.P . I 153 1.. . . gag .35; 2! n ‘. .. Fr! 1.55%“. .wr « 1mm «maymuvmwuwmw . m. swan.” d5: v.1 3 1-,, :5? ~ . . 4 .. A 1. In: If an}... . SS“ 1:... :33 1.. 3.3.6.2.. This is to certify that the dissertation entitled Correlation of Analyses of Odor Profiles of HDPE Films Coated with Different Adhesives Using Electronic Nose, Sensory Evaluation, and GC-MS presented by LI XIONG .LlBRARY Michigan State Umversrty has been accepted towards fulfillment of the requirements for the PhD. degree in Packaging Jma, Jfl Major Professor’s Signature Date MSU is an Affirmative Action/Equal Opportunity Institution PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 6/07 p:IClRC/Date0ue,indd-p,1 CORRELATION OF ANALYSES OF ODOR PROFILES OF HDPE FILMS COATED WITH DIFFERENT ADHESIVES USING ELECTRONIC NOSE, SENSORY EVALUATION, AND GC-MS By Li Xiong A DISSERTATION Subnfifledto Michigan State University in Partial fulfillment of the requirements for the degree of DOCTOR OF- PHILOSOPHY SCHOOL OF PACKAGING 2006 ABSTRACT CORRELATION OF ANALYSES OF ODOR PROFILES OF HDPE FILMS COATED WITH DIFFERENT ADHESIVES USING ELECTRONIC NOSE. SENSORY EVALUATION, AND GC-MS By Li Xiong The two dimensional PCA (Principle Component Analysis) module of the E-nose system differentiated successfully five groups of HDPE films (one uncoated film and four films coated with different adhesives) and a DI (Discrimination Index) of 82 was obtained. A DFA (Discriminative Function Analysis) model with a POR (Percentage of Recognition) of 94 was built and proved effective in identifying unknown samples as one of the training samples. Both the painrvise ranking test and the quantitative affective consumer test indicated the control sample (HDPE base film) had the weakest odor while one type of coated film had the strongest odor among the five groups of samples. In addition, there was no significant difference in terms of the intensity of the perceived odor among the other three types of coated films. The “Acceptability” and “Intensity" sensory scores of the odor profiles were also obtained. Two different sample preparation techniques, DH-TD (Dynamic Headspace - Thermal Desorption) and SPME (Solid Phase Microextraction), were used to concentrate volatile compounds released from the HDPE film samples before the compounds were analyzed by GC-MS (Gas Chromatography — Mass Spectrometry). The DH-TD technique seemed to collect more volatile compounds from the sample matrix but its repeatability was affected by the connection between the transfer line of the thermal desorption unit and the gas chromatograph. In comparison, the SPME technique proved to be easier, faster, and Cheaper to use. The same groups of potential volatile compounds were identified in the odor profiles of all five HDPE films, including ketones, aldehydes, aromatics, and hydrocarbons, among which were acetone, hexanal, benzaldehyde, nonanal, and 1,3—di-tert-butylbenzene. An SGE ODO II olfactory detector was used in the sniffing tests to describe the odor profiles of two coated HDPE films. Sniffing tests proved 1,3-di-tert-butylbenzene was not the crucial compound responsible for the stronger odor of one of the coated films. PLS (Partial Least Square) models with correlation coefficients of 0.92 - 0.99 in the E-nose system were proved effective and robust in predicting the “Acceptability” sensory scores when the models were built for pairs of samples. On the other hand, the effectiveness and robustness of PLS models for predicting the amounts of acetone and nonanal in the odor profiles were affected by which pair of samples and what volatile compound were being investigated. Nevertheless, the E-nose system showed its potential as a quality control tool in packaging systems if the correlation between the E-nose system, sensory evaluation, and GC/MS was established. Moreover, the study Showed the importance of complementing the three analytical techniques with one another. Copyright by Li Xiong 2006 Iii iii-g To Those Whom I love, Who Love and Sacrifice for Me ACKNOWLEDGMENTS I would like to thank the School of Packaging for the six-year journey, during which I Ieamed, developed, and eventually come to this point of opening another new page of my life. No words can express my gratitude to my advisor, Dr. Susan Selke. Her knowledge and guidance helped me tremendously throughout my study. Her kindness, trust, patience, and support to me have undeniably made an un- washable mark in my life. I want to thank my other committee members, Dr. Bruce Harte, Dr. Janice Harte, Dr. Theron Downes, and Dr. Randolph Beaudry for their insightful advice. If I would want to become an educator one day, I would like to be like them, who showed great devotion to their profession, care for their students, and respectful principles of life and work. I want to take this opportunity to say “thank you” and “I love you" to my family, especially my mom, for their unconditional love and sacrifice for me. Without their support, it would not have been possible for me to accomplish what I did. Life is full of ups and downs. Looking back to the past six years, I am proud of myself. In the meantime, I feel grateful for everything, good and happy ones, as well as unfortunate ones. There are so many people who have helped me and whose name I cannot list here, but their kindness will be with me forever. VI TABLE OF CONTENTS LIST OF TABLES ........................................................................... XII LIST OF FIGURES ........................................................................ XVI CHAPTER 1 INTRODUCTION ...................................................... 1 BACKGROUND INFORMATION .................................................................................... 1 SIGNIFICANCE OF WORK .......................................................................................... 6 OBJECTIVES ............................................................................................................ 7 CHAPTER 2 LITERATURE REVIEW ............................................ 9 ADHESIVES ............................................................................................................. 9 Pressure-Sensitive Adhesives ....................................................................... 12 Formulations of Pressure-Sensitive Adhesives ........................................ 12 Forms of Pressure-Sensitive Adhesives .................................................. 15 Solvent-Based Adhesives ................................................................... 15 Water-Based Adhesives ..................................................................... 16 Hot-Melt Adhesives ............................................................................. 17 Off-odor from Packaging Systems Containing Adhesives ............................. 19 ANALYSIS OF OFF-ODOR IN PACKAGING SYSTEMS ................................................... 22 Challenges ..................................................................................................... 22 Sample Preparation ....................................................................................... 29 Dynamic Headspace and Thermal Desorption ......................................... 29 Theory of DH-TD ................................................................................ 30 Applications of DH-TD in Packaging Analysis ..................................... 38 Solid-Phase Microextraction (SPME) ....................................................... 40 Theory of SPME ................................................................................. 41 Thermodynamics of Direct and Headspace SPME ............................. 46 Kinetics of Direct and Headspace SPME ............................................ 48 Parameters in SPME .......................................................................... 50 Applications of SPME in Packaging Analysis ...................................... 54 SENSORY EVALUATION .......................................................................................... 57 Human Senses .............................................................................................. 58 Types of Sensory Evaluation ......................................................................... 60 Discriminative Tests ................................................................................. 60 Descriptive Tests ...................................................................................... 62 Affective Tests .......................................................................................... 63 Sensory Evaluation in Packaging Applications .............................................. 64 VII ELECTRONIC NOSE ................................................................................................ 67 Electronic Nose and Human Nose ................................................................. 68 Electronic Sensors ......................................................................................... 70 Metal Oxide Semiconductor Sensors (M.O.S. Sensors) .......................... 72 Conducting Polymer Sensors ................................................................... 73 Quartz Crystal Microbalance Sensors (Q.C.M. Sensors) ......................... 75 Pattern Recognition Systems ........................................................................ 76 Principle Component Analysis (PCA) ....................................................... 77 Discriminative Function Analysis (DFA) ................................................... 79 Partial Least Squares (PLS) ..................................................................... 79 Good-Bad Analysis .................................................................................. 81 APPLICATIONS OF THE ELECTRONIC NOSE SYSTEM ........................... 81 ELECTRONIC NOSE AND OTHER ANALYSIS TECHNIQUES ................... 83 Electronic Nose and Sensory Evaluation ................................................. 83 Electronic Nose and GC-MS .................................................................... 85 CHAPTER 3 ANALYSIS OF HEADSPACES OF HDPE FILMS USING ELECTRONIC NOSE .......................................................... 88 INTRODUCTION AND OBJECTIVES ............................................................................ 88 MATERIALS AND METHODOLOGY ............................................................................. 90 RESULTS AND DISCUSSION .................................................................................... 95 Parameter Optimization for the E-nose System ............................................. 95 Justification and Criteria ........................................................................... 95 Initial Settings of Acquisition Parameters ................................................. 96 Effect of Sample Size ............................................................................... 98 Effect of Incubation Time ........................................................................ 100 Sensor Response Pattern of Control Sample ......................................... 104 PCA Based on the E-nose Method .............................................................. 106 Application of Principle Component Analysis (PCA) .............................. 109 Discriminative Function Analysis (DFA) ....................................................... 112 CHAPTER 4 EVALUATIONS OF ODOR PROFILES OF HDPE FILMS WITH SENSORY EVALUATION TESTS ........................... 117 INTRODUCTION AND OBJECTIVES .......................................................................... 1 17 MATERIALS AND METHODOLOGY ........................................................................... 118 Sample Preparation ..................................................................................... 118 Pairwise Ranking Test ................................................................................. 119 Quantitative Affective Consumer Test ........... ' .............................................. 121 RESULTS AND DISCUSSION .................................................................................. 123 Painrvise Ranking Test ................................................................................. 123 Quantitative Affective Consumer Test ......................................................... 127 VIII ——in Justification ............................................................................................ 127 Mixed Model and Validation of Residuals Assumption ........................... 129 Paired comparison of treatment means (acceptability and intensity scores) ............................................................................................................... 131 Effectiveness and Efficiency of RCBD ................................................... 137 CHAPTER 5 ANALYSIS OF VOLATILES USING DH-TD COUPLED WITH GC-MS .............................................................. 139 INTRODUCTION AND OBJECTIVES .......................................................................... 139 MATERIALS AND METHODOLOGY ........................................................................... 141 DIP (Direct Insertion Probe) Analysis .......................................................... 141 Selection of Thermal Desorption Tube ........................................................ 142 DH-TD GC/MS ............................................................................................. 144 RESULTS AND DISCUSSION .................................................................................. 148 DIP (Direct Insertion Probe) Analysis .......................................................... 148 Selection of Thermal Desorption Tube ........................................................ 157 DH-TD GCD (GC/MS) Analysis ................................................................... 165 Repeatability of DH-TD/GCD Analyses .................................................. 166 Potential Identities of Volatile Compounds ............................................. 173 CHAPTER 6 ANALYSIS OF VOLATILES USING SPME COUPLED WITH GC-MS .............................................................. 175 INTRODUCTION AND OBJECTIVES .......................................................................... 175 MATERIALS AND METHODS ................................................................................... 176 Optimization of SPME Parameters .............................................................. 177 Comparisons of Gas Chromatogram Profiles .............................................. 179 Description of Odor Profile with Sniffing Test .............................................. 180 RESULTS AND DISCUSSION ................................................................................... 183 Optimization of SPME Parameters .............................................................. 183 Effect of SPME Sampling Time .............................................................. 184 Effect of SPME Sampling Temperature ................................................. 187 Comparisons of Gas Chromatogram Profiles .............................................. 190 Confirmation of 1,3-Di-tert-Butylbenzene ............................................... 190 Origin of 1,3-Di-tert-Butylbenzene .......................................................... 192 Potential Identities of Volatile Compounds ............................................. 193 Description of Odor Profile with Sniffing Test .............................................. 196 Comparison of Two Retention Times ..................................................... 196 Descriptions of Odor Profiles .................................................................. 199 CHAPTER 7 CORRELATION OF ANALYSES OF E-NOSE, SENSORY EVALUATION, AND Gc-Ms ...................................... 203 IX INTRODUCTION AND OBJECTIVES ........................................................................... 203 E-NOSE AND SENSORY EVALUATION ..................................................................... 203 Instrument Sensitivity versus Human Nose Sensitivity ................................ 203 Subjective versus Objective Judgment ........................................................ 204 Partial Least Squares Using Sensory Scores .............................................. 207 PLS Based on Five Groups of Samples ................................................. 207 PLS Based on Three Groups of Samples .............................................. 210 PLS Based on Pairs of Samples ............................................................ 212 E-NOSE AND GC-MS ANALYSIS ............................................................................ 216 PLS Based on Five Groups of Samples ................................................. 217 PLS Based on Three Groups of Samples .............................................. 218 PLS Based on Pairs of Samples ............................................................ 219 PLS Based on Transformed Data .......................................................... 221 CHAPTER 8 SUMMARY AND CONCLUSIONS ........................ 226 ELECTRONIC NOSE .............................................................................................. 226 SENSORY EVALUATIONS ...................................................................................... 227 GC-MS ANALYSIS ............................................................................................... 228 CORRELATION OF E-NOSE, SENSORY EVALUATION AND GC-MS ANALYSIS .............. 230 RECOMMENDATIONS FOR FUTURE WORK .............................................................. 232 APPENDICES ............................................................................... 234 APPENDIX 1 WORKSHEET USED IN THE PAIRWISE RANKING TEST ............................. 234 APPENDIX 2 WORKSHEET AND DATA IN THE QUANTITATIVE AFFECTIVE CONSUMER TEST .......................................................................................................................... 235 APPENDIX 3 SCORESHEET USED IN THE PAIRWISE RANKING TEST ............................ 242 APPENDIX 4 SCORESHEET USED IN THE QUANTITATIVE AFFECTIVE CONSUMER TEST243 APPENDIX 5 SAS PROGRAM AND ITS OUTPUT ........................................................ 244 APPENDIX 6 IIC CHROMATOGRAPH OF SAMPLE CONT IN DIP ANALYSIS (2ND SET) ....... 257 APPENDIX 7 IIC CHROMATOGRAPH OF SAMPLE PCTA IN DIP ANALYSIS (2ND SET) ...... 257 APPENDIX 8 TIC CHROMATOGRAMS OF ODOR PROFILES FROM HDPE FILMS ANALYZED WITH SPME/GC-MS ........................................................................................... 258 X APPENDIX 9 DESCRIPTION OF ODOR PROFILES USING GC/MS WITH ODO II SNIFFING PORT ................................................................................................................. 263 APPENDIX 10 PLS MODELS BASED ON PAIRS OF SAMPLES AND THEIR VALIDATION DATA .......................................................................................................................... 266 APPENDIX 11 RESPONSE AREAS OF ACETONE AND NONANAL BASED ON THE DATA FROM SPME/GC-MS ANALYSIS .................................................................................... 269 APPENDIX 12 PLS MODELS BASED ON TRANSFORMED DATA OF PAIRS OF SAMPLES...273 BIBLIOGRAPHY..... ....................... . ............................................. . 275 XI LIST OF TABLES Table 2.1 Chronological developments of adhesives in the US. ........................ 11 Table 2.2 Comparison of acrylic— and rubber/resin-based pressure-sensitive adhesives ............................................................................................................ 14 Table 2.3 Elastomers in common use ................................................................. 15 Table 2.4 Advantages and disadvantages of pressure-sensitive adhesives ....... 19 Table 2.5 Taste thresholds of halophenols and haloanisoles ............................. 23 Table 2.6 The effect of medium on the thresholds of several aromatic compoung: Table 2.7 Different odor thresholds for hexanal in water ..................................... 25 Table 2.8 Effect of concentration on the taste description of trans-2-nonenal in water ................................................................................................................... 26 Table 2.9 Odor/flavor description of various compounds .................................... 28 Table 2.10 SPME compared with other sample preparation techniques ............. 41 Table 2.11 Various SPME fiber coatings and coating thickness and their recommended applications ................................................................................. 43 Table 2.12 Comparison of kinetics and applications of direct and headspace SPME .................................................................................................................. 50 Table 2.13 The human senses and their perceptions ......................................... 59 Table 2.14 Various test methods used in sensory evaluation ............................. 60 Table 2.15 Comparison of different methods in the discriminative sensory test .61 Table 2.16 Characteristics of various affective sensory tests ............................. 64 Table 2.17 Various electronic nose systems and their sensor technologies ....... 68 Table 2.18 Comparison of detection threshold values of several Chemical compounds in water determined by an electronic nose system (Alpha MOS Fox 3000) and human nose ....................................................................................... 70 7"able 2.19 Comparison of different sensors used in electronic nose systems....74 XII Table 3.1 Codes and formulas of adhesives and base film ................................. 90 ‘ Table 3.2 Sensor array used in Alpha MOS Fox 3000 E-nose system ............... 92 Table 3.3 MOS sensors and their specificity to organic compounds ................... 93 Table 3.4 Initial settings of the data acquisition parameters for the E-nose systesgrtt3 Table 3.5 Modification to the data acquisition parameters for the E-nose ........... 98 Table 3.6 Finalized settings of data acquisition for the E-nose system ............. 106 Table 3.7 Reproducibility of sensors and sample groups .................................. 107 Table 3.8 Euclidean distances between each two groups of HDPE films ......... 108 Table 3.9 Validation of "unknown" samples and their projected identities ........ 116 Table 4.1 Sample codes and pair codes used in the pairwise ranking test ....... 120 Table 4.2 Sample codes used in the quantitative affective consumer test ........ 121 Table 4.3 Worksheet used in the quantitative affective consumer test ............. 122 Table 4.4 Data analysis of the pairwise ranking test to rank the intensity of perceived odor of five HDPE films .................................................................... 123 Table 4.5 Rank sum of the tested samples in the pairwise ranking test ............ 124 Table 4.6 ANOVA analyses of the acceptability and intensity scores ............... 132 Table 4.7 Paired comparisons of the acceptability scores and the intensity scores of the sensed odors of the five HDPE films ....................................................... 134 Table 5.1 DH-TD and GC analysis parameters ................................................ 158 Table 5.2 Major peaks identified for sample CONT in the DH-TD GC analysis with Carbotrap 400 ............................................................................................ 159 Table 5.3 Main peaks and their area percentages of sample CONT in DH-TD analysis using Carbotrap 300 ........................................................................... 161 Table 5.4 Main peaks and their area percentages of sample PATA in DH-TD analysis using Carbotrap 300 ........................................................................... 161 Table 5.5 Main peaks and their area percentages of sample PBTA in DH-TD analysis using Carbotrap 300 ........................................................................... 