+ All Categories
Home > Documents > Contaminant levels in recycled PET plastic · I am particularly grateful to my supervisors Dr R. F....

Contaminant levels in recycled PET plastic · I am particularly grateful to my supervisors Dr R. F....

Date post: 18-May-2018
Category:
Upload: ngotram
View: 221 times
Download: 0 times
Share this document with a friend
319
NOTE This online version of the thesis may have different page formatting and pagination from the paper copy held in the Swinburne Library.
Transcript

NOTE

This online version of the thesis may have different page formatting and pagination from the paper copy held in the Swinburne Library.

CONTAMINANT LEVELS

IN

RECYCLED PET PLASTIC

By

LIDIA KONKOL

A thesis submitted in fulfilment of the requirements for the

degree of Doctor of Philosophy

November, 2004

Environment and Biotechnology Centre

Swinburne University of Technology

Victoria 3122, Australia

Acknowledgments

ACKNOWLEDGMENTS

I would like to dedicate this thesis to my loving family members and partner for all their

support, understanding, optimism and encouragement throughout my academic years.

I am particularly grateful to my supervisors Dr R. F. Cross, Dr I. Harding and Dr E.

Kosior for their guidance and valuable suggestions. A special thank-you goes out to Dr

Reg Cross for motivating and assisting me in writing papers.

I would like to express my gratitude to my friend Larry Bautista from Philip Morris for

assisting me with the static headspace work and for being a great friend.

Finally I would like to thank my fellow postgraduate students and Swinburne staff,

especially Savithri Galappathie, Sheila Curtis and Andrew Smairl, for their friendship

throughout my academic years.

i

Preface

PREFACE

I hereby declare that, to the best of my knowledge, this thesis contains no material

previously written or published by another person except where reference is made in

the text. I also declare that none of this work has been previously submitted for a

degree or similar award at another institution.

ii

Table of contents

TABLE OF CONTENTS

ACKNOWLEDGEMENTS i

PREFACE ii

TABLE OF CONTENTS iii

LIST OF FIGURES x

LIST OF TABLES xv

ABBREVIATIONS xviii

ABSTRACT xx

CHAPTER 1: INTRODUCTION 1

CHAPTER 2: LITERATURE REVIEW 3

2.1 BACKGROUND 3 2.1.1 Definition of PET and its applications in the food industry 3

2.1.2 Manufacture of PET bottles 4

2.1.3 Improving gas barrier properties of PET 5

2.2 FOOD CONTACT CONSIDERATIONS FOR VIRGIN PET 7

2.2.1 Introduction 7

2.2.2 Sorption considerations in food contact applications 8

2.2.3 Factors contributing to the degree of sorption 9

2.2.3.1 Properties of sorbate 10

2.2.3.2 Polymer chemical and morphological properties 11

2.2.3.3 Solubility parameter 12

2.2.3.4 Polymer physical properties 13

2.2.3.5 Temperature 14

2.2.3.6 Time 15

2.2.4 Migration considerations in food contact applications 16

2.2.5 Factors affecting the extent of migration 18

iii

Table of contents

- External factors 18

- Polymer and migrant factors 19

2.2.6 Potential migrants resulting from the manufacture of PET 21

- Residual compounds resulting from manufacture identified in PET 21

2.2.7 Reaction by-products formed during PET manufacture 25

- Acetaldehyde 25

- Oligomers in PET 26

- Oligomer migration from PET 28

2.2.8 Additives 29

2.2.9 Global migration 30

2.2.10 Other compounds identified in PET 30

2.3 FOOD CONTACT CONSIDERATIONS FOR RECYCLED PET 31

2.3.1 Introduction to recycling 31

2.3.2 Modes of recycling 31

2.3.2.1 Re-use: Zeroth order recycling 32

2.3.2.2 Primary recycling 32

2.3.2.3 Physical reprocessing: Secondary recycling 32

- The Visy process 33

2.3.2.4 Tertiary recycling 33

2.3.3 Recycled PET for food contact purposes 34

2.3.3.1 Consumer misuse/reuse 35

2.3.3.2 Sorption from the original contents of the bottle 35

2.3.4 Threshold of regulation 39

2.3.5 Validation of recycling process – the challenge test 40

2.3.5.1 Introduction 40

2.3.5.2 Challenge test process 42

2.3.5.3 Challenge test studies 42

- Refillable plastic bottles 42

- Secondary recycled plastic bottles 44

2.3.6 Estimated level of real contaminants in recycled PET 47

2.3.7 Methods of reducing contamination 48

2.3.7.1 Functional barrier 48

iv

Table of contents

2.4 EXTRACTION AND ANALYSIS OF POLYMERS 50

2.4.1 Modes of extraction 50

2.4.2 Parameter optimisation 51

2.4.2.1 Time 51

2.4.2.2 Temperature 52

2.4.2.3 Pressure 54

2.4.2.4 Nature of extraction solvent 54

2.4.2.5 Particle size 57

2.4.2.6 Migrant shape/size 58 2.4.3 Modes of separation and analysis 59

2.5 PURPOSE OF THIS THESIS 60

2.6 OUTLINE OF THIS THESIS 62

CHAPTER 3: MATERIALS AND METHODS 64

3.1 METHOD FOR CHAPTER 4 64

3.1.1 Chemicals 64

3.1.2 Preparation of stock standards 67

3.1.3 Soxhlet calibration standards 67

3.1.4 Dissolution calibration standards 67

3.1.5 Gas chromatography-mass spectroscopy (GC-MS) analysis 68

3.1.6 Commercial Visy treatment of curbside PET 68

3.1.7 Laboratory preparation of polymer before analysis 69

3.1.8 Soxhlet extraction conditions 69

3.1.9 Sonication 70

3.1.10 Total dissolution extraction conditions 70

3.1.10.1 Total dissolution by TFA – Qualitative analysis 70

3.1.10.2 Total dissolution by TFA – Quantitative analysis 70

3.1.10.3 Total dissolution by HFIP – Qualitative analysis 70

v

Table of contents

3.1.11 Crystallinity analysis 71

3.2 METHOD FOR CHAPTER 5 71

3.2.1. Chemicals 71

3.2.2. Crystallinity analysis 71

3.3 METHOD FOR CHAPTER 6 71

3.3.1 Preparation of stock standards 71

3.3.2 Soxhlet calibration standards for external standardisation 72

3.3.3 SPME 72

3.3.4 Static Headspace (SHS) 73

3.3.5 Quantitative analysis by SHS 73

3.3.6 GC/MS Conditions – SPME 73

3.3.7 GC/MS Conditions – SHS 74

CHAPTER 4: SEMI-VOLATILE CONTAMINANTS 75

AND LEVELS OF OCCURRENCE IN WASHED AND

DRIED SHREDDED PET

4.1 GENERAL INTRODUCTION 75

4.1.1 Purpose of the chapter 75

4.1.2 Brief outline of chapter 76

4.1.3 Selecting the right extraction solvent for Soxhlet extraction 76

4.2 QUALITATIVE STUDY OF CONTAMINANTS IN WASHED 79

AND DRIED PET FLAKE

4.2.1 Introduction to Soxhlet extraction of washed and dried flake 79

4.2.2 Choosing a suitable low boiling solvent 79

4.2.3 GC/MS analysis of DCM extracts of washed and dried flake 84

4.2.4 Qualitative analysis of washed and dried flake extracted by 93

total dissolution

4.2.5 Running the extracts on polar column 98

vi

Table of contents

4.2.6 Possible origin of the components 99

4.3 QUANTITATIVE STUDY OF CONTAMINANTS IN WASHED 102

AND DRIED PET FLAKE

4.3.1 Introduction 102

4.3.2 Study of extraction kinetics for flake ground to 0-300 µm 103

4.3.3 Particle size variation 116

4.3.4 Kinetic studies for the larger particle sizes 121

4.3.5 Comparison of contaminant levels in different 70 g grabs from the 140

original 15 kg sample

4.3.6 Validation of the Soxhlet extraction methodology 142

4.3.6.1 Total dissolution compared with Soxhlet extraction 142

4.3.7 Particle size range and degree of crystallinity 148

4.3.8 Investigation of the relative levels of contaminants in the two types 153

of flake

4.3.9 Representative sampling 159

4.3.10 Levels of contaminants in flake and the threshold of regulation 160

CHAPTER 5: SEMI-VOLATILE CONTAMINANTS 162

AND LEVELS OF OCCURRENCE IN EXTRUDED PET PELLETS FROM CURBSIDE COLLECTION

5.1 GENERAL INTRODUCTION 162

5.1.1 Purpose of the chapter 162

5.1.2 Brief outline of this chapter 164

5.2 KINETICS OF SOXHLET EXTRACTION FROM EXTRUDED 165

AND ANNEALED PET

5.2.1 Pellets ground to 0-300 µm 165

5.2.1.1 Pellets ground to 0-300 µm: The relationship between 171

extraction kinetics and contaminant molecular weight.

5.2.2 Annealed pellets ground to >300-425µm 174

vii

Table of contents

5.2.3 Annealed pellets ground to >425-700 µm 176

5.2.4 Unground annealed pellets 178

5.2.5 The effect of particle size reduction upon measured contaminant levels 182

in extruded and annealed pellets

5.3 ANNEALED VERSUS AMORPHOUS EXTRUDED PELLETS 189

5.3.1 Kinetics of extraction from amorphous pellets 189

5.3.2 Variation of DCM uptake with PET crystalline structure 195

5.3.3 Contaminant diffusion coefficients out of amorphous and annealed PET 199

5.3.4 Contaminant loss during the annealing of pellets 204

5.3 FLATTENING AMORPHOUS PET PELLETS 205

5.4 LEVELS OF CONTAMINANTS IN PELLETS AND 208

THRESHOLD OF REGULATION

5.5 CONCLUSIONS 211

CHAPTER 6: VOLATILE CONTAMINANTS AND 213

LEVELS OF OCCURRENCE IN EXTRUDED PET FLAKE

AND PELLETS FROM CURBSIDE COLLECTION 6.1 GENERAL INTRODUCTION 213

6.1.1 Purpose of the chapter 213

6.1.2 Background to thermal extraction 213

6.1.3 Brief outline of chapter 216

6.2 QUALITATIVE SPME STUDY OF CONTAMINANTS 217

IN PET EXTRUDED PELLETS

6.2.1 Comparison of the compounds extracted by different fibres 217

6.2.2 Effect of temperature on extraction 223

viii

Table of contents

6.2.3 Effect of mass of sample on extraction 235

6.2.4 Effect of adsorption time 242

6.2.5 Effect of extraction time on extraction 243

6.3 QUANTITATIVE SPME AND STATIC HEADSPACE STUDY 249

OF RECYCLED PET

6.3.1 Quantitation using the CX/PDMS fibre 249

6.3.2 SPME using PDMS, an absorption fibre 249

6.3.3 Static headspace analysis (SHS) 253

6.3.4 Quantitative analysis of PET 258

6.3.5 Multiple headspace extraction (MHE) 258

6.3.6 External standardisation (ES) 260

6.4 CONCLUSION 262

CHAPTER 7: CONCLUSIONS 264

APPENDIX 268

BIBLIOGRAPHY 269

ix

List of figures

LIST OF FIGURES

Figure

2.1 Formation of PET (polyethylene terephthalate) 2

2.2 Structure of PEN 5

2.3 Drawing of the bottom part of a PET soft drink bottle 6

illustrating sorption, migration and permeation

2.4 A summary of the disadvantages of “flavour scalping” 8

2.5 Sorbate, polymer and external factors effecting sorption in PET 9

2.6 Formation of dimethyl terephthalate and terephthalic acid from 22

xylene

2.7 Formation of bis-(2-hydroxyethyl)terephthalate (BHET) 24

from dimethyl terephthalate and ethylene glycol

2.8 Formation of PET from BHET 25

2.9 Formation of acetaldehyde from PET 26

2.10 Cyclic oligomers identified in PET 27

2.11 Chemolysis reactions used in tertiary recycling 34

4.1 Plot of the number of mmole of solvent absorbed at 3 h versus 83

(δPET - δsolvent)

List of figures

4.2 Chromatogram of DCM extract for washed and dried flake 87

4.3 Mass spectrum and structure of (a) cyclic dimer and (b) dimer ether 92

4.4 Chromatogram of TFA/heptane extract for washed and dried flake 97

4.5 Chromatogram of HFIP extract for washed and dried flake 99

4.6 Schematic presentation of the three subsequent steps in solvent 105

extraction

4.7 Soxhlet extraction kinetic study of washed and dried flake ground to 106

0-300 µm. Compounds identified at levels below 200 ppb

4.8 Soxhlet extraction kinetic curves of trimethylnaphthalene isomers 107

extracted from washed and dried flake ground to 0-300 µm

4.9 Soxhlet extraction kinetic study of washed and dried flake ground 108

to 0-300 µm. Compounds identified at levels above 200 ppb

4.10 The standard deviations associated with data points defining 109

kinetic curves that do not follow the general trends of Figure 1

4.11 Soxhlet extraction kinetic study of washed and dried flake ground

to 0-300 µm. Ethylene glycol analysed on an EC-Wax Econo-cap

column

4.12 Ratio of amount extracted at 8 h (A8) to amount extracted after 24 h (Ae)

(as a percentage) versus contaminant molecular weight

4.13 “Venn diagram” grouping the contaminants according to their functional

group types

List of figures

4.14 Ratio of amount extracted at 8 h (A8) to amount extracted after 24 h (Ae)

(as a percentage) versus estimated solvent strength parameter

4.15 Amount of contaminant extracted from flake ground to different particle

sizes (compounds below 200 ppb)

4.16 Amount of trimethylnaphthalene contaminants extracted from flake ground

to different particle sizes

4.17 Amount of contaminant extracted from flake ground to different particle

sizes (compounds above 200 ppb)

4.18 The standard deviations associated with data points defining curves that do

not follow the general trends of Figure 4.15 – 4.17

4.19 Soxhlet extraction kinetics of flake ground to >300-425 µm. Contaminants

below 120 ppb

4.20 Soxhlet extraction kinetics of flake ground to >300-425 µm. Contaminants

between 120 ppb and 300 ppb

4.21 Soxhlet extraction kinetics of flake ground to >300-425 µm. Contaminants

above 400 ppb

4.22 Soxhlet extraction kinetics of flake ground to >300-425 µm.

Trimethylnaphthalene isomers

4.23 Soxhlet extraction kinetics of flake ground to >425-700 µm. Contaminants

below 100 ppb

4.24 Soxhlet extraction kinetics of flake ground to >425-700 µm.

Trimethylnaphthalene isomers

List of figures

4.25 Soxhlet extraction kinetics of flake ground to >425-700 µm. Contaminants

above 200 ppb

4.26 Soxhlet extraction kinetics of whole flake. Contaminants below 70 ppb

4.27 Soxhlet extraction kinetics of whole flake. Trimethylnaphthalene

isomers

4.28 Soxhlet extraction kinetics of whole flake. Contaminants between 70 ppb

and 200 ppb

4.29 Soxhlet extraction kinetics of whole flake. Contaminants above 200 ppb.

4.30 Soxhlet extraction kinetics of whole flake. Ethylene glycol

4.31 Log-log plot of levels of contaminants determined by total dissolution

versus levels extracted by sonication and comparison with the ideal

relationship (y=x): for flake ground to small particle sizes

4.32 Variation in contaminant levels between two 70 g grabs of flake from the

original 15 kg bag. Analyses were performed on PET ground to the 0-300

µm particle size in each case

4.33 Log-log plot of levels of contaminants determined by total dissolution

versus levels extracted by Soxhlet and comparison with the ideal

relationship (y=x): for flake ground to medium and large particle sizes and

for unground flake

4.34 Log-log plot comparing the contaminant concentrations extracted from

unground flake by TFA versus the amount extracted from unground flake by

Soxhlet extraction

List of figures

4.35 Percentage of amount extracted from the >425-700 µm particle size range to

the amount extracted from the 0-300 µm particle size range plotted versus

contaminant molar mass

4.36 Log amount extracted from crystalline particles versus log amount extracted

from amorphous particles, for each particle size range.

4.37 Log amount extracted from whole amorphous pellets versus log amount

extracted from flattened amorphous particles.

5.1 Soxhlet extraction kinetic study of annealed pellets ground to 0-300 µm.

Compounds identified at levels below 10 ppb.

5.2 Soxhlet extraction kinetic study of annealed pellets ground to 0-300 µm.

Trimethylnaphthalene isomers.

5.3 Soxhlet extraction kinetic study of annealed pellets ground to 0-300 µm.

Compounds identified at levels between 11 ppb and 130 ppb.

5.4 Percentage of contaminant extracted at 1 h versus molecular weight.

5.5 A log-log plot of the amounts of contaminants extracted at 24 h versus the

amounts extracted at 3 h for >300-425µm.

5.6 A log-log plot of the amounts of contaminants extracted at 24 h versus the

amounts extracted at 3 h for particles >425-700 µm.

5.7 Soxhlet extraction kinetic study of unground annealed pellets. Compounds

identified at levels below 2 ppb.

5.8 Soxhlet extraction kinetic study of unground annealed pellets. Compounds

identified at levels below 13 ppb.

List of figures

5.9 Soxhlet extraction kinetic study of unground annealed pellets. Compounds

identified at levels below 70 ppb.

5.10 Typical variations in contaminant levels measured from the same batch of

annealed pellets ground to the three particle sizes.

5.11 Extreme examples of the variation in contaminant levels.

5.12 An example of the experimental spread (means ± standard deviation) for

divergent measurements of a contaminant in the three particle sizes derived

from the same batch of annealed pellets.

5.13 Log-log plot of the amounts of contaminants extracted from >425-700 µm

particles versus the amounts extracted from 0-300 µm particles.

5.14 A log-log plot of the amounts extracted at 3 h versus the amounts extracted at

24 h for unground amorphous pellets.

5.15 Soxhlet extraction kinetic study of unground amorphous pellets. Compounds

identified at levels below 4 ppb.

5.16 Soxhlet extraction kinetic study of unground amorphous pellets. Compounds

identified at levels below 11 ppb.

5.17 Soxhlet extraction kinetic study of unground amorphous pellets. Compounds

identified at levels below 13 ppb.

5.18 Soxhlet extraction kinetic study of unground amorphous pellets.

Trimethylnaphthalene compounds.

5.19 Sorption kinetics of DCM into amorphous and annealed pellets.

List of figures

5.20 A plot of the (amount of DCM sorbed /amount sorbed at equilibrium) versus

the square root of time.

5.21 A plot of At/Ae (amount extracted/amount extracted at equilibrium from

annealed pellets) versus the square root of time (a representative plot;

naphthalene).

5.22 A plot of calculated diffusion coefficients versus molecular weights (for

annealed pellets).

5.23 A plot of fraction extracted at 2 h versus molecular weights (for amorphous

pellets).

5.24 A log-log plot of the amounts of contaminants extracted from amorphous

pellets versus the amounts extracted from ground annealed pellets.

6.1 Ground annealed pellets: contaminant area (abundance) versus extraction

temperature for three different particle sizes using the CX/PDMS fibre.

6.2 Effect of incubation temperature on extraction of 6g of unground

extruded pellets using the CX/PDMS fibre.

6.3 Effect of incubation temperature on extraction of 1g of unground

extruded pellets using the CX/PDMS fibre.

6.4 Effect of incubation temperature on extraction of 0.3g of unground

extruded pellets using the CX/PDMS fibre.

6.5 Effect of extraction time on abundance.

6.6 Superimposed chromatograms obtained from the analysis of pellets using the

PDMS (bold) and CX/PDMS (fine) fibres.

6.7 Effect of incubation temperature on extraction of 6g of unground extruded

pellets using the 100 µm PDMS fibre.

List of figures

6.8 Chromatograms for extruded pellets obtained by (a) SHS and (b) SPME using

the CX/PDMS fibre.

6.9 Effect of incubation temperature for the SHS of extruded PET pellets.

6.10 Multiple headspace analysis of flake ground to 425–700 µm.

LIST OF TABLES

Table

2.1 Comparative properties of PET versus PEN.

2.2 A list of FDA food simulants.

2.3 A list of EEC food simulants.

2.4 Threshold values for the maximum permitted contaminant

concentration in polymers and food simulant.

2.5 Surrogates used in a challenge test performed by Franz and Welle

(1999a).

2.6 The estimated level of contaminants in PET after each recycling stage.

2.7 Hildebrand solubility parameters for different solvents and polymers.

3.1 Contaminants identified in washed and dried PET flake and the

standards used.

4.1 Hildebrand solubility parameters of some solvents and PET.

4.2 Hildebrand solubility parameters of hexane, 2-propanol, ethanol and

PET.

4.3 Compounds identified in ground washed and dried PET flake [“x”

denotes presence of compound in virgin (V) and recycled (R) PET].

4.4 Compounds extracted from washed and dried flake by total dissolution

using TFA/heptane [“x” denotes presence of contaminant in virgin (V)

and recycled (R) PET].

4.5 Constituents of PET were also extracted by total dissolution using

HFIP [“x” denotes presence of contaminant in virgin (V) and recycled

(R) PET].

4.6 Soxhlet extract run on an EC-Wax Econo-cap column.

4.7 The percentages of naphthalene derivatives extracted at 8 h alongside

their molecular structure.

4.8 Contaminant levels (and standard deviations) [in ppb, in flake ground

to 0-300 µm] determined by total dissolution with TFA, compared to

extraction by sonication in DCM.

4.9 Contaminant levels (and standard deviations) [in ppb, in flake ground

to 0-300 µm] determined by total dissolution with TFA, compared to

extraction by sonication in DCM: anomalies for (a) m-cymene

[TFA>DCM] and (b) limonene, cineole and γ-terpinene [DCM>TFA].

(All levels are in ppb by mass.)

4.10 Flake ground to >300-425 µm particle size and extracted for 3 h and

then re-extracted for another 24 h.

4.11 Levels of contaminants (and their standard deviations) determined by

Soxhlet extraction with dichloromethane (DCM) compared with total

dissolution by trifluoracetic acid (TFA) followed by extraction with

heptane: for flake ground to medium and large particle sizes, and for

unground flake. (All levels are in ppb by mass.)

4.12 Levels of contaminants (and their standard deviations) determined by

Soxhlet extraction with dichloromethane (DCM) compared with total

dissolution by trifluoracetic acid (TFA) followed by extraction with

heptane: anomalies for (a) m-cymene [TFA>DCM] and (b) limonene,

cineole and γ-terpinene [DCM>TFA]. (All levels are in ppb by mass.)

4.13 Percentages of crystallinity for amorphous and crystalline fractions of

washed and dried flake ground to different particle sizes.

4.14 Percentages of crystallinity for two batches of unsegregated washed

and dried flake ground to different particle sizes.

4.15 Mass of amorphous and crystalline flake ground to different particle

sizes.

4.16 Levels of contaminants in amorphous and crystalline flake ground to

different particle sizes (analysed by sonication in DCM for 3 h). (All

levels are in ppb by mass.)

5.1 Amounts of contaminants extracted from annealed pellets

ground to 0-300 µm by Soxhlet extraction and sonication (standard

deviation, n=3 for 3 h; n=2 for 15 h).

5.2 Flattened and whole amorphous pellets extracted by sonication and

Soxhlet extraction.

5.3 Levels of contaminants in ground flake (0-300 µm), unground flake

and extruded pellets.

6.1 Compounds extracted by four different fibres from ground flake (x

indicates assignment and n/a = “not analysed” due to the inclusion of a

solvent delay time).

6.2 Area of benzene and limonene after reducing the fibre exposure time

from 30 minutes to 5 minutes.

6.3 Comparison of concentrations determined in flake and pellets by

Soxhlet

and static headspace analysis. Standard deviations are in parentheses. All

values are in ppb.

6.4 Concentrations (ppb) of three compounds determined by static headspace

but not Soxhlet. Standard deviations are in parentheses.

Abbreviations

ABBREVIATIONS

A list of abbreviations for words not defined in the main body of the thesis.

MDPE medium density polyethylene

LDPE low density polyethylene

LLPE linear low density polyethylene

PP polypropylene

PE polyethylene

PS polystyrene

HDPE high density polyethylene

PVC polyvinyl chloride

PMMA poly(methyl methacrylate)

MEG monoethylene glycol

DEG diethylene glycol

GPC gel permeation chromatography

TLC thin layer chromatography

HPLC high performance liquid chromatography

SFC supercritical fluid chromatography

MAE microwave accelerated extraction

SFE supercritical fluid extraction

ASE accelerated solvent extraction

HPLC-UV high performance liquid chromatography-

ultra violet detection

SEC size exclusion chromatography

GC/MS gas chromatography/mass spectrometry

GC/FID gas chromatography/flame ionisation

detection

BHT butylated hydroxy toluene

DEHP di-(2-ethyl hexyl) phthalate

DEP diethyl phthalate

DiOP diisooctyl phthalate

BEHA bis-(2-ethyl hexyl) adipate

xviii

Abbreviations

BHA 4-(1-methyl-1-phenylethyl)-phenol

Tinuvin P 2-(2’-hydroxy-5-

methylphenyl)benzotriazole

DiBP diisobutyl phthalate

DBP dibutyl phthalate

DOA dioctyl adipate

SEC-HPLC size exclusion chromatography-high

performance liquid chromatography

XRD X-ray Diffraction

SML Specific migration limit

xix

Abstract

ABSTRACT

The purpose of this thesis was to determine which contaminants were present in washed

and dried shredded poly(ethylene terephthalate) (PET, flake) obtained from curbside

collection and to determine whether their concentrations were above the US FDA

threshold of 215 ppb. Over thirty semi-volatile contaminants were extracted from the

treated flake by Soxhlet extraction using dichloromethane as a PET swelling solvent

and gas chromatography-mass spectroscopy for identification and quantification.

Soxhlet extraction of flake ground to 0-300 µm was effectively completed by 24 h,

whereas sonication reduced the extraction time to 3 h. In contrast Soxhlet extractions

on flake ground to a larger particle size range (>300-425 µm and >425-700 µm) were

completed within four hours, possibly due to less aggregation in the extraction thimble.

In the finely ground flake (0-300 µm) the levels of most contaminants were below 215

ppb, but six were not. Dodecanoic acid was present at about 1200 ppb, 2-butoxyethanol

was approximately 1000 ppb, limonene, benzophenone and methylsalicylate were above

800 ppb and 2-methylnaphthalene near 215 ppb. After analogous method development

the levels of all diffusible compounds in extruded PET pellets were below the threshold

of 215 ppb.

The Soxhlet extraction technique was validated by comparison with total dissolution by

TFA for two of the three particle size ranges obtained by grinding the PET flake (>300-

425 µm and >425-700 µm) and for the unground flake. Further validation was achieved

by the comparison of contaminant levels determined by total dissolution with TFA and

sonication with DCM using flake ground to the 0-300 µm size range. The levels of

contaminants were found to increase with decreasing particle size range, but XRD

measurements of degrees of crystallinity were similar for each PET particle size range,

thus showing that the differences in contaminant levels were not due to variable

percentages of the amorphous material from the tops and bottoms of shredded bottles,

relative to the amounts of crystalline PET from the mid-sections of the bottles. Hence it

was postulated that the variations in contaminant levels were due to selective grinding

of the more highly contaminated surfaces, whilst the larger particles incorporated the

less contaminated interior material.

xx

Abstract

The analysis of the more homogenous annealed (extruded) pellets indicated that

contaminant levels between the analogous particle size ranges were equivalent.

This observation validated our interpretation of the high levels of contaminants found in

finely ground flake being due to selective surface grinding where high levels are

expected.

When analysing volatiles, static headspace analysis was performed on flake and

extruded pellets due to the limitations surrounding SPME. External standardisation was

used as the method of quantification and the levels of toluene, undecane and p-xylene in

extruded pellets were found to be below 38 ppb and therefore within the 215 ppb FDA-

set threshold for flake and pellets.

xxi

Chapter 1

CHAPTER 1 INTRODUCTION

The accumulation of plastic waste in landfill together with the increasing market share

of plastic soft drink bottles has encouraged recycling industries around the world to

consider recycling post consumer PET (polyethylene terephthalate) for food contact

applications.

Although recycling addresses the environmental concerns regarding waste disposal,

there are serious health issues associated with the use of recycled polymers for soft

drink storage.

Due to the permeable nature of plastic, it is feared that recycled PET could contain

compounds sorbed during the initial use of plastic bottles. For example, the sorption of

flavour compounds during soft drink storage. More seriously, the polymer could be

contaminated with compounds sorbed during the consumer reuse of soft drink bottles

for storing automobile, household or garden chemicals.

These sorbed “post-consumer contaminants” could have the potential to re-migrate

from the recycled polymer into the beverage at concentrations detrimental to consumer

health.

To obtain accreditation for recycled bottle manufacture, recycling industries are

required to prove that the levels of post-consumer contaminants in their recycled PET

bottles are below the regulated thresholds that assure consumer safety. For example,

for recycled PET soft drink bottles to be granted food contact approval in Australia, the

cleansing efficiency of the recycling process must be such that the levels of

contaminants in the polymer falls below the US Food and Drug Administration (US

FDA) ‘threshold of regulation’ of 215 ppb (FDA 1992). If this condition is not

satisfied, the migrating level into soft drink simulant (10% ethanol) must be

demonstrated not to exceed 10 ppb (FDA 1992, Komolprasert et al. 1997, Begley

1997).

When monitoring the cleansing efficiency of a recycling process, researchers generally

adopt the “challenge test” approach specified by the US FDA (FDA 1992,

Komolprasert et al. 1997, Franz et al. 1998). This procedure involves deliberately

contaminating a PET batch with exaggerated levels of surrogate contaminants and then

analysing the decontaminating capability of the recycling process. Our co-workers

1

Chapter 1

Harding et al. (unpublished) have found that the level of surrogates remaining in PET

after recycling is sometimes above the 215 ppb threshold. Subsequent migration tests

into food simulants proved that – according to the US FDA definitions - the PET did

not pose a health risk when used as a food grade plastic (Cross et al. unpublished). In

order to ascertain whether the levels of real contaminants in recycled PET fall below

the FDA threshold, their analysis in treated post-consumer PET was instigated. Hence

the general aim of this thesis was to determine what volatile and semi-volatile

contaminants are present in post-consumer PET and whether the identified

contaminants exceed the 215 ppb “threshold of regulation” set by the US FDA in order

to satisfy food safety requirements.

2

Chapter 2

CHAPTER 2 LITERATURE REVIEW

2.1 BACKGROUND

2.1.1 Definition of PET and its applications in the food industry PET (polyethylene terephthalate) resin is a polyester polymer produced by the

reaction of ethylene glycol with either terephthalic acid or dimethyl terephthalate and

in the presence of catalysts including salts of manganese, cobalt, cadmium, calcium,

lead, zinc, antimony, titanium and germanium (Figure 2.1). Its manufacture involves

several steps, which are described in greater detail in Section 2.2.6.

Polyesters, such as PET, are produced worldwide by companies such as Du Pont,

Eastman, Monsanto, and Hoechst and are used in the manufacture of textile fibres,

film, bottles and molding compounds (Ulrich, 1993). In 1999, over 5 million tonnes

of PET was processed worldwide for these applications and the trend has been

growing due to the polymer’s superiority over glass for food packaging (Kosmidis et

al. 2001).

Figure 2.1: Formation of PET (polyethylene terephthalate).

+

HO(CH2)2OH +

O O

C OCO

terephthalic acid ethylene glycol

H2O

CH2CH2

O

CH2CH2

O

C OCO

O O

C OCOCH2CH2

PET

HH

catalysts

1

Chapter 2

The use of PET as a film and bottle has been successfully incorporated into the food-

packaging industry due to the polymer’s strength, light-weight, flexibility, clarity,

resistance to high temperature and its negligible permeability to carbon dioxide,

which is especially important in the packaging of carbonated soft drinks (Begley and

Hollifield 1990a, Ashby 1988).

Bottles, for storing soft drinks, mineral waters, edible oils, wines, fruit juices and

spirits, are one of the main uses of PET involving direct contact with foods (Ashby

1988). PET bottles are also used in non-food contact applications, such as storage for

toiletries, cosmetics and other household products (Ulrich 1993).

As it will not thermally deform below 220ºC, PET is used as a film for roasting bags

and containers for reheating, cooking and crisping food. Metallised PET film

(thermally conducting susceptor film) is used in microwave cooking for browning

applications such as pastries, potato fries and popcorn. In non-food contact

applications, PET film is used for X-ray and other photographic film, magnetic tape

and electrical insulation.

2.1.2 Manufacture of PET bottles

The manufacture of PET bottles involves two parts: injection moulding of the PET

resin and blow moulding of the resultant preforms. These steps can be performed

simultaneously as a one-stage process or separately as a two-stage process.

Injection moulding involves melting PET under vacuum and injecting the melt into

mould cavities. Rapid cooling then takes place and the preform, which possesses a

test-tube like form, is ejected (Pó et al. 1995).

Blow moulding involves heating and mechanically stretching the preform in its axial

direction and finally blow moulding it into the bottle shape using hot air (Pó et al.

1995).

During stretching and blow moulding, polymer chains align themselves closely in the

direction of the stretch, improving the gas barrier and mechanical properties (e.g.

tensile strength, Young’s modulus, elongation at break) of the bottle (Mc Evoy et al.

1998). The top and bottom of the bottle are amorphous, like the preform, whilst the

2

Chapter 2

mid-section is crystalline and biaxially oriented, resulting from the blow-moulding

stage of the bottle manufacture (Sadler et al. 1996, Nielsen 1994).

2.1.3 Improving gas barrier properties of PET

In the case of food packaging, oxygen and carbon dioxide permeation needs to be

minimised in order to prevent carbon dioxide loss during carbonated soft drink

storage, and oxygen entry, which can lead to bacterial spoilage, off-flavours and/or

colour change in the food/beverage.

Apart from the obvious changes to molecular orientation and crystallinity, the

addition of co-monomers during the formation of the PET resin can also improve

many of the final physical properties of the polymer. For example, the addition of

isophthalic acid, cyclohexane dimethanol, diethylene glycol, or 2,6-napthalene

dicarboxylic acid modifies the polymer’s crystallinity, its melt temperature and glass

transition temperature, its melt strength and melt viscosity, its tensile strength and

elasticity modulus and its gas permeability (Schumann and Thiele 1996).

In the beverage industry, PEN (polyethylene naphthalate)-PET co-polyesters and

blends are becomming popular since PEN (Figure 2.2) enhances the gas barrier

properties of PET. The improved gas barrier performance of PEN derives from the

double ring structure of the naphthalene molecule, which increases the intermolecular

bonds between polymer chains. These stronger intermolecular bonds give rise to

further improved properties of PEN over PET, which are shown in Table 2.1

(Schumann and Thiele 1996). The higher glass transition temperature and thermal

resistance of PEN makes this polymer extra suitable for use as oven containers.The

greater cost of dimethyl naphthalene dicarboxylic acid (required for PEN) compared

to dimethyl terephthalate (required for PET) limits the commercial use of the PEN

homopolymer (Pó et al. 1995). However by combining the economics of PET with

the superior properties of PEN, the container’s mechanical requirements are met at a

lower cost in relation to using PEN as a homopolymer.

Figure 2.2: Structure of PEN.

3

Chapter 2

O

OCH2CH2 CO

C O

O

C

O

OCH2CH2O C

CH2CH2

The gas permeation of PET bottles can be further reduced by a multilayer technique,

whereby a superior barrier material is sandwiched between two PET layers (e.g. the

use of nylon in PET beverage bottles). The presence of polyamides (PA), ethylene

vinyl alcohol (EVOH) or active oxygen absorbents in the centre layer will reduce the

level of gas permeation of the polymer bottle. PA and EVOH are sensitive to

moisture, therefore are protected from the aqueous environment by two outer layers of

PET (Feigenbaum et al. 1993).

An alternative method is to coat the PET walls with siloxane, epoxy resins or

amorphous carbon (e.g. the coating of PET beer bottles with an epoxy amine resin).

Table 2.1: Comparative properties of PET versus PEN.

Property PET PEN

Glass transition

temperature

69°C 113°C

Thermal resistance 120°C 155°C

Oligomer Extraction 15 mg/m2hr 2 mg/m2hr

Resistance to hydrolysis 50 hr 200 hr

Resistance to irradiation 2 MGY 11 MGY

Breakdown voltage 400 V/µm 400 V/µm

Tensile strength 45 kg/mm2 80 kg/mm2

Young’s modulus 1200 kg/mm2 1800 kg/mm2

CO2 permeation 16 [cm3mm/(m2d.bar)] 2 [cm3mm/(m2d.bar)]

O2 permeation 4 [cm3mm/(m2d.bar)] 0.5 [cm3mm/(m2d.bar)]

The addition of “barrier” additives to PET raises concerns during recycling, which

requires that the PET be of reasonable purity. The presence of “barrier” polymers in

post-consumer PET feed may have a detrimental effect on the final properties of

4

Chapter 2

recycled PET; therefore it is desirable that their levels are kept at a minimum and

seldom used.

PET recycling and the extent of its purification with respect to post-consumer

contamination will form the basis of this thesis in the area of food safety.

2.2 FOOD CONTACT CONSIDERATIONS FOR VIRGIN PET

2.2.1 Introduction

In the area of food-contact polymers there are three polymer-food interaction

mechanisms which could potentially affect the integrity of the food and/or polymer.

As already discussed, one of these mechanisms is the permeation of gases,

specifically carbon dioxide and oxygen through the package. The other two forms of

interaction, schematically illustrated in Figure 2.3, are:

• The migration of substances from the package into the food.

• The sorption of substances from the food into the package.

Figure 2.3: Drawing of the bottom part of a PET soft drink bottle illustrating sorption,

migration and permeation.

Yellow spheres: Aroma compounds

Green spheres: Migrants of polymer

Soft drink

PET bottle wall

Purple spheres: penetrants e.g. oxygen and carbon dioxide

Sorption

Re-migration

Permeation

5

Chapter 2

Before PET is used in food contact applications, research must be undertaken

reinforcing the safety and suitability of the polymer as an item of food packaging.

This worldwide obligation erupted with the realisation that although plastics could be

used towards protecting food from bacterial and environmental contamination, they

themselves could also represent a source of contamination to the food by means of

migrating polymer constituents (Tice and McGuinness 1987). Additionally, unlike

glass, polymers can act as a sink removing some of the essential constituents of the

food (Paik 1992, Tavss et al. 1988, Gavara et al. 1997). Flavour constituents

(myrcene and limonene) have been shown to remain in PET and PC after washing and

this was confirmed by a strong smell of orange from the plastic bottles (Nielsen

1994). Such data not only indicates a potential problem of the flavour being lost on

storage, but also indicates the possibility of carry-over of flavour from one product to

another if the PET were to be re-used, a topic which will be discussed in detail

shortly.

2.2.2 Sorption considerations in food contact applications

One of the polymer-matrix interaction mechanisms, which could have an impact on

the integrity of the contacting media, is the sorption or “scalping” of food components

by polymers. For example, flavour compounds (e.g. terpinenes, esters, aldehydes and

alcohols) from orange juice or soft drink could sorb into the PET plastic in turn

affecting the organoleptic properties of the beverage, which is perceived as a loss in

fresh-like quality (Imai et al. 1990).

In addition, the sorbed components could act as migratory contaminants if the bottle is

recycled or reused for food contact applications (Nielsen 1994), which in turn could

have an impact on the taste and smell of the receiving medium. In subsequent

sections we will elaborate on this point by addressing the issues associated with

contaminants in recycled PET, arising from previous use and consumer abuse,

remigrating into soft drink.

Apart from affecting the organoleptic properties of food, flavour sorption has also

been shown to have an influence on the gas barrier (van Willage et al. 2002, Sadler

and Braddock 1991, Mannheim et al. 1987) and mechanical properties (Tawfik et al.

1998) of polymers. Fortunately, PET has the advantage of a low sorption capacity

6

Chapter 2

compared to polyolefins, PC (polycarbonate) and EVOH making it more suitable for

use as a food packaging (Nielsen et al. 1992, Nielsen 1994, van Willige et al. 2002,

Imai et al. 1990, Gavara et al. 1997). However, previous authors have concluded that

in spite of the very small extent of sorption, the physical properties of PET (Tawfik et

al. 1998), and the organoleptic properties (which are those associated with the five

senses) of the refill when the polymer is reused (Nielsen 1994), could still be affected.

Where gas barrier properties are concerned, van Willage et al. (2002) showed that

oxygen permeability in PET was not significantly affected by absorption of flavour

compounds. However, rubbery polymers with low glass transition temperatures (Tg)

such as LDPE and PP swelled in the presence of flavour compounds, increasing the

free volume and thus the oxygen permeability.

Figure 2.4: A summary of the disadvantages of “flavour scalping”.

Sorbed compounds could migrate during polymer reuse

Sorption could alter gas barrier and mechanical properties of PET

Disadvantages of sorption

Sorption could affect the organoleptic integrity of the food

2.2.3 Factors contributing to the degree of sorption

The degree of sorption is dependent upon sorbate shape, size, polarity and

concentration of the sorbate (Sadler and Braddock 1991, Brody 1989, Nielsen 1991a,

Shimoda 1988, Gavara et al. 1997, Reynier et al. 2001) as well as polymer

morphological and chemical characteristics such as polarity, crystallinity, axial

orientation, cohesive energy density, packing of polymer chains, degree of cross-

7

Chapter 2

linking and glass transition temperature (Ackermann et al.1995, Brody 1989,

Shimoda 1988, Charara et al. 1992, Fayoux et al. 1997, Gavara et al. 1997, Miltz et

al. 1997, Nielsen et al. 1992). Physical properties of the polymer (e.g. thickness and

surface area) and external factors such as temperature and storage time also influence

the amount of sorbate absorbed.

Figure 2.5: Sorbate, polymer and external factors effecting sorption in PET.

Sorbate factors Polymer factors

Temperature Time pH

Co-sorbate

CO

Cohesi

Glass tra

Shape Size

Polarity Concentration

External factors

Factors effecting sorption in PET

2.2.3.1 Properties of sorbate

Since the diffusion coefficient (D) o

sorption is generally expected to d

(Limm and Hollifield 1996, Nir et a

which is diffusion controlled. In ter

crawl through the polymer matrix, d

have been reported to move by s

Feigenbaum et al. 1993). A general

compounds with aromatic structure

sorbed into PET more readily than o

heptane, phenyl cyclohexane). There

Polarity rystallinity rientation

ve energy density Packing

nsition temperature

f a molecule is inversely proportional to its size,

ecrease with increasing diameter of a sorbate

l.1996, Begley et al. 2002), at least for sorption

ms of molecular shape, linear molecules, which

iffuse faster than spherical molecules. The latter

lower sequential jumps (Reynier et al. 2001,

observation made by Franz et al. (1997) was that

(e.g. phenol, toluene and chlorobenzene) were

ther non-aromatic compounds (e.g. limonene, n-

are exceptions to the general rule that sorption

8

Chapter 2

decreases with increasing molecular size. Discrepancies in the relationship between

sorption and molecular size occur when molecular polarity plays a more important

role in sorption than sorbate volume. For example, Shimoda et al. (1988), Ikegami et

al. (1991) and Nielsen et al. (1992) observed that as the carbon number in the straight

chain of esters, aldehydes and/or alcohols increased, the degree of sorption also

increased until a certain carbon number is reached, beyond which sorption decreases.

The increase in sorption was explained in terms of polarity; the longer the chain, the

less polar the compounds and the easier the compounds are sorbed into non-polar

polymers (Nielsen et al. 1992). Shimoda et al. (1988) explained the subsequent

decline in sorption beyond an optimum carbon number in terms of steric hindrance.

From the results of these studies it can be concluded that a balance exists between the

effects of sorbate polarity and steric hindrance with increasing carbon number.

The importance of sorbate polarity on the degree of sorption has been demonstrated in

other cases. For example, it is argued that the lipophilic nature of limonene makes it

more inclined to sorb into polyolefins than into PET (Kwapong and Hotchkiss 1987).

In addition, polar volatiles such as short chain aldehydes and alcohols are sorbed into

polyolefins to a lesser extent than non-polar hydrocarbon compounds such as

limonene (Sadler and Braddock 1991, Charara et al. 1992). A similar dependence on

sorbate polarity was confirmed by Shimoda et al. (1988) who made the observation

that esters were sorbed into LDPE to a greater extent than aldehydes, which in turn

were sorbed more than alcohols. This sorption order was moreover observed for

MDPE but not for EVOH, which is hydrophilic and therefore more likely to sorb the

polar compounds (alcohols and aldehydes) than the terpene hydrocarbons (Ikegami et

al. 1991).

2.2.3.2 Polymer chemical and morphological properties

As mentioned earlier, diffusion is not only influenced by sorbate characteristics but

also by polymer character. A container that is crystalline, biaxially orientated, and

has both a high transition temperature and cohesive energy density (strong

intermolecular bonds) impedes diffusion due to the lack of free volume (“holes”) in

the polymer. The favourable cohesive energy, polarity and intermolecular packing of

PET provide a degree of resistance to sorption (or migration) whilst acting as a barrier

9

Chapter 2

to diffusion (Gavara et al. 1997, Arora and Halek 1994). In addition, the high glass

transition temperature of PET hampers the free vibration and rotational motion of

PET at room temperature, affecting the diffusion of sorbates (van Willige et al. 2002,

Paik 1992). By increasing the degree of crystallinity and orientation in PET the size,

shape and distribution of microcavities in PET will decrease, further obstructing the

path to diffusion (Nir et al. 1996, Miltz et al. 1997). It has been suggested that

diffusion of analytes only occurs in the amorphous regions of PET (Fayoux et al.

1997, Charara et al. 1992), therefore sorption is expected to occur most in the top and

bottom of a PET soft drink bottle, which is more amorphous than its biaxially

orientated mid-section (Sadler et al. 1996, Nielsen 1994, Jetten et al. 1999). Begley

et al. (2002) demonstrated the effect of polymer orientation and crystallinity on the

sorption of lindane into PET. It was observed that nine times less sorption occurred

into orientated and crystalline PET than into amorphous PET.

2.2.3.3 Solubility parameter

The difference in the extent of sorption of different types of compounds into a

polymer can be predicted by the similarities in Hildebrand solubility parameter (δ), or

cohesive energy density, between the sorbate and polymer. However, the effects of

hydrogen bonding and polarity as well as sorbate shape and size must also be taken

into account since these are sometimes ignored, or poorly treated, when calculating

solubility parameters (Nielsen et al. 1992, Konczal et al. 1992, Arora and Halek

1994). As shown by Paik (1992) the solubility parameter difference does not predict

well the relative solubility of more polar compounds, as can be the case with PET.

Equation 2.1 is used to calculate the solubility parameter (δ) for a substance where Le

is the molar latent heat of vaporisation of the liquid and V is the molar volume, both

at absolute temperature, T. The dimensions are (cal/cm2) ½, also called 1 Hildebrand

(1H) for convenience. If the Le value is not known it is calculated from Small’s molar

attraction constants, which could be obtained from the CRC Handbook of Chemistry

and Physics (Weast and Melvin, 1979).

δ = [(Le – RT)/V]1/2 Equation 2.1The difference in solubility parameters

between two components is described as the “heat of mixing”. The lower the heat of

10

Chapter 2

mixing (or difference in solubility parameters) the higher the solubility of a sorbate in

the polymer. This interaction leads to the swelling (or plasticisation) of a polymer,

which can increase its oxygen permeability (van Willige et al. 2002) or facilitate the

extraction of polymeric components. The theory of solvation has been applied to the

solvent extraction of migratory constituents out of polymers intended for food contact

applications (Vandenburg et al. 1999, Feigenbaum et al. 2002) and will be addressed

in subsequent sections of the current thesis.

Polymer swelling opens up the polymer structure and therefore facilitate sorption of

other components from the contacting solution. The effect of acetone, a PET-swelling

solvent, on the penetration of lindane has been already studied (Begley et al. 2002).

There have been a few studies undertaken explaining sorption results in terms of

differences in solubility parameter.

As demonstrated by Nielsen et al. (1992) PET absorbed smaller quantities of esters

and aldehydes than polyolefins (i.e. LDPE, LLDPE and PP) as a result of the large

solubility differences between the sorbates and PET. Although the solubility

parameters for alcohols matched those of PET, sorption was negligible probably

because of the hydrogen bonding differences between the alcohols and PET. In the

latter study, differences in the amount sorbed between the polyolefins resulted from

morphological differences between the different polymers, such as variations in

crystallinity and interchain packing. In terms of interchain packing, it was observed

that the longer the branches on a polymer the smaller the packing density and

therefore the greater the extent in sorption, as for LDPE.

In another study, Arora and Halek (1994) established that PET sorbed much more

fatty acids than PP. The opposite was true for the triglycerides, which were sorbed

into PP to a greater extent than into PET. The author attributed these observations to

similarities in solubility parameters as well as morphological effects such as

crystallinity and interchain packing enforced by the zig-zag planar structure of PET.

PP has a helical structure that does not pack as efficiently. These two studies

highlight the importance of considering factors such as polymer morphology and

sorbate hydrogen bonding as well as solubility parameters when interpreting the

extent of sorption into a polymer.

11

Chapter 2

2.2.3.4 Polymer physical properties

Other than polymer morphological and chemical properties, the polymer physical

properties such as surface area and film thickness also have an impact on the degree

of sorption (Ikegami 1991, Shimoda 1988, Nielsen 1994).

Ikegami (1991) and Shimoda (1988) both established that an increase in polymer

thickness caused a rise in the amount of compound sorbed. This result is expected

given that film thickness describes the sorbate capacity of a polymer, but only if

enough contact time has been involved such that capacity limits the extent of sorption

and not kinetic factors.

Nielsen (1994) observed the effects of increasing polymer surface area on the sorption

of limonene and myrcene into PET. With increasing polymer surface area, sorbate

diffusion occurs from more directions, resulting in a greater degree of sorption. An

increase in sorption with surface area, results in an improved rate of migration or

extraction. In this case, kinetic factors presumably allow for a greater sorption even

though the sorption capacity (polymer thickness) is the same. Alternatively, if

sorption is really an adsorption, rather than absorption, process then an increase in

surface area will genuinely increase the adsorption capacity [Harding and Healy

1979]. The effect of polymer surface area on extraction will be discussed in

subsequent chapters.

2.2.3.5 Temperature

External factors that influence sorption are time, temperature, pH and the presence of

co-sorbents (Fayoux et al. 1997).

Nielsen (1994) found that three times more limonene was sorbed into PET at 25°C

than at 4°C. Kwapong and Hotchkiss (1987) also observed that the sorption of

limonene into LDPE increased with temperature. This effect has been observed for a

number of adsorbates, for example benzene sorption into PET (Sadler et al. 1996).

The increase in sorption with temperature is attributed to a different equilibrium

constant and/or a faster diffusion process at higher temperatures. The reason behind

the latter stems from the fact that the diffusion coefficient (D) is exponentially related

to temperature (T) as indicated by the Arrhenius equation (Equation 2.2) (Cotton et al.

1993). The diffusion coefficient (whose units are cm2/s) of a sorbate/migrant depends

12

Chapter 2

on two factors controlled by temperature (whose units are K) - the vibrational motions

of the (a) polymer chains and (b) sorbates.

In the Arrhenius equation, D0 (cm2/s) is a constant related to the entropy of

activationand E (kJ/mol) is the activation energy of diffusion, which relates to the

energy required to make an opening between polymer chains large enough to allow a

sorbate molecule to pass through.

D = D0 exp (-E/RT) Equation 2.2

Contrary to the results obtained in liquid phase, sorption of molecules in the vapour

phase has been shown to decrease with temperature illustrating a different sorption

mechanism (Fayoux 1997, Sadler et al. 1996).

2.2.3.6 Time

The relationship between the extent of sorption, and/or of migration, and time is

described by equations 2.3 and 2.4, where Mt is the amount sorbed (or migrated) at

time t (g/cm2), M∞ is the amount sorbed at equilibrium (g/cm2), D is the diffusion

coefficient (cm2/s) and l is the thickness of the film (cm).

Equation 2.3 represents short-term migration/sorption well before the saturation level

is reached (for Mt/M∞ < 0.6) whilst Equation 2.4 represents long-term

migration/sorption (for Mt/M∞ > 0.6). In these equations, the rate is assumed to be

governed by the local migrant gradients and therefore the rate-controlling step is

diffusion of analytes through the polymer from high concentration to low

concentration (Fick’s law of diffusion).

If Fick’s law of diffusion is obeyed, a straight line is generated when Mt/M∞ is plotted

against t1/2/L (short-term migration/sorption) or when ln (1- Mt/M∞) is plotted against

t/L2 (long-term migration/sorption). The slope of the straight line, from which the

diffusion coefficient can be derived, is represented by 4(D/π)½ for short-term

sorption/migration and π2D for long-term sorption/migration.

Mt/M∞ = 4 (Dt/πL2) ½ Equation 2.3

13

Chapter 2

Mt/M∞ = (1-8/π2) exp (-π2Dt/L2) Equation 2.4

A plot of the natural log of the calculated diffusion coefficients versus the inverse of

temperature has been shown to result in a straight line, suggesting an Arrhenius

relationship between both variables for the migration (rather than sorption) of

antioxidants from polyolefins into fatty simulants (Lickly et al. 1990). A similar

Arrhenius plot was constructed for the sorption of acetone (Shan and Tsu-Shang

1999) and the sorption of benzene into PET (Patton et al. 1984). Deviation from

linearity can occur. For example, glassy polymers above their glass transition

temperature do not give a linear Arrhenius plot. This is because the activation energy,

which is given by the slope of the Arrhenius curve, does not vary in a linear fashion

with temperature for glassy polymers above their glass transition temperature (Begley

and Hollifield 1990a).

Fick’s law of diffusion is obeyed for amorphous rubbery polymers in the absence of

swelling. Diffusion kinetics for PET is not expected to follow Fick’s law of diffusion

because it is a glassy polymer. In fact, past researchers have observed a pseudo-

Fickian behaviour for dichloromethane sorption (Liu and Neogi, 1992) and other

organic solvent sorption into PET (Nir et al. 1996). The swelling stresses created by

the solvent penetration are thought to have contributed to a deviation from Fickian

behaviour. For pseudo-Fickian diffusion the “½” in Equation 2.3 is substituted by a

smaller fraction (< ½) whereas for non-Fickian diffusion it is replaced by “1”.

