1. A.R. McLauchlin1, O. Ghita1 and B. Tübke2
Detection of Contaminants in Recycled Poly(ethylene terephthalate) (PET) by Near Infrared Spectroscopy
A.R. McLauchlin1, O. Ghita1 and B. Tübke2
1College of Engineering, Mathematics and Physical Sciences, University of Exeter, UK 2Institute for Chemical Technology, Fraunhofer Institute, Karlsruhe, Germany

2. Contents Introduction: The SupercleanQ Project Materials and Methods
Results
Conclusions

3. SupercleanQ Aim Role of spectroscopy
A post-market challenge test for validation of recycled food contact materials with 100% reliability
rPET must comply with EC 282/2008
Role of spectroscopy
Offline measurement of contaminants in rPET
Inline measurement of degradable contaminants in rPET during injection moulding

4. Contaminants: Examples
Limonene
Fruit Juice Carbonated Beverages
Acetaldehyde
By-product of PET synthesis
Benzophenone
UV stabiliser

5. Concerns Over Purity of rPET
If compounds can diffuse into PET, can they diffuse out again?
Franz, R., Mauer, A., & Welle, F. (2004). Food additives and contaminants, 21(3), 265–286.
Cecchi, T., Passamonti, P., & Cecchi, P. (2009). Is It Advisable to Store Olive Oil in PET Bottles? Food Reviews International, 25(4), 271–283.

6. Hansen Solubility Parameter
Surrogate
Mol. Wt g/mole
B.Pt °C
Dipole Moment D
Hansen Solubility Parameter
ACETALDEHYDE
44.05
20.2
2.7
20.9
BENZOPHENONE
182.22
305.4
2.98
22.1
PET
21.6
Sum of dispersive, polar and hydrogen bonding properties of a molecule
Predictor of miscibility of two molecules
Acetaldehyde and benzophenone likely to be absorbed by PET

7. Current Methods Thermal desorption / Solvent extraction
Gas or Liquid Chromatography + MD*
For: Sensitive, accurate
Against: Slow, expensive
Need for rapid, sensitive method
*Franz, Mauer & Welle (2004)

8. Materials and Methods Previous work: “Single Dose”
Aim: To achieve a range of concentrations
Volatility (b.pt. Acetaldehyde = 20°C)
Precludes melt compounding
Chosen approach:
Uptake from aqueous solutions of various concentrations

9. Impregnation Pristine juice bottles Acetaldehyde or Benzophenone
2mg/l - 400mg/l
10% Methanol in all Benzophenone solutions
5 50°C °C in the dark

10. Analysis Bottle wall only was sampled PET analysis: Water analysis:
Benzophenone: TD-GC-MS
Acetaldehyde: TD + 2,4-DNPH then HPLC-DAD
Water analysis:
HPLC-DAD
Acetaldehyde: 2,4-DNPH derivative
Benzophenone: no derivatisation

11. Spectroscopy Apparatus 21 spectra / bottle Bruker Matrix
Ocean Optics QP600 probes
21 spectra / bottle
16 scans
cm-1
4cm-1 resolution

12. Data Analysis Predictive Models: Model Tested Acetaldehyde:
OPUS 6.5
Quant2 Optimisation
15 out of 21 spectra used
Outliers deselected
Model Tested
5 additional spectra
Not used in the validation
Acetaldehyde:
No spectral preprocessing
Benzophenone
Straight Line Subtraction

13. Data Analysis Spectral ranges: Both: 10000-7500cm-1
2nd and 3rd overtone C-H
Benzophenone: cm-1
1st overtone CH CH2 CH3
Acetaldehyde: cm-1
Aldehyde groups
< 0.12AU

14. PET Analysis: Benzophenone
Benzophenone detected in all but 2 treatments

15. PET Analysis: Acetaldehyde
Acetaldehyde not detected in any sample
Acetaldehyde + ethylene glycol = acetal product: 1-methyl-2,3-dioxolane

16. Validation Curve: Acetaldehyde
RMSECV = 11.1
Initial rank = 9
Final rank = 6

17. Validation Curve: Benzophenone
RMSECV = 13.6
Initial rank = 9
Final rank = 7

18. Factor Analysis 1 Maxima coincide with absorbance maxima for PET
Spectral variation arises from interaction of analyte with PET.
Factors 1+2+3= 95% of spectral variance

19. % of total variance represented by:
Factor Analysis 2
% of total variance represented by:
Analyte
Factor 1
Factor 2
Factor 3
Factor 4
Factor 5
Total
Acetaldehyde
87.5
6.8
0.8
0.5
0.4
95.9
Benzophenone
87.3
7.2
0.7
96.0
Factors 1 & 2 accounted for most of the variance

20. Factor Analysis 3: Score Plots
Acetaldehyde
Benzophenone

21. Testing the Model: Acetaldehyde
Coefficient of 0.897: underestimates Acetaldehyde conc.
Detection limit in PET could not be defined

22. Testing the Model: Benzophenone
Coefficient of : good correspondence
Effective range mg/kg PET

23. Conclusions NIR detected presence of acetaldehyde and benzophenone
Linear PLS models obtained in both cases
Good prediction of benzophenone
Underestimate of acetaldehyde
Response to acetaldehyde may have been due to reaction product with ethylene glycol

24. Acknowledgements
The research leading to these results was performed as part of the SupercleanQ Project, funded by the European Union’s Seventh Framework Programme under grant agreement no