162 XIII Table 5.6 Main peaks and their area percentages of sample PBTB in DH-TD analysis using Carbotrap 300 ........................................................................... 163 Table 5.7 Main peaks and their area percentages of sample PCTA in DH-TD analysis using Carbotrap 300 ........................................................................... 164 Table 5.8 DH-TD and GC-MS analysis parameters .......................................... 165 Table 5.9 Potential Identities of volatile compounds from the DH-TD/GCD analyses ............................................................................................................ 174 Table 6.1 Parameters tested in optimizing SPME method ................................ 178 Table 6.2 Parameters used in the SPME/GC-MS with constant column gas flow .......................................................................................................................... 179 Table 6.3 Response areas for specific ions of volatile compounds in analyzing sample PCTA with SPME/GC-MS under different sampling times (sampling temperature 35°C) ............................................................................................ 185 Table 6.4 Peak response areas for acetone peak in sample PCTA with SPME/GC-MS at different sampling temperatures (sampling time 30 min) ...... 187 Table 6.5 Peak response areas for 1,3-di-tert-butylbenzene peak in sample PCTA with SPME/GC-MS at different sampling temperatures (sampling time 30 min) ................................................................................................................... 187 Table 6.6 Areas of m/z 175 ion at retention time 271 - 272 seconds ................ 191 Table 6.7 Confirmation of volatile compounds by comparing their retention times WIth those of standards ..................................................................................... 194 Table 6.8 Volatile compounds tentatively identified from the HDPE film samples USIng SPME/GC-MS analysis ........................................................................... 195 Table 6.9 New parameters used in the SPME/GC-MS with ramped pressure Program ............................................................................................................ 197 Table 6.10 Comparison of retention times of volatiles determined by the two detectors using SPME/GC-MS and CDC I| sniffing port ................................... 198 Table 7.1 Using 5 groups of validation samples to evaluate the PLS model ..... 209 Table 7.2 Using 3 groups of validation samples to evaluate the PLS model ..... 212 Table 7.3 Correlation coefficients of PLS models based on different pairs of 8amines ............................................................................................................ 212 Tab'e 7.4 Validate the PLS model based on sample PCTA and PBTA ............. 214 XIV Table 7.5 Validate the PLS model based on sample PATA and PBTB ............. 215 Table 7.6 Average response areas of acetone and nonanal detected in the odor profiles of different HDPE film samples in SPME/GC-MS analysis ................... 217 Table 7.7 Predicted response areas of acetone and nonanal based on the PLS model of five groups of samples ....................................................................... 218 Table 7. 8 Predicted response areas of acetone and nonanal based on the PLS model of three groups of samples ..................................................................... 219 Table 7.9 Predicted response areas of acetone and nonanal based on the PLS model of samples CONT and PCTA ................................................................. 220 Table 7.10 Predicted response areas of acetone and nonanal based on the PLS model of samples PATA and PCTA .................................................................. 220 Table 7.11 Predicted response areas of acetone and nonanal based on the PLS model of samples PBTA and PCTA .................................................................. 220 Table 7.12 Predicted response areas of acetone and nonanal based on the PLS model of samples PBTB and PCTA .................................................................. 221 Table 7.13 Predicted response areas of acetone and nonanal of samples CONT and PCTA based on the PLS model without outliers ........................................ 223 Table 7.14 Predicted response areas of acetone and nonanal of samples PATA and PCTA based on the PLS model without outliers ........................................ 223 Table 7.15 Predicted response areas of acetone and nonanal of samples PBTA and PCTA based on the PLS model without outliers ........................................ 224 Table 7.16 Predicted response areas of acetone and nonanal of samples PBTB and PCTA based on the PLS model without outliers ........................................ 224 Table 7.17 Comparison of percentages of difference before and after outliers were eliminated from PLS models .................................................................... 225 XV LIST OF FIGURES fFigure 2.1 Structure of thermoplastic rubber ...................................................... 18 Figure 2.2 Dynamic thermal stripper model 1000 ............................................... 33 Figure 2.3 Carrier gas flow directions in sampling and desorption ...................... 35 Figure 2.4 Dynatherrn thermal description unit model 890 .................................. 36 Figure 2.5 Six-port valve dictating sample preparation (A) and sample desorption (B) paths ............................................................................................................. 37 Figure 2.6 SPME fiber assembly ......................................................................... 42 Figure 2.7 SPME extraction and desorption procedure ...................................... 45 Figure 2.8 Extraction time profile curve in SPME ................................................ 51 Figure 2.9 Role of sensory evaluation department in a food or consumer product company ............................................................................................................. 66 Figure 2.10 Recognition of Brazilian coffee by the E-nose and the human nose 69 Figure 2.11 Different types E-nose sensors ........................................................ 71 Figure 2.12 Data matrix created by analyzing n samples with an electronic nose system with an array of 12 sensors ..................................................................... 77 Figure 2.13 Differentiate two samples with a PCA with three principle components in an electronic nose analysis ......................................................... 78 Fi9Ure 2.14 Correlation of the E-nose nose, sensory evaluation, and GC-MS....87 Fi9ure 3.1 Alpha MOS Fox 3000 E-nose system with HS-100 auto-sampler and M03 sensors ...................................................................................................... 92 l:59ure 3.2 Sensor response patterns of the 12 E-nose sensors in analyzing sampje PBTA under the acquisition parameters listed in Table 3.4 .................... 97 FiQure 3.3 Sensor response patterns of the 12 E-nose sensors in analyzing sample PBTA under acquisition parameters listed in Table 3.5 .......................... 99 Ifigure 3.4 Effect of sample size on the sensor response patterns of the 12 E- 039 Sensors in analyzing sample PBTA .......................................................... 100 XVI Figure 3.5 Effect of incubation time on the sensor response patterns of the 12 E- nose sensors in analyzing sample PBTA .......................................................... 101 Figure 3.6 PCA of HDPE film samples after different incubation time ............... 103 Figure 3.7 Comparison of sensor response patterns of the 12 E-nose sensors in analyzing sample CONT with different sample size .......................................... 104 Figure 3.8 Comparison of sensor response patterns of the 12 E-nose sensors of sample CONT and an "air" sample ................................................................... 105 Figure 3.9 Using PCA module of the E-nose to differentiate five different HDPE films .................................................................................................................. 107 Figure 3.10 Using PCA module of the E-nose to differentiate the headspaces of HDPE films with different sample sizes ............................................................. 111 Figure 3.11 DFA training model of five different HDPE films ............................ 113 Figure 3.12 Validation of DFA model by projecting "unknown" samples ........... 115 Figure 4.1 Rank sums of five HDPE film samples in the pairwise ranking test .126 Figure 4.2 Residuals versus sample ID in analyzing acceptability scores ........ 130 Figure 4.3 Residuals versus sample ID in analyzing intensity scores ............... 131 Figure 4.4 Average acceptability scores of the sensed odor of samples .......... 134 Figure 4.5 Average intensity scores of the sensed odor of samples ................. 135 FiQure 5.1 Carbotrap 400 multi-bed thermal desorption tube (Supelco, 1998a)142 Fi9ure 5.2 Carbotrap 300 multi-bed thermal desorption tube (Supelco, 1998a)143 Figure 5.3 Connecting nickel transfer line of TDU to GC .................................. 145 Fi9ure 5.4 Modified connection between TDU and GCD system ...................... 147 Fi9ure 5.5 TIC and RTIC Chromatogram from mass spectrometer ................... 149 Figure 5.6 RTIC Chromatogram of sample CONT in DIP analysis .................... 150 Figure 5.7 RTIC Chromatogram of sample PCTA in DIP analysis ..................... 150 Figure 5.8 IIC Chromatogram of sample CONT in DIP analysis (set 1) ............. 151 FiQUre 5.9 IIC Chromatogram of sample PCTA in DIP analysis (set 1) ............. 151 XVII Figure 5.10 Mass spectrum of scan #117 of sample CONT in DIP analysis ..... 153 Figure 5.11 Mass spectrum of scan #155 of sample CONT in DIP analysis ..... 153 Figure 5.12 Mass spectrum of scan #181 of sample CONT in DIP analysis ..... 154 Figure 5.13 Mass spectrum of scan # 196 of sample PCTA in DIP analysis ....155 Figure 5.14 Mass spectrum of scan # 916 of sample PCTA in DIP analysis ....155 Figure 5.15 Mass spectrum of scan # 1163 of sample PCTA in DIP analysis .. 156 Figure 5.16 Mass spectrum of scan # 613 of sample PCTA in DIP analysis 1 56 Figure 5.17 Mass spectrum of scan # 660 of sample PCTA in DIP analysis ....157 Figure 5.18 Triplicates of TIC Chromatogram of sample PATA using DH-TD/GCD analysis ............................................................................................................. 166 Figure 5.19 Triplicates of TIC Chromatogram of sample PBTA using DH-TD/GCD analysis ............................................................................................................. 168 Figure 5.20 Effect of split-less time on TIC Chromatogram profiles of sample CONT in DH-TD/GCD analysis ......................................................................... 171 Figure 5.21 Effect of split-less time on TIC Chromatogram profiles of sample PCTA in DH-TD/GCD analysis .......................................................................... 172 Figure 5.22 TIC Chromatogram of sample PBTB from DH-TD/GCD analysis 173 Figure 6.1 ODO II module controls carrier gas to flow splitter and humidified air to sniffing nose cone ............................................................................................. 180 Figure 6.2 Heated transfer line connected to sniffing nose cone in ODO I l module .......................................................................................................................... 182 l=iQure 6.3 Average response areas of acetone and 1,3-di-tert-butylbenzene determined by SPME/GC-MS for sample PCTA versus SPME sampling time .185 Figure _6.4 Average response areas of acetone and 1,3-di-tert-butylbenzene determined by SPME/GC-MS for sample PCTA versus SPME sampling temperature ...................................................................................................... 188 Figure 6.5 Mass spectrum of standard compound 1,3—Di-tert-butylbenzene ..... 190 fIiiigure 6.6 Average area of m/z 175 ion detected in the headspaces of HDPE ms eXCept sample PCTA using SPME/GC-MS .............................................. 192 XVIII Figure 6.7 Mass spectrum of volatile compound detected at RT 271-272 seconds in SPME/GC-MS analysis ................................................................................. 194 Figure 6.8 Odor profile of sample PATA detected in the sniffing test ................ 200 Figure 6.9 Odor profile of sample PCTA detected in the sniffing test ................ 200 Figure 7.1 Side-by-Side comparison of E-nose sensor responses to headspaces of sample PBTA and PCTA .............................................................................. 206 Figure 7.2 PLS plot of "Acceptability" scores of odor profiles of 5 groups of samples with training data for the E-nose Shown only ...................................... 208 Figure 7.3 PLS plot of "Acceptability" scores of odor profiles of 3 groups of samples with training data for the E-nose shown only ...................................... 211 Figure 7.4 PLS plot of "Acceptability" scores of odor profiles of sample PCTA and PBTA with both training data and validation data .............................................. 213 Figure 7.5 PLS plot of "Acceptability" scores of odor profiles of sample PATA and PBTB with both training data and validation data .............................................. 215 Figure 7.6 Using PCA to detect outliers before building PLS models ............... 222 Note: Images in this dissertation are presented in color. XIX CHAPTER 1 INTRODUCTION BACKGROUND INFORMATION The food packaging market has seen increasing applications that require adhesives to be in direct or indirect contact with food. Several examples include fruit labeling, microwave popcorn bags, paper food wraps and plastic food wraps. These applications require various levels of FDA regulation compliance. The level is dependent upon the type of food (aqueous, fatty, dry, and acidic) and the conditions in which the adhesive will come into contact with food. In the past, much attention has been given to the safety of components used in food packaging systems. In most cases an extraction test is completed. However, of equal importance is the effect of components in a food packaging SYstem on the odor and/or flavor quality of the food. Off-odor and off—flavor can become a big food quality issue and can be the result of processing, storage, and preparation of the food packaging system. In many cases, the problem ori9inates from the adhesives used in the package, especially when the package I3 exposed to high temperatures, such as in microwave heating, in which more volatiles are generated. Adhesives are being used in increasing amounts in food packaging Sltuations where there may be direct contact of the adhesive with the product, or other Circumstances which permit migration of substances from the adhesive to the contained food. Compliance with FDA regulations does not in itself ensure that volatile components of the adhesive will not have an adverse impact on food odor. Off-odor is often associated with transfer of small quantities and complex mixtures of volatiles, which are difficult to detect and characterize using traditional analysis methods such as gas chromatography. Use of sensory panels can provide valuable information about odor and taste concerns, but responses of panels can be highly variable. Further, the panel response does not in itself provide any information about the source of the objectionable taste or odor, and therefore is not always of value in efforts to remediate the problem. Since the introduction of electronic nose technology, such systems have been used with considerable success to differentiate between problematic and acceptable samples of products of a variety of types. Patterns of responses to the set of e-nose sensors provide qualitative differentiation between samples. Correlation of these responses to results from sensory panels can then allow the e-nose to be used as a quality control tool. Further, combining e-nose technology with gas chromatography/mass spectrometry or other suitable ana'ysis tools can permit identification of particular compounds as those Predominantly responsible for the odor and flavor problems. Willing et al (1998) investigated the correlation between sensory panel evaluation and electronic nose analysis of odors associated with paperboard. Several groups of e-nose responses were identified which had good correlation with problem odors in the paperboard. Some problem odors did not correlate to any pattern of e-nose responses, and some e-nose patterns did not match any smells identified by the test panel. No attempt was made to identify the compounds responsible for the odors. The e-nose was considered to be a useful tool for such characterization, although further optimization was recommended. The work also identified a statistical technique for correlation of e-nose responses with test panel responses. Culter (1999) surveyed the use of electronic nose technology in quality control of products and packages. He found that most of the work with the technology had been for food products, flavor ingredients, and water. Little had been published about use in evaluation of packaging materials and problems associated with package/product interaction. He presented a procedure for development of methods for such use, since suitable standard methods were not YBt available. The discussion includes sample preparation, purging and eQuilibration time, sampling, and statistical procedures, along with discussion of reproducibility. Examples cited include identification using e-nose of paperboard from different mills, and identification of laminations done by different processes. Gruner and Piringer (1999) studied migration of adhesive components in paper and paperboard packaging into foods. Adhesives studied were an EVA (ethyiene vinyl acetate) hot melt, dextrin, starch, a PVAC (polyvinyl acetate) homopolymer dispersion, and a VAE (vinyl acetate-ethylene) copolymer dispersion. Components in the adhesives which could potentially migrate were determined by extraction using iso-octane and ethanol. Simulation of actual migration from the adhesives applied to paperboard was also carried out. Maximum global migration for several sample package/product systems was then calculated. In most cases, no attempt was made to identify particular migrants. Galotto and Guarda (1999) examined overall migration from plastic packaging materials intended to be in cOntact with foods during thermal and microwave treatment. Microwave heating (compared to thermal treatment) was found to increase overall migration for PVC samples, but not for the other samples studied. These included polypropylene, and four different multilayer systems containing adhesives. Heydanek (1978) examined the prediction of flavor effects of packaging materials, concluding that gas chromatography could be used to help forecast off-flavors and maintain product quality. Examples Cited include a pine or spruce-like odor associated with waxed glassine cereal liners, and an insecticide or plastic off-flavor associated with cereal products stored in polystyrene foam containers. Hollifield (1980) studied off-flavor in maple syrup associated with container-derived contamination. Headspace GC and GC-MS were used to determine the trace volatiles and verify that the methyl methacrylate/styrene/butadiene copolymer containers were the source of the taste and odor problem. The technique was successful in verifying the presence of methyl methacrylate, styrene, and toluene, and showed its usefulness in conducting difficult analyses of volatile contaminants in foods. Ziegleder (1998) studied volatiles extracted from unprinted paperboard using steam distillation and analyzed by GC-MS. A list of 50 volatile compounds commonly found in paperboard was presented. Those found to be Significant in contributing to odor intensity include 2,4-decadienals, 2-nonenal, 2-octen-3-ol, and a variety of aldehydes and short-Chain fatty acids. Ho et al (1994) used purge-and-trap GC-MS to identify 47 volatile compounds, belonging to alkane, alkene, aldehyde, ketone, phenolic, olefin, and paraffin groups, released from blow-molded HDPE bottles, as well as evaluate the effectiveness of anti-oxidants in reducing off-odor and off-taste associated with these compounds. Aldehydes and ketones were found to be the most important odor compounds. Taste and odor tests using an untrained panel were conducted. Freire et al (1998) used Tenax trapping and GC-MS to examine volatiles released at high temperatures from PET-based packaging materials. Most volatiles were found to likely have originated from printing inks and adhesives. Materials tested included PET bottles, roasting bags, susceptor film, PET-coated paperboard, and several PET laminates. SIGNIFICANCE OF WORK Adhesives are widely used in food packaging, especially in primary packaging applications such as paperboard/seal material/flange, sealing pouches and bags, and bonding susceptors in microwave packaging (IOPP, 1995). Potentially, using adhesives poses problems for the odor quality of packaged food. Fast and accurate analysis of off-odor due to adhesives is very important for quality control purposes. Using an electronic nose makes it possible to detect the pattern of responses associated with off-odor as the result of adhesives used In food packaging systems. Off-odor associated with components from food packaging is a significant concern. In many cases, the origin is adhesives used in the package. This concern is particularly great for packages which are exposed to high temperatures, such as in microwaving, which increases the generation of volatiles, but packages used at room temperature or lower are not immune. Identification and characterization of such volatiles is complicated by the fact that many have effects at extremely low concentrations. Using electronic olfactory sensing technology enables detection of patterns of response that are associated with such problems, without requiring identification and quantification of individual components. Further, it permits very rapid detection of potential problems. Therefore, if a relationship between the response of e-nose systems and the presence of objectionable odor can be determined, e-nose systems can be used as an efficient and effective quality control measure in packaging systems. It is possible that the methodology could be used as an on-line monitoring system. Further, through the correlation of electronic olfactory sensing, organoleptic testing, and GC-MS analysis, components of adhesives which contribute to objectionable odors may be identified. Successful completion of this research would permit consideration of their elimination or minimization in adhesive systems, and will also allow electronic nose technology to be used as a potential quality control measure in adhesive manufacturing and application. OBJECTIVES The adhesive/food systems selected for the study are pressure-sensitive formulations used in coating HDPE film for food packaging applications. The Alpha MOS Fox 3000 Electronic Nose system is used to determine the response patterns of the E-nose sensors to the headspaces of HDPE films coated with different adhesives. The resultant odor profiles will be evaluated with sensory evaluation tests; with the suspect component volatile compounds being identified using GC-MS., By correlating the analyses of the odor profiles with the E-nose system, sensory evaluation, and GC-MS, this study will investigate the potential of the E- nose system as a quality control tool. CHAPTER 2 LITERATURE REVIEW ADHESIVES Adhesives are substances that are used to bond materials together. Adhesives take various forms including solids, liquids, or pressure sensitive formulations. They can be of natural origin, such as collagens, starches, dextrins, casein, rubber, etc. They can be synthesized, such as synthetic rubber, block copolymer, thennosetting resin, and thermoplastic resin adhesives, etc. Depending on the applications of adhesives, the formulations vary dramatically and are usually proprietary information. Some factors affecting an adhesive formulation include (IOPP, 2002): Manufacturing process Tackifiers Material base pH additives Adhesive polymer base binders Viscosity additives Carriers Extenders Anti-foaming agents Fillers Anti-mold growth additives Plasticizers Humectants Antioxidants Adhesives are widely used in electronics, wood, pharmaceuticals, health care, automotive industries, and packaging applications (Gutcho, 1983; Pizzi 8 Mittal, 1994). In the late 1980s, about 35% of adhesives used around the world went to packaging-related applications such as paper, paperboard, glass, metal, and plastics (Brody 8r Marsh, 1997). One packaging-related application of adhesives is in food packaging. Examples are adhesives used in tube Iidding, microwave packaging, fruit labeling and food wraps. Adhesives that are widely used in packaging applications include starches, dextrins, resin emulsions, hot melts, and pressure-sensitives. In early years, there were more nature-derived adhesives, but synthesized adhesives find a wider range of applications nowadays due to their variety. Moreover, natural adhesives are prone to attacks by microorganisms and their performance is usually sensitive to environmental factors such as temperature and humidity. In recent years, the adhesive industry is being pushed into two new trends (Brody & Marsh, 1997). One iS the elimination of solvents or solvent- based adhesives, mostly because of the pressure from both governmental agencies and consumer concerns. The other trend is more and more need for recyclable adhesives, especially in recycled packaging applications. 