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

Significance of the topic

The rapid, non-destructive identification and quantification of polymers is critical for quality control in plastics manufacturing and for effective sorting in recycling streams. Polyethylene (PE) is one of the highest-volume polymers with multiple grades distinguished primarily by density and branching (LLDPE, MDPE, HDPE). Near-infrared Fourier transform spectroscopy (FT‑NIR) provides a fast analytical route to classify PE grades and to quantify copolymer content in related polymers such as ethylene‑propylene copolymers, supporting process control, product specification verification, and material separation for recycling.

Objectives and overview of the study

This application note reports two complementary studies using an Antaris II FT‑NIR analyzer with a Method Development Sampling (MDS) system: 1) classification of polyethylene samples into density-based classes (LLDPE, MDPE, HDPE) and quantitative prediction of MDPE density; 2) quantification of ethylene content in ethylene‑propylene copolymer films. The goals were to evaluate qualitative discrimination capability and to develop chemometric models (discriminant analysis and partial least squares regression) yielding quantitative predictions with laboratory-grade accuracy and minimal sample preparation.

Methodology

Study 1 (PE density classification and MDPE density prediction):

• Sample sets comprised three density ranges: LLDPE (~0.917–0.920 g·cm⁻3), MDPE (~0.926–0.940 g·cm⁻3), and HDPE (>0.941 g·cm⁻3).
• Spectral acquisition: 10,000–4,000 cm⁻1 using an integrating sphere with spinning sample cup to minimize sampling heterogeneity.
• Chemometrics: discriminant analysis used first‑derivative spectra between 6,000 and 5,700 cm⁻1 with a Norris derivative smoothing filter for classification; PLS regression for MDPE density used unprocessed spectra between 10,000 and 6,200 cm⁻1 with a one‑point baseline correction at 8,840 cm⁻1.

Study 2 (ethylene in polypropylene copolymers):

• 28 random and impact polypropylene film samples, ethylene content spanning ~2–16% (wt%).
• Spectral acquisition: 9,000–4,500 cm⁻1, unprocessed spectra with a one‑point baseline at 9,029 cm⁻1.
• Chemometrics: PLS regression to predict ethylene fraction in the copolymers.

Used instrumentation

• Thermo Scientific Antaris II FT‑NIR analyzer equipped with Method Development Sampling (MDS) system.
• Integrating sphere module with spinning sample cup (to average sample inhomogeneity and reduce packing effects).
• TQ Analyst software for discriminant analysis and PLS model development and validation.

Main results and discussion

Classification (Study 1):

• All polyethylene samples in the test set were correctly assigned to their density classes (LLDPE, MDPE, HDPE) by the discriminant model. Principal component score plots showed clear separation between classes, indicating distinct NIR spectral signatures related to density/branching differences.

Quantitation (Study 1 — MDPE density):

• A PLS model developed for 11 MDPE samples (density range 0.9340–0.9395 g·cm⁻3) produced a RMSEP of 0.0005 g·cm⁻3 and a calibration/validation correlation coefficient approximately 0.977, demonstrating high accuracy for density prediction within the MDPE subset.

Quantitation (Study 2 — ethylene in PP copolymers):

• The PLS model predicting ethylene content in polypropylene copolymer films achieved RMSEP ≈ 0.386% (reported as <0.4%) and a correlation coefficient ≈ 0.9976 across a 2–16% ethylene range, indicating excellent linearity and predictive performance.

Interpretation and practical considerations:

• Distinct NIR absorbance patterns tied to molecular composition and crystallinity/branching enable both reliable classification and accurate quantitation for these polymer systems.
• Use of an integrating sphere and spinning cup improves spectral reproducibility for heterogeneous polymer samples (pellets, films) by mitigating specular reflection and packing variation.
• Model performance depends on representative calibration sets; narrow-range quantitative models (MDPE) give superior precision but require coverage of expected variability (additives, fillers, processing history, temperature).

Benefits and practical applications

• Fast, non‑destructive analysis without chemical sample preparation — suitable for at‑line QC, incoming goods inspection, and support of sorting in recycling facilities.
• High classification accuracy enables automated sorting of PE grades (LLDPE/MDPE/HDPE), reducing cross‑contamination in recycling streams and improving recycled material quality.
• Quantitative determination of copolymer content (ethylene in PP) supports formulation control, melt processing setpoint adjustment, and specification compliance (e.g., melting point control via ethylene content).

Limitations and factors to control

• Calibration transfer between instruments or different sample geometries (films vs. pellets) requires attention; standardization or transfer methods may be necessary.
• Spectral interferences from additives, fillers, pigments, moisture, or surface contamination can degrade model accuracy if not represented in the calibration set.
• Temperature, packing density, and sample thickness influence NIR spectra and should be controlled or included during model development.

Future trends and potential applications

• Expansion of NIR polymer libraries and robust calibration transfer methods will enable broader deployment across recycling centers and production sites.
• Integration with portable/handheld FT‑NIR or process‑inline NIR probes for real‑time monitoring and automated sorting lines.
• Advanced chemometrics and machine learning approaches (nonlinear models, domain adaptation) can improve prediction across wider sample variability and mixed compositions.
• Combining NIR spectroscopy with hyperspectral imaging could enable spatially resolved identification of mixed materials on conveyor belts for automated sorting.

Conclusion

The Antaris II FT‑NIR system with an MDS sampling accessory demonstrates robust capability to distinguish polyethylene density classes and to provide accurate quantitative predictions of MDPE density and ethylene content in polypropylene copolymers. The approach offers rapid, non‑destructive analysis suitable for QC and recycling applications, provided calibration sets reflect expected sample variability and appropriate sampling geometry and temperature controls are used.

References

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