2.2.4 Migration considerations in food contact applications

Previous research has shown the presence of low molecular weight components in

virgin PET and other polymers that result from the polymers' original manufacture

(Monteiro et al. 1996, Kim et al. 1990, Costley et al. 1997, Ezquerro et al. 2003).

Many of the residual compounds could have high diffusibilities and thus the potential

to migrate into food by passing randomly through microscopic voids created by the

movement of polymer chains (Kashtock and Breder 1980, Startin et al. 1987, Morelli-

Cardoso et al. 1997, McNeal and Hollified 1993, Castle et al. 1996, Tawfik and

Huyghebaert 1988).

14

Chapter 2

The contamination of food by migratory components in PET raises health concerns

and/or could affect the food’s organoleptic properties. Even if these problems are

eventually shown not to create a significant health risk they could, if found to be true,

result in diminished sales due to negative consumer perception (Ackermann et

al.1995, Ezquerro et al. 2003). It is therefore desirable to monitor the safety of PET

by monitoring what contaminants can be extracted or migrated out of it. The former

involves quantitative analysis of contaminants present in the PET. The latter involves

contacting the polymer with a food, or food simulant, determining the quantity of

components migrating from the plastic and deciding whether the thresholds enforced

by the FDA or EEC are exceeded. Such tests are not quantitative measures of the

contaminants in PET but are realistic measures of the amount of contamination which

using PET might result in. In order to simplify the migration experiment in terms of

extraction and instrumental analysis, food simulants that reproduce migration are

usually used rather than foods themselves (Table 2.2 and Table 2.3 show a

comprehensive list of recommended food simulants). The storage conditions for

migration testing using these simulants are 40°C /10 days (ECC) and 49°C /10 days

(FDA). Alternative food simulants to those recommended by the FDA and ECC can

be used, provided migration tests are performed demonstrating their ability to

reproduce the food of interest (Baner et al. 1994a, Hamdani and Feigenbaum 1996,

Tawfik and Huyghebaert 1998). As demonstrated by Riquet and Feigenbaum (1997),

it is also possible to tailor the aggressivity of food simulants, using mixtures of

solvents.

Table 2.2: A list of FDA food simulants.

Food Type Recommended Food Simulating Solvent Aqueous and acidic foods

10% Ethanol (in specific applications water and 3% acetic acid simulant)

Low and high alcoholic foods 10 or 50% Ethanol

Fatty foods Food oil (e.g. corn oil), HB307, or Miglyol 812™1

1 HB307 is a mixture of synthetic triglycerides (primarily C10, C12,C14). Miglyol 812™ is derived from coconut oil, which also consists of triglycerides (C8,C10).

15

Chapter 2

Table 2.3: A list of EEC food simulants.

Food Type Recommended Food Simulating Solvent Aqueous foods with pH >4.5

Distilled water

Aqueous foods with pH < 4.5 3% acetic acid in water

Alcoholic beverages of alcoholic strength equal to or exceeding 5% volume

15% ethanol or 50%

Fatty foods Food oil (e.g. corn oil), HB307

2.2.5 Factors affecting the extent of migration

As with sorption, the extent of migration into food/simulant depends on external

factors (e.g. temperature, time and interactivity of matrix with the polymer) as well as

polymer characteristics (e.g. crystallinity, thickness and glass transition temperature)

and migrant factors (e.g. concentration, size, shape, polarity and solubility in the

food/simulant).

External factors

Migrating ability increases with temperature, time and in the presence of “aggressive”

or polymer-interactive foods, which could act as migrant extractants by swelling the

polymer (Ashby 1988). For example, Tawfik and Huyghebaert (1998) showed that

there was a relationship between the extent of styrene migration from polystyrene

cups into milk and the fat content of that milk (milk fat is a polymer interactive).

These authors also concluded that the level of migration depends on storage

temperature and time. Likewise, Snyder and Breder (1985) monitored the migration

of styrene from polystyrene into various solvents (potential food simulants), with

respect to time. The results were modelled using the standard migration equation

(Equation 2.3) and the diffusion coefficient was determined for migration into each

solvent. Solvent penetration by the more aggressive solvents (e.g. decanol and 50%

ethanol) contributed to an increase in styrene migration. This solvent penetration was

more pronounced at a higher temperature, (70°C compared to 40°C) and therefore so

was the migration of styrene. Since the migration equation is theoretically

inapplicable during solvent penetration, the diffusion coefficients calculated were

16

Chapter 2

thought of as “effective diffusion coefficients”. The most suitable fatty food simulant

at 70°C was 8% ethanol (decanol, 50% ethanol and 20% ethanol were too

aggressive).Lickly et al. (1990) studied the migration of an antioxidant from HDPE

and PP into a series of potential fatty food simulants and demonstrated that an

Arrhenius-type relationship existed between the diffusion coefficient and temperature

for the migration of antioxidant. Similarly, Goydan et al. (1990) found an Arrhenius

correlation between diffusion coefficient and temperature whilst studying the

migration of antioxidants from HDPE and LDPE into 95% ethanol, corn oil and

aqueous simulants (water and 8% ethanol). The amount migrating into the fatty

simulants (95% ethanol and corn oil) was higher than into aqueous simulants.Begley

and Hollifield (1990a) found an increase in the migration of cyclic trimer from PET

microwave susceptor trays into corn oil with increasing temperature however

concluded that the Arrhenius plot was non-linear above the PET glass transition

temperature (see Figure 2.10 for a definition of what is meant by "cyclic" trimer).

Baner et al. (1994a) and Vijayalakshmi et al. (1999) confirmed the kinetic "rule of

thumb" that the kinetic rate (proportional to the diffusion coefficient) doubles for

every 10°C increase in temperature for the migration of additives from polymers,

including PET. This indicates an Arrhenius type relationship between temperature

and diffusion coefficient for temperatures below 70°C.

Polymer and migrant factors

Other factors that influence the amount of migrant entering the food phase are

polymer crystallinity, thickness, and glass transition temperature of the polymer, as

well as migrant concentration, size, polarity and solubility in the food simulant of the

migrant (usually a contaminant). Miltz (1998) established that the diffusion

coefficient for the migration of toluene from PET into water increased with initial

toluene concentration. Hamdani et al. (1997) further demonstrated the mathematical

relationship describing the dependence of migration on migrant concentration in the

polymer (Equation 2.5). Equation 2.5, which directly relates migration to the initial

concentration in the polymer, is a derivation of Equation 2.3.

Mf,t/A = 2Cp,0(Dt/π)1/2 Equation 2.5

Mf,t = Amount of substance migrating into food at time t (g)

17

Chapter 2

A = Material’s food contact surface area (cm2)

Cp,0 = Initial concentration of migrant in the polymer (g/cm3)

D = Diffusion coefficient of the migrant in the polymer (cm2/s)

Castle et al. (1988) suggested the use of high molecular weight plasticisers to reduce

their migration into food, implying a dependence of molecular size on migration.

Baner et al. (1994b) modelled this effect showing, that diffusion decreased with

increasing molecular weight. A formula relating the diffusion coefficient of a migrant

to its molecular weight was developed in order to predict the extent of migration at

any given temperature and time (Equation 2.6). Begley and Hollifield (1990a)

observed that the percent migration of cyclic trimer oligomer migrating into corn oil

exceeded that of the cyclic tetramer which, in turn, exceed that of the cyclic hexamer.

This result is almost certainly attributable to differences in molecular size.

D ≤ 10,000 exp (A - a M – b 1/T) [cm2/s] Equation 2.6

D = diffusion coefficient in the polymer [cm2/s]

M = molecular weight of migrant

T = temperature in K

A = dimensionless polymer specific constant

a and b = constants

Hamdani and Feigenbaum (1996) and Feron et al. (1994) have shown that the polarity

and solubility of the migrant and therefore its affinity for the polymer-contacting

medium is a factor governing migration. For example, Hamdani and Feigenbaum

(1996) found that TEHTM [tris(2-ethylhexyl)trimellitate] migrates more into

isooctane and oil, but less into ethanol presumably because TEHTM prefers the more

non-polar solvents. The opposite was observed for ESBO (epoxidised soybean oil),

which contains polar groups and therefore migrates more into ethanol than into

isooctane and oil. Similarly, Devlieghere et al. (1998) observed that global (sum)

migration into fatty food simulants was higher than that into aqueous simulants. This

observation was attributed to the higher solubility of organic migrants into a fatty

medium compared with an aqueous one.

18

Chapter 2

Feron et al. (1994) suggested that the low solubility of lindane could be the limiting

factor for its remigration from the bottle into the soft drink simulant.

Riquet and Feigenbaum (1997) discussed the effects of hydrogen bonding on the

migration of amino TEMPO out of PVC containing ESBO, which interacts with the

migrant through hydrogen bonding. To encourage migration, a simulant is used that

competes for interaction with the migrant and displaces ESBO.

Ashby (1988) presented the effects of increasing film thickness and crystallinity on

the overall migration from PET into olive oil. Global migration was shown to

decrease with increasing crystallinity and orientation. Conversely, migration was

shown to increase in a linear fashion with increasing film thickness until a limiting

polymer thickness was reached beyond which the rate of migration became was

reduced. It was assumed that migration from thick samples was reduced because the

olive oil failed to penetrate the thicker sample. Baner et al. (1994a) also discussed

similar implications of film thickness on migration.

Begley and Hollifield (1990b) observed that crystalline PET trays exhibit lower

migration rates than paperboard PET trays and attributed this to differences in

polymer crystallinity.

Begley et al. (1995) concluded that migration was lower from a PET tray designed for

oven use than a nylon-roasting bag owing to the higher glass transition temperature of

PET.

2.2.6 Potential migrants resulting from the manufacture of PET

Components inherited during PET manufacture are the primary source of

contamination in PET food packaging situations, apart from

when the container is recycled or reused.

Potential migrants in virgin PET include compounds that are added to assist in

polymer formation or enhance the polymer’s final properties (e.g. additives, catalysts

and starting materials) and compounds that result from the extreme conditions of the

polymerisation process (e.g. monomers, oligomers, reaction bi-products, acetaldehyde

and additive breakdown products). In order to monitor the food safety of PET

containers, studies have been carried out to screen for components in the (virgin)

polymer which can migrate into its contacting food or food-simulant. From the

19

Chapter 2

results of these experiments and their comparison to set regulations, conclusions can

be drawn as to when, if at all, PET poses a threat to the consumer when used as a

food-grade plastic.

Residual compounds resulting from manufacture identified in PET

The manufacture of PET involves three steps, each having the potential for

introducing migratory components. The first stage involves the manufacture of

ethylene glycol, terephthalic acid and/or dimethyl terephthalate, all from crude oil

using catalysts, pressure and heat. In the case of the latter two compounds, p-xylene

from the naphtha fraction of crude oil is either oxidised to terephthalic acid or

oxidised and esterified (with methanol) to produce dimethyl terephthalate (Figure

2.6). Ethylene glycol is manufactured by oxidation of ethylene from the gas fraction

of crude oil to ethylene oxide (oxirane), which is subsequently hydrolysed with water.

The oxidation of ethene to oxirane takes place in the presence of a silver catalyst.

The potential migratory components resulting from this step are catalysts (cobalt-

manganese salt, silver); p-xylene and other components of crude oil; ethylene glycol;

terephthalic acid and/or dimethyl terephthalate; p-toluic acid; p-toluic acid methyl

ester; p-terephthalate; and monomethyl ester.

Figure 2.6: Formation of dimethyl terephthalate and terephthalic acid from xylene.

20

Chapter 2

CH3

CH3

COOH

COOH

CH3

COOH

CH3

COOCH3 COOCH3

COOH COOCH3

COOCH3

O2

O2

CH3OH HNO3 CH3OH

p-xylene

p-toluic acid p-toluic acidmethyl ester

p-terephthalatemonomethyl ester

dimethyl terephthalate

terephthalic acid

The presence of catalysts, p-xylene, ethylene glycol, terephthalic acid and dimethyl

terephthalate as residues in PET and as migrants in food/simulants is well

documented. For example, Freire et al. (1998) identified p-xylene, terephthalic acid,

dimethyl terephthalate and other volatile compounds in PET, including samples used

as multilayer films, bottles, susceptors and roasting bags. The levels of the volatiles

were concluded to be low, indicating no hazard to public health. Hillery et al. (1989)

identified the presence of terephthalic acid and ethylene glycol in a PET beverage

bottle, microwavable tray and two commercial resins. Both compounds were also

identified at acceptable levels in a commercial amber PET bottle wall (Kim et al.

1990). The migration of ethylene glycol from PET bottles into food simultant (3%

acetic acid) was also studied by Kashock and Breder (1980). The quoted levels of

ethylene glycol were 5 ppm and 0.1 ppm in the polymer and food simulant

respectively. The EEC regulation for the migration of ethylene glycol into food is 30

ppm, therefore this packaging is considered appropriate for food contact applications

in terms of ethylene glycol migration.

Morelli-Cardoso et al. (1997) performed a specific migration study concerning the

migration of ethylene glycol from PET bottles into aqueous food simulants (distilled

21

Chapter 2

water, 3% w/v aqueous acetic acid and 5% v/v aqueous ethanol. The levels of

ethylene glycol migrating were below the maximum method detection limit (2.2 ppm

for 3% aqueous acetic acid). Therefore, the PET bottles were concluded to be of a

suitable quality whilst demonstrating migration levels below the 30-ppm upper limit.

Morelli-Cardoso et al. (1997) performed a specific migration study concerning the

migration of ethylene glycol from PET bottles into aqueous food simulants (distilled

water, 3% w/v aqueous acetic acid and 5% v/v aqueous ethanol. The levels of

ethylene glycol migrating were below the maximum method detection limit (2.2 ppm

for 3% aqueous acetic acid). Therefore, the PET bottles were concluded to be of a

suitable quality whilst demonstrating migration levels below the 30-ppm upper limit.

Fordham et al. (1995) identified the presence of the cobalt metal ion amongst other

catalyst residues in PET and discovered that the level of this metal migrating into food

simulants was below the proposed EEC limit of 100 ppb. In fact, for cobalt, the levels

migrating into 3% acetic acid, 15% ethanol and olive oil were in the low ppt range.

The second stage of PET manufacture involves the formation of monomer BHET via

either the esterification of terephthalic acid with ethylene glycol or the

transesterification of dimethyl terephthalate with ethylene glycol (Figure 2.7) under

heat and high pressure. Catalysts are only used during transesterification and include

the acetates of calcium, manganese, cobalt, cadmium, lead or zinc. Since these

catalysts promote degradation of the polyester during polymerisation, phosphorous

compounds (inhibiting stabilisers) are later added to deactivate such unnecessary

function.

The potential migrants resulting from this step are the catalysts and any bis-(2-

hydroxyethyl)terephthalate (BHET) remaining after later steps [the latter has been

quantified in commercial PET by Begley and Hollifield (1989)]. Monohydroxy

ethylene terephthalic acid (MHET), a bi-product of this reaction, and terephthalic acid

were also quantified by these researches. The concentrations of terephthalic acid,

MHET and BHET in the beverage bottle were 6.9 ppm, 34.4 ppm and 49.1 ppm

respectively. The significance of these results with regards to potential human health

risks was not evaluated by Begley and Hollifield (1989). However, one can reason

that the level of terephthalic acid in the bottle wall does not threaten human health

because it is below the SML of 7.5 ppm. Since the food-contact regulations for

22

Chapter 2

MHET and BHET are not known, it is not known whether the levels of these

compounds in the bottle wall pose a health threat.

In 1990 Begley and Hollifield (Begley and Hollifield 1990b) determined the amount

of BHET migrating into oil from PET microwavable trays (0.046 µg/cm2) whilst

Shiono (1979) qualitatively identified this compound in commercial PET film.

Figure 2.7: Formation of bis-(2-hydroxyethyl)terephthalate (BHET) from dimethyl

terephthalate and ethylene glycol.

+ CH3OH2HOCH2CH2C OCO

O

CH2CH2OH

O

2 HO(CH2)2OH +CH3

O

CH3

O

C OCO

catalysts

The third, and final, stage of PET manufacture involves manufacture by melt

polymerisation (Figure 2.8) of the monomers followed by solid-state polymerisation.

The latter process is important in the production of beverage bottles since it involves

the vacuum and high temperature treatment of PET granules at low moisture and

oxygen levels. This treatment increases the molecular weight of the polymer and

removes migratory volatile reaction products such as acetaldehyde. Polymer

molecular weight, in addition to sidewall thickness, interchain bonding, crystallinity

and orientation are important factors that reduce the gas permeation of PET. An

added bonus of this procedure, however, is the removal of many volatile and semi-

volatile contaminants which would otherwise be significant migrants.

The catalysts used during melt polymerisation are usually substances of antimony,

germanium, titanium or lead. Ashby (1988) and Fordholm et al. (1995) studied the

migration of antimony and germanium from PET into food simulants and concluded a

low level of migration for these metals.

Figure 2.8: Formation of PET from BHET.

23

Chapter 2

O

CH2CH2OH

O

C OCO

CH2CH2 C OCO

OO

C OCO

O

CH2CH2

O

HO(CH2)OH+

CH2CH2

CH2CH2HO2

catalysts

2.2.7 Reaction by-products formed during PET manufacture

Compounds with a potential to migrate do not only include those directly involved in

the manufacture of PET, such as monomers and catalysts. Mobile molecules

originating from side reactions during polymer production could also migrate. In the

production of PET, by-products resulting from side reactions include acetaldehyde

and oligomers.

Acetaldehyde

Prolonged heat treatment during polymerisation results in the formation of carboxyl

and vinyl ester end-groups. A reaction between carboxyl and vinyl ester end-groups

generates vinyl alcohol, which exists as a tautomer with acetaldehyde (Figure

2.9).The reported organoleptic detection limits for acetaldehyde are very low ranging

from 4 to 65 ppb (Lorusso 1985), thus their presence in food contact containers such

as soft drink bottles raises particular concerns. For this reason acetaldehyde has been

analysed in PET bottles (Dong et al. 1980, Wyatt 1983, Franz and Welle 1999b) and

beverage simulants in contact with PET (Wyatt 1983, Lorusso et al. 1985, Ashby

1988, Trinh Vu-Duc 1998, Eberhartinger et al. 1990). Acetaldehyde levels in water

were found to be below the EEC standard value of 100 ppb (Lorusso et al. 1985,

Ashby 1988) when stored at 40°C for 10 days in PET bottles. However, the threshold

24

Chapter 2

was exceeded when stored for more than 30 days at this temperature (Lorusso et al.

1985). At room temperature, the latter sample was stable for up to six months.

Figure 2.9: Formation of acetaldehyde from PET.

+C OHC

O

CH2

+

O

CO

O

C OCH2CH2

OCH2CH2

CH3CHO

CHOHCH2

CH2CH2OHC

CH2

O

OHCC+C OH

Othermal degradation

C

O

CO

O

O

Oligomers in PET

Other reaction by-products formed during the manufacture of BHET and melt

polymerisation are oligomers. The analysis of oligomers in PET has been a long

existing, but highly problematic, objective. Extraction studies date back to 1954

(Hudgins et al. 1978), when the cyclic trimer (Figure 2.10) was extracted from PET

film. Since then most researchers have found that the major extractable original

component is the cyclic trimer (Goodman and Nesbitt 1960, Shiono 1979, St küppers

1992) although the "cyclic" dimer (Monteiro et al. 1996, Triantafyllou et al. 2002)

and higher molecular weight oligomers up to the cyclic decamer have also been

identified (Hudgins et al 1978, Begley and Hollifield 1989, Barnes et al 1995, Costley

et al. 1997). The presence of ether oligomer counterparts, which have one MEG unit

replaced by a DEG unit (Goodman and Nesbitt et al. 1960, Monteiro et al. 1996,

Triantafyllou et al. 2002), and linear oligomers (Hudgins et al. 1978, Begley and

Hollifield 1989) has also been reported. The overall level of oligomers in PET

generally does not exceed 3% (Goodman and Nesbitt 1960, Peebles et al. 1969,

Hudgins et al. 1978). Although the relative proportions of cyclic oligomers compared

with linear ones are not commonly quoted, it was observed from a liquid

25

Chapter 2

chromatogram produced by Begley and Hollifield (1989) that the areas of the peaks

representing the cyclic oligomers were larger than those for the linear oligomers.

In some of the early studies, separation of oligomers from the PET extract involved

fractional extraction (Goodman and Nesbitt 1960, Peebles et al. 1969). The

precipitated oligomers were then further separated by TLC and/or column

chromatography and identified by physical and chemical analysis. Nowadays, PET

extracts are separated and analysed more efficiently by GPC, HPLC and SFC, but not

generally by GC/MS, because of the low volatility of the higher molecular weight

oligomers (Dulio et al. 1995, Shiono 1979, Barnes et al. 1995). Oligomers have been

hydrolysed and methylated into dimethyl terephthalate prior to analysis by GC/MS

(Gramshaw et al. 1995).

Figure 2.10: Cyclic oligomers identified in PET.

C

C

O

OCH2CH2OC

C

O

OCH2CH2O

n

n=1 (dimer)

n=2 (trimer)

n=4 (pentamer)

O O

Extraction of oligomers from PET has also been simplified in terms of labour

intensity and solvent consumption, with the introduction of automated techniques

(MAE, SFE and ASE) to replace manual methods such as Soxhlet extraction and total

dissolution procedures often employed to extract oligomers from PET. Polymer

extraction methods will be discussed more specifically in subsequent sections.

26

Chapter 2

Oligomer migration from PET

Vacuum treatment during solid-state polymerisation decreases the amount of

oligomers present in PET by polymerising them into larger molecules. However

despite this polymerisation process, migration of oligomers into food and simulants

has been observed.

Begley et al. (1990) have studied the migration of PET oligomers from susceptor

packaging under actual microwave use conditions into food (popcorn, pizza, French

fries, fish sticks, and waffles). Food was extracted by solvent extraction and the

extract was separated and analysed by HPLC-UV. The cyclic trimer, tetramer and

pentamer were quantified and of the three oligomers the cyclic trimer migrated most,

whilst being the smallest in size (of the three cyclic oligomers) and the most abundant

of all oligomers in the polymer. The bulk of oligomers migrated from the PET bowl

into popcorn (31.3 µg/in2). French fries also had a high overall oligomer migration of

28.5 µg/in2 from susceptor packaging. However, these authors concluded that in a

previous study by Begley and Hollified (1990) it was observed that the level of

oligomers migrating into oil was five times the amount migrating into popcorn. This

result was attributed to the increased contact area with the oil compared to popcorn.

Castle et al. (1989) additionally determined the level of PET oligomer migration into

foods (e.g. lasagna, sausages, stewed apple), beverages and oil during microwave and

conventional cooking. Oligomers were measured as total levels whilst being

converted to terephthalic acid prior to separation and analysis by SEC and GC/MS.

Migration was higher for fatty foods and for susceptor packaging as opposed to the

PET tray because the former gets much hotter. The highest level of migration (1.47-

2.73 ppm) occurred into French fries that were microwave-heated in a susceptor

carton. Migration into aqueous beverages was slightly lower than that into the more

polymer-aggressive alcoholic beverages, with the levels in both cases falling below 80

ppb. Migration into food, beverage and oil was lower than the proposed global

migration limit of 10 mg/dm2 (60 ppm).

In another migration study, Buiarelli et al. (1993) identified the presence of the cyclic

dimer and its analogous ether in different brands of mineral water bottled in PET by

separating the solvent extract by HPLC and analysing the collected fractions by

GC/MS. The migrating levels of these compounds ranged from <10 ppb to 115.7 ppb

27

Chapter 2

for the cyclic dimer and from 42.8 ppb to 85.7 ppb for the analogous dimer ether.

2.2.8 Additives

Polymer additives such as plasticisers, thermal stabilisers, slip additives, light

stabilisers and antioxidants are added to polymers during manufacture in order to

improve their performance (O.-W. Lau and S.-K. Wong 2000, Nielsen 1991a).

Although PET has been reported to be relatively free from additives compared with

other polymers, some researchers have been able to identify, and in some cases

quantify, the presence of additives in food-grade PET and its contacting food medium.

Kim et al. (1990) analysed extracts from PET bottle using GC/MS and identified the

presence of fatty acids, the antioxidant BHT, and phthalate (e.g. DEHP, DEP, DBP,

DiOP) and adipate (BEHA) plasticisers. The concentrations of these potential

migrants, however, were all determined to be below the limits set by the FDA.

Monteiro et al. (1996) used an SEC-HPLC system followed by GC/MS to separate

and identify the components extracted from a PET bottle. The polymer extracts

consisted of plasticisers such as adipates, phthalates and erucamide; antioxidants

including BHA and BHT; and the UV stabiliser Tinuvin P. The amount of Tinuvin P

present in the top, body and bottom of PET bottles was determined and the author

concluded that there was no significant difference in the quantity in the different parts

of the bottle. Monteiro et al. (1998) later performed an analogous study without using

the SEC-HPLC system for separation prior to GC/MS analysis and obtained the same

result. The average level of stabiliser (Tinuvin P) in PET was determined to be 131

ppm, which was argued to be within the limits established by legislation [FDA?].

Buiarelli et al. (1993) also determined the amount of erucamide, added to polymers as

a plasticiser, migrating from PET bottles into mineral water, and the levels ranged

from 2.0 ppb to 182.0 ppb, depending on the brand of mineral water.

Freire et al. (1998) analysed volatiles in PET packaging materials and identified the

presence of DEHP - a plasticizer - and degradation products of BHT.

Nerin et al. (2000) also identified phthalate and adipate plasticisers including DEP,

DBP, DiBP and DOA. However their levels in PET were very low and the polymer

was therefore declared suitable for contact applications. Other compounds identified,

but not quantified in this study, were lubricants such as oleamide, erucamide and fatty

acid esters.

28

Chapter 2

2.2.9 Global migration

The migration studies discussed thus far were specific for the compounds of interest.

In order to determine whether the polymer poses a health risk when used as a food

grade plastic, the results from specific migration studies are compared to specific

migration limits (SML) enforced by the EEC and FDA for different migrants.

Alternatively, total migration tests, based on gravimetric determinations, can be

performed. The EEC and FDA both regulate global migration, enforcing a 60 ppb

and 50 ppb migration limit respectively.

Global migration usually involves the evaporation of the food simulant followed by

gravimetric analysis of the residue. Since heat is applied during simulant evaporation,

the resulting residue comprises of semi- and non-volatile compounds such as

oligomers. Monarca et al. suggested a method to account for volatile migrants, based

on freeze-drying prior to gravimetric analysis, however such techniques have not been

legislated for. Global migration results clearly depend on the nature of the extractant

or solvent. Gravimetric analysis is particularly applicable to aqueous and other polar

solvents. For non-polar solvents (oils), global migration usually involves weighing

the polymer before and after contact with oil. However, since oil has been shown to

absorb into the polymer, such weight difference techniques are subject to considerable

uncertainty (Baner et al. 1994a, Ashby 1988).

2.2.10 Other compounds identified in PET

Other than those already specified, compounds identified in virgin PET include

(Gramshaw et al. 1995):

Benzaldehyde

Butoxybenzene

2-Phenyl-1, 3-dioxolan

2-Methyl-1, 3-dioxolan

Dodecanoic acid

Dimethyl cyclohexane-1, 4-dicarboxylate

Toluene

Hexamethyl-cyclotrisiloxane

29

Chapter 2

Octamethyl-cyclotetrasiloxane

Decamethyl-cyclopentasiloxane

These compounds were extracted from PET by dynamic headspace at 200°C and

analysed by GC/MS. The author reported that PET had few migratable compounds

relative to the other polymers analysed [thermoset polyester, polyethersulphone,

poly(4-methyl-pent-1-ene)], again giving confidence in the use of PET as the plastic

of choice for food contact. The cyclic and linear siloxanes were assumed to be

constituents of silicon oil lubricants. The origins of the other compounds in this list

were not mentioned in by Gramshaw et al. (1995).

2.3 FOOD CONTACT CONSIDERATIONS FOR RECYCLED PET

2.3.1 Introduction to recycling

With the increasing market share of soft drink bottles made from PET, a means for

combating the amount of post-consumer PET bottles reaching landfills has become

more and more important. The obvious, and simplest, means is to reuse the PET

containers. Although this can, and is, done by the consumer there are considerable

barriers to the commercial reuse of PET containers. The most serious of these is the

inherent public health danger involved in reusing bottles which may have been mis-

used (and therefore considerably contaminated) prior to reuse.

A more viable, and commonly used, method is to recycle the PET, generally in non-

food contact applications such as clothing, pillow fillers, carpets, furniture, road

construction materials, automotive parts, film and food-contact and non-food contact

containers. Unlike reuse, the recycling of bottles for these purposes involves

destroying the original package and generating a new decontaminated resin.

About 25,000 tonnes of the 80,000 tonnes (31 per cent) of PET produced in Australia

each year is recycled, compared with about 50 per cent of all milk bottles (HDPE).

30

Chapter 2

2.3.2 Modes of recycling

There are three different methods for recycling post-consumer polymers – primary,

secondary and tertiary recycling. The reuse of post-consumer bottles is not

technically a form of recycling, as it does not involve destroying the original package,

but has been previously termed “zeroth order” recycling (US FDA 1992). Both

recycling and reuse have common goals - to decrease the build up of waste in landfills

– therefore they frequently appear as a part of the “recycling” category.

2.3.2.1 Re-use: Zeroth order recycling

Re-use involves visual and volatile screening of the post-consumer polymer batch

using colour scanners and “sniffers” prior to washing and sanitising the batch under

mild caustic, detergent and temperature conditions. The sanitised material can

practically only serve the function of its primary use (e.g. as a soft drink bottle) and

therefore must retain its structural integrity after washing. To ensure the bottles are

not distorted, the cleansing conditions (e.g. temperature) are milder during zeroth

order “recycling” than during other forms of recycling. Consequently, zeroth order

“recycling” is less capable of reducing contamination and therefore poses a

considerable health risk. For example, 45% of limonene and 31% of myrcene have

been removed from PET bottles by zeroth order recycling (Nielson 1994). In contrast

over 99% of model contaminants were removed from PET by secondary recycling

(Komolprasert and Lawson 1995, Franz et al. 1998).

2.3.2.2 Primary recycling

Primary recycling involves the physical reprocessing of industrial plant shavings and

off-cuts that have never had consumer exposure. Since they have not had consumer

exposure, they are unlikely to suffer from post-manufacture misuse and therefore have

a quality similar to that of virgin PET.

2.3.2.3 Physical reprocessing: Secondary recycling

Secondary recycling involves shredding the post-consumer polymer into flake, which

is then washed, dried and melted under vacuum before the polymer is reformed. Note

that the chemical structure of the polymer is not destroyed in this process. Secondary

recycling is more refining than zeroth order recycling and has greater economic

31

Chapter 2

incentive than tertiary recycling. Hence it is the most common form of recycling for

food contact purposes.

The Visy process

Visy Plastics has developed a novel, patented process for recycling PET. The Visy

secondary recycling process involves the following steps in sequence:

• The removal of outside contamination by tumbling the incoming bottles in hot

water (90-100°C) with cleaning agents (caustic soda and detergent) for a

typical time of 5 minutes. Further tumbling in a cylinder drains the water

containing dirt, contamination and labels.

• The use of molecular sensors to detect the presence of specific plastics.

Plastics such as PVC, HDPE, PP and PS are ejected after being detected by X-

ray absorption or infrared absorption.

• The grinding and intense washing of PET bottles. This involves grinding the

PET against rotating knives and a cutting screen in warm water (10-40°C)

containing caustic soda, surfactants and antifoaming additives. The flake is

then fed into the hot wash reactors, where it is washed in caustic detergent-

containing water for 10 – 20 minutes at 75-90°C.

• The sink-float separation of PP and PE from PET based on their differences in

density relative to water.

• The drying of the rinsed flake at 140-185°C in an atmosphere of flowing gas

for 5-8 hours.

• Melting and vacuum decontamination (extrusion).

• Subjecting the amorphous pellets to solid-state condensation to further

improve its purity for use in food-grade applications.

This process has been shown to result in "safe" PET suitable for recycle as a food

contact container. As a result, recycled PET is now included as raw feed stock in the

manufacture of food contact containers such as soft drink bottles. It is worth noting

that this use of secondary recycled PET coincided with the initial stages of this thesis.

32

Chapter 2

2.3.2.4 Tertiary recycling

Although tertiary recycling (also known as chemical reprocessing) is the best form of

recycling in terms of decontamination, it is seldom applied due to the process’ high

expense. Tertiary recycling involves destroying the polymer structure by

depolymerisation and regenerating monomers or oligomers, which are subsequently

purified via vacuum distillation and polymerised to regenerate the recycled PET resin.

The depolymerisation processes by hydrolysis, methanolysis, glycolysis and diolysis

remove any contaminants bound to the polymer chain. Figure (2.11) summarises the

main chemolysis reactions used in tertiary recycling (Scheirs 1998).

Figure 2.11: Chemolysis reactions used in tertiary recycling.

Glycolysis Methanolysis Hydrolysis

G H2O

+

Low MW

2.3.3 Recycled

There have b

manufacturers

food containers

manufacture in

combination o

Companies are

E

T

polyols E

PET for food contact pu

een two main reasons

to increase their market

. The first is simple e

creases and the cost

f the legislative, altru

more and more expected

MeOH

T A

BHE DM

+ G

rposes

for an increase i

-share of secondary

conomics as, with ti

of recycling decre

istic, and consum

to undertake the "cr

TP

n

re

me

as

er

ad

+

EG

interest on behalf of

cycled PET for use as

, the cost of monomer

es. The second is a

pressures to recycle.

le to grave" philosophy

33

Chapter 2

of manufacture where the company is responsible for the ultimate fate of the product

(waste plastic). However, the recycling of plastics for food-contact purposes has

raised its own concerns regarding the contamination of food by foreign and unwanted

substances that may migrate from these plastic into the food.

The presence of such migratory contaminants in recycled PET could arise from

sorption during the initial use of the plastic and/or consumer misuse/reuse. Of less

current interest, but potentially even more serious, is the potential for recycled or

reused PET containers to pose a public health threat due to microbial contamination

(Devlieghere and Huyebaert, 1997).

2.3.3.1 Consumer misuse/reuse

PET containers such as soft drink bottles are often re-used by the consumer for storing

substances like pesticides, herbicides, automotive fuels, household chemicals and

industrial chemicals (5,27,91,94, 99)??. The chemical constituents of these substances

may sorb into PET bottles and consequently pose a health risk after the containers are

recycled and used for food purposes.

Although a potential exists for the consumer to contaminate the plastic bottle with

household and garden chemicals, in practice this happens rarely. In fact, it is

estimated that contamination from consumer misuse is as low as 1 bottle per 10,000

(Bayer 1997). Furthermore, the degree of sorption for any given contaminant depends

on the nature of that contaminant, as discussed in Section 2.2.3, is rarely higher than

5-10% and often as low as 0.1% (Bayer 1997). Fortunately, PET has the advantage of

having a relatively low diffusivity (Franz et al. 1998). This property of PET acts as a

double safety factor, not allowing a high degree of sorption whilst also not allowing a

high degree of migration of the contaminant into food.

Due to the dilution effect in the plastic bottle feed stream, contaminants in post-

consumer PET are not expected to arise from consumer misuse but more likely from

the original contents of the bottle.

Nonetheless, it can be argued that any such contamination is highly problematic to

secondary recycling (as opposed to simple reuse) since it might contaminate the entire

batch of recycled plastic and therefore pose a threat to a large number of consumers.

34

Chapter 2

2.3.3.2 Sorption from the original contents of the bottle

As indicated in former studies (253,256,260 [make sure you remember to change the

form of the references here]), contaminants in recycled plastic do not only arise from

consumer negligence. Other forms of contamination derive from the environment

[e.g. contamination of LDPE milk bottles by naphthalene in air (Lau and Wong

1994)] as well as the original contents of the plastic containers. For example, many

chemicals present in personal hygiene products, cleaning agents and foods and

beverages could also sorb into PET plastic. Not only do such components pose a

health risk, they could also influence the organoleptic properties of the re-filled

foodstuffs (Jetten et al. 1999). Greater than 99% of the Visy recycling stream

comprises of carbonated soft drink bottles, fruit juice and vegetable oil bottles and

less than 0.9% of the total stream derives from non-food bottles such as mouth wash,

shampoo and detergent bottles. Since the amount of bottles used for household

solvents and cleaners represent only 0.1% of all bottles, the majority of contaminants

are likely to result from food, although of course these contaminants are less likely to

provide a serious health risk compared with any which do arise from the non-food

bottles

Huber and Franz (1997) have used a dissolution approach to extract limonene at 3-

ppm levels from recycled PET. In the same study, Soxhlet extraction was used to

extract limonene from HDPE, PP, and PS at levels higher than in PET. All extracts

were analysed and quantified by GC/MS and the polymers were additionally

subjected to odour analysis. It was found that for direct food contact use these

materials would not be in compliance with the food law.

Franz and Welle (1999b) extracted volatile compounds from washed and dried flake,

intended for extrusion, by static headspace GC/FID. Limonene and acetaldehyde

were quantitatively evaluated and found to exceed the threshold of regulation (see

section 2.3.4) suggesting that the washed and dried PET was not suitable for food

contact applications. The levels of limonene in the treated PET ranged from 1.5 ppm

to 11.0 ppm.

35

Chapter 2

Bayer (2002) used headspace and total dissolution to extract a total of 121 different

compounds from five different post-consumer PET feed-streams, each containing

different proportions of food containers and non-food containers. Since some of the

compounds could be traced back to the original contents of specific containers, the

relative amounts of food containers and non-food containers in each of the feed-

streams could be ascertained. The concentrations of compounds present above a

certain integrated count were determined in two feed-streams after the commercial

washing process. Limonene, present at 18 ppm, was the predominant contaminant in

the “deposit” feed-stream (100% beverage bottles). Methyl salicylate, present at 15.3

ppm, was the predominant contaminant “food-grade" PET used in non-food

applications (100% non-food containers such as mouthwash, detergents, cleaners).

Both of these compounds could be traced back to the original use of the container -

soft drink and mouthwash respectively. With the maximum total concentration of

contaminants in washed PET at ~40 ppm, the author concluded that regardless of the

source of the feed-stream, this PET could be re-used in food contact applications.

Triantafyllou et al. (2002) used total dissolution and solvent extraction to extract

potential migrants from recycled PET. The contaminants identified by GC/MS were

mainly soft drink flavourings such as limonene, γ-terpinene and p-cymene, originating

from the initial use of the packaging. The levels of limonene, the predominant

species, in washed and dried PET flake ranged between 2.5 and 15 ppm. After solid

state post-condensation, there were no detectable amounts of limonene or other

potential migrants detected in the PET, emphasising the efficiency of the recycling

process.

Another polymer, which is recycled and widely used for food contact purposes, is

HDPE. However, HDPE is likely to contain a higher level of contaminants than PET

since it has lower barrier properties. Therefore the efficiency of the recycling process

in terms of decontaminating HDPE is very important if the recycled PET is to be used

for food contact purposes. The decontamination of HDPE represents the worst-case

36

Chapter 2

scenario relative to the cleansing of PET, since the latter is renown for having lower

sorption character compared with HDPE.

There have been numerous studies focussing on the extraction of contaminants from

recycled HDPE. Recycled (caustic washed and steam/air stripped) HDPE milk bottles

have been evaluated for food applications after Soxhlet extraction with

dichloromethane and GC-FID analysis (Devlieghere et al. 1998). In addition, global

and specific migration tests into aqueous and fatty food simulants have been

performed (Devlieghere et al. 1998). The major compounds identified in the treated

HDPE were characterised by an even number of carbon atoms (C12 to C22), which

were presumed to derive from an incomplete polymerisation reaction. Oxidised

hydrocarbons, possibly produced during the pelletization process, were also

identified. As a result of the similarities in migration levels and extraction

chromatograms between virgin HDPE and steam stripped HDPE the recycled PET

was considered comparable to the virgin material. Oxidised hydrocarbons are known

to be odour responsible; therefore their screening in polymers is practised (Ezquerro

et al. 2002, Hakkarainen et al. 1997)

In another study performed by Huber and Franz (1997), contaminants in recycled

(ground, washed and extruded) HDPE produced from household waste were extracted

by Soxhlet extraction with dichloromethane and analysed by GC/MS. Most of the

compounds identified originated from personal hygiene products and cleaning agents,

foods and beverages, and residues from other polymers. Examples of compounds

extracted include limonene and hexanal (aroma compounds), esters of myristic and

palmitic acids (compounds for personal hygiene and cosmetic products), toluene

(industrial cleaning agent or solvent) and hexyl salicylate (preservative).

The highest concentrations were identified for limonene, di(ethylhexyl) phthalate

(additive or residue from other polymers) and the isopropyl ester of myristic and

palmitic acids, present at concentrations between 50-200 ppm. Many of the aroma

and preservative compounds were detected at levels between 0.5 ppm and 10 ppm.

Due to the high content of contaminants, it is argued that recycled HDPE should only

be used as packaging for non-food products.

37

Chapter 2

Camacho and Karlsson (2001) determined the quality of recycled HDPE and PP by

GC-MS after microwave-assisted extraction. Examples of fragrance and flavour

compounds extracted include limonene, 3-carene and betamyrcene. Other compounds

identified were from the following groups of compounds: carboxylic acids, aromatic

hydrocarbons, aliphatic hydrocarbons, esters, alcohols and ketones. Many of the

substances identified rose from personal hygiene products, cleaning agents, foods and

beverages. Some compounds were identified in the virgin polyolefins as well as in

the recycled polyolefins (e.g. aliphatic and aromatic hydrocarbons).

The similarity of contaminants found in post-consumer PET compared to post-

consumer HDPE and the relationship between these contaminants and the original

ingredients of the bottle gives confidence to the assumption that most contaminants

arise from the "normal" use of plastic bottles and not consumer mis-use.

2.3.4 Threshold of regulation

In order to insure that contaminants in recycled PET are unlikely to induce toxic

effects, the FDA has considered toxicity data from a large range of compounds, from

which it has established a threshold level for contaminants in PET. The threshold

level is that which, after the inclusion of a 200 to 2,000-fold safety factor, might result

in a toxic effect. Compliance with the “threshold of regulation” effectively means

that contaminants in the container do not exceed a dietary concentration of 0.5 µg/kg

(0.5 ppb) when contacted with food (Begley 1997).

An estimate of the exposure of a contaminant in the diet requires a combination of

migration amounts (Mi) with food type distribution factors (ft) and the fraction of the

daily diet expected to contact specific packaging materials (consumption factors, CF).

The food type distribution factors represent the fraction of all food that is aqueous,

acidic, alcoholic and fatty. The concentration of the migrant in each type of food is

obtained by multiplying the individual ft value by the measured amount of migration

(Mi) in food type i. Multiplying the sum of the concentration of migrant in each type

of food by CF gives the concentration of the migrant in the total diet (Equation 2.7)

(Begley 1997). 4

Dietary Concentration = CF ×

(Mi × ft) Equation 2.7 1

Σi=

38

Chapter 2

From Equation 2.7, the maximum migration values that will produce a dietary

concentration of 0.5 ppb were calculated for PET, PS, PVC, HDPE, PP, LDPE and

PC (Table 2.4). Table 2.4 also lists the threshold values for the maximum permitted

contaminant concentration in the polymer, determined from the calculated maximum

migration amounts, assuming 100% migration and that 1 in2 polymer contacts 10 g of

food. A further assumption was that the mass-to-surface area ratio for PET is 460 mg/

in2.

10 ppb migration into food = 10 ng/g of food

=100 ng/10g of food

= 100 ng/ in2 of packaging

= 100 ng/460 mg of packaging

≈ 215 ng/g of packaging

Subsequently, if 215 ppb of contaminant were present in the PET container and if it

totally migrated into food, the concentration of the contaminant in the daily diet

would be 0.5 ppb. Therefore, to ensure safety from induce toxic effects, the level of

specific contaminants in PET should not exceed 215 ppb.

Table 2.4: Threshold values for the maximum permitted contaminant concentration in

polymers and food simulant.

Polymer Maximum

concentration in the

polymer (ppb)

Maximum migration

amount (ppb)

Polyethylene terephthalate (PET) 215 10

Polystyrene (PS) 180 6

Polyvinyl chloride (PVC) 90 5

High density polyethylene (HDPE) 123 4

Polypropylene (PP) 778 25

Low density polyethylene (LDPE) 92 3

Polycarbonate (PC) 256 10

39

Chapter 2

2.3.5 Validation of recycling process – the challenge test

2.3.5.1 Introduction

In Australia, the US FDA regulation for food-packaging safety is accepted. For PET

to be suitable for direct food contact applications, the FDA requests that the cleaning

efficiency of the recycling process be such that a maximum of 215 ppb of

contaminant remains in recycled PET and/or a maximum of 10 ppb migrates into the

food simulant (Begley 1997, Komolprasert et al. 1997).

Since the levels and types of contaminants can vary significantly from batch to batch

the US FDA does not rely on analysis of "real life" contaminants in PET. Rather,

they require a "challenge" test. This test involves deliberately contaminating PET

with model contaminants (“surrogates”) at exaggerated levels to simulate worst-case

consumer misuse, subjecting the contaminated PET to the proposed recycling process

and examining the cleansing efficiency of various recycling stages (US FDA 1992).

The surrogates are chosen such that they represent five general categories of

compounds (volatile and non-polar; volatile and polar; non-volatile and non-polar;

non-volatile and polar; metallic/organometallic), as well as range of functional groups

and molecular weights, in order to model all the different chemical and physical

properties of real-life contaminants (Franz and Welle 1997, US FDA 1992, Begley

1997, Jetten et al. 1999, Komolprasert and Lawson 1995).

As an example, the surrogates used in a challenge test performed by Franz and Welle

(1999a) are listed in Table 2.5. These surrogates correspond to the recommendations

of the FDA document “Points to consider for the Use of Recycled Plastics in Food

Packaging” (US FDA 1992). Model contaminants used in other challenge tests were

components of household pesticides such as malathion and diazinon (Komolpraset et

al. 1995), lindane, chloroform, 1,1,1-trichloroethane, 1-octadecanol, squalene, copper

(II)-2-ethylhexanoate and zinc stearate (Scheirs 1998, Franz et al. 1998, Komolprasert

and Lawson 1997, Bayer 1997).

40

Chapter 2

Table 2.5: Surrogates used in a challenge test performed by Franz and Welle (1999a).

Surrogate name Molecular

weight

Functional group Properties

Acetone 58.1 aliphatic ketone volatile, polar,

water soluble

Toluene 92.1 aromatic hydrocarbon volatile, non-polar

Chlorobenzene 112.6 halogenated aromatic

hydrocarbon

volatile, medium-

polar, very

aggressive to PET

phenylcyclohexane 160.3 aromatic hydrocarbon non-volatile, non-

polar

benzophenone 182.2 aromatic ketone non-volatile, polar

methyl stearate 298.5 aliphatic ester non-volatile, polar

2.3.5.2 Challenge test process

The challenge test process involves soaking the plastic (bottles or flake) in a cocktail

of surrogates for 2 week at 40°C. The mixture is then drained and the polymer is

subjected to the recycling process. The challenge test presents a worst-case scenario

because:

• 100% of the polymer is contaminated in the challenge test, whereas normally

0.01% (1 in 10,000) of misused bottles and 0.1% of household solvent/cleaner

bottles enter the recycling stream. Thus a safety factor of 10,000 and 1,000,

respectively, is incorporated.

• The challenge test solutions are more concentrated than any likely commercial

solution (source of contamination in the "real" world), a factor of 10 times the

anticipated strength.

2.3.5.3 Challenge test studies

A range of challenge tests has been performed in the area of zeroth order recycling

(bottle reuse) and secondary recycling (physical reprocessing). The challenge tests

41

Chapter 2

carried out in the area of zeroth order recycling mainly involve optimisation of the

washing procedure whilst those in the area of secondary recycling involve the

assessment of every recycling stage in terms decontamination. The efficiency of

secondary recycling with respect to decontamination is generally superior to zero

order recycling as evidence by the level of contaminants migrating into simulants

from “reused” bottles exceeded the 10 ppb migration threshold. Conversely,

migration from reprocessed PET bottles often does not exceed the migration threshold

(Cross et al. unpublished).

Refillable plastic bottles

Feron et al. (1994) contaminated refillable PET soft drink bottles with 62 test

substances and passed the bottles through sanitization by washing. Studies on the

migration of these contaminants into simulated soft drink were performed and it was

concluded, after a detailed hazard assessment, that none of the tested substances

migrated into the simulant at levels that would pose a health concern. The hazard

assessment involved comparing the Acceptable Daily Intakes (ADIs), set by the Joint

FAO/WHO Committee on Pesticide Residues, with the “maximum potential

exposure” (MPE) calculated from the amount migrated per bottle. Comparisons of

the MPEs with the estimated non-toxic doses (ENTDs) were also made during the

hazard assessment.

Nielson (1994) studied the effects of different washing procedures on the removal of

orange flavour components (myrcene and limonene) from PET bottles. In order to

study the effects of caustic soda concentration and temperature on decontamination,

the different washing conditions considered were:

• 1.5% NaOH solution, 60°C, 15 minutes.

• 3% NaOH solution, 60°C, 15 minutes.

• 1.5% NaOH solution, 70°C, 15 minutes.

It was found in this study that about 32% of the myrcene and 22% of the limonene

sorbed into the PET bottles were removed at 60°C at both caustic concentrations

(1.5% and 3% NaOH). Increasing the temperature by 10°C caused a further 13% and

42

Chapter 2

11% removal of myrcene and limonene respectively (i.e. a total of 45% of myrcene

and 31% of limonene were removed), again for both caustic concentrations. It was

therefore concluded that temperature plays a vital role in cleaning whilst the

concentration of caustic soda (given that it was fairly concentrated in the first place)

does not. Even so, the final concentrations of myrcene and limonene were very high

and the washed bottles were deemed inappropriate for refilling. In addition, it was

discovered that washing virgin PET bottles prior to filling decreased the level of the

flavour components sorbed, possibly due to the increase in crystallinity with washing.

Devlieghere et al. (1997) examined bottle wash processes in removing limonene and

chloroxylenol from PET and other polymers. The effect of temperature, caustic soda

and commercial additive concentration was determined. The residual contamination

after washing was determined by means of migration into a beverage simulant. It was

shown that caustic soda concentration and temperature had the greatest effect on

decontamination. In general, higher temperatures resulted in higher decontamination.

However above 75°C, and at high caustic concentration, the cleaning effectiveness

declined. For PET, the optimal washing conditions were 70°C and 2-2.8% (w/v)

NaOH. Washing the bottles prior to filling had no effect on the amount of surrogate

absorbed.

The levels of limonene and chloroxylenol migrating were high above the 10 ppb limit

set by the FDA; therefore the washed bottles would not be suitable for refill.

Komolprasert and Lawson (1997) simulated contamination of PET bottles with

various model contaminants (benzene, butyric acid, lindane and malathion) and

determined the effects of washing on their removal.

The results were that 26 - 49% of surrogates were removed after washing at 74ºC and

then drying with an IR lamp. Since the levels of the contaminants remaining (29 -818

ppm) exceeded the 215 ppb threshold, their migration into 8% aqueous ethanol was

also determined. The migrating levels of benzene (2000 ppb) and butyric acid (260

ppb) after 10 days at 49ºC exceeded the 10 ppb threshold.

43

Chapter 2

Secondary recycled plastic bottles

Komolprasert and Lawson (1995) intentionally contaminated PET chips and bottles

with benzene, tetracosane, lindane, butyric acid, malathion and copper (II) 2-

ethylhexanoate. In order to evaluate the efficiency of decontamination, the

contaminated chips were washed at 90°C with aqueous 4% NaOH and subsequently

dried for 4 hours at 160-170°C prior to extraction and analysis using chromatographic

or spectroscopic techniques. Such high temperatures are not used during zeroth order

recycling and might be expected to result in higher efficiencies of removal.

It was found that washing alone (without the addition of NaOH) removed 14 - 91% of

surrogates from the chips and 26 - 99% of surrogates from the bottles. Tetracosane

was the easiest to remove, whilst benzene was the most resistant to cleaning. It was

assumed that during sorption, the non-polar, non-volatile tetracosane remained on the

surface of PET whilst the non-polar, volatile benzene diffused into the PET, making it

less accessible to cleaning.

The addition of NaOH further increased the cleaning efficiency of the PET chips to 30

- 91%. However, caustic soda had no effect on reducing the level of tetracosane

removal from PET; its percentage removal remained at 91%. The combination of

washing and drying removed more than 99% of organic surrogates from the polymer

chips. Despite the high percentage removal, some contaminants exceeded the 215

ppb threshold (e.g. benzene 6.2 ppm, butyric acid 0.27 ppm, malathion 120 ppm).

Therefore, additional recycling steps (e.g. extrusion, solid-phase condensation) and

migration studies should be performed to further purify the polymer and determine

whether it is suitable for food contact use.

In another study, Komolprasert et al. (1997) deliberately contaminated PET chips

with benzene, butyric acid, dodecanoic acid, dodecane, octadecane, tetracosane,

diazinon, lindane, and copper (II) ethyl hexanoate. The contaminated polymer was

passed through the recycling process. The levels of contaminants in PET exceeded

the FDA threshold after extrusion. However, the level of the contaminant migrating

into heptane and 8% ethanol was below the 10 ppb threshold.

Franz et al. (1998) performed a further challenge test involving the deliberate

contamination of PET with surrogates at different concentration levels. The

44

Chapter 2

contaminated polymer was passed through the extrusion and solid phase condensing

stages of the recycling process. The latter stage involved the use of a 12-hour high

vacuum and temperature program. Efficient removal of the volatile compounds (e.g.

toluene and chlorobenzene) by extrusion was observed. Solid-phase condensation

further reduced the amounts of these contaminants to undetectable levels.

The removal of non-volatile compounds (e.g. phenyl cyclohexane, benzophenone,

octadecanol and methyl stearate) was less efficient. Residual levels were as high as

25 – 50% for these surrogates after extrusion. The subsequent condensation step

reduced these levels to 2 - 4% for all non-volatile compounds except benzophenone,

which had the highest residual level of 8%. Benzophenone has a similar solubility

parameter to PET, therefore is the most persistent to remove (Harding et al.

unpublished). The introduction of a wash and dry stage before feeding the

contaminated flake into the extruder results in an overall efficiency of more than 99%

for all surrogates, including benzophenone. Despite the high percent removal, the

levels of residual model contaminants exceeded the 215 ppb threshold due to the high

initial level of contamination. However, the authors concluded that since the real life

contamination levels fall in the low ppm range, the recycled material would

theoretically not exceed the threshold of regulation and thus be suitable for food

contact applications.