10 Table 2.1 Chronological developments of adhesives in the US. Year Material 1814 Glue from animal bones (patent) 1872 Domestic manufacture of fish glues (isinglass) 1874 First US. fish glue patent 1875 Laminating of thin wood veneers attains commercial importance 1909 Vegetable adhesive from cassava flour (F.G. Perkins) 1912 Phenolic resin to plywood (Baekeland-Thurlow) 1915 Blood albumen in adhesives for wood (Haskelite Co.) 1917 Casein glues for aircraft construction 1920 - 1930 Developments in cellulose aster adhesives and alkyd resin adhesives 1927 Cyclized rubber in adhesives (Fisher-Goodrich Co.) 1928 Chlorized rubber in adhesives (McDonald-B. b. Chemical Co.) 1928 — 1930 Soybean adhesives (I. F. Laucks Co.) 1930 Urea-formaldehyde resin adhesives 1930 — 1935 Specialty pressure-sensitive tapes: rubber base (Drew- Minnesota Mining 8 Mfg. Co.) 1935 Phenolic resin adhesive films (Resinous Products & Chemical Co.) 1939 Poly (vinyl acetate) adhesives (Carbide & Carbon Chemicals Co.) 1940 Chlorinated rubber adhesives 1941 Melamine-formaldehyde resin adhesives (American Cyanamid Corp.) and Redux by de Bruyne (Aero Research Ltd.) 1942 Cycleweld metal adhesives (Saunders-Chrysler Co.) 1943 Resorcinol-formaldehyde adhesives (Penn. Coal Products Co.) 1944 Meltbond adhesives (havens, Consolidated Vultee-Aircraft Corp.) 1945 Furane resin adhesives (Delmonte, Plastics Inst.) and Pliobond (Goodyear Tire 8 Rubber Co.) 1946 Neoprene-phenolic adhesives 1948 Epoxy adhesives 1949 - 1952 NitriIe-phenolic and Nylon-phenolic adhesives 1960 Nylon-epoxy adhesives, modified epoxy-phenolic adhesives (service temperature ~ 550°C) 1962 Polybenzimidazole (service temperature ~ 1000°C) 1966 Polyimide 1969 Polymercaptan sealant commercialized 1974 Polybenzothiazoleiservice temperature ~ 1000°C) Modified, (Delmonte, 1947; Pizzi 8r Mittal, 1994) 11 Pressure-Sensitive Adhesives According to a survey by Business Trend Analysts in 1990, pressure- sensitive adhesives accounted for 44.6% of adhesives sold in the United States with a sales value of $4.9 billion (Pizzi 8r Mittal, 1994). They remain tacky (i.e. sticky) at room temperature and form the bond with the substrate by applied pressure, which makes them easy to use. The three most traditional applications of pressure-sensitive adhesives are packaging tape, labeling, and identification and special marking labels (IOPP, 2002; Pizzi & Mittal, 1994). Recently, they are being used in plastic food wrap in the consumer products market. Formulations of Pressure-Sensitive Adhesives The US FDA regulates the ingredients used in formulating pressure- sensitive adhesives for food applications, in which adhesives can be in either direct or indirect contact with food. Approved ingredients can be found in various sections of the Code of Federal Regulations, 21 CFR 175 and 21 CFR 176 (Rosenberg, 1985). Typical ingredients used in pressure-sensitive adhesives are (IOPP, 2002): Resin or rubber base Tackifiers Plasticizers Fillers 12 Antioxidants Carn'er if pressure-sensitive is solvent or water borne Two major base formulations for pressure-sensitive adhesives are acrylics and rubber/resin blends (IOPP, 2002; Soroka, 2002). Acrylic-based systems are usually composed of acrylic acid ester monomers and other co-monomers. The selection of monomer, ratios of co-monomer, and degree of polymerization directly affect the performance of adhesives including cohesion and adhesion. Rubber/resin blends are usually composed of block copolymers with tackifying resins, oils, and antioxidant. The selection of rubber and resin and the choice of tackifiers control the performance of adhesives (Soroka, 2002). In the book “Adhesives in Packaging” published by IOPP in 2002, the two major formulations of pressure-sensitive were compared (See Table 2.2) (IOPP, 2002): 13 Table 2.2 Comparison of acrylic- and rubber/resin-based pressure-sensitive adhesives Property Acrylics and . Rubber/Resin Acrylic Co-polymer Tack Good Excellent Peel Good Excellent Hold, Room Temp. Good Excellent Hold, Higher Temp. Good Fair Aging Excellent Fair Clarity, Non-Discoloration Excellent Poor (Sunlight, UV) Resistance Oxidation Excellent Poor Oils, Plasticizers Excellent Poor Solvents Polar Fair Good Non-polar Good Poor Humidity (High) Good Good Adhesion Polar Plastics, Metals Excellent Fair Non-polar plastics Fair Good Low Temperature Good Poor Cost Medium-High Low-Medium Reprint from (IOPP, 2002) 14 Coulding (1994) listed the most commonly used elastomers: Table 2.3 Elastomers in common use Elastomer Used in Rubbers Natural rubber Solvent-based and water-based Butyl rubber Solvent-based Styrene-butadiene rubber Solvent-based and water-based Block copolymers Styrene-butadiene-styrene Solvent-based and hot-melt Styrene-isoprene-styrene Solvent-based and hot-melt Other polymers Polybutene Solvent-based and hot-melt Poly(vinyl ether) Solvent-based and water-based Acrylic Solvent-based and water-based Ethylene-vinyl acetate Hot-melt Atactic polypropylene Hot-melt Silicon Solvent-based Reprint from (Goulding, 1994) Forms of Pressure-Sensitive Adhesives Three physical forms exist for pressure-sensitive adhesives: solvent- based, water-based, and hot-melt adhesives. In the book edited by Pizzi and Mittal, a variety of examples of formulations of solvent-based, water-based, and hot-melt pressure-sensitive adhesives are listed (Pizzi 8r Mittal, 1994). Solvent-Based Adhesives Solvent-based adhesives have three major components: an elastomer, the tackifier, and the carrier. Examples of elastomers are natural rubber and 15 synthetic rubbers such as butyl rubber, styrene-butadiene rubber, poly-isoprene, acrylic polymers, block-copolymers of styrene with butadiene or isoprene, silicone elastomer, vinyl ethers, and poly-isobutylene. Two common tackifying resins nowadays are wood rosin derivatives and hydrocarbon resins. The former is usually made through hydrogenation or esterification, and the latter is usually aliphatic, aromatic, or terpenes. The choice of the elastomer and the tackifier, as well as their ratio, determines the suitability for the end-use application, with the tackifier being mainly responsible for tack, peel, and Shear properties (Goulding, 1994). Solvent-based pressure-sensitive adhesives used to dominate the market. However, with more available synthetic elastomers, along with the health and environment concern toward the use of solvent, pressure-sensitive adhesives in the form of either water-based or hot-melt have cut into the market (Brody 8r Marsh, 1997; Goulding, 1994; IOPP, 2002; Soroka, 2002). V_Vg_ter-Ba_sed Adhesives Water-based adhesives are usually in the form of dispersions that are composed of at least two monomers. The combination of two monomers and their ratio give a wide range of properties (Goulding, 1994). Compared to solvent—based adhesives, water-based adhesives have advantages of resistance to heat, UV light, and oxidation. Thus antioxidant can be eliminated from the 16 formulations. However, drying speed can sometimes be a concern in certain applications. Hot-Melt Adhesives A distinction between hot-melt and solvent-based adhesives is the mechanism to control the viscosity. With solvent-based adhesive, the viscosity is controlled with solvents, while the viscosity of hot-melts is controlled through temperature or the appropriate selection of tackifier resin. Unlike water-based adhesives, antioxidant is essential for hot-melt adhesives because they are normally applied at higher temperatures. The primary advantage of hot-melt pressure-sensitive adhesives is their ability to develop bonds very fast, which make them suitable for high-speed processing (Wieczodek, 1990). 17 Different from solvent-based and water-based adhesives, pressure- sensitive hot-melt adhesives have a two-phase structure as shown below: Polystyrene domains (regions of ___> associated - polystyrene end-blocks) Selected polystyrene end-blocks highlighted for clarity \ Continuous Selected polydiene rubber network mid-blocks highlighted (shaded region) for clarity Reprint from (Goulding, 1994) Figure 2.1 Structure of thermoplastic rubber Goulding explained how the two-phase structure displayed in Figure 2.1 works and gives holt-melt adhesives vulcanized rubber-like properties at room temperature and the ability to flow at elevated temperatures or in solvent: “thermoplastic regions of styrene end blocks lock the elastomeric mid-sections of butadiene or isoprene at room temperature but allow the elastomer to move freely at elevated temperatures or in solvent” (Goulding, 1994). Table 2.4 Advantages and disadvantages of pressure-sensitive adhesives Solvent-based Water-based Hot-melt Advantages Quick drying Easy Cleaning Very fast setting Good adhesion to non- Good adhesion to polar No solvent waste polar substances substances Good key on certain Good heat and aging Environmentally plastics resistance acceptable Versatile Environmentally acceptable 100% active High solids Ready to use Disadvantages Flammability Slow drying High equipment cost Toxicity Requires heat to dry Requires heat Relatively low solids Less easy to clean Poor on non-polar surfaces Thermal degradation Difficult to clean Can melt substrate Difficult to package Reprint from (GoUIding, 1994) Off-odor from Packaging Systems Containing Adhesives Flavor scientists attribute off-flavor/odor in a food system to the following sources (Baigrie, 2003): Packaging Materials Microorganisms Oxidative rancidity Millard reaction Interactions between food components Cleaning and disinfecting agents 19 In packaging-related applications, especially in food packaging, the packaging materials are sometimes exposed to high temperatures and/or extended storage periods, which may lead to off-odor. Printing inks, additives, and adhesives, as part of most packaging systems, all can contribute to off—odor because of their low-molecular weight components (Lord, 2003). Kim et al (1988) used purge-and-trap and GC/MS techniques to characterize the off-odor released from two different PVC films. Various volatile compounds were identified, including alcohols, aldehydes, and ketones. It was concluded that those volatile compounds originated from the degradation of the bis-(diethylhexyl) phthalate, one major plasticizer used in PVC film. Freire et al (1998; 1999) studied the formation of volatiles from various food packaging forms including laminates, bottles, and roasting bags made of PET (polyethylene terephthalate). It was found printing inks and adhesives were probably the packaging components that caused the formation of volatile compounds during the thermal processing, while PET itself was not a main factor in the formation of volatiles. McNeal and his colleagues (1993) investigated and identified various volatile compounds released from several commercially-available microwaveable packaging systems with different susceptor designs, which were composed of 20 metalized polyester film, adhesives, and paper packaging materials. It was concluded every component of the susceptors, metallized PET film, adhesives, and paper materials, all contribute to the formation of the complicated volatile profile. In a Similar study, Booker and his co-workers (1989) studied and compared the volatiles released from microwave-interactive paperboard packaging materials that were heated in a microwave oven and in a conventional oven. Solvents from. adhesives were quoted as one of the sources of volatile compounds such as 1,1,1-trichloroethane. Czamecki (1997) focused his work on the solvent retention and its contribution to odor in flexo packaging systems. It was concluded odor is an important attribute of any packaging system, and solvent used in printing inks Should be selected carefully to alleviate the odor problem. In his work, Czamecki also mentioned some water-based adhesives degraded and generated a strong vinegar-like odor. In Anderson’s work (1988), samples from a therrnoforrned and microwaveable container made of PP/Saran/PP were enclosed in a vial and heated in a microwave oven. The released volatiles were collected from the headspace, and four hydrocarbons and BHT (butylated hydroxytoluene) were identified and quantified. 21 In the article authored by Larson (1991 ), off-flavors and off-odor originated from food packaging systems were quoted as one of the major factors that contributed to less-than-satisfactory food quality. Printing inks, adhesives, paperbOard, and plastics were all mentioned as major sources of volatiles. ANALYSIS OF OFF-ODOR IN PACKAGING SYSTEMS Challenges Even though the importance of off-odor in food and consumer products has been widely accepted, the analysis of volatile compounds in an off—odor system has never been an easy task due to the following facts. First, a lot of volatile compounds can be perceived by a human nose at a very low concentration, sometimes even being in the range of ppb (parts per billion) to ppt (parts per trillion), which makes it extremely difficult to detect those compounds with analytical instruments with moderate sensitivity (Baigrie, 2003; Marsili, 1997, 2002). 22 Table 2.5 Taste thresholds of halophenols and haloanisoles Compound , Parts per billion (10"), in water 2-Cholorophenol 0.1 2-Bromophenol 0.03 2,6-DiChlorophenol 0.3 2,6-Dichloroanisole 0.04 (odour) 2,6—Dibromophenol 5 x 10“ 2,4,6—Trichlorophenol 2 2,4.6-Trichloroanisole 0.02 2,4,6-Tribromophenol 0.6 2,4,6-Tribromoanisole 8 x 10'6 (odour) Reprinted from (Kilcast, 2003) As shown in Table 2.5, the taste thresholds of 2,6-dibromophenol and 2,4,6-tribromoanisole are in the range of 10“ and 106 ppb respectively, which are both out of the detection limits of most analytical instruments. Second, the analysis of off-odor is complicated by the inconsistency of reported thresholds. Threshold, a term commonly used by scientists studying flavors and odors, is the concentration of a compound “in a specified medium that is detected by 50% of a specified population" (Kilcast, 2003). As hinted by . the definition, the value of the threshold of a Specific compound can vary dramatically from one reference to another (Devos et al., 1990; Kilcast, 2003; Saxby, 19963), because it is significantly affected by the nature of the medium such as temperature and pH, the test method and procedure, experience and number of test subjects, etc. 23 Table 2.6 lists the thresholds of several compounds above water and in water (Fazzalari, 1978; Lord, 2003; Saxby, 1992, 19963). Table 2.6 The effect of medium on the thresholds of several aromatic compounds Compound Odor threshold Taste threshold ppm in water* ppm in water*“ 2,4,6-Trichloroanisole 3 x IOT 2 x 10'5 2,3,4,6-Tetrachloroanisole 4 x 10’6 2 x 10‘4 Chlorophenol 1.2 6 x 10'3 2,4-Dichlorophenol 0.2 3 x 101 2,4,6-Trichlorophenol 0.8 2 x 10'3 * Odor thresholds above water; ** Taste thresholds in water. Modified from (Lord, 2003) As shown in Table 2.6, a threshold is affected by the means by which human sensesinteract with the compound. In addition, the threshold of the same compound in different mediums varies significantly. The British Standards Institute expanded the meaning of “threshold” and defined two different thresholds; one is the “Detection threshold” and the other is the “Recognition threshold” (BSI, 1992). The former is the lowest concentration of a chemical entity that can be perceived and the latter is the lowest concentration that can be correctly identified. 24 Table 2.7 Different odor thresholds for hexanal In water Threshold Value/range (pan) Odor detection 0.19 - 30.0 Odor recognition 4.5 - 400 Taste detection 0.2 - 10 Modified from (Kilcast, 2003) Third, the analysis of an off-odor system is especially difficult considering the fact that most such systems can be composed of hundreds of volatile compounds, interacting with and affecting each other (Acree & Teranishi, 1993; Jackson & Linskens, 2002). For example, more than 114 chemical components were identified in various citrus essential oils (Ruberto, 2002), 200 different odor regions were detected from a concentrated Chardonnay wine extract (Ferreira et al., 2002), and 850 compounds were reported responsible for the unique taste of beer (Meilgaard, 1982). The aroma of coffee is composed of 791 unique compounds, which belong to 18 classes of Chemical entities such as hydrocarbons, alcohols, aldehydes, ketones, acids, esters, phenols, amines, sulfur compounds, etc. (Partiment, 1997). The sensed odor Characteristic of one volatile compound can vary because of the presence of another volatile compound. In a similar way, the threshold of one volatile compound might change because of the presence of another compound. For example, it was reported that taste detection thresholds of 2, 3, 6-trichloroanisole in tea, whisky, and blancmange were 0.016, 100, and 500 ppb, respectively. The reason is quite simple; that is, presence of other flavor compounds in the medium affects the sensory Characteristics of 2, 3, 6-trichloroanisole (Kilcast, 1996). 25 Adding to this complexity is the fact that the perceived odor or taste characteristics of a compound can Change, as its concentration in a medium changes. Table 2.8 Effect of concentration on the taste description of trans-2-nonenal in water Concentration (jig/I) Taste 0.2 Plastic 0.4 — 2.0 Woody 8 - 40 Fatty 1000 ' Cucumber Reprint from (Saxby, 1996b As shown in Table 2.8, the perceived taste of the same compound changes from an unpleasant plastic-like to a pleasant cucumber-like, as its concentration increases. Fourth, a volatile compound in a food or consumer product system can come from a variety of sources, including ingredients, manufacturing process, packaging system, environment during storage and distribution, etc. This makes it extremely difficult to identify the origin of volatile compounds. For example, it was reported that both the characteristic aroma of rose flowers and the bitter taste of soybean protein vary at their different ripening stages, and thus Chemical compounds responsible for their aroma and taste change during their life cycle too (Helsper et al., 2002; Maehashi & Arai, 2002). 26 Last, the description of an odor system is always complicated and unreliable (Kilcast, 2003). Griffith asked a group of panelists to describe the sensed odor of 2, 3-dichloroanisoles, 30%, 20%, 15%, and 20% of panelists described it as “musty", “medicinal”, “solvent and alcoholic”, and “sweet and fruity", respectively (Griffith, 1974). This can be understood easily considering the experience and sensing capability of one panelist might be totally different from that of another. Moreover, as mentioned earlier, an odor system is usually composed of hundreds of volatile components, and thus the perceived odor is the integrated interactions among those components. This is why screening and training a panel is so important for an accurate and consensus description of the odor system in a descriptive analysis. Nevertheless, odor and/or flavor descriptions of various volatile compounds have been published (Acree & Teranishi, 1993; Baigrie, 2003; Saxby, 1996a). Table 2.9 lists the odor/flavor description of various chemical compounds. Terry Acree and Heinrich Am at Cornell University set up an online database called Flavomet (www.flavomet.org) with 738 odorants. A similar database with more than 1500 entries was built by Don Mottram at University of Reading in UK (http://www.odour.org.uk/). 27 Table 2.9 Odor/flavor description of various compounds Compound 1 ,1-Diethoxyethane 2-Ethyl-5,5-dimethyI-1 ,3-dioxane 3-lsopropyl-2-methoxypyrazine 4,4,6-Trimethyl-1 ,3-dioxane 2,2,6-Trimethyl-1 ,5-dioxane 2,2,4,5-Tetramethyl-1 ,3-dioxane 2-EthenyI-2,5-dimethyI-1 ,3-dioxane 4-Methyl-4-mercaptopentan-2-one 4-Phenylcyclohexene Acetaldehyde Benzophenone Aliphatic acids Alkyl acetates (ethyl-, propyl-, butyl.- acetate) Alkyl substituted benzenes d-Methyl styrene Butyl acetate Chlorocresol Cumene (lsopropyl benzene) Cyclohexanone Di/tribromophenol Di/trichlorophenol Dichlorobenzene Diphenyl sulfide Glycol ethers, e.g. 2-butoxyethanol Guaiacol Hexanal Isophorone Methyl benzaldehyde Methyl benzoate n-propyl benzene Naphthalene p-Cresol Pentan-1 ,2—dione Styrene Thioglycollic acid alkyl esters Toluene Tribromoanisoles Description Jasmine odor Sweet, nutty, woody Musty odor Musty odor Sweet, camphor odor Camphor, liniment odor Musty, liniment odor Catty urine odor Synthetic latex odor Pear-like odor taste Geranium odor Short chain lengths particularly odorous, e.g. butyric acid has a rancid off odor Fruity odor Hydrocarbon Hydrocarbon plastic Pear drop odor, fruity taste Medicinal odor and taste Hydrocarbon Sweet pungent odor Medicinal taint Medicinal taint Medicinal taste Cabbage-like odor Soapy taste Smoky phenolic Board/mown grass odor Pungent brown sugar odor Almond odor Pungent herbal odor Hydrocarbon Petroleum odor/taste Phenolic ' Medicinal, chemical taint DIY fiber glass car repair odor Pungent strong stale beer Petroleum odor/taste Musty odor Modified from (Lord, 2003) 28 Sample Preparation As discussed earlier, concentrations of volatile compounds in an odor system are usually too low to be detected directly with most analytical instruments. Thus a sample preparation step is necessary to separate, purify, and concentrate analytes of interest from a sample matrix. Many sample preparation techniques are available, including solvent extraction, steam distillation, direct (or static) headspace, dynamic headspace (DH), purge and trap, stripping, direct thermal desorption (DTD), solid-phase extraction (SPE), solid-phase micro-extraction (SPME), and supercritical fluid extraction (SFE) (Marsili, 1997, 2002; Zhang et al., 1994). Some of these techniques involve the use of toxic organic solvents, multiple steps, costly equipment, or comprehensive training of the operator. In his report, Zhang listed the criteria for an ideal sample preparation technique: “solvent-free, simple, inexpensive, efficient, selective, and compatible with a wide range of separation methods and applications” (Zhang et al., 1994). Dynamic Headspace and Thermal Desorption The prototype of dynamic headspace and thermal desorption (DH-TD) technique was invented in the 19603 by a group of scientists in California but the technique did not draw much attention until Tenex (poly-2,6-diphenyl-p- phenylene oxide) was introduced as a universal adsorbent material by Zlatkis 29 and his colleagues at the University of Houston in 1970s (Ettre, 2001 ). Since then the technique has been widely accepted asan effective way to concentrate analytes from various sample matrices, such as air, water, soil, pharmaceutical, food, and packaging materials. Compared to traditional sample preparation techniques, DH-TD is advantageous because it eliminates the use of solvent. Though several versions of thermal desorption techniques exist, including direct thermal desorption, short-path desorption, automatic thermal desorption, and thermal desorption cold trap—injection, there are always two stages involved. The first stage is to trap analytes of interests by using a thermal desorption tube filled with single- or multi-bed adsorbent material, and the second step is to desorb the trapped analytes to the separation instruments such as GC (Gas Chromatography) by heating the desorption tube. Theory of Dfi-LQ The term “dynamic headspace” is used to differentiate itself from static headspace, in that an inert carrier gas is continuously flushed into the sample vial and thus the headspace above the sample is always Changing. In static headspace sampling, no sample will be taken until equilibrium has been established between the sample matrix and the headspace, and thus concentrations of volatile compounds have been stabilized. 