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

Significance of the topic

The accurate, rapid and representative measurement of additives in polymer pellets is critical for modern polymer production quality control. Heterogeneous solid samples such as pellets can show local compositional variability, and traditional laboratory methods (extraction, titration, GC) are time-consuming, destructive and require skilled personnel. Fourier-transform near-infrared (FT-NIR) spectroscopy combined with appropriate sampling accessories offers at-line or near-line, non-destructive, reagent-free analysis with results in minutes, improving production feedback and reducing scrap and rework.

Objectives and overview of the study

This application note compares two diffuse-reflectance sampling approaches for quantifying a UV stabilizer additive in polystyrene pellets using FT-NIR: (1) an automated Sample Cup Spinner accessory that continuously rotates the sample cup during acquisition to integrate over a large sample volume, and (2) manual multiple single-point measurements taken at different cup positions with manual rotation between acquisitions. The goal was to determine which sampling strategy yields the most accurate, reproducible and time-efficient prediction of additive concentration using a single chemometric calibration model.

Methodology and chemometrics

- Samples: 17 polystyrene pellet lots from a proprietary source; 13 used for calibration (additive 42–58 wt%), 4 used as independent validation samples.
- Sampling modes compared: Sample Cup Spinner (continuous rotation; single integrated spectrum per sample) versus manual single-point sampling (multiple spectra averaged by manual repositioning; ~40° rotation between measurements).
- Spectrometer and acquisition: Thermo Scientific Antaris FT-NIR with Integrating Sphere Module; internal gold background reference used with the sample cup in place. Spectra collected at 8 cm-1 resolution with 16 scans; typical collection time ~10 seconds (total analysis time per sample ~10–15 s).
- Preprocessing and model: Norris second derivative (5-segment, 0-gap) pretreatment; two-term Stepwise Multiple Linear Regression (SMLR) model built using two spectral variables at 7332 cm-1 (first overtone region) and 5091 cm-1 (combination band region). Cross-validation performed with leave-one-out protocol.
- Software: Thermo Scientific TQ Analyst for chemometrics and RESULT data acquisition software.

Used instrumentation

• Thermo Scientific Antaris FT-NIR analyzer with Integrating Sphere Module
• Sample Cup Spinner accessory with 47.8 mm quartz window (solid sampling cup)
• Internal gold reference for the integrating sphere
• Thermo Scientific RESULT software and TQ Analyst quantitative analysis package

Main results and discussion

- Calibration performance (Sample Cup Spinner data): correlation coefficient r = 0.9995; RMSEC = 0.147 wt%; RMSECV (leave-one-out) = 0.179 wt%. The final SMLR model used two wavelengths (7332 and 5091 cm-1).
- Prediction (validation) performance: RMSEP for validation samples analyzed with the Sample Cup Spinner = 0.302 wt%.
- Reproducibility: For each of four validation samples, the Sample Cup Spinner produced predicted concentrations with substantially lower standard deviation than manual single-point sampling. The single-point technique showed approximately twice the standard deviation (i.e., ~2× higher variability) compared to the spinner measurements across 30 repeated acquisitions.
- Practical observation: Spectra collected with the Sample Cup Spinner were visually more consistent than spectra from single-point measurements. The spinner’s continuous rotation integrates signals from a larger sample volume (largest volume passing through the NIR beam in one revolution), reducing the influence of local heterogeneities that bias single-point spectra.
- Time and workflow: The Sample Cup Spinner reduces operator intervention and overall analysis time by avoiding repeated manual cup rotations and multiple acquisitions needed to approximate bulk composition.

Benefits and practical use of the method

- Representative sampling: The spinner yields spectra that better represent the bulk pellet lot, improving prediction accuracy for heterogeneous solids.
- Speed and at-line utility: Rapid acquisition (seconds) enables at-line or near-line monitoring to provide timely production feedback and reduce out-of-spec production.
- Non-destructive, minimal preparation: No chemical reagents or destructive sample prep required; samples remain available for other analyses.
- Operator reproducibility: Automated rotation reduces operator-to-operator variability compared with manual single-point protocols.
- Simplicity of model: Robust calibration using a small number of well-chosen spectral features can yield high precision when combined with representative sampling.

Limitations and practical recommendations

• Calibration range and representativeness: Calibrations must cover the expected additive concentration range and account for pellet size/shape variations; here calibration spanned 42–58 wt%.
• Sample cup/window maintenance: Clean quartz window and consistent packing/leveling of pellets in the cup are necessary to avoid artifacts and maintain repeatability.
• Spectral preprocessing and validation: Use appropriate derivative or scatter-correction pretreatments and rigorous cross-validation (leave-one-out or better) and independent validation sets.
• When single-point sampling is used, increase the number of positions and replicate measurements to partially mitigate heterogeneity effects, at the cost of longer analysis time.

Future trends and possibilities for application

- Instrument evolution: Faster, higher-performance FT-NIR instruments (e.g., Antaris II and successors) and improved detector sensitivity will reduce acquisition times and further improve precision.
- Inline/online integration: Development of robust inline probes, automated sampling devices and process integration will enable continuous monitoring of additives during compounding and pelletizing.
- Advanced chemometrics and machine learning: Multivariate calibration transfer, regularized regression and machine-learning approaches can improve robustness against batch-to-batch variability and complex matrix effects.
- Hybrid sampling strategies: Combining automated rotation with controlled sample agitation or flowing streams could further minimize heterogeneity impacts for challenging solids.
- Standardization and calibration networks: Wider adoption of standardized procedures for cup geometry, acquisition parameters, and validation protocols will ease model transfer between sites and instruments.

Conclusion

This study demonstrates that a representative sampling accessory—the Sample Cup Spinner—significantly improves the precision and reliability of FT-NIR quantitation of a UV stabilizer in polystyrene pellets relative to manual single-point sampling. The spinner reduces measurement variability (roughly twofold improvement in standard deviation in this study), shortens analysis time, and provides spectra that better reflect bulk composition. When combined with a simple, well-validated chemometric model, FT-NIR with automated cup rotation is an effective at-line tool for rapid additive monitoring in polymer manufacturing.

References

The application note data and methodology are based on an internal Thermo Scientific Antaris FT-NIR evaluation described in the provided application note. No external literature references were listed in the source document.