Franz and Welle (1999a) performed a challenge test study, whereby extraction and

migration experiments confirmed the suitability of the recycled PET material for all

types of foodstuffs. The flake, contaminated with six model contaminants (acetone,

toluene, chlorobenzene, phenylcyclohexane, benzophenone, methyl stearate), was

passed through four steps of the recycling process (washing, drying, re-extrusion and

solid-state polycondensation). Extraction was accomplished by total dissolution

whilst specific migration was measured into 95% ethanol, 10% ethanol and 3% acetic

acid. The extraction experiments showed that the removal of volatile compounds

such as acetone, toluene and chlorobenzene was more than 99.9% complete after the

dry stage of the recycling process. Phenylcyclohexane, which is non-volatile and

non-polar, exhibited the same degree of cleansing, but only after extrusion.

Benzophenone and methyl stearate, which are polar and non-volatile, were the most

challenging surrogates. After extrusion, the levels of benzophenone and methyl

45

Chapter 2

stearate remaining in the polymer were 1.46 ppm and 0.15 ppm respectively. These

levels decreased to 0.987 ppm for benzophenone and 0.069 ppm for methyl stearate

after bottle formation.

All compounds, other than benzophenone, fell below the 215 ppb threshold after the

extrusion step. The level of benzophenone exceeded the 215 ppb threshold even after

bottle formation. Fortunately, migration experiments demonstrated that the levels of

migration for all compounds, including benzophenone, fell below the 10 ppb FDA

threshold. Therefore, the FDA considered the recycled PET bottles suitable for food

contact use.

Of direct interest to this thesis is the Visy recycling process, which has also proven

efficient in the cleaning of surrogate compounds out of deliberately contaminated

PET intended for soft drink use. The polymer was passed through the recycling

process and analysed for residual contaminants after each cleansing stage. The

process, which included washing, drying and extrusion (with vacuum

decontamination), was found to be capable of removing from 95 – 99.7% of the

contamination. However, the level of contaminants in the recycled bottles again

exceeded the FDA threshold of 215 ppb, with benzophenone the most challenging

compound (20 ppm of benzophenone was present after bottle formation).

Fortunately, the level migrating into 10% ethanol fell below the 10 ppb threshold.

This final result made the recycled bottles suitable for soft drink applications, given

that the FDA is less concerned with the level of contaminant in the container than it is

with the amount migrating into food.

2.3.6 Estimated level of real contaminants in recycled PET

From the challenge test results obtained for the Visy recycling process and the

assumption that the challenge test included a safety factor 10,000, the realistic level of

contaminants (from consumer misuse) after each stage of the Visy recycling process

can be estimated, assuming the realistic levels in the treated polymer are 10,000 times

lower than the levels determined by the challenge test. The results for this estimation

(in ppb) are shown in Table 2.6. From this table the realistic levels of contaminants

resulting from consumer misuse in the recycled bottles do not exceed 2.0 ppb, which

is well within the threshold limit of 215 ppb.

46

Chapter 2

The number of bottles resulting from household solvent/cleaners is ten times the

number resulting from consumer misuse (1 bottle in 1000 bottles), therefore the levels

of contaminants in the recycled PET bottle resulting from non-food contact

applications is predicted not to surpass 20 ppb. This is further reduced to 2 ppb when

the exaggerated user strength factor of 10 is taken into account.

Table 2.6: The estimated level of contaminants in PET after each recycling stage.

Contaminant

Level of

contaminant in unwashed

PET flake (ppb)

Level of

contaminant in washed PET flake

(ppb)

Level of

contaminant after

extrusion (ppb)

Level of

contaminant after moulding

into bottles (ppb)

Chloroform 61.3 5.2

0.5 0.2

Toluene 176.8 36.1 < 0.5 < 0.5

Benzophenone 71.3 17.5 3.6 2.0

Methyl stearate

8.1 1.6 < 0.5 < 0.5

Copper octanoate

23.0 0.8 0.4 0.4

2.3.7 Methods of reducing contamination

Strategies to lower the amount of contaminant migration include blending the

recycled polymer with virgin resin or including a functional barrier in the bottle

structure. The latter involves inserting the post-consumer recycled material between

two layers of virgin resin. Thus there is a virgin polymer barrier between the food

and the recycled resin, whose presence will delay the migration of contaminants from

the recycled layer.

In Australia, Coca-Cola Amatil manufactures beverage bottles with 25% recycled

content whilst ACI Petalite manufactures multilayer (functional barrier) beverage

bottles.

47

Chapter 2

2.3.7.1 Functional barrier

Any functional barrier must reduce the contaminant migration to an acceptable level

that would not endanger the consumer. The degree of migration (Mt) through a

functional barrier depends on the contaminant diffusion coefficient (D), the thickness

of the virgin layer (l), the concentration of contaminant in the recycled layer (C0) and

the time (t). Equation 2.9 describes migration through a functional barrier, however it

is too complex for practical use and a simplified equation (Equation 2.8) describing

migration through a virgin layer has been developed (Begley and Hollifield 1993,

Kuznesof and Van Derveer ??).

τ = Dt/l2 Equation 2.8

Mt = Dt/l2 – 1/6 – 2/π

It was calculated, using E

bi-layer migration, that w

perspective on the diffusio

food), the amount migrati

thereby indicating that th

Hollifield 1993).

Compared to the migratio

the amount migrating from

any given concentration o

in the case of bi-layer p

concentration of less tha

contaminant in the recycle

layer packaging, the co

Alternatively, if the conce

µg/cm3, the migration lev

the bi-layer package.

Mitz et al. (1997) further

deliberately contaminated

2 Σ (-1n/n2) exp (-n2π2Dt/l2) C0 l Equation 2.9

quation 2.3 for monolayer migration and Equation 2.9 for

hen τ > 0.6 (τ is a unitless variable that provides a

n rate for a contaminant through the functional barrier into

ng begins to approximate that of the monolayer packaging

e functional layer has no barrier properties (Begley and

n=1

n level calculated from the monolayer migration equation,

a bi-layer package was calculated to be much smaller for

f contaminant in the recycled polymer layer. For example,

ackaging (for τ = 0.15), in order to obtain a migration

n 10 ppb in the food, the maximum concentration of

d PET must not exceed 100 µg/cm3. In contrast, for mono-

ntaminant concentration must not surpass 7.5 µg/cm3.

ntration of contaminant in a single-layer package were 100

el would be 130 ppb compared to the 10 ppb obtained for

observed that migration of the surrogate, toluene, from a

film was two orders of magnitude larger than that migrating

48

Chapter 2

from a multilayer packaging containing the same amount of surrogate in the centre

layer.

Begley and Hollifield (1995) studied the migration of diethylene glycol dibenzoate

(DEGDB) from spiked paperboard through a virgin PET layer into oil. The oil was

analysed with time and a graph of the amount of DEGDB migrated versus time was

plotted. The “lag-time” technique was used to calculate the diffusion coefficient.

This involved determining the lag-time from the plot of “amount migrated versus

time” and substituting the value into Equation 2.10 to calculate the diffusion

coefficient.

Lag time = l 2 /6D Equation 2.10

The calculated diffusion coefficient was entered into Equation 2.9, along with the

contaminant concentration in the recycled layer (C0), the time and various virgin layer

thicknesses, in order to calculate the amount migrating for different thicknesses of

virgin layer. It was found that for the migration of DEGDB not to exceed the FDA

threshold of 10 ppb in 1 hour, the thickness of the virgin layer must surpass 3 mm

when the initial concentration in the recycled resin is 10 ppm and the temperature is

150°C.

The above examples illustrate that equations 2.8, 2.8 and 2.10 can be used to estimate

the minimum thickness or the minimum percentage of a virgin functional barrier

required to give a safe product. Although this optimises expense on the part of the

manufacture, it is always more expensive to use a functional layer than not to. For

this reason, interest should also be applied to economical methods of cleaning

recycled PET such that it meets the requirements of a food contact product. Emphasis

also needs to be placed on the level of contaminants present in real examples of

recycled PET and methods for determining such.

2.4 EXTRACTION AND ANALYSIS OF POLYMERS

49

Chapter 2

2.4.1 Modes of extraction

The first step to the analysis of components in PET involves an extraction facilitated

by solvent strength, supercritical fluid state and/or temperature. Plastic samples have

been traditionally solvent extracted using Soxhlet extraction (Huber and Franz 1997a,

Costley et al. 1997, Kim et al. 1990), sonication (Pochivalov 2000, Nerin et al. 1998),

boiling under reflux (Vandenberg 1997) or total dissolution (Komolprasert and

Lawson 1997, Franz and Huber 1997, Nerin et al. 2000, Franz and Welle 1999a).

These techniques are either time consuming, use large quantities of solvents and/or,

due to the requirement of a post-extraction concentration step, are only applicable to

the analysis of semivolatile compounds. The desire to reduce extraction time and

solvent volume has triggered the development of automated extraction techniques.

These are microwave-assisted extraction (MAE) (Camacho and Karlsson 2001,

Costley et al. 1997), accelerated solvent extraction (ASE) (Vandenburg et al. 1997,

Lou et al. 1997) and supercritical fluid extraction (SFE) (Bartle et al. 1990, Lou et al.

1996, Nerin et al. 2000). Although very expensive in terms of equipment, these

techniques decrease analysis time and are less "man-power" intensive. They increase

the diffusion of analytes from the polymer matrix into the extraction fluid by using

several complementary kinetic parameters, including temperature, pressure and

suitable polymer-swelling solvents. By heating the sample in a closed system at

elevated pressures, the extractant and volatile components are not lost during the

automated extraction. Subsequent analysis is thus more reliable than non-automated

extractions. Furthermore, when extracting polymers in the traditional way, additional

headspace techniques are necessary to account for the omitted volatile components.

Examples of headspace methods applied in the polymer industry are static (Dong et

al. 1980, Komoprasert and Lawson 1995, Hakkarainen et al. 1997, Paik 1992) and

dynamic headspace (Komoprasert et al. 2001, Komoprasert et al. 1994, Bayer 2002),

solid phase microextraction (SPME) (Bart 2001, Hakkarainen et al. 1997) and thermal

desorption analysis (TDA) (Komolprasert et al. 2001, Bayer 2002).

Another advantage of automated techniques over the long-established traditional

extractions is that large volumes of solvent are unnecessary. Apart from the economic

advantage this provides, there are also advantages in terms of toxicology and that the

concentration of organic interferences from solvent impurities in the extract is

reduced.

50

Chapter 2

Despite these advantages of techniques, the conventional approach is often viewed to

be more attractive in terms of equipment availability/pricing and for the analysis of

thermolabile compounds (Camacho and Karlsson 2001, Eskilsson and Björklund

2000). Another disadvantage of the automated methods is that the PET particles can

fuse or collapse at temperatures > 125ºC whilst undergoing a swelling effect with the

solvent. This behaviour, which impinges on polymer surface area, has been shown to

cause a decline in the extraction efficiency (Eskilsson and Björklund 2000,

Vandenburg et al. 1997, Lou et al. 1997).

2.4.2 Parameter optimisation

Each of the mentioned extraction techniques requires method validation in the quest

to optimise extraction. In the area of polymer extraction, the external parameters that

affect the degree of extraction are time, temperature, pressure, particle size, flow rate

and the nature of extraction solvent. Polymer and migrant properties affecting the

level of extraction are shape and size (Garde et al. 1998) as well as polymer

crystallinity (Spell and Eddy 1960) and glass transition temperature (St. Küppers

1992). These have already been discussed in terms of migration studies and the same

principles apply to efficiency of analysis:

2.4.2.1 Time

Extraction involves the mass transfer of analytes though the bulk polymer matrix into

the extracting solvent. The movement of compounds through the polymer into the

solvent takes time, depending on the migrant’s ability to diffuse.

The time required to reach a plateau concentration in the solvent is considered the

optimum because such a plateau is assumed to be due to reaching total extraction

(Feigenbaum et al. 2002).

Wim and Swarin (1975) monitored the Soxhlet extraction process with respect to time

for two additives from polypropylene pellets and concluded that a plateau was

reached by 24 hours (using tetrahydrofuran as the extracting solvent). Similarly,

Komolprasert et al. (2001) extracted PET sheets by Soxhlet using dichloromethane

for 24, 48, 72 and 96 hours, and concluded that the optimum extraction time was 24 h.

Shorter extraction times were not considered in that study.

51

Chapter 2

Kinetic extraction studies can be used to determine the diffusion coefficients during

extraction. Feigenbaum et al. (2002) determined the “effective” diffusion coefficient

for the solvent extraction of aromatic migrants out of different polymers using

Equation 2.3 and the plot of per cent migrated versus time. These values were

calculated for comparative purposes only, as they have no direct physical meaning.

The optimisation of polymer extraction with respect to time has been carried out

during heat-facilitated extractions such as SFE (Garde et al. 1998), ASE (Lou et al.

1997), MAE (Camacho and Karlsson 2000) and SPME (Ezquerro et al. 2002).

Temperature is a diffusion enhancing parameter, therefore extraction time is reduced,

compared with Soxhlet extraction, during high temperature extraction. However, it

was discovered that extraction yields decreased with time during MAE of MDPE

(middle density polyethylene) at high temperatures due to the thermal degradation of

the analyte (Camacho and Karlsson 2000). Therefore extended time periods during

high temperature extraction could prove to be unfavourable for thermolabile

compounds.

2.4.2.2 Temperature

The rate of diffusion follows an exponential Arrhenius form with temperature, where

rate ∝ A exp (-E/RT), as described in Section 2.2.3.5. This indicates that increasing

the temperature should exponentially increase the diffusion rate. Temperature also

increases the solubility of the solutes in the extractant and reduces their interaction

with the polymer surface (Lou et al. 1997).

The temperature of Soxhlet extraction is restricted to the boiling point of the

extracting solvent; therefore unless a high boiling point solvent is used, extraction can

be time consuming (Lou et al. 1997). Therefore high boiling point solvents such as

xylene and 1-methyl naphthalene have been used for the extraction of oligomers from

PET (Hudgins et al. 1978, Costley et al. 1997, Cooper and Semlyen 1973). Hudgins

et al. (1978) concluded that extraction in a “Parr Bomb” using chloroform was a more

efficient way of extracting oligomers out of PET than Soxhlet extraction with

chloroform or xylene. This is because chloroform, recognized as a PET swelling

solvent, was heated above its boiling point under high pressure.

52

Chapter 2

Similarly automated solvent extractions such as MAE, SFE and ASE are performed in

closed systems and often at high pressures, which allow the solvent to be heated well

above its normal boiling point without the loss of volatile analytes. Therefore, the

ideal solvent, which frequently has a low boiling point, can be used at high extraction

temperatures.

Lou et al. (1997) and Vanderburg et al. (1999) extracted polymeric samples by ASE.

It was postulated by Lou et al. (1997) that when diffusion is the rate-limiting step, as

opposed to solubility, the effects of temperature on the amount extracted are

prominent. Conversely, when solvent flow rate has a positive effect on extraction, the

extraction rates are controlled by solubility. The extraction of an additive

(caprolactam) from nylon-6 and oligomers from PBT [poly(1,4-butylene

terephthalate)] was monitored with increasing temperature and flow rate. Since the

amount extracted increased with temperature, but not with flow rate, it was argued

that the ASE extraction rate was controlled by diffusion as opposed to solubility.

Similarly, higher temperatures lead to higher extractions rates for SFE of polymers,

where solubility in the supercritical fluid is not rate limiting. However, increasing the

temperature decreases the density of the supercritical fluid, which inevitably makes

solubility the rate-limiting factor. Thus, extraction initially increases with

temperature until the supercritical fluid density decreases such that solubility becomes

rate limiting (Lou et al. 1996). The temperature at which the amount extracted starts

to decline depends on the properties of the analyte and the matrix as well as the

supercritical fluid pressure and flow rate (Lou et al. 1996).

St. Küppers (1992) observed a sharp increase in the amount of trimer oligomer

extracted from PET at the glass transition temperature. Diffusion at the Tg is

accelerated because this is the temperature when the polymer structure changes from

its rigid glassy form to a rubbery form, relieving the diffusion of analytes through the

matrix. However, St. Küppers (1992) also demonstrated that after the Tg has been

reached, the amount of trimer extracted began to stabilise indicating that extraction

has become solubility-limited. Higher pressures and the addition of modifier were

shown to increase amount of trimer extracted at high temperatures, probably due to

the enhanced solubility initiated by these two parameters.

53

Chapter 2

2.4.2.3 Pressure

Pressure has been shown to have a positive effect on the SFE of polymers, especially

when the extraction rate is solubility limited (e.g. at high temperatures), increasing the

supercritical fluid density, which in turn improves the solubility (St Küppers 1992,

Lou et al. 1996).

Increased pressure can also increase the degree of plasticisation of the polymer, thus

increasing diffusion and the extraction rate (Vandenberg et al. 1997). For example,

the swelling of nyon-6 and PBT caused by the high-pressured supercritical fluid, was

the reason for the absence of a sharp rise in extraction recovery at the Tg (Lou et al.

1996). This is because the pasticisation of polymers by CO2 lowers the Tg

(Vandenberg et al. 1997).

Daimon and Hirata (1991) performed the selective extraction of additives from PP by

adjusting pressure and temperature. The authors found that different additives could

be extracted at different temperatures and pressures, depending on the diffusivity and

solubility of the additive. Similarly, St Küppers (1992) selectively extracted

oligomers from the surface of PET fibres. The temperature and pressure was then

increased to further extract oligomers from the inner core of the polymer matrix.

2.4.2.4 Nature of extraction solvent

The nature of the solvent is usually chosen such that it interacts with the polymer,

thereby swelling it. However, swelling the polymer at high temperatures can be

detrimental to extraction efficiency because it can lead to the coalescence of the

ground polymer particles, reducing their surface area (Vandenburg et al. 1999). It is

also possible for the fused polymer to lose viscosity and block instrument transfer

lines.

Vandenburg et al. (1999) has designed an ASE method to prevent the coalescence of

polymer particles at high temperatures which makes use of the chemical capatability

between the polymer and solvent. Since the swelling of a polymer can be described in

terms of the Hildebrand solubility parameter similarity between the solvent and the

polymer (Table 2.7), the initial part of the method involves extracting the polymer at

high temperatures with a solvent whose solubility parameter differs from that of the

54

Chapter 2

polymer (i.e. a poor solvent for the polymer). The temperature at which maximum

extraction occurs without polymer coalescence is determined and used. A solvent

with a solubility parameter similar to that of the polymer is then added (in increments)

causing the polymer to swell and enhancing diffusion. An optimum amount of

swelling solvent was determined before the onset of coalescence.

Costey et al. (1997) had identified that dichloromethane, theoretically the best solvent

for PET, fused the polymer at 120°C. Therefore Vandenburg et al. (1999) first

optimised the extraction temperature using hexane, a poor solvent for PET, and then

added ethyl acetate, a good solvent for PET. It was found that significant swelling did

not occur at 190°C until 100% ethyl acetate was reached.

Due to the swelling capability of dichloromethane and chloroform towards PET, these

chlorinated solvents are often used in neat form during Soxhlet extraction/maceration

(Shiono 1979, Komolprasert et al. 2001, Triantafyllou et al. 2002) or as modifiers

during SFE (St Küppers 1992).

Modifiers in SFE are important for two reasons: (a) they can swell the polymer,

thereby enhancing diffusion and (b) they can increase the solvent strength of the

supercritical extractant thus increasing solubility (Lou et al. 1996).

The addition of dichloromethane (DCM) into the supercritical fluid has been shown to

have a positive effect on the solubility of cyclic trimer in the extractant at high

temperatures (St Küppers 1992). The sorption of DCM into the PET matrix caused

high oligomer extraction at low temperatures. Methanol and isopropyl alcohols, on

the other hand, were found to be poor modifiers for PET and completely prevented

extraction.

The effects of modifiers have also been studied during the SFE of other polymers

(Lou et al. 1996). It was found that the effects of modifiers on the extraction rate are

different depending on whether diffusion or solubility governs the extraction rate.

Modifiers were determined to be more effective at lower temperatures, when the

extraction rate is diffusion limited. At high temperatures, modifiers sometimes have a

negative effect on extraction. At low temperatures and in diffusion-limited situations,

the contact time between the polymer and the modifier is important because polymer

55

Chapter 2

swelling is a slow process. The modifier amount is another important factor in

diffusion and solubility limited situations.

Garde et al. (1998) optimised the extraction of antioxidants from PP. Two modifiers

were tested - hexane to swell the polymer and liberate the antioxidants and methanol

to increase the solubility of the antioxidants in the supercritical fluid at high

temperature.

Spell and Eddy (1960) emphasized the importance of the nature of the solvent during

the shake-flask extraction of additives from PE pellets. These authors found that 2.5

hours were required to recover the antioxidant Santanox [4,4-thio-bis(6-tert-m-

cresol)] using carbon disulfide whereas 76 hours was required to recover the

antioxidant using iso-octane. Similarly, Wims and Swarin (1975) extracted additives

from talc-filled PP using tetrahydrofuran (THF), chloroform and DCM. The most

efficient extraction solvent was THF. It gave complete recovery in 24 hours as

opposed to the 72 hours required using chloroform. Using DCM only 50% of the

additives were extracted in 24 hours. Lou et al. (1997) also observed the effects of

using different solvents when extracting compounds from nylon-6 and PBT.

Methanol for nylon-6 and chloroform for PBT were more efficient than hexane in

Soxhlet extraction. However, extraction with hexane was improved at elevated

temperatures, suggesting that a poor extraction solvent in a Soxhlet extraction could

be a good solvent in ASE at high temperatures, which facilitate diffusion.

The nature of solvents used during total dissolution extraction depends on the analytes

of interest. The extraction of non-polar compounds out of the PET matrix involves

dissolving the polymer in trifluoroacetic acid (TFA) and partitioning the liberated

analytes into hexane/heptane (Bayer 2001, Komolprasert and Lawson 1995). Polar

compounds are usually extracted by dissolving the PET in a mixture of

hexafluoroisopropanol (HFIP)/DCM followed by polymer precipitation with

methanol/acetone/isopropanol (Begley and Hollifield 1989, Triantafyllou et al. 2002,

Komolprasert et al. 1995). An alternative method for extracting polar compounds

involves dissolving PET in TFA and partitioning the liberated analytes into methyl

tertiarybutyl ether (MTBE) (Bayer 2001, Barnes et al. 1995).

56

Chapter 2

The advantage of dissolution and re-precipitation over liquid-solid extraction is that

(theoretically) no analyte remains bound in the polymer network. This is because the

solvent effectively breaks up the polymer matrix. Therefore optimisation of

diffusion-enhancing parameters is not necessary.

Table 2.7: Hildebrand solubility parameters for different solvents and polymers.

Material Solubility

parameter (Mpa1/2)

Material Solubility

parameter (Mpa1/2)

Hexane 14.9 Ethanol 26.0

Cyclohexane 16.8 Methanol 29.7

Ethyl acetate 18.6 Polypropylene 16.6

Chloroform 19.0 PVC 19.5

Dichloromethane 19.8 PET 20.5

Acetone 20.3 Nylon 6,6 28.0

2-Propanol 23.8 PMMA 19.0

2.4.2.5 Particle size

As mathematically demonstrated in Equation 2.11, the amount of analyte migrating

from a polymer into a solvent is inversely proportional to the path length (L), i.e.

Amount migrating ∝ D/L2 Equation 2.11

Therefore, a decrease in L causes an increase in the amount of analyte migration. In

order to reduce the path length for the mass transfer of analytes from the polymer core

to its surface, the particle size is generally reduced by cryogenic grinding or flattening

prior to extraction (Garde et al. 1998, Lou et al. 1997, Vandenburg et al. 1997, Bartle

et al. 1990, Hunt and Dowle 1991). In the area of polymer extraction, grinding is a

widely accepted sample preparation technique for path length reduction (Huber and

Franz 1997 a, b, Nielson 1991, Daimon and Hirata 1991). Another advantage of

grinding polymer samples is that it reduces the inhomogeneous distribution of

57

Chapter 2

analytes when small masses of samples are used (Daimon and Hirata 1991, Hunt and

Dowle 1991).

Work by Perlstein (1983) and Spell and Eddy (1960) demonstrated that powdering

PVC and PE pellets had a positive effect on Soxhlet and shake-flask extraction time

and recovery. In addition, Spell and Eddy (1960) found that extraction time required

for complete recovery was also affected by polymer density and crystallinity.

Garde et al. (1998) have presented the effect of decreasing path length on the SFE of

antioxidants from PP. These authors observed that, in general, as particle diameter

decreased, the extraction percentage increased. However, since extraction from films

was lower than that from powder, results were better correlated with surface to

volume ratio rather than directly to size. Ashraf-Khorassani et al. (1991) and Hunt

and Dowle (1991) also found that faster SFE of additives is obtained as the surface

area of the PE and PVC matrix increases.

2.4.2.6 Migrant shape/size

As Soxhlet extraction is mainly diffusion-limited, extraction recovery and time is

dependent on migrant size. This is because D ∝ M as shown in Equation 2.6. Apart

from molecular weight, migrant shape and interaction with the polymer could also

play a role towards its diffusion through the polymer matrix. It has been shown by

Spell and Eddy (1960) that during the solvent extraction of antioxidants out of PE,

diffusion of BHT was faster than that of Santonox possibly due to the larger MW of

the latter. Similarly, Perlstein (1983) found that for this reason Uvinul N-539 was

extracted more efficiently than Tinuvin 320 and Cyasorb UV-9 from PVC.

Garde et al. (1998) extracted antioxidants from PP by SFE and demonstrated that

differences in extraction recoveries were caused by the different diffusivity of each

antioxidant. The antioxidants that were the largest in size and/or had voluminous

groups were the slowest to diffuse through the PP matrix into the extraction fluid. An

increase in temperature could assist the diffusion of the larger compounds through the

PP matrix. Daimon and Hirata (1991) presented the effect of temperature on the

selective SFE of antioxidants out of PP.

58

Chapter 2

In addition, Daimon and Hirata (1991) showed that, in SFE, antioxidant size could

impinge on solubility, which in turn could affect the amount extracted. Therefore, the

pressure of the supercritical fluid must rise in order to increase solubility and thus

facilitate the extraction of the larger, less soluble compounds. The manipulation of

pressure and temperature in extracting antioxidants of interest underlines the high

selectivity of SFE.

Similarly, Lou et al. (1996) showed the effect of oligomer size on solubility during

the SFE of dimer and trimer from PBT. It was found that the amount of the dimer

extracted was the highest at 150°C and for the trimer at 110°C. Therefore, the

extraction of the larger and less-soluble trimer becomes solubility-limited at a lower

temperature (i.e. when the density of the supercritical fluid is higher).

2.4.3 Modes of separation and analysis

Once the polymer is solvent extracted, the extract is generally concentrated by solvent

evaporation prior to instrumental separation and analysis.

The most common instrumental forms of extract separation and analysis are

HPLC/UV and GC/MS.

HPLC is used in the analysis of non-volatile, polar and/or thermolabile polymer

constituents such as oligomers, residual reactants and their bi-products (e.g.

terephthalic acid, MHET and BHET) and additives.

In the absence of HPLC, the structures of non-volatile and polar compounds could be

chemically made suitable for GC analysis. Kim et al. (1990) demonstrated that polar

compounds such as ethylene glycol and terephthalic acid could be made more volatile

by derivatisation (trimethylsilylation) to esters/ethers prior to GC analysis (Atkinson

and Calouche 1971). Furthermore, involatile oligomers could be broken down into

dimethyl terephthalate by hydrolysis and subsequent methylation prior to GC analysis

(Gramshaw et al. 1995, Castle et al. 1990, Castle et al. 1988). The more volatile low

molecular weight oligomers of PET (dimer and dimer ether) have been successfully

identified by GC/MS without chemical modification (Buirelli et al. 1993, Monteiro et

al. 1996).

59

Chapter 2

Size exclusion chromatography (SEC) can be used to "clean up" extracts that require

removal of the higher molecular weight compounds prior to GC and HPLC analysis

(Monteiro et al. 1996, Staetin et al. 1987, Gramshaw et al. 1995, Gilbert et al. 1982,

Castle et al. 1989, Shiono 1979). In a similar way, fractions can be collected during

HPLC and further separated and identified by GC-MS analysis (Biuarelli et al. 1993).

SEC has very poor resolution and therefore is not recommended as a single means of

separation and analysis (Munteanu et al. 1987), although some authors have separated

and quantitated polymer additives in this way (Shiono 1979).

Compounds extracted by SFE have been analysed by GC (Nielsen et al. 1992, 1991b,

Garde et al. 1998), HPLC (Bartle et al. 1991) and SFC (Bartle et al. 1991, Ashraf-

Khorassani and Levy 1990) in the past. SFC extends the range of molar mass above

that available by GC and has advantages in separation efficiency compared with

HPLC, especially of closely related compounds such as oligomers (Bartle et al. 1991).

2.5 PURPOSE OF THIS THESIS

The general purpose of this thesis is to determine what volatile and semi-volatile

contaminants are present in post-consumer PET and whether the identified

contaminants exceed the 215 ppb “threshold of regulation” set by the US FDA in

order to satisfy food safety requirements (the “threshold of regulation” is discussed in

Section 2.3.4). Although this threshold level was originally established as the upper

limit of contamination for challenge compounds (surrogates) in recycled PET, it is

analogously used in this thesis as a regulation for actual contaminants in recycled

PET.

Migration testing, such as that performed on virgin and recycled PET in the past (see

Sections 2.2.4 - 2.2.10 for virgin PET and Section 2.3.5.3 for recycled PET) was not

performed in this study because the levels of “contaminants” identified in recycled

Visy PET were very low (in the low ppb range); therefore the level migrating was

expected to be below the 10 ppb threshold set by the FDA for migration.

60

Chapter 2

The “contaminants” which are expected to be predominantly present in post-consumer

PET are those sorbed during the initial use of the container such as soft drink

additives (e.g. limonene, benzoic acid, γ-terpinene, carvone) and

mouthwash/detergent/shampoo additives (e.g. cineole, menthone, methyl salicylate),

as contamination from consumer reuse/misuse is less likely. In fact it is estimated that

contamination from consumer misuse is as low as 1 bottle per 10,000 (Section

2.3.3.1) whilst in contrast greater than 99% of the Visy recycling stream comprises of

carbonated soft drink bottles, fruit juice and vegetable oil bottles (Section 2.3.3.2).

Bayer (2002) has already identified the mentioned soft drink and

mouthwash/detergent/shampoo additives, amongst many other chemicals including

alcohols, aldehydes, ketones, esters, carboxylic acids, aromatics and alkanes, in post-

consumer PET using thermal and total dissolution extraction methods.

The presence of compounds inherited during the initial manufacture of PET (e.g.

monomers, additives, oligomers, breakdown products) may also be identified during

the screening of post-consumer PET. Sections 2.26 – 2.2.8 discussed these

compounds and their earlier detection in PET and food simulants.

Nerin et al. (2003) also identified numerous compounds in recycled PET flakes.

These were classified by chemical group: aroma compounds (e.g. limonene and p-

cymene), aliphatic aldehydes (e.g. hexanal, heptanal, octanal, nonanal and decanal),

aromatic aldehydes (e.g. benzaldehyde), esters (e.g. vinylbenzoate, methyl

dodecanoate and methyl stearate), aliphatic acids (acetic acid, hexanoic acid and

nonanoic acid), aromatic compounds (e.g. p-xylene, isopropyltoluene and toluene),

alkanes (e.g. dodecane), plasticisers (e.g. dibutylphthalate and dioctylphthalate),

ketones (e.g. benzophenone) and alcohols (e.g. dodecanol and hexadecanol). Similar

compounds were identified by Bayer et al. (2002) in an independent study, therefore

it is highly likely that a selection of these compounds will be identified in the post-

consumer PET we extracted.

The decontamination efficiency of the Visy secondary recycling process will be

determined in our study after extracting the post-consumer PET following each PET-

cleansing stage, as carried out previously during challenge tests (see Section 2.3.5.3).

Soxhlet extraction, sonication, total dissolution and headspace analysis (SPME and

static headspace analysis) will be the extraction techniques used to isolate the

contaminants from the polymer matrix prior to their instrumental analysis (see Section

61

Chapter 2

2.4 for a description of the extraction methods). The polymer will be ground to

different particle size ranges to facilitate extraction and all extracts will be run on a

GC/MS (which accounts for the volatile and semi-volatile contaminants) and

quantified using calibration standards. The involatile compounds cannot be

accounted for by these techniques. Therefore, some other means of analysis such as

HPLC should be considered in future work.

Throughout our study there will be a major focus drawn on extraction method

development in the quest to determine optimum extraction conditions. Time, particle

size and temperature are three external parameters that will be monitored for the

different extraction techniques used throughout this thesis. The effects of polymer

and “contaminant” properties on the level of extraction such as contaminant shape and

size as well as polymer crystallinity and orientation will be studied in order to obtain a

broad overview of the extraction process, which is comparable to migration.

References will be made to Equation 2.3, which relates the amount of contaminant

migrated to time and the diffusion coefficient.

In summary, the general aim of the thesis is to determine what contaminants are

present in recycled PET and whether the levels exceed the FDA threshold of

regulation for food-contact use. In fulfilling this aim, extraction method development

will be carried out involving the solvent extraction of different particle sizes for

different time intervals. Temperature optimisation will be carried out in the area of

headspace analysis. The effects of polymer crystallinity on extraction will also be

investigated.

2.6 OUTLINE OF THIS THESISThe results of this thesis are segregated into three

separate chapters (Chapter 4, 5 and 6). The first chapter of this series (Chapter 4)

involves the Soxhlet extraction and GC/MS screening of the washed and dried flake

for foreign components.

The chapter discusses the method development and the quantitative results found

from the extraction of semivolatile contaminants from washed and dried PET flake,

using Soxhlet extraction.

62

Chapter 2

The primary goal of Chapter 4 was to exhaustively extract contaminants from treated

curbside PET. Two parameters – particle size and time – were varied in search of the

optimum conditions. The Soxhlet extraction technique was validated by comparison

with total dissolution using TFA.

The second chapter of this series (chapter 5) presents the Soxhlet extraction kinetics

of annealed extruded pellets ground to three different particle size ranges. The effect

of particle size on the level of contaminant extracted was subsequently investigated.

In addition, the amorphous pellets were flattened using a hydraulic press (pressure = 8

tons) to reduce the path-length of the amorphous pellets prior to extraction.

Whole amorphous and annealed extruded pellets were extracted with time and the

effects of crystallinity on extraction and sorption kinetics were discussed.

The aim of Chapter 6 is to account for the volatile contaminants present in recycled

PET using headspace techniques (static headspace analysis and SPME). Extraction

temperature optimisation was the main focus of this chapter.

63

Chapter 3

CHAPTER 3 MATERIALS AND METHODS

3.1 METHOD FOR CHAPTER 4 3.1.1 Chemicals

The standards used are presented in Table 3.1. In this Table, Aldrich standards were

purchased from Milwaukee, WI, USA; Ajax standards from Auburn, NSW, Australia;

May and Baker standards from Dagenham, England; Lancaster standards from

Eastgate, White Lund, Morecambe, England; BDH Chemicals from Poole, England;

and Merck from Hohenbrunn, Germany.

Other chemicals purchased were dichloromethane (99.5%), heptane (99.5%), acetone

(99.5%), ethyl acetate (99.5%), isopropanol (99.7%), ethylene glycol (99%) and

trifluoroacetic acid (TFA, 99%). These solvents were manufactured by Merck

(Kilsyth, Victoria, Australia) whilst hexafluoroisopropanol (HFIP, 99.5%), hexane

(95%) and chloroform (99.8%) were manufactured by Fluka biochemika (Neu-Ulm,

Switzerland), LabScan (Bangkok, Thailand) and BDH chemicals (Poole, England)

respectively.

Table 3.1: Contaminants identified in washed and dried PET flake and the standards used.

Contaminant Retention Time

Selected Ion

Standard used Purity Company

2-Butoxyethanol 5.77 100 2-Butoxyethanol 95% Ajax Chemicals

1,2,4-, Trimethylbenzene

7.90 105 1,2,4-, Trimethylbenzene

99% Aldrich

m-Cymene*

8.82 119 o-Cymene* 99% Aldrich

(R)-(+)-Limonene

8.96 136 (R)-(+)-Limonene 97% Aldrich

Cineole

9.06 154 Cineole 99% Aldrich

γ-Terpinene

10.76 93 γ-Terpinene 97% Aldrich

3-Ethyl-o-xylene*

10.80 119 o-Cymene* 99% Aldrich

64

Chapter 3

1,2,3,5- Tetramethyl benzene-

11.91 119 1,2,3,5- Tetramethyl benzene-

80% Aldrich

(-)-Menthone

13.48 112 (-)-Menthone 90% Aldrich

Methyl salicylate 14.90 152 Methyl salicylate 99% May and Baker

Benzene, 1-methoxy, 4-(1-propenyl-)

15.18 148 Not quantified

4-n-Propylanisole

15.46 121 4-n-Propylanisole 96% Lancaster

Naphthalene

14.54 128 Naphthalene 99% Aldrich

Internal standard 1

14.54 136 Naphthalene-d8 99% Aldrich

n-Dodecane 15.33 170 n-Dodecane 99% BDH Chemicals

(S)-(+)-Carvone

17.08 82 (S)-(+)-Carvone 96% Aldrich

2-Methylnaphthalene 19.16 141 2-Methylnaphthalene

97% Aldrich

1- Methylnaphthalene 19.78 141 1-Methylnaphthalene

95% Aldrich

Biphenyl

22.73 154 Biphenyl 99% Merck

1-Ethylnaphthalene

23.33 141 1-Ethylnaphthalene 98% Aldrich

2,6- Dimethylnaphthalene

23.83 141 2,6-Dimethylnaphthalene

99% Aldrich

Tetradecane

23.92 198 Tetradecane 99% Aldrich

1,7- Dimethylnaphthalene

24.39 141 1,7-Dimethylnaphthalene

99% Aldrich

1,6- Dimethylnaphthalene

24.57 141 1,6- Dimethylnaphthalene

99% Aldrich

1,4- Dimethylnaphthalene

25.25 141 1,4- Dimethylnaphthalene

95% Aldrich

1,2- Dimethylnaphthalene

25.80 141 1,2- Dimethylnaphthalene

95% Aldrich

Cyclooctane, 1,5-dimethyl

27.07 55 Not quantified

Trimethylnaphthalene isomers

27.96-31.14

170 2,3,5-Trimethylnaphthalene standard used

98% Aldrich

65

Chapter 3

Hexadecane 28.12 57 Not quantified Benzoic acid, 4-methyl-, 2-methyl propyl-

29.41 119 Not quantified

Dodecanoic acid

30.88 129 Dodecanoic acid 99.5% Aldrich

n-Hexylbenzoate

31.47 105 n-Hexylbenzoate 98% Lancaster

Benzophenone

32.94 182 Benzophenone 99% Aldrich

Internal standard 2 32.94 110 Benzophenone-2,3,4,5,6-d5

99% Aldrich

* o-Cymene produces the same ion with similar intensity as m-Cymene and 3-Ethyl-o-xylene.

66

Chapter 3

3.1.2 Preparation of stock standards

Removable needle, teflon-tipped plunger, Hamilton (Reno, Nevada, USA) gastight

microsyringes of 25 µl; 50 µl; 100 µl; 250 µl; 500 µl; 1000 µl were used for all

sampling and transfers. Approximately 100 ppm solutions of each contaminant and

internal standard were prepared using an analytical balance and separate volumetric

flasks, using DCM as the solvent. Aliquots from each of these flasks were then pooled

together, resulting in one stock calibration standard (SCS) comprising all contaminants,

each at 1 ppm.

3.1.3 Soxhlet calibration standards

Appropriate amounts of the combined SCS were individually diluted for each

contaminant to eight different concentrations to create a calibration curve that

bracketed the unknowns. A 30 µl aliquot of a 75 ppm deuterated internal standard

solution was added before the calibration standards were made up to the mark with

DCM. The two internal standards (naphthalene-d8 and benzophenone-2,3,4,5,6-d5)

eluted at the extremes of the gas chromatogram and were structurally related to some of

the compounds identified in PET. The calibration standards were then concentrated in

the same way as the polymer extracts (see Section 3.1.8 “Soxhlet extraction

conditions”) and analysed by GC/MS. R2 for the standard curves exceeded the value of

0.99, which is approved by the U.S. FDA and the U.S. Environmental Protection

Agency (EPA) (US EPA, 1996).

3.1.4 Dissolution calibration standards

Different amounts of the 1 ppm stock calibration solution were added to separate

containers, each containing 20 ml TFA and 30 µl of the 75 ppm internal standard

solution. The spiked TFA was then extracted with 2 X 20 ml volumes of heptane. The

extracts were combined, washed with milli-Q water, dried over anhydrous sodium

sulphate and concentrated as for the PET samples (see Section 3.1.10 “Total

Dissolution Extraction Conditions”). R2 for the standard curves once again exceeded

0.99.

67

Chapter 3

3.1.5 Gas chromatography-mass spectroscopy (GC-MS) analysis

For the analysis of less polar contaminants, GC-MS was performed using a Hewlett

Packard 5890 Series II gas chromatograph coupled to a Hewlett Packard 5971A mass

spectrometer. Ultrahigh purity helium carrier gas (BOC gases, NSW, Australia) was

used. Qualitative analysis was achieved in TIC mode whilst SIM mode was required

for quantitative analysis for sensitivity and selectivity. The selected ions chosen for

each contaminant are presented in Table 3.1.

Gas chromatograph conditions: Column, DB-5MS, 30 m x 0.25mm, 0.25 µm film

thickness (J. & W. Scientific, USA); injection port temperature, 250°C. The initial

oven temperature was 35°C for 1min. Temperature programming was then employed

at 10°C/min until 70°C and at 3°C/min until 280°C (final hold time 25 mins).

MS conditions: solvent delay, 5 mins; temperature, 280°C.

For the analysis of more polar contaminants, GC/MS was performed using a Hewlett

Packard 6890 gas chromatograph coupled to a Hewlett Packard 5973 mass

spectrometer. Ultrahigh purity helium carrier gas (BOC gases, NSW, Australia) was

used. Qualitative analysis was achieved in TIC mode whilst SIM mode was required

for quantitative analysis. Only 2-butoxyethanol and ethylene glycol were quantified by

SIM (the selected ions were 100 and 31 for 2-butoxyethanol and 1,2-ethanediol

respectively).

Gas chromatograph conditions: Column, EC-Wax Econo-cap, 30 m x 0.25mm, 0.25

µm film thickness (Alltech Associates, Inc., Deerfield, IL); injection port temperature,

250°C; Split ratio, 1:1. The initial oven temperature was 60°C. Temperature

programming was then employed at 12°C/min until 220°C (final hold time 5 mins).

MS conditions: solvent delay, 5.4 mins; temperature, 250°C.

3.1.6 Commercial Visy1 treatment of curbside PET

The curbside stock, contaminated with polypropylene (PP) from closures and neck

rings, was shredded (by an industrial cutter) and intensely washed with detergent for

10-20 minutes at 75-95 °C under caustic conditions. After the flake was washed, the

1 The recycling process utilised by Visy is the subject of international patent application PCT/AU00/016131 and Australian patent application PQ2946.

68

Chapter 3

excess wash water was removed by centrifugation. Subjecting the flake to a sink-float

separation eliminated the PP and PE contaminants, which have a lower density than

PET.

The PET was then dried using air and elevated temperatures (140-185°C) for 5-8 hours.

3.1.7 Laboratory preparation of polymer before analysis

All analyses on washed and dried PET flake throughout our studies were performed on

samples taken from a single (approximately 15 kg) bag collected from the Visy

recycling plant (Reservoir, Vic, Australia) in June 2000. Each sample of washed and

dried PET flake (a random grab of approximately 70g) was cryogenically ground with

liquid nitrogen using a steel blender (CBL20, Breville, Botany, NSW, Australia).

Grinding involved chipping the flake against the walls of the blender. The ground

material was subsequently separated into three different particle size ranges using

manual sieves: 0-300 µm (small, around 4.6% of the total ground material by weight),

300-425 µm (medium, about 2.5%) and 425-700 µm (large, approximately 5.9%).

There was a large amount of ungrounded chipped flake remaining in the blender after

each grinding. (In the vicinity of 87% of the total material subjected to grinding). This

residue was not reduced in particle size, irrespective of the time of grinding.

3.1.8 Soxhlet extraction conditions

A cellulose Soxhlet extraction thimble (25 mm x 80 mm, Whatman, Maidstone,

England) was filled with 10 g of washed and dried PET flake ground to the small

particle size range (0-300µm). The ground flake was extracted for different times (8 h,

16 h, 24 h, 48 h and 72 h) using 200 ml of boiling DCM. Extractions were performed as

single measurements for 8 h and 16 h; duplicates for 24 h and 72 h; and triplicates for

48 h. A 50 µl microsyringe was used to transfer 30 µl of internal standard solution (75

ppm) to the extract before the DCM was distilled off to provide a 10 ml concentrate.

The extract was further concentrated to approximately 1 ml under a stream of nitrogen

and filtered through a SGE Teflon membrane syringe filter (pore size 0.45 µm,

Ringwood, Victoria, Australia) prior to GC-MS analysis.

Masses of the medium and large particle size ranges (7 g and 6 g respectively) were

analogously extracted for 24 h prior to further kinetic studies.

69

Chapter 3

3.1.9 Sonication

Small flake (1.5-5 g) was ground and sonicated for 3 h in 70 ml DCM with occasional

swirling to prevent aggregation. A 30 µL aliquot of internal standard solution (75 ppm)

was added to the PET/DCM mixture prior to sonication (Ultrasonics, Sydney, NSW,

Australia, 50Hz). The sonicated mixture was vacuum filtered with intermittent

washings using DCM. The mother liquor was concentrated and filtered as for the

Soxhlet extracts prior the GC-MS analysis.

3.1.10 Total dissolution extraction conditions

3.1.10.1 Total dissolution by TFA – Qualitative analysis

Extraction of the PET polymer by TFA dissolution for qualitative analysis involved

sonicating 10 g of polymer (virgin PET pellets, washed and dried flake, and extruded

pellets) in 50 ml of TFA until it completely dissolved. The TFA extract was then

extracted with three 35 ml portions of heptane, which were drawn off into another

container, washed twice with 50 ml of milli-Q water, passed over anhydrous sodium

sulphate and concentrated to 10 ml using a fractional distillation apparatus. The

volume of the concentrate was further reduced to 1 ml with a stream of nitrogen before

GC-MS analysis

3.1.10.2 Total dissolution by TFA – Quantitative analysis

Ground PET flake (1.5 g – 5 g) was dissolved in TFA. The volume of the latter was

equivalent to 10 times the mass of plastic. A 30 µl aliquot of internal standard solution

was added to the PET/TFA solution, which was subsequently sonicated until total

dissolution of the polymer (approximately 1 hour). The mixture was shaken with a

volume of heptane equivalent to that of TFA. The layers were allowed to separate

before drawing off the top layer (heptane) into another container. This procedure was

repeated once again and the extracts were combined.

The heptane layer was washed twice with milli-Q water, passed over anhydrous sodium

sulphate and concentrated to 10 ml using a fractional distillation apparatus. The

volume of the concentrate was further reduced to 1 ml with N2 before GC-MS analysis.

3.1.10.3 Total dissolution by HFIP – Qualitative analysis

HFIP (25 ml) was added to 10g of PET together with 50 ml of dichloromethane. The

mixture was sonicated for 2 hours before an additional 190 ml of dichloromethane was

70

Chapter 3

added. The extract was precipitated with 300 ml of methanol, filtered under vacuum

and the supernatant was evaporated to 10 ml by fractional distillation and then to 1 ml

under a stream of nitrogen before GC- MS analysis.

3.1.11 Crystallinity analysis

Differential scanning calorimetry (DSC) was performed in a nitrogen atmosphere on a

2920 DSC manufactured by TA instruments. The ramp conditions were: 10°C/min

from 50°C to 300°C.

X-ray diffraction patterns were collected on a Siemans D-5000 automated diffraction.

An aluminium sample holder and graphite monochromator (26.6 °) were used. Since

the SIROQUANT databank did not include the structure of PET, a graphite structure

was used to simulate the main peaks in the PET XRDs.

3.2 METHOD FOR CHAPTER 5

3.2.1. Chemicals

The chemicals used, procedures for the preparation of standards and curbside samples,

and, extraction and GC/MS analysis conditions were all as described in Section 3.1.

3 2.2 Crystallinity analysis

See Section 3.1.11.

3.3 METHOD FOR CHAPTER 6

3.3.1 Preparation of stock standards

Removable needle, teflon-tipped plunger, Hamilton gastight microsyringes (Reno,

Nevada, USA) of 25 µl; 50 µl; 100 µl; 250 µl; 500 µl; 1000 µl were used for all

sampling and transfers.

As described previously (Section 3.1.2), 100 ppm of each selected standard was

prepared separately in DCM. Appropriate volumes of these standards were then pooled

into one standard flask in order to produce a 1-ppm calibration standard. Additional

chemicals that were investigated by SHS in PET were undecane (>99%, Aldrich,

Milwaukee, WI, USA), toluene (>99%, Fluka chemika, Neu-Ulm, Switzerland) and p-

xylene (99+%, Sigma-Aldrich, Milwaukee, WI, USA).

71

Chapter 3

3.3.2 Soxhlet calibration standards for external standardisation

A five-point calibration curve was constructed from the pooled stock calibration

standard. Volumes of 20 µl, 50 µl, 75 µl, 100 µl, 150 µl and 250 µl were placed in

headspace 20 ml vials, half filled with glass beads (size = 2mm x 2mm) to simulate

similar headspace volumes as for the samples.

3.3.3 SPME

A manual SPME holder was used with four different fibre types; 100 µm and 7 µm

PDMS, 75 µm CX/PDMS, and 85 µm PA. All of these fibres including the holder were

purchased from Supelco (Bellefonte, PA, USA).

In the first experiment approximately 3 g of washed and dried flake (ground to 425µm –

700 µm) was placed in 25 ml static headspace vial and capped. The vial was heated in a

temperature-controlled paraffin oil bath set at 90°C for 15 minutes and then the

CX/PDMS fibre was immersed into the headspace and allowed to equilibrate for 30

minutes. The fibre was then subjected to GC/MS analysis. This procedure was

repeated for the remaining fibres and the results were qualitatively compared. The

CX/PDMS fibre was chosen for the extraction temperature optimisations.

Due to the heterogeneous nature of flake, temperature optimisations were performed on

the more homogenous annealed extruded pellets ground to three different particle size

ranges (0-300 µm; 300-425 µm and 425-700 µm), in a pursuit to determine the optimum

extraction temperature and particle size. Approximately 8 g of annealed ground and

unground pellets were placed in headspace sample vials and capped. In an attempt to

determine maximum extraction, heating was carried out at different temperatures (90°C,

130°C, 160°C, 194°C, 214°C, 238°C, 260°C) for one hour using a laboratory oven. The

samples were then immediately transferred into an oil bath thermostatted at 90°C and

the CX/PDMS fibre was immersed into the headspace for 30 minutes prior to GC/MS

analysis. The individual areas for each selected ion were plotted against extraction

temperature for the three ground particle sizes and the unground pellets. To confirm

the distinctive shapes of the curves, the thermodynamic procedure was repeated for

unannealed extruded pellets. In this case many more temperature measurements were

taken and the resultant extraction the thermodynamic were comparable to that of the

annealed pellets.

72

Chapter 3

In an attempt to reduce the degree of competitive adsorption, smaller quantities (0.3g

and 1 g) of ground annealed PET were analysed for different times. A shorter fibre

exposure time of 5 mins was also considered.

In order to determine whether similar competitive effects arise when using “absorption”

fibres, extraction thermodynamics were similarly studied on extruded PET pellets using

PDMS (100µm).

3.3.4 Static Headspace

With the CX/PDMS fibre undergoing competitive adsorption, another technique was

sought for the analysis of volatile contaminants in recycled PET. SHS does not

incorporate the use of a sorbent therefore it was selected as an alternative means of

extraction.

As for the SPME investigation, the first step to SHS analysis was to optimise the

extraction temperature in an attempt to attain exhaustive extraction. In this case 6 g of

pellets were heated for an hour at 70°C, 90°C, 120°C, 140°C, 180°C and 200°C and

then analysed by SHS. Since extraction was incomplete during the course of this study,

a set temperature and time were selected for quantitative analysis (See Section 3.3.5).

3.3.5 Quantitative analysis by SHS

Multiple headspace extraction was the first quantitative method investigated. This

method involved heating 0.3 g of ground washed and dried flake (425 µm –700 µm) at

180°C for 30 minutes and then analysing the headspace. The same sample was then

reheated at 30-minute intervals and re-analysed until the attainment of 190 mins.

Pellets and ground-annealed pellets (425 µm –700 µm) were extracted.

As unsatisfactory outcomes eventuated, no standardisation procedure was finalised for

the multiple headspace extraction method.

External standaridisation involved extracting pellets, ground flake, virgin PET and

external standards (glass beads spiked with undecane, toluene and p-xylene) at 180°C

for 20 minutes. All samples were analysed in triplicate.

73

Chapter 3

3.3.6 GC/MS Conditions – SPME

GC/MS was performed using a Hewlett Packard 5890 Series II gas chromatograph

coupled to a Hewlett Packard 5971A mass spectrometer. Ultrahigh purity helium

carrier gas (BOC gases, NSW, Australia) was used. Qualitative analysis was achieved

in TIC mode.

Gas chromatograph conditions: Column, DB-5MS, 30 m x 0.250mm, 0.25 µm film

thickness (J. & W. Scientific, USA); injection port temperature, 250°C. The initial

oven temperature was 35°C for 1min. Then temperature programming was employed

at 10°C/min until 70°C and at 3°C/min until 280°C (final hold time 25 mins).

MS conditions: solvent delay, 2 mins; temperature, 280°C.