30 In contrast, volatile compounds are instantly swept out of the headspace and trapped in the adsorbent tube at the exit when an Inert gas is continuously flushed into the sample vial in dynamic headspace sampling. As a result, the thermodynamics favors the transfer of volatiles from the sample matrix to the headspace, and an exhaustive extraction becomes possible. A mathematical model was quoted in a paper authored by Nunez to calculate the theoretical time required to strip 95% of an analyte out of the sample matrix, Tags (Nunez 8 Gonzalez, 1984): 3 T005 = 'I:(VG +KiVL) where F is the volumetric flow rate of the purge gas in the sampling process, VL and V3 are the volumes of the sample and the gaseous phase, and K; is the capacity factor, which is dependent on the analyte and the trapping tube. In the same report, a second theoretical model was quoted to predict the value of the breakthrough volume for a particular compound (Nunez 8 Gonzalez, 1984): V=V§(1—2/f1\7) where N is the number of theoretical plates of the trapping tube and VRi is the retention volume of a compound i, which is dependent on the properties of the compound itself, the trapping tube and the purge gas. 31 “Breakthrough volume" is used as a limiting factor in the trapping process (Nunez 8 Gonzalez, .1984; Reid, 2003). As the purge gas continuously flows through the trapping tube, trapped volatile compounds will slowly move Upward and eventually reach the end of the tube and begin to be eluted. Several factors were mentioned as important in dictating the value of the breakthrough volume (Nunez 8 Gonzalez, 1984), including 1) properties of the trapping tube, including the size of the tube, the adsorbent material used in the tube such as porosity, surface area, polarity, amount, and interaction with the analytes; 2) properties of the purge gas, including flow rate, temperature, and purity; 3) properties of the analytes, including concentrations, Chemical structures, and sample matrix. If the sample matrix contains water, moisture carried over and retained in the adsorbent bed can lead to difficulty in the following thermal desorption process. However, because the breakthrough volume for water is usually much lower than those for volatile compounds, the trapping tube can be purged with dry clean gas after the sampling to eliminate the retained moisture from the sample matrix (Reid, 2003). Figure 2.2 is a picture of a Dynamic Thermal Stripper Model 1000 used in dynamic headspace sampling to trap volatile and semi-volatile compounds, in which the oven temperature can be maintained between ambient and 150°C. 32 Thermal Desorption Tube '3 <— Heatable Jacket Purge Gas ——... Heated Airbath Oven (30 - 200°C) Injection Port Reprint from (Supelco, 1998a) Figure 2.2 Dynamic thermal stripper model 1000 33 Various adsorbent materials are available (Nunez 8 Gonzalez, 1984; Reid, 2003), including Poropak series (Horwood et al., 1981 ), Chromosorb series (Whitfield et al., 1983), Tenax-GC (Durst 8 Laperle, 1990; Kwo, 1991; Mazza 8 Pietzak, 1990; Wellnitz—Ruen et al., 1982), Tenax-TA (Kanavouras, 2003; Morales et al., 1997; Werkhoff 8 Bretschneider, 1987a), and multi-bed tubes such as Carbotrap 300 (Kanavouras, 2003) and other Carbotrap series (Supelco, 1998a) Compared to Single-bed tubes, multi-be‘d-tubes are more efficient in trapping a wider spectrum of volatile compounds with varying polarity in a single sampling. In addition, the order of different adsorbent materials in a tube is in such a way that the least volatile compounds will be trapped by the least active layer and the most volatile compounds will be trapped by the most tenacious layer. This arrangement ensures high molecular weight compounds will not be irreversibly adsorbed by the most tenacious adsorbent layer in the tube, which will lead to a very slow desorption process (Supelco, 1998a). It also ensures low molecular weight compounds, Which usually have lower breakthrough volume, can be effectively retained without being eluted out of the trapping tube before the sampling process is completed. Sampling Carrier Gas Flow Least Active Layer — Most Tenacious Layer — Trap High M.W. Compounds Trap Low M.W. Compounds t Desorption Carrier Gas Flow Figure 2.3 Carrier gas flOw directions in sampling and desorption Following the sampling is the thermal desorption step, in which the desorption tube with retained analytes is loaded onto the thermal desorption unit and heated to a high temperature instantly so that the analytes are released very quickly from the tube to the head of a GC column for separation. Figure 2.4 is the Dynatherrn Thermal Desorption Unit Model 890, in which the tube chamber can be heated up to 399°C very quickly to help desorb trapped analytes from thedesorption tube to the head of the GC column in a few seconds. In addition, all transfer lines can be heated to 250°C to avoid condensation of compounds with higher boiling points. 35 Figure 2.4 Dynatherm thermal description unit model 890 The unit features a multi-port valve design, as shown in Figure 2.5, which makes two distinct flow paths possible; one is for sample preparation (or focusing) and the other is for sample desorption. 36 ._, Focusing Tu'e GC Sample Desorption Path Carrier Gas Sample Preparation Path Split Vent for Sample Saver Focusing Tube Carrier Sample Preparation Path Gas Sample Desorption Path Split Vent for Sample Saver —> GC Figure 2.5 Six-port valve dictating sample preparation (A) and sample desorption (8) paths Path A is used to clean the desorption tube after analytes have been eluted to the GC column and before the tube is used for the next sampling. Alternatively, path A can be used to focus analytes by desorbing them from the relatively bigger thermal desorption tube, which is 4 1/2” x 6 mm O.D., to a small- bore adsorbent tube. The smaller internal volume of the focusing tube helps improve desorption efficiency, especially when a capillary column is used in the GC (Dynatherm, 1989). 37 Path B is used to introduce thermally desorbed analytes from the desorption tube to the head of the GC column for separation. An optional split path is connected to a sample saving vent, where a portion of desorbed analytes from the original desorption tube can be re-collected.by a second tube. Applications of DH- TD in Packaging Analysis In packaging, the dynamic headspace and thermal desorption technique has been used to study the odor systems originating from various food and consumer products packaged in different packaging materials, as well as the permeability properties of polymer materials. Werkhoff and his co-workers (1987a; 1987b) optimized operating parameters in dynamic headspace and thermal desorption, including sampling temperature, desorption temperature, and desorption gas flow rate. Several flavor and fragrance applications using the thermal desorption technique were demonstrated. Kwo ( 1991) used a Carbotrap-300 thermal desorption tube to trap volatile compounds released from a heated susceptor and then thermally desorbed the compounds to a GC column for analysis. Six volatile compounds were confirmed in her study. 38 A modified purge and trap/thermal desorption system was used to measure the organic vapor permeability of various high barrier polymer membranes (Chang, 1996). The approach proved to be more sensitive than a standard isostatic procedure using a MAS 2000TM permeation instrument. In his study investigating the effect of co-permeant on the permeability of various binary organic vapor mixtures through OPP and PVdC coated OPP films, Laoharavee (1998) used a purge and trap/thermal desorption system to measure the concentrations of permeated organic vapors. The author concluded that the compositions of the studied binary vapor mixtures did not affect the mass transfer properties of the co-permeant through the two films used in the study. Kanavouras (2003) evaluated flavor compounds generated from packaged olive oil with the dynamic headspace and thermal desorption approach using Carbotrap-300 and Tenax-TA desorption tubes. It was concluded the method was capable of isolating and concentrating flavor compounds in olive oil, which were then identified with a coupled GC/MS. instrument. Direct thermal desorption and short-path thermal desorption, two other revised versions of the thermal desorption technique, were used by some researchers to study various flavor and fragrance problem, from food packaging films (Hartman et al., 1991 a), to forest products (Coello—Perez et al., 1997), to air and soil samples (Manura 8 Hartman, 1992), to food and food products (Hartman 39 et al., 1991b; 1991c; Manura 8 Hartman, 1992), and to pharmaceutical applications (Manura 8 Hartman, 1992). Solid-Phase Microextraction (SPME) SPME (Solid-Phase Microextraction) is a solvent-free sample preparation technique that was first introduced in 1990 at the University of Waterloo in Canada (Arthur 8 Pawliszyn, 1990). Compared to DH-TD, SPME involves easier sampling procedures and requires simpler instmment setup. SPME has almost all the qualities of an ideal sample preparation technique (Zhang et al., 1994): “solvent-free, simple, inexpensive, efficient, selective, and compatible with a wide range of separation methods and applications”. Moreover, SPME has more advantages such as linear results over a wide range of concentrations of analytes (Arthur et al., 1992/1993; 1992a; 1992b; 1992c; Potter 8 Pawliszyn, 1992, 1994; Supelco, 1998b), automatic sample introduction to CC or HPLC, fast sampling process, reusability, and its applicability to field sampling (Supelco, 2005). Table 2.10 compares SPME and other sample preparation techniques (Supelco, 1998b). 4O Table 2.10 SPME compared with other sample preparation techniques Detection Precision Expense Time Solvent Simplicity Limit . (% RSD)“ Use Purge 8 Trap ppb 1 - 30 High 30 min No No Stripping ppt 3 — 20 High 2 hr No No Headspace ppm N/A Low 30 min No Yes Liquid-Liquid Extraction ppt 5 — 50 High 1 hr Yes Yes Solid Phase Extraction ppt 7 - 15 Medium 30 min Yes Yes SPME ppt < 1 — 12 Low 5 min None Yes Reprinted from (Supelco, 1998b) * RSD%, percentage of relative standard deviation, also called CV (Coefficient of Variation), equals the ratio of the standard deviation to the mean. In the past sixteen years, SPME has been widely accepted as a powerful and easy-to-use technique in extracting odor and flavor compounds from solid, gaseous, and liquid sample matrices, environmental analysis, forensic analysis, and toxicology applications (Supelco, 2001a; 2005). Theory of SPME The core of the SPME technique is a Chemically-inert and stable fused- silica fiber coated with liquid-phase polymer material, and in some cases mixed 41 with a solid adsorbent. Various coating materials are available, with varying polarities for extracting different volatile compounds (see Table 2.11). The coated silica fiber is very fragile and thus needs some mechanical protection. The SPME device is designed in such a way that the fiber Is connected to a stainless steel plunger via a spring. The fiber assembly is then contained in a syringe-like holder whose end is a hollow needle (see Figure 2.6). Ad usable noodle - a ~----'- Reprinted from (Zhang et al., 1994) Figure 2.6 SPME fiber assembly 42 Table 2.11 Various SPME fiber coatings and coating thickness and their recommended applications ‘ Fiber - Recommended use PDMS (Polydimethylsiloxane) 7 pm 30 pm 100 pm Moderately polar to non-polar high molecular weight compounds (MW 125 — 600) Non-polar semi-volatiles (MW 80 — 500) Volatiles (MW 60 — 275) _ PDMS/DVB (PolydimethylsiloxanelDivinylbenzene) 60 um 65 um CAR/PDMS (Carboxeanolydimethylsiloxane) . 75 um, 85 pm PA (Polyacrylate) 85 pm CW/DVB (Carbowax/Divinylbenzene) 65 um, 70 pm Amines and polar compounds, for HPLC only - Polar volatiles, amines and nitro-aromatic compounds (MW 50 — 300) Trace-level volatiles, and gases and low molecular weight compounds (MW 30 — 225) Polar semi-volatiles (MW 80 — 300) Alcohol and polar analytes (MW 40 - 275) ’ DVBICARIPDMS (DivinylbenzeneICarboxen On Polydimethylsiloxane) 50/30 um Flavor compounds: volatiles , and semi-volatiles, C3 — C20 (MW 40 — 275) CWIT PR (CarbOwax/I’emplated A Resin) ‘ j j f 7 _ . 50 pm Surfactants (for HPLC only) Modified from (Supelco, 1998b; 2005) 43 Similar to the HD-TD technique, two steps are involved in SPME. The first is sampling and the second is desorption. In the first step, the coated silica fiber is either exposed to the headspace of a solid or liquid sample matrix or directly submerged into the gaseous or aqueous sample to extract analytes of interest onto the fiber. The former is called ‘Headspace SPME’ and the latter is termed “Direct SPME’. In the second step, the fiber assembly is inserted into a GC injection port or an HPLC interface port and the fiber is exposed again to release analytes by heat (GC) or by the mobile phase (HPLC). 44 Extraction Procedure Me septum Retract IIbe/IIVItthaN needle. on sanple contaner. SPME flberiextract analytes. —-> DesorptIon Procedure Ram we” Pierce septum in CC inlet (or withdraw needle. introduce needle Into SPME/HPLC Interface). Expose fiber/desorb analytes. Reprinted from (Supelco, 1998b) Figure 2.7 SPME extraction and desorption procedure 45 Thermodynamics of Direct and Headspace SPME The extraction process in SPME sampling is essentially a partitioning process of analytes between the coating on the fiber and the extraction medium (headspace or the sample matrix itself). Equilibrium will not be reached until the chemical potentials of the analyte in all phases are equal (Zhang et al., 1994). Theoretical models were proposed to describe the thermodynamics of the partition process. In direct SPME, the equilibrium involves two phases: the sample matrix (gaseous or aqueous) and the liquid coating on the fused silica fiber. The extracted amount of a particular compound can be calculated using this theoretical equation (Yang 8 Peppard, 1994; Zhang 8 Pawliszyn, 1993; Zhang et al.,1994): K V V n: f9 f S C0 VS+KfSVf in which n is the mass of a compound adsorbed by the coating after equilibrium has been reached; V: and Vs are the volumes of the coating and the sample; Krs is the partition coefficient of the compound between the coating and the sample matrix; and Co is the initial concentration of the compound in the sample matrix. 46 This equation indicates the linear relationship between n and Co. Moreover, if the affinity between the compound of interest and the coating is strong, the value of K13 is high, which leads to good sensitivity and concentrating effect. In the extreme case where Kgi >> Vs, the above equation can be simplified as: n = VSCO and an exhaustive extraction is reached, which further underlines the importance of Choosing the right coating material for a particular analysis. However, Krs cannot be realistically large enough for every volatile compound in an odor or flavor system. Another conclusion can be deduced from this equation, which is the suitability of SPME to field sampling. If the sample volume is extremely large, e.g. open air, a lake, or a river, Vs >> KrsVr and the equation is simplified as: n = KfszCO As indicated by this new equation, n is not dependent on the sample volume anymore, making the SPME technique a suitable tool for field sampling purposes. 47 On the other hand, three phases are involved in the thermodynamics of headspace SPME and a different model was proposed (Yang 8 Peppard, 1994; Zhang 8 Pawliszyn, 1993): _ K fSVfVS n — C0 VS +KfS‘Vf + KgSVg in which V3 is the volume of the headspace and Kgs is the partition coefficient of the analyte between the headspace and the sample matrix. Comparing the two equations for direct and headspace SPME, the only difference is the term Kgsvg, which means It based on headspace SPME is always smaller than n based on direct SPME. In the other words, direct SPME is always more sensitive than headspace SPME. However, Kgs is relatively small for most analytes (Zhang 8 Pawliszyn, 1993), which makes the amount of extracted compound, designated as ‘n’ in these equations, in headspace SPME similar to that in direct SPME, if V9 (headspace volume) is much smaller than Vs (sample volume) as well (Supelco, 1998C). Kinetics of Qirect and Headspace SPME The kinetics of the extraction process in SPME is controlled by the mass transfer of the analyte among the sample matrix, the headspace, and the coating on the fiber. 48 In direct SPME, the coated fiber is immersed in the gaseous or the aqueous sample. Equilibrium can be reached very quickly if the mass transfer is controlled by the diffusion of analytes in the thin coating (3 — 100 um) (Louch et al., 1992) and if the sample is gaseous, which leads to a large diffusion coefficient. Vigorous agitation, e.g. magnetic stirring or sonication, is necessary to help reach equilibrium quicker in an aqueous sample. However, the equilibrium time is usually longer for an aqueous sample in reality even with agitation because of a thin static aqueous layer adjacent to the coating on the fiber (Arthur et al., 1992/1993; Zhang et al., 1994). Headspace SPME needs to be adapted for solid samples or other sample matrices where direct SPME is not suitable, such as oil, wastewater, and sludge. The mass transfer in headspace SPME is more complicated because two inter- phases among three phases (matrix, headspace, and coating) are involved. It is easier to extract low molecular weight compounds, which have lower boiling points and high volatility, than to extract semi-volatiles. The speed of extraction in these cases is usually dependent on how fast analytes can escape from the sample matrix into the headspace (Pawliszyn, 1993; Zhang 8 Pawliszyn, 1993). One way to speed up this process is to heat the sample as well as stirring the sample to continuously create a fresh surface between the sample matrix and the headspace (Zhang 8 Pawliszyn, 1993; Zhang et al., 1994). 49 Table 2.12 Comparison of kinetics and applications of direct and headspace SPME SPME Kinetics Applications Direct SPME Controlled by the diffusion of the Gaseous or Clean/low- analyte in the coating viscosity aqueous sample (if the solution is well stirred) Headspace Controlled by the mass transfer Any sample matrix SPME from the sample to the headspace More detailed theoretical analysis of the mass transfer process in both direct SPME and headspace SPME can be found in other papers (Koziel et al., 2000; Louch et al., 1992; Martos 8 Pawliszyn, 1997; Zhang 8 Pawliszyn, 1993). Parameters in SPME Because the extraction in SPME is a partitioning process, the amounts of extracted analytes will not be stabilized until equilibrium is reached. In some applications, equilibrium can be reached in less than 1 min (Arthur et al., 1992/1993; Louch et al., 1992). However, amuch longer time (2 — 30 min) is usually required in most applications (Supelco, 1998b). Figure 7 explains how the extraction time affects the amount of analyte extracted by the fiber in SPME. 50 u Pre-Equlllbrlum /* * — ,_\.\ Time Control Not As Critical. Small chmge in time results Ana'ym '0 small or no change in Absorbed analyte absorbed. Time Control i Eqnilibrium ‘ Is Critical. ‘ Reached = Small change in , time-results in large change in analyte . s absorbed. ; Extraction Time > Reprinted from (Supelco, 2001a) Figure 2.8 Extraction time profile curve in SPME Time As shown in Figure 2.8, the amount of extracted analytes will become independent of the extraction time once equilibrium has been reached. On the other hand, that amount is dependent on the extraction time during the pre- equilibrium stage. It becomes very critical to keep the extraction time consistent ifworking in the pre-equilibirum period. 51 Temmratu re It is easily understood that temperature is another critical parameter because it affects the kinetics of the extraction process. Partition coefficients of analytes between the headspace and the sample matrix increase with temperature at first, and then start to decrease after reaching an optimum temperature. In the other words, more analytes are released from the sample matrix as the temperature rises, which leads to higher concentration of analytes in the headspace and favors the SPME sampling process. This trend will start to reverse after passing an optimum temperature, at which partition coefficients of analytes between the coating on the fiber and the headspace start to decrease. Coating The type and the thickness of the coating polymeric materials on the fiber, along with the addition of adsorbent materials such as Carbowax and Carboxen, affect the effectiveness of the extraction process, as well as the amount of analytes that can be extracted after equilibrium is reached (see Table 2.11 and thermodynamics equations). Generally speaking, a thick coating is used to extract volatile compounds, and a thin coating is preferred for extracting semi- volatile compounds because a thick coating in this case will prolong the desorption time and semi-volatile compounds can be carried over to the next sampling (Supelco, 1998b). 