Sampling Considerations for the Measurement of a UV Stabilizer in Polymer Pellets Using FT-NIR Spectroscopy

Significance of the topic

Polymers produced at high throughput require rapid, reliable at-line quality control to monitor additive levels. Near-infrared (NIR) spectroscopy can deliver fast, non-destructive results without sample preparation, reducing time-to-answer versus classical wet-chemistry or chromatographic assays. For particulate and heterogeneous samples such as polymer pellets, representative sampling is critical: small or localized measurements may not reflect lot-level composition and can cause large variability in results. This study evaluates two diffuse-reflectance sampling approaches for FT-NIR analysis of a UV stabilizer in polystyrene pellets and demonstrates how an automated Sample Cup Spinner improves representativity, reproducibility, and throughput.

Objectives and overview of the study

• Compare two sampling methods for diffuse-reflectance FT-NIR of polystyrene pellets: (a) automated Sample Cup Spinner that continuously rotates the sample cup through the NIR beam to integrate over a large sample volume, and (b) manual multiple single-point measurements (rotating the cup manually by ~40° between measurements) followed by averaging.
• Develop a single chemometric calibration for quantification of a UV-stabilizer additive in polystyrene and validate prediction performance for unknown samples using both sampling modes.
• Assess analysis time, prediction error, and measurement variability to determine the most efficient and accurate sampling approach for heterogeneous pellets.

Methodology

• Samples: 17 polystyrene pellet samples supplied by a proprietary source; additive concentration in the set ranged approximately 42–58 weight %; pellet shapes and sizes varied.
• Calibration/validation split: 13 samples for calibration, 4 samples reserved for validation.
• Spectroscopy: Thermo Scientific Antaris FT-NIR analyzer with the Integrating Sphere Module in diffuse-reflectance mode; spectra collected with 8 cm-1 resolution and 16 scans. Acquisition times were short (spectra collection <10 s) with typical total analysis times reported near 15 s per sample.
• Sampling modes:
  • Sample Cup Spinner: the cup (47.8 mm quartz window) continuously rotates during data collection so the NIR beam probes many portions of the bulk in a single revolution, producing one integrated spectrum representative of the whole cup.
  • Manual single-point measurements: operator manually rotates cup ~40° between acquisitions; multiple spectra are recorded at discrete locations and averaged to obtain a reported value.
• Preprocessing and modeling: Norris second derivative (5-segment, 0-gap) as spectral pretreatment; Stepwise Multiple Linear Regression (SMLR) with two predictors (wavenumbers at 7332 cm-1 — first overtone region — and 5091 cm-1 — combination band region). Chemometric modeling and cross-validation were performed using Thermo Scientific TQ Analyst software; leave-one-out cross-validation used to assess model robustness.
• Validation protocol: four unknown samples were measured 30 times each by both sampling methods and results compared in terms of RMSEP, standard deviation, relative standard deviation (RSD), and prediction range.

Instrumention used

• Thermo Scientific Antaris FT-NIR analyzer with Integrating Sphere Module.
• Sample Cup Spinner accessory (solid-sampling cup with 47.8 mm quartz window and rotating mechanism).
• Thermo Scientific RESULT data-acquisition software.
• Thermo Scientific TQ Analyst quantitative chemometric software.
• Note: the study used an earlier Antaris model; Thermo Scientific later released Antaris II with improved speed and performance.

Main results and discussion

• Calibration performance: the two-term SMLR model achieved an excellent fit with correlation coefficient ≈ 0.9995, RMSEC = 0.147 wt%, and RMSECV (leave-one-out) = 0.179 wt%.
• Prediction performance: for the validation set, RMSEP using the Sample Cup Spinner was 0.302 wt%.
• Reproducibility and variability: repeated (n = 30) predictions on the same validation samples showed substantially lower variability with the Sample Cup Spinner versus manual single-point sampling. The standard deviation of predicted values from manual single-point measurements was approximately two times larger than that from the Sample Cup Spinner. Corresponding relative standard deviations were roughly 0.59% (spinner) vs 1.27% (manual) for a representative sample, and prediction ranges were consistently narrower for the spinner method.
• Representative sampling explanation: the Sample Cup Spinner integrates over many pellet positions in one acquisition, reducing the influence of local heterogeneity; single-point spectra sample a small fraction of material and typically do not capture lot-level variability, causing increased scatter and lower reproducibility.
• Time efficiency: the Sample Cup Spinner reduces operator involvement and overall analysis time by removing the need for repeated manual repositioning and multiple acquisitions to approximate bulk composition.

Benefits and practical applications of the method

• Fast, non-destructive quantification of additives in polymer pellets with minimal or zero sample preparation.
• Improved at-line/near-line process control: rapid feedback (seconds to tens of seconds) enables quicker corrective actions compared with lab-based wet-chemistry or GC workflows.
• Enhanced measurement representativity and reproducibility for heterogeneous solids through automated integration of larger sample volumes, reducing need for multiple sub-samples or operator-dependent procedures.
• Reduced consumable and labor costs by eliminating reagents and extensive sample preparation.

Future trends and potential uses

• Integration of modern FT-NIR instruments (e.g., Antaris II and successors) with faster electronics and improved optics will further shorten acquisition times and improve signal-to-noise, enabling tighter prediction errors and higher throughput in production environments.
• Automation and in-line adaptations: coupling continuous sampling or automated feed systems to integrating-sphere FT-NIR could enable true in-line monitoring of pellet production for real-time process control.
• Advanced chemometrics and machine learning: multivariate models that exploit full-spectrum information, non-linear regression, or transfer-learning strategies could broaden applicability to lower-concentration additives or more complex formulations while maintaining robustness to pellet heterogeneity.
• Standardization of sampling accessories and protocols for particulate materials to ensure comparability across plants and instruments.

Conclusion

The study demonstrates that diffuse-reflectance FT-NIR combined with an automated Sample Cup Spinner accessory provides rapid, accurate, and reproducible quantification of a UV stabilizer in polystyrene pellets. The Sample Cup Spinner outperforms manual single-point sampling by producing more representative spectra, halving measurement variability, and reducing operator time and complexity. FT-NIR with appropriate sampling design is a practical at-line solution for polymer production quality control, and instrument and chemometric advances will expand its utility further.

Reference

• Thermo Fisher Scientific application note AN50790_E (2022): Sampling considerations for measurement of a UV stabilizer in polymer pellets using FT-NIR (Antaris FT-NIR analyzer, Sample Cup Spinner, TQ Analyst). For research-use application note material from Thermo Fisher Scientific.