3.3.7 GC/MS Conditions – SHS

Static Headspace GC/MS was performed using a Hewlett Packard 6890 gas

chromatograph coupled to a Hewlett Packard 5973 mass spectrometer and an

automated HS apparatus (HP 7694E) directly coupled to the GC. Ultrahigh purity

helium carrier gas (BOC gases, NSW, Australia) was used. Qualitative analysis was

achieved in TIC mode whilst SIM mode was required for quantitative analysis. The

selected ions (SI) were: 55, 78, 69, 91, 104, 105, 120, 77, 119, 67, 57, 115, 93, 128,

141, 121, 56, 81, 170, 154, 73, 98, 100. The compounds that were investigated were 1-

methylethyl benzene (SI = 105, 120), m-cymene (SI = 119), Cyclopentane, 1-methyl-2-

propyl- (SI = 57), 1,2,4-trimethylbenzene (SI = 105), naphthalene (SI = 128), 2,4,6-

trimethyloctane (SI = 57), benzaldehyde (SI =105, 120), 3-ethyl-o-xylene (SI = 119),

cineole (SI = 154), propylanisole (SI = 121), biphenyl (SI = 154), propylbenzene (SI =

91, 120), 2-ethylfuran (SI = 81), limonene (SI = 69), toluene (SI = 91) and 1,2,3,5-

tetramethylbenzene (SI = 119) and p-xylene (SI = 91).

Gas chromatograph conditions: Column, HP-5MS, 30 m x 0.25mm, 0.25 µm film

thickness (Alltech); injection port temperature, 250°C; Split ratio, 2:1. The initial oven

temperature was 35°C (initial time: 1 minute). Temperature programming was then

employed at 20°C/min until 270°C (final hold time 2 mins).

MS conditions: solvent delay, 2 minsϕ; temperature, 250°C.

ϕ A solvent delay was included to exclude the air peak from the chromatogram.

74

Chapter 4

CHAPTER 4 SEMI-VOLATILE CONTAMINANTS AND LEVELS OF OCCURRENCE IN

WASHED AND DRIED SHREDDED PET

4.1 GENERAL INTRODUCTION

4.1.1 Purpose of the chapter

The general purpose of the work in the following chapter was to determine which

contaminants are present in washed and dried shredded PET (flake) obtained from curbside

collection and to determine whether their concentrations are above the US FDA threshold

of 215 ppb. The steps undertaken in achieving this goal – qualitative analysis and

extraction method development - have been segregated into two sections (Section 4.2 and

Section 4.3) of the existing chapter.

The quantitative results presented in this chapter (for the extraction of washed and dried

flake) are the first step in determining whether recycled PET is suitable for food contact

use. In subsequent chapters extraction of contaminants from further treated (extruded) PET

will be addressed and each recycling stage will be assessed with respect to its

decontamination efficiency, as previously carried out during challenge tests (Franz and

Welle 1999a, Harding et al. unpublished).

It is presumed that the concentrations of the semi-volatile and non-volatile contaminants in

recycled PET would represent a “worst case scenario” surpassing the levels of volatile

contaminants, which are more likely to be removed by the high temperatures and vacuum

extrusion used during recycling.

Therefore, from a consumer-safety perspective, it would conceivably be of greater

significance to recognize the levels of semi-volatiles and non-volatiles in recycled PET.

Soxhlet extraction, ASE, MAE and SFE are extraction methods generally used to extract

semi-volatiles and non-volatiles from polymers. Throughout this thesis the Soxhlet

extraction method has been selected as the principal extraction technique alongside GC/MS

analysis such that the presence of semi-volatile compounds in washed and dried flake could

be identified and quantified. The analysis of non-volatile contaminants in post-consumer

PET will not be initiated, as the gas chromatograph is limited to the analysis of volatile and

75

Chapter 4

semi-volatile compounds. In future work it may be desirable to analyze non-volatile

compounds in recycled PET plastic by liquid chromatography, even though the migration

of large compounds into food is slow relative to the diffusion of smaller, more volatile

compounds through the polymer matrix.

4.1.2 Brief outline of chapter

The first main theme of this chapter (Section 4.2) discusses the Soxhlet extraction and

GC/MS screening of the washed and dried flake for foreign components. Dichloromethane,

the recommended extraction solvent for PET, was used because it is known to swell the

polymer matrix and potentially enhance the diffusion of its constituents (Feigenbaum et al.

2002). This behaviour is attributed to the solubility parameter similarity between DCM and

PET (Vandenburg et al. 1999, St. Küppers 1992) and the large diffusion coefficient of

DCM in PET (Sadler et al. 1996). Huber and Franz (1997a) employed a similar extraction

and screening method to identify the contaminants in recycled (ground, washed and

extruded) HDPE produced from household waste.

Comparisons of the post-consumer PET extracts with the virgin PET extracts were made in

order to decide which components are foreign and which are hereditary to the polymer, and

hence not regarded as post-consumer contaminants.

Two total dissolution techniques were employed in order to theoretically extract any

contaminants that could be entrapped in the polymer matrix and therefore not accessible by

the Soxhlet extraction solvent. It was proposed that highly interactive solvents such as

TFA and HFIP extract contaminants out of the polymer matrix more indiscriminately.

The second main theme of this chapter (Section 4.3) discusses the method development and

the quantitative results found from the extraction of semivolatile contaminants from washed

and dried PET flake, using Soxhlet extraction.

The primary goal of Section 4.3 was to exhaustively extract contaminants from treated

curbside PET. Two parameters – particle size and time – were varied in search of the

optimum conditions. The Soxhlet extraction technique was validated by comparison with

total dissolution using TFA.

76

Chapter 4

4.1.3 Selecting the right extraction solvent for Soxhlet extraction

Some other solvents, which are expected to be suitable extraction solvents as a result of

their compatible solubility coefficients with PET, are acetone, xylene, 1,4-dioxane,

chloroform, ethyl acetate and meta-cresol (Table 4.1). Although the swelling action of

acetone has been documented (Moore and Sheldon 1961, Begley et al. 2002) and xylene,

1,4-dioxane and chloroform have been used to extract oligomers from PET (Hudgins et al.

1978, Costley et al. 1997, Goodman and Nesbitt 1960), the higher boiling points of these

solvents, relative to dichloromethane, was the initial deterrent to their use during the current

investigation. Low boiling point extraction solvents were sought to prevent the loss of

analytes through volatisation during extraction and solvent removal (evaporation) prior to

GC/MS analysis.

Table 4.1: Hildebrand solubility parameters of some solvents and PET (Vandenburg et al.

1999, Brandrup and Immergut 1989, Weast and Melvin 1979).

Material Solubility

parameter (δ

Mpa1/2)

Boiling point (°C)

Xylene 18.2 140

Ethyl acetate 18.6 77

Chloroform 19.0 61

Dichloromethane 19.8 40

Acetone 20.3 56

1,4- Dioxane 20.5 101

m-Cresol 20.9 203

PET 20.5

Nonetheless, from Table 4.1, the boiling points of acetone and chloroform are reasonably

low (56°C and 61°C respectively), even though they exceed the boiling point of DCM

(40°C). As a result, the likelihood of semi-volatile contaminant loss would still be

expected to be negligible during extraction using acetone and chloroform. However,

77

Chapter 4

acetone and chloroform were not selected as extraction solvents in preference to DCM

because:

1. Costley et al. (1997) reported that DCM was more efficient than acetone during the

microwave extraction of cyclic trimer out of PET. St Küppers (1992) made an analogous

observation whilst extracting cyclic trimer from PET by SFE using four different modifiers

(methanol, isopropyl alcohol, DCM and acetone). Only the DCM modifier gave an

increase in the amount of trimer extracted.

2. Chloroform has a greater molecular weight than DCM, and therefore was expected to

diffuse through PET at a slower rate.

3. Besides the higher boiling point of ethyl acetate compared to DCM (77°C and 40°C

respectively), another reason why ethyl acetate was not considered as the prime extraction

solvent in preference to DCM was that according to Vandenburg et al. (1999) and Costley

et al. (1997), DCM swelled PET to the point of fusion at a lower temperature (120°C) than

that of ethyl acetate (190°C) under high pressure conditions. This led to the assumption that

DCM could be more powerful than ethyl acetate in swelling PET.

In the extraction of non-volatile components from PET, the use of higher boiling solvents

could be advantageous during Soxhlet extraction, especially if the solvents boil above the

Tg of the polymer (69°C for PET). Above the Tg, a glassy polymer becomes rubbery,

which facilitates extraction through the increased movement of polymer chains. Since the

Tg is reduced when a polymer is plasticized, a solvent that swells PET but boils a few

degrees below the original Tg could also be effective. Alternatively, the extraction

temperature could be increased above the boiling point of the solvent in an accelerated

solvent extractor, without the loss of volatiles. Due to the unavailability of automated

extraction techniques such as ASE, SFE and MAE, solvent extraction was performed by

means of manual extraction methods (e.g. total dissolution and Soxhlet extraction).

78

Chapter 4

4.2 QUALITATIVE STUDY OF CONTAMINANTS IN WASHED AND DRIED PET

FLAKE

4.2.1 Introduction to Soxhlet extraction of washed and dried flake

In order to efficiently extract PET by Soxhlet extraction, an appropriate solvent must be

used; one which swells the polymer matrix and liberates the potential migrants. The

extracted constituents could then be separated and identified by GC/MS.

As mentioned in Section 4.1.3, a relatively low boiling solvent with a solubility parameter

similar to that of PET is the initial criterion for selecting the right solvent for the extraction

of semi-volatiles. However, the solubility parameter is not the only decisive factor that

impinges on the extent of polymer swelling. Other contributing factors are solvent

molecular shape, size and affinity towards the polymer functional groups. In Section 4.1.3,

it was reasoned from previous extraction studies that DCM would be a more suitable

extraction solvent than ethyl acetate, chloroform and acetone, despite all of these solvents

having similar solubility parameters to that of PET (Table 4.1). It is believed that DCM is

more efficient in swelling PET than the other solvents and the reasons behind this

presumption are not solely related to the solubility parameter.

4.2.2 Choosing a suitable low boiling solvent

In order to confirm the hypothesis that DCM swells PET more readily than three solvents

with similar δ to PET (acetone, ethyl acetate, chloroform) and three solvents with dissimilar

δ to PET (hexane, ethanol, 2-propanol), a simple gravimetric experiment was undertaken

during which extruded PET pellets were subjected to each of the test solvents and the

polymer was weighed after a particular exposure time. Similar gravimetric swelling

experiments have been performed for PET in the past (Moore and Sheldon 1961, Begley et

al. 2002, Jameel et al. 1981). The disadvantage of weight uptake measurements is that they

can be altered by weight loss due to the migration of polymer constituents (Feigenbaum et

al. 1991). It is assumed that the level of migration is low relative to the extent of sorption

due to PET being relatively free of mobile compounds such as additives (Sauvant et al.

1995, Castle et al. 1989).

79

Chapter 4

The solvent swelling effects of PET pellets were visually identified for solvents with

solubility parameters similar to that of PET (dichloromethane, acetone, ethyl acetate and

chloroform). These solvents lead to solvent-induced crystallization of the amorphous

pellets thus causing them to transform from transparent to opaque.

DCM and chloroform produced the greatest degree of opaqueness of all the solvents tested.

The color change instigated by the sorption of acetone and ethyl acetate was minor relative

to the chlorinated solvents. No color change was observed for isopropanol, hexane and

ethanol, emphasizing the lack of interaction with the polymer.

The solvent induced crystallization of PET by acetone and dimethylformamide have been

reported in the past (Jameel et al. 1981, Ouyang et al. 1998, Moore and Sheldon 1961).

Crystallinity in polymer swelling solvents theoretically occurs because the solute lowers the

Tg and provides enough mobility to the polymer chain segments for them to form crystals

(Liu and Neogi 1992).

Figure 4.1 is a plot of the number of mmole of solvent absorbed at 3 h (Soxhlet extraction)

versus the difference in solubility parameters between PET and the corresponding solvent

(δPET - δsolvent). It is interesting to note that although the solubility parameter difference

between PET and acetone is smaller than that between PET and any other solvent, the

number of mmoles of acetone absorbed by PET is six times lower than for DCM and over 4

times lower than for chloroform. A possible explanation for the stronger affinity of DCM

towards PET could be attributed to the smaller molecular size of DCM compared with

acetone. Using a computer molecular modeling program (Sybyl 6.8), the volume of acetone

(59.6 A3) was calculated to be 10.2 A3 greater than that of DCM (49.4 A3).

80

Chapter 4

Figure 4.1: Plot of the number of mmole of solvent absorbed at 3 h versus (δPET - δsolvent).

0

10

20

30

40

50

60

-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7

PET - solvent

amou

nt s

orbe

d (m

mol

es)

Hexane Acetone Dichloromethane ChloroformEthyl Acetate 2-Propanol Ethanol

Since chloroform (60.8 A3) and acetone (59.6 A3) were calculated to have similar

molecular sizes, the rationale behind chloroform’s superior sorption by PET cannot be

linked to size. A more credible explanation is related to functional groups’ interactions.

Chloroform has acidic properties in the Lewis sense and could potentially solvate the basic

carbonyl polymer groups. Moore and Sheldon (1961) have already demonstrated that

hydrogen bonding between acidic liquids and basic polymer groups cause greater polymer

swelling than analogous interactions between basic ketones such as acetone and the acidic

hydrogen atoms of the CH2 groups adjacent to oxygen atoms in PET. The aggressiveness

of chlorinated and fluorinated compounds (e.g. trichloroethane, chlorobenzene,

trichloroacetic acid, HFIP and TFA) towards PET has been documented (Demertzis et al.

1997, Franz and Welle 2002, Chidambaram et al. 2003); therefore the superior sorption

action of DCM and chloroform by PET is not surprising. One exception, where contact

with a chlorinated compound did not result in any swelling was with carbon tetrachloride

(Moore and Sheldon 1961). This solute did not absorb into PET presumably because of its

large molecular size, low solubility parameter relative to PET (∆δ = 2.9 Mpa1/2) and the

absence of an “acidic” proton to interact with “basic” polymer groups.

81

Chapter 4

Ethylacetate, which is clearly larger than acetone, chloroform, DCM and has a basic ketone

group and larger ∆δ, does not absorb appreciably.

Swelling occurs in liquids whose solubility parameters do not differ vastly. It is likely that

the remaining solvents in Figure 4.1 (hexane, 2-propanol, ethanol) did not absorb into PET

efficiently because of the large differences in solubility parameters (Table 4.2). The

difference in solubility parameter (∆δ) between each of the solvents in Table 4.2 and PET

exceeds 3, which is considered significant in the interactive sense.

Table 4.2: Hildebrand solubility parameters of hexane, 2-propanol, ethanol and PET

(Vandenburg et al. 1999).

Material Solubility parameter (Mpa1/2)

Hexane 14.9

2-propanol 23.8

Ethanol 26.0

PET 20.5

Refluxing the pellets (7 g) with equal volumes of DCM, chloroform, ethyl acetate and

acetone and weighing the pellets after a 10-minute exposure time provided a similar

sorption order to that obtained by a 3 h Soxhlet extraction (with one exception). The

number of moles of DCM, chloroform, ethyl acetate and acetone sorbed were 19.3 mmol,

14.8 mmol, 2.80 mmol and 2.51 mmol, respectively. Therefore, the amount of DCM and

chloroform absorbed during reflux exceeded the amount of ethyl acetate and acetone

absorbed, as for the Soxhlet extraction.

However, the amount of ethyl acetate absorbed approximated the amount of acetone sorbed

during reflux. In contrast, during Soxhlet extraction, the level of acetone absorbed (7.76

mmole) exceeded the level of ethyl acetate absorbed (4.31 mmole) by the pellets. During

Soxhlet extraction, the extraction temperature may not be equivalent to the boiling point of

the solvent, since the boiled solvent condenses before dropping into the extraction thimble.

Therefore, the extraction temperature is not expected to be a major parameter controlling

the differences in the sorption of acetone and ethyl acetate during Soxhlet extraction.

82

Chapter 4

Instead, the solubility difference (∆δ) and sorbent molecular size could be affecting the

amount of ethyl acetate and acetone sorbed during Soxhlet extraction. Since ∆δ and

molecular size are smaller for acetone (∆δ = 0.2 Mpa1/2) than is for ethyl acetate (∆δ = 1.9

Mpa1/2), the level of acetone absorbed by the pellets is higher.

During reflux (as opposed to Soxhlet extraction), the extraction temperature is equivalent to

the boiling point of the solvent, since the polymer pellets are in direct contact with the

refluxing solvent. Therefore, both ∆δ and extraction temperature could have an effect on

the degree of sorption during reflux. Since the boiling point of ethyl acetate (77°C) is

higher than that of acetone (56°C), the amount of ethyl acetate absorbed approximated the

amount of acetone absorbed, despite the latter solvent having a smaller ∆δ and molecular

size.

An observation that was made during the reflux sorption analysis was that the pellets

clumped together and turned white after the addition of DCM or chloroform. However,

when the pellets were in contact with ethyl acetate and acetone, they were slightly opaque

and separated from one another in solution. The DCM and chloroform reflux extracts were

either cloudy (chloroform) or contained white solid particles (DCM), whereas the ethyl

acetate and acetone were clear. The white particulates could be extracted oligomers, which

may be more soluble in chloroform than in DCM. Weighing the white solid particles after

evaporating off the solvent supported the supposition that chloroform extracted more

oligomers from PET than DCM, possibly due to solubility effects. The percentage (by

mass) of solid extracted was 0.46%, 1.13%, 0.05%, and 0.06% for DCM, chloroform, ethyl

acetate and acetone respectively.

Since DCM showed the best swelling character during the sorption studies, it was selected

as the solvent for the Soxhlet extraction of washed and dried flake. A solvent that swells

the polymer appreciably would presumably extract the “freely diffusible” substances which

are dissolved in the amorphous region of the polymer matrix. However, “sorbed” or

“bound” substances localized at active sites or “holes” within the glassy polymeric matrix

may not migrate into the extraction solvent. Total dissolution using aggressive solvents

such as TFA and HFIP would theoretically dissipate the polymer chains away from each

other and extract the entrapped components. However, past researchers have not

83

Chapter 4

investigated the efficiency of total dissolution towards extracting constituents confined in

“holes” of the polymer matrix. There is the chance that some contaminants may be

entrapped in “holes” of entangled chains or crystallites, whose attraction forces cannot be

overwhelmed by interaction with the solvent. Undissipated polymer chains may not be

visually identified, hence just because the polymer appears totally dissolved, in reality there

may be some polymer chains grouped together.

The presence of PEN as a co-polymer of PET could further complicate extraction. PEN

increases glass transition temperature, improves static chain packing and decreases local

segmental mobility (Mc Dowell et al. 1998). The reduction in local scale mobility and

rotational motion of the bulky, stiff 2,6-naphthalate units results in a lower free volume and

thus lower diffusion coefficients of penetrants and migrants (Mc Dowell et al. 1998). Also

the strong intermolecular forces between chains of PEN may inhibit solvent penetration and

therefore the extraction of molecules that are “entrapped” in cavities between the PEN

chains. In this case, a contaminant could be residing in a cavity waiting for an unlikely

neck opening to another cavity. Mc Dowell et al. (1998) demonstrated that the diffusivity

and solubility of acetone in PET/PEN decreased with the increased level of PEN. In

addition, samples rich in PEN could not be dissolved in TFA. The strong intermolecular

bonds may be preventing solvent molecules from establishing adequate interactions with

the whole polymer, thus preventing the molecules being carried off into solution.

The components entrapped in the polymer are likely to be those originating from the

original manufacture of the polymer. These could be localised between the biaxially

orientated chains during bottle manufacture. Post-consumer contaminants are likely to

diffuse throughout the amorphous region of the polymer and therefore have the potential to

be extracted by DCM, unless washing and drying at mild temperatures leads to thermally

induced crystallization, thus potentially trapping the formerly “diffusible” substances.

Theoretically, diffusion is not feasible in crystalline domains of the polymer matrix.

4.2.3 GC/MS analysis of DCM extracts of washed and dried flake

Qualitative analysis of washed and dried flake involved extracting the ground flake (0-300

µm) by Soxhlet for 24 h and analyzing the concentrated DCM extract by GC/MS. The GC

84

Chapter 4

chromatograms of the flake and blank extracts were superimposed and the peaks with mass

spectra unique to the washed and dried flake are indicated in Figure 4.2, which is a

chromatogram of the flake extract. It was anticipated that by grinding the polymer to a

smaller particle size, the path length for contaminant extraction would be reduced. This

would result in increased extraction efficiency and henceforth, improved detection of the

extracted contaminants.

Figure 4.2: Chromatogram of DCM extract for washed and dried flake.

The peak numbers in Figure 4.2 correspond to the compounds in Table 4.3. The PET

cyclic dimer and its corresponding ether eluted after 30 minutes. The peaks that were not

assigned in Figure 4.2 were present in the blank and therefore are not exclusive to the flake.

85

Chapter 4

Table 4.3: Compounds identified in ground washed and dried PET flake [“x” denotes

presence of compound in virgin (V) and recycled (R) PET].

Contaminant V R Contaminant V R 1 2-Butoxyethanol 20 Biphenyl# x 2 1,2,4-

Trimethylbenzene# 21 1-Ethylnaphthalene# x

3 m-Cymene x 22 Benzoic acid, butyl ester

x x

4 (R)-(+)-Limonene# x 23 2,6-Dimethylnaphthalene#

x

5 1,8-Cineole# 24 Tetradecane# 6 γ-Terpinene# 25 Benzoic acid, 4-

methyl-, 2-methyl propyl-

x

7 Nonanal# x 26 1,7-Dimethylnaphthalene#

8 3-Ethyl-o-xylene 27 1,6-Dimethylnaphthalene#

9 1,2,3,5-Tetramethyl benzene-#

28 1,4-Dimethylnaphthlalene#

10 (1)-Menthone# 29 1,2-Dimethylnaphthalene#

11 Methyl salicylate# x 30 Cyclooctane, 1,5-dimethyl-

12 Benzene, 1-methoxy, 4-(1-

propynyl-)

31 Trimethylnaphthalene isomers (5 peaks)

x

13 4-n-Propylanisole# x 32 Ethanol, 2-[4-(1,1-dimethylethyl-2-methylpenoxy]-

x x

14 Naphthalene# x 33 Hexadecane

15 Benzoic acid x 34 Dodecanoic acid# 16 n-Dodecane# 35 n-Hexylbenzoate# 17 (S)-(+)-Carvone# 36 Benzophenone# x 18 2-

Methylnaphthalene# x 37 Cyclic dimer x x

19 1-Methylnaphthalene#

x 38 Cyclic dimer ether x x

# Identity confirmed by retention time of a standard.

The large number of peaks in the blank corresponds to the impurities in dichloromethane

and the thimble intensifying after a 200-fold concentration step. The presence of

86

Chapter 4

interfering peaks during chromatography is one of the disadvantages of solvent extraction.

Thermal extraction methods (e.g. static headspace, SPME, thermal desorption analysis) are

employed to circumvent solvent interferences and reduce sample preparation and extraction

time. However, they are not as suitable as the solvent extraction methods for the analysis

of semi-volatiles and non-volatiles. Automated solid-fluid techniques such as SFE and

ASE utilize smaller amounts of solvent compared with manual techniques such as Soxhlet

extraction and ultrasonication, therefore are expected to generate extracts with less

interference. Despite this advantage, Soxhlet extraction was selected as the prime

extraction method due to the unavailability of the alternatives. The fact that Soxhlet

extraction involves the replenishment of the sample compartment with fresh extraction

solvent at regular time intervals makes it more appealing than ultrasonication, whose

extraction efficiency could succumb to the effects of a decreasing concentration gradient

with time.

In addition, Soxhlet extraction of non-volatile components using a high boiling solvent,

could be more beneficial than ultrasonication, whose extraction temperature would not

reach that imposed by Soxhlet extraction. This theory assumes the effects of temperature

on diffusion and solubility during solvent extraction is greater than that of ultrasonic waves.

The compounds that were identified in ground virgin PET (particle size = 0-300 µm) or

recycled (extruded) PET pellets are denoted by an “x” in Table 4.3. Those compounds that

are not detected in the virgin polymer are considered foreign (“contaminants”) to the post-

consumer PET. As indicated in Table 4.3, there were 34 contaminant peaks exclusively

identified in the post-consumer PET. Five trimethylnaphthalene isomers were identified,

however they are grouped together as “trimethylnaphthalene isomers” (peak 31) because

they were not unequivocally distinguished from one another. Due to the similarity in mass

spectra for the trimethylnaphthalene isomers, there was not enough confidence in the MS

library search results to distinguish between the isomers. Ideally, standards should be run

to unequivocally distinguish between the isomers.

This large number of “contaminants” identified (34) is not surprising due to the permeable

nature of PET (relative to glass) and the history of the pre-used containers. However, Huber

and Franz (1997a) extracted 74 contaminants from extruded HDPE granules by the same

87

Chapter 4

extraction method and GC/MS. This result underlines the higher permeability of HDPE

relative to PET and/or suggests that a more diverse range of packaging types enter the

HDPE recycling stream providing alternative sources of contamination relative to the PET

recycling stream, whose feed is generally limited to beverage bottles. Devlieghere et al.

(1997) also observed that polyolefins (PP, HDPE) and PVC retained significantly more

contaminants compared with the more polar polymers (PC and PET). It is presumed that

the favorable cohesive energy, polarity, intermolecular packing and high Tg of PET provide

superior resistance to contaminant sorption (Gavara et al. 1997, Arora and Halek et al.

1994, Van Willige et al. 2002, Paik 1992).

Table 4.3 implies that virgin PET is relatively free of semi-volatile migratory compounds.

This finding supports the well-established fact that virgin PET is generally free of additives

compared with other, less stable polymers, which require additives to maintain their

stability and improve their flexibility. In fact Eastman, the manufacturer of the virgin PET

analysed in the current study, have acknowledged the absence of additives in their polymer.

Therefore the presence of mobile components in virgin PET is only expected to arise from

the polymer’s original manufacture (e.g. starting materials, monomers, reaction by-

products), as already presented by researchers (St Küppers 1992, Begley and Hollifield

1989).

Although PET has the reputation for being generally free of additives and other mobile

compounds, there have been some solvent extraction studies (e.g. maceration with

dichloromethane; Soxhlet extraction with ethanol) that have found hydrocarbons, BHT,

BHA, Erucamide, Tinuvin P, phthalates and adipates in PET bottles (Monteiro et al. 1996,

Kim et al. 1990, Van Lierop 1997). As a result, PET obtained from different manufacturers

could contain different mobile constituents, depending on what the producer adds to the

polymer during its manufacture. Provided these constituents are not transferred into food at

levels that could endanger human health or affect the quality of food, their presence in PET

is not of great concern.

Since the post-consumer PET bottles that enter the recycling stream derive from an

extensive range of virgin PET manufacturers, some of the compounds identified in the

washed and dried flake, but not in our virgin PET, could nonetheless arise from the original

88

Chapter 4

manufacture of the polymer, depending on its source. However, it is impossible to analyze

the different virgin PET available in order to unequivocally verify whether the migrants

derive from post-consumer use or original polymer manufacture.

As presented in Table 4.3, two low molecular weight cyclic oligomers (peaks 37 and 38 in

Table 4.3), possibly formed during melt polymerization, were identified in the virgin and

post-consumer PET. These oligomers were presumed to be that of the cyclic dimer (peak

37) and dimer ether (peak 38), whose mass spectra and structures are shown in Figures 4.3

(a) and (b). The mass spectra for these peaks are not present in the spectral library.

However, they are closely matched with those reported by Buiarelli et al. (1993) and

Monteiro et al. (1996). These authors suggested that the ions of m/e 341 and m/e 385 in

the mass spectra could be due to M-43 (i.e., M – CH2CHO). Another prominent ion in both

spectra was m/e 296, which could be due to the (CO [C6H4COOC H2]2) fragment.

The remaining two compounds identified in virgin PET (peaks 22 and 32, Table 4.3) could

have possibly resulted from the degradation, oxidation, recombination and rearrangement

of the starting materials. Their library match results were only about 72 and 40 per cent for

peaks 22 and 32 respectively, therefore the identification of these compounds was not

reliable and speculation on their origin will not be pursued. Conversely, the assignment of

the “contaminant” peaks in Table 4.3 by computer matching on the MS database was more

reliable. For all significant peaks, the highest probability match was assumed for the

assignments of identity by the database. For 83% of peaks there was a greater than 90%

confidence limit, whilst for 13% the confidence limit was 80-90% and for the remaining

4% the confidence limit was <80% but >70%. Standards were purchased for most of these

compounds and identity was confirmed by coincidence of retention time. This was

especially important for the naphthalene isomers, which had different retention times but

similar mass spectra. As mentioned earlier, of all the naphthalene isomers identified, only

the trimethylnaphthalene isomers were not distinguished from one another by means of

standards.

89

Chapter 4

Figures 4.3: Mass spectrum and structure of (a) cyclic dimer and (b) dimer ether.

(a)

O

CH2CH2 O

C O

O

C

C O C O

O CH2CH2 O

(b)

O

CH2CH2 O

C O

O

C

C OO

C OO

CH2 CH2 O CH2 CH2

The number and the relative area of contaminants identified in recycled PET were

reproducibly lower than that in the ground flake. At this stage, the reasons behind this

observation could be two-fold: (a) the extraction efficiency from extruded pellets is lower

than that from the powdered flake, and (b) the extrusion step of the recycling process

further reduces the level of contaminants in the post-consumer PET. As concluded from

90

Chapter 4

subsequent chapters of the current thesis, the likely reason for this observation is (b), seeing

that the Soxhlet extraction method has been demonstrated to be exhaustive for “freely

diffusible” contaminants in PET. Extrusion has already been shown to be effective in

reducing the levels of contaminants in post-consumer washed and dried PET, especially the

levels of volatile contaminants, which readily volatilise under extrusion conditions

(Triantafyllou et al. 2002, Harding et al. unpublished).

Compounds directly involved in polymer formation such as ethylene glycol, terephthalic

acid, dimethyl terphthalate, BHET and MHET were not identified in the Soxhlet extracts

for one or a combination of the following reasons:

(a) The compounds are too polar (e.g. alcohols and carboxylic acids), therefore must be

derivatised and/or run on a more polar column (e.g. EC-Wax Econo-cap column) to

improve GC/MS sensitivity (Kim et al. 1990, Atkinson et al. 1971). For example,

dicarboxylic acids tend to give poor gas chromatographic peaks, broadening and

tailing and are therefore derivatised to their trimethylsilyl or methyl derivatives

(Atkinson et al. 1971, Gramshaw 1995). Previously, HPLC has been a successful

technique for analyzing all of these compounds - other than ethylene glycol, which

does not posses a chromophore - without derivatisation (Begley and Hollifield

1989).

(b) The components are entrapped within the polymer matrix and cannot be

accessed/extracted by the solvent. Total dissolution techniques are employed to

disperse the polymer matrix and extract these compounds.

(c) The level of these compounds is below the limit of detection.

(d) The compounds may be held in the polymer matrix by hydrogen bonding, whose

attraction forces cannot be overcome by the presence of a non-polar or weakly polar

extraction solvent.

91

Chapter 4

(e) Interfering peaks may have obscured the presence of these compounds in the

GC/MS chromatogram. An extraction method that uses smaller quantities of

solvent (e.g ASE or SFE) or no solvent (e.g. thermal extraction techniques) should

be used to exclude interfering peaks from analysis.

Other compounds that have been previously detected in virgin PET are p-xylene

acetaldehyde, high molecular weight oligomers and metal catalysts. These compounds

were not identified in our extracts for the following reasons:

(a) High molecular weight oligomers are non-volatile and therefore unlikely to be

identified by GC/MS. HPLC is normally used to separate and analyse oligomers

(Begley and Hollifield 1989). In some cases oligomers can be broken down by

hydrolysis; methylation of the hydrolysed extracts then makes them appropriate for

GC analysis (Gramshaw et al. 1995).

(b) Acetaldehyde is very volatile (boiling point = 21ºC), therefore volatilises during

extraction, prior to GC/MS analysis. Acetaldehyde is normally analysed by static

headspace to avoid loss during extraction (Dong et al. 1980, Wyatt 1983, Franz and

Welle 1999b).

(c) p-Xylene could be co-eluting with the solvent peak, which is not analysed by the

MS. Thermal extraction techniques (e.g. static headspace, SPME) could be used to

analyse p-xylene and other volatile compounds that co-elute with the solvent peak.

(d) Metals cannot be identified by GC/MS. Spectroscopic methods such as atomic

absorption (AA), or inductively coupled plasma (ICP) must be employed for the

analysis of metals in the acid-digested extract of PET.

These points underline the limitations of the Soxhlet extraction technique and GC/MS

method and call for a diverse range of techniques to be used in order to cover all types of

constituents in PET.

92

Chapter 4

4.2.4 Qualitative analysis of washed and dried flake extracted by total dissolution

In order to theoretically extract any compounds resisting Soxhlet extraction from the PET

matrix, the washed and dried flake was first dissolved in either TFA (for the analysis of

non-polar compounds) or HFIP (for the analysis of polar compounds) to disperse the

polymer matrix and theoretically release the entrapped contaminants. The components

were then solvent extracted with heptane for the analysis of non-polar compounds in the

case of TFA dissolution. For HFIP, methanol was added to re-precipitate the polymer prior

to GC/MS analysis, as outlined in Chapter 3 (Section 3.1.10.3).

The dissolution technique has been readily used for the extraction of surrogate compounds

from PET during challenge tests (Komolprasert and Lawson 1995, Harding et al.

unpublished). Dissolution of polymers involves two stages. Firstly, solvent molecules

gradually diffuse into the polymer producing a swollen gel. This gel then gradually

disintegrates as yet more solvent enters the gel and as the molecules of solvated polymer

gradually leave the gel and are carried out into solution. This latter stage is sped up by

ultrasonication, which has been used in our method.

Cross-linking, crystallinity, hydrogen bonding and the absence of chain branching could

hinder the polymer dissolution (Nicholson 1991). Therefore, it is unclear whether the

dissolution procedure is 100% effective towards extracting contaminants from PET; the

contaminants could be residing in “holes” between polymer chains, which may not be

readily accessed by the solvent. However, one can assume the extraction efficiency for the

total dissolution is at least the same, if not greater, than that for Soxhlet extraction using

non-dissolving solvents such as DCM.

Table 4.4 lists the compounds extracted from washed and dried flake by total dissolution

using TFA and followed by liquid-liquid extraction with heptane.

93

Chapter 4

Table 4.4: Compounds extracted from washed and dried flake by total dissolution using

TFA/heptane [“x” denotes presence of contaminant in virgin (V) and recycled (R) PET].

Contaminant V R Contaminant V R 1 m-Cymene x 12 Benzene, 1-methyl-4-

(1-methylethyl)-2 -nitro

x

2 Nonanal x 13 1,7-Dimethylnaphthalene

x

3 Benzene, 1-methoxy, 4-(1-propynyl-)

x 14 1,6-Dimethylnaphthalene

4 Naphthalene x 15 Benzoic acid, 4-methyl-, 2-

methylpropyl-

x

5 4-n-Propylanisole x 16 1,4-Dimethylnapthlalene

x

6 2-Methylnaphthalene x 17 1,2-Dimethylnaphthalene

x

7 1-Methylnaphthalene x 18 Trimethylnapthalene isomers (5 peaks)

8 1,2-Ethanediol, dibenzoate

x x 19 Benzophenone x

9 Biphenyl x 20 Cyclic dimer x x 10 1-Ethylnaphthalene x 21 Cyclic dimer ether x x 11 2,6-

Dimethylnaphthalene x

The corresponding chromatogram for the heptane extract is shown in Figure 4.4. The

cyclic dimer and its corresponding ether are not present in Figure 4.4 because they eluted

later in the chromatogram (after 30 minutes). A total of 21 compounds were identified in

the flake’s chromatogram after it was superimposed with the blank’s chromatogram. The

discrepancy in the number of components determined by Soxhlet extraction (38 peaks) and

total dissolution using TFA (21 peaks) could have occurred due to any of the following

reasons:

(a) The TFA is aggressive and therefore could cause some contaminants to breakdown.

Nerin et al. (2002) addressed the severity of the total dissolution procedure towards

limonene in view of the lack of GC/MS response.

94

Chapter 4

(b) Heptane has more interfering impurity peaks than DCM, even though it was

distilled prior to liquid-liquid extraction. Selected ion monitoring (SIM) with

retention time coincidence has been used in the next section (Section 4.3) to verify

the presence of some contaminants identified in the total ion chromatogram (TIC) of

the Soxhlet extract.

(c) The concentrations of some contaminants in the ground flake (extracted by Soxhlet)

could be greater than in the unground flake (extracted by total dissolution).

Despite the discrepancy in the number of contaminants extracted by both techniques, the

presence of many contaminants in post-consumer flake extracted by Soxhlet extraction was

confirmed by the identification of the same contaminants in the heptane extract.

Figure 4.4: Chromatogram of TFA/heptane extract for washed and dried flake.

The compounds extracted by total dissolution with TFA but not by means of Soxhlet

extraction are peak 8 (1,2-ethanediol, dibenzoate) and peak 12 [benzene, 1-methyl-4-(1-

methylethyl)-2 –nitro]. Peak 8 was also present in virgin PET; therefore it is assumed to be

a polymerization by-product. Peak 12 was absent from virgin PET, however in its absence,

a similar compound (benzene, 3-dimethyl-2-nitro-) was identified in virgin PET.

95

Chapter 4

Therefore, the likelihood of peak 12 being a contaminant in flake was weakened by the

presence of the analogue in virgin PET. There exists the possibility that the components

corresponding to peaks 8 and 12 are entrapped through electrostatic bonding or in “holes”

within the polymer matrix and therefore cannot be extracted unless aggressive solvents

such as TFA disperse the polymer’s structure.

The constituents of PET were also extracted by total dissolution using HFIP, as performed

in past studies (Begley and Hollifield 1989, Triantafyllou et al. 2002, Komolprasert et al.

1995). The compounds identified in the flake extract after comparison with the blank are

presented in Table 4.5. The corresponding peaks are presented in the flake chromatogram

in Figure 4.5.

Table 4.5. Constituents of PET were also extracted by total dissolution using HFIP [“x”

denotes presence of contaminant in virgin (V) and recycled (R) PET].

Contaminant V R Contaminant V R 1 Benzoic acid 5 Benzene, 1,1’-(1,2-

ethenediyl) bis- x x

2 Methyl salicylate 6 Terephthalic acid, methyl vinyl ester

x x

3 Isoquinoline x x 7 Dimer x x 4 Benzophenone x 8 Dimer ether x x

The number of compounds identified in the HFIP extract was much lower than the numbers

determined in the Soxhlet and TFA/heptane (total dissolution) extracts. This could be a

consequence of the large number of interfering peaks in the HFIP extract resulting from the

large volumes of solvents used during extraction. Most of the compounds listed in Table

4.5, apart from peak 5, contain polar functional groups, for which the extraction method is

supposedly specific. Another way of extracting polar compounds by total dissolution was

reported by Bayer (2002), who dissolved the PET in TFA and partitioned the polymer

constituents into methyl tertiarybutyl ether.

96

Chapter 4

Figure 4.5: Chromatogram of HFIP extract for washed and dried flake.

Peaks number 3, 5 and 6 were not identified in the Soxhlet or TFA/heptane extracts. There

is a chance that these compounds, which are also present in virgin PET, were entrapped in

the polymer matrix and could not be extracted by DCM during Soxhlet extraction. The

moderate polarity of [terephthalic acid, methyl vinyl ester] could have prevented efficient

extraction with TFA/heptane, which is a liquid-liquid technique that relies on a favorable

partition coefficient into a non-polar solvent. Alternatively, [terephthalic acid, methyl vinyl

ester] and [benzene, 1,1’-(1,2-ethenediyl) bis-] may have been broken down by TFA, which

is assumed to add across the double bonds of molecules.

Isoquinoline is charged in an acidic environment therefore is unlikely to be transferred into

the heptane extract during liquid-liquid extraction.

The components foreign to the flake extract were benzoic acid, methyl salicylate and

benzophenone. All of these compounds were identified in the Soxhlet extract of post-

consumer PET.

The HFIP extracts were run on a polar column (EC-Wax Econo-cap), however no unique

peaks were identified.

97

Chapter 4

4.2.5 Running the extracts on polar column

The Soxhlet extract for the flake and blank (with DCM) were analysed on an EC-Wax

Econo-cap (polyethylene glycol) column. The compounds identified in the flake are listed

alongside their chromatogram retention times in Table 4.6.

Table 4.6: Soxhlet extract run on an EC-Wax Econo-cap column.

Retention time

Contaminant Retention Time

Contaminant

1 5.883 Limonene 14 13.299 1-Methylnaphthalene 2 5.980 Cineole 15 13.724 Cyclopropane, nonyl-3 6.233 Benzene, 1-ethyl-2-

methyl- 16 13.893 Dimethylnaphthalene

isomer 4 6.765 m-Cymene 17 14.165 Dimethylnaphthalene

isomer 5 6.915 Benzene, 1,3,5-

trimethyl- 18 14.219 Dimethylnaphthalene

isomer 6 7.002 Undecane 19 14.138 Biphenyl 7 8.257 Ethanol, 2-butoxy-# 20 14.138 Biphenyl, 3-methyl- 8 9.166 1-Hexanol, 2-ethyl- 21 15.128 Trimethylnaphthalene

isomer 9 9.245 Tritetracontane 22 15.068 Trimethylnaphthalene

isomer 10 10.647 1,2-Ethanediol# 23 19.189 Benzophenone 11 12.216 Methyl salicylate 24 19.630 Benzoic acid, butyl

ester 12 12.626 Benzene, 1-methoxy

4-(1-propenyl)- 25 20.221 Dodecanoic acid

13 12.93 2-Methylnaphthalene # Identity confirmed by retention time of standard.

Many of the contaminants identified were also observed on a semi-polar 5% phenyl

polymethylsiloxane column (Table 4.3), however there were several exceptions. These

exceptions were peaks 6, 8, 9, 10, 20. The different selectivity of the column may have

contributed to the separation of these peaks away from interfering peaks. For example, [1-

hexanol, 2-ethyl-] (peak 8) was identified as a co-eluting peak (with m-cymene) on the non-

polar column whilst it was totally separated on the polar column.

The identification of 1,2-ethanediol was also possible on the polar column. The high

polarity of this compound could have made it less compatible on the non-polar column. It

98

Chapter 4

is proposed that due its the lack of interaction with the stationary phase 1,2-ethylene glycol

eluted early in the run, during the solvent delay time. This compound is normally

derivatised prior to analysis when using a non-polar column (Kim et al. 1990). However,

Morelli-Cardoso et al. (1997) and Kashock and Breder (1980) analysed it in aqueous food

simulants by GC using a polar column (EC-Wax Econo-cap).

Other polar compounds involved in the manufacture of PET (e.g. terephthalic acid, BHET,

MHET) were not identified.

The identities of most compounds in Table 4.6 were not confirmed by coincidence in

retention time with standards. Therefore, the dimethylnaphthalene and

trimethylnaphthalene isomers were not distinguished from each other.

4.2.6 Possible origin of the components

As already mentioned in Section 2.3.3, foreign components in post-consumer PET (prior to

recycling) could arise from three possible sources – the original use of the packaging,

consumer abuse/misuse and microbial contamination. Ironically, contaminants could also

be introduced during recycling (e.g. labels running during the wash and dry stage of the

recycling process, contamination from detergents etc).

Many of the compounds identified in PET flake (Table 4.3) could be traced back to the

original contents of the PET container. For example, methyl salicylate, menthone, (+)-

carvone and 1,8-cineole are found in mouthwash whilst limonene and γ-terpinene are

constituents of soft drinks. Some of these compounds could also have been mixed into

household cleaners, such as dishwashing detergents, in order to transmit a citrus fragrance

or provide antiseptic properties (Bayer 2002). All of these compounds have been reported

as contaminants in post-consumer PET by other researchers (Bayer 2002, Franz and Welle

1999b).

There were other contaminants whose origin was not so evident, such as naphthalene and

its methyl ethyl derivatives. Naphthalene has already been identified in low-density

polyethylene (LDPE) at 0.7 and 2 ppm levels (Lau and Wong 1994, 1995). It was

presumed in those studies that naphthalene contamination resulted from a polluted

environment, for example, from lacquer, paint and mothballs. Gramshaw et al. (1995)

99

Chapter 4

identified naphthalene in poly(ethersulphone) and thermoset polyester. The possible origin

of these two compounds was not discussed.

Naphthalene, along with 1-methylnaphthalene and 1,6-dimethylnaphthalene, was also

detected in PET multilayer films (Freire et al. 1998). The origin of these naphthalene

derivatives was suggested to be printing inks, but we have extracted the only evident source

of printing inks (labels) and have not detected naphthalenes. From our knowledge of the

industry, PET sources, etc. it is believed that the above origins are not responsible for the

presence of naphthalenes in our PET. In 1998 Coca Cola Australia manufactured two

million bottles from poly(ethylene naphthalate) at its Newcastle plant (NSW). Two years

later when we sampled the washed and dried flake, there can be little doubt that a

significant amount of PEN would still be in curbside collections as old PEN bottles. The

alkylnaphthalene isomers and naphthalene in PEN could also have formed during the

preparation of 2,6-dimethylnaphthalene, a precursor of 2,6-naphthalene dicarboxylic acid,

which is reacted with ethylene glycol to produce PEN.

Additionally, it is possible that there were some naphthalene-based acids present as

impurities in feedstock at the point of manufacture. As virgin PET is imported into

Australia from several Southeast Asian countries, we have been unable to establish the

veracity or frequency of this. In the one batch of virgin PET we have monitored, there

were no naphthalenes.

It is also feasible that naphthalene sulphonate derivatives could have been used during the

wash process (in detergents) as emulsifiers.

Dodecanoic acid – which we will see later to be the highest-level contaminant from the

washed and dried shredded flake – does not appear to have any logical source associated

with the PET per se. The very high levels consistently detected suggest to us that it is

probably derived from detergents used in the washing process. However, we do not have

any data to verify this. Kim et al. (1990) identified three fatty acids (palmitic acid, oleic

acid and stearic acid) in a commercial amber PET bottle wall. Gramshaw et al. (1995)

identified dodecanoic acid in PET and other fatty acids in thermoset polyester and

poly(ethersulphone) (e.g. hexanoic acid, heptanoic acid, benzoic acid). Benzoic acid was

also identified in the Visy washed and dried flake, presumably as a break down product of

terephthalic acid. Benzoic acid is occasionally added as a preservative to soft drink such as

100

Chapter 4

Coca Cola, thus there was a chance that this compound diffused into PET during soft drink

storage. Bayer (2002) reported the occurrence of benzoic acid amongst many other

compounds in PET feedstock entering the recycling stream. Other compounds that were

identified by this author and that were also present in our analysis of washed and dried

flake were methyl salicylate, (+)-carvone, 1,8-cineole, p-cymene, limonene, (γ)-terpinene,

nonanal, undecane, hexadecane, benzophenone, menthone, [1-hexanol, 2-ethyl-] and

[benzene, 1-methoxy, 4-(1-propenyl-)]. A total of 121 components were identified from

five feedstock materials by thermal desorption GC/MS analysis. The absence of interfering

solvent peaks could account for the large number of components identified by thermal

analysis relative to the solvent extraction carried out in the current chapter.

Benzophenone is primarily used as a photoinitiator and fragrance enhancer; therefore it is

often used in polymer production and added to beverages. This compound, however, was

not present in the virgin PET we analysed, therefore the latter source seemed more likely.

Nevertheless, the virgin PET used to make the post-consumer soft drink bottles could have

originated from a diverse range of manufacturers that use benzophenone in their PET

production.

Benzophenone was used as a ‘non-volatile polar surrogate’ during challenge tests and is

generally the most challenging compound to be removed by recycling (Harding et al.

unpublished, Franz et al. 1998, Franz and Welle 1999a). As benzophenone has a low

volatility and similar solubility parameter to PET, it is not surprising that it was the most

persistent (Harding et al. unpublished).

Another high-level contaminant, 2-butoxyethanol, could logically be derived from the

ethylene glycol PET reagent during degradation of PET, especially at elevated

temperatures. On the other hand, this contaminant was not present in the virgin PET we

have analysed. Gramshaw et al. (1995) identified 2-ethoxyethanol in thermoset polyester

and speculated that it could be a solvent residue.

101

Chapter 4

4.3 QUANTITATIVE STUDY OF CONTAMINANTS IN WASHED AND DRIED

PET FLAKE

4.3.1 Introduction

Unless contaminant levels are proven to be acceptably low (i.e., below the US FDA

threshold of 215 ppb), post-consumer PET could pose a health threat to the consumer if the

components migrate into food. Therefore, before food-contact safety could be officially

declared, the level of each contaminant in recycled PET must be determined and proven to

fall below the threshold of regulation. If the levels are above 215 ppb, migration testing

into food simulants should be instigated and the amount of each contaminant migrating

must not exceed 10 ppb to assure food safety.

The purpose of this section was to determine the level of foreign components in washed

and dried flake. Soxhlet extraction was the prime extraction method used because it

accounted for the identification of more contaminants relative to the total dissolution

extraction methods (Section 4.2). In addition, DCM is not as aggressive as TFA and HFIP,

which could lead to the molecular breakdown of some polymer constituents (e.g.

limonene).

Soxhlet extraction involves the diffusion of components out of the core of the polymer

matrix into the extraction solvent; therefore the amount of contaminant extracted out of

PET is dependent on three diffusion related parameters: particle size, temperature and time.

The effects of particle size and time on extraction were investigated in this section and the

optimum conditions were selected for further extractions of washed and dried flake.

Other processes that impede extraction once the contaminants defuse to the surface of the

polymer particles is their transfer from the surface and solubility in DCM. As shown in

Figure 4.6, the polymer particles are thought to be in contact with a stagnant layer of

solvent through which the contaminants have to pass to enter the “free” solvent. Lou et al.

(1997) suggested that raising the temperature could increase diffusion of contaminants

through the stagnant layer and improve solubility in DCM. It is also thought that

102

Chapter 4

sonication could improve migration through the stagnant layer. Sonication will be another

solvent extraction technique addressed throughout this thesis.

Figure 4.6: Schematic presentation of the three subsequent steps in solvent extraction.

M

Stagnant solvent layer

Polymer particle

The temperature used during Soxhlet extraction

which is 40°C. Using a solvent with a higher

temperature, however this was not considere

components, as discussed in Section 4.1.3. Th

already been demonstrated during the SFE, M

Lou et al. 1997; St. Küppers 1992; Camacho an

4.3.2 Study of extraction kinetics for flake grou

During this investigation the washed and drie

size efficiently possible (0-300 µm). The grou

specified in Figures 4.7-4.10, which display th

contaminants in ground washed and dried flak

that 24 h was adequate to completely extract

As seen in Figure 4.7, the two possible excepti

both of which appear not to have reached max

of extraction times. However, as illustrated

incomplete extraction of both contaminants i

realized that the error bars represent only one

determined is approximately twice as large.

103

DC

was restricted to the boiling point of DCM,

boiling point could increase the extraction

d due to the concern with losing volatile

e effects of temperature on extraction have

AE and ASE of polymers (Lou et al. 1996;

d Karlsson 2000).

nd to 0-300 µm

d flake was ground to the smallest particle

nd polymer was then extracted for the times

e kinetics of the extraction process for the

e. From this kinetic study it was observed

most contaminants out of the ground flake.

ons are 4-n-propylanisole and (-)-menthone,

imum extraction throughout the entire range

in Figure 4.10, error bars indicate that the

s probably incorrect, especially when it is

standard deviation and the range of values

Chapter 4

Figure 4.7: Soxhlet extraction kinetic study of washed and dried flake ground to 0-300 µm.

Compounds identified at levels below 200 ppb.

0

20

40

60

80

100

120

140

160

180

200

0 20 40 60 8Time (hours)

Am

ount

Ext

ract

ed (p

pb)

0

1,2,4-Trimethylbenzene Dodecane1,6-Dimethylnaphthalene 2,6-Dimethylnaphthalene1,7-Dimethylnaphthalene n-Hexylbenzoate1-Methylnaphthalene TetradecaneCineole 1,4-Dimethylnaphthalenegamma-Terpinene Naphthalene4-n-Propylanisole 1-EthylnaphthaleneBiphenyl (-)-Menthone1,2-Dimethylnaphthalene m-Cymene3-Ethyl-o-xylene 1,2,3,5-, tetramethylbenzene

104

Chapter 4

Figure 4.8: Soxhlet extraction kinetic curves of trimethylnaphthalene isomers extracted

from washed and dried flake ground to 0-300 µm.

0

5

10

15

20

25

30

35

40

0 20 40 60 8

Time (hours)

Con

cent

ratio

n (p

pb)

0

Trimethyl Naphthalene Isomer 1 Trimethyl Naphthalene Isomer 2Timethyl Naphthalene Isomer 3 Trimethyl Naphthalene Isomer 4Trimethyl Naphthalene Isomer 5 Trimethyl Naphthalene Isomer 6

105

Chapter 4

Figure 4.9: Soxhlet extraction kinetic study of washed and dried flake ground to 0-300 µm.

Compounds identified at levels above 200 ppb.

0

200

400

600

800

1000

1200

1400

0 20 40 60 8

Time (hours)

Am

ount

Ext

ract

ed (p

pb)

0

Dodecanoic Acid 2-Butoxy Ethanol

Limonene Methyl salicylate

2-Methyl Naphthalene Benzophenone

106

Chapter 4

Figure 4.10: The standard deviations associated with data points defining kinetic curves

that do not follow the general trends of Figure 4.7.

0

20

40

60

80

100

120

140

160

180

200

0 20 40 60 8

Time (hours)

Am

ount

Ext

ract

ed (

ppb)

0

Dodecane 1,2,4-TrimethylbenzeneTetradecane 4-n-Propylanisole(-)-Menthone

107

Chapter 4

Student t-tests comparing the amounts of n-propylanisole extracted at 24 and 48h, 48 and

72 h, and, 24 and 72 h yielded tcalc values of 0.97, 1.28 and 2.28, respectively. As the t

values at the 5% confidence level are either 2.78 or 3.28 (depending upon the number of

replicate analysis at each time), the differences in the amounts extracted are not significant.

The same is true for (-)-menthone where the equivalent tcalc values for the same number of

replicates are 1.18, 1.40 and 1.90. Figure 4.10 also displays the error bars for the analytes

exhibiting erratic recoveries verses time (in Figure 4.7) and whose irregular shapes are an

indication of irreproducibility. Again the student t-tests confirm that there is no significant

difference at the 5% level.