52 Other factors Other important factors in the SPME sampling process include: sample volume, headspace volume, sample vial size, pH and salt content of the sample ' solution, sample agitation, and fiber immersion depth (Supelco, 1998b; 2001a; 2001b). Important factors in the desorption process include: time‘between extraction and desorption, fiber positioning during injection, injector temperature, desorption time, etc (Su pelco, 1998b; 1998c; 2001 a; 2001 b). In SPME sampling, it is} more important to keep these sampling parameters consistent than to reach a full equilibrium before taking samples. Various studies have proved that a reliable, precise and reproducible quantitative analysis using the SPME technique is possible when the sampling practice is well controlled and kept consistent (Martos 8 Pawliszyn, 1997; Supelco, 1998b, 2001 a; Zhang et al., 1994). In one study, ten duplications of SPME/GC analysis were conducted with a solution of Chlorinated pesticides, which contained 20 compounds. The results showed the % RSD (percentage of relative standard deviations) of the relative response of the 20 compounds in the gas chromatographs ranged from 4.8% to 28.6% (Supelco, 1998b). When Potter and his colleagues used a 15 pm PDMS- coated SPME fiber to extract volatile and semi-volatile compounds from water 53 samples, it was concluded the detection limits in the SPME analysis exceeded the regulatory requirements of the standard analysis method US EPA 525 (Potter 8 Pawliszyn, 1994). Applications of SPME in Packaging Analysis Supelco compiled a SPME applications guide with over 500 references (Supelco, 2001c), which covers a large variety of applications using the SPME sampling technique, from food to packaging materials, to pharmaceuticals, to toxicology, to forensics, to environmental analysis (air, wastewater, pesticides, surfactants, and soil), and to field sampling. A 75 um CAR/PDMS fiber was used to extract US EPA method 624 volatile organic compounds from water in field sampling (Supelco, 2004). Concentrated compounds adsorbed on the fiber were then stored at -40°C for 3 days before being analyzed with a GC. The results proved the fiber was capable of retaining compounds effectively. Excellent linearity was obtained (R2 of 0.953 to 0.990) for a standard mixture composed of 9 volatile compounds including acrylonitrile, 1,3-butadiene, and vinyl chloride, whose concentrations were in the range of 1 and 400 ppb. Various flavor analyses using the SPME sampling technique were quoted in the report by Supelco, including fnJit juice beverage, whole fruit, flavor oils, punch flavor in the presence of glycerin, rancid corn oil, odor in wines (trichloroanisole), and off-flavor compounds from light-oxidized milk (Supelco, 2000). One study conducted by Song and Beaudry at Michigan State University attributed the characteristic flavor of fresh, ripe tomato to Z-3-hexenal, E-2- hexenal, 1-pentene-3-one, 2-isobutylthiazole, and 6-methyl-5-hepttene-2-one, and the flavor of strawberries to methyl butanoate, ethyl butanoate, methyl hexanoate, hexyl acetate, ethyl hexanoate, 2,5-dimethyl-4-hydroxy- 3(2H)furanone and its methyl ether, 2,5-dimethy-4-methoxy-3(2H)furanone. In total 30 and 34 volatile compounds were identified for tomato and strawberry, respectively, using SPME sampling coupled with time-compressed gas Chromatography (TCGC) and time-of-flight mass spectrometry (T OFMS). The study proved that the SPME sampling can significantly cut short the total analysis time compared to purge-and-trap and/or simultaneous steam distillation techniques (Supelco, 2000). Chai et al (1993) compared the sensitivity of direct SPME and headspace SPME and investigated the linearity of concentrations of volatile chlorinated hydrocarbons in air and water samples. It was concluded the two techniques had comparable sensitivity. Results showed direct SPME held linearity over a wider range of concentrations but required longer sampling time relative to headspace SPME. 55 Potter and Pawliszyn (1992) used a 100 pm PDMS fiber to extract BTEX compounds (benzene, toluene, ethyl benzene, and xylene) in water. The results showed the detection limit of benzene in water was 15 pglml, which was well below the requirement in the US EPA method 524.2 (30 — 80 pglml). The study also proved the SPME sampling technique coupled with ion-trap mass spectrometry provided good linearity, which extended over five orders of magnitude, and good reproducibility with a RSD% in the range of 2.7 to 7.5% for the concentrations of the BTEX compounds. Four different SPME fiber assemblies, 100 pm PDMS, 65 um PA, 65 pm PDMS/DVB, and 65 pm CW/DVB, were used to investigate the flavor additives used in tobacco products (Clark 8 Bunch, 1997). The headspace sampling technique was proved effective for the application and a total of 31 characteristic tobacco flavor compounds were detected. Tombesi and Freije (2002). used SPME-GC-MS to monitor the concentration of BHT (Butylated hydroxytoluene) in fifteen commercial branded water samples packaged in plastic containers. The study focused on the fiber exposure time,.detection limits, linearity and precision of the analysis. Hill and Smith (2000) investigated volatile and semi-volatile sulphur compounds from beer at trace levels with various SPME fibers using headspace SPME sampling combined with GC and found the CAR/PDMS fiber was the most 58 effective among the fibers studied. The technique proved simple, effective and provided good sensitivity for the flavor analysis of beer. SENSORY EVALUATION Analysis of off-odor is always complicated because an odor system might be composed of hundreds of volatile compounds. Nevertheless, both quantitative and qualitative tools are available for investigating and describing off- odor and/or off-flavor problems. Quantitative analyses such as GC-MS focus on locating and identifying volatile compounds in the odor system, while qualitative analyses such as sensory evaluation focus more on the description of the odor system. The Sensory Evaluation Division of the Institute of Food Technologists defined sensory evaluation as “the scientific discipline used to evoke, measure, analyze and interpret those reactions to characteristics of foods and materials as perceived through the senses of sight, smell, taste, touch and hearing” (IFT, 1975). Sensory evaluation is considered “a child of industry” (Lawless 8 Heymann, 1998) because its systematic formation as a scientific discipline in the middle of the 20‘" century was the result of rapid growth of consumer product and 57 food industries. However, sensory evaluation is widely recognized as a useful and effective tool to help improve quality of consumer and food products. Human Senses As explained in the definition (IFT, 1975), human senses are used in sensory evaluation to evaluate the quality of products in terms of color, shape, sound, taste, mouth-feel, texture, flavor, etc. In his book published in 1996, Schiffman described the process of perceiving a sensory attribute as: stimulus triggers sensation by the sense organ. The signal is transferred to the brain via the nerve system and processed there. By comparing the processed signal (the sensation) with the database, which is generated via past experiences, saved in the brain, a perception results (Schiffman, 2001). Compared to most commonly used analytical instruments, a well-trained panelist can be more sensitive (Grimm et al., 2002; MacRae 8 Falahee, 1995; Peled 8 Mannheim, 1977). For example, it was reported (Baigrie, 2003) that a human nose is capable of detecting some volatile compounds with a concentration as low as 0.01 ppb (parts per billion), making a human nose a powerful tool in studying odor/flavor problems. 58 Table 2.13 lists the major human senses and their corresponding perceived sensory attributes. Table 2.13 The human senses and their perceptions Sense Perception Vision Appearance (color, shape) Hearing Sound Gustation Flavor/taste/mouth feel Olfaction Odor/aroma/fragrance Chemical/trigeminal Irritant Touch Texture and consistency Modified from (Kilcast, 2003) Some people use the term “odor" and “flavor” interchangeably. However, most sensory experts agree there is a distinction between them (Baigrie, 2003; Lawless 8 Heymann, 1998; Meilgaard et al., 1999). Odor is sensed when a volatile compound enters the nasal passage, either via the nose. or via the retro-nasal path in the mouth, and stimulates the olfactory epithelium found in the roof of the nasal cavity (Baigrie, 2003; Kilcast, 2003; Meilgaard et al., 1999). In comparison, flavor involves more complicated processes and is the combination of aromatics perceived by the olfactory system, tastes perceived by the gustatory system, and the chemical feelings perceived by the Chemical/trigeminal system (Meilgaard et al., 1999). 59 Types of Sensory Evaluation ‘ There are three major categories of test methods used in sensory evaluation, and they serve different purposes (Lawless 8 Heymann, 1998; Poste etaL,1991) Table 2.14 Various test methods used in sensory evaluation , Category Discriminative Descriptive Affective ‘ How do products Question of Interest Type of Test Test Methods Are products different in any way? (is. determine whether a difference exists) wAnaNuc" “Analytic” different in specific sensory characteristics? (i.e. determine the nature and intensity of the difference) How well are products liked or which products are preferred? (i.e. determine relative preference) “Hedonic” Triangle test Duo-trio test 2-out-of-5 test Paired comparison test Ranking test (Friedman) Scaling methods Descriptive analysis Paired comparison preference test Hedonic scaling test Ranking test Modified from (Lawless 8 Heymann, 1998) Discriminative Tests In the discriminative category, the main purpose is to decide whether or not one test sample is different from another. Either trained or untrained panel 60 can be used in this type of test. The advantage of discriminative tests is they can be used as a quick way to screen panelists to imprOve the accuracy and precision of later sensory evaluation results. The disadvantage is they do not answer questions like how big the difference is. Among the test methods in the discriminative category, each has certain pros and cons (Poste et al., 1991). Table 2.15 Comparison of different methods in the discriminative sensory test Test Method Advantages Disadvantages Triangle test Used in quality control Samples must be Used to screen panelists homogeneous (i.e. samples can only be different in one aspect) No nature or intensity of difference is determined Do not specify any particular characteristic Duo-trio test Easier than triangle test , Same as triangle test Statistically less effective than triangle test Always one-tailed test 2-out-of-5 Statistically more effective than Strongly affected by sensory test triangle test fatigue Suitable for visual, auditory, and Not suitable for flavor and tactile analysis odor analysis Paired Same as triangle test No indication of the intensity comparison Determine if a difference exists of the difference test in a characteristic Statistically less effective Can be either one-tailed or two- than triangle test tailed test Ranking test Carl study several samples at No indication of the intensity once of the difference Used to screen samples Results are sample group Significance study possible dependent Fixed ranking unit Modified from (Poste et al., 1991) 61 Descriptive Tests Different from discriminative tests, descriptive tests answer questions such as how big the difference of the perceived sensory attribute is, or what sensory characteristics lead to the detected difference. More than one sensory characteristic can be evaluated at once, which is considered more advantageous than discriminative tests. In addition, they can be used to describe the profile of the perceived sensory attribute such as flavor and texture. In scaling methods, both structured scaling and unstructured scaling are used. In structured scaling, the determined intensity of difference does not necessarily disclose the true intensity of difference because the numerical distance between any two consecutive anchor points is always fixed as one. Moreover, because each panelist might have different psychological perceptions of descriptive words and the structured scale, the panel is usually trained so everyone on the panel can agree upon the meaning of those terms and thus an accurate and meaningful result can be obtained. However, a trained panel does not necessarily represent a typical consumer group. As a result, the findings from this test can not always be generalized as the opinion of consumers. In comparison, panelists have the freedom to mark the perceived intensity of the investigated sensory attribute anywhere on the unstructured scale. 62 Nevertheless, the significance of the difference can be determined with the appropriate statistical analysis in both structured and unstructured scaling tests. Affective Tests In order to obtain a more complete profile of the sensory attributes of a product, a descriptive analysis method should be used. A lot of such methods have been developed, including flavor profile, texture profile, quantitative descriptive analysis (QDA), spectrum descriptive analysis, time-intensity descriptive analysis, and free-choice profiling (Meilgaard et al., 1999; Poste et al., 1991). Usually screening and extensive training of panelists are involved in descriptive analysis. Affective tests are also referred to by many as hedonic tests. AS disclosed by the term itself, this type of tests focuses mainly on the subjective opinion, such as preference, liking, and acceptance, of panelists toward a product based on its perceived sensory attributes. Usually a large number of untrained panelists are used so the result can be generalized as the opinion of consumers. HoWever, the panelists can be selected from a specific region, age group, gender, income level, and so on, so that a target market can be studied for the product. Three most commonly used methods in this category are paired comparison preference test, hedonic scaling test, and ranking test (Poste et al., 1991). 63 Table 2.16 Characteristics of various affective sensory tests Test Characteristics Paired comparison preference test Hedonic scaling test Ranking test Used to investigate the preference by the panel; Either one-tailed or two-tailed test; No data on the size of the difference; No indication of what characteristic the preference is based on. Used to measure the degree of liking; Either unstructured or stmctured scale; Order of sample presentation is randomized for each panelist. Either overall preference or a specific sensory attribute can be studied; Uniform distance between two consecutive rankings; Only study the relative nature of the rank; No indication of degree of difference; Results are sample group dependent. Modified from (Poste et al., 1991) Sensory Evaluation in Packaging Applications Sensory attributes are inherently crucial parts of the overall quality of consumer products and processed food products, which are almost definitely contained and protected in some forms of packages. Naturally, sensory evaluation is used in packaging applications as a complimentary and helpful tool to study the off-odor and off-flavor of various products due to packaging materials. Different from other instrumental analysis, subjective results are obtained by the panelists in a sensory evaluation to disclose the liking, preference, and acceptance by consumers. Maneesin (2001) stored water samples in HDPE containers and glass bottles for 6 months and evaluated the off-flavor of the water by a pre-screened and trained panel, using the Difference-from-Control Test (6-point ranking test with a reference sample). The results helped the researcher investigate the effect of different packaging materials and additives such as antioxidants on the flavor quality of water samples. Das (2003) used an untrained sensory panel in his study to rank water samples based on their perceived off-flavor after the water samples were stored in direct contact nine different food grade HDPE resins at 40 :I: 2°C for one week. Sensory evaluation results were used to provide subjective analysis of HDPE resins, in addition to the objective analysis of those resins using an electronic nose system. Chung (2004) focused her work on the off-flavors of reduced fat milk and Cheddar cheese that was exposed to light oxidation. The effect of packaging materials, including glass, HDPE, HDPE-TiOz, PET, and PE-coated paper cartons, was also investigated. Sensory evaluation was used to study the perceived off-flavor using both consumer and trained panelists. Chung’s work proved that sensory evaluation is a powerful and complimentary tool to other instrumental analysis such as GC-MS. 65 Besides being used as an analytical tool, sensory evaluation is also used in packaging design to improve the quality of packaged products (Anon, 2006). Figure 2.9 (Lawless 8 Heymann, 1998) explains how a sensory evaluation department interacts with other functional groups in a food or consumer product company in the process of research and development, quality control, packaging design, etc. A Legal Sensory _ . Servrces EVa'uatlafl k] Packaging 8 Design Engineering 8 Process _ Quality / Control Interacting Departments: U.S. Foods and Consumer Products Industries Reprinted from (Lawless 8 Heymann, 1998) Figure 2.9 Role of sensory evaluation department in a food or consumer product company 66 ELECTRONIC NOSE Gardner and Bartlett defined the electronic nose (E-nose) as “an instrument, which comprises an array of electronic Chemical sensors with partial specificity and an appropriate pattern recognition system, capable of recognizing simple or complex odors” (Mielle, 1996). Since the first discovery by Taguchi in 1971, electronic sensors have gained more and more attention due to their superior advantages, which include consistent, objective, nondestructive, prompt responses and lower running cost compared to a trained sensory panel (Mielle, 1996). Later, some researchers found that arrays of these sensors are capable of recognizing complex food aroma systems (Bartlett et al., 1997). By now, the electronic nose has been widely used in the food, personal care, cosmetics, tobacco, and automobile industries, and, of course, in packaging applications (Chung, 2004; Culter, 1997, 1998; Gardner 8 Persaud, 2000; Harper, 2001; Haugen, 2001; Hurst, 1998; Maneesin, 2001; Mielle, 1996; Siripatrawan, 2002; Siripatrawan et al., 2004a, 2004b,2006a,2006b) Table 2.17 lists some commercially available electronic nose systems and their sensor technologies. 67 Table 2.17 Various electronic nose systems and their sensor technologies Electronic Nose System Sensors Alpha MOS Metal oxide sensors (6 — 18 sensors) Aroma Scan Organic polymer (32 sensors) Bloodhound Organic polymer (6 — 12 sensors) Moses ii ’ Combination array (multiple sensors) Hewlett Packard Headspace with mass selective detector Neotronics Organic polymer (12 sensors) Perkin-Elmer Quartz crystal sensors Modified from (Harper, 2001) Electronic Nose and Human Nose In the human nose, a vast number of Chemical sensors are activated and react with the compounds when the odorous compounds enter the nasal cavity. This process is called data acquisition. The chemical reactions are transformed into a photoelectrical signal, which is transferred to the brain via the chemosensory nerves in the nasal cavity and the central neural system (Boudreau, 1983). The signal is analyzed and sorted by the brain, which is called data processing. By comparing the odor with the stored Information in neurons, we can recognize an unknown simple or complex odor system. 68 Recognized as BRAZIle .9 77'. ‘ é \_ u / l . l 7‘ i ‘ Data Comparison ot'thc odour RF LT % [Processing] [ by the neural system ] \| Raw signals I I Processed signal I Neural Networks fr " .IO Recognized as , . BRAZILIAN " _ . _, l , Reprinted from (Alpha—MOS, 1999) Figure 2.10 Recognition of Brazilian coffee by the E-nose and the human nose The term electronic nose originated from the fact that it works in a similar way to the human nose. The electronic nose recognizes an unknown odor system via the process of data acquisition, data analysis, comparison and recognition (see Figure 2.10). Electronic chemical sensors resemble the chemical sensors in the human nasal cavity. Pattern recognition is similar to that occurring in the brain of a human being. Moreover, the electronic nose does not separate the Individual components in the odor system. Instead, similar to the human nose, it 69 recognizes the odor system via a global analysis by comparing the unknown sample with the database, which is input into the electronic nose system during the training period. According to a report by Harper (1998), sensitivity of the sensors used in the Alpha MOS Fox 3000 electronic nose system is comparable to that of the human nose (see Table 2.18). Table 2.18 Comparison of detection threshold values of several Chemical compounds in water determined by an electronic nose system (Alpha MOS Fox 3000) and human nose ‘ Compound Fox 3000 Human Nose Ethanol . < 25 ppm 24.9 - 100 ppm 1-octen-3-ol 10 ppb 1 ppb -1ppm Octanal 1 ppm . 0.7 - 0.8 ppb Butyric Acid 1 ppm 50 ppb — 2.15 ppm Diacetyl 100 ppb 4 - 15 ppm p-mesol ' 50 ppb 55 - 680 ppb Nookatone 1 ppb , 1 ppb - 1ppm Modified from (Harper 8 Kleinhenz, 1998) Electronic Sensors Because most electronic sensors only offer partial specificity, an array of sensors is commonly used (Bartlett et al., 1997; Mielle, 1996). The electronic sensor array is the heart of the electronic nose system. The various sensors in the array are used to create an olfactory picture of the system by reacting with the odorous compounds. There are three main types of sensors that are 70 commercially available and widely used. These are M.O.S. (Metal Oxide Sensors), conducting polymer sensors, and Q.C.M. (Quartz Crystal Microbalance) sensors. Metal Oxide Sensors Oxidizing gas Lead -\ _L Electrode I ' Electrode eVs With presence Heater I' Stintered tin oxide eVs in air of an oxidizing gas Ceramic former m Quartz Crystal Microbalance Conducting Polymers duaing polyme Change Of Coatin Isolation layer frequency I W — g I EI-md- - —IG°”I 0:0:9:o:9:9:9:o — Summ- Quartz o...o.o.o.o.o.o ’9’... o o o 9’ Electrode PM. ’c’o’o‘o‘c’¢°o’ v — / l—l 20p Reprinted from (Alpha-M.O.S., 1997) Figure 2.11 Different types E-nose sensors Deventer and Mallikarjuman (2002a) published a comparison of performance of three electronic nose systems based on three different sensors: Alpha M.O.S. Fox 3000 system based on metal oxide sensors, Cyranose 320 system based on conducting polymer sensors, and QMBG system based on quartz crystal microbalance sensors. They concluded all three systems were 71 capable of differentiating film samples with different levels of retained solvents. However, metal oxide sensors and conducting polymer sensors demonstrated better discriminatory ability. Metal Oxide Semiconductor Sensors (M.O.S. Sensors) Metal oxide sensors were firstly discovered by Taguchi in 1971 (Mielle, 1996). They are the most commonly used ones nowadays (Aishima, 1991a, 1991b; Tan et al., 1995). The detection principle is based on the measure of the variation of the resistance of the sensors (Culter, 1999). These sensors are made of a ceramic coated with a semi-conducting film, usually in a thickness of 50 pm (Alpha-MOS, 1997). Selectivity of the sensors toward different chemical compounds can be obtained by either doping the film with noble catalytic metals or by modifying the working temperature of the sensing element in the range of 50-400°C. In addition, the selectivity can be affected by the particle size of the polycrystalline semiconductor (Alpha-M.O.S., 1997; Mielle, 1996). The resistance of the sensor is changed with the reaction of an odorant with the sensor, which generates a measurable electronic signal. The magnitude of the response depends on the nature of the volatile molecules and the type of metal oxide. 72 These sensors offer good sensitivity in the ppm or even ppb range to a very wide range of Chemical compounds (Table 1). They are relatively resistant to humidity and to aging. However, they can be easily poisoned by sulphurous compounds and weak acids and they are extremely sensitive to ethanol, which could “blind” the sensors to any other analyte of interest. Caution is needed when high molecular weight compounds are analyzed because these compounds result in slow baseline recovery (Alpha-M.O.S., 1997; Culter, 1999; Mielle, 1996). Conducting Polymer Sensors Conducting polymer sensors are made of a thin layer of polymer across the gap between electrodes via electro-polymerization (Mielle, 1996). Such conductive polymers include polypyrroles, polyanilines, and poly(3- methylthiophenes). It was found that doping the polymer with various ions could enhance its conductivity (Culter, 1999; Vetenskapsakademien, 2000). Similar to M.O.S., the adsorption of the volatile compoUnd into the polymer will change the polymer’s resistance (Culter, 1999; Mielle, 1996). 