processing history to improve generalization.
• Inline and at‑line implementations: integrating NIR probes or conveyor scanning for real‑time sorting and process control.
• Advanced chemometrics and machine learning: employing non‑linear or hybrid models to capture complex compositional or morphological effects.
• Hyperspectral imaging: combining spatial and spectral information to sort heterogeneous waste or identify contaminants.
• Measurement of other quality attributes: monitoring additives, degradation products, moisture or fillers to expand QC capability.

Conclusion

The application note demonstrates that FT‑NIR spectroscopy with appropriate preprocessing and chemometric models can both qualitatively classify polyethylene by density and quantitatively predict MDPE density and ethylene content in polypropylene copolymers with high accuracy. These capabilities make FT‑NIR a practical tool for polymer quality control, material sorting in recycling operations and compositional monitoring in production.

Reference

Pásztor J., Tenkl L., Strother T., Hirsch J., Density and Copolymer Content in Polyethylene Samples by FT‑NIR Spectroscopy, Application Note 51663, Thermo Fisher Scientific, 2008.