Similarly, Wim and Swarin (1975) monitored the extraction process of two additives from

polypropylene pellets and concluded that a plateau was reached by 24 h (using

tetrahydrofuran as an extracting solvent). Komolprasert et al. (2001) extracted PET sheets

with DCM for 24, 48, 72 and 96 hours, and concluded that the optimal extraction was 24 h.

Shorter extraction times were not considered in that study.

Further documentation of extraction kinetics of polymers by Soxhlet extraction is scarce.

There has been, however, greater emphasis in optimizing other, more time efficient

extraction techniques in polymer research, such as SFE (Bartle et al. 1990, Daimon and

Hirata 1991, Hunt and Dowle 1991).

The extraction kinetics of ethylene glycol are shown in Figure 4.11. Ethylene glycol is

unlikely to be a post-consumer contaminant and presumably originates from the original

manufacture of virgin PET. The apparent source of this ethylene glycol in the polymer may

explain why the amount extracted (approximately 30 ppm) exceeded that of all the

contaminants quantified. It is not unusual that the level of a compound inherited by the

polymer exceeded those that diffused into the PET bottle during its commercial use. This is

because ethylene glycol is directly added to PET during manufacture and therefore its

concentration in the polymer is not limited by its diffusion coefficient, as it is during

sorption in commercial use. The EEC regulation for the migration of ethylene glycol into

food is 30 ppm, therefore from these results, the washed and dried ground flake could be

considered appropriate for food contact applications in terms of ethylene glycol migration

since the amount extracted with DCM (30 ppm) is expected to surpass the level of the

actual migration into soft drink (i.e. the level of migration is expected to fall below 30

108

Chapter 4

ppm). This is because DCM is a more efficient extraction medium than soft drink whilst

causing polymer swelling.

Figure 4.11: Soxhlet extraction kinetic study of washed and dried flake ground to 0-300

µm. Ethylene glycol analysed on an EC-Wax Econo-cap column.

0

5000

10000

15000

20000

25000

30000

35000

0 10 20 30 40 50 60 70

Time (hours)

Am

ount

Ext

ract

ed (p

pb)

Ethylene Glycol

The dense packing of the polymer particles in the extraction thimble along with the

swelling or “softening” instigated by the solvent may have caused agglomeration of the

polymer particles. During polymer swelling the polymer chains become mobile and the

consequential softening or dissolution of the particle surface could result in the particles

blending into one another, especially if they are crammed side-by-side.

This coalescence was observed when attempting to remove the packed particulate matter

from the thimble – the particles were clumped together and no longer independent of one

another. Therefore, it was proposed that having PET fine particles in the extraction thimble

could have paradoxically extended the duration of extraction due to the decrease in the

surface area resulting from particulate fusion. Further extraction studies in this section will

validate this theory.

109

Chapter 4

A plot of the percentages of extraction at 8h (A8/Ae) versus contaminant molecular weight

is presented in Figure 4.12 (for compounds in figures 4.7-4.10). Note that in this plot

averages of the dimethylnaphthalene and trimethylnaphthalene isomers have been taken.

Figure 4.12: Ratio of amount extracted at 8 h (A8) to amount extracted after 24 h (Ae) (as a

percentage) versus contaminant molecular weight.

y = -0.0016x2 + 0.3438x + 30.955R2 = 0.6098

20

25

30

35

40

45

50

55

100 120 140 160 180 200 220 240Molecular Weight

(A8/

Ae)

*100

A R2 value of 0.61 for a second order polynomial of best fit indicates a weak correlation

between the variables. A linear fit (y = -0.18x + 73.11) gave a lower R2 value of 0.58.

110

Chapter 4

A better correlation would be expected if solvent penetration did not govern the rate of

contaminant migration, as it does in our case where the solubility parameters of DCM and

PET are closely matched. When this is not so, migration is generally Fickian and

predominantly dependent on the diffusion coefficient of the contaminant molecule, which is

proportional to size (Riquet et al. 1991, Feigenbaum et al. 1993). When a polymer is

swollen it is expected that there is less discrimination in the diffusion of migrants with

different size because the polymer matrix is more open up.

There are, however, other factors besides molecular weight that effect diffusion. These

include molecular shape and thus effective volume, and, functional group interactions

between the contaminant and the polymer.

The effective volumes of the naphthalene derivatives will be determined by their cross-

sectional areas perpendicular to the direction of diffusion and therefore there will be

differences in the resistance to diffusion, depending upon which cross sectional area is

presented. Table 4.7 lists the percentages of naphthalene derivatives extracted at 8 h and

alongside their molecular structure. Assuming diffusion in the horizontal direction, the

naphthalene derivative which is expected to have the most difficulty diffusing through the

polymer matrix (horizontally) is 1,4-dimethylnaphthalene because it has two methyl

substituents “para” to each other, increasing its effective volume and thus hindering its

movement between polymer chains in the horizontal direction.

Yet the fraction of 1,4-dimethylnaphthalene extracted at 8 hours (48.0%) was higher than

that of 2,6-dimethylnaphthalene (39.9%), 1,2-dimethylnaphthalene (44.6%), 1-

ethylnaphthalene (45.5%), and 1,7-dimethylnaphthalene (46.9%). The lowest percentage

extracted was observed for 2,6-dimethylnaphthalene (39.9%), which is likely to have a

smaller effective volume than 1,4-dimethylnaphthalene.

The fastest diffusion was expected for naphthalene since it has no substituents that could

obstruct its movement. In our study, the highest percentage extracted was observed for 1,6-

dimethylnaphthalene (51.0%). Therefore, in conclusion, the effective volumes of

naphthalene derivatives were not shown to have an impact on their level of extraction.

111

Chapter 4

Table 4.7: The percentages of naphthalene derivatives extracted at 8 h alongside their

molecular structure.

Compound Structure (A8/Ae) 100 2,6-Dimethylnaphthalene CH3

CH3

39.9%

1,2-Dimethylnaphthalene CH3

CH3

44.6%

1-Ethylnaphthalene CH2CH3

45.5%

1,7-Dimethylnaphthalene C H3C H3

46%

1,4-Dimethylnaphthalene C H 3

C H 3

48.0%

Naphthalene

49.5%

1,6-Dimethylnaphthalene CH3

CH3

51.0%

The deviation from the expected extraction order could be attributed to experimental

uncertainty as only one extraction was performed at 8 hours. In addition, contaminants

could be diffusing in the vertical direction (as opposed to the horizontal direction) through

the polymer matrix, which would present a different effective volume. For example 2,6-

dimethylnaphthalene is expected to have the highest effective volume when assuming

diffusion in the vertical direction. Therefore, it is expected to experience the greatest

difficulty moving vertically through the polymer matrix and accordingly have the smallest

extraction efficiency. As expected, 2,6-dimethylnaphthalene had the smallest extraction

efficiency (39.9%) relative to the other naphthalene derivatives. Conversely, 1,6-

dimethylnaphthalene, which had the highest extraction efficiency of the naphthalene

112

Chapter 4

derivatives listed in Table 4.7, was expected to have a lower extraction percentage than

naphthalene, 1,4-dimethylnaphthalene and 1-ethylnaphthalene. Once again, no clear

correlation between effective volume and percent extracted was observed for diffusion of

naphthalene derivatives in the vertical direction (as opposed to the horizontal direction).

The rate of diffusion is expected to be an average of the overall rate with all molecular

orientations.

In order to establish whether a correlation exists between contaminant polarity and percent

extracted Figure 4.12 was converted into a “Venn diagram”, grouping the contaminants

according to their functional group types (Figure 4.13). From Figure 4.13 it could be seen

that all the non-polar (hydrocarbons) are much more rapidly extracted, the weakly polar

ethers not quite as fast, the moderately polar ketones are extracted more slowly again and

the dodecanoic acid with the highly polar carboxylic group is extracted much more slowly

again. The highly polar methyl salicylate and the moderately polar other ester (n-hexyl

benzoate) are the two obvious exceptions to this. Otherwise there is a clear trend that

shows an unequivocal dependence upon polarity. To test the strength of this dependence

the percent extracted was plotted against an estimated solvent strength parameter for each

of the contaminants (Figure 4.14). The solvent parameter for each contaminant was

estimated from the solvent strength parameter of a representative compound sharing the

same functional group (Johnson and Stevenson 1979).

A correlation coefficient (R2) of 0.50 for Figure 4.14 implies significant scatter about the

quadratic fit and a weak correlation. The fact that there is a negative gradient suggests that

there may be functional group interactions taking place between the contaminant and the

polymer, causing a decrease in percent extracted as contaminant polarity increases. A

quadratic fit rather than a linear fit was selected because a higher correlation coefficient

was obtained with the former (R2 = 0.46 for linear fit y = -9.3091 + 48.555; R2 = 0.50 for

quadratic fit y = 4.16x2 - 17.3x +50.33).

113

Chapter 4

Figure 4.13: “Venn diagram” grouping the contaminants according to their functional group

types.

25

30

35

40

45

50

55

90 110 130 150 170 190 210 230MW

Hydrocarbons Ethers Methyl salicylate Ketones Acids Esters(A

8/A

eq)*

100

114

Chapter 4

Figure 4.14: Ratio of amount extracted at 8 h (A8) to amount extracted after 24 h (Ae) (as a

percentage) versus estimated solvent strength parameter.

y = 4.1648x2 - 17.3x + 50.328R2 = 0.4971

20

25

30

35

40

45

50

55

0 0.5 1 1.5 2 2Estimated solvent strength parameter

(A8/

Ae)

100

.5

4.3.3 Particle size variation

In order to determine whether the Soxhlet extraction rate was limited by diffusion, larger

particle sizes were extracted for 24 h. It was presumed that larger PET particles sizes

would give rise to longer diffusion paths for the mass transfer of contaminants from the

polymer core to its surface. The rate of diffusion is proportional to D/L2, where D is the

diffusion coefficient and L is the length of the shortest dimension (Vandenburg et al. 1997).

115

Chapter 4

Therefore, increasing L should theoretically decrease the rate of diffusion out of the matrix.

Work by Perlstein (1983) and Spell and Eddy (1960) support this theory. It was concluded

in their studies that powdering the PVC and PE pellets respectively had a positive effect on

extraction time and recovery. Ashraf-Khorassani et al. (1991) and Salafranca et al. (1999)

also found that faster SFE is obtained as the surface area of the polyolefin matrix increases.

Figures 4.15 – 4.17 show a general decrease in the amount of contaminant extracted as

particle size increases. The large decrease in measured contaminant levels from the

smallest particle size (0-300 µm) to the intermediate particle size (>300-425 µm) is clear

and consistent (for 16 of the 19 contaminants), whilst the effect of increasing the particle

size further to the large sizes (>425-700 µm) is much smaller – and due to this reduced

difference between the levels for these two particles sizes – is only clear for just over half

of the contaminants. However, a student t-test on the means of the contaminant levels for

the 19 analytes shown in Figure 4.15 for the medium and large particle sizes yields tcalc =

2.53. As t = 2.10, a significant decrease in the amount extracted between the medium and

large particles is supported (at the 5% level). Figure 4.16 is an equivalent plot for the six

trimethylnaphthalene isomers and the large decrease in contaminant levels from small to

medium particle sizes, followed by a small decrease to large particle sizes is visually

unambiguous in this case. For the six contaminants extracted at greater than 200 ppb, the

results (Figure 4.17) are more scattered, but are generally consistent with the above

interpretation of the data in Figures 4.15 and 4.16. For the small number of exceptions to

the general trends, these tend to be for analytes with large standard deviations in the

amounts extracted (Figure 4.18). [Further extraction work on extruded (recycled) PET

pellets has provided smaller standard deviations for many compounds, presumably due to

the contaminants being uniformly blended into the polymer network within the melt phase

prior to pellet manufacture. This work will be presented in the next chapter].

116

Chapter 4

Figure 4.15: Amount of contaminant extracted from flake ground to different particle sizes

(compounds below 200 ppb).

0

20

40

60

80

100

120

140

small medium large

particle size

Am

ount

Ext

ract

ed a

fter 2

4h (p

pb)

Dodecane 1,6-Dimethylnaphthalene2,6-Dimethylnaphthalene 1,7-Dimethylnaphthalenen-Hexylbenzoate 1-MethylnaphthaleneTetradecane Cineole1,4-Dimethylnaphthalene gamma-TerpineneNaphthalene 4-Propylanisole1-Ethylnaphthalene Biphenyl(-)-Menthone 1,2-Dimethylnaphthalenem-Cymene 3-Ethyl-o-xylene1,2,3,5-Tetramethylbenzene

117

Chapter 4

Figure 4.16: Amount of trimethylnaphthalene contaminants extracted from flake ground to

different particle sizes.

0

5

10

15

20

25

30

35

small medium large

particle size

Am

ount

Ext

ract

ed a

fter 2

4h (p

pb)

Trimethyl Naphthalene Isomer 1 Trimethyl Naphthalene Isomer 2

Trimethyl Naphthalene Isomer 3 Trimethyl Naphthalene Isomer 4

Trimethyl Naphthalene Isomer 5 Trimethyl Naphthalene Isomer 6

118

Chapter 4

Figure 4.17: Amount of contaminant extracted from flake ground to different particle sizes

(compounds above 200 ppb).

0

200

400

600

800

1000

1200

small medium large

particle size

Am

ount

Ext

ract

ed a

fter 2

4h (p

pb)

Limonene BenzophenoneDodecanoic acid Methyl salicylate2-Methylnaphthalene 1,2,4-Trimethylbenzene

119

Chapter 4

Figure 4.18: The standard deviations associated with data points defining curves that do not

follow the general trends of Figure 4.15 – 4.17.

0

50

100

150

200

250

300

350

small medium large

particle size

Am

ount

Ext

ract

ed (p

pb)

gamma-Terpinene1,2,3,5-Tetramethylbenzene1,2,4-Trimethylbenzene3-Ethyl-o-xylene

120

Chapter 4

4.3.4 Kinetic studies for the larger particle sizes

Figures 4.19-4.22 show results for the kinetics studies on the >300-425 µm particle size

range. Due to single measurements being taken at each extraction time in this case℘, the

plots are not as well defined as those in Figures 4.7-4.11. However, for each analyte the

values appear to fluctuate about a mean and thus indicate that extraction was completed

before the first 4 hours of Soxhlet extraction. Similar kinetic trends were observed for the

unground flake (Figure 4.26-4.29), which are also single measurements at each extraction

time. For the >425-700 µm ground flake (Figures 4.23-4.25), however, there were only

three points taken, one for each extraction time (1 h, 3 h and 24 h). Therefore, it is difficult

to determine whether ascending trends (e.g. for biphenyl, naphthalene, dodecanoic acid) are

due to sample variation or an increase in extraction efficiency with time. There are also a

few compounds (e.g. 1,7-dimethylnaphthalene, 2,6-dimethylnaphthalene, 1-

methylnaphthalene, 1,4-dimethylnaphthalene, 1-ethylnaphthtalene, 1,2,3,5-

tetramethylbenzene) that do not appear to be fully extracted until 3 h. Some other

compounds (e.g. dodecane, tetradecane, trimethylnaphthalene isomer 7) show descending

extraction trends with time, which may result from sample variation. Student t-tests were

performed in order to determine whether a significant difference exists between the mean

contaminant concentrations at 1 h and 3 h; 3 h and 24 h; and 1 h and 24 h and the tcalc

values were 0.07, 0.16 and 0.08 respectively. Since the t-test value is 2.042 at the 5%

confidence level, it can be concluded that, in general, there is no significant difference in

the amount of contaminant extracted at the three different times.

The extraction kinetics of dimethylterephthalate (Figure 4.29) and ethylene glycol (Figure

4.30) in unground flake were also determined. Despite both of these compounds arising

from the polymer’s original manufacture, the sample-to-sample variation (i.e. standard

deviation) of these compounds was large.

These compounds were present at higher concentrations in the ground flake (0-300 µm).

The levels were 12 ± 3 ppm (unground flake) and 38 ± 7 ppm (flake ground to 0-300 µm)

for ethylene glycol and 2.3 ± 1.3 ppm (unground flake) and 3.8 ± 0.2 ppm (flake ground to

0-300 µm) for dimethylterephthalate.

℘ Replicates at each time are unnecessary if the points are fluctuating about a mean behaviour and generating a general trend.

121

Chapter 4

Figure 4.19: Soxhlet extraction kinetics of flake ground to >300-425 µm. Contaminants

below 120 ppb.

0

20

40

60

80

100

120

0 10 20 30 40 50 60Time (hours)

Am

ount

Ext

ract

ed (p

pb)

1,2,4-Trimethylbenzene 1,7-Dimethylnaphthalene2,6-Dimethylnaphthalene Cineole1-Methylnaphthalene Tetradecane1,6-Dimethylnaphthalene Dodecanoic acidm-Cymene gamma-TerpineneNaphthalene 1,4-Dimethylnaphthalene1-Ethylnaphthalene Biphenyl4-Propylanisole (-)-Menthone1,2-Dimethylnaphthalene Benzene, 1,2,3,5, Tetramethyl-3-Ethyl-o-xylene n-Hexylbenzoate

122

Chapter 4

Figure 4.20: Soxhlet extraction kinetics of flake ground to >300-425 µm. Contaminants

between 120 ppb and 300 ppb.

0

50

100

150

200

250

300

0 10 20 30 40 50 60Time (hours)

Am

ount

Ext

ract

ed (p

pb)

2-Methylnaphthalene Dodecane Methyl salicylate

123

Chapter 4

Figure 4.21: Soxhlet extraction kinetics of flake ground to >300-425 µm. Contaminants

above 400 ppb.

0

200

400

600

800

1000

1200

0 10 20 30 40 50 60

Time (hours)

Am

ount

Ext

ract

ed (p

pb)

Limonene Benzophenone

124

Chapter 4

Figure 4.22: Soxhlet extraction kinetics of flake ground to >300-425 µm.

Trimethylnaphthalene isomers.

0

2

4

6

8

10

12

14

16

0 10 20 30 40 50 60Time (h)

Am

ount

Ext

ract

ed (p

pb)

Trimethylnaphthalene Isomer 1 Trimethylnaphthalene Isomer 2Trimethylnaphthalene Isomer 3 Trimethylnaphthalene Isomer 4Trimethylnaphthalene Isomer 5 Trimethylnaphthalene Isomer 6Trimethylnaphthalene Isomer 7

125

Chapter 4

Figure 4.23: Soxhlet extraction kinetics of flake ground to >425-700 µm. Contaminants

below 100 ppb.

0

10

20

30

40

50

60

70

80

90

100

0 5 10 15 20 25

Time (h)

Am

ount

Ext

ract

ed (p

pb)

Dodecane m-Cymene2-Methylnaphthalene 1,7-DimethylnaphthaleneTetradecane gamma-Terpinene2,6-Dimethylnaphthalene 1,6-Dimethylnaphthalene1-Methylnapthalene 4-Propylanisole1,2,4-Trimethylbenzene Naphthalene1,4-Dimethylnaphthalene (-)-MenthoneBiphenyl 1-Ethylnaphthalene1,2-Dimethylnaphthalene 3-Ethyl-o-xylene1,2,3,5-Tetramethylbenzene n-Hexylbenzoate

126

Chapter 4

Figure 4.24: Soxhlet extraction kinetics of flake ground to >425-700 µm.

Trimethylnaphthalene isomers

127

3

5

7

9

11

13

15

0 5 10 15 20 25Time (h)

Am

ount

Ext

ract

ed (p

pb)

Trimethyl Naphthalene Isomer 1 Trimethyl Naphthalene Isomer 2Trimethyl Naphthalene Isomer 3 Trimethyl Naphthalene Isomer 4Trimethyl Naphthalene Isomer 5 Trimethyl Naphthalene Isomer 6Trimethyl Naphthalene Isomer 7

Chapter 4

Figure 4.25: Soxhlet extraction kinetics of flake ground to >425-700 µm. Contaminants

above 200 ppb.

0

200

400

600

800

1000

1200

1400

0 5 10 15 20 25Time (h)

Am

ount

Ext

ract

ed (p

pb)

Limonene Benzophenone Dodecanonic Acid

128

Chapter 4

Figure 4.26: Soxhlet extraction kinetics of whole flake. Contaminants below 70 ppb.

0

10

20

30

40

50

60

70

2 4 6 8 10 12 14 16 18 20 22 24Time (h)

Am

ount

Ext

ract

ed (p

pb)

1,2,3,5-, tetramethyl benzene Naphthalene1-Ethylnaphthalene Biphenyl3-Ethyl-o-xylene 2,6- Dimethylnaphthalene1,4- Dimethylnaphthalene 1-Methylnaphthalene1,2- Dimethylnaphthalene Dodecanen-Hexylbenzoate Tetradecane(-)-Menthone 1,2,4-, Trimethylheptane

129

Chapter 4

Figure 4.27: Soxhlet extraction kinetics of whole flake. Trimethylnaphthalene isomers.

0

5

10

15

20

25

30

35

40

45

0 2 4 6 8 10 12 14 16 18 20 22 24Time (h)

Am

ount

Ext

ract

ed (p

pb)

Trimethyl Naphthalene Isomer 1 Trimethyl Naphthalene Isomer 2Trimethyl Naphthalene Isomer 3 Trimethyl Naphthalene Isomer 4Trimethyl Naphthalene Isomer 5 Trimethyl Naphthalene Isomer 6

130

Chapter 4

Figure 4.28: Soxhlet extraction kinetics of whole flake. Contaminants between 70 ppb and

200 ppb.

0

20

40

60

80

100

120

140

160

180

200

0 2 4 6 8 10 12 14 16 18 20 22 24Time (h)

Am

ount

Ext

ract

ed (p

pb)

m-Cymene1,7-, Dimethylnaphthalene & 1, 6 Dimethylnaphthalene2-MethylnaphthaleneDodecanoic Acid

others

131

Chapter 4

Figure 4.29: Soxhlet extraction kinetics of whole flake. Contaminants above 200 ppb.

0

500

1000

1500

2000

2500

3000

0 2 4 6 8 10 12 14 16 18 20 22 24Time (h)

Am

ount

Ext

ract

ed (p

pb)

Limonene Benzophenone Dimethyl Terephthalate

132

Chapter 4

Figure 4.30: Soxhlet extraction kinetics of whole flake. Ethylene glycol.

0

2000

4000

6000

8000

10000

12000

14000

16000

0 5 10 15 20 25Time (h)

Am

ount

Ext

ract

ed (p

pb)

Ethylene Glycol

133

Chapter 4

Kashtock and Breder (1980) extracted ethylene glycol (15 ppm) by Soxhlet extraction (with

DCM) out of unused PET bottles ground to 700 µm particles. Likewise Kim et al. (1990)

extracted approximately 15 ppm of ethylene glycol from a commercial amber PET bottle

wall by Soxhlet extraction (with ethanol).

The results are contrary to expectation based upon path length as the larger particles size

ranges are extracted sooner (by Soxhlet) than the 0-300 µm particle size range.

We assume that for the 0-300 µm particle size range, packing in the extraction thimble was

more compact than for the larger particle sizes and thus, due to hampered solvent access

and polymer aggregation, equilibrium was not attained until 24 h. To confirm this theory, a

sample of the 0-300 µm particle size was dispersed in DCM and sonicated for 3h with

intermittent stirring to reduce coagulation. As shown in Table 4.8, the concentrations

obtained for the DCM sonicaton extractions were mostly equivalent to those obtained by

total dissolution for the 0-300 µm particle size1, thus demonstrating complete extraction in

3 h by sonication. In fact a gradient of 0.96 (Figure 4.31) was obtained for a plot of amount

extracted by total dissolution versus the amount extracted by sonication for the small

particle size. [Vandenburg et al. (1999) also noted the inhibition of extraction caused by

excessive close packing and used sand as a dispersant to prevent PVC pellets from

coagulation in an ASE thimble]. Comprehensive comparisons between the ultrasonication

and the Soxhlet extraction methods were made during the extraction of extruded pellets and

the results are presented in Chapter 5. In contrast, the effects of coagulation was not

observed for the finely ground crystallized extruded pellets, possibly due to their higher

crystallinity relative to the flake.

The results in Table 4.8 and Figure 4.31 exclude the small number of compounds that could

not be analysed by total dissolution. These compounds are presented in Table 4.9 (a) and

(b). Limonene, cineole and γ-terpinene are assumed to react with TFA, which might

explain their absence from chromatograms. Likewise, Nerin et al. (2000) addressed the

severity of another total dissolution procedure (using dimethylformamide) towards

limonene in view of the lack of GC/MS response. Komolprasert et al. (1995) also discussed

the decomposition of malathion and diazinon during total dissolution with

1 This excludes the small number of compounds that cannot be analysed by the total dissolution technique. These will be discussed in Section 4.3.6 on “Validation of extraction procedures”.

134

Chapter 4

Figure 4.31: Log-log plot of levels of contaminants determined by total dissolution versus

levels extracted by sonication and comparison with the ideal relationship (y=x): for flake

ground to small particle sizes.

y = 0.9616x + 0.0781R2 = 0.9846

0

0.5

1

1.5

2

2.5

3

3.5

4

0 0.5 1 1.5 2 2.5 3 3.5 4

Log [Amount Extracted by Sonication (ppb)]

Log

[Am

ount

Ext

ract

ed b

y To

tal D

isso

lutio

n (p

pb)]

0-300 um Linear (0-300 um)

y=x

135

Chapter 4

Table 4.8: Contaminant levels (and standard deviations) [in ppb, in flake ground to 0-300

µm] determined by total dissolution with TFA, compared to extraction by sonication in

DCM.

1,2,4- Trimethyl-benzene

3-Ethyl-o-xylene

1,2,3,5- Tetramethyl-benzene

(-)-Menthone Methyl salicylate

4-n-propylanisole

DCM TFA DCM TFA DCM TFA DCM TFA DCM TFA DCM TFA 81.3 (11.4)

87.4 (7.9)

14.9 (1.7)

13.3 (1.0)

33.4 (2.9)

21.7 (0.9)

6.1 (0.2)

9.8 (1.1)

552 (64)

616 (99)

39.1 (5.2)

40.0 (3.6)

Naphthalene Biphenyl 2-Methyl

naphthalene

1-Methyl naphthalene

1-Ethyl naphthalene

2,6- Dimethyl naphthalene

DCM TFA DCM TFA DCM TFA DCM TFA DCM TFA DCM TFA 88.8 (3.1)

77.9 (1.3)

56.3 (1.9)

56.1 (2.7)

304 (25)

300 (26)

158 (13)

145 (3)

60.5 (2.6)

58.9 (2.6)

148 (1)

145 (4)

1,7- Dimethyl naphthalene & 1,6- Dimethyl naphthalene

1,4- Dimethyl naphthalene

1,2- Dimethyl naphthalene

Trimethyl naphthalene isomer 1

Trimethyl naphthalene isomer 2

Trimethyl naphthalene isomer 3

DCM TFA DCM TFA DCM TFA DCM TFA DCM TFA DCM TFA 437 (8)

417 (5)

76.6 (10)

84.3 (3.5)

31.7 (4.7)

31.0 (2.8)

21.6 (2.8)

23.7 (2.9)

18.2 (2.4)

20.7 (2.0)

25.5 (1.9)

24.8 (1.3)

Trimethyl naphthalene isomer 4 & 5

Trimethyl naphthalene isomer 6

Trimethyl naphthalene isomer 7

Benzophenone

DCM TFA DCM TFA DCM TFA DCM TFA 43.1 (2.7)

46.7 (2.4)

28.9 (3.5)

29.4 (2.1)

11.6 (0.3)

15.8 (0.7)

1236 (66)

1234 (135)

136

Chapter 4

Table 4.9: Contaminant levels (and standard deviations) [in ppb, in flake ground to 0-300

µm] determined by total dissolution with TFA, compared to extraction by sonication in

DCM: anomalies for (a) m-cymene [TFA>DCM] and (b) limonene, cineole and γ-

terpinene [DCM>TFA]. (All levels are in ppb by mass.)

(a)

m-Cymene

Particle size

DCM TFA

Small

(0-300 µm)

54.4

(0.0)

500

(58)

(b)

Limonene

Cineole γ-Terpinene Particle size

DCM TFA DCM TFA DCM TFA

Small (0- 300 µm)

1121 (40)

ND 94.5 (9.5)

ND 55.6 (3.7)

ND

137

Chapter 4

hexafluoro-2-isopropanol (HFIP). In additional studies limonene and cineole were

presumed to degrade in the presence of HFIP, due to their absence from chromatograms.

It is presumed that the double bonds of limonene and γ-terpinene react with the carbonyl

group of TFA, forming polar compounds, which may not be extracted by the non-polar

heptane. The presence of electron withdrawing fluorine atoms on the carbon adjacent to

the carbonyl group of TFA exert a positive charge on the carbonyl carbon, which is

prepared for nucleophilic attack.

The reason behind the large concentrations determined by TFA relative to DCM for m-

cymene is not so obvious. It is assumed that the TFA causes the breakdown of other

compounds (analogs of m-cymene) in the flake, forming m-cymene. There is also the

possibility that m-cymene is trapped between polymer chains and only released after the

total dissolution procedure which destroys the polymeric network and breaks some chains.

However, the latter proposition seems unlikely because none of the other contaminants

have behaved in an analogous way to m-cymene and there is no distinct feature of m-

cymene, which would lead to the nonconforming results.

In another kinetic experiment, the >300-425 µm particle size was extracted for 3 h and then

re-extracted for another 24 h with DCM by Soxhlet extraction (Table 4.10). There was no

significant increase in the amount extracted, indicating that extraction was completed after

the first extraction. Nielson et al. (1997) analogously confirmed that sorbed surrogate

contaminants were completely extracted in the first extraction from PET by shaking with

DCM for 24 h.

138

Chapter 4

Table 4.10: Flake ground to >300-425 µm particle size and extracted for 3 h and then re-

extracted for another 24 h.

Contaminant

Level of contaminant

from 1st extraction (ppb)

Level of contaminant

from 2nd extraction (ppb)

Percent increase from

3h to 24h

Limonene 725.6 7.6 1.0 Cineole 29.5 ND ND

m-Cymene 50.0 4.3 8.6 3-Ethyl-o-xylene 13.1 0.4 3.1

γ-Terpinene 32.3 0.3 0.9 1,2,3,5-, Tetramethyl

benzene 10.2 0.5 4.9

Menthone 17.1 1.0 5.8 Dodecane 105.7 1.4 1.3

Naphthalene 28.7 0.5 1.7 Methyl salicylate 498.1 1.8 0.4 4-n-Propyl anisole 41.4 0.4 1.0

Biphenyl 22.1 0.2 0.9 Tetradecane 73.5 5.2 7.1

2-Methylnaphthalene 123.8 0.6 0.5 1-Methylnaphthalene 64.4 0.6 0.9 1-Ethylnaphthalene 28.4 0.2 0.7

2,6- Dimethylnaphthalene 103.4 1.0 1.0 1,7- Dimethylnaphthalene 82.9 0.6 0.7 1,6- Dimethylnaphthalene 68.3 0.6 0.9 1,4- Dimethylnaphthalene 18.4 0.2 1.1 1,2- Dimethylnaphthalene 10.4 0.2 1.9

Tetradecane 105.7 1.4 1.3 Trimethylnaphthalene

Isomer 1 10.3 ND ND

Trimethylnaphthalene Isomer 2

9.7 ND ND

Trimethylnaphthalene Isomer 3

12.3 ND ND

Trimethylnaphthalene Isomer 4

10.6 ND ND

Trimethylnaphthalene Isomer 5

7.7 ND ND

Trimethylnaphthalene Isomer 6

12.4 ND ND

n-Hexylbenzoate 4.1 0.2 4.9 Lauric acid 151.3 6.4 4.2

Benzophenone 231.1 1.3 0.6

139

Chapter 4

4.3.5 Comparison of contaminant levels in different 70 g grabs from the original 15 kg

sample

The data in Table 4.8 and Figures 4.7-4.10 enable us to gain some insight into the

variability of the distribution of contaminants. The data in Figures 4.7-4.10 were obtained

from one random grab, grinding and analysis, whilst the Table 4.8 results were obtained

from a later random 70 g grab. As both 70 g grabs were taken from the original 15 kg sack

of washed and shredded PET sampled continuously from the recycling plant in June 2000,

the differences represent a very short-term variation in composition. Figure 4.32 is a plot

of the amount extracted from 0-300 µm particles derived from the later 70 g grab (by

sonication with DCM), versus the amount extracted from the 0-300 µm particles derived

from the earlier 70g grab (by Soxhlet with DCM). The log-log mode has been used to even

out the distribution of data points and allow equal inspection of the majority of the points

occurring at the low levels of contamination. The most obvious feature of this plot is that

the line of best fit to the points (continuous line, y = 1.0539x – 0.1265) and the theoretical

line of perfect agreement – on average – between the separate grabs (dashed line, y=x) are

virtually coincident. In fact, without the low-lying outlier of the set, it is unlikely that any

difference could be discerned. This is perhaps not surprising. The 15 kg of shredded PET

(from which both grabs were taken) would be derived from a finite albeit large number of

bottles and the two grabs, with many hundreds of flakes each (est. 1800), would

substantially contain differing proportions of the same bottles. On the other hand, R2 =

0.84 indicates significant scatter around the line of best fit, as can readily be seen.

Furthermore, it must be remembered that the log-log scale has the effect of appearing to

minimize difference. For example, the two most concentrated contaminants at the top

right-hand end of the plot lie about 0.2 log units above and below the y = x line and are thus

around 1.58 times larger and smaller (respectively) in the later grab than in the earlier grab.

A 58% difference is considerable, although the majority of contaminants are not present in

the two grabs in levels that are this different. [It should be noted that the spread of data in

Figure 4.32 is far more than can be attributed to differences in extraction efficiencies of

alternative extraction techniques. This is apparent from both Table 4.8 and from validation

studies to be presented in the next section (Section 4.3.6)].

140

Chapter 4

Figure 4.32: Variation in contaminant levels between two 70 g grabs of flake from the

original 15 kg bag. Analyses were performed on PET ground to the 0-300 µm particle size

in each case.

y = 1.0539x - 0.1265R2 = 0.8362

0

0.5

1

1.5

2

2.5

3

3.5

4

0 1 2 3 4

Log[amount extracted from earlier grab (ppb)]

Log[

amou

nt e

xtra

cted

from

late

r gra

b (p

pb)]

0-300 um Linear (0-300 um)

y=x

141

Chapter 4

4.3.6 Validation of the Soxhlet extraction methodology

4.3.6.1 Total dissolution compared with Soxhlet extraction

Table 4.11 shows the results of Soxhlet extraction using DCM compared with levels

obtained by total dissolution using TFA and subsequent partitioning into heptane (for

extraction from the two larger particle size ranges and from the unground flake).

The values in Table 4.11 are for those contaminants with very similar concentrations

determined by both methods [with the exception of the concentrations in column TFA (2)

whose inconsistent results will be discussed later]. Amongst the data, excluding the values

in column TFA (2), there are clearly two extreme outliers, but only in the case of the analysis

on the unground flake (methyl salicylate and lauric acid). The relationships between the

results for the two methods are the opposite for these two contaminants, and because the

data is so extreme, we have excluded them from further considerationφ.

Figure 4.33 is a plot of the amount extracted after total dissolution versus that extracted by

Soxhlet, excluding the values in column TFA (2). The log-log mode has again been used to

even out the distribution of data points and allow equal inspection of the majority of the

results, which occur at the lower levels of contamination. And again the most obvious

features of these plots are that the data sets overlap strongly with each other and with the

theoretical line for perfect agreement between the techniques (y=x), and the R2 values are

reasonable – especially for the two particle size ranges, for which the regression lines are

not significantly different. On the other hand, the data for the unground PET flake is more

scattered. This is to be expected, as the small number of flakes analysed by each technique

are more likely to have different contaminant levels, whereas the large number of small

particles obtained after grinding will more closely approach a representative sample by

averaging over a mass approximately 10 times as large. Once again, the need for particle

size reduction is demonstrated (Cross 2000).

There are some exceptions to the general agreement between concentrations obtained by

total dissolution and those determined by Soxhlet [Table 4.12 (a) & (b)]. For the medium

particle size (>300-425 µm), the large particle size (>425-700 µm) and the unground flake,

respectively, the contaminant levels (in ppb) determined by Soxhlet were 918±162,

φ Methyl salicylate is very polar and its extraction in heptane poor. Conversely, lauric acid is better extracted in heptane (after TFA dissolution) than by Soxhlet extraction with DCM.

142

Chapter 4

1102±169 and 1139±274 for limonene, 77±12, 104±51 and 75±75 for cineole, and, 41±7,

44±8 and 42±10 for γ-terpinene. After total dissolution with TFA, none of these

contaminants were detected in any of these samples. The unground flake analyses for

cineole appear to be an example of the problem of unrepresentative sampling referred to in

the previous paragraph as the individual analyses were 139, 16, 139 and 5 ppb.

Table 4.11: Levels of contaminants (and their standard deviations) determined by Soxhlet

extraction with dichloromethane (DCM) compared with total dissolution by trifluoracetic

acid (TFA) followed by extraction with heptane: for flake ground to medium and large

particle sizes, and for unground flake. (All levels are in ppb by mass.)

PARTICLE SIZE >300-425 µm >425-700 µm Unground flake

CONTAMINANT DCM TFA DCM TFA DCM TFA (1) TFA (2)

1,2,4-Trimethyl-benzene

3-Ethyl-o-xylene

1,2,3,5- Tetramethyl-benzene

(-)-Menthone

Methyl salicylate

4-n-propylanisole

Naphthalene

Biphenyl

2-Methylnaphthalene

102 (6)

8.9

(1.1)

10.4 (1.3)

24.2 (7.9)

302 (47)

25.8 (5.3)

33.9 (2.6)

23.9 (3.5)

146 (25)

50.1 (7.6)

7.6

(0.1)

9.4 (0.3)

35.0 (3.5)

317 (70)

31.0 (6.9)

38.7 (1.9)

22.9 (1.2)

141 (8)

25.7 (6.7)

5.2

(1.3)

4.4 (0.8)

17.7 (5.1)

138 (60)

30.9 (4.0)

28.5 (2.9)

16.2 (3.6)

69.7 (6.4)

21.8 (6.6)

3.2

(0.3)

4.7 (0.5)

19.5 (9.3)

95.1

(84.7)

36.0 (9.9)

23.9 (1.0)

17.0 (4.3)

71.9 (1.3)

20.0 (17.3)

4.5

(1.8)

13.9 (1.6)

1.2

(0.4)

897 (644)

94.3

(26.9)

32.6 (1.9)

18.4 (1.9)

96.6 (4.7)

53.3 (10.5)

11.2 (1.1)

15.1

(10.5)

3.0 (0.2)

56.8

(27.3)

42.6 (13.1)

25.6 (3.9)

18.4 (1.7)

82.9 (14.5)

43.6 (9.3)

33.3 (3.8)

32.0 (4.4)

n/a2

177.6 (25.8)

48.1

(22.2)

38.3 (1.0)

54.1 (9.0)

105 (2)

2 N/a = not analysed

143

Chapter 4

1-Methylnaphthalene

1-Ethylnaphthalene 2,6-Dimethylnaphthalene

1,7- & 1,6- Dimethylnaphthalene

1,4- Dimethyl naphthalene

1,2- Dimethyl naphthalene

Trimethyl naphthalene

isomer 1

Trimethyl naphthalene isomer 2

Trimethyl naphthalene

isomer 3

Trimethyl naphthalene isomer 4 & 5

Trimethyl naphthalene

isomer 6

Trimethyl naphthalene isomer 7

Benzophenone

Lauric acid

n-Hexyl benzoate

74.0 (7.6)

26.6 (4.2)

84.3

(13.6)

206 (6)

37.3 (4.5)

17.3 (3.2)

10.0 (1.0)

10.9 (1.1)

12.1 (2.2)

19.0 (2.3)

13.3 (1.7)

6.6

(0.8)

664 (106)

62.2

(10.2)

3.1 (0.2)

73.5 (2.4)

27.3 (2.3)

88.9

(11.6)

192 (20)

39.9 (3.3)

19.8 (1.3)

11.9 (1.2)

11.7 (1.8)

12.8 (1.2)

23.6 (1.9)

15.7 1.4)

7.7

(0.8)

663 (55)

53.5 (9.8)

3.6

(1.2)

38.9 (1.3)

13.6 (0.5)

43.5 (2.2)

99.5 (7.0)

21.4 (1.7)

12.1 (1.4)

6.3

(0.7)

6.2 (0.5)

8.0

(0.8)

11.4 (0.9)

8.2

(0.5)

3.6 (0.4)

311 (5)

703

(101)

2.1 (0.2)

35.8 (2.7)

13.7 (0.7)

46.1 (1.6)

98.3 (1.0)

20.6 (1.3)

13.4 (0.6)

6.4

(0.4)

5.2 (0.7)

5.8 (1)

12.6 (0.9)

8.2

(0.9)

4.1 (0.3)

275 (12)

570

(119)

3.5 (2.2)

46.8 (5.0)

21.0 (1.5)

69.0

(12.2)

136 (24)

21.4 (3.7)

10.7 (0.5)

6.2

(0.8)

5.4 (0.6)

7.9

(1.4)

14.1 (2.5)

8.9

(1.5)

4.2 (0.8)

309 (24)

12.0 (0.9)

18.9

(15.0)

42.5 (6.9)

18.6 (3.1)

62.2 (4.5)

119 (18)

25.3 (2.5)

9.1

(1.5)

6.6 (1.2)

4.5

(1.1)

5.1 (1.6)

11.4 (2.7)

6.9

(1.4)

3.6 (1.1)

258 (32)

683

(310)

16.0 (0.5)

66.1 (5.8)

53.1 (6.3)

211 (32)

242 (24)

83.7 (8.1)

27.9

(18.8) 41.7 (3.3)

33.2 (5.9)

33.3 (4.1)

57.0 (6.5)

36.1 (3.2)

14.1 (1.4)

257 (24)

740

(235)

2.9 (1.5)

144

Chapter 4

Figure 4.33: Log-log plot of levels of contaminants determined by total dissolution versus

levels extracted by Soxhlet and comparison with the ideal relationship (y=x): for flake

ground to medium and large particle sizes and for unground flake.

y = 0.8535x + 0.1821R2 = 0.8952

y = 0.9332x + 0.1098R2 = 0.9655 y = 0.9513x + 0.0488

R2 = 0.9806

0

0.5

1

1.5

2

2.5

3

3.5

0 0.5 1 1.5 2 2.5 3 3.5

LOG(Amount Extracted by Soxhlet (ppb))

LOG

(Am

ount

Ext

ract

ed b

y To

tal D

isso

lutio

n (p

pb))

unground >300-425 um>425-700 um Linear (unground)Linear (>300-425 um) Linear (>425-700 um)

y=x

145

Chapter 4

Table 4.12: Levels of contaminants (and their standard deviations) determined by Soxhlet

extraction with dichloromethane (DCM) compared with total dissolution by trifluoracetic

acid (TFA) followed by extraction with heptane: anomalies for (a) m-cymene [TFA>DCM]

and (b) limonene, cineole and γ-terpinene [DCM>TFA]. (All levels are in ppb by mass.)

(a)

m-Cymene

Particle size

DCM TFA Medium

(> 300-425 µm)

46.5 (4.4)

439.9 (4.2)

Large (> 425-700 µm)

67.4 (16.0)

439.1 (41.4)

Whole flake 77.2 (4.1)

650.7 (107.6)

(b)

Limonene

Cineole γ-Terpinene Particle size

DCM TFA DCM TFA DCM TFA Medium

(> 300-425 µm)

918 (162)

ND 76.5 (12.3)

ND 40.5 (7.0)

ND

Large (> 425-700

µm)

1102 (169)

ND 104.3 (50.7)

ND 43.6 (8.4)

ND

Whole flake

1139 (274)

ND 78.6 (79.0)

ND 41.8 (10.3)

ND

There was an independent occasion on the same batch of flake, during which unground

flake was extracted in triplicate for a second time by total dissolution using TFA (Table

4.11; see column TFA(2)). However, this time, the levels of most contaminants

reproducibly exceeded those determined by Soxhlet extraction for the unground flake in

Table 4.11.

146

Chapter 4

Figure 4.34 is a log-log plot comparing the contaminant concentrations extracted from

unground flake by TFA on the two different occasions (average of triplicate values for each

contaminant) versus the amount extracted from unground flake by Soxhlet extraction (all

values in Table 4.11). From this graph it can be seen that on the second occasion the level

of contaminants extracted from unground flake by total dissolution exceeded the amount

extracted by Soxhlet extraction. Conversely, on the first occasion the level of contaminants

extracted from unground flake by total dissolution approximated the amount extracted by

Soxhlet extraction. Amongst the data there were three outliers, which have been excluded

from the plot (n-hexyl benzoate, lauric acid and methyl salicylate).

It is possible that, due to sample-to-sample variation, the unground flake extracted by total

dissolution the second time could have contained a higher level of contaminants. For

example, the higher level of naphthalene derivatives could be resulting from the presence of

more polyethylene naphthalate (PEN) in the unground flake extracted the second time by

total dissolution. However, this does not explain the discrepancy for the other compounds.

The gradient of 0.62 in Figure 4.34 suggests that - on average - about 62% of each

contaminant was extracted during the Soxhlet extractions relative to the second set of total

dissolution extractions. However, a correlation coefficient (R2) of 0.64 indicates that the

points are extremely scattered about the line of best fit. The reasons behind the discrepancy

between the first and second set of total dissolution extractions are not known.

On full inspection of the data in Table 4.11, it is interesting that the level of benzophenone

was reproducibly equivalent for all extractions. However, given the relatively small

standard deviations in the multiple extractions (10% for the flake), experimental error

should logically be discounted and the equivalence for the two total dissolutions should be

attributed to coincidence.

147

Chapter 4

Figure 4.34: Log-log plot comparing the contaminant concentrations extracted from

unground flake by TFA versus the amount extracted from unground flake by Soxhlet

extraction.

y = 0.6171x + 0.9412R2 = 0.6415

y = 1.0479x - 0.0605R2 = 0.8168

0

0.5

1

1.5

2

2.5

3

3.5

0 0.5 1 1.5 2 2.5 3 3.5

Log (amount extracted by Soxhlet extraction)

Log

(am

ount

ext

ract

ed b

y to

tal d

isso

lutio

n)

Second TFA extraction First TFA extraction

Linear (Second TFA extraction) Linear (First TFA extraction)

x=y

Where food contact is concerned, two more compounds other than methyl salicylate,

benzophenone and lauric acid were extracted at levels above the 215 ppb threshold during

the second set of total dissolution extractions. These were 1,6-dimethylnaphthalene and 1,7-

dimethylnaphthalene (242 ppb when measured together).

148

Chapter 4

4.3.7 Particle size range and degree of crystallinity

The observed differences in concentration of the contaminants with the PET particle size

could have resulted from the grinding procedure selectively milling the less crystalline

uniaxially orientated fractions (bottle top and bottom) more readily than the more

crystalline biaxially orientated part (bottle mid-section) of the flake. Although polymer

brittleness is renowned for increasing with polymer crystallinity (Schumann and Thiele

1996), the biaxial orientation of the bottle mid-section prevents it from rapid grinding.

Conversely, despite being more amorphous, the bottle top and bottom is uniaxially

orientated and possesses some crystallinity (Table 4.13), which allows the polymer to

acquire brittleness for superior grinding.

The more amorphous regions could potentially be more contaminated than the crystalline

and biaxially oriented regions, as discussed in previous sorption studies (Mitz et al. 1997).

In the current study, DSC experiments confirmed that (the thicker and more transparent

flakes from) the top and bottom are more amorphous than are (the thinner and more opaque

flakes from) the midsection (Table 4.13), and, as can be seen from Table 4.14, the grinding

process appears not to alter the average degrees of crystallinity of the particles in each of

the particles size ranges.

However, for typical ‘grabs’ of the PET flake without segregation into the top and bottom

material and the mid-section flakes, XRD showed there to be little difference in percentage

crystallinity between each particle size range. In particular, it is clear from Table 4.14 that

there was no significant difference in percent crystallinity between the small and medium

particle size ranges, for which the contaminant levels were most different (Figures 4.15-

4.16). Therefore, crystallinity does not account for the differences in contaminant

concentrations between each size here. Furthermore, where there does appear to be a small

but clear increase in percentage crystallinity (between the medium and large-sized

particles), differences in contaminant levels are minimal (but do decrease), so the degree of

crystallinity could be a contributing factor in this case, albeit a weak one.

The other clear aspect of Table 4.14 is that the unsegregated, ground PET has a degree of

crystallinity much closer to that of amorphous flake (tops and bottoms of the bottles, Table

4.13) than to the midsections of the bottles. Assuming that the crystallinity values in Table

4.14 are a simple average of the components given in Table 4.13, then (averaged over the

149

Chapter 4

two batches), there is 93% of the amorphous material in the two smaller particle size ranges

(0-300 µm, >300-425 µm) in the unsegregated, ground PET and 70% in the large particle

size range (>425-700 µm). The calculations for these percentages are shown in the

appendix 2. The percentage of amorphous material in the unsegregated ground flake was

also estimated gravimetrically by randomly obtaining a handful of flake, segregating the

flake into ‘amorphous’ and ‘crystalline’ material by visual inspection, and then separately

grinding the material, sieving it into the three particle size ranges and weighing the isolated

particles. The masses of each of the particle size ranges for the crystalline and amorphous

flake are shown in Table 4.15. The percentages resulting from the amorphous fraction of

the random handful of flake were calculated to be 68%, 64% and 57% for the small,

medium and large particle size ranges respectively. This experiment further shows that the

amorphous material of the flake is more easily ground than the biaxially oriented crystalline

material, considering the initial amount of crystalline and amorphous material was 48g and

19g respectively.

Table 4.13: Percentages of crystallinity for amorphous and crystalline fractions of washed

and dried flake ground to different particle sizes.

Mid section of bottle crystallinity

(75.4% by mass in total sample)

Top and bottom of bottle crystallinity

(24.6% by mass in total sample)

Ground to 0-300 µm

48.41 % 19.61 %

Ground to >300-425 µm

51.85 % 19.22 %

Ground to >425-700 µm 48.93 % 12.94 % Table 4.14: Percentages of crystallinity for two batches of unsegregated washed and dried

flake ground to different particle sizes.

Batch 1 Batch 2

Ground to 0-300 µm

20.6 % 22.8 %

Ground to >300-425 µm

20.9 % 22.2 %

Ground to >425-700 µm 23.0 % 23.6 %

150

Chapter 4

Table 4.15: Mass of amorphous and crystalline flake ground to different particle sizes.

Amorphous (19 g) Crystalline (48 g)

% Due to amorphous

Ground to 0-300 µm

1.61 g (8.5%) 0.75 g (1.6%) 68

Ground to >300-425 µm

1.59 g (8.4%) 0.88 g (1.8%) 64

Ground to >425-700 µm 3.12 g (16.4%) 2.41 g (5.0%) 57

Ungrounded

12.7 g (67%)

43.9 g (91%)

22

Similarly, from Table 4.13, given that the original percentage of the more amorphous

material was only 24.6%, there is no doubt that flakes derived from the softer, amorphous

sections of the top and bottom of the PET bottles are selectively ground in the particle size

reduction process. However, as the degrees of crystallinity are generally the same for each

particle size range (Table 4.14), differences in crystallinity do not account for differences in

contaminant concentrations between each particle size.

It is therefore suggested that the increase in contaminant concentration with smaller particle

size could be attributed to the size reduction process selectively grinding the highly

contaminated surface of the flake to 0-300 µm. The material ground to larger particles

sizes would then contain a proportionately larger quantity of material derived from the

interior of the plastic where it is less likely to have significant contact with impurities.

As a possible test of the hypothesis, the percentages extracted from the largest particle size

relative to that extracted from the smallest particles size have been plotted versus

contaminant molecular weight in Figure 4.35.

Two of the 32 compounds (1,2,4-trimethyl benzene and 3-ethyl-o-xylene) have percentages

greater than 100. As these are two of the compounds with large standard deviations (A0-300

µm/A>300-425 µm = 1.5 ± 0.7 for both compounds) and are contrary to the general trend, they

have been eliminated. Also, for simplicity, the results for the four dimethylnaphthalene

isomers have been averaged to yield one representative point on the graph. Similarly, the

ratios for the six trimethylnaphthalene isomers have been averaged.

151

Chapter 4

Figure 4.35: Percentage of amount extracted from the >425-700 µm particle size range to

the amount extracted from the 0-300 µm particle size range plotted versus contaminant

molar mass.

y = 630.52e-0.018x

R2 = 0.4326

0

10

20

30

40

50

60

70

80

90

100

100 120 140 160 180 200 220Contaminant MW

100(

Am

ount

Ext

ract

ed fr

om 4

25-7

00 u

m/A

mou

nt in

< 3

00 u

m)

There is a weak negative correlation (R2 = 0.43), but one that is consistent with the original

contamination process of the PET bottles, if the small particles are disproportionately

derived from the surface of the bottles.

The alternative possible explanation for the correlation in Figure 4.35 lies in the diffusion

of contaminants into crystalline regions being smaller than into amorphous regions of the

bottle and decreasing as the size of the contaminant increases (Begley et al. 2002).

However, as we have already pointed out, the data in Table 4.14 indicates that there is no

significant change in crystallinity between the small and medium size particles where the

152

Chapter 4

decrease in contaminant levels is large. On the other hand, the small increase in

crystallinity between the medium and large size particles may contribute to the small

decrease in contaminant levels. (The equivalent plot of the percentage extracted from the

medium particle size relative to that extracted from the smallest particle size versus

contaminant molecular weight had R2 close to 0.25, demonstrating a weaker correlation.)

4.3.8 Investigation of the relative levels of contaminants in the two types of flake

Washed and dried flake was segregated into amorphous and crystalline particles, ground

individually to the three particle size ranges and then analysed by the sonication method.

Figure 4.36 is the log-log plot of amounts of contaminants in the more crystalline particles

versus the amounts in the amorphous particles for each particle size range and presents an

overview of the comparison.