73 Table 2.19 Comparison of different sensors used in electronic nose systems Metal oxide sensor Low-medium selectivity High sensitivity (ppb-ppm) Low to medium sensitivity to humidity Low temperature dependence Medium desorption time Fast recovery time Long lifetime (18 — 36 months) Conducting polymer Medium to high selectivity sensors Good sensitivity (ppm) High sensitivity to humidity High temperature dependence Long desorption time Slow recovery time Shorter lifetime (6 - 9 months) Q.C.M. sensors Medium to high selectivity High sensitivity (ppb-ppm) Dependent on humidity (coating) Moderate temperature dependence Quick desorption process Slow recovery time Reasonable lifetime (9 — 12 months) Modified from (Alpha-M.O.S., 1997; Haugen, 2001) Their main advantages are that they operate at room temperature and in some cases they are more selective for certain compounds. However, they are not good for high temperature applications because the inherent shift with temperature is very high. Moreover, they are very sensitive to humidity, which means external variables such as temperature and humidity must be controlled in using these sensors (Bartlett et al., 1997). The analysis time could be extended dUe to long response time (Culter, 1999; Mielle, 1996). Compared to metal oxide sensors, conducting polymer sensors are expensive and have poor reproducibility. 74 The success of the analysis using conducting polymer sensors is critically dependent on the choice of polymer sensors. In addition, the choice of appropriate signal preprocessing and subsequent Choice of pattern recognition technique is of paramount importance, which could mean either success or failure for a particular application (Bartlett et al., 1997). Motion et al (2004) used an electronic nose system based on conducting polymer sensors to identify wood chip samples of various species. It was found the system was capable of differentiating different samples but this capability was affected by the ratio of mixtures if two different wood chip samples were mixed and analyzed with the electronic nose system. Gameau et al (2004) used the same type of electronic nose system to successfully differentiate wood samples of different species. Quartz Crystal Microbalance Sensors (Q.C.M. Sensors) The sensing element for this type of sensor is a coated quartz resonator. The sensor is made by depositing a gas-sensitive coating such as silicone or polyglycol onto a quartz support. The volatile compounds can absorb or adsorb onto the coating upon exposure, which leads to a change of mass of the crystal and correspondingly a Change of the frequency of the oscillation. 75 The sensitivity and selectivity of the Q.C.M. sensors can be controlled by the coating material and the quantity deposited. Compared to the other two electronic sensors, this type of sensor is less used (Culter, 1999). Deventer and Mallikarjunan (2002b) analyzed volatiles released from printing inks used in 9 different food packaging films using an electronic nose system based on an array of 6 resonating quartz sensors. Pattern Recognition Systems In most cases, an odor system is composed of hundreds of volatile compounds and each of them will react with each sensor in the sensor array to a different extent. Thus, the response of each sensor is the integrated result of reactions with every volatile compound contained in the odor system. An array of sensors will thus give a global profile of the odor system. The database created for analysis is generally presented as a matrix of n rows and p columns with each row standing for one analyzed odor sample (objects) and each column equivalent to one sensor response (variables) (see Figure 2.12). This database is created by measuring the interaction of the electronic nose with the odor samples. It is used for comparison and identification purposes with the help of a pattern recognition system. 76 aI aI ai aI,4 ai.5 31.6 q] ai,8 aI,9 aIJO al.” 3|,” ali alfl 32.3 a2.4 82.5 all) a2.7 %.8 82.9 aZJO a2,“ aZJZ Objects M A A A A A A A A A A M anti alrLZ arrlj an-i.4 all-1.5 arr-HI and] affix aft—L9 an'IJO anti] aIT‘IJZ a anS ané n7 anti an9 anl0 an,” an.2_l _C11501 12 Figure 2.12 Data matrix created by analyzing n samples with an electronic nose system with an array of 12 sensors The most popular pattern recognition systems include principle component analysis, discriminative function analysis, partial least squares, and good-bad analysis. Principle Component Analysis (PCA) One easy way to discriminate two samples is to visually compare the sensor response profiles for these two samples. As mentioned above, a sensor array will give a matrix that iS difficult to represent in the three-dimensional world. Principle component analysis is used to reduce the number of variables by linearly combining the original variables with maximal variance. In other words, 77 the technique seeks a dimension along which the observations are maximally separated or Spread out (Alpha-M.O.S., 1999; Krzanowski, 1988; Rencher, 1995). Then it is possible to compare two samples by projecting the original data (matrix of n x p) into a two or three dimensional space, with the Chosen principle components (the linear combination of original variables) being the vectors (see Figure 4). By doing so, we can assess the Similarities between samples or the relation between sensors. Sample] Samplez , C2 e3" Figure 2.13 Differentiate two samples with a PCA with three principle components in an electronic nose analysis 78 Discriminative Function Analysis (DFA) Discriminative function analysis is a dimension reduction technique similar to principle component analysis. However, a mathematical model is developed and used to classify unknown samples into one of the training sample groups. In order to do so, the first step involves defining new variables, the linear combination of the descriptive variables, which has the Characteristic of separating as much as possible the different groups of samples. Then the technique can be used to place a new unknown sample into the known training groups by using computation of the distance between the centers of gravity between different groups (Alpha-M.O.S., 1999). Culter (1999) reported work using discriminative function analysis to Classify different paperboard samples and study the effect of humidity on the odor profile of paperboard. Partial Least Squares (PLS) Partial Least Squares (PLS) is a quantitative technique. Two types of quantitative information are related to volatile compound analysis. One is the concentration of specific compounds in the analyzed sample. The other is the sensory evaluation score. The objective of PLS is to build a model with which 79 such quantitative results can be obtained through the sensor response profile in the E-nose system (Alpha-M.O.S., 1999). PLS is based on linear regression analysis. If Y is the matrix of the database of the quantitative measurements (e.g. concentration of a specific compound or sensor score), Y' is the matrix of the corresponding prediction values, and X is the matrix of database from the sensors’ response, then partial least squares is the technique to find a matrix B that is used to predict the quantitative values by minimizing the distance between Y and Y’ with Y’ = XB. The scientists at Alpha M.O.S. used the electronic nose Fox 4000 system to build a mathematical model to predict the concentration of peppermint flavor in drinking water in the range of 0.01% to 0.50% (Alpha-M.O.S., 2001). A high correlation coefficient of 0.97 was obtained and the model was validated with two different samples. A similar study was conducted at the School of Packaging at Michigan State University to predict the concentration of nonanal, which was claimed by the investigator to be the main odorous compound in drinking water stored in HDPE containers. A correlation coefficient of 0.99 was obtained (Maneesin, 2001). 80 Good-Bad Analysis This technique is used to optimize a model and then use this model to Classify the unknown sample as either good or bad product. This technique needs an extensive training process in which all the possible differences between samples due to formulations, production conditions, and batch-to-batch variations must be taken into consideration and are presented to the electronic nose system during the training process (Alpha-M.O.S., 1999). Alpha M.O.S. used this technique to study the quality of the raw material used in producing cherry fragrance during a food treatment process (Alpha- M.O.S., 2001). Six batches of samples were Classified as acceptable and the technique was found successful in recognizing unknown good samples and unknown bad samples. APPLICATIONS OF THE ELECTRONIC NOSE SYSTEM There have been various applications of the electronic nose system in the fields of food, cosmetics, automobiles, environment, pharmaceuticals, packaging, etc. (Bartlett et al., 1997; Chen 8 Cu, 2001; Chung, 2004; Culter, 1999; Maneesin, 2001; Mielle, 1996). 81 Maneesin (2001) analyzed the off-flavor compounds in water stored in two different plastic containers made of HDPE with two different antioxidant system, using an electronic nose system, sensory evaluation, and GC-MS. Siripatrawan (2002) investigated the possibility of using an electronic nose system to detect and identify pathogens including E. Coli and Salmonella growing on various food systems. Chung (2004) used an electronic nose as a tool to investigate the quality of light-oxidized reduced fat milk and cheddar Cheese. Off-flavors were analyzed with the electronic nose system as well as other traditional analytical methods such as GC-MS. The off-flavors were also evaluated with human sensory evaluation and the results were correlated with the sensor responses of the electronic nose system. Haugen (2001) listed a variety of food applications using the electronic nose system, including odor analysis, lipid oxidation, freshness and spoilage, taints and of-flavor, food safety, process control, and packaging. At Ohio State University, Harper and his team (2001) worked with three electronic nose systems, Aroma Scan, Neotronics, and Alpha MOS Fox 2000, to study flavors and quality of various dairy products such as milk, butter, cheese, and whey protein. 82 ELECTRONIC NOSE AND OTHER ANALYSIS TECHNIQUES There are often misconceptions about the electronic nose technology. One is that the electronic nose is a replacement for sensory evaluation. Another is that the electronic nose is superior to the traditional separation techniques such as gas chromatography (GC), and thus can replace the latter. As we will discuss here, neither is valid. Electronic Nose and Sensory Evaluation The electronic nose is pictured by some people as similar to the human nose. From some standpoints, they are right because the electronic nose system can do some of the jobs a human nose does, such as recognizing an unknown odor system. A human nose can make subjective judgments based on personal life experience. An individual can recognize Brazilian coffee because he has encountered it in the past and stored the “image” of the coffee odors in his brain. In a similar way, an individual can judge whether an unknown sample is good or bad quality because he has built up a database from his past experience, which helps to make such subjective judgments later on. Human olfactory systems can be made very sensitive and reliable by a pre-screening and training process. A trained panel can give precise and 83 consistent results. For the last half century, evaluation of odor and flavor of food and food packaging systems has been done by organoleptic evaluation technology, which gives complex sensations via the interactions of human senses and the tested items (Lawless 8 Heymann, 1998; Meilgaard et al., 1999; Poste et al., 1991). However, panelists are prone to off—days, colds and hay- fever. Their judgments are always subjective and susceptible to their body health and mood. Besides, human beings are not suitable in the case of harmful chemicals. Most importantly, it takes time and it is costly to train an effective panel and run human sensory evaluation tests. On the other hand, running the electronic nose system is lower cost, and it offers objective judgments, prompt response, consistent results, and comparable sensitivity to the human nose in some applications. However, it must be noted that the electronic nose itself only can make objective judgments. Sometimes, subjective judgment is preferred or required. The untrained electronic nose system cannot judge which sample is good or which is bad. As discussed earlier, PCA can decide whether two samples are similar or not, without any training. But in order to identify an unknown sample (DFA), or decide the sample is good 'or bad (Good-Bad Analysis), or determine the concentration of a specific compound in the unknown sample (PLS), the electronic nose system must be trained in advance. In other words, the system must be told which training samples are good or bad, what they are, or the concentration of the compound of interest. Human sensory evaluation can help the training process of the electronic nose system by offering information such as good or bad judgments, or sensory scores based on the perceived sensory characteristics of the samples. Such information can be used for the purpose of DFA or PLS functions in the electronic nose system. Willing et al (1998) studied the odor from paperboard samples with an. electronic nose. The correlation between the results from the electronic nose system to the results from human sensory evaluation was built with a PLS model. A similar approach was used in Maneesin’s and Chung’s work (Chung, 2004; Maneesin, 2001). Electronic Nose and GC-MS Some people think the electronic nose system is superior to 60 analysis. The fact is the electronic nose system does not and cannot separate a compound mixture into individual components. As mentioned earlier, the electronic nose system acquires the global odor profile, based on which the pattern recognition system is applied. 60, on the other hand, is used to separate the mixture into individual components. 85 With the help of gas chromatography-mass spectrometry (GC-MS), two types of information can be obtained and used in the electronic nose system. First, volatile compound(s) contributing to the objectionable odor may be identified. Moreover, the concentration of volatile compound(s) can be determined. Such information can be used in DFA or PLS in the electronic nose system. Actually, the electronic nose technique, human sensory evaluation, and GC-MS can be correlated with each other to get a more comprehensive analysis of an odor system. Sensory evaluation offers subjective judgment and quantitative results (sensory scores) and GC-MS offers qualitative as well as quantitative results. Such information can be used for training purposes for the electronic nose system. After the training process, the electronic nose is ready for quick, easy and objective analysis. Moreover, the results obtained with GC-MS can be used in the sensory evaluation test. Identification of major compounds and their concentration can help in the training process by selecting the appropriate training standards. Figure 2.14 demonstrates the idea of correlation between these three analysis techniques. 86 Sensory Panel Modified from (Hodgins, 1997) Figure 2.14 Correlation of the E-nose nose, sensory evaluation, and GC-MS Forsgren et al (1997; 1999) used GC, an electronic nose system, and human sensory analysis to study the volatile compounds released from two food packaging board products. Data obtained from GC was processed with multivariate analysis and was correlated satisfactorily with the results from the electronic nose analysis and human sensory evaluation. It was concluded that the electronic nose system could be potentially used as an on-line instrument for monitoring the quality of board products. Heinio et al (2002) studied volatile compounds from printed paperboard samples using an electronic nose system, GC-MS, and sensory evaluation. PLS (Partial Least Square) models were built to correlate the sensory evaluation data with the sensor responses obtained from the electronic nose. 87 CHAPTER 3 ANALYSIS OF HEADSPACES OF HDPE FILMS USING ELECTRONIC NOSE INTRODUCTION AND OBJECTIVES Adhesives are widely used in food packaging, especially in primary packaging applications such as sealing lidding stock, paperboard/seal material/flange, sealing pouches and bags, and bonding susceptors in microwave packaging (IOPP, 1995). In the past, focus has been on the adhesive’s compliance with FDA regulations, which emphasize the'safety of adhesives in indirect contact with food. Little attention was paid to the contribution of adhesives to off-odor and/or off- flavor of food packaging systems. Predictably, identification and characterization of such volatile compounds could help us to resolve the problem, improve the formulation of the adhesive, and optimize the conditions in which the adhesive will come into contact with food. In the last half century, evaluation of odor and/or flavor of food and food packaging systems has often been done by organoleptic evaluation techniques, which give complex sensations via the interaction of human senses and the tested item (Poste et al., 1991). Human olfactory systems can be made very sensitive by pro-screening and training procedures. Trained panels can provide precise and consistent results, but they are prone to off-days, colds and hay- 88 fever. Their judgments are always subjective and human beings are not suitable in the case of harmful chemicals. More importantly, it takes time to train an effective panel and running human sensory evaluation tests is costly (Poste et al., 1991) Since the first discovery by Taguchi in 1971, electronic chemical sensors have gained more and more attention due to their superior advantages, which include consistent, objective, non-destructive, prompt responses and lower running costs compared to a trained sensory panel (Mielle, 1996). Researchers have found that arrays of these sensors are capable of recognizing complex food aroma systems (Merrnelstein, 1997). So far, the electronic nose has been widely used in the food industry, personal care, cosmetic industry, tobacco industry, automobile industry, and packaging applications (Mielle, 1996). Gardner and Barlett defined the electronic nose (E-nose) as “an instrument, which comprises an array of electronic chemical sensors with partial specificity and an appropriate pattern recognition system, capable of recognizing simple or complex odors” (Mielle, 1996). Using an electronic nose makes it possible to detect the pattern of responses associated with off-odor and off-flavor as the result of adhesives used in food packaging systems, without requiring Identification of the individual components. Furthermore, if a relationship between the response of the E-nose 89 and the objectionable odor and/or flavor (presence and/or concentration) could be established, the E-nose system could then be used as an efficient and effective on-line QC measure. The objectives of this part of the research were to: 1) develop and optimize the data acquisition parameters so a meaningful and processable sensor response pattern can be obtained for the analyzed sample headspace; 2) determine the response patterns of odor associated with adhesives used to coat HDPE films using an Alpha MOS Fox 3000 electronic nose system; 3) differentiate the four adhesive formulas and the base film with PCA (Principle Component Analysis); 4) develop a DFA model and validate the model by Identifying “unknown” HDPE film samples. MATERIALS AND METHODOLOGY Four different pressure-sensitive adhesive formulations were used to coat the control sample HDPE film, which were all provided by National Starch, Inc. Table 3.1 Codes and formulas of adhesives and base film Code from NSC* Polymer Tackifier Code used in the study 11753 - 36A A A PATA 11753 — 368 B A PBTA 11753 - 360 B B PBTB 1 1753 — 45A C A PCTA 7613 - 48 - 1 I / CONT (HDPE base film) * National Starch 8 Chemical 90 More details of the formulations are: 11753-36A: rubber block copolymer, food grade rosin ester, mineral oil, and antioxidant; 11753-363: different rubber block copolymer, same food grade rosin ester as 1 1753-36A, mineral oil, and antioxidant; 11753-36C: same polymer as 11753-36B, food grade terpene, mineral oil, and antioxidant; 11753-45A: rubber polymer (random architecture), same food grade rosin ester as 11753-36A, mineral oil, and antioxidant. Figure 3.1 is a pictUre of an Alpha MOS Fox 3000 electronic nose system with HS-100 auto-sampler, which has an array of 12 sensors (see Table 3.2), and MOS sensors. 91 I ‘—_-—-——- HS-100 headspace auto-sampler Alpha MOS sensors Agitato . ‘ _ - . '- Injection - - y__ — _p_o_rt Sample vial Sensor chamber (Reprinted from Alpha MOS picture database, with permission) Figure 3.1 Alpha MOS Fox 3000 E-nose system with HS-100 auto-sampler and MOS sensors Table 3.2 Sensor array used in Alpha MOS Fox 3000 E-nose system Set Sensor 1 Sensor 2 Sensor 3 Sensor 4 Sensor 5 Sensor 6 1 SY—LG SY—G SY—AA SY-gH SY-gctL SY—gcT 2 T30/1 P1 0/ 1 P1 0/2 P40/ 1 T70/2 PA2 Even though MOS sensors generally have low to medium selectivity, each Sensor does have specificity to certain compounds (see Table 3.3). 92 Table 3.3 M08 sensors and their specificity to organic compounds Sensors 7 Specificity to organic compounds Applications T30l1 Polar compounds ethanol Liquors, beers P10/1 P10/2 Hydrocarbons such as methane and Cooking, roasting of SY-AA propane coffee, petro-Chemistry SY-gcT P40l1 Fluorinated and chlorinated Environment, SY-LG compounds, aldehydes packaging T7012 - Alcohol compounds, food aroma and Petro-chemistry, volatiles natural aromas, coffee PA2 Alcohol, solvents Alcoholic perfumes, SY-gctL fermentation SY-G » Amines, amine containing compounds, Various and ammonia derivatives SY-gH AICOhol and aromatic compounds Paint, polymer ' (toluene) industry, smoke detection Modified from (Alpha-M.O.S., 1996) As shown in Figure 3.1, the Alpha MOS Fox 3000 E-nose system consists of three major parts: an agitator, a HS-100 auto-sampler, and the sensor Chamber. HDPE samples were cut and sealed into 10ml glass vials, which were loaded onto the sample tray. Glass vials were transported individually and consecutively by the HS-100 headspace auto-sampler to the agitator, where the sample vial was heated at a pre-spec'rfied temperature for a pre-specified period 01' time before a sample syringe took a sample of the headspace from the vial 93 and introduced it into the injection port automatically. The sample vial was then taken back to the sample tray and a new sample vial was moved to the incubator for the next analysis. Volatile compounds from the headspace were transferred into the sensor chamber by the carrier gas (zero-grade air composed of 80% N2 and 20% 02), where they reacted with the MOS sensors, whose resistances changed due to the reactions. Variations of resistances were then used to generate measurable electric signals, which were recorded and processed by the software of the E-nose system. The generated response-pattem for each headspace analysis by the Alpha MOS Fox 3000 E-nose system was a plot that was composed of 12 curves. Each curve represents the variation of the sensor resistance with time, calculated as AR/Ro, where AR equals (R — R0), and R0 and R are the sensor resistance att = 0 and time t, respectively. Because the compositions of the headspaces (the odor systems) of different samples vary both qualitatively and quantitatively, the resultant E-nose sensor response patterns will be different as well, based on which the headspaces and thus the original samples can be differentiated from each other and/or recognized, if the E-nose system is pre-trained. 94 RESULTS AND DISCUSSION Parameter Optimization for the E-nose System Justification and Criteria The first step of using the electronic nose system is always developing the data acquisition method (Culter, 1999). The most important parameters are incubation temperature, incubation time, acquisition time, delay, and sample size. Incubation temperature and time are the temperature and time the sample vial is held in the agitator. Acquisition time is the time the E-nose system records generated Signals. Delay is the time after the last analysis and before the next headspace sample is injected into the E-nose system, which is necessary for the sensor responses to come back to the baseline. These parameters should be studied and selected based on the following considerations (Alpha-M.O.S., 1996): 1) Return of sensor responses to the baseline is rapid and stable; 2) Able to acquire the peak (maximum) of the response curve; 3) Sensors are not saturated (front panel values of the sensor readings do not reach 0 or 4095); 4) Sensors react sufficiently (minimum of 5% and maximum 90% AR/Ro). 95 Initial Settings of Acquisition Parameters Table 3.4 lists the initial settings of the data acquisition parameters for the E-nose system, including all the crucial parameters as well as minor parameters, which were based on recommendations by Alpha MOS (Alpha-M.O.S., 1999). The incubation temperature was set at 100°C based on the potential applications of the adhesive-coated HDPE films proposed by National Starch, Inc. Table 3.4 Initial settings of the data acquisition parameters for the E—nose system Crucial Parameters Incubation time (s) 600 Incubation temperature (°C) 100 Acquisition time (s) 1140 Delay (5) 0 Sample size 1 strip of 10" x 1” HDPE film Minor Parameters Agitation speed (rpm) 500 Syringe type (ml) 5 Fill speed (pl/s)1 500 Syringe temperature (°C) 5 degrees higher than incubation temperature Flushing time (s)2 120 Vial type (ml) 10 Injection volume (pl)3 5000 Injection. speed (ml/min)3 2500 Acquisition period (s)4 1 Flow (ml/minf 300 1. The speed at which the syringe takes the headspace sample from the vial; 2. The time the syringe flushes the remaining headspace sample out of itself after the injection; 3. The volume and the speed that the sampled headspace is injected into the E- nose system; 4. The frequency at which signals of sensors are recorded; 5. The flow rate of the carrier gas. 96 “Delay” was set at zero so the total run time equals the acquisition time, which made it possible to visually observe profiles of the sensor responses all through the analysis. 02-re -22-- -- - -22 -22- 22-2 d~~~~~~ O E c: < ,1?» l Pfi;»\ 01- j” “e i '\\ 7 “r :TNh-hm f Fhu‘x “-» NH .7-¥ mtg-34:..E; '0'1 'l l I I I I l I I I l l 0.0 103.0 200.0 300.0 400.0 500.0 500.0 700.0 800.0 300.0 1000.0 1140.0 (original data was generated on 02/22/01) Figure 3.2 Sensor response patterns of the 12 E-nose sensors In analyzing sample PBTA under the acquisition parameters listed in Table 3.4 Figure 3.2 is the sensor reSponse pattern for the headspace sample from film PBTA. It was observed that an acquisition time of 600 seconds was long enough so that all the peaks of the sensor responses would be recorded. However, a total run time (acquisition time plus delay) of 1140 seconds was not long enough for the sensors to go back to the baseline. 97 Moreover, intensities of some sensor responses for some film samples were lower than 5% (data not shown), indicating insufficient reaction of those sensors with volatile compounds to give a reliable detection. Based on these observations, the acquisition time was increased to 1500 seconds and the sample size was doubled in each glass vial to 2 strips of 10" x 1" HDPE films. Effect of Sample Size Table 3.5 lists the modified data acquisition parameters based on the initial test results discussed in the previous section. Table 3.5 Modification to the data acquisition parameters for the E-nose Crucial Parameters“ Incubation time (s) 600 Incubation temperature (°C) 100 Acquisition time (s) 1500 Delay (s) 0 Sample size 2 strips of 10” x 1" HDPE films * All the minor parameters were same as those listed in Table 3.4 “Delay" was set as zero again but the total run time was extended to 1500 seconds to ensure all the sensors would have enough time to go back to the baseline before the next headspace sample was injected for analysis. Moreover, sample size was doubled to get higher sensor response intensities. 98 AR/Ro x‘m - en a 7 _ fire—:‘i ‘— 7 ha-S‘...‘ r... Time (S) 0.0 250.0 500.0 750.0 1000.0 1250.0 15000 (Original data was generated on 03/13/01) Figure 3.3 Sensor response patterns of the 12 E-nose sensors in analyzing sample PBTA under acquisition parameters listed in Table 3.5 As shown in Figure 3.3, a run time of 1500 seconds was long enough for the sensors to go back to the baseline. Figure 3.4 is the side-by-side comparison of the sensor response patterns of PBTA with different sample sizes. By doubling the sample size, intensities of sensor responses were increased to satisfactory levels, though not doubled. 99 02_ _. Incubation time: 600 s '3‘“. Acquisition time: 1500 or 1140 s \. Delay: 0 l s a AR/Ro , (Original data were generated on 03/13/01 and 02/22/01 respectively) Figure 3.4 Effect of sample size on the sensor response patterns of the 12 E- nose sensors In analyzing sample PBTA Effect of Incubation Time Another concern was whether the incubation time of 600 seconds was long enough for equilibrium to be established between the HDPE film sample and the headspace inside the glass vial. There is no easy way to predict the time required to reach equilibrium. However, the effect of the incubation time on the sensor response pattern can be investigated using PCA (Principle Component Analysis). Two groups of samples were analyzed with the E-nose system. One was incubated for 600 seconds and the other for 1200 seconds. Each group consisted of five different HDPE film samples (see Table 3.1) and each film sample had three duplicates. 100 All the samples (2 groups x (5 samples/group x 3 duplicates/sample) = 30) were analyzed on the same day and presented to the E-nose randomly to eliminate the effect of sensor drifts. The thirty samples generated thirty profiles of the sensor responses, based on which a library was built and analyzed with PCA. Figure 3.5 shows the side-by-side comparison of the sensor response profiles of film PBTA under incubation times of 600 seconds and 1200 seconds. Sample size: 2 strips of films Acquisition time: 600 s 0_2_ Delay: 900 s r:\\ Incubation time: 1200 s o 3’ \\ E - Incubation time: 600 s I \ E Q i 5597—55. . __ . 3"“qu / \2, IA. ~« ‘r: I v a “a: H.._ k r \\ I “Q.- " «~— A.\ . ., '/ \§"‘~—:___:h:_ _ ——‘ _ Time (8) (Original data were generated on 04/05/01) Figure 3.5 Effect of incubation time on the sensor response patterns of the 12 E- nose sensors in analyzing sample PBTA 101 As shown in Figure 3.5, sensor responses were almost doubled with the doubled incubation time, indicating the latter played an important role in the formation and thus the composition of the headspace of the HDPE film sample. The effect of the incubation time was further investigated with the PCA module of the E-nose system (see Figure 3.6). The library consisted of the 30 profiles of the sensor responses corresponding to the 30 samples analyzed by the E-nose system. Each triangle in the PCA plot represented one group of HDPE film samples with three duplicates, which were pre-defined by the investigator before applying the PCA module to the library. For example, the triangle tagged with “PBTA040501-2-20" was headspaces of 2 strips of film samples of PBTA that were generated by being incubated at 100°C for 20 minutes (1200 seconds) in the agitator of the E-nose on April 5, 2001. The database in the library was a 30 x 12 (objects x variables) matrix, with the row and the column representing 30 samples and 12 sensors, respectively. The x-axis and y-axis were the primary and secondary components of the principle components, which accounted for 97.31% and 1.73% of the total variance, respectively. Ten (10) groups of samples were separated from each other in the PCA plot. That along with a Discrimination Index (DI) of 78 indicated the PCA module differentiated the headspaces of the same HDPE film generated at different incubation times, eg. the control base film at 600 seconds and at 1200 seconds were detected as two separate groups. The PCA plot confirmed 102 the significance of the incubation time in optimizing the data acquisition parameters. By balancing the real-life situation and the requirement for reliable E-nose analysis, an incubation time of 1200 seconds was determined to be more appropriate. 0015— ‘ 2 "RT—\marwmsorzz 0&01040 i . p.31... . ._ HIM-020 F. 01115- emanasm 210} ' t g omfimotfism-g-zo . PCTA 2 ‘ l I 220 l l -- 070 122110 l i . El $0 L ftrmezn I i I 'O.M- ' 'T 7' ' 7 ‘ 3 3 3 ’i ’3 T 0010- 2 ‘ = 3 . 3_ a ‘ 3, ; I 1 Psz-10 1 :3 ~ ‘ . ; ‘ I 0.015— 3 ', z N ‘ JAE®LU ’ } -D_Dm— »‘ ‘ l l ‘ \“Qv Y ‘ 2t. l 025 2 . 2 l l I l l ; a I 010 0'00 0'05 0'04 002 0.00 0.02 0.04 0.05 0.00 0.10 0.12 0.i4 0.15 0.13 0.20 01:97.31: (Original data were generated on 04/05/01) Figure 3.6 PCA of HDPE film samples after different incubation time 103 Sensor Response Pattern of Control Sample During the investigation, it was found the intensities of the sensor responses for the control sample, base HDPE film, were always low and sometimes even lower than 5%. 0.0a i Incubation time: 600 s l -. ,3 Acquisition time: 1140 or 15005 cl 1" .‘ -’ ( Delay:0 5‘ ‘ F Model 1 03 5506.06 53.46 4.04 <.0001 Sample 4 551.18 137.79 10.42 <.0001 Panelists 99 4954.88 50.05 3.78 <.0001 Error 396 5237.74 13.23 Corrected Total 499 10743.80 Intensity , V Source DF Sum of Squares Mean Square F-Value Pr > F Model 103 3484.93 33.83 3.34 <.0001 Sample 4 873.52 218.38 21.54 <.0001 Panelists 99 261 1.42 26.38 2.60 <.0001 Error 396 4014.27 10.14 Corrected Total 499 7499.20 As shown in Table 4.6, P-values for the samples (treatments) in analyzing for both the acceptability and intensity were below a (= 0.05), indicating there were significant differences among the mean acceptability scores, as well as among the mean intensity scores, of the sensed odor of the five HDPE film samples. 132 Table 4.7 summarizes the pair-wise comparison of the average acceptability and intensity scores of the odor of the samples perceived by the panelists based on the Tukey—Kramer test (see Appendix 5). HSD (Honestly Significant Difference) base on a RCBD can be calculated as: HSD = qa,dfe,a=0.05 se(_ _ ) . _ ., J— y.] y.j 2 where q is the critical statistic in the Studentized range distribution (Ott & Longnecker, 2001) for a total number of treatments a, a degree of freedom of error dfe, and or = 0.05; and se (y. j - y, ,~) is the standard error between any two average sensory scores based on treatments. HSD for the acceptability and intensity scores was calculated as: HSDacceptability= -3——-‘/§_6 X 0.5143=1.40 HSDi 3 86 ><.O 4503: 1.23 ntensity= 7— The difference between two average sensory scores must be higher than its appropriate HSD value to be considered significant. 133 Table 4.7 Paired comparisons of the acceptability scores and the intensity scores of the sensed odors of the five HDPE films Control PATA PBTA PBTB PCTA Acceptability 10.22 a 9.893 9.07 a 9.623 7.25” 2.933 6.36b Intensity 3.01 a 2.91 a 3.80 a Note: Based on the significance level of a = 0.05. Different letters on the same row indicate significant statistical differences. As shown in Table 4.7, any two scores carrying different superscripts indicate a significant difference between them. Othenrvise, there was no proved significant difference between them. There was no significant difference among the acceptability scores of CONT, PATA, PBTA, and PCTA but the acceptability score of PCTA was significantly lower than others. Acceptability PBTA 9.07 PATA 9.89 PCTA 7.25 l o 7.5 1 15 Not at all CONT 1022 v ch e mu acceptable PBTB 9-62 acceptable Figure 4.4 Average acceptability scores of the sensed odor of samples As shown in Figure 4.4, four samples’ average acceptability scores were found to be above the central point of the scale and in the range of 9.07 - 10.22 These four samples were CONT, PATA, PBTA, and PCTA. By comparing their 134 distance to the central point of the scale to the HSD value (1.40), their sensed * odor profiles were considered “acceptable” by the panelists. On the other hand, the average acceptability score for sample PCTA was 7.25 and very close to the central point of the scale, indicating that opinions of the panelists to this sample’s acceptability were neutral to slightly unacceptable. In other words, the panelists neither found the odor of PCTA acceptable or unacceptable. Table 4.7 shows sample PCTA was perceived by the panelists as having a significantly more intense odor, which agreed with the conclusions drawn in the pairwise ranking test. Intensity PBTA 3.80 CONT301 l PCTA 6.36 I . I I l m C I l o If 7.5 15 Extremely PBTB 2.93 Extreme” weak PATA 2.91 strong Figure 4.5 Average intensity scores of the sensed odor of samples Moreover, Figure 4.5 shows that the average intensity scores for sample CONT, PATA, PBTA, and PBTB were determined to be in the lower range of the unstructured scale (2.91 - 3.80), while the average intensity score for sample 135 PCTA was 6.36. However, all five samples were perceived as having weak odor profiles by comparing their distance to the central point to the HSD value (1.23). These results brought up a very interesting issue: the correlation between the intensity and the acceptability of the sensed odor. The evaluations of “acceptability” and “intensity" of the sensed odor profiles of sample CONT, PATA, PBTA, and PBTB agreed with each other very well. Namely, the sensed odor was considered acceptable if its intensity was perceived as weak. However, sample PCTA was different. ltwas evaluated as “has a weak intensity of the sensed odor" but the panel also evaluated it as “has neutral and slightly unacceptable sensed odor". This further proved the two questions asked in the quantitative affective consumer test were appropriate and effective because it helped disclose information that was not addressed in the first sensory evaluation test, the painrvise ranking test. In terms of odor analysis, the threshold values of various volatile compounds as well as the interactions among them all affect the perceived odor profile as a whole. It was suspected there was some volatile whose threshold 136 value was low in sample PCTA and thus the sample’s odor profile was perceived as “neutral to slightly unacceptable” . Effectiveness and Efficiency of RCBD As shown in the SAS program output type II SS (Sum Square) in Appendix 5), the p-values of “judge” (panelists) were found to be < 0.0001 , indicating using each panelist as one block in the RCBD was an effective way to control the variability. Again, this approach was employed because each panelist had different sensitivity and accuracy in evaluating the sensed odor of the HDPE film samples. Moreover, the efficiency of the blocking can be measured by E, which was defined as (Ott & Longnecker, 2001) : 2 E— O'CRD "' 2 O'RCBD where the two variance terms, ocaoz and Cacao?“ are the residual variances from CRD (Complete Random Design) and RCBD (Randomized Complete Blocking Design), respectively. It is used to indicate how many more replicates would be needed for a CRD design to get the same standard error of any treatment mean difference as that in a RCBD (Ott & Longnecker, 2001). It can be estimated by using: 137 E = (n -1)MSBL + n(a —1)MSE (na —1)MSE in which n is the number of blocks; a is the number of treatments; MSBL is the mean square of blocks; and MSE is the mean square of errors. Using the ANOVA table for the acceptability scores as one'example, E of RCBD was estimated as (see Table 4.6 for the data): (100—1)x50.05+100><(5—1)x13.23 ~ 2 E: (100x5—1)x13.23 indicating that about 2 times more replicates per treatment would have been required if CRD were used for this sensory evaluation test. 138 CHAPTER 5 ANALYSIS OF VOLATILES USING DH-TD COUPLED WITH GC-MS INTRODUCTION AND OBJECTIVES As discussed earlier, concentrations of volatile compounds in an odor system are usually too low to be detected directly with most analytical instruments, which makes a sample preparation step necessary to separate, purify, and concentrate analytes of interest from a sample matrix. Many sample preparation techniques are available, among which dynamic headspace — thermal desorption (DH-TD) has been widely used to concentrate analytes from various sample matrices, including air, water, soil, food, pharmaceutical, and packaging materials. Compared with other sample preparation techniques, DH-TD eliminated the use of organic solvents. Instead, a multi-bed adsorbent tube is used to adsorb volatiles released from the sample matrix, which is then heated to a high temperature very quickly to release the concentrated volatile compounds. There have been many published papers about using Thermal Desorption coupled with GC—MS to identify volatile compounds. Esteban (1993) determined volatile compounds for different plants. Werkhoff and Bretschneider (1987a; 1987b) presented the principles and application, and the effect of sampling and 139 desorption parameters on recovery using dynamic headspace gas chromatography. All of them used the thermal desorption technique to collect and concentrate volatiles before GC-MS analysis. Sunesson et al (1992) used the multivariate optimization method to determine the experimental conditions for direct thermal desorption and gas chromatography. Maneesin (2001) used the multi-bed adsorbent tube Carbotrap 400 from Supelco to collect volatiles from HDPE bottles and from water samples stored in the HDPE bottles for 6 months. In collecting volatiles released from HDPE bottles, 0.19 of HDPE flakes were incubated at 100°C for 2 minutes in a thermal stripper before the GC analysis. In studying water samples, 250 ml of water was heated at 100°C and flushed with pure helium gas at a flow rate of 100 ml/min for 10 minutes so that the volatiles were carried away and trapped by the desorption tube in the thermal stripper. Parameters mentioned in the sample preparation and introduction to the GC column included carrier gas, temperature, and time. Kanavouras (2003) used two different desorption tubes, Tenax-TA and Carbotrap-300, in studying volatiles generated from fresh olive oil and oxidized olive oil samples. Investigated parameters in the sample preparation and analysis included temperature, time, carrier gas, and the ratio of headspace/sample in the sample vial. 140 In Chapter 3, it was found the odor profiles generated from five different HDPE film samples (one control and four coated with different adhesives) were differentiated by the electronic nose system. In Chapter 4, sensory evaluation determined that sample PCTA had significantly stronger odor compared with the other four HDPE samples. This chapter reports use of the dynamic headspace - thermal desorption (DH — TD) technique to collect volatile compounds released from the HDPE film samples before the compounds were analyzed and identified by a GC-MS system. MATERIALS AND METHODOLOGY The same HDPE film samples used in Chapter 3 were used (see Table 3.1): the control sample HDPE film (CONT) and four film samples coated with different adhesives (PATA, PBTA, PBTB, and PCTA). DIP (Direct Insertion Probe) Analysis Two different film samples were studied: the control sample CONT (HDPE base film) and sample PCTA (the base film coated with adhesive PCTA). A sample of film measured 0.8 cm x 0.6 cm was cut and inserted into the heating tube in the DlP instrument located in the Mass Spectrometry Center at Michigan 141 State University. The film sample was heated (temperature started and stayed at 30°C for 2 minutes, increased at 16°C/min to 300°C, and was kept at 300°C for 2 minutes) to generate the odor system. The generated gaseous phase was ionized by the El (Electron Impact) method, and ions were scanned and recorded by the mass spectrometer JEOL JMS-AX505 H (Jeol USA, Peabody, MA). The mass scan range was 45 to 750 Th (Thompsons, m/z ratio). Selection of Thermal Desorption Tube Two types of thermal desorption tubes were used: Carbotrap 400 and Carbotrap 300 (11.5 cm long and ID. 4 mm) manufactured by Supelco (Bellefonte, PA) (see Figures 5.1 and 5.2). Sample Flow Desorption Flow _____, 4——— Carbotrap F Carbotrap C Carbotrap B Carboxen-569 Adsorbent Bed Adsorbent Bed Adsorbent Bed Adsorbent Bed II II I\\ // Glass Frit Silanized Glass Wool Figure 5.1 Carbotrap 400 multi-bed thermal desorption tube (Supelco, 1998a) 142 Sample Flow Desorption Flow —, *— Carbotrap C Carbotrap B Carbosieve S-Ill Adsorbent Bed Adsorbent Bed Adsorbent Bed I I I Silanized Glass Wool Figure 5.2 Carbotrap 300 multi-bed thermal desorption tube (Supelco, 1998a) The thermal desorption (TD) tubes were pre-conditioned at 350°C under inert helium gas flow for 8 hours before use. Two strips of 10” x 1” HDPE films were contained in a 9-ml stripping glass vial, which was then placed into a Dynamic Thermal Stripper Model-1000 (Dynatherm Analytical Instruments, Inc., Kelton, PA). Helium gas was flushed into the vial while the vial was heated. Volatile compounds were released and carried away by the helium gas and trapped in the thermal desorption tube that was attached to the exhaust port of the Stripper (see Figure 2.2). Upon the completion of the stripping process, the TD tube was transferred to a Dynatherm Thermal Desorption Unit Model-890 (see Figure 2.4) (Dynatherm Analytical Instruments, Inc. Kelton, PA), where the tube was heated to a pre- 143 specified high temperature in a short period of time and volatile compounds were released to the GC column through a heated transfer line (1-meter long nickel tubing of 1/16" diameter that is connected to the fused silica capillary column via a 1/16" stainless steel union). The CO was a HP 5890 Series II GC (Hewlett Packard, Philadelphia, PA) equipped with a fused silica capillary column SPB-5 (30 m x 0.32 mm ID, 1 pm thickness of coating) from Supelco (Bellefonte, PA) and an FlD (Flame Ionization DetectOr). Integration of the chromatography peaks was performed by an HP 3395 integrator (Hewlett Packard, Philadelphia, PA). DH-TD GCIMS The same procedures described in the previous section were followed to collect and concentrate volatile compounds released from the HDPE film samples using a Carbotrap 300 TD tube from Supelco (Bellefonte, PA). Trapped volatiles were separated and identified by a GC/MS system, HP G1800B GCD Plus system equipped with a fused capillary column HP—5 (30 m x 0.25 mm ID, 1 pm thickness of coating) and an EI mass quadrapole detector and software GCD Plus ChemStation (G1074B version A.01.00) from Agilent Technologies, Inc. (Palo Alto, CA) in the Department of Chemistry at Michigan State University. 144 Ideally the TDU (Thermal Desorption Unit) and the GC system with a capillary fused silica column should be connected through a nickel transfer line, which enters the GC oven through an accessing opening on the top of the oven instead of through the injector (see Figure 5.3). The transfer line outside the TDU valve compartment should be wrapped by a heating jacket. Moreover, the heated portion of the transfer line should extend inside the GC oven by V.” - ‘/z” to avoid cold spots along the line (Dynatherm, 1989). 1116" Stainless Steel Ferrules 1116” Nickel To valve Low Dead ‘ transfer line volume Heated L portion of transfer Extend fused silica 2-3” inside nickel tubing Detector 1I16" Vespel Graphite ferrule with 1116” hole 1116" sis union 1116" Vespel Graphite ferrule with appropriate hole for column 0.0. Reprinted from (Dynatherm, 1989) Figure 5.3 Connecting nickel transfer lineof TDU to GO 145 Due to the logistics of the GCD system, this connection was not allowed. A modified connection had to be adopted to connect the TDU with the GCD system. The solution was to insert one deactivated fused silica tubing inside the nickel transfer line of the TDU. One end of the silica tubing was connected to the valve behind the thermal desorption tube inside the TDU valve compartment and the other end was connected to a custom-made aluminum connector tube (1/8" OD.) which had a flat skirt on one end and sat on the top of the septum inside the injection port. This custom-made aluminum connector was then screwed tight and held in place by the septum retainer nut (Agilent catalog# 18740-60835). The extended part of the connector tubing was connected to one end of a Swagelock union and locked in place by a ferrule, and the other end of the union was connected to the transfer line’s heating jacket and locked in place by another ferrule. The fused silica line inside the heating jacket and nickel tubing went all the way through the Swagelock union, the custom-made connector tube and GC injection port, and then extended another 7-8” into the injector (see Figure 5.4). 