The group of points centrally located below the lines of best fit are all related to three

compounds: cineole, (-)-menthone and n-hexylbenzoate, the last of which compounds is

also responsible for the two points on the baseline near the origin. (See Table 4.16 for

individual details.) Analyses for these compounds would be clearly compromised by the

selective grinding of the amorphous PET flakes because of the very different levels

apparent in the amorphous and crystalline flake. A much more rigorous procedure would

be required for representative sampling in these cases (Cross 2000). For the rest of the

contaminants, the vast majority of the data points are uniformly spread about the ideal

(y=x) line in a random fashion, although, given the log-log scale it is clear that thee are still

significant differences in the measured values for some contaminants.

Close examination of Table 4.16 shows, for example, that cymene and limonene seem to be

absorbed more strongly onto the crystalline phase and would again require more careful

sampling procedures after sample size reduction. (Triplicate Soxhlet analyses of unground

flake after segregation into the amorphous and more crystalline shreds strongly support this

finding, with 100x(the amount in the more crystalline material/the amount in the

amorphous phase) = 314% for limonene and 514% for cymene.) A possible explanation for

this finding is that these compounds derive from soft drink and the contact times between

them and the crystalline and amorphous parts of the bottle are not the same (the top of the

bottle, which is amorphous is not usually in permanent contact with the soft drink).

153

Chapter 4

For 2-methylnaphthalene and 1-ethylnaphthalene Table 4.16 indicates a slight bias towards

absorption into the amorphous phase and the triplicate Soxhlet analyses again support the

findings, with 74% and 70% respectively being found in the more crystalline phase, relative

to the amorphous phase. On the other hand, from Table 4.16 the six trimethylnaphthalene

isomers appear to be ideal analytes in that there is negligible discrimination between the

levels measured in the two phases, and, for the remainder of the contaminants there appears

not to be clear discrimination. In these cases, the tripicate Soxhlet analyses generally

indicated the amorphous phase to be favored, with relative amounts in the more crystalline

phase varying between 62% and 86%.

154

Chapter 4

Figure 4.36: Log amount extracted from crystalline particles versus log amount extracted

from amorphous particles, for each particle size range.

y = 1.1436x - 0.3332R2 = 0.8395

y = 1.034x - 0.1084R2 = 0.9051

y = 0.9148x + 0.0256R2 = 0.8079

0

0.5

1

1.5

2

2.5

3

3.5

4

0 1 2 3 4

log [Amount in Amorphous]

log

[Am

ount

in C

ryst

allin

e]0-300 um >300-425 um>425-700 um Linear (0-300 um)Linear (>300-425 um) Linear (>425-700 um)

y=x

155

Chapter 4

Table 4.16: Levels of contaminants in amorphous and crystalline flake ground to different particle sizes

(analysed by sonication in DCM for 3 h). (All levels are in ppb by mass.)

Amorphous Flake (top and bottom of the bottle)

Crystalline Flake (midsection of the bottle)

Contaminants

0- 300 µm

>300-425 µm

>425-700 µm

0-300 µm

>300-425 µm

>425-700 µm

1,2,4-, Trimethylbenzene

42.5 17.7 14.0 31.9 18.8

12.5

m-Cymene 54.0 34.4 33.1 100.7 62.3 46.4

(R)-(+)-Limonene

786.7 410.9 286.7 1133 460.7 407.5

Cineole

42.1 48.9 116.5 10.7 10.1 6.3

γ-Terpinene

61.1 34.4 28.5 61.1 23.1 18.8

3-Ethyl-o-xylene

10.7 5.4 4.2 8.9 4.3 3.1

Benzene, 1,2,3,5- tetramethyl-

21.4 13.7 11.7 18.7 10.3 8.6

(-)-Menthone

53.7 25.7 22.4 11.8 37.3 4.1

Methyl salicylate

n/a n/a n/a n/a n/a n/a

4-n-Propylanisole

40.8 17.4 15.0 36.6 30.5 29.3

Naphthalene

93.8 35.5 29.6 81.0 43.9 30.7

n-Dodecane 211.3 122.4 172.9 192.1 135.7 142.9

2-Methylnaphthalene

174.7 94.9 82.8 155.1 66.7 70.2

1- Methylnaphthalene

225.9 121.6 68.6 257.3 68.1 50.3

Biphenyl

30.9 11.9 12.6 23.4 15.6 13.7

1-Ethylnaphthalene

63.0 28.9 22.0 56.3 26.2 17.8

2,6- Dimethylnaphthalene

157.6 71.2 46.9 159.9 57.6 46.9

156

Chapter 4

Tetradecane

284.4 196.3 225.6 313.8 192.3 238.9

1,2- Dimethylnaphthalene

84.5 50.9 26.1 102.0 37.6 19.3

1,7- & 1,6-Dimethylnaphthalene

407.5 216.3 155.0 468.8 166.3 112.5

1,4- Dimethylnaphthalene

161.8 85.0 51.9 201.2 56.1 36.6

Trimethylnaphthalene Isomer 1

29.7 14.7 10.5 30.6 13.5 10.5

Trimethylnaphthalene Isomer 2

22.5 10.7 6.6 24.9 9.1 7.2

Trimethylnaphthalene Isomer 3

31.8 18.3 11.3 28.4 13.9 11.2

Trimethylnaphthalene Isomer 4

20.1 10.7 6.8 23.4 10.4 6.5

Trimethylnaphthalene Isomer 5

22.6 9.2 5.6 23.4 8.8 6.2

Trimethylnaphthalene Isomer 6

29.1 16.0 9.4 33.3 13.7 8.7

Dodecanoic acid

n/a n/a n/a n/a n/a n/a

n-Hexylbenzoate

33.3 2.3 0.0 3.4 0.0 0.0

Benzophenone 955.2 473.9 578.4 1164.2 720.1 453.4

However, there are some further compounds indicated by the triplicate Soxhlet analyses to

be favored more in the more crystalline phase: lauric acid 143%, n-hexyl benzoate 221%,

1,2,4,-trimethylbenzene 164% and 3-ethyl-o-xylene 118%. However, notwithstanding the

Soxhlet analysis being averages of triplicates, where these do not agree with the results in

Table 4.13, care should be taken in placing too much emphasis upon them since the

sampling error is likely to be significant. In three cases (cineole, methyl salicylate and 4-n-

157

Chapter 4

propylanisole) the measured values varied by as much as an order of magnitude, with the

ranges for the two phases overlapping. These are clear examples of the heterogeneity

between flakes. Thus we believe the Table 4.16 data to be generally more reliable.

The gradients of the individual plots in Figure 4.36 approximate 1, but can be seen to

systematically decrease as the particle size increases. However, the intercepts tend to

compensate for the deviations in the gradients relative to 1, and we interpret these

deviations in the gradient as the result of the relatively few data points at the extremities of

the sets, combined with the effects of outliers. Generally, the levels of contaminants in the

crystalline and amorphous particles resemble each other, so that the selective grinding of

the amorphous flake will not lead to significant errors in the analysis for most of the

contaminants.

The detailed sonication data in Table 4.16 very strongly support the Soxhlet data in Figures

4.15-4.17 in that the levels of contaminants in both amorphous and crystalline PET

decreased significantly from the 0-300 µm particle size range to the larger particle size

ranges. In turn, the hypothesis of selective grinding of the more contaminated flake surface

to a smaller size relative to the less contaminated inner core was further strengthened.

In order to circumvent the effect of selective grinding when studying the influence of path-

length on extraction, the amorphous PET from the washed and dried flake was flattened

using a hydrolytic press and extracted by DCM sonication. The GC/MS results for this

extract were compared to those determined for the unflattened (amorphous) flake, which

were also extracted by sonication.

Figure 4.37 is the log-log plot of amounts of contaminants in the unflattened particles

versus the amounts in the flattened particles for each particle size range. The gradient of

this plot approximates unity, indicating that there is negligible difference in the amount of

contaminant extracted from unflattened amorphous pellets relative to the flattened

amorphous pellets. This result further supports the presumption that the variation in

contamination with particle size could have resulted from selective grinding rather than a

decrease in extraction efficiency with increasing particle path length.

158

Chapter 4

Figure 4.37: Log amount extracted from whole amorphous pellets versus log amount

extracted from flattened amorphous particles.

y = 0.9897xR2 = 0.9659

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

-0.5 0 0.5 1 1.5 2 2.5 3 3.5

Flattened amorphous (log conc)

Who

le a

mor

phou

s (lo

g co

nc)

4.3.9 Representative sampling

Obtaining representative samples small enough for analysis, from bulk materials that are

heterogeneous in nature, is always problematic and multi-layered (Cross 2000). For used

PET, first there is the heterogeneity of the bulk shredded, washed and dried, curbside

collected material. This heterogeneity is itself bi-faced; the bottles have variable histories

and thus contaminant profiles and concentrations and the amorphous PET in the tops and

bottoms of the bottles may or may not contain the same levels of contaminants as the

midsections of the same bottles (Table 4.16). Putting aside the low frequency occurrences

of contamination by pesticides, herbicides and other poisons, lubricating oils etc. – all of

which would be diluted and eventually internalized in the recycling extrusion process, and

none of which have been detected in these studies – a large sample of shredded flake is

required to reasonably represent the variable histories of the collected bottles. From the

159

Chapter 4

previous section, it seems that some contaminants are more concentrated in either of the

structurally different parts of the bottle and some are not. Therefore it depends upon the

contaminants of interest as to whether the material from the tops and bottoms of the bottle,

and the shredded midsections must be sampled in proportion to their amounts. If

proportional sampling is needed, the second problem is how to achieve it, although the

sampling of bulk materials is well understood (Smith and James 1981).

Whichever is the case, particle size reduction is generally essential. Our analyses of

unground flake by alternative methods led to far greater discrepancies than were observed

for the same procedures compared for the large or medium sized particles (Figure 4.33).

This is simply a reflection of the number of particles (flakes) sampled. The only way to

circumvent this problem and eliminate the need for particle size reduction is analyse the

same mass of material as would be used to generate the ground material. In terms of

dissolution or sonication this involves economically and environmentally unacceptable

amounts of solvents, but with the capital-expensive accelerated solvent extraction, large

volume extraction cells are a very attractive alternative that removes the need for particle

size reduction and the subsequent sampling of the distribution of particle sizes generated.

Disregarding the few contaminants at the top of Table 4.11 (that tend to be misleading) and

those in column TFA (2), for the majority of compounds in this study, the levels of

contaminants in the unground flake appear to lie between those measured in the medium

and large sized particles or not far below that range of concentrations. Therefore, from

Table 4.16 there are two compounds whose levels relative to the FDA threshold will

depend upon the particle sizes analysed; 1-methylnaphthalene and the 1,6- and 1,7-

dimethylnaphthalene isomers, if measured together.

4.3.10 Levels of contaminants in flake and the threshold of regulation

The contaminants identified in the flake above the 215 ppb FDA threshold were limonene,

methyl salicylate, benzophenone, lauric acid and dodecanoic acid. Benzophenone, a well-

known surrogate used in challenge tests, has been reported by researchers in the past to be

the most difficult analyte to remove, possibly due to the good solubility match between

benzophenone and PET (Franz et al. 1998, Harding et al. unpublished). Limonene has

been found to occur at low ppm levels (1.5 ppm to 11.0 ppm) in washed and dried

160

Chapter 4

commercial flake by Franz and Welle (1999b). Bayer (2002) reported even higher levels of

limonene (18 ppm) and methyl salicylate (15.3 ppm) after commercially washing the PET.

Triantafyllou et al. (2002) identified limonene in washed and dried PET flake at levels

between 2.5 and 15 ppm. Despite the high concentrations obtained for limonene, PET has

been shown to absorb the least amount of aroma components compared with low-density

polyethylene, and an ionomer (Gavara et al. 1997). Van Willige et al. (2002) obtained

similar results for the sorption of flavor compounds onto PET, low-density polyethylene,

polypropylene and polycarbonate. Fortunately, PET has the advantage of having a low

sorption capacity compared to polyolefins, PC (polycarbonate) and EVOH making it more

suitable for use as a food packaging (Nielsen et al. 1992, Nielsen 1994, Van Willige et al.

2002, Imai et al. 1990, Gavara et al. 1997).

The amounts of contaminants obtained by Soxhlet extraction represent the maximum level

of migration. In the actual contact situation such an amount may never be attained, or

excessively long-term storage may be required to reach that level. During the challenge

tests carried out by Visy (Harding et al. unpublished, Cross et al. unpublished) it was

observed that the levels of contaminants present in deliberately contaminated washed and

dried flake exceeded the real levels of contaminants determined during this study; yet the

recycled material was deemed suitable for food contact use due to low migration levels.

In the next Chapter we will examine whether the levels of these contaminants exceed the

FDA threshold after extrusion into pellets.

161

Chapter 5

CHAPTER 5 SEMI-VOLATILE CONTAMINANTS AND LEVELS OF OCCURRENCE IN

EXTRUDED PET PELLETS FROM CURBSIDE COLLECTION

5.1 GENERAL INTRODUCTION

5.1.1 Purpose of the chapter It was reported by Harding et al. (unpublished) that the wash stage of the Visy Plastics

recycling process removes surrogate contaminants by 75.5 % - 96.5%. The subsequent

extrusion step, which involves melting the washed and dried shredded PET (flake)

under vacuum, further reduced the residual contamination levels, achieving overall

reduction of the initial levels by 93.8% - 99.7%. Therefore this latter stage is viewed as

an important treatment in terms of decontaminating simulated post-consumer PET. It

should be realised, however, that high removal efficiency does not always result in a

low residual concentration. As determined previously by Cross et al. (unpublished),

further migration experiments were necessary because the level of each surrogate in the

final recycled material nonetheless exceeded the 215 ppb threshold. It was concluded in

that paper that the level of each “specific” contaminant migrating into a soft drink

simulant proposed by the FDA (10% ethanol) was within the maximum migration

specification of 10 ppb, deeming the material suitable for food-contact use. Other

researchers have reported similar results from their migration studies (Franz and Welle

1999a, Komolprasert et al. 1997).

In terms of food contact approval these were encouraging results, however the reality

was that the levels of specific surrogate chemicals in the recycled material were still

above the 215 ppb threshold.

It has been already shown in Chapter 4 that the levels of most contaminants in curbside

washed and dried shredded PET (flake) were below the 215 ppb limit. Contaminants

exceeding this limit were limonene, methyl salicylate, benzophenone, lauric acid and a

total of the dimethylnaphthalene derivatives. It was anticipated that the level of these

contaminants would decrease below 215 ppb after extrusion of the flake due to the

extreme temperature and vacuum conditions involved in this treatment and the low

levels of post-consumer contaminants in washed and dried flake. The major reason

why the concentration of the challenge compounds (surrogates) was not reduced below

215 ppb (Harding et al. unpublished) is because the initial concentrations used in

162

Chapter 5

challenge tests are intentionally exaggerated to represent a worst-case scenario (FDA

1992, Komolprasert et al. 1997, Franz et al. 1998). Such large concentrations from

consumer misuse/reuse are not expected to occur in real life. In fact, a 10-fold safety

factor is fixed per bottle with 100% of the bottles deliberately contaminated with

challenge chemicals (FDA 1992). In a real situation it is estimated that 0.1% of bottles

(1 in 1000 bottles) is contaminated by consumer abuse whilst the contaminant level

resulting from the portion of bottles used for household cleaners and detergents is 1%

(1 in 100 bottles). Therefore a factor of 1,000 - 10,000 is included in the overall

challenge test.

This chapter focuses on the optimization of the Soxhlet extraction of contaminants from

extruded (melted at 170-215°C under vacuum) PET pellets before and after they are

thermally crystallised in the quest to determine whether these high temperature

purification steps decontaminate curbside washed and dried flake to acceptable

contaminant levels (below 215 ppb).

As was the case in our studies of the washed and dried flake (Chapter 4), the

parameters optimised (for the pellets) were particle size and extraction time in an effort

to ensure complete recovery. As already shown in Chapter 4, extraction with DCM

was considered appropriate for the extraction of free, diffusible contaminants out of

washed and dried flake; therefore it was subsequently used as the extraction solvent for

extruded PET pellets. In addition, as presented in Chapter 4, weight uptake

experiments confirmed that DCM was the most aggressive solvent towards swelling

PET pellets (relative to hexane, acetone, chloroform, ethyl acetate, 2-propanol and

ethanol).

It was suggested in previous papers (Feigenbaum et al. 2002, Riquet and Feigenbaum

1997) that if a solvent is soluble in a plastic, it opens up the polymer matrix and

therefore facilitates the diffusion of potential migrants. Sorption rates (i.e., diffusion

coefficients) of DCM into amorphous and crystallised PET pellets were determined

assuming Fickian behaviour and discrepancies were interpreted in terms of differences

in polymer morphology. Analogously, extraction rates (i.e., diffusion coefficients) of

contaminants out of amorphous and crystallised PET into DCM were determined.

163

Chapter 5

5.1.2 Brief outline of this chapter

This chapter presents the Soxhlet extraction kinetics of annealed extruded pellets

ground to three different particle size ranges (0-300µm, >300-425µm and >425µm -

700µm). The effects of extraction time and particle size on the levels of contaminants

extracted were subsequently investigated.

Whole amorphous and annealed extruded pellets were extracted with time and the

effects of crystallinity on contaminant extraction and DCM sorption kinetics were

discussed using calculated diffusion coefficients.

In addition, the amorphous pellets were flattened using a hydraulic press (pressure = 8

tons) to reduce the path-length of the amorphous pellets prior to extraction. The effect

of flattening on the levels of contaminants was studied.

Finally, comparisons in contaminant levels were made between washed and dried flake

and extruded (amorphous) pellets, and conclusions were drawn regarding the use of the

recycled material for food-contact purposes.

164

Chapter 5

5.2 KINETICS OF SOXHLET EXTRACTION FROM EXTRUDED AND

ANNEALED PET

The importance of reducing polymer particle size in extraction studies is well

documented (Perlstein 1983, Spell and Eddy 1960, Ashraf-Khorassani et al. 1991).

Smaller particle sizes give rise to shorter diffusion paths for the mass transfer of

contaminants to the surface of the polymer particle. Therefore extraction is normally

more rapid when PET is ground (Vandenburg et al. 1997).

Since completely amorphousφ extruded pellets are not naturally brittle enough to grind

(polymer brittleness increases with crystallinity), the effect of particle size on the

extraction of amorphous pellets – and thus the examination of possible variations in

concentrations on the surface and in the bulk – could not be determined. Hence in order

to assist with the grinding process, amorphous pellets were made brittle by thermal

crystallization known as “annealing”, which involves heating the pellets in an oven set

at 150ºC for 1 hour with occasional stirring. Successful grinding meant that the effect

of pellet size reduction on the extraction kinetics of extruded PET could be examined.

Furthermore, since the annealing of extruded PET pellets normally precedes soft drink

bottle manufacture, the levels of contaminants determined in heat-treated extruded PET

are highly relevant. Hence it is of interest to determine whether “annealing” is an

efficient decontamination step.

5.2.1 Pellets ground to 0-300 µm

In our first experiments, finely ground annealed pellets (0-300 µm) were Soxhlet

extracted for 1, 3, 8, 24, 48 and 72 h to determine the minimum time required for

exhaustive extraction of contaminants. Figures 5.1 – 5.3 indicate that 1 h Soxhlet

extraction is clearly inadequate to entirely extract contaminants out of the fine-ground

pellets. All of the plots exhibit (at least) a localized maximum at 3 h, most display a

minimum at 8 h followed by higher values at 24 and 48 h and finally a reduced value at

72 h. These features appear to be artifactual and are no doubt due to a combination of

the fact that only a single experiment was done to determine each point, and, the

presence of some sample heterogeneity between the samples used for each time interval.

It seems clear that 3 h is sufficient for total extraction and that the best estimates of the φ The reference by Schumann and Thiele (1996) states that the PET melt and polymer chips are completely amorphous. They can then be made crystalline by temperature or mechanical deformation.

165

Chapter 5

levels of the contaminants from these experiments will be the average of the last five

data points.

Figure 5.1: Soxhlet extraction kinetic study of annealed pellets ground to 0-300 µm.

Compounds identified at levels below 10 ppb.

0

2

4

6

8

10

12

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75

Time (h)

Am

ount

Ext

ract

ed (p

pb)

m-Cymene Cineole3-Ethyl-o-xylene 1,2,3,5-Tetramethylbenzene(-)-Menthone Dodecane4-n-Propylanisole Tetradecane2-Methylnaphthalene 1-Methylnaphthalene1-Ethylnaphthalene 2,6-Dimethylnaphthalene1,7-Dimethylnaphthalene 1,6-Dimethylnaphthalene1,4-Dimethylnaphthalene 1,2,4-Trimethylbenzenen-Hexylbenzoate 1,2-Dimethylnaphthalenegamma-Terpinene

166

Chapter 5

Figure 5.2: Soxhlet extraction kinetic study of annealed pellets ground to 0-300 µm.

Trimethylnaphthalene isomers.

0

0.5

1

1.5

2

2.5

3

0 10 20 30 40 50 60 70

Time (h)

Am

ount

Ext

ract

ed (p

pb)

Trimethylnaphthalene Isomer 1 Timethylnaphthalene Isomer 2Trimethylnaphthalene Isomer 3 Trimethylnaphthalene Isomer 4Trimethylnaphthalene Isomer 6 Trimethylnaphthalene Isomer 5

167

Chapter 5

Figure 5.3: Soxhlet extraction kinetic study of annealed pellets ground to 0-300 µm.

Compounds identified at levels between 11 ppb and 130 ppb.

0

20

40

60

80

100

120

140

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80

Time (h)

Am

ount

Ext

ract

ed (p

pb)

Naphthalene Benzophenone LimoneneDodecanoic acid Methyl salicylate Biphenyl

However, to confirm these conclusions, triplicate Soxhlet extractions were carried out at

3 h and compared with duplicates at 15 h (for another batch of the smallest ground

particles). To further validate this, 1 h sonication experiments were also performed in

triplicate. Table 5.1 shows the results. In general there is excellent agreement between

168

Chapter 5

the 15 h Soxhlet extractions and sonication results, indicating that complete extraction

has occurred under these conditions. And for 27 of the 30 compounds, student t-tests

comparing the amounts of contaminants extracted at 3 h and 15 h yielded tcalc values

lower than the tref value of 2.78 at a 5% confidence level. These results confirmed that a

3 h Soxhlet extraction was adequate to extract 90% of the investigated contaminants

from the finely ground (0-300 µm) annealed pellets.

For the remaining three compounds - benzophenone, 1-ethylnaphthalene and

trimethylnaphthalene isomer 2 - the tcalc values were 3.45, 5.20 and 3.29 respectively.

Therefore the tref value of 2.78 was exceeded, implying that extraction was unachievable

in 3 h for these compounds. However, a closer examination of the extraction kinetics

(Figures 5.1 - 5.3) does not reveal any significant difference between these three

compounds and the rest. We conclude that 3 h is sufficient for the full extraction of all

contaminants from the small particles.

Comparatively, extraction by sonication proved to be more time efficient relative to

Soxhlet extraction, reducing the extraction time to 1 h (Table 5.1). A possible

explanation for this result is attributed to the compact packing of the polymer powder in

the Soxhlet extraction thimble compared to particulate dispersion during sonication.

When polymer particles are isolated from each other in DCM there is greater solvent

accessibility into the polymer matrix and hence migration is facilitated. This

accessibility is hindered when the particles are crammed together as there is less surface

area exposed for the solvent to penetrate.

It is also possible that the agitating motion of the sonication process moves solvent into

the polymer matrix and therefore liberates contaminants out of it more readily than by

Soxhlet extraction.

Contrary to the Soxhlet extraction of fine-ground annealed pellets (particle size = 0-300

µm) that took 3 h to extract, Soxhlet extraction of fine-ground flake was not completed

until 24 h (Chapter 4, Section 4.3.2). This may have resulted from the ground flake

being slightly more amorphous (percent crystallinity = 20.6-22.8%) than the powdered

annealed pellets (percent crystallinity = 24.4%) thus undergoing coalescence during

swelling.♣

♣ Another possible explanation for the faster extraction of the fine-ground annealed pellets relative to the fine-ground flake is that depending on the shape and size of the particles, the effective contact time with DCM is different due to the different effective surface of the particles.

169

Chapter 5

Table 5.1: Amounts of contaminants extracted from annealed pellets ground to 0-300 µm

by Soxhlet extraction and sonication (standard deviation, n=3 for 3 h; n=2 for 15 h).

1,2,4- Trimethyl-benzene

3-Ethyl-o-xylene

1,2,3,5- Tetramethyl-benzene

(-)-Menthone

Methyl salicylate

4-n-propyl anisole

3 h Soxhlet 6.2 (0.3) 1.2 (0.1) 0.9 (0.0) 2.4 (0.2) 14 (2) 6.6 (0.3) 15 h Soxhlet 6.3 (0.1) 1.2 (0.1) 0.9 (0.1) 2.6 (0.1) 15 (1) 6.9 (0.1) 1 h Sonication 6.9 (0.9) 1.3 (0.2) 0.9 (0.0) 2.6 (0.1) 15 (0) 6.8 (0.1)

m-Cymene

Limonene Cineole γ-Terpinene Naphthalene Biphenyl

3 h Soxhlet 10.5 (0.3) 131 (22) 6.9 (0.3) 6.6 (1.0) 50 (1) 16 (1) 15 h Soxhlet 9.9 (0.9) 117 (4) 6.8 (0.6) 5.3 (0.2) 50 (1) 17 (0) 1 h Sonication 11 (1) 122 (0) 7.3 (0.3) 5.4 (0.3) 50 (0) 15 (1)

2-Methyl naphthalene

1-Methyl naph.

1-Ethyl naph.

2,6- Dimethyl naph.

1,7- & 1,6- Dimethyl naph.

1,4- Dimethyl naph.

3 h Soxhlet 9.5 (0.3) 4.6 (0.3) 2.7 (0.1) 6.8 (0.3) 19 (1) 1.9 (0.2) 15 h Soxhlet 11 (1) 5.2 (0.3) 3.0 (0.0) 7.8 (0.0) 21 (0) 1.9 (0.1) 1 h Sonication 9.9 (0.8) 5.1 (0.6) 2.5 (0.1) 6.9 (0.5) 20 (1) 1.9 (0.2)

1,2- Dimethyl naph.

Trimethyl naphthalene Isomer 1

Trimethyl naph. Isomer 2

Trimethyl naph. Isomer 3

Trimethyl naph. Isomer 4

Trimethyl naph. Isomer 5

3 h Soxhlet 1.6 (0.0) 1.2 (0.1) 1.3 (0.1) 1.4 (0.1) 1.3 (0.1) 1.2 (0.1) 15 h Soxhlet 1.7 (0.1) 1.2 (0.1) 1.6 (0.1) 1.7 (0.2) 1.5 (0.1) 1.5 (1.4) 1 h Sonication 1.7 (0.1) 1.2 (0.1) 1.5 (0.1) 1.5 (0.2) 1.5 (0.1) 1.4 (0.0)

Trimethyl naph. Isomer 6

Benzo- phenone

Tetradecane Lauric acid⊕

n- Hexyl benzoate

Dodecane

3 h Soxhlet 1.9 (0.1) 80 (2) 15 (6) 44 (21) 5.3 (1.4) 16 (5) 15 h Soxhlet 2.2 (0.3) 92 (5) 9.2 (2.7) 42 (25) 5.0 (0.5) 15 (4) 1 h Sonication 2.2 (0.1) 91 (4) 15 (5) 53(7) 4.0 (0.5) 18 (5)

⊕ Lauric acid is also known as “dodecanoic acid”.

170

Chapter 5

This fusion of the fine-ground flake was observed when attempting to empty the

extraction thimble following Soxhlet extraction.

The tight packing of the fine powder in the Soxhlet thimble enhances this fusing effect,

which has been observed by others in the presence of high temperature during

accelerated solvent extraction (Eskilsson and Björklund 2000, Vandenburg et al. 1997,

Lou et al. 1997).

5.2.1.1 Pellets ground to 0-300 µm: The relationship between extraction kinetics and

contaminant molecular weight.

A plot of the amounts of contaminants extracted (from annealed pellets ground to 0-300

µm) at 1h divided by the amounts extracted at equilibrium (an average of five values at

3, 8, 24, 48 and 72 h) versus the contaminant molecular weight is presented in Figure

5.4. The extraction levels for the dimethyl naphthalene isomers and trimethyl

naphthalene isomers have been averaged. At a first glance, there does not appear to be a

correlation between both parameters. However, when the three points signifying >80%

extraction are excluded, we obtain a weak correlation coefficient (R2) of 0.39 for the

remaining points.

It is presumed that a better R2 would be attained in the absence of solvent penetration

(polymer swelling), as migration into non-swelling solvents is predominantly dependent

on the diffusion coefficient, which is inversely proportional to the size of the molecule

(equation 5.1).

Q = -D(dc/dx) (5.1)

In equation 5.1, Q is the amount of contaminant exiting the PET during a time unit; D is

the diffusion coefficient of the contaminant in the plastic; and -dc/dx is the

concentration gradient along the depth of PET (Feigenbaum et al. 1993).

171

Chapter 5

Figure 5.4: Percentage of contaminant extracted at 1 h versus molecular weight.

30

40

50

60

70

80

90

100

110

120

80 100 120 140 160 180 200 220

Contaminant MW

100(

Am

ount

at 1

h/A

mou

nt a

t equ

ilibr

ium

)

At the other extreme, when the solvent totally swells the packaging material, the

concentration of the migrants in the solvent is governed by the partition coefficient (γ)

between the solvent and plastic (equation 5.2) because diffusion through the polymer

matrix is no longer the rate-determining step. Contaminants are expected to diffuse out

of the polymer at the same rate, irrespective of their size. This is because the size of the

transient voids in-between the PET polymer chains increase with the penetration of

DCM and result in less discrimination towards the movement of analytes through the

matrix. In fact the efficiency of extraction is determined by the equilibrium distribution

into the solvent. The maximum possible concentration of migrant in the solvent, C*f is

given by:

172

Chapter 5

C*f = CoPγVf/(VP+γVf ) (5.2)

where VP and Vf are the volumes occupied by the plastic and the solvent respectively, γ

is the partition coefficient of the migrant between the plastic and solvent and CoP is the

initial concentration of the migrant in the polymer (Feigenbaum et al. 1993).

In our study partial swelling was expected at the one-hour extraction time interval

therefore only a weak correlation between percentage extracted and molecular weight

was observed. This weak correlation presumably indicates that migration of

contaminants from PET into DCM is a process intermediate between the purely kinetic,

diffusion-controlled case and the thermodynamic, partitioning alternative. The other

factor contributing to the extraction rate (as well as the partitioning) is the analyte

polarity. After the exclusion of one outlier (for dodecanoic acid), a plot equivalent to

that of Figure 5.4 but with contaminant polarity (estimated solvent strength parameter)

on the abscissa, led to a weaker correlation coefficient (R2) of 0.31. Therefore,

contaminant polarity did not greatly contribute to the fraction extracted.

The three isolated points in Figure 5.4, signifying >80% extraction, happen to be for

straight chain organic compounds. It was postulated that the efficient extraction of

these linear compounds relative to the cyclic compounds resulted from the ability of

elongated molecules to diffuse faster in polymers. This result underlines the potential

significance of molecular shape on extraction efficiency. Reynier et al. (2001)

observed that size is not the only factor contributing towards the diffusion of

compounds in polymers. Other parameters that correlate with the diffusion coefficient

include the compound’s minimum cross section, shape, interaction with the polymeric

matrix and its flexibility. In terms of molecular shape, linear molecules - which crawl

through the polymer matrix - diffuse faster than spherical molecules. The latter have

been reported to move by slower sequential jumps (Reynier et al. 2001).

A less pronounced tendency towards the extraction of linear contaminants was observed

in our previous studies on flake (Chapter 4, Section 4.3.2). Therefore the possibility of

the isolated points in Figure 5.4 being outliers was not ruled out, keeping in mind that

the values at 1h represent single measurements.

173

Chapter 5

5.2.2 Annealed pellets ground to >300-425µm

The kinetics of extraction from particles ground to this size were not determined.

Rather, after the extensive investigations of the kinetics of extraction from small

particles (0-300 µm, previous section), the kinetics of extraction from the large particles

(>425-700 µm, next section) were completed and 3 h was again found to be sufficient

for complete extraction. Hence it was clearly indicated that only 3 h would be

necessary to quantitatively extract the medium sized particles.

Figure 5.5 is a log-log plot of the amounts of contaminants extracted from >300-425 µm

particles versus the amounts extracted from 0-300 µm particles, yielding a gradient of

0.994. As can be seen from Figure 5.5, there is significant scatter even with R2 = 0.985,

particularly when the log-log nature of the plot is taken into account. Nonetheless, the

gradient implies that the variables on average are in reasonable agreement with each

other, and generally demonstrates that further particle size reduction (from >300-425

µm to 0-300 µm) did not systematically influence the level of contaminants.

174

Chapter 5

Figure 5.5: A log-log plot of the amounts of contaminants extracted from >300-425 µm

particles versus the amounts extracted from 0-300 µm particles.

y = 0.9937xR2 = 0.9849

-0.5

0

0.5

1

1.5

2

2.5

3

-0.5 0 0.5 1 1.5 2 2.5 3

Log

[am

ount

in >

300-

425

um (p

pb)]

Log[amount in 0-300 um (ppb)]

x=y

175

Chapter 5

5.2.3 Annealed pellets ground to >425-700 µm

Kinetic studies for a different batch of annealed pellets ground to >425-700 µm were

carried out. Figure 5.6 is a log-log plot of the amounts of contaminants extracted at 3 h

versus the amounts extracted at 24 h, with the exception of one outlier; dodecanoic acid

again. Although this outlier was excluded from this plot, student t-tests on the raw data

established that there was no significant difference between the amount extracted at 3 h

and 24 h at the 5% confidence level (tcal = 0.89; tref = 2.57).

Log-log plots are used throughout this thesis to even out the distribution of data points

and allow equal inspection of most of the results, which occur at low concentrations.

As shown on Figure 5.6, the theoretical line of perfect agreement between both

extraction times (y=x) is superimposed on the experimental line, demonstrating that on

average there is an insignificant difference between both extraction times. Thus

contaminant extraction was generally completed in 3 h. The correlation coefficient (R2)

of 0.99 suggests minimum scatter. There is only one point (for n-hexylbenzoate with

coordinates 0.73, 0.44) that is clearly above the line of best fit. However, once again,

student t-tests on the raw data established that there was no significant difference in the

amount of this contaminant extracted at 3 h and 24 h at the 5% confidence level (tcal =

1.96; tref = 2.57) due to the great spread in the 24 h data for n-hexyl benzoate (2.8 ± 2.2

ppb).

176

Chapter 5

Figure 5.6: A log-log plot of the amounts of contaminants extracted at 24 h versus the

amounts extracted at 3 h for particles >425-700 µm.

y = xR2 = 0.989

-0.5

0

0.5

1

1.5

2

2.5

-0.5 0 0.5 1 1.5 2 2.5

log (amount extracted at 24 h)

log

(am

ount

ext

ract

ed a

t 3 h

)y=x

177

Chapter 5

5.2.4 Unground annealed pellets

Figures 5.7 - 5.9 are plots demonstrating the extraction kinetics of contaminants from

unground annealed pellets. Some compounds, which show similar trends, were

excluded from these plots for the sake of clarity. The curves are modeled by a three

parameter exponential function1, generally giving correlation coefficients (R2) between

0.81-0.99. As indicated by the dashed lines in Figure 5.7, complete extraction (±

random experimental error) was achieved in 7.6 h. One exception was benzophenone

(Figure 5.9), where the amount extracted appears to still be increasing at 24 h. This

conclusion is heavily dependent upon the last data point collected – and of course, may

be due to sampling variation since fresh pellets are necessarily utilized for the extraction

done for each time interval. On the other hand, benzophenone may not have been fully

extracted from the large-size particles (>425-700 µm). Hence, there is doubt about the

complete extraction of benzophenone in 7.6 h, but we did not investigate this further.

The levels of the elongated compounds (dodecane, tetradecane and dodecanoic acid) did

not exceed the levels identified in the blank. Therefore their extraction kinetics were

not studied in the pellets.

The less favorable kinetics for the annealed pellets relative to their ground counterparts

is attributed to the larger path length for the pellet. DCM takes longer to diffuse into the

matrix and contaminants are slower to diffuse out, thus there is a decrease in overall rate

of mass transfer.

1 Sigmaplot software package was used to generate the curves using the function y = y0 + a(1-bx).

178

Chapter 5

Figure 5.7: Soxhlet extraction kinetic study of unground annealed pellets. Compounds

identified at levels below 2 ppb.

0

0.5

1

1.5

2

0 2 4 6 8 10 12 14 16 18 20 22 24Time (h)

Am

ount

Ext

ract

ed (p

pb)

1,2,3,4-Trimethylbenzene (-)-Menthone

n-Propylanisole 2-Methylnaphthalene

1-Ethylnaphthalene

179

Chapter 5

Figure 5.8: Soxhlet extraction kinetic study of unground annealed pellets. Compounds

identified at levels below 13 ppb.

0

2

4

6

8

10

12

14

0 2 4 6 8 10 12 14 16 18 20 22 24

Time (h)

Am

ount

Ext

ract

ed (p

pb)

m-Cymene Cineole

3-Ethyl-o-xylene Biphenyl

gamma-Terpinene Trimethylnaphthalene Isomer 1

1,2,4-Trimethylnapthalene

180

Chapter 5

Figure 5.9: Soxhlet extraction kinetic study of unground annealed pellets. Compounds

identified at levels below 70 ppb.

0

10

20

30

40

50

60

70

80

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28

Time (h)

Am

ount

Ext

ract

ed (p

pb)

Limonene Naphthalene Benzophenone

181

Chapter 5

5.2.5 The effect of particle size reduction upon measured contaminant levels in

extruded and annealed pellets

Figures 5.10 (a and b) and 5.11 are plots of the contaminant levels measured in each

particle size range (small, 0-300 µm; medium, >300-425 µm; large, >425-700 µm)

derived from the same batch of extruded pellets after annealing and grinding. Figure

5.10 (a) and (b) contains typical plots representative of the majority of analytes. It can

be seen that there is no regular pattern to the variations from one particle size to the

next, and most importantly, no discernable trend. This result is different to that

observed for washed and dried PET flake (Chapter 4), where the contaminant levels

dropped by anywhere between two and six times from the small to medium size

particles; small decreases were then systematically observed from the medium to large

particles. The rationale behind this observation was explained in terms of the grinding

process selectively grinding the more contaminated surface of flake to the smallest

particle size. The larger particle size consists of the inner core of the flake, which is

theoretically less contaminated.

Figure 5.10 (a) and (b) indicates that melting during the extrusion process gives rise to a

more-or-less uniform distribution of the contaminants throughout the pellets and the

degree of grinding or any bias towards the surface is irrelevant. Figure 5.11 shows the

extreme examples of variability of contaminant levels between particle sizes. The

worse case is for tetradecane, but as Figure 5.12 shows, the standard deviations are so

large for the small and medium particle size that no particular meaning can be attached

to these results and the general conclusion drawn from Figure 5.10 – that there is no

systematic dependence upon particle size – remains valid.

In order to further demonstrate that, on average, particle size reduction (from >425-700

µm to 0-300 µm) did not systematically influence the level of contaminants, a log-log

plot of the amounts of contaminants extracted from >425-700 µm particles versus the

amounts extracted from 0-300 µm particles was plotted (Figure 5.13). Given the

significant scatter (R2 = 0.953) the gradient of 0.952 in Figure 5.13 can be taken to

imply that the variables, on average, are in reasonable agreement with each other, and

generally support the theory that particle reduction did not influence the level of

contaminants.

182

Chapter 5

Figure 5.10: Typical variations in contaminant levels measured from the same batch of

annealed pellets ground to the three particle sizes.

(a)

2

3

4

5

6

7

8

9

10

11

12

Medium

Am

ount

Ext

ract

ed (p

pb)

1,2,4-Trimethylbenzene (-)-Menthone4-n-Propylanisole m-CymeneCineole gamma-Terpinene2-Methylnaphthalene 1-Methylnapththalene1-Ethylnaphthalene 2,6-Dimethylnaphthalenen-Hexyl benzoate

Small Large

183

Chapter 5

(b)

0

0.5

1

1.5

2

2.5

Medium

Am

ount

Ext

ract

ed (p

pb)

3-Ethyl-o-xylene1,2,3,5-Tetramethylbenzene1,4-Dimethylnaphthalene1,2-Dimethylnaphthalene

LargeSmall

184

Chapter 5

Figure 5.11: Extreme examples of the variation in contaminant levels.

5

7

9

11

13

15

17

19

21

23

Am

ount

Ext

ract

ed (p

pb)

Methyl salicylateBiphenyl1,7- & 1,6- DimethylnaphthaleneTetradecaneDodecane

Small LargeMedium

185

Chapter 5

Figure 5.12: An example of the experimental spread (means ± standard deviation) for

divergent measurements of a contaminant in the three particle sizes derived from the

same batch of annealed pellets.

0

5

10

15

20

25

30

Am

ount

Ext

ract

ed (p

pb)

Tetradecane

Small Medium Large

186

Chapter 5

Figure 5.13: Log-log plot of the amounts of contaminants extracted from >425-700 µm

particles versus the amounts extracted from 0-300 µm particles.

y = 0.9523xR2 = 0.9533

-0.5

0

0.5

1

1.5

2

2.5

3

-0.5 0 0.5 1 1.5 2 2.5 3

Log[amount in 0-300 um (ppb)]

Log

[am

ount

in >

425-

700

um (p

pb)]

x=y

187

Chapter 5

The three compounds that are significantly below the line of best fit in this plot are

dodecanoic acid, dodecane and tetradecane. Student t-tests were performed comparing

the amounts of these contaminants extracted from the 0-300 µm and >425-700 µm

particles and the calculated t values at the 5% confidence level were lower than the

reference t value of 2.45 for all the compounds except dodecane (tcal = 2.97). Therefore

the difference in the amounts extracted from both particle sizes was significant only for

this compound.

An analogous plot to that of Figure 5.13, except correlating the amounts of

contaminants extracted from the 0-300 µm and >300-425µm particles was shown in

Figure 5.5 and a similar conclusion was reached (see Section 5.2.2).

188

Chapter 5

5.3 ANNEALED VERSUS AMORPHOUS EXTRUDED PELLETS

5.3.1 Kinetics of extraction from amorphous pellets

Particle size is not the only factor influencing migration rate. It is an established fact

that percent crystallinity plays an important role in polymer extractions and in migration

and sorption studies (Nir and Ram 1996, Nielsen, 1994). In order to confirm this

theory, extraction kinetics were carried out on amorphous PET pellets. Figure 5.14 is a

log-log plot of the amounts of contaminants extracted at 3 h versus the amounts

extracted at 24 h for amorphous pellets. A gradient of 0.985 suggests both variables are

in excellent agreement with each other. The correlation coefficient (R2=0.993) of

Figure 5.14 is very good; the only point significantly below the line of best fit,

signifying incomplete extraction at 3 h was for n-hexylbenzoate. However, due to the

large size of the experimental error, t-tests for this compound verified that there was no

significant difference in the amounts extracted during both times at the 5% confidence

level (tcal = 2.39; tref = 2.57).

Therefore, extraction was generally completed in 3 h for the amorphous pellets,

compared to 7.6 h for the annealed pellets♣. Further experiments concluded that

extraction time could not be reduced below 3 h for the amorphous pellets (Figures 5.15-

5.18).

It was postulated that variations in polymer crystallinity cause differences in polymer

swelling and thus influence the extraction kinetics. As polymer crystallinity increases

the shapes and sizes of microcavities available for solvent movement decreases,

affecting the extraction kinetics (Limm and Hollifield, 1996).

♣ Even benzophenone was extracted in 3 h from the amorphous pellets, as for all of the other compounds (Figure 5.14, data points 2.06, 2.03, arrowed). For the annealed (and more crystallized) pellets, there was some doubt from the raw data (Figure 5.9) as to whether benzophenone was completely extracted even after 24 h. The above result from the amorphous less crystallized pellets again indicates that the 24 h data point for benzonphenone in Figure 5.9 is erroneous.

189

Chapter 5

Figure 5.14: A log-log plot of the amounts extracted at 3 h versus the amounts extracted

at 24 h for unground amorphous pellets.

y = 0.9845xR2 = 0.9929

-0.5

0

0.5

1

1.5

2

2.5

3

-0.5 0 0.5 1 1.5 2 2.5 3

log (amount extracted at 24 h)

log

(am

ount

s ex

tract

ed a

t 3 h

)

y=x

Benzophenone

190

Chapter 5

Figure 5.15: Soxhlet extraction kinetic study of unground amorphous pellets.

Compounds identified at levels below 4 ppb.

0

0.5

1

1.5

2

2.5

3

3.5

4

0 1 2 3 4 5 6 7 8

Time (mins)

Am

ount

Ext

ract

ed (p

pb)

(-)-Menthone 2-Methylnaphthalene1-Methylnaphthalene 4-n-Propylanisolegamma-Terpinene 2,6-Dimethylnaphthalene3-Ethyl-o-xylene 1,2,3,5-Tetramethylbenzene1,4-Dimethylnaphthalene 1,2-Dimethylnaphthalene

191

Chapter 5

Figure 5.16: Soxhlet extraction kinetic study of unground amorphous pellets.

Compounds identified at levels below 11 ppb.

0

2

4

6

8

10

12

0 1 2 3 4 5 6 7 8

Time (mins)

Am

ount

Ext

ract

ed (p

pb)

Cineole Dodecane 1,7-Dimethylnaphthalene Biphenyl

192

Chapter 5

Figure 5.17: Soxhlet extraction kinetic study of unground amorphous pellets.

Compounds identified at levels below 13 ppb.

0

5

10

15

20

25

30

35

40

45

50

0 1 2 3 4 5 6 7 8

Time (mins)

Am

ount

Ext

ract

ed (p

pb)

Benzophenone Naphthalene1,2,4-Trimethylbenzene Tetradecane

193

Chapter 5

Figure 5.18: Soxhlet extraction kinetic study of unground amorphous pellets.

Trimethylnaphthalene compounds.

0

0.2

0.4

0.6

0.8

1

1.2

0 1 2 3 4 5 6 7

Time (mins)

Am

ount

Ext

ract

ed (p

pb)

8

Trimethyl Naphthalene Isomer 1 Trimethyl Naphthalene Isomer 2

Trimethyl Naphthalene Isomer 3 Trimethyl Naphthalene Isomer 4

Trimethyl Naphthalene Isomer 5 Trimethyl Naphthalene Isomer 6

194

Chapter 5

5.3.2 Variation of DCM uptake with PET crystalline structure

It has been postulated that variation in polymer crystallinity causes differences in

polymer swelling and thus influence the extraction kinetics. As polymer crystallinity

increases the shapes and sizes of microcavities available for solvent movement

decreases, slowing the extraction kinetics (Limm and Hollifield 1996). Thus it would

be expected that DCM is more aggressive towards the more amorphous PET and

weight-uptake experiments were carried out on amorphous and crystalline PET.

The DCM sorption kinetics for the amorphous and annealed pellets are shown in Figure

5.19 and demonstrate that maximum sorption is achieved at significantly different

times. For the amorphous pellets maximum sorption was achieved in 2 h and maximum

extraction in 3 h, compared to approximately 15 h (maximum sorption) and 7.6 h

(maximum extraction) for the more crystallized (annealed) pellets. The maximum

sorbed amount was also greater for amorphous pellets (≈ 4 g) compared to crystallised

pellets (≈ 3 g). It was observed that the amorphous pellets turned white during the

course of sorption due to solvent-induced crystallisation.

The kinetics of sorption is described by Equation 5.3 where n = 0.5 for Fickian

diffusion, n < 0.5 for pseudo-Fickian diffusion and n = 1 for non-Fickian diffusion. In

order to compare diffusivities an approximation of the diffusion coefficient (D) was

made assuming Fickian behaviour despite Liu and Neogi (1992) observing pseudo-

Fickian behaviour for semicrystalline PET due to the elastic stresses of swelling.

At/Ae ≅ 4(Dt/πL2)n (5.3)

In Equation 5.3 At is the amount sorbed at time t, Ae is the amount sorbed at

equilibrium, D is the diffusion coefficient and L is the thickness of the pellet.

195

Chapter 5

Figure 5.19: Sorption kinetics of DCM into amorphous and annealed pellets.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 2 4 6 8 10 12 14 16 18 20 22 24

Time (h)

Am

ount

sor

bed

(g)

Crystalline pellets Amorphous pellets

196

Chapter 5

Figure 5.20 is a plot of At/Ae (amount sorbed at time t/amount sorbed at equilibrium)

versus the square root of time for amorphous and crystalline pellets. From the slopes of

the steep portions of these curves, with the aid of equation 5.3 and n= 0.5, setting

L=0.35 cm the D for the sorption of DCM into amorphous pellets (D = 7.75 x 10-3

cm2/h) was calculated to be about 4 times greater than that for the annealed pellets (D =

1.87 x 10-3 cm2/h). This result agrees with the longer extraction times required for the

extraction of contaminants out of the annealed pellets.

Feigenbaum et al. (2002) performed an analogous study for polyolefins. The D values

for DCM sorption ranged between 4.68 x 10-4 cm2/h and 1.33 x 10-3 cm2/h. Our D

values for PET exceeded these values, possibly due to the better solubility compatibility

between DCM and PET. Vandenburg et al. (1999) reported that the solubility parameter

(δ) difference between PET and DCM (∆δ = 0.7 Mpa1/2) is significantly smaller than

between polypropylene and DCM (∆δ = 3.2 Mpa1/2).

Similarly, Nir and Ram (1996) studied the sorption kinetics of toluene, benzyl alcohol,

heptane and propylene glycol (as solvents) in amorphous and biaxially oriented PET

film. Our estimated D values for the sorption of DCM into PET were higher than those

quoted for the mentioned solvents by the latter authors. The D values determined by Nir

and Ram (1996) ranged between 1.08 x 10-6 cm2/h (for benzyl alcohol sorption in

biaxially oriented PET) and 5.40 x 10-6 cm2/h (for benzyl alcohol sorption in amorphous

PET). No detectable sorption was observed for heptane and 1,2-propanediol. The

larger solubility parameter differences between the PET and the non-chlorinated

solvents could be responsible for the smaller diffusion coefficients obtained by Nir and

Ram (1996).

197

Chapter 5

Figure 5.20: A plot of the (amount of DCM sorbed /amount sorbed at equilibrium)

versus the square root of time.

198

0

0.2

0.4

0.6

0.8

1

1.2

0 1 2 3 4 5

Square root time (h)

At/A

e

6

Crystalline pellets Amorphous pellets

Chapter 5

5.3.3 Contaminant diffusion coefficients out of amorphous and annealed PET

The results in Figure 5.7-5.9 (Soxhlet extraction kinetic study of annealed pellets) were

re-plotted as At/Ae (amount extracted/amount extracted at equilibrium) versus the square

root of time. Figure 5.21 is an example of a representative plot for naphthalene.

Figure 5.22 is a plot of the calculated diffusion coefficients for each contaminant

(determined from the slopes in curves analogous to that of Figure 5.21) versus the

reciprocal of molecular weight in the annealed pellets. A linear line was fitted with a

weak correlation (R2) of 0.61 after the exclusion of 2 outliers and the points for the

linear molecules, which were extracted before 2 hours (dodecane, tetradecane and

dodecanoic acid).

The average D was calculated to be 2.51 x 10-3 cm2/h for the contaminants included in

Figure 5.22.

For the amorphous (un-annealed) pellets the average calculated D was 8.00 x 10-3

cm2/h. Figure 5.23 is a plot of the calculated diffusion coefficients for each contaminant

(determined from the slopes in curves analogous to that of Figure 5.21) versus the

reciprocal of molecular weight in the amorphous pellets. A linear line (with a positive

slope) was fitted with a weak correlation (R2) of 0.50.

The D values for amorphous (un-annealed) and annealed pellets indicate that diffusion

is more rapid in amorphous pellets than in the more crystalline annealed pellets. The

relationship Dannealed = Damorph.(1-Xcryst.)℘ (see Appendix 3 for derivation) can be used to

calculate the “effective” diffusion coefficient for crystalline pellets (DC) assuming

diffusion occurs only in amorphous regions of the annealed pellets. When substituting

the diffusion coefficient for amorphous (un-annealed) pellets (Damorph. = 8.00 x 10-3

cm2/h) and the fraction crystallinity for the annealed pellets (Xannealed = 0.24) into this

equation, the predicted (“effective”) Dannealed for annealed pellets equals 6.08 x 10-3

cm2/h. In summary, assuming the un-annealed pellets are 100% amorphous (Schumann

and Thiele 1996) and have a diffusion coefficient of 8.00 x 10-3 cm2/h, then a 24%

crystalline pellet would give an “effective” diffusion coefficient of 6.08 x 10-3 cm2/h,

assuming diffusion only takes place in the amorphous regions of the pellet.

℘In this equation (1-X) signifies the fraction of amomphous region in the pellet. Damorph. and Dannealed signify the diffusion coefficients in the amorphous and annealed pellets respectively.

199

Chapter 5

The experimental Dannealed was 2.51 x 10-3 cm2/h, which is 2.4 times smaller than the

predicted (“effective”) value. The discrepancy between the experimental and predicted

value may have resulted from the DCM inducing greater crystallinity in annealed pellets

than in amorphous pellets (Nir and Ram, 1996).

Figure 5.21: A plot of At/Ae (amount extracted/amount extracted at equilibrium from

annealed pellets) versus the square root of time (a representative plot; naphthalene).

y = 0.3837x - 0.0707R2 = 0.9788

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5

square root time (h)

At/A

e

200

Chapter 5

Figure 5.22: A plot of calculated diffusion coefficients versus the reciprocal of

molecular weights (for annealed pellets).

y = 1.0841x - 0.005R2 = 0.6135

0

0.0005

0.001

0.0015

0.002

0.0025

0.003

0.0035

0.004

0.0045

0.005 0.006 0.007 0.008 0.009

1/MW

Diff

usio

n co

effic

ient

(cm

2/h)

201

Chapter 5

Figure 5.23: A plot of calculated diffusion coefficients versus the reciprocal of

molecular weights (for amorphous pellets).

y = 3.4641x - 0.015R2 = 0.4698

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.005 0.0055 0.006 0.0065 0.007 0.0075 0.008 0.0085

1/MW

Diff

usio

n co

effic

ient

(cm

2/h)

5.3.4 Contaminant loss during the annealing of pellets

In earlier work with surrogates specified by the US FDA challenge test (Harding et al.

unpublished), the effect of moulding the PET into bottles (after the vacuum extrusion at

the conclusion of the recycling process) indicated a diminution in the levels of semi-

202

Chapter 5

volatile contaminants. There were two alternative vacuum extrusion processes utilized

and the levels of benzophenone decreased from 33 to 21 ppm in one case and from 36 to

19.7 ppm in the other. Methyl stearate was the second semi-volatile surrogate in our

study, but its level was below the limit of quantitation (of 5 ppm) for the GC-FID

method used before moulding. Thus on the basis of the benzophenone results, the

moulding process – and from the analytical perspective in our current studies – the

annealing carried out to impart hardness to the extruded pellets to allow grinding before

sampling and analysis, decreases in contaminant levels would be expected.

Figure 5.24 is a log-log plot of the amount of contaminant extracted from amorphous

pellets versus the amount extracted from ground-annealed pellets. The gradient of

0.9827 for the regression is so close to the ideal value of 1.000 (x=y), and with the

correlation between the two sets of data so strong (R2 = 0.9928), the plot demonstrates

that the levels of contaminants measured are in excellent agreement with one another.

Nonetheless, since it is a log-log plot, small deviations from the x=y line are more

significant than is apparent. In the worst case (the data point 0.591, 0.730), there is a

38% difference. However, for this point and for the other deviating points that lie

below the x=y line, the annealed pellets have higher contaminant levels (than those

pellets not subjected to heat treatment; the amorphous pellets). This can only be

explained by significant variations in some initial contaminant levels in PET from

which the sub-batches were drawn. However, the majority of contaminants have very

similar levels in the amorphous and the annealed pellets. Therefore our conclusions are

that semi-volatile contaminant loss is insignificant during annealing of extruded pellets

at 150°C and this post-extrusion step is not an important purification step.

203

Chapter 5

Figure 5.24: A log-log plot of the amounts of contaminants extracted from amorphous

pellets versus the amounts extracted from ground annealed pellets.

y = 0.9827xR2 = 0.9928

-0.5

0

0.5

1

1.5

2

2.5

3

-0.5 0 0.5 1 1.5 2 2.5 3

log[amount extracted from ground (annealed) pellets (ppb)]

log[

amou

nt e

xtra

cted

from

am

orph

ous

pelle

ts (p

pb)]

x=y

5.3 FLATTENING AMORPHOUS PET PELLETS

Apart from grinding, flattening could also reduce the path-length of a PET pellet prior

to extraction. Since amorphous pellets are not brittle enough to grind unless heat-

treated (i.e., annealed), an alternative means of reducing pellet path-length was sought

involving flattening the pellets using a hydraulic press (8 tons). The extraction results

for the whole and flattened amorphous pellets are shown in Table 5.2. Flattening

significantly effected the extraction of some compounds, namely, 3-ethyl-o-xylene,

1,2,3,5-tetramethyl benzene, naphthalene, biphenyl, 2-methyl naphthalene, 1-methyl

204

Chapter 5

naphthalene, 1-ethyl naphthalene, the dimethyl naphthalene isomers and

trimethylnaphthalene isomers. And the compound that was most dramatically affected

by the flattening process was biphenyl; whose level in flattened pellets averaged 60

times the level in whole pellets. The levels of all other compounds in the flattened

pellets did not exceed five fold the level in non-flattened pellets. Only biphenyl

exceeded the 215 ppb threshold in the flattened pellets.

Interestingly, as mentioned in Section 5.2.5, grinding did not have this effect on the

amount extracted. Therefore it was proposed that extraction was not primarily diffusion

controlled. It may have been that the mechanical action of the flattening process

dispersed the polymer chains away from each other, releasing trapped contaminants.

Not all compounds (e.g. the flavor compounds such as limonene, cineole, γ-terpinene)

were affected by this mechanical action because not every contaminant is trapped

between the polymer chains. It is possible that the compounds, which are trapped

between polymer chains, derive from the original manufacture of the polymer. The

highly variable nature of the results for the flattened pellets may be attributed to

partially irreproducible flattening of the PET pellets.

Table 5.2: Flattened and whole amorphous pellets extracted by sonication and Soxhlet

extraction. (concentrations in ppb.)

m-Cymene

Limonene Cineole 3-Ethyl-o-xylene

1,2,3,5- Tetramethyl benzene

(-)-Menthone

4-n-propyl anisole

3h sonication

5.4 (0.4)

87.0 (5.4)

17.1 (0.5)

2.6 (0.1)

8.8 (0.2)

3.9 (0.4)

3.9 (0.0)

24h Soxhlet (whole pellets)

5.7 (0.1)

100.2 (7.3)

14.9 (0.1)

2.3 (0.1)

10.6 (1.1)

2.8 (0.2)

4.1 (0.1)

Sonication (flat pellets) 1h 3h 15h 28h 48h

4.9 4.8 5.6 5.6 9.1

81.8 84.2 87.1 85.8 82.5

14.8 15.5 16.9 16.5 17.3

7.3 7.8

10.0 10.5 8.2

14.3 18.8 11.0 18.9 18.6

2.8 2.5 2.2 2.5 2.3

4.1 4.4 4.9 4.6 4.5

205

Chapter 5

Naphthalene Biphenyl 2-Methyl Naph.

1-Methyl Naph.

1-Ethyl Naph.

2,6-Dimethyl Naph.

1,7- & 1,6- Dimethyl Naph.

3h sonication

18.7 (2.6)

9.1 (0.8)

12.3 (1.0)

7.4 (0.8)

3.0 (0.3)

9.0 (0.3)

24.4 (0.0)

24h Soxhlet (whole pellets)

21.1 (2.2)

8.6 (0.3)

13.5 (1.4)

7.6 (0.5)

2.7 (0.2)

8.4 (0.1)

23.0

(1.7)

Sonication (flat pellets) 1h 3h 15h 28h 48h

76.9 71.2

112.5 101.7 51.4

494.5 372.8 845.6 676.0 152.7

43.6 41.4 53.7 56.3 45.3

23.7 19.3 29.5 24.8 16.7

7.9 7.7

13.0 11.3 6.0

15.9 17.9 19.6 18.7 15.7

36.9 38.9 38.0 40.4 37.2

1,4- Dimethyl Naph.

1,2- Dimethyl Naph.

Trimethyl Naph. isomer 1

Trimethyl Naph. isomer 2

Trimethyl Naph. isomer 3

Trimethyl Naph. isomer 4

Trimethyl Naph. isomer 5

3h sonication

6.1 (0.4)

4.2 (0.1)

2.1 (0.3)

2.6 (0.0)

2.1 (0.3)

2.5 (0.5)

3.1 (0.2)

24h Soxhlet (whole pellets)

5.2 (0.3)

4.1 (0.3)

1.7 (0.1)

2.3 (0.2)

2.1 (0.2)

2.4 (0.1)

2.7 (0.0)

Sonication (flat pellets) 1h 3h 15h 28h 48h

8.7 9.8 8.5 8.0 7.4

6.1 7.7 6.8 6.8 6.1

4.3 4.0 4.3 5.1 ND

4.2 3.5 4.1 4.6 ND

3.7 3.2 4.1 3.9 ND

4.5 3.2 4.3 4.9 ND

4.7 5.2 5.2 6.8 ND

Trimethyl Naph. isomer 6

Benzophenone γ-Terpinene Dodecane Tetradecane

3h sonication 3.5 (0.3)

28.8 (3.2)

3.5 (0.1)

26.4 (7.4)

29.4 (16.9)

24h Soxhlet (whole pellets)

3.7 (0.1)

35.4 (1.1)

4.5 (2.0)

14.8 (7.2)

8.4 (5.4)

Sonication (flat pellets) 1h 3h 15h 28h 48h

6.9 6.4 6.8 8.1 ND

37.6 36.3 45.2 44.3 40.6

2.4 2.5 2.6 3.0 4.9

26.0 36.8 26.7 52.7 36.3

18.2 17.4 27.3 46.7 34.8

206

Chapter 5

5.4 LEVELS OF CONTAMINANTS IN PELLETS AND THRESHOLD OF

REGULATION

Table 5.3 gives an overview of measured contaminant levels in curbside collected PET.

In general, the levels of contaminants measured in the ground flake (0-300 µm)

exceeded those in unground flake and extruded pellets. However, there were three

exceptions where the levels in the ground flake were not the lowest (limonene, methyl

salicylate and 4-n-propylanisole). This anomaly could be justified by batch-to-batch

variation.

The generally higher concentrations in the ground flake (0-300 µm) relative to the

unground flake were presumed to have arisen from the selective grinding of the surfaces

of the washed and dried flake to the smallest particle size (Chapter 4). These

concentrations represent the worst-case scenario for the washed and dried flake as far as

food contact is concerned. Therefore, for six contaminants where the level exceeds 215

ppb, it would be necessary to undertake migration testing to obtain approval from the

US FDA.

Compared with the unground flake, the levels of contaminants in the extruded pellets

are considerably reduced, with the average reduction around 71% (see Table 5.3). This

reduction in the level of contamination is attributed to the high temperature and vacuum

used in vacuum extrusion.

A final feature of Table 5.3 is that the relative standard deviation for the flake (23%) is

higher than that for the ground flake (7%) and whole pellets (7.3%). It is expected that

grinding the polymer would cause homogenisation due to the larger number of particles

allowing more reproducible determinations. Despite this better reproducibility, the

levels in the ground flake were less representative of the overall composition whilst

only representing the surface of the flake.

In contrast, melting the flake causes homogeneity within the polymer matrix and the

smaller variation in contamination from pellet-to-pellet allows more reproducible

determinations. This homogeneity within the polymer matrix is consistent with our

general findings of similar levels across particle size ranges after grinding the annealed

pellets and tends to validate our interpretation of the high levels of contaminants found

in finely ground flake being due to selective surface grinding where high levels are

expected.

207

Chapter 5

208

Chapter 5

Table 5.3: Levels of contaminants in ground flake (0-300 µm), unground flake and extruded pellets. (Table continued on next page)

0-300µm ground flake Unground flake Pellets % CONTAMINANT ppb σn-1 % RSD ppb σn-1 % RSD ppb σn-1 % RSD Reduction

Dodecanoic acid 1196 91 8 12 0.9 8 n/d Limonene 898 69 8 1139 274 24 100 7 7 91

Benzophenone 796 28 4 309 24 8 35 1 3 89 Methyl salicylate 839 35 4 897 644 72 15 1.5 10 98 n-Hexyl benzoate 117 5 4 19 15 79 n/d

Cineole 82 5 4 75 75 100 15 0.1 1 80 1,2,4- Trimethyl-benzene 150 12 8 20 17 85 6.3 0.1 2 69

γ-Terpinene 58 5 9 42 10 24 4.5 2 44 89 Biphenyl 44 2 5 18 2 11 8.6 0.3 3 52

4-n-propylanisole 58 4 7 94 27 29 4.1 0.1 2 96 1,2,3,5- Tetramethyl-benzene 19 1 5 14 2 14 11 1 9 21

3-Ethyl-o-xylene 21 2 10 4.5 1.8 40 2.3 0.1 4 49 (-)-Menthone 38 6 16 1.2 0.4 33 2.8 0.2 7 -133

(Average)

n/d = not detected

208

Chapter 5

Table 5.3 (continued): Levels of contaminants in ground flake (0-300 µm), unground flake and extruded pellets.

0-300µm ground flake Unground flake Pellets % CONTAMINANT ppb σn-1 % RSD ppb σn-1 % RSD ppb σn-1 % RSD Reduction

Naphthalene 61 6 10 33 2 6 21 2 10 36 2-Methylnaphthalene 245 9 4 97 5 5 14 1 7 86 1-Methylnaphthalene 109 4 4 47 5 11 7.6 1 7 84 1-Ethylnaphthalene 51 1 2 21 1.5 7 2.7 0.2 7 87

2,6-Dimethylnaphthalene 138 6 4 69 12 17 8.4 0.1 1 88 1,7- & 1,6- Dimethylnaphthalene 272 6 2 136 24 18 23 2 9 83

1,4- Dimethyl naphthalene 72 3 4 21 4 19 5.2 0.3 6 75 1,2- Dimethyl naphthalene 25 3 12 11 1 5 4.1 0.3 7 63

Trimethyl naphthalene isomer 1 23 1 4 6.2 0.8 13 1.7 0.1 6 73 Trimethyl naphthalene isomer 2 25 1 4 5.4 0.6 11 2.3 0.2 9 57 Trimethyl naphthalene isomer 3 33 1 3 7.9 1.4 18 2.1 0.2 10 73

Trimethyl naphthalene isomer 4 & 5 42 1 2 14 3 21 2.4 0.1 4 83 Trimethyl naphthalene isomer 6 28 1 4 8.9 1.5 17 2.7 0.0 0 70 Trimethyl naphthalene isomer 7 12 1 8 4.2 0.8 19 3.7 0.1 3 12

(Average) (5) (23) (7.3) (71)

209

Chapter 5

5.5 CONCLUSIONS

This study has shown that particle size reduction of annealed extruded pellets to less

than 700 µm reduces Soxhlet extraction times from 7.6 h (for unground pellets) to 3 h

(for ground pellets). Therefore, in the case of analysing crystalline (i.e., annealed)

extruded pellets, it is desirable to grind the pellets prior to extraction. However, since

the extraction kinetics (and contaminant levels) for amorphous pellets were not

significantly different from the ground (annealed) pellets, sampling of the whole

amorphous extruded pellet (without the need for annealing and grinding) is indicated

to be a sufficient method for determining contaminant levels in curbside-collected,

recycled PET.

The times taken for quantitative extraction of the contaminants from the amorphous

extruded pellets (3 h) and the annealed pellets (7.6 h) are consistent with the larger

values of the calculated diffusion coefficients of the contaminants (out of amorphous

pellets) and faster DCM uptake into the amorphous pellets than into the annealed (and

therefore crystallized) PET pellets.

In contrast to the results obtained for the flake, the levels of contaminants across the

particle size ranges after grinding the annealed pellets were similar, which tends to

suggest that the extruded pellet matrix is homogenous - in other words, there is a

uniform distribution of contaminants within the pellet and therefore selective grinding

of its surface (as in the case of the more contaminated flake’s surface) is irrelevant.

In terms of recycling PET for food-contact applications, this study has shown that

vacuum extrusion at elevated temperatures decontaminates PET sufficiently to permit

the curbside-collected polymer to be used for food-contact purposes, as the levels of

contaminants in amorphous PET pellets were below the FDA threshold of 215 ppb in

all cases.

As the levels of contaminants were within the obligatory limit, migration tests were

not officially required and therefore not carried out.

210

Chapter 5

Furthermore, previous challenge tests (Harding et al. unpublished) of the same

recycling process suggested that even when the concentration of surrogate

contaminants was at the ppm level, the amount migrating into 10% ethanol was below

10 ppb (Cross et al. unpublished).

210

Chapter 6

CHAPTER 6 VOLATILE CONTAMINANTS AND LEVELS OF OCCURRENCE IN EXTRUDED

PET FLAKE AND PELLETS FROM CURBSIDE COLLECTION

6.1 GENERAL INTRODUCTION

6.1.1 Purpose of the chapter

Chapters 4 and 5 dealt with the solvent extraction of semivolatile contaminants out of

recycled PET plastic intended for soft drink bottle manufacture. The aim of those chapters

was to verify that the levels of semivolatile contaminants in recycled PET were not in

breach of the 215 ppb threshold legislated by the US FDA.

For a more comprehensive investigation on the extent of contamination in post-consumer

PET, it is also important to account for the volatile contaminants present in recycled PET

since their smaller size makes them more inclined to migrate through the polymer matrix

into soft drink.

Hence the purpose of the current chapter is to determine the level of volatile contaminants

present in recycled PET flake and pellets using thermal extraction techniques, which for

reasons specified in Section 6.1.2, are more appropriate than solvent extraction for the

analysis of volatile compounds in polymers.

Due to post-consumer PET undergoing thermal and vacuum treatment during recycling, the

overall level of semivolatiles is expected to exceed the total concentration of volatile

compounds, which are more likely to volatilise out of PET during high temperature

conditions. This hypothesis, however, assumes that the initial concentrations of volatile

contaminants in the pre-recycled post-consumer PET do not significantly exceed the level

of semivolatile contaminants.

6.1.2 Background to thermal extraction

It is well established that some traditional solvent extraction techniques such as Soxhlet,

sonication and total dissolution/reprecipitation have major drawbacks, making heat-

extraction such as headspace analysis appear more attractive. Some disadvantages related

to non-automated solvent extraction are the use of large amounts of solvent, which raises

213

Chapter 6

cost, disposal and environmental concerns; long experimental times resulting from the slow

diffusion process and chemical handling steps; the loss of volatile components during

solvent concentration; and the presence of chromatogram-interfering species in the solvent

which makes identification and quantification less reliable. These issues could be

minimised or eliminated by means of solvent-less headspace techniques, which employ

heat to facilitate extraction.

To date there have been a variety of thermal methods that have been used to extract

components out of plastic. These comprise of thermal desorption (TD), static and dynamic

headspace (SHS and DHS) and solid phase microextraction (SPME).

TD involves heating the PET under a stream of inert gas, which transports the liberated

analytes directly onto a sorbent or GC column. The sorbent is then heated and the retained

volatiles are released and cryofocussed before GC entry. Komolprasert et al. (2001)

extracted irradiated and non-irradiated PET by TD in the quest to determine a difference in

the chemical composition between the treated and untreated plastic. Bayer (2002) used TD

alongside other extraction techniques to determine the composition of five different types

of post-consumer PET feedstreams. TD has also been used to analyse the extent of aroma

sorption by three polymer films (Hernandez-Munoz et al. 2001); the thermal stability of

UV treated films (Fortin and Lu 2001); and the volatiles given off during the extrusion

coating of low-density polyethylene (Villberg and Veijanen 2001).

Dynamic headspace (purge and trap) works in a similar way, however it always involves

trapping onto a sorbent and was originally designed to extract volatiles from aqueous

samples such as wastewater, blood and urine at ambient temperature. Its function is

relatively simple to explain: the sample is bubbled with an inert gas and the vapour is

passed through a sorbent column, where the volatile components are adsorbed. The sorbent

column is then heated and backflushed with the gas to desorb the components onto a gas

chromatographic column.

In order to extract contaminants from solids such as soil, food and plastic, the sample tube

is placed in an oven to facilitate extraction. Gramshaw et al. (1993) extracted potential

migrants from samples of dual-ovenable plastics, including PET, by means of DHS. A

DHS procedure was also employed to determine the levels of residual solvents in food

214

Chapter 6

packaging printed film (Kolb et al. 1981) and benzene residues in recycled PET

(Komolprasert et al. 1994).

A simpler, cheaper and more portable form of TD is SPME, which is not a dynamic system

but involves the introduction of a polymer-coated fused silica fibre into the headspace of a

heated solid sample and sorbing the liberated organic analytes onto a coating specific for

the compounds of interest. The analytes are then desorbed into the GC injection port and

analysed. The fibre could also be immersed directly into an aqueous sample (Shirey 2000,

Valor 2001, Barrionuevo and Lanças 2000, Batlle et al. 1999a,b). This technique has

already been used as a means to isolate degradation products in polymers (Hakkarainen et

al. 1997, Khabbaz et al. 1998) and in the determination of acetaldehyde (Huynh and Vu-

Duc 1998) and butylated hydroxytoluene (Tombesi and Hugo, 2002) in drinking water

stored in PET bottles. Apart from polymers, some of the other uses of SPME include the

analysis of organic compounds in indoor air (Jia et al. 2000) and plants (Bicchi et al. 2000);

volatile aroma compounds in pork (Elmore et al. 2000) and fruit (Augusto et al. 2000);

polychlorinated byphenyls in soils (Llompart et al. 1999); and for the characterisation of

cheeses (Pérès et al. 2001).

The three heat driven techniques described involve concentrating headspace analytes onto a

selective fibre prior to GC analysis. Conversely, SHS involves withdrawing an aliquot of

the headspace vapour with a syringe and injecting it into a GC. The most significant

limitation of the static headspace technique is lack of sensitivity. This is because once

equilibrium is reached no more analytes are released from the polymer matrix. SHS has

been used to determine the amount of flavour absorbed into plastic packaging (Tavss et al.

1988); residual acetaldehyde in PET (Dong et al. 1980) and PET-bottled mineral water

(Lorusso et al. 1985); residual solvents in food packaging films (Kolb et al. 1981);

degradation products in polymers (Hakkarainen et al. 1997); contaminants in post-

consumer PET flake (Franz and Welle 1999a); degradation compounds in irradiated PET

(Komolprasert et al. 2001) and the determination of migrants from food packaging

materials in aqueous food simulants and real food (Nerin et al. 2002).

6.1.3 Brief outline of chapter

215

Chapter 6

In the current study we used SPME and automated SHS to extract volatile compounds out

of recycled PET flake and pellets. The study was initiated by optimising SPME in terms of

fibre selection and extraction temperature for the chosen fibre. The fibres investigated

were polyacrylate (PA), polydimethylsiloxane (PDMS) and carboxen/polydimethyl

siloxane (CX/PDMS). The latter fibre, which is porous and extracts analytes by adsorption,

was selected for temperature optimisation because it was shown to extract volatile

compounds more efficiently than the absorbing fibres, PA and PDMS.

It is well established that an increase in temperature increases the diffusion rate of a

compound to the polymer surface. This is because thermal energy assists in the movement

of analytes through the polymer matrix and forms transient voids, created by the increased

motion of the polymer chains (Cotton et al. 1993).

The effect of increasing temperature on the extraction efficiency of polymers has already

been shown when optimising the parameters of accelerated solvent extraction (ASE),

supercritical fluid extraction (SFE) and microwave-assisted extraction (MAE). Lou et al.

(1996, 1997) discussed the temperature optimisation behind the ASE and SFE of monomers

and oligomers from polymeric samples. St Küppers (1992) and Cotton et al. (1993)

optimised extraction temperature whilst removing cyclic trimer from PET by SFE.

Camacho and Karlsson (2001) applied an optimised temperature to extract low molecular

weight contaminants from HDPE and PP by MAE.

Similar temperature optimisations have been performed in our study for the SHS and

SPME of contaminants out of recycled PET in a quest to achieve total extraction for

quantitation by external standardisation.

Due to CX/PDMS displaying competitive adsorption behaviour with increasing extraction

temperature as a result of the enhanced headspace concentrations, it was considered to be

an inappropriate fibre for quantitative analysis. Therefore, quantitative analysis was

achieved by means of SHS (instead of SPME) and selected ion-monitoring (SIM) GC/MS.

216

Chapter 6

6.2 QUALITATIVE SPME STUDY OF CONTAMINANTS IN PET EXTRUDED

PELLETS

6.2.1 Comparison of the compounds extracted by different fibres

The types of compounds extracted by the four different fibres were compared for washed

and dried flake ground to >425 µm-700 µm and are presented in Table 6.1. This Table

represents the qualitative analysis of flake samples analysed in triplicate by different fibres.

Retention times and the quality of library match for each peak are listed in this table. Up to

55 compounds are listed; hydrocarbons, organic acids, alcohols, ketones and aldehydes.

For some compounds the library matches were low, possibly due to incomplete library,

peak overlap and/or small peak size (below the universally accepted signal:noise ratio of 3)

preventing unequivocal identification based on library search. For example, the peak at

0.834 minutes (identified in Table 6.1 as “3-pentanol” with a library match quality of “35”)

has different mass spectra (and therefore library matches) at discrete points of the peak,

suggesting that this peak consists of co-eluting compounds with different mass spectra.

Further analysis using the “peak purity” option of the GC/MS software deduced that there

were six components making up this peak. Therefore, this peak could not be identified

with accuracy.

There were other examples where peak overlap affected the identification process (e.g. the

peaks at 1.004 and 1.010 minutes, which overlapped with each other and with other

compounds; the peak at 4.628 which overlapped with other compounds, as suggested by

different mass spectra at discrete points of the peak). Such overlapping peaks –

occasionally having low library match qualities (below a quality of 80) - were not excluded

from further analysis as “selected ion extraction” allowed the isolation of individual

chromatographic peaks for semi-quantitative purposes. Unfortunately, any qualitative

rationalization for the chemical behaviour of these compounds based on their molecular

structure could not be achieved, as the mass spectral library could not provide an

unequivocal structural identification due to the low match quality.

Paradoxically, it should also be kept in mind that a high library match quality for a

particular peak does not necessarily guarantee unequivocal identification either; thus one

should keep an open mind towards identification based on library match. Coincidence in

217

Chapter 6

retention time between the unknown peak and a standard is a confirmatory means of peak

identification. Similar compound identification using standards was carried out in Chapter

4 during the extraction of contaminants out of flake.

Out of the four fibres tested (see Table 6.1), the adsorptive CX/PDMS fibre was selected

for further analyses, as it was the only fibre that accounted for the large number of very

volatile compounds eluting early in the gas chromatogram (prior to 6.040 minutes). The

most inefficient fibre in terms of number of contaminants extracted was the non-polar

absorptive 7 µm PDMS fibre, possibly due to its thin film and limited capacity for

absorbing compounds. The 100 µm PDMS fibre was more effective in terms of the number

of compounds extracted; however this fibre is inappropriate for the analysis of the early-

GC eluting compounds, and for naphthalene and methylnaphthalene derivatives (which can

be seen from Table 6.1 not to be detected). The polar PA fibre was as efficient as the

CX/PDMS fibre for extracting semivolatiles, including the naphthalene compounds whose

polarisability (from the electron density inflicted by the π-bonds) presumably permits their

adsorption into the polar fibre. Again, there were no early eluting analytes with retention

times prior to 6.040 minutes that could be identified for the PA fibre; therefore, like the

PDMS fibres, it was not selected for further analysis.

GC/MS analysis of the PDMS and PA fibres included a 2-minute solvent delay to exclude a

large swamping dichloromethane peak, which was particularly absorbed by these fibres. It

was presumed that the presence of dichloromethane in the polymer could have resulted

from environmental contamination given that this solvent was extensively used during our

solvent extraction studies (Chapters 4 and 5). Hence it was excluded from further analysis.

Table 6.1: Compounds extracted by four different fibres from ground flake (x indicates

assignment and n/a = “not analysed” due to the inclusion of a solvent delay time).

RT M/Z Compound (library match quality in PDMS/CX PDMS 1 PDMS 2 PA

218

Chapter 6

(mins) parentheses) (100 µm) (7 µm) 0.834 55, 56 3-Pentenol (35)

1-Pentene, 3,4-dimethyl- (33) Cyclopropane, 1,1-dimethyl- (14)

x n/a n/a n/a

1.004 55 1-Butene, 3-methyl- (40) 1,5-Hexadiene, 3-methyl- (9)

x n/a n/a n/a

1.010 73 1,3-Dioxolane (70)

x n/a n/a n/a

1.279 78, 51 Benzene (90)

x n/a n/a n/a

2.051 91, 65 Toluene (72) Ethylene glycol (3)

x n/a n/a n/a

2.294 55 1-Octene (59) 2-Octene (50) Cyclopropane, 1-butyl-2-pentyl-, trans (50)

x

3.202 91, 106 Benzene, ethyl- (81) 3,5-Heptadiyn-2-one (38) 1-Cycloocten-5-yne (38)

x

3.350 91, 106 Benzene, 1,4-dimethyl- (83) Benzene, 1,3-dimethyl- (81) Benzene, 1,2-dimethyl- (81)

x

3.667 104, 78 Propanedinitrile, methylene-(86) 2-Butenedinitrile (72)

x

4.628 91, 120 Benzene, propyl- (52) Benzacetaldehyde (50)

x

4.743 105, 120 Benzene, 1-ethyl-2-methyl- (91) Benzene, 1,2,3-trimethyl- (87)

x

5.017 105, 120 Benzaldehyde (80)

x

5.29 81, 53 2-Ethylfuran (56) 1H-Pyrrole, 1-Methyl- (50)

x

5.470 57, 85, 71

Octane, 2,4,6-trimethyl- (64) Octane, 2,3-dimethyl- (47) Decane, 2,5,6-trimethyl- (42)

x

5.608 57, 55 1,3-Cyclopentanediol, cis- (38) 1,3-Cyclopentanediol, trans- (32) Octanal (32)

x

5.946 105, 120 Benzene, 1-ethyl-2-methyl- (49) Benzene, 1-ethyl-3-methyl- (43)

x

219

Chapter 6

Benzene, 1-methylethyl- (40)

6.040 119 m-Cymene (90)

x x x

6.125 67, 68, 93

Limonene (80)

x x x x

6.190 57, 83 1-Hexanol, 2-ethyl- (32) 1-Pentanol, 2-ethyl-4-methyl- (37) Ethanol-2-[(2-ethyl hexyl) oxy]- (37)

x x x

6.747 93 γ-Terpinene (58)

x x x

7.571 57, 71, 85

Undecane (91) Undecane, 3,6-dimethyl- (91) Dodecane, 3-methyl- (72) Nonane, 2-methyl- (72)

x x x x

7.687 57, 70

Hepten-1-ol (53) Nonanal (50) Cyclopentane, 1-methyl-2-propyl- (47) 9-Dodecenol (35)

x x x x

9.167 128 Naphthalene (89) Azulene (81)

x x

9.240 58, 71 2-Decanone (59) Undecanone (50) 2-Undecanone (45)

x x x

9.288 120,152 Salicylic acid, ME (72) Salicylic acid, AO (32)

9.325 57, 71, 85

Undecane (83) Tritetracontane (83) Nonane, 2-methyl- (64) Tridecane, 6-methyl- (59) Dodecane, 2,7,10-trimethyl- (59)

x x x x

9.478 121, 150 n-Propylanisole (64)

x x x

9.537 57, 71 Dodecane, 6-methyl- (83) Undecane, 3,6-dimethyl- (74) Decane, 2,6,8-trimethyl- (50) Octane, 2,5-dimethyl- (47)

x x x x

9.998 55, 83 Cyclohexane, 1,1’-(1,4-butanediyl)bis- (78) Cyclohexane, (1,3-dimethyl butyl)- (50) 1-Azabicyclo (3.1.0) hexane (47)

x x x

10.372 57, 71, Heptadecane, 2,6,10,15-tetramethyl- x x

220

Chapter 6

85 (80) Decane, 2,9-dimethyl- (72)

10.464 57, 71, 85

Octane, 2,3,7-trimethyl- (72) Octane, 2,6-dimethyl- (59) Octane, 2-methyl- (53) Dodecane, 2,7,10-trimethyl- (50)

x x x x

10.627 60, 73, 87

Octanoic acid, silver (1+) salt (72) Octanoic acid (64) Nonanoic acid (28)

x x x

10.900 57, 71 84 Tridecane (90) Decane, 2,3,5-trimethyl- (83) Undecane, 3,6-dimethyl- (64) Dodecane, 2,7,10-trimethyl- (72) Undecane, 2,9-dimethyl- (64)

x x x x

9.07 60, 73, 84

Octanoic acid (38) Nonanoic acid (25) D-Lyxose (23)

x x

10.947 115, 141, 142

Benzocycloheptatriene (87) Naphthalene, 1- methyl- (64)

x x

11.194 115, 141, 142

1,4-Methanonaphthalene, 1,4-dihydro- (91) Benzocycloheptatriene (58) Naphthalene, 2-methyl- (38)

x x

11.586 55, 83, 82

Cyclohexane, undecyl- (78) Cyclohexane, octyl- (64) Cyclohexane, 1,1’-(1,3-propanediyl)bis- (64) Cyclohexane, 2-propenyl- (64)

x x x

11.657 57, 71, 85

Dodecane, 2,6,11-trimethyl- (64) Heptadecane, 2,6-dimethyl- (64)

x x x

11.731 92, 119, 120

Benzoic acid, 2-amino-, methyl ester (93) 2-Picadine, 6-nitro- (50)

x x

11.854 57, 71, 85

Undecane, 3,9-dimethyl- (72) Undecane, 2,9- dimethyl- (40) Dodecane, 3-methyl- (33) 1-Hexene, 3,5,5-trimethyl- (28)

x x x

11.923 57, 70 Hexatriacontane (64) Hexane, 3-ethyl-4-methyl- (43) Heptane, 2,2,3,4,6,6-hexamethyl- (43) Decane, 2,6,7-trimethyl- (35)

x x x

12.012 57, 71, 85

Heptadecane, 2,6,10,14-tetramethyl- (90)

x x x x

221

Chapter 6

Dodecane, 2,6,11-trimethyl- (53) 1-Octanol, 2-butyl- (47) Nonadecane (47)

12.199 154 Naphthalene, 2-ethenyl- (78) 1,1’-Biphenyl (46) 2-Quinolinecarbonitrile, 1-oxide (46)

x

12.363 57, 71, 84

Tetradecane (91) Heptadecane (83) Undecane, 5,6-dimethyl- (64) Octane, 2,4,6-trimethyl- (64) Undecane (64)

x x x

12.541 55, 57 Tetradecanal (64) Hexadecanol (58) Tridecanal (59)

x x x

12.603 141, 142 Naphthalene, 1,3-dimethyl- (70) Naphthalene, 1,6-dimethyl- (62) Naphthalene, 2,3-dimethyl- (62)

x x

12.762 141, 142 Naphthalene, 1,6-dimethyl- (93) Naphthalene, 2,6-dimethyl- (92) Naphthalene, 1,5-dimethyl- (83)

x x

12.825 141, 142 Naphthalene, 1,5-dimethyl- (87) Naphthalene, 2,6-dimethyl- (84) Naphthalene, 1,3-dimethyl- (64)

x x

13.089 83, 82 Octane, 2-cyclohexyl- (64) Cyclohexane, 1-propyl- (50) Cyclohexane, 1,1’-methylene bis- (50)

x x

13.428 55, 69, 83

Cyclotetradecane (78) Cyclododecane (78) 2-Dodecene (64)

x x x

13.726 57, 71, 85

Decane, 2,3,5-trimethyl- (90) Dodosane (83) Heptane, 2,6-dimethyl- (64) Octane, 2,7-dimethyl- (64) Nonadecane (64)

x x x x

14.900 71, 111, 173

Propanoic acid, 2-methyl-, 1-(1,1-dimethyl)- (64)

x x

15.001 57, 71, 85

Tetradecane (86) Docosane (78) Nonadecane (64) Docosane (59) Octane, 2,4,6- trimethyl- (59)

x x x x

15.548 57, 71, 85

Heptadecane, 2,6-dimethyl- (64) Tetratetracontane (22)

x x x

222

Chapter 6

Tetradecane, 2,5-dimethyl- (22) Decane, 2,6,8-trimethyl- (22)

16.217 57, 71, 85

Hexacosane (86) Nonane, 3,7-dimethyl- (80) Heptadecane, 2,6,10,15-tetramethyl- (72) Tritetracontane (64)

x x x x

17.416 57, 71, 85, 97

Hexacosane (80) Undecane (53) Decane, 2,4-dimethyl- (50) 1-Octanol, 2-butyl- (50) Tetracontane, 3,5,24-trimethyl- (50)

x x x

6.2.2 Effect of temperature on extraction

Temperature optimisations (referred to as “thermodynamic extraction studies”) on PET

were performed using a CX/PDMS fibre. It is well established that adsorption is an

exothermic process and desorption is an endothermic process that dominates with

increasing temperature. However, it was anticipated that the high temperatures would

release the analytes from the polymer matrix into the headspace. Therefore, after each

succeeding extraction temperature (e.g. 130°C, 160°C, 195°C…) the temperature was

decreased to 90°C for SPME sampling in order to reduce the effects of analyte-fibre

desorption occurring with elevated temperatures. Once the temperature was decreased to

90°C, a large proportion of the analytes would remain in the headspace ready to be

adsorbed onto the fibre and analysed.ℵ Thus the effects of extraction effectiveness are

demonstrated rather than the effects of analyte-fibre desorption with increasing

temperature. Some of the plots obtained for contaminant peak area (referred to as

“abundance”) versus extraction temperature for ground and unground annealed pellets

(whole pellet size = 3mm x 3mm) are shown in Figures (6.1a – 6.1j).

The purpose of these plots was to demonstrate the effects of particle size and temperature

on the extraction efficiency of some compounds. The “abundances” in Figures 6.1a-6.1j

were determined by selecting a representative ion for the peak in the gas chromatogram.

This avoided the problems associated with integrating overlapping peaks.

223

Chapter 6

Several of the compounds in these plots present an adsorption maximum followed by a

decrease of the absorbed quantity, with increasing temperature (e.g. benzene; 2,4,6-

trimethyloctane; limonene). Of all compounds in Figures (6.1a – 6.1j) the ones with a

lower library match quality than 80 were from Figures 6.1 (a), (e) and (f) (see Table 6.1 for

the library match qualities of the corresponding compounds). The lowest match quality

was observed for “3-pentanol” [represented by (a) in Figure 6.1], whose identity was

improbable, most likely due to peak overlap. The reason for the low match qualities for

compounds represented by (e) and (f) in Figure 6.1 could stem from the insignificant size

of the chromatographic peak as well as peak overlap.

Figures 6.1a-6.1j: Ground annealed pellets: contaminant area (abundance) versus extraction

temperature for three different particle sizes using the CX/PDMS fibre. Plots for selected

analytes are shown in Figure 1a - 1j. ∆ = Pellets, = >425-700µm, X = 0-300µm.

(a)

3-Pentanol

0

4000000

8000000

12000000

16000000

70 120 170 220

Extraction Temperature (Degrees)

Abundance

(b)

ℵ As the volatiles released at higher temperatures condense on the surface of PET at 90°C, this system is only valid when the amount of compounds released is very high so that the concentration in the vapour phase is enough to surpass the detection limit.

224

Chapter 6

Benzene

0

100000000

200000000

300000000

75 95 115 135 155 175 195 215 235 255 275

Extraction Temperature (Degrees)

Abu

ndan

ce

(c)

Benzene, ethyl-

0200000400000600000800000

100000012000001400000

75 95 115 135 155 175 195 215 235 255 275

Extraction Temperature (Degrees)

Abu

ndan

ce

(d)

p-Xylene

0

5000000

10000000

15000000

20000000

25000000

75 125 175 225 275

Extaction Temperature (Degrees)

Abu

ndan

ce

(e)

225

Chapter 6

Benzene, propyl-

0

300000

600000

900000

1200000

75 95 115 135 155 175 195 215 235

Extraction Temperature (Degrees)

Abu

ndan

ce

(f)

Octane, 2,4,6-trimethyl-

0500000

10000001500000200000025000003000000

75 95 115 135 155 175 195 215 235 255

Extraction Temperature (Degrees)

Abu

ndan

ce

(g)

Limonene

0

2000000

4000000

6000000

8000000

65 85 105 125 145 165 185 205 225 245 265

Extraction Temperature (Degrees)

Abu

ndan

ce

(h)

226

Chapter 6

Naphthalene

0

5000000

10000000

15000000

20000000

75 95 115 135 155 175 195 215 235 255

Extraction Temperature (Degrees)

Abu

ndan

ce

(i)

Toluene

0

5000000

10000000

15000000

75 125 175 225 275

Extraction Temperature (Degrees)

Abu

ndan

ce

(j)

227

Chapter 6

m-Cymene

0

500000

1000000

1500000

2000000

75 125 175 225

Extraction Temperature (Degrees)

Abun

danc

e

The effect of temperature on extraction efficiency was investigated in greater depth for

extruded pellets and the results, which often reach an adsorption maximum and then

decrease with temperature, are shown in Figures (6.2a – 6.2d). This behaviour, which is

mainly associated with “adsorptive” fibres, has already been observed in the past by other

researchers, who attributed the distinctive shapes of the thermodynamic plots to

displacement effects arising from competitive adsorption (Semenov et al. 2000; Tuduri et

al. 2001; Murray 2001).

It should be further mentioned that when heating polymers at high temperatures (over

130°C) the degradation of the polymer increases and more compounds may appear, as a

result of the breaking of chains and also from the degradation of some additives (Nerin et

al. 1998b; 1996). Therefore, in some cases the rise in compound levels with increasing

temperature could result from the degradation of thermally liable compounds in the

polymer or the polymer itself.

228

Chapter 6

Figure 6.2: Effect of incubation temperature on extraction of 6g of unground extruded

pellets using the CX/PDMS fibre. Curves for compounds are plotted in Figures 6.2a –

6.2d.

(a)

0

5000000

10000000

15000000

20000000

85 105 125 145 165 185 205 225 245 265

Temperature (degrees)

Abu

ndan

ce

3-Pentenol 1-Butene, 3-methyl-Benzene Toluenep-Xylene NaphthaleneEthylbenzene Limonene1-Methyl-2-propylcyclopentane

229

Chapter 6

6.2 (b)

0

200000

400000

600000

800000

1000000

1200000

1400000

1600000

80 100 120 140 160 180 200 220 240 260 280

Temperature (degrees)

Abu

ndan

ce

n-Propylanisole Biphenyl2-Dimethylnaphthalene 1-Methylnaphthalenem-Cymene gamma-Terpinene1,2,4-Trimethylbenzene Propylbenzene1-Ethyl-,2-methyl-,benzene Undecane

230

Chapter 6

6.2 (c)

0

500000

1000000

1500000

2000000

2500000

3000000

3500000

4000000

75 95 115 135 155 175 195 215 235 255 275

Temperature (Degrees)

Abu

ndan

ce2-Ethylfuran cis-1,3-Cyclopentanediol2,4,6-Trimethyloctane 1-OcteneTritetracontane TetradecaneTridecane Octanoic Acid2-Decanone

231

Chapter 6

6.2 (d)

0

20000000

40000000

60000000

80000000

100000000

120000000

80 100 120 140 160 180 200 220 240 260Temperature (degrees)

Abun

danc

eBenzaldehydePropanedinitrile, methylene-1,3-Dioxolane

232

Chapter 6

As temperature increases the total contaminant concentration in the headspace increases

and there is rising potential for reversible binding and displacement of low molecular

weight contaminants by high molecular weight contaminants (Jia et al. 2000, Semenov et

al. 2000; Tuduri et al. 2001). This is because larger molecules generally have a higher

affinity for the adsorptive fibre than smaller molecules. Therefore, when contaminant

concentrations are similar, the larger molecules have a longer range, which is

approximately linear but (steadily) increasing over a wide temperature range. Long “linear”

ranges were observed for n-propylanisole, biphenyl, 2-dimethylnaphthalene, 1-

methylnaphthalene, benzaldehyde (two consecutive “linear” sections), tridecane,

tetradecane. Comparatively smaller compounds that did not undergo the expected

competitive adsorption until 250°C were 3-pentenol∗; 1-butene, 3-methyl-∗; and p-xylene.

These contaminants could be present in comparatively large quantities, impeding their

displacement until the level of displacing compound increases significantly in the

headspace. This is because the competitive adsorption process follows an exchange

mechanism, which is concentration and affinity dependent. Equations 6.1- 6.4 describe the

displacement of analyte “X” by analyte “Y”, where K is the equilibrium constant that

describes the affinity one compound has to the surface relative to the competing compound.

Y + X:SURFACE ↔ X + Y:SURFACE (Equation 6.1)

K = [Y:SURFACE][X] / [Y][X:SURFACE] (Equation 6.2)

∴[Y:SURFACE] = K[Y][X:SURFACE]/ [X] (Equation 6.3)

and

[X:SURFACE] = [Y:SURFACE] [X]/ K[Y] (Equation 6.4)

∗ Due to the poor library matches obtained for “3-pentanol” and “1-butene, 3-methyl-”, the correct identity of these two earliest eluting compounds was uncertain. The compounds were assumed to be volatile relative to the other compounds because they eluted very early in the gas chromatogram. In gas chromatography retention time is a reflection of analyte boiling point and stationary phase-analyte interactions.

233

Chapter 6

It is implied by these equations that “X” type compounds with a lower affinity for the

surface and high headspace concentrations (high K values) will not be displaced by “Y”

type compounds with higher affinities until their headspace concentrations ([Y]) reach a

sufficiently high level.

In Figures (6.2a – 6.2d) for a few of the contaminants showing signs of early competitive

adsorption (e.g. benzene, limonene, m-cymene, undecane, propylbenzene and 1-ethyl-2-

methyl-benzene), the amount of contaminant adsorbing onto the fibre begins to increase

after the steep decrease, which signifies an exchange in displacement. It is presumed that

this change in trend results from high extraction temperatures releasing more volatile

contaminants from the internal body of the polymer into the headspace at a faster rate than

the displacing compounds, which are mainly semivolatiles. This increase in the headspace

concentration ratio between the analyte of low affinity and the analyte of high affinity

causes the replacement of some high affinity contaminants by some low affinity

contaminants. An alternative explanation was made by H.-J. Cho et al. (2003) who

suggested that capillary condensation at high concentrations could be responsible for the

observed increase in area.

The effect of decreasing particle size on amount extracted is difficult to summarise from

Figures (6.1a – 6.1j). General remarks are difficult to make because the particle size order

changes with increasing temperatures (curves cross over). One generalisation that could be

made at low temperatures is that the “∆” (representing the pellets) are always lowest-lying

before 160°C. This is consistent with the theory that least contaminant is liberated from the

internal body of the largest particle size to the headspace.

Variation in particle size order with increasing temperature could result from coalescence

between the compactly packed smaller particles taking place at higher temperatures,

reducing the amount of contaminant liberated. In addition, because each point represents

only one experimental value, it is possible that experimental error is another factor

contributing to the changes in particle size order. Another feasible explanation is that the

total level of analytes in the headspace is the largest for the smallest particle size (due to

larger surface are per unit mass of PET) therefore the number of available sites on the fibre

decreases rapidly and molecules start competing for them. The effect of increasing total

analyte concentration on adsorption has previously been demonstrated by H.-J. Cho et al.

234

Chapter 6

(2003). It is also possible that the compounds in the internal body of the largest particles

reach the surface and are released when temperature increases.

6.2.3 Effect of mass of sample on extraction

The amount of recycled PET pellets analysed by SPME was reduced from 6 g to 1 g in an

attempt to reduce analyte concentration in the headspace and therefore potential saturation

and competitive adsorption. Ezquerro et al. (2002) demonstrated that the size of the vial

could also effect the headspace concentration and therefore the amount of analyte adsorbed.

Figures 6.3a – 6.3c present the thermodynamic results obtained for 1 g of pellets. The

compounds presented in these graphs are those that demonstrated competitive adsorption in

Figures 6.2a – 6.2d. The extraction temperatures sampled were between 160°C and 218

°C. Limonene (Figure 6.3a) and benzene (Figure 6.3b) started to exhibit displacement

after 180°C. In Figures 6.2a-6.2d the same compounds started to undergo competitive

adsorption after 160°C and 168°C respectively. Therefore competitive adsorption was

postponed when using a smaller sample mass. Ethylbenzene, which displayed competitive

adsorption between 168°C and 250°C when 6 g of pellets were analysed (Figure 6.2a), did

not display signs of displacement when 1g was sampled (Figure 6.3c). A similar result was

observed for undecane, which did not show any displacement effects when 1g of PET was

analysed.

In Figure 6.2b, m-Cymene underwent competitive adsorption after 180°C whereas in

Figure 6.3a there was no decrease in adsorption prior to 210°C.

A change in adsorption was not observed until 210°C for 2,4,6-trimethyloctane (Figure

6.3a) and not at all for 1-ethyl-2-methyl-benzene (Figure 6.3a). Conversely, in Figures 6.2c

and 6.2b respectively, adsorption started to decline for these compounds after 168°C and

180°C. A similar delay in competitive adsorption was observed for 2-ethylfuran, which did

not demonstrate competitive adsorption prior to 200°C (Figure 6.3a).

In an attempt to prevent the displacement of limonene; m-cymene; benzene, 2,4,6-

trimethyloctane; 2-ethylfuran; and methylene propanedinitrile an even smaller mass of

pellets was analysed (0.3 g). Figures (6.4a – 6.4b) indicate that the decrease in mass caused

an increase in the “linear” range (range over which there is a steady increase in abundance

235

Chapter 6

with temperature) for 2,4,6-trimethyloctane, 2-ethylfuran and methylene propanedinitrile.

There was no change in the thermodynamic pattern for the remaining three compounds

(limonene; m-cymene; benzene).

236

Chapter 6

Figure 6.3a-c: Effect of incubation temperature on extraction of 1g of unground extruded

pellets using the CX/PDMS fibre.

(a)

0

500000

1000000

1500000

2000000

2500000

3000000

3500000

4000000

4500000

150 170 190 210 230

Temperature (Degrees)

Abu

ndan

ce

Limonene m-Cymene1-Ethyl-,2-methyl-,benzene 2-EthylfuranUndecane 2,4,6-TrimethyloctanePropanedinitrile, methylene-

237

Chapter 6

6.3 (b)

Benzene

0

50000000

100000000

150000000

200000000

250000000

300000000

350000000

400000000

450000000

150 170 190 210 230Temperature (Degrees)

Abu

ndan

ce

238

Chapter 6

6.3 (c)

0

1000000

2000000

3000000

4000000

5000000

6000000

7000000

150 170 190 210 230Temperature (Degrees)

Abu

ndan

ce

Ethylbenzene

239

Chapter 6

Figure 6.4: Effect of incubation temperature on extraction of 0.3g of unground extruded

pellets using the CX/PDMS fibre. Curves for compounds are plotted in Figures 6.6a –

6.6b.

(a)

0

200000

400000

600000

800000

1000000

1200000

1400000

150 170 190 210 230Temperature (Degrees)

Abu

ndan

ce

Propanedinitrile, methylene- m-CymeneLimonene 2,4,6-Trimethyloctane2-Ethylfuran

240

Chapter 6

6.4 (b)

Benzene

0

20000000

40000000

60000000

80000000

100000000

120000000

140000000

160000000

180000000

150 170 190 210 230Temperature (Degrees)

Abu

ndan

ce

241

Chapter 6

6.2.4 Effect of adsorption time

The fibre exposure time of 30 minutes was originally selected in order to achieve fibre-

headspace equilibrium, which is required to maintain consistency in results. When working

in the pre-equilibrium period, time control is critical because a small change in time results

in large change in analyte absorbed.

The adsorption time was reduced from 30 minutes to 5 minutes and extraction of pellets

was carried out at 160°C and 200°C. The contaminants analysed were limonene and

benzene because they underwent competitive adsorption before 200°C in Figures 6.3a and

6.3b and were thus considered the worst-case compounds in terms of competitive

adsorption. Table 6.2 indicates that there was a decrease in the amount absorbed for both

compounds with increasing temperature. This implied that competitive adsorption

nonetheless occurred at the reduced sorption time.

Table 6.2: Area of benzene and limonene after reducing the fibre exposure time from 30

minutes to 5 minutes.

Incubation Temperature

(°C)

Contaminant

Benzene (m/e 78) Limonene (m/e 68)

160

200

160873631

112310430

3093681

1403324

242

Chapter 6

6.2.5 Effect of extraction time on extraction

The extraction time is a critical parameter in the SPME sampling process - the longer the

extraction time, the larger the distribution of the analytes into the headspace.

Ground annealed pellets (>425-700µm) were extracted at 160°C for different times (0.5h –

7.2 h) in an attempt to determine the kinetics of extraction. In Figures 6.5a – 6.5e

consistent increase in abundance with time was observed for tetradecane; 1-ethyl, 2-

methyl-benzene; 2-decanone; cis-1,3-cyclopentanediol; and tritetracontane. The

abundances for the remaining analytes were shown to either increase abruptly after 5h (e.g.

ethylbenzene; 1-butene, 3-methyl), remained fairly constant throughout the study (e.g. p-

xylene; n-propylanisole) or showed signs of competitive adsorption (e.g. limonene;

tridecane). Due to the limitations associated with the use of the CX/PDMS fibre, the

effects of extraction time on extraction were not accurately demonstrated for all

compounds. A simple explanation for the different shapes cannot be made because

displacement on the CX/PDMS is not only dependent on the analyte size and vapour

pressure (H.-J. Cho et al., 2003) but also on its relative headspace concentration, which is

not known.

243

Chapter 6

Figure 6.5a-e: Effect of extraction time on abundance.

(a)

2500000

3500000

4500000

5500000

6500000

7500000

8500000

9500000

10500000

11500000

0 1 2 3 4 5 6 7 8

Time (h)

Abu

ndan

ce3-Pentenol 1-Butene, 3-methyl-p-Xylene Propanedinitrile, methylene-1,2,4-Trimethylbenzene 2,4,6-TrimethyloctaneLimonene Undecane

244

Chapter 6

6.5 (b)

550000

1050000

1550000

2050000

2550000

3050000

3550000

4050000

0 1 2 3 4 5 6 7

Time (h)

Abu

ndan

c

8

e

1,2,4-Trimethylbenzene 1-Ethyl-,2-methyl-,benzene2-Ethylfuran m-CymeneNaphthalene TetradecaneTridecane

245

Chapter 6

6.5 (c)

0

200000

400000

600000

800000

1000000

1200000

1400000

1600000

1800000

2000000

0 1 2 3 4 5 6 7 8Time (h)

Abu

ndan

ce

1-Octene EthylbenzenePropylbenzene Propylbenzenecis-1,3-Cyclopentanediol gamma-TerpineneBiphenyl n-Propylanisole2-Decanone Tritetracontane

246

Chapter 6

6.5 (d)

100000000

120000000

140000000

160000000

180000000

200000000

220000000

240000000

260000000

280000000

0 1 2 3 4 5 6 7 8Time (h)

Abu

ndan

ceBenzene

247

Chapter 6

6.5 (e)

8000000

9000000

10000000

11000000

12000000

13000000

14000000

15000000

16000000

17000000

0 1 2 3 4 5 6 7 8Time (h)

Abu

ndan

ceToluene

248

Chapter 6

6.3 QUANTITATIVE SPME AND STATIC HEADSPACE STUDY OF RECYCLED

PET

6.3.1 Quantitation using the CX/PDMS fibre

As Murray (2001) pointed out, the relative proportions of the analytes adsorbed onto the

fibre depend on their ratio in the headspace. Therefore during external quantitation, only a

standard containing all components at the same relative concentrations is appropriate.

Small errors in composition of standards and sample can result in large errors in

quantitation. Therefore, the identity and concentration of every component in the sample

that might displace low molecular weight compounds from the fibre must be known so that

an identical standard is prepared. Otherwise an absorptive fibre or static headspace must be

used for quantitation of multicomponent matrices. External calibration could really only be

used with the CX/PDMS fibre when the concentrations of the analytes are low and

interfering compounds are absent (H.-J. Cho et al. 2003).