146 Nickel heating jacket Fused silica line Swagelock union 4 Septum retainer nut . Custom-made aluminum connector Ill LI Septum V Figure 5.4 Modified connection between TDU and GCD system 147 RESULTS AND DISCUSSION DIP (Direct Insertion Probe) Analysis DIP analysis was used as a quick and simple technique to compare the odor systems generated from sample CONT and PCTA. Figure 5.5 shows how the data was collected to get the IIC (Individual Ion Current) Chromatogram and RTIC (Reconstructed Total lon Current) Chromatogram. Based on the RTIC chromatograms of the control sample CONT and sample PCTA (see Figure 5.6 and 5.7),.it was concluded that different odor systems were generated from the two film samples. More volatiles were released from sample PCTA because the average abundance of TIC (Total Ion Current) was much higher than that of the control sample. Moreover, maximum abundance of the control sample happened around RT (Retention Time) 1 - 3 min (corresponding to scan number 100 - 300), while for sample PCTA, maximum abundance happened around RT 5 — 8 min (corresponding to scan number 300 — 500), which further indicated the compositions of these two odor systems were different from each other. 148 Time << I—— Intensity mlz Add up intensity (ion current) in each scan file yielding a total ion current (TIC) IIC (Individual Ion Current) for a TIC vs time -) RTIC particular mlz value vs time -) Mass (Reconstructed Total Ion Chromatogram of mlz 73 for example Current) Chromatogram E 9 h I— 5 O . o: c O 2 O . O I L. Time Time Figure 5.5 TIC and RTIC chromatogram from mass spectrometer 149 Abundance 188 Max 3432.01 1.0- 15 RT '1 Mag. Abund. N I I 4 80‘ , 60* 4 40« 4 20‘ *1.B 3432.81 430 EEC (:63 15:38 1208 Scan Figure 5.6 RTIC chromatogram of sample CONT in DIP analysis Abundance 108‘ Max 3597.40 1.0 RT 2,, Mag. Abund. I 88‘ TIC ME) -I’”. I. '1 .0 359?. 49 203 458 see 552: 1669 1288 Scan Figure 5.7 RTIC chromatogram of sample PCTA in DIP analysis Two sets of IIC chromatograms were obtained. One included mass charge ratios (mlz) of 57, 71, 60, 210, 237, 310, 338 and 450 (see Figure 5.8 and 5.9), and the other featured m/z of 173, 239, 266, 357, 620, 634 and 659 (see Appendix 6 and 7). The high mlz values could be the result of high temperatures the samples were exposed to during the analysis (up to 300°C). More attention should be given to the lower mlz values because they originated from volatile compounds. Direct comparison of the scan numbers cannot be made due to the difference of the starting time in the two scans in Figures 5.6 and 5.7. Max 250.073 RT 8 Mag. Abund. s 148 1,5 A a L I TIC . (11.0 3432.81 ‘ 858 1#131.9 1.8962 x45.8 5.4564 Abundance ['39.6 6.3176 18.1 38.9295 t4.5 55.2981 7'1.5.3 16.3955 61.3 194.366 . _ . x1 .0 256.073 3 206 4293 660 266 1006 1260 Scan Figure 5.8 IIC chromatogram of sample CONT in DIP analysis (set 1) Max 162.726 RT Mag, Abund. *1.9 3597.48 )”9.5 2.073? '35.5 4.5859 Abundance '32.? 4.9426 *8.‘ 19.2626 *3.3 78.086! *8.8 20.3185 ‘1.4 115.198 :1.0 162.726 _‘6 4130 662 see mm 1260 \1 Scan Figure 5.9 IIC chromatogram of sample PCTA in DIP analysis (set 1) 151 Comparing the first set of IIC chromatograms of samples CONT and PCTA, it was noticed that the elution time of individual mlz value was generally shorter for CONT. For both samples CONT and PCTA, the IIC profile of mlz 60 was different from those of mlz 57 and 71, indicating that mlz 60 might have originated from different compound(s) while m/z 57 and 71 could be fragment ions from the same source compound(s). This was true for mlz 210 and 237 as well, indicating these two fragment ions were from different source compounds. All these indicated that compositions of the odor systems generated from sample CONT and PCTA were different from each other. The presence of alkanes with various chain lengths in the odor system from sample CONT was confirmed in the mass spectra of scans number 117, 155, and 181 (see Figure 5.10, 5.11, and 5.12). These mass spectra all showed low mass ion clusters of mlz 57, 71, 85, and 99, which were separated from each other by a mass of CH2 group (mlz 14).and among which the base peak happened at mlz 57 (C4H9"). Lots of molecular ions (even mlz ions) were detected. For example, mlz 310 in the mass spectrum of scan number 155 could be the C22H45+ molecular ion. The presence of ion mlz 210 in the mass spectrum of scan number 117 was considered unusual because it could not be a molecular ion of alkane, which indicated that secondary fragmentation could have occured. 152 100 5? . I i 80‘ 71 C , > m in 26” ° ‘ :2 3 1 > g 9 7 I Q—l D LD ‘10 . 0 I 0 b D: l 2162132 . .1 .. J m 300 350 ABC mlz Figure 5.10 Mass spectrum of scan #117 of sample CONT in DlP analysis in... SF' . I I 881 71 8 ‘ I C 1 to < as 'o c 60-1 . 3 1 .o < I Q) q .2 1°. .g-a I <0 7;, mac . m .- f 222--. A A. . - .I 108 220 302 490 $08 602 mlz Figure 5.11 Mass spectrum of scan #155 of sample CONT in DIP analysis 153 .— 'C 71 $19.0 Relative Abundance A Q L3 IzAALAYA -v . 2.2.. a 188 208 398 408 566 600 Figure 5.12 Mass spectrum of scan #181 of sample CONT in DIP analysis For sample PCTA, the mass spectrum of scan number 196 (see Figure 5.13) had a base peak at mlz 181 while the mass spetra of both scan number 916 and 1163 (see Figure 5.14 and 5.15) had a base peak at m/z 256 and all other mass spectra had a base peakat mlz 57, indicating that these fragment ions originated from non-alkane volatile compounds. In the mass spectra of scan number 613, 660, 916 and 1163 (see Figure 5.16 and 5.17), the abundances of mlz 239 and mlz 256 were unusually high, which were not considered characteristic of the mass spectra of alkanes either. All these observations pointed out that the volatiles released from sample PCTA were different from those from sample CONT in terms of both composition and concentration, resulting in different characteristic fragment ions in the mass spectra in the DIP analysis. 154 100‘ 131 804 8 I 496 c L L 3 “I 5 3 I C I g 5 I 3 Bl 2,.8 i 4 2‘. I .7 I r a I 57 J - . r 76 j I *18.8 , i3 26} 97 1 2 r a: , 1:9 » 4 i I 2 7 b I ' . . I 'I I I” -I " I ' l ‘l‘ 1 4.‘ v—fl _l v—u 166 266 366 406 566 cm mlz Figure 5.13 Mass spectrum of scan # 196 of sample PCTA in DIP analysis 166 1 255 6:6 659 I, . s: . 66- L . 71 <0 ‘ .f g < I N 60‘ 85 I' '0 ‘ 17" I C . _ 5 3, 3 9' 239 '. < 49 .‘I 0 I .2 n 11 :- ,_, I I i.“ I l a) 26+ . 9 i I. . I (.IH‘Ii-isxr SKA]: _ . 3 3 . ; 6 ' ‘ ' .._i 160 266 366 466 566 666 766 mlz Figure 5.14 Mass spectrum of scan # 916 of sample PCTA in DIP analysis 155 108 2 6 L , 66 . 8 i c ‘ ' (u 60‘ 173 . E ‘ r 3 u 2 46 57 z I9 I 9 7‘ I * 6:: . (D I 85 185 i 375 b 6 284 I 97 9 'i 7 ' tr ' 1 11 1 9 % 1?? 2 3 I > ‘ i . i I i Z S I" ’ é .1; i. 611:5. lat-9 :l..'." r." .‘I'Iéwtn Lu. . ..-.. 1' .mlII. . «1;. ..-."... -... . ' n. ._ . 21:. "a 3 7 I : 56 :66 156 266 256 366 356 mlz Figure 5.15 Mass spectrum of scan # 1163 of sample PCTA in DIP analysis 106 5, 1 I I CI. 66- ’1 - 8 I 97 . g go. 93 .. E . I 3 I . 1'! b .0 I i < 46" 1:5 '- 9 I ‘ 'fi; 2 l 137 Z '03 26+ I . (I H 15 7 29 26 . m I Q ‘ I 1 9 1 3 I I Q I " I l I~¥ ' I z 6 G .I I ‘ ' I II .1: II LI ..IIII Ii I .II II I£.‘I;II .I .II'?“-t 56 166 156 2.0 256 366 mlz Figure 5.16 Mass spectrum of scan # 613 of sample PCTA in DIP analysis 156 108 r 5? _ r 89‘ . 8g 71 g ‘ I 97 U . g} 68 95 .0 < 1 1 l G) 40 x 3 I | 1 s .r ‘6 -; — l : 0 ‘ 1 1 7 ;' l ‘r t l 15 g .. i ‘l‘ u. ‘ I - 4 . -.« l . I I I: ii i .l l I ll. ll . . l l. 53 130 152 208 250 309 Figure 5.17 Mass spectrum of scan # 660 of sample PCTA in DIP analysis Selection of Thermal Desorption Tube Two types of thermal desorption tubes, Carbotrap 400 and Carbotrap 300, were used to collect and concentrate volatile compounds in the headspace of the HDPE film samples. Based on the comparisons of their performance and the repeatability of the results, one of the two tubes would be selected for the subsequent analysis. 157 Table 5.1 lists the parameters used in the DH-TD and GC analysis in order to choose the appropriate thermal desorption tube. Table 5.1 DH-TD and GC analysis parameters Dynamic headspace preparation Helium flow rate Temperature and time Sample size Thermal desorption (path B) Helium flow rate Temperature and time (for Carbotrap 400) Temperature and time (for Carbotrap 300) Desorption tube recovery (path A) Helium flow rate Temperature and time (for Carbotrap 400) Temperature and time (for Carbotrap 300) Other temperatures Transfer line Valve 60 Helium flow rate Detector (F lD) . Temperature (for Carbotrap 400) Temperature (for Carbotrap 300) 40 cc/min 100°C for 20 min 2 strips of 10” x 1” HDPE films 6 CC/min 310°C for 8 min 260°C for 6 min 2.5 cc/min 320°C for 25 min 350°C for 60 min 290°C 250°C 1.0 cc/min 300°C 40°C for 5 min, increased at 5°C/min to 220°C and then kept for 10 min 40°C for 8 min, increase at 10°C/min to 100°C and then at 6°Clmin to 220°C, kept for 5 min Table 5.2 lists the results of the DH-TD GC analysis when Carbotrap 400 was used to collect volatile compounds generated from analyzing triplicates of sample CONT. 158 Table 5.2 Major peaks identified for sample CONT in the DH-TD GC analysis with Carbotrap 400 Replicate Total number of RT (min) of major Area (%) of the peaks identified peaks peaks* ’ 12.216 1.79 13.774 1 .89 1 77 16.707 20.06 19.169 31.69 21.565 2.44 18.877 2.03 24.559 20.97 2 83 29.585 - 32.45 33.950 6.72 18.851 2.46 24.483 18.13 3 74 29.500 29.34 33.785 ' 2.52 * Peaks whose percentage of area was more than 1%. Results in Table 5.2 showed a low repeatability in terms of the distribution and intensity of peaks obtained in the gas chromatography analyses. Repeatability is important in the DH-TD GC analysis because it is the first step and the basis of the separation and identification of volatile compounds. Many factors affect the results of DH-TD GC analysis, one of which is the selection of the appropriate thermal desorption tube. Carbotrap 400 is usually suitable for analyzing aqueous samples while Carbotrap 300 is for gaseous samples (Supelco, 1998a). They consist of different adsorbent materials with different physical Characteristics such as mesh 159 size, surface-area, and pore size (see Figure 5.1 and 5.2). Carbotrap F in Carbotrap 400 has a larger pore size and thus a smaller surface area, which makes it suitable for trapping larger molecules. Similarly, the Carboxen-569 used in Carbotrap 400 has a larger mesh size and thus smaller particle size than the Carbosieve used in Carbotrap 300, making the former more suitable for applications involving larger molecules. The use of adsorbents designed to capture higher carbon-Chain analytes, along with the fact that aqueous samples are more likely to have higher concentrations of the larger compounds than air samples, makes Carbotrap 400 better suited for aqueous samples. Volatile compounds in the headspace generated from HDPE film samples were to be studied and thus Carbotrap 300 seemed a better candidate for the thermal desorption process. 160 Tables 5.3 to 5.7 list the major peaks and their area percentages for sample CONT, PATA, PBTA, PBTB, and PCTA when Carbotrap 300 was used. Table 5.3 Main peaks and their area percentages of sample CONT in DH-TD analysis using Carbotrap 300 Component Replicate 1 Replicate 2 Replicate 3 RT (min) Area (%) RT (min) Area (%) RT (min) Area (%) 1 6.635 1 .06 6.834 0.29 6.495 0.46 2 10.730 0.67 10.803 0.33 10.688 0.26 3 12.525 0.11 12.557 0.09 12.507 1.10 4 13.211 0.98 13.233 0.63 13.195 0.61 5 14.535 0.21 14.555 0.17 14.536 0.20 6 15.182 ~ 2.17 15.186 1.09 15.169 0.63 7 16.479 0.19 16.472 0.14 16.467 0.09 8 17.141 3.87 17.140 3.58 17.130 3.25 9 21.282 5.98 21.282 5.47 21.281 5.93 10 25.341 2.81 25.344 3.07 25.344 3.07 11 29.134 . 0.26 29.132 0.32 29.129 0.34 12 32.706 0.16 32.675 0.17 32.683 0.22 Table 5.4 Main peaks and their area percentages of sample PATA in DH-TD analysis using Carbotrap 300 Component Replicate 1 Replicate 2 Replicate 3 RT (mm Area (‘79) RT (min) Area (%) RT (min) Area (%) 1 5.898 1.46 6.016 0.27 6.010 1.26 2 10.491 0.85 10.522 0.67 10.524 0.58 3 12.420 0.21 12.429 0.19 12.429 0.18 4 13.127 1.84 13.134 1.90 13.089 1.04 5 14.496 0.33 14.498 0.32 14.499 0.29 6 15.135 1.50 15.134 1.28 15.135 0.84 7 16.514 0.23 16.561 0.49 16.578 0.56 8 17.105 3.13 17.100 3.67 17.097 2.77 9 21.264 4.80 21 .259 5.58 21.259 5.40 10 22.919 1.28 22.911 0.82 22.915 1.49 11 23.554 6.29 23.542 2.24 23.592 2.95 12 25.329 3.04 25.327 4.53 25.335 5.93 13 29.119 0.52 29.116 1.03 29.125 1.75 14 32.696 0.47 32.680 0.47 32.690 0.58 161 Table 5.5 Main peaks and their area percentages of sample PBTA in DH-TD analysis using Carbotrap 300 Component Replicate 1 Replicate 2 Replicate 3 - f RT (min) Area (%) RT (min) Area (%) RT (min) Area (%) . 1 6.299 0.36 6.026 0.93 6.015 1.03 2 10.617 0.53 10.539 0.46 10.535 0.43 3 12.473 0.14 12.440 0.15 12.439 0.12 4 13.094 0.72 13.092 0.89 13.109 1.11 5 13.166 0.76 13.145 0.69 13.146 0.67 6 14.518 0.27 14.506 0.26 14.505 0.23 7 14.748 0.33 14.760 0.54 14.790 0.57 8 15.153 0.91 15.146 0.86 15.147 0.98 9 16.544 0.36 16.599 0.63 16.647 0.84 10 17.107 2.50 17.112 2.25 17.113 3.05 11 21.271 6.68 21.255 4.42 21.254 3.70 12 22.917 0.67 22.910 0.60 22.910 0.35 13 23.597 2.50 23.592 3.04 23.583 0.75 14 25.341 6.61 25.320 2.93 25.319 2.73 15 29.126 1.60 29.116 0.51 29.119 0.58 16 32.701 0.59 32.679 0.34 32.705 0.27 162 Table 5.6 Main peaks and their area percentages of sample PBTB in DH-TD analysis using Carbotrap 300 Component Replicate 1 Replicate 2 Replicate 3 RT 1min) Area (%) RT (min) Area (%) RT (min) Area (%) 1 5.990 0.69 5.957 0.06 5.805 0.16 2 10.525 0.35 10.508 0.31 10.470 0.25 3 12.427 0.09 12.421 0.09 12.407 0.08 4 13.010 0.57 12.996 0.58 13.002 0.63 5 13.135 0.56 13.131 0.51 13.125 0.44 6 14.495 0.16 14.490 0.13 14.489 0.11 7 14.716 0.30 14.710 0.37 14.737 0.39 8 15.133 0.39 15.129 0.36 15.130 0.38 9 16.547 0.34 16.535 0.53 16.597 0.66 10 16.926 0.06 16.943 0.25 16.944 0.24 11 17.098 1.72 17.095 1.64 17.100 1.51 12 21.240 3.06 21.236 3.01 21.240 2.39 13 22.898 0.32 22.894 0.26 22.901 0.17 14 23.587 4.03 23.575 2.43 23.580 1.50 15 25.307 2.37 25.309 3.09 25.316 2.86 16 27.473 2.06 27.474 1.95 27.480 1.63 17 27.720 0.62 27.718 0.59 27.739 0.50 18 27.886 0.66 27.886 0.60 27.890 0.53 19 28.025 1.05 28.015 1.09 28.020 0.92 20 28.289 3.70 28.285 3.61 28.291 3.29 21 28.429 1.27 28.427 1.21 ' 28.441 1.06 22 28.575 1.65 28.573 1.66 28.580 1 .49 23 28.728 2.38 28.722 2.37 28.725 2.17 24 28.890 1.12 28.887 1.15 28.898 1.06 25 29.011 0.73 29.005 0.66 29.015 0.57 26 29.129 1.12 29.132 1.66 29.148 1.71 27 29.390 4.20 29.385 4.20 29.393 3.99 28 29.554 1 .42 29.555 1 .56 29.556 1 .38 29 29.808 2.94 29.805 3.16 29.810 3.21 30 30.080 2.69 30.081 3.17 30.095 3.22 31 30.430 1 .43 30.422 1 .64 30.438 1 .65 32 30.725 0.73 30.719 0.90 30.729 0.95 33 30.817 0.46 30.804 0.54 30.813 0.57 34 31.029 0.75 31.028 1.01 31.038 1.08 35 31.386 0.48 31.393 0.66 31.401 0.73 36 31.590 0.35 31.585 0.45 31.594 0.48 37 31.812 0.11 31.812 0.17 31.818 0.18 38 32.703 0.28 32.684 0.39 32.699 0.34 163 Table 5.7 Main peaks and their area percentages of sample PCTA in DH—TD analysis using Carbotrap 300 Component Replicate 1 Replicate 2 Rglicate 3 RT (min) Area (%) RT (min) Area (%) RT (min) Area (%) 1 1.940 2.30 1.958 13.65 / l 2 6.296 0.36 6.248 0.04 6.408 2.96 3 10.719 0.31 10.682 0.13 10.654 0.73 4 13.190 0.70 13.179 0.36 13.185 1.76 5 14.533 0.18 14.531 0.11 14.553 0.26 6 15.160 1.15 15.153 0.41 15.163 2.36 7 16.235 0.13 16.232 0.08 16.248 0.31 8 17.100 2.67 17.095 1.79 17.116 3.22 9 18.218 5.10 18.215 5.23 18.234 4.05 10 19.782 0.74 19.781 0.51 19.799 0.58 11 21.262 6.30 21.255 4.11 21.270 5.86 12 22.669 2.1 1 22.667 1.65 22.682 1.32 13 22.925 1.32 22.920 0.93 22.939 0.87 14 23.380 0.48 23.376 0.36 23.387 0.43 15 23.553 7.79 23.545 4.63 23.559 2.14 16 25.335 5.52 25.325 3.76 25.345 6.48 17 29.132 2.21 29.125 1.45 29.143 2.37 18 32.710 ' 0.66 32.708 0.30 32.700 0.07 The chromatograms of the five HDPE film samples were different from one another in terms of retention times of the main peaks and their relative abundance expressed as area percentage. Interestingly, sample PBTB seemed have the most complex volatile profile, even though sample PCTA was perceived as having the strongest odor by the sensory panel. By comparing the 60 results with the sensory evaluation results, it proved that the complexity of a volatile system is not always directly correlated to the intensity of the perceived odor, probably because some volatile compounds were not odorous or their concentrations were below their threshold values. 164 DH-TD GCD (GCIMS) Analysis Table 5.8 listed the parameters used in the dynamic headspace and thermal desorption coupled with gas Chromatograph and mass spectrometry. Table 5.8 DH-TD and GC-MS analysis parameters Dynamic headspace preparation Thermal desorption tube Helium flow rate Temperature and time Sample size Thermal desorption (path B) Helium flow rate Temperature and time DeSorption tube recovery (path A) Helium flow rate Temperature and time (for Carbotrap 300) Other temperatures Transfer line Valve GC Helium flow rate Split-less time Injection port temperature Detector (El mass quadrapole detector) Mass range Temperature (for Carbotrap 300) Supelco Carbotrap 300 40 cc/min 100°C for 20 min 2 strips of 10" x 1" HDPE films 6 CC/min at 60 psi 260°C for 6 min 25Cdmm 350°C for 60 min 290°C 250°C 1.0 CC/min at 7.1 psi 2 min 300°C 300°C 45 - 450 mlz 40°C for 8 min, increase at 10°C/min to 100°C and then at 6°C/min to 220°C, kept for 5 min Triplicates of each HDPE film were analyzed (CONT, PATA, PBTA, PBTB, and PCTA) with the DH-TD couple with the GCD system (GC/MS). 165 Repeatability of DH-TDIGCD Analyses To investigate the repeatability of the DH-TD/GCD analyses, triplicates of sample PATA and of sample PBTA were analyzed with clean thermal desorption tubes (confirmed with blank runs to ensure the tubes were clean). The results are shown in Figures 5.18 and 5.19. Chromatograms showed that the TIC profiles of the triplicates were similar but the peak abundances varied significantly. ”0°00 I Abundance f 2.61 90000 : I‘ - I 000005 5 70000{ I 50000{ I E 3 500°C ’ 0000:} 3 2 .96 30000~ ,. r ’. 14.00 200004 I 13~Q‘16. 017. 90 : fl 1517 If 29265.09 20000 3 » ‘\ 11: 3‘ ‘ 3? 91°, 28 L ' 23 s . . : \ ”"1398: I 3 I Trme(min) c5 .12... .,-.3/.\/\.- ”I s..L . 3. . ....-.., .. ,..... -..... . 5.00 10.00 15.00 20. 00 25.00 30.00 35.00 Figure 5.18 Triplicates of TlC chromatogram of sample PATA using DH-TD/GCD analysis 166 Figure 5.18 (Cont’d). . 2 .04 ‘ Abundance 90000 I 000000 700001 I l J l 50000 1 ,I ' i 50000 ‘ 5.00 90000 2.b6 Abundance 80000 70000' l I j l 2 1 l l GOOOOI I 4 4 500°C xfl5/N. ._..-.m-k? 10. 00 14 42 :1 15.02 {I 15.00I' 21:99 25,96 23.50 24.37 I i 5 I 21. 4321150 _. ”I“ ~Nflm M1-~hl.4 \s- ”A- , Time (min) 20.00 25. 00 30. 00 35. 00 103,10 17I91 22.00 119.2: Time (min) 20.00 25.00 30.00 35.00 167 1100004 3, 5 Abundance 100000 90000 .0000 70000 30000 __ w- w 2.. ---.__._——.._ -—..--.l w..- “w-M4-JMH‘..O.LL MM “— 50000 1 .. 10.03 00000; 25.90 I1 17;3§.37 20.37! I ~- ‘6th 4-'~.‘.. A. .._. ‘ 30000 3' l 20000I ' t ‘0...- 1: 21 I. I , \’\ "1 'F, . F .“L . - ime min 0: .i- .,. “HA.“ ,2, 5.403\ . , .‘ ‘ LL54,“ . I01 L ”I . T (. ). 5.00 10.00 15.00 20.00 25.00 30.00 35.00 Figure 5.19 Triplicates of TIC chromatogram of sample PBTA using DH-TD/GCD analysis Figure 5.19 (Cont'd). 100000- 13i" 3 Abundance 16.02 120003: 100000 - 231° . I 30000: 3 7 g; . .3 24.30 00000; 40003: I . I 22.5I25.90 20003. I . I l 2 .0. . . . . Time min 0 .’ .“hI—Aiq\w:wwwi .. I ,. I. 4, wet-‘5, -,. ..~~~ 0 , ,_ ,. 1,“. -.-A( ). 5.00 25.00 30.00 35.00 168 Figure 5.19 (Cont’d). I 2. 9 ”0°03 IIAbundance 700001 I 2 i I .. 000004 I i I 50000+ I 10.04 000003 ' I I ’3 I 2””20 37 g .-16.03 . : 300001 . . I I ' I 17.91 3 %5I 13.90 20000; t I J 25.09 10000 I I I III; 3? 1 “1°! 21' 123.5 I '2 ,.- 1110 I; ?&7;6. 3 . ‘ 7 K I 13 160: ’ I I II I - - 0.: . I; , ,. A43. , CuntIifij’fM I .%-‘a.“; . .I,.I..... -... .‘Au ,. .. .MT!me(.n:“!]2-..- 5.00 10.00 15.00 20.00 25.00 30.00 35.00 The low repeatability of the analysis happened for several possible reasOns. The transfer line of the TDU was recommended to enter the GC oven via an access hole and be connected to the capillary column through the stainless steel union (see Figure 5.3). Making the transfer line go through the injection port was not suggested due to the dead volume of the injection port and poor heat transfer (Dynatherm, 1989). Secondly, two carrier flows enter into the top of the capillary column with the modified connection instead of one single carrier gas source as displayed in Figure 2.5. In the modified connection, one flow came from the transfer line of the TDU, which went into the injection port of the 60 through the custom-made connector (see Figure 5.4), and the other flow came from the carrier gas for the 60 itself. The two carrier gas flows met and collided with each other inside the 169 injection port of the GC. which could have affected the consistency of the carrier gas flow and thus the transfer of volatile compounds to the GC column. Moreover, the split-less injection mode on the GCD system might have affected the repeatability of the results as well. The split-less time is the time at the beginning of an injection when there is no flow out of the injection port vent, during which the entire injected sample is vaporized and allowed to flow onto the 60 column. Because the transfer line entered the GC via the injection port and all thermally-desorbed volatiles were carried over to the GC column from the injection port, the split-less mode and its time setting took over the purge time of the TDU. In the other words, the injection port vent would be turned to open after the set split-less time (maximum 2 minutes), which was controlled by the ChemStation software solely and could not be changed by the control panel on the GC. Even though the purge time of the TDU was set to allow the desorption tube to be heated and flushed with carrier gas for 6 minutes, the effective purge time of the TDU was actually decreased from 6 minutes to 2 minutes. Volatile compounds that were not released from the thermal desorption tube or did not travel to the GC injection port within 2 minutes after the GOD system was started, would have been flushed out of the injection port vent and could not reach the GC column for analysis. 170 Figure 5.20 and 5.21 showed the effect of split-less time on the TIC chromatograms of sample CONT and PCTA. 0 11000’ I Abundance Split-less time 1 min .' I I 10000; I a 13"’15.01 ; I 31.35 9000; . 5 . I I III II . g . 0000; . 7oooI i i 17.90 9 I I 21.99 0000? I . 19:39 I ‘ 35.04 A‘ W —. » I 5000; g 4000i I . I I 17.05 I AJII C“... I . 7. 1 . 3000‘ 3 7 7. 7 11 0* ‘ . ::::; =0" . =3 I; .2... oi . II 3% Ist fwfilfifi k” agnIiz Time ("1'“) 5.00 10. 00 15. 00 20. 00 30. 30 35. 00 flow-z, 1400004 22I00 g Abundance I I f 120000; 16:01 20.41 Split-less time 2 min i I 100000; a 00000; I 60000; a I 1 as... I : 25.90 00000; :0 M 20000. ‘ I : L 11721 W I : - . .. 120 $004.... 1 O 1 .. m\—\ .4 s... .s “3...... -4LM5‘P‘1 ‘ 4* *4. Y ane (min) 5. DC 10. 00 15. 00 20. 00 25. 00 30. 00 35. 00 Figure 5.20 Effect of split-less time on TIC chromatogram profiles of sample CONT in DH-TDIGCD analysis 171 I 1’I13 500000? . . . I Abundance Split-less timei min 050000; 000000; 350000; 14:05 300000? 250000 ; 200000; 16.00 I 150000~I 14.39 I I 100000f |300°C L I 29.51 32.79 c. . _ . . , I ‘ ITime(min) 5.00 10.00 15.00 20.01 25. 00 30.00 35.00 1000000 «3 Abundance I Split-less time 2 min 1000000} 1400000.. ”.1, 12000005 1000000} 000000{ 2‘““ 6042000 i 000000: [25.95 29.55 : 917. 93 200000 - 32.03 2.135 I}, !91 I, . ' - 0..-...40‘-——. ........-., , . «1.-....‘0fiés 17 :‘r .\"~WI- ,I T'meunln) 5.00 10. 00 15. 00 20. 00 25.00 30.00 35.00 Figure 5.21 Effect of split-less time on TIC chromatogram profiles of sample PCTA in DH-TD/GCD analysis Figure 5.22 was the TIC chromatogram of sample PBTB from the DH- TD/GCD analysis. 172 14.39 9 “‘°°°°‘. Abundance . I :I 120000- : I. 1000003 3 22.01 °°°°°j 10.00 _ 13.00 ' 11.91 60000- I! I I I I I I (0000: ‘1 19 29 . I 11.32 ‘1 .0coc- 42. 30 1'05 13.118 10 Is? . , - : , Fifi