Theoretically, the quantitative method of standard additions could be successful whilst

using the CX/PDMS fibre because it insures that the ratio of analytes in the headspace

remains constant for standards and samples. However the amount of each compound must

not exceed the linear range for accurate quantification. In a past study by Ezquerro et al.

(2003) the method of standard additions was not effective during the SPME extraction of

packaging materials using the CX/PDMS fibre, therefore this method was not investigated

here.

6.3.2 SPME using PDMS, an absorption fibre

Due to the effects of competitive adsorption observed with adsorption-type fibres,

thermodynamic studies were carried out using PDMS, the conventional absorption-type

fibre. The PDMS fibre was far less sensitive (estimated x 60) than the CX/PDMS fibre for

the analysis of extruded pellets, as indicated by the superimposed chromatograms (Figure

6.6). Only limonene and a few hydrocarbons were identified. As shown in Figure 6.7a-

6.7b, the contaminants analysed by PDMS present a weak relationship (R2) between

abundance and extraction temperature for most of the compounds, which do not always

have ascending trends, even at low temperatures and possibly due to the large degree of

249

Chapter 6

scatter. The large spread is thought to result from the irreproducibility associated with the

PDMS fibre for the analysis of contaminants present in trace quantities. Due to the

limitations surrounding SPME, a different thermal extraction method was sought which

accounted for the low boiling point compounds and was even simpler in terms of

equipment – static headspace analysis.

Figure 6.6: Superimposed chromatograms obtained from the analysis of pellets using the

PDMS (bold) and CX/PDMS (fine) fibres. c

Αbundan

250

Time (min)

Chapter 6

Figure 6.7 (a-b): Effect of incubation temperature on extraction of 6g of unground extruded

pellets using the 100 µm PDMS fibre.

(a)

y = 2800.7x + 80720R2 = 0.1744

y = -1066x + 388532R2 = 0.1988

0

100000

200000

300000

400000

500000

600000

700000

800000

900000

1000000

80 130 180 230Temperature (Degrees)

Abu

ndan

ce

Limonene Cyclopentene, 1-methyl-2-propyl-

251

Chapter 6

6.7 (b)

y = 1836.3x - 169456R2 = 0.5481

y = 1180.8x - 120351R2 = 0.5348

y = -146.55x + 78174R2 = 0.0294

y = 1303.1x - 122395R2 = 0.5034

0

50000

100000

150000

200000

250000

300000

80 130 180 230

Temperature (Degrees)

Abun

danc

eUndecane TritetracontaneDodecane, 6-methyl- TridecaneLinear (Tridecane) Linear (Undecane)Linear (Dodecane, 6-methyl-) Linear (Tritetracontane)

252

Chapter 6

6.3.3 Static headspace analysis (SHS)

Owing to the displacement behaviour of analytes on the CX/PDMS fibre, we were unable

to observe the optimum temperature required for the complete extraction of volatile

contaminants out of PET matrix. Therefore an analogous thermodynamic study was carried

out by SHS. Ground washed and dried flake was preliminarily screened by SHS to

determine what contaminants were present in the flake and their retention times. Since the

concentration of contaminants in recycled pellets was clearly a lot smaller, GC/MS analysis

incorporating selected ion monitoring (SIM) and retention time recognition was

subsequently used to analyse a selection of contaminants. The sensitivity of the static

headspace technique was shown to be significantly smaller than the SPME method using

the CX/PDMS fibre for the analysis of the recycled pellets (Figure 6.8a – 6.8b). Note that

the abundance scale in Figure 6.8b is ten times coarser than in Figure 6.8a. The drawback

of the static headspace method is that it only accounts for the highly volatile compounds in

trace analysis. Selected ion monitoring (SIM) GC/MS was therefore used to improve the

sensitivity of the less volatile analytes during quantitation. Note that the retention times for

the peaks obtained by both methods do not correspond as the chromatograms were obtained

on different columns. The retention times and library matches for the chromatographic

peaks obtained by the CX/PDMS fibre (Figure 10a) are listed in Table 6.1.

Figures 6.9a – 6.9c include the abundance-temperature plots for several volatile compounds

in PET pellets. These graphs denote a strong dependence upon the extraction temperature

and the amount extracted. All of the correlation coefficients (R2) for the second order

polynomials (Figure 6.9a and 6.9b) in these graphs were above 0.94, signifying strong

relationships between both variables. Second order polynomial curves were not fitted for

the analytes in Figure 6.9c, whose points form different trends.

253

Chapter 6

Figure 6.8: Chromatograms for extruded pellets obtained by (a) SHS and (b) SPME using

the CX/PDMS fibre (1=dichloromethane; 2=trichloromethane; 3=1,3-dioxane, 2-methyl-;

4=benzene; 5=limonene).

(a)

1

4

3

2 5

(b)

4

5

254

Chapter 6

Figure 6.9 (a-c): Effect of incubation temperature for the SHS of extruded PET pellets.

(a)

0

20000

40000

60000

80000

100000

120000

140000

65 85 105 125 145 165 185 205

Temperature (degrees)

Abu

ndan

ce1-Methylethyl benzene 1,2,4-Trimethylbenzenem-Cymene NaphthaleneCyclopentane, 1-methyl-2-propyl- 2,4,6-Trimethyloctane

255

Chapter 6

6.9 (b)

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

20000

50 70 90 110 130 150 170 190 210

Temperature (Degrees)

Abu

ndan

ceBenzaldehyde 3-Ethyl-o-xylene Cineolen-Propylanisole Biphenyl Propylbenzene2-Ethylfuran

256

Chapter 6

6.9 (c)

0

20000

40000

60000

80000

100000

120000

50 70 90 110 130 150 170 190 210

Temperature (degrees)

Abu

ndan

ceLimonene Toluene p-Xylene

257

Chapter 6

6.3.4 Quantitative analysis of PET

As complete extraction was not achieved during attempted temperature optimisation,

quantitation by external standardisation (ES) was not primarily considered. Instead,

multiple headspace extraction (MHE) was employed. This method, which allows for

matrix effects, involves sampling repeatedly the same sample at equal time intervals to

obtain the decay of the concentration of analytes. MHE has already been used for the

analysis of residual solvents (Kolb et al., 1981; Penton 1999) and flavours (Tavss et al.,

1988) in food packaging materials and for the determination of benzene residues in

recycled PET (Komolprasert et al., 1994). Standard additions have been applied in the

analysis of volatile organic compounds in packaging (Ezquerro et al., 2003). In this study,

the method of standard additions was found to be an inappropriate hence it was not

investigated.

6.3.5 Multiple headspace extraction (MHE)

Figure 6.10 presents a non-exponential decrease in contaminant area with time for flake.

Therefore flake cannot be quantified by this method.

258

Chapter 6

Figure 6.10: Multiple headspace analysis of flake ground to 425–700 µm.

0

2000

4000

6000

8000

10000

20 70 120 170 220Time (mins)

Abu

ndan

ce

Toluene m-CymeneEthylbenzene n-Propylanisolep-Xylene gamma-Terpinene1,2,3-Trimethylbenzene Naphthalene

259

Chapter 6

6.3.6 External standardisation (ES)

External standardisation has been applied in the past to quantify degradation products in

PET sheets by static headspace and thermal desorption (Komolprasert et al., 2001). It has

also been used to analyse contaminants in PET flakes (Nerin et al. 2003a). Ezquerro et al.

(2003) used external standardisation amongst other quantitative methods to determine the

composition of volatile organic compounds in packaging materials by SPME.

In this study flake ground to >425-700 µm and recycled pellets were extracted by static

headspace and quantified by external standardisation. Table 6.3 compares the results

obtained for static headspace and Soxhlet extraction for a few contaminants. Soxhlet

extraction was carried out as described in Chapter 3 (Materials and Methods). For flake,

there is no significant difference between both methods in the concentrations of 1,2,4-

trimethylbenzene, m-cymene, limonene and cineole. However as the molecules increase in

size (e.g. naphthalene, n-propylanisole and biphenyl), there is a significant difference

between both methods with the amount extracted being higher for Soxhlet extraction. For

recycled pellets, the amount of contaminant extracted by static headspace was on average

10 times smaller than that obtained by Soxhlet extraction.

In order to approximate the level of some volatile compounds identified by SHS but not by

the Soxhlet extraction method in recycled pellets, concentrations determined by static

headspace could be multiplied by ten to obtain a theoretical Soxhlet extraction result. For

example, the experimental concentration for toluene (Table 6.4) in recycled pellets was

determined to be 3.7 ppb by static headspace; therefore the concentration expected by the

Soxhlet extraction method is estimated to be 37 ppb (3.7 ppb x 10).

To quantify the concentration of volatiles in flake we use the static headspace value as is,

without introducing a scaling factor (concentration of toluene in flake = 73.0 ppb). Other

volatiles that were quantified by SHS were p-xylene and undecane (Table 6.4). Benzene

and ethylbenzene could not be quantified due to their standard curves displaying a large

degree of scatter, possibly due to the low concentrations investigated. The quantified

compounds had correlation coefficients (R2) were between 0.847 (toluene) and 0.998

(naphthalene).

260

Chapter 6

Table 6.3: Comparison of concentrations determined in flake and pellets by Soxhlet and

static headspace analysis. Standard deviations are in parentheses. All values are in ppb.

Contaminant Flake (ppb) Pellets (ppb) Static

headspace Soxhlet

Static

Headspace Soxhlet

1,2,4-Trimethylbenzene

24 (3)

26 (7)

0.7 (0.2)

10 (1)

m-Cymene 60

(3)

67 (16)

0.7 (0.1)

11 (1)

Limonene 1290 (189)

1102 (169)

14 (14)

126 (16)

1,2,3,5-

Tetramethylbenzene 13 (1)

4.4 (0.8)

0.1 (0.0)

1.3 (0.1)

Cineole 78 (8)

104 (51)

0.0 (0.0)

10.7 (1.2)

Naphthalene 13 (1)

29 (3)

4.0 (0.8)

32.1 (3.0)

n-Propylanisole 11 (0)

31 (4)

0.4 (0.1)

1.4 (0.1)

Biphenyl 4.7 (1.3)

16 (4)

0.4 (0.1)

5.7 (0.1)

261

Chapter 6

Table 6.4: Concentrations (ppb) of three compounds determined by static headspace but not

Soxhlet. Standard deviations are in parentheses.

Contaminant Flake (ppb) Pellet (ppb)

Toluene

73 (28)

37 (23)

p-Xylene 6.1 (2.4)

3.7 (1.8)

Undecane 21

(11) 7.9

(0.2)

6.4 CONCLUSION

The 75-µm CX/PDMS fibre was qualitatively the most efficient for analysing volatiles in

PET. Incubation and time optimisation results suggested the competitive replacement of

smaller molecules by larger molecules with increasing analyte concentrations in the

headspace. Decreasing the sample mass and reducing the fibre exposure time was not

sufficient to totally eliminate competitive behaviour with increasing temperature. The use

of PDMS (100-µm), an absorptive-type fibre, meant the absence of competitive adsorption,

but many compounds could not be accounted for and that scatter in results was

unacceptably large with increasing temperature. Due to the limitations surrounding SPME,

quantification was sought using SHS. Incubation profiles by SHS suggested incubation for

1 h at 200°C was inadequate to completely extract compounds out of PET pellets.

However, as multiple headspace extraction did not prove to be satisfactory, external

standardisation, which does not account for matrix effects was used as the method of

quantification. Comparison with Soxhlet extraction results indicated that concentrations of

all analytes were underestimated by SHS of the pellets. The concentrations of volatile, or

262

Chapter 6

early eluting, compounds in the flake approximated those determined by Soxhlet (Table

6.3).

Subsequently, the levels of toluene, undecane and p-xylene were found to be well below

the 215 ppb FDA-set threshold for flake and pellets.

263

Appendices

APPENDIX 1

Mass = density x Volume = 1.48g/cm3 (2.54 cm3/in3) (20 x 0.001 in) = 0.4588 g ~ 460 mg

268

Appendices

APPENDIX 2

Y = fraction of amorphous material

100-Y = fraction of crystalline material

21.7 = average of % crystallinity for Batch 1 and Batch 2 for flake ground

to 0-300 µm (Table 4.14)

Y (19.61/100) + (100-Y)48.41/100 = 21.7

1/100 (19.61Y + 4841 - 48.41Y) = 21.7

19.61Y + 4841 -48.41Y = 2170

Y = 93%

269

Appendices

APPENDIX 3

Annealed pellets comprise of both crystalline and amorphous phases.

Xcryst = fraction of crystalline phase

Dcryst = diffusion coefficient in crystalline phase

Xamorphous = fraction of amorphous phase

Damorph.= diffusion coefficient in crystalline phase

D annealed = Xcryst Dcryst+ Xamorph. Damorph.

Since Dcryst in the crystalline phase = 0

D annealed = Xamorph. Damorph.

= (1- Xcryst) Damorph.

270

Bibliography

BIBLIOGRAPHY

Ackermann, P., Jägerstad, M., and Ohlsson, T., 1995, Foods and packaging

materials: chemical interactions, Royal Society of Chemistry, Cambridge,

England, pp12-22.

Arora, A.P., and Halek, J.W., 1994, Structure and cohesive energy density of

fats and their sorption by polymer films. Journal of Food Science, 29, 1325-

1327.

Ashby, R., 1988, Migration from polyethylene terephthalate under all

conditions of use. Food Additives and Contaminants, 5, Supplement no. 1,

485-492.

Ashraf-Khorassani, M., and Levy, J.M., 1990, Quantitative analysis of

polymer additives in low density polyethylene using supercritical fluid

extraction-supercritical fluid chromatography. Journal of High Resolution

Chromatography, 13, 742-747.

Ashraf-Khorassani, M., Boyer, D.S., and Levy, J.M., 1991, Optimisation of

experimental parameters for the determination of polymer additives using

online SFE-SFC. Journal of Chromatographic Science, 29, 517-521.

Atkinson, E.R., and Calouche, S.I., 1971, Analysis of polyethylene

terephthalate prepolymer by trimethylsilylation and gas chromatography.

Analytical Chemistry, 43 (3), March, 460-462.

Augusto, F., Valente, A.L.P., Tada, E., and Rivellino, S.R., 2000, Screening

of Brazilian fruit aromas using solid phase microextraction-gas

chromatography-mass spectrometry. Journal of Chromatography A, 873,

117-127.

269

Bibliography

Baner, A.L., Franz, R., and Piringer, O., 1994a, Alternative methods for the

determination and evaluation of migration potential from polymeric contact

materials. Deutsche Lebensmittel-Rundschau, 90 (5), 137-143.

Baner, A.L., Franz, R., and Piringer, O., 1994b, Alternative methods for the

determination and evaluation of migration potential from polymeric contact

materials. Deutsche Lebensmittel-Rundschau, 90 (6), 181-185.

Barnes, K.A., Damant, P.D., Startin, J.R., and Castle, C., 1995, Qualitative

liquid chromatographic-atmospheric-pressure chemical-ionisation mass

spectrometric analysis of polyethylene terephthalate oligomers. Journal of

chromatography A, 712, 191-199.

Barrionuevo, W.R., and Lanças, F.M., 2000, Solid phase microextraction of

pyrethroid pesticides from water at low and sub-ppt levels at different

temperatures. Journal of High Resolution Chromatography, 23, 485-488.

Bart, J.J.C., 2001, Direct solid sampling methods for gas chromatographic

analysis of polymer/additive formulations. Polymer Testing, 20, 729-740.

Bartle, K.D., Clifford, A.A., Hawthorne, S.B., Langenfeld, J.J., Miller, D.J.,

and Robinson, R., 1990, A model for dynamic extraction using a

supercritical fluid. The Journal of Supercritical Fluids, 3, 143-149.

Bartle, K.D., Clifford, A.A., Hawthorne, S.B., Langenfeld, J.J., Miller, D.J.,

and Robinson, R., 1990, A model for dynamic extraction using a

supercritical fluid. Journal of Supercritical Fluids, 3, 143-149.

Bartle, K.D., Boddington, T., Clifford, A.A., and Cotton, N.J., 1991,

Supercritical fluid extraction and chromatography for the determination of

oligomers in poly(ethylene terephthalate) films. Analytical Chemistry, 63,

2371-2377.

270

Bibliography

Batlle, R., Sánchez, C., and Nerin, C., 1999a, A systematic approach to

optimise solid phase microextraction. Determination of pesticides in

ethanol/water mixtures used as food simulants. Analytical Chemistry, 2417-

2422.

Batlle, R., Salafranca, J., and Nerin, C., 1999b, The use of solid phase

microextraction for the analysis of Bisphenol A and Bisphenol A Diglycidyl

Ether in food simulants. Journal of Chromatography A, 864, 137-144.

Bayer, F.L., 2002, Polyethylene terephthalate recycling for food contact

applications: testing, safety and technologies: a global perspective. Food

Additives and Contaminants, 19, Supplement, 111-134.

Begley, T.H., and Hollifield H.C., 1989, Liquid Chromatographic

determination of residual reactants and reaction by-products in polyethylene

terephthalate. Journal of the Association of Official Analytical Chemists, 72

(3), 468-470.

Begley T.H., and Hollifield H.C., 1990a, Evaluation of polyethylene

terephthalate cyclic trimer migration from microwave food packaging using

temperature-time profiles. Food Additives and Contaminants, 7 (3), 339-346.

Begley T.H., and Hollifield H.C., 1990b, High-performance liquid

chromatographic determination of migrating poly(ethylene terephthalate)

oligomers in corn oil. Journal of Agricultural Food Chemistry, 38, 145-148.

Begley T.H., Dennison, J.L., Hollifield, H.C., 1990, Migration into food of

polyethylene terephthalate (PET) cyclic oligomers from PET microwave and

susceptor packaging. Food Additives and Contaminants, 7 (6), 797-803.

Begley T.H., and Hollifield H.C., 1993, Recycled Polymers in food

packaging: migration considerations. Food technology, 47 (11), 109-112.

271

Bibliography

Begley T.H., and Hollifield H.C., 1995, Plastics, Rubber, and Paper: A

Pragmatic Approach, edited by C.P. Rader, S.D. Baldwin, D.D. Cornell,

G.D. Sadler and R.F. Stockel, ACS Symposium Series 609 (Washington,

DC: ACS), pp. 445-457.

Begley T.H., Gay, M.G., Hollifield, H.C., 1995, Determination of migrants

in and migration from nylon food packaging. Food additives and

contaminants, 12 (5), 671-676.

Begley, T.H., 1997, Methods and approaches used by FDA to evaluate the

safety of food packaging materials. Food Additives and Contaminants, 14

(6-7), 545-553.

Begley T.H., McNeal, T.P., Biles J.E., and Paquette, K.E., 2002, Evaluating

the potential for recycling all PET bottles into new food packaging. Food

Additives and Contaminants, 19, Supplement, 135-143.

Bicchi, C., Drigo, S., and Rubiolo, P., 2000, Influence of fibre coating in the

headspace solid phase microextraction-gas chromatographic analysis of

aromatic and medicinal plants. Journal of Chromatography, 892, 469-485.

Brandrup, J., and Immergut, E.H., 1989, Polymer Handbook 3rd Edition,

John Wiley and Sons, Canada, pp. 519-557.

Brody, A.L., 1989, Flavour interacts with packaging. Prepared Foods, Sept.,

158 (10), 128-131.

Buiarelli, F., Giampaolo, C., Coccioli, F., 1993, HPLC and GC-MS

detection of compounds released to mineral waters stored in plastic bottles of

PET and PVC. Annali di Chemica, 83, 93-104.

Camacho, W., and Karlsson, S., 2001, Quality determination of recycled

plastic packaging waste by identification of contaminants by GC/MS after

272

Bibliography

microwave assisted extraction. Polymer Degradation and Stability, 71, 123-

134.

Castle L., Mercer, A.J., and Gilbert, J., 1988, Gas chromatographic-mass

spectrometric determination of adipate-based polymeric plasticizers in foods.

Journal of the Association of Official Analytical Chemists, 71 (2), 394-396.

Castle, L., Mayo, A., Crews, C., and Gilbert, J., 1989, Migration of

poly(ethylene terephthalate) (PET) oligomers from PET plastics into foods

during microwave and conventional cooking and into bottled beverages.

Journal of Food Protection, 52 (5), 337-342.

Castle, L., Jickells, S.M., Gilbert, J., and Harrison, N., 1990, Migration

testing of plastics and microwave-active materials for high-temperature

food-use applications. Food Additives and Contaminants, 7 (6), 779-796.

Castle L., Price, D., and Dawkins, J.V., 1996, Oligomers in plastics

packaging. Part 1: Migration tests for vinyl chloride tetramer. Food

Additives and Contaminants, 13 (3), 307-314.

Charara, Z.N., Williams, J.W., Schmidt, R.H., and Marshall, M.R., 1992,

Orange flavour absorption into polymeric packaging materials. Journal of

food science, 57 (4), 963-966.

Chidambaram, D.,Venkatraj, R., and Manisankar, P., 2003, Solvent-induced

modifications in polyester yarns. II. Structural and thermal behaviour.

Journal of Applied Polymer Science, 89, 1555-1566.

Cho, H.-J., Baek, K., Lee, H.-H., Lee, S.-H., and Yang, J.-W., 2003,

Competitive extraction of multi-component contaminants in water by

Carboxen-polymethylsiloxane fibre during solid-phase microextraction.

Journal of Chromatography, 988, 177-184.

273

Bibliography

Cooper, D.R., and Semlyen, J.A., 1973, Equilibrium ring concentrations and

the statistical conformations of polymer chains: Part 11. Cyclics in

poly(ethylene terephthalate). Polymer, 14, May.

Costley, C.T., Dean, J.R., Newton, I., and Carroll, J., 1997, Extraction of

Oligomers from poly(ethylene terephthalate) by microwave-assisted

extraction. Analytical Communications, 34, 89-91.

Cotton N.J., Bartle, K.D., Clifford, A.A., Dowle, C.J., 1993, Rate and extent

of supercritical fluid extraction of additives from polypropylene: diffusion,

solubility, and matrix effects. Journal of Applied Polymer Science, 48, 1607-

1619.

Cross, R.F., Harding, I.H., Cass, P.J., Mudumba, R., and Kosior, E., 2001,

Evaluation of a new poly(ethylene terephthalate) recycling process for food

contact applications: Part B – migration studies, unpublished.

Daimon, H., and Hirata, Y., 1991, Directly coupled supercritical-fluid

extraction/capillary supercritical fluid chromatography of polymer additives.

Chromatographia, 32(11/12), 549-554.

Demertzis, P.G., Johansson, F., Lievens, C., and Franz, F., 1997, Studies on

the development of a quick inertness test procedure for multi-use PET

containers – Sorption behaviour of bottle wall strips. Packaging technology

and science, 10, 45-58.

Devlieghere F., De Meulenaer, B., Sekitoleko, P., Estrella Garcia, A.A., and

Huyghebaert, A., 1997, Evaluation, modelling and optimisation of the

cleaning process of contaminated plastic food refillables. Food Additives and

Contaminants, 14 (6-7), 671-683.

Devlieghere, F., De Meulenaer, B., Demyttenaere, J., and Huygherbaert, A.,

1998, Evaluation of recycled HDPE milk bottles for food applications. Food

Additives and Contaminants, 15 (3), 336-345.

274

Bibliography

Dong, M., DiEdwardo, A.H., and Zitomer, F., 1980, Determination of

residual acetaldehyde in polyethylene terephthalate bottles, preforms and

resins by automated headspace gas chromatography. Journal of

Chromatographic Science, 18, 242-246.

Dulio, V., Po, R., Borrelli, R., Guarini, A, Santini, C., 1995, Low-molecular-

weight oligomers in recycled PET. Die Angewandte Makromolekulare

Chemie, 225,109-122.

Eberhartinger, S., Steiner, I., Washuttl, J., and Kroyer, G., 1990,

Untersuchungen zur migration von acetaldehyd aus polyethylenterephthalat-

flaschen fur kohlensaurehaltige erfrischungsgetranke. Lebensmittel

Untersuchung Forschung, 191, 286-289.

Elmore, J.S., Mottram, D.S., and Hierro, E., 2000, Two-fibre solid phase

microextraction combined with gas chromatography-mass spectrometry for

the analysis of volatile aroma compounds in cooked pork. Journal of

Chromatography A, 905, 233-240.

Eskilsson, C.S., and Björklund, E., 2000, Analytical-scale microwave-

assisted extraction. Journal of Chromatography A, 902, 227-250.

Ezquerro, Ó., Pons, B., and Tena, M., T., 2002, Development of a headspace

solid-phase microextraction-gas chromatography-mass spectrometry method

for the identification of odour-causing volatile compounds in packaging

materials. Journal of Chromatography A, 963, 381-392.

Ezquerro, Ó., Pons, B., Tena, M., 2003, Direct quantitation of volatile

organic compounds in packaging materials by headspace solid-phase

microextraction-gas chromatography-mass spectrometry. Journal of

Chromatography A, 985, 247-257.

275

Bibliography

Fayoux, S.C., Seuvre, A.M., and Voilley, A.J., 1997, Aroma transfer in and

through plastic packaging: orange juice and d-limonene. A review. Part I:

Orange juice aroma sorption. Packaging Technology and Science, 10, 69-82.

FDA, 1992, Points to consider for the use of recycled plastics in food

packaging: Chemistry considerations, May 1992. US Food and Drug

Administration, Center for Food Safety and Applied Nutrition, (HFS-247),

Washington, DC.

FDA, 1995, Recommendations for chemistry data for indirect food additive

petitions. US Food and Drug Administration, Center for Food Safety and

Applied Nutrition, Washington, DC.

Feigenbaum, A.E., Ducruet, V.J., Depal, S., and Wolff, N., 1991, Food and

Packaging Interactions: Penetration of fatty food simulants into rigid

poly(vinyl chloride). Journal of Agricultural Food Chemistry, 39, 1927-

1932.

Feigenbaum, A., Riquet, A., and Ducruet, V., 1993, Safety and quality of

foodstuffs in contact with plastic materials. Journal of Chemical Education,

70 (11), 883-886.

Feigenbaum, A., Scholler, D., Bouquant, J., Brigot, G., Ferrier, D., Franz, R.,

Lillemark, L., Riquet, A.M., Petersen, J.H., Van Lierop, B., and Yagoubi,

N., 2002, Safety and quality if food contact materials. Part I: Evaluation of

analytical strategies to introduce migration testing into good manufacturing

practice. Food Additives and Contaminants, 19 (2), 184-201.

Feron, V.J., Jetton, J., De Kruijf, N., and Van Den Berg, F., 1994,

Polyethylene terephthalate bottles (PRBs): a health and safety assessment.

Food Additives and Contaminants, 11 (5), 571-594.

Fordham, P.J., Gramshaw, J.W., Crews, H.M., and Castle, L., 1995, Element

residues in food contact plastics and their migration into food simulants,

276

Bibliography

measured by inductively-coupled plasma-mass spectrometry. Food Additives

and Contaminants, 12 (5), 651-669.

Fortin, J.B., and Lu, T.M., 2001, Ultraviolet radiation induced degradation of

poly-para-xylene (parylene) thin films. Thin Solid Films, 397 (1-2), 223-228.

Franz, R., Demertzis, P. G., Johansson, F., Lievens, C., 1997, Studies on the

development of a quick inertness test procedure for multi-use PET

containers – sorption behaviour of bottle wall strips. Packaging Technology

and Science, 10, 45-58.

Franz, R., and Huber, M., 1997, Identification of migratable substances in

recycled high-density polyethylene collected from household waste. Journal

of High Resolution Chromatography, 20, 427-430.

Franz, R., Huber, M., and Welle, F., 1998, Recycling of post-consumer

poly(ethylene terephthalate) for direct food contact application – a feasibility

study using a simplified challenge test. Deutsche Lebensmittel-Rundschau,

94 (9), 303-308.

Franz, R., and Welle, F., 1999a, Post-consumer poly(ethylene terephthalate)

for direct food contact applications – final proof of food law compliance.

Deutsche Lebensmittel-Rundschau, 95 (10), 424-427.

Franz, R., and Welle, F., 1999b, Analytisches screening und bewertung von

marküblichen postconsumer-PET-recyclaten für die erneute anwendung in

lebensmittel/verpackungen. Deutsche Lebensmittel-Rundschau, 95 (3), 94-

100.

Franz, R., and Welle, F., 2002, Recycled poly(ethylene terephthalate) for

direct food contact applications: challenge test of an inline recycling process.

Food Additives and Contaminants, 19 (5), 502-511.

277

Bibliography

Freire, A., Castle, L., Reyes, F.G.R., and Damant, A.P., 1998, Thermal

stability of polyethylene terephthalate food contact materials: formulation of

volatiles from retail samples and implications for recycling. Food Additives

and Contaminants, 15 (4), 473-480.

Garde, J.A., Catalá, R., and Gavara, R., 1998, Analysis of antioxidants

extracted from polypropylene by supercritical fluid extraction. Food

Additives and Contaminants, 15 (6), 701-708.

Gavara, R., Catalá, R., and Hernández-Muñoz, P., 1997, Study of aroma

scalping through thermosealable polymers used in food packaging by inverse

chromatography. Food Additives and Contaminants, 14 (6-7), 609-616.

Goodman, I., and Nesbitt, B.F., 1960, The structures and reversible

polymerisation of cyclic oligomers from poly(ethylene terephthalate).

Journal of Polymer Science, XLVIII, 423-433.

Goydan, R., Schwope, A.D., Reid, R.C., Cramer, G., 1990, High-

temperature migration of antioxidants from polyolefins. Food Additives and

Contaminants, 7 (3), 323-337.

Gramshaw, J.W., Vandenburg, H.J., and Lakin, R.A, 1995, Identification of

potential migrants from sample of dual-ovenable plastics. Food Additives

and Contaminants, 12 (2), 211-222.

Hakkarainen, M., Albertsson, A., and Karlsson, S., 1997, Solid phase

microextraction (SPME) as an effective means to isolate degradation

products in polymers. Journal of Environmental Polymer Degradation, 5

(2), 67-73.

Hamdani, M., and Feigenbaum, A., 1996, Migration from plasticized

poly(vinyl chloride) into fatty media: importance of simulant selectivity for

the choice of volatile fatty simulants. Food Additives and Contaminants, 13

(6), 717-730.

278

Bibliography

Hamdani, M., Feigenbaum, A., and Vergnaud, J.M., 1997, Prediction of

worst-case migration from packaging to food using mathematical models.

Food Additives and Contaminants, 14 (5), 499-506.

Harding, I.H., Cross, R.F., Cass, P.J., Mudumba, R., and Kosior, E., 2001,

Evaluation of a new poly(ethylene terephthalate) recycling process for food

contact applications: Part A – decontamination efficiency, Submitted to

Food Additives and Contaminants, Unpublished.

Hernandez-Munoz, P., Catala, R., and Gavara, R., 2001, Food aroma

partition between packaging materials and fatty food simulants. Food

Additives and Contaminants, 18 (7), 673-682.

Huber, M., and Franz, R., 1997a, Identification of migratable substances in

recycled high density polyethylene collected from household waste. Journal

of High Resolution Chromatography, 20, 427-430.

Huber, M., and Franz, R., 1997b, Studies on contamination of post-consumer

plastics from controlled resources for recycling into food packaging

applications. Deutsche Lebensmittel-Rundschau, 93, 10, 328-331.

Hudgins, W.R., Theurer, K., and Mariani, T., 1978, Journal of Applied

Polymers Science: Applied Polymer Symposium, 34, 145-155.

Hunt, T.P., and Dowle, C.J., 1991, Analysis of poly(vinyl chloride) additives

by supercritical fluid extraction and supercritical fluid chromatography.

Analyst, 116, 1299-1304.

Huynh, C.K., and Vu-Duc, T., 1998, Un procédé automatique de

determination de l’acétaldéhyde dans l’eau minérale en bouteilles PET par

CPG et micro-extraction en phase solide dans l’espace de tête. Travoux de

Chimie alimentaire et d’hygiene, 89, 705-714.

279

Bibliography

Ikegami, T., Nagashima, K., Shimoda, M., Tanaka, Y., and Osajima, Y.,

1991, Sorption of volatile compounds in aqueous solution by ethylene-vinyl

alcohol copolymer films. Journal of Food Science, 56 (2), 500-503.

Imai, T., Harte, B.R., and Giaicin, J.R., 1990, Partition distribution of aroma

volatiles from orange juice into selected polymeric sealant films. Journal of

food science, 55 (1), 158-161.

Jameel, H., Waldman, J., and Rebenfield, L., 1981, The effects of orientation

and crystallinity on the solvent induced crystallisation of poly(ethylene

terephthalate). I. Sorption and diffusion-related phenomena. Journal of

Applied Polymer Science, 26, 1795-1811.

Jetten, J., de Kruijf, N., and Castle, C., 1999, Quality and safety aspects of

reusable plastic food packaging materials: A European study to underpin

future legislation. Food Additives and Contaminants, 16 (1), 25-36.

Jia, M.Y., Koziel, J., and Pawliszyn, J., 2000, Fast field sampling/sample

preparation and quantification of volatile organic compounds in indoor air

by solid-phase microextraction and portable gas chromatography. Field

Analytical Chemistry and Technology, 2-3, 73-84.

Kashtock, M., and Breder, C.V., 1980, Migration of ethylene glycol from

polyethylene terephthalate bottles into 3% acetic acid. Journal of the

association of official analytical chemists, 63 (2), 168-172.

Khabbaz, F., Albertsson, A.-C., and Karlsson, S., 1998, Trapping of volatile

low molecular weight photoproducts in inert and enhanced degradable

LDPE. Polymer Degradation and Stability, 61, 329-342.

Kim, H., Gilbert, S.G., And Johnson, J.B., 1990, Determination of potential

migrants from commercial amber polyethylene terephthalate bottle wall.

Pharmaceutical Research, 7 (2), 176-179.

280

Bibliography

Kolb, B., Pospisil, P., and Auer, M., 1981, Quantitative analysis of residual

solvents in food packaging printed films by capillary gas chromatography

with multiple headspace extraction. Journal of Chromatography. 204, 371-

376.

Komolprasert, V., Hargraves, W.A., and Armstrong, D.J., 1994,

Determination of benzene residues in recycled PETE by dynamic headspace-

gas chromatography. Food Additives and Contaminants, 11 (5), 605-614.

Komolprasert, V., and Lawson, A.R., 1995, Residual contaminants in recycled

poly(ethylene terephthalate)-Effects of washing and drying. Plastics,

Rubber, and Paper: A Pragmatic Approach, edited by C.P. Rader, S.D.

Baldwin, D.D. Cornell, G.D. Sadler and R.F. Stockel, ACS Symposium

Series 609 (Washington, DC: ACS), pp. 435-444.

Komolprasert, V., and Lawson, A.R., 1997, Considerations for reuse of PET

bottles in food packaging: migration study. Journal of Agricultural and

Food Chemistry, 45, 444-448.

Komolprasert, V., Lawson, A.R., and Begley, T.H., 1997, Migration of

residual contaminants from secondary recycled poly(ethylene terephthalate)

into food-simulating solvents, aqueous, ethanol, and heptane. Food

Additives and Contaminants, 14 (5), 491-498.

Komolprasert, V., McNeal, T.P., Agrawal, A., Adhikari, C., and Thayer,

D.W., 2001, Volatile and non-volatile compounds in irradiate semi-rigid

crystalline poly(ethylene terephthalate). Food Additives and Contaminants,

18 (1), 89-101.

Konczal, J.B., Harte, B.R., Hoojjat, P., and Giacin, J.R., 1992, Apple juice

flavour compound sorption by sealant films. Journal of Food Science, 57

(4), 967-966 and 972.

281

Bibliography

Kosmidis, V.A., Achilias, D.S., and Karayannidis, G.P., 2001, Poly(ethylene

terephthalate) recycling and recovery of pure terephthalic acid. Kinetics of a

phase transfer catalised alkaline hydrolysis. Macromolecular Material

Engineering, 286, 640-647.

Kuznesof, P.M., and Van Derveer, M.C., 1995, Recycled plastics for food-

contact applications. Plastics, Rubber, and Paper Recycling: a Pragmatic

Approach, edited by C.P. Rader, S.D. Baldwin, D.D. Cornell, G.D. Sadler

and R.E. Stockel. AcS Symposium Series No, 609 (Washington, DC:

American Chemical Society), pp. 389-403.

Kwapong, O.Y., and Hotchkiss, J.H., 1987, Comparative sorption of aroma

compounds by polyethylene and ionomer food-contact plastics. Journal of

food science, 52 (3), 761-763, 785.

Lau, O., and Wong, S., 1994, Naphthalene contamination of sterilised milk

drinks contained in low-density polyethylene bottles. Part 1. Analyst, 119,

1037-1042.

Lau, O., and Wong, S., 1995, Naphthalene contamination of sterilised milk

drinks contained in low-density polyethylene bottles. Part 2. Effect of

naphthalene vapour in air. Analyst, 120, 1125-1129.

Lau, O.-W., and Wong, S.-K., 2000, Contamination in food from packaging

material. Journal of Chromatography A, 882, 255-270.

Lickly, T.D., Bell, C.D., and Lehr, K.M., 1990, The migration of Irganox

1010 antioxidant from high-density polyethylene and polypropylene into a

series of potential fatty-food simulants. Food Additives and Contaminants, 7

(6), 805-814.

Limm, W., and Hollifield, H. C., 1996, Modelling of additive diffusion in

polyolefins. Food Additives and Contaminants, 13 (8), 949-967.

282

Bibliography

Liu, C.-P.A., and Neogi, P.,1992, Sorption of methylene chloride in

semicrystalline polyethylene terephthalate. Journal of Macromolecular

Science – Physics, B31 (3), 265-279.

Llompart, M., Li, K., and Fingas, M., 1999, Headspace solid-phase

microextraction for the determination of polychlorinated biphenyls in soils

and sediments. Journal of Microcolumn Separations, 11 (6), 397-402.

Lorusso, S., Gramiccioni, L., Di Marzio, S., Milana, M.R., Di Prospero, P.,

and Papetta, A., 1985, Acetaldehyde migration from poly(ethylene

terephthalate) (PET) containers. GC determination and toxicological

assessment. Annali di Chimica, 75, 403-414.

Lou, X., Janssen, H., and Cramers, C.A., 1996, Effects of modifier addition

and temperature variation in SFE of polymeric materials. Journal of

Chromatographic Science, 34, 282-290.

Lou, X., Janssen, H., and Cramers, C.A., 1997, Parameters affecting the

accelerated solvent extraction of polymeric samples. Analytical Chemistry,

69 (8), 1598-1603.

Mannheim, C.H., Miltz, J., and Letzter, A., 1987, Interaction between

polyethylene laminated cartons and aseptically packed citrus juices. Journal

of Food Science, 52 (3), 737-740.

McDowell, C.C., Freeman, B.D., McNeely, G.W., Haider, M.I., and Hill

A.J., 1998, Synthesis, physical characterization, and acetone sorption

kinetics in random copolymers of poly(ethylene terephthalate) and

poly(ethylene 2,6-naphthalate). Journal of Polymer Science, Polymer

Physics Ed., 36 (16), 2981-3000.

McEvoy, J.P., Armstrong, C.G., and Crawford, R.J., 1998, Simulation of the

stretch blow molding process of PET bottles. Advances in Polymer

Technology, 17 (4), 339-352.

283

Bibliography

Mc Neal, T.P, and Hollifield, T.C., 1993, Determination of volatile

chemicals released from microwave-heat-susceptor food packaging. Journal

of AOAC International, 76 (6), 1268-1275.

Miltz, J., Ram, A., and Nir, M.M., 1997, Prospects for application of post-

consumer used plastics in food packaging. Food Additives and

Contaminants, 14 (6-7), 649-659.

Miltz, J., 1998, Approaches to deal with the issue of plastic packages and the

environmental. Macromolecular Symposium, 135, 265-275.

Monarca, S., Fusco, R., Biscardi, D., Feo, V., Pasquini, R., Fatigoni, C.,

Moretti, M., and Zanardini, A., 1997, Studies of migration of potentially

genotoxic compounds into water stored in PET bottles. Food Chemistry and

Toxicology, 783-788.

Monteiro, M., Nerin, C., and Reyes, F.G.R., 1996, Determination of UV

stabilizers in PET bottles by high performance-size exclusion

chromatography. Food Additives and Contaminants, 13 (5), 575-586.

Monteiro, M., Nerin, C., Rubio, C., and Reyes, F.G.R., 1998, A GC/MS

method for determining UV stabilizers in polyethyleneterephthalate bottles.

Journal of High Resolution Chromatography, 21, May, 317-320.

Moore, W.R., and Sheldon, R.P., 1961, The crystallisation of polyethylene

terephthalate by organic liquids. Polymer, 2, 315-321.

Morelli-Cardoso, M.H.W., Tabak, D., Cardoso, J.N., Pereira, A.S., 1997,

Application of capillary gas chromatography to the determination of

ethylene glycol migration from PET bottles in Brazil. Journal of High

Resolution Chromatography, 20, March, 183-185.

284

Bibliography

Munteanu, D., Isfan, A., Isfan, C., Tincul, I., 1987, "High-performance

liquid chromatographic separation of polyolefin antioxidants and light-

stabilisers", Chromatographia, 23 (1), 7-14.

Murray, R. A., 2001, Limitations to the use of solid-phase microextraction

for quantitation of mixtures of volatile organic sulfur compounds. Analytical

Chemistry. 73 (7), 1646-1649.

Nerín, C., Salafranca, J., Rubio, C., and Cacho, J., 1994, Separation of

polymer and on-line determination of several antioxidants and UV-

stabilizers by coupling size-exclusion and normal-phase high-performance

liquid chromatography columns. Journal of Chromatography A, 690, 230-

236.

Nerín, C., Salafranca, J., Rubio, C., and Cacho, J., 1998a, Multicomponent

recycled plastics: considerations about their use in food contact applications.

Food Additives and Contaminants, 15 (7), 842-854.

Nerín, C., Tornes, A.R., Batlle, R., and Cacho, J., 1998b, Pesticides in

recycled plastics. A new environmental challenge. Quimica Analitica, 17

(4), 177-184.

Nerín, C., Asensio, E., Fernández, C., Batlle, R., 2000, Supercritical fluid

extraction of additives and degradation products from both virgin and

recycled PET. Química Analítica, 19, 205-212.

Nerín, C., Phio, M.R., Salafranca, J., and Castle, L., 2002, SPME-HPLC as a

tool for the determination of migrants from food packaging materials in

aqueous food simulants and real food. Journal of Chromatography A, 963

(1-2), 375-380.

Nerin, C., Albinana, J., Philo, M.R., Castle, L., Raffael, B., and Simoneau,

S., 2003a, Evaluation of some screening methods for the analysis of

285

Bibliography

contaminants in recycled polyethylene terephthalate flakes. Food Additives

and Contaminants, 20 (7), 668-677.

Nerin, C., Fernandez, C., Domeno, C., Salafranca, J., 2003b, Determination

of potential migrants in polycarbonate containers used for microwave ovens

by high-performance liquid chromatography with ultraviolet and

fluorescence detection. Journal of Agricultural and food chemistry, 51,

5647-5653.

Nicholson, J.W., 1991, The chemistry of Polymers, The Royal Society of

Chemistry, Cambridge, pp75-89.

Nielsen, T.J., 1991, Extraction and quantitation of polyolefin additives.

Journal of Liquid Chromatography, 14 (3), 503-519.

Nielsen, T.J., Margaretha Jägerstad, I., Öste, R.E., and Sivik, B.T.G., 1991,

Supercritical fluid extraction coupled with gas chromatography for the

analysis of aroma compounds absorbed by low-density polyethylene.

Journal of Agricultural Chemistry, 39, 1234-1237.

Nielsen, T.J., Margaretha Jägerstad, I., Öste, R.E., and Wesslen, B.O., 1992,

Comparative absorption of low molecular aroma compounds into commonly

used food packaging polymer films. Journal of Food Science, 57 (2), 490-

492.

Nielsen, T.J., 1994, Limonene and Myrcene sorption into refillable

polyethylene terephthalate bottles and washing effects on removal of sorbed

compounds. Journal of Food Science, 59 (1), 227-230.

Nielson, T.J., Damant, A.P., and Castle, L., 1997, Validation studies of a

quick test for predicting the sorption and washing properties of refillable

plastic bottles. Food Additives and Contaminants, 14 (6-7), 685-693.

286

Bibliography

Nir, M.M., Ram, A., and Miltz, J., 1996, Sorption and migration of organic

liquids in poly(ethylene terephthalate). Polymer engineering and science, 36

(6), 862-868.

Ouyang, H., Chen, C., Sanboh, L., and Yang, H., 1998, Acetone transport in

poly(ethylne terephthalate) and related phenomena. Journal of Polymer

Science: Part B: Polymer Physics, 36, 163-169.

Paik, J.S., 1992, Comparison of sorption in orange flavour components by

packaging films using the headspace technique. Journal of Agricultural

Food Chemistry, 40, 1822-1825.

Patton,C., J., Felder, R.M., and Koros, W.J., 1984, Sorption and transport of

benzene in poly(ethylene terephthalate). Journal of Applied Polymer

Science, 29, 1095-1110.

Peebles, L.H., Huffman, M.W., and Ablett, C.T., 1969, Isolation and

identification of the linear and cyclic oligomers of poly(ethylene

teephthalate) and the mechanism of cyclic oligomer formation. Journal of

Polymer Science, Part A-1, 7, 479-496.

Penton, Z., SPME Application Note Number 17, Varian, Walnut-Creek,

USA, 1999, Determination of residual solvents and monomers in polymers

with solid phase microextraction (SPME) and GC/MS.

Pérès, C., Viallon, c., and Berdagué, J.-L., 2001, Solid-phase

microextraction-mass spectrometry: A new approach to the rapid

characterisation of cheeses. Analytical Chemistry. 73, 1030-1036.

Perlstein, P., 1983, The determination of light stabilisers in plastics by high

performance liquid chromatography. Analytica Chimica Acta, 21-27.

287

Bibliography

Pochivalov, K.V., Pronin, A.M., Mizerovskii, L.N., and Kozlov, S.N., 2000,

Effect of ultrasound on the kinetics of extraction of the polymer component

from a porous composite material. Fibre Chemistry, 32 (4), 298-302.

Pó, R., Occhiello, E., Giannotta, G., Pelosini, L., and Abis, L., 1995, New

polymeric materials for containers manufacture based on PET/PEN

copolyesters and blends. Polymers for Advanced Technologies, 7, 365-373.

Reynier, A., Dole, P., and Feigenbaum, A., 2001, Additive diffusion

coefficients in polyolefins II. Effect of swelling and temperature on the

D=f(M) correlation. Journal of Applied polymer Science, 82, 2434-2443.

Riquet, A.M., Sandray, V., Akermann, O., and, Feigenbaum, A., 1991, ESR

study of the evolution of rigid PVC: Influence of heat and contact with fatty

or aqueous simulating liquids. Sciences Des Aliments, 11, 341-359.

Riquet, A.M., and Feigenbaum, A., 1997, Food and packaging interactions:

tailoring fatty food simulants. Food Additives and Contaminants, 14 (1), 53-

63.

Sadler, G., and Braddock, R.J., 1991, Absorption of citrus flavour volatiles

by low-density polyethylene. Journal of Food Science, 56 (1), 35-37.

Sadler, G., Pierce, D., Lawson, A., Suvannunt, D., and Senthil, V., 1996,

Evaluating organic compound migration in poly(ethylene terephthalate): a

simple test with implications for polymer recycling. Food Additives and

Contaminants, 13 (8), 979-989.

Salafranca , J., Cacho, J., and Nerin C., 1999, Considerations about the

supercritical fluid extraction (SFE) feasibility on the determination of

antioxidants and UV-stabilizers in virgin and recycled polyolefins. Journal

of High Resolution Chromatography, 22 (10), 553-558.

288

Bibliography

Sauvant, M.P., Pepin, D., and Bohatier, J., 1995, Chemical and in vitro

toxicological evaluations of water packaged in polyvinyl chloride and

polyethylene terephthalate bottles. Food Additives and Contaminants, 12 (4),

567-584.

Scheirs, J., 1998, Polymer Recycling, John Wiley and Sons Ltd., West

Sussex, England, pp. 121-182.

Schumann, H.-D., and Thiele, U.K., 1996, Polyester producing plants:

Principals and technology. Verlag moderne industrie, 5-61.

Semenov, S. N., Koziel, J.A., and Pawliszyn, J., 2000, Kinetics of solid-

phase extraction and solid-phase microextraction in thin adsorption layer

with saturation sorption isotherm. Journal of Chromatography A. 873, 39-

51.

Shimoda, M., Ikegami, T., and Yutaka, O., 1988, Sorption of flavour

compounds in aqueous solution into polyethylene film. Journal of Food and

Agriculture, 42, 157-163.

Shiono, S., 1979, Determination of poly(ethylene terephthalate) oligomers in

refrigeration oils by absorption column chromatography-gel permeation

chromatography. Analytical Chemistry, 51 (14), December 1979.

Shirey, R.E., 2000, Optimisation of extraction conditions for low-molecular-

weight analytes using solid phase microextraction. Journal of

Chromatographic Science. 38, 109-116.

Snyder, R., C., and Breder, C., V., 1985, New FDA migration cell used to

study migration of styrene from polystyrene into various solvents. Journal of

the Association of Official Analytical Chemistry, 68 (4), 1985.

289

Bibliography

Spell, H.L., and Eddy, R.D., 1960, Determination of additives in

polyethylene by absorption spectroscopy. Analytical Chemistry, 32 (13),

1811-1814.

Startin, J.R., Parker, I., Sharman, M., and Gilbert, J., 1987, Analysis of di-(2-

ethylhexyl)adipate plasticiser in foods by stable isotope dilution gas

chromatography-mass spectrometry. Journal of Chromatography, 387, 509-

514.

St. Küppers, 1992, The use of temperature variation in supercritical fluid

extraction of polymers for the selective extraction of low molecular weight

components from poly(ethylene terephthalate). Chromatographia, 33

(9/10), 434-440.

Tavss E.A., Santalucia, J., Robinson, R.S., and Carroll, D.L., 1988, Analysis

of flavour absorption into plastic packaging materials using multiple

headspace extraction gas chromatography. Journal of Chromatography, 438,

281-289.

Tawfik M.S., Devlieghere, F., and Huyghebaert, A., 1998, Influence of D-

limonene absorption on the physical properties of refillable PET. Food

Chemistry, 61 (12), 157-162.

Tawfik, M.S., and Huyghebaert, A., 1988, Polystyrene cups and containers:

styrene migration. Food Additives and Contaminants, 15 (5), 592-599.

Tice, P.A., and McGuinness, J.D., 1987, Migration from food contact

plastics. Part I. Establishment and aims of the PIRA project. Food Additives

and Contaminants, 4 (3), 267-276.

Tombesi, N.B., and Hugo, F., 2002, Application of solid-phase

microextraction combined with gas chromatography-mass spectroscopy to

the determination of butylated hydroxytoluene in bottled drinking water.

Journal of Chromatography. 963, 179-183.

290

Bibliography

Tuduri, L., Desauziers, V., and Franlo, J. L., 2001, Potential of solid-phase

microextraction fibres for the analysis of volatile organic compounds in air.

Journal of Chromatographic Science. 39, 521-529.

Triantafyllou, V.I., Karamani, A.G., Akrida-Demertzi, K., and Demertzis,

P.G., 2002, Studies on the usability of recycled PET for food packaging

applications. European Food Research Technology, 215, 243-248.

Ulrich, K.T.,1993, Introduction to industrial polymers, 2nd ed., Munich ;

New York: Hanser Publishers.

US EPA METHOD 8000B, 1996, revision 2. Determinative Chromatographic

Separations, section 7.5.2, Washington, DC, p.21.

Valor, I., Pérez, M., Cortada, C., Apraiz, D., Moltó, J.C., and Font, G., 2001,

SPME of 52 pesticides and polychlorinated biphenyls: Extraction

efficiencies of the SPME coatings poly(dimethylsiloxane), polyacrylate,

poly(dimethylsiloxane)-divinylbenzene, carboxen-poly(dimethylsiloxane),

and carbowax-divinylbenzene. Journal of Separation Science. 24, 39-48.

Vandenburg, H.J., Clifford, A.A., Bartle, K.D., Carroll, J., Newton, I.,

Garden, L.M., Dean, J.R., and Costley, C.T., 1997, Analytical extraction of

additives from polymers. Analyst, 122 (101R-115R).

Vandenburg, H.J., Clifford, A.A., Bartle, K.D., Carlson, R.E., Carroll, J., and

Newton, I.D., 1999, A simple solvent selection method for accelerated

solvent extraction of additives from polymers. Analyst, 124, 1707-1710.

Van Lierop, J.B.H., 1997, Enforcement of food packaging legislation. Food

Additives and Contaminants, 14 (6-7), 555-560.

Van Willige, R.W.G., Lissen, J.P.H., Meinders, M.B.J., Van Der Stege, H.J.,

and Voragenn, A.G.J., 2002, Influence of flavour absorption on oxygen

291

Bibliography

permeation through LDPE, PP, PC and PET plastics food packaging. Food

Additives and Contaminants,19 (3), 303-313.

Vijayalakshmi, N.S, Baldev, R., Ravi, P., and Mahadeviah, M.,1999, Effect

of time and temperature on the overall migration of additives from plastics

into food simulants. Deutsche Lebensmittel-Rundschau, 1, 22-25.

Villberg, K., and Veijanen, A., 2001, Analysis of a GC/MS thermal

desorption system with simultaneous sniffing for determination of off-odour

compounds and VOCs in fumes formed during extrusion coating of low-

density polyethylene. Analytical Chemistry. 73 (5), 971-977.

Weast, R.C., and Melvin, J.A., 1979, CRC Handbook of Chemistry and

Physics, The Chemical Rubber Company, C732-735.

Wim, A.M., and Swarin, S.J., 1975, Determination of antioxidants in

polypropylene by liquid chromatography. Journal of Applied Polymer

Science, 19, 1243-1256.

Wyatt, D.M., 1983, Semi-automation of headspace GC as Applied to

determination of acetaldehyde in polyethylene terephthalate beverage

bottles. Journal of Chromatographic Science, 21, November.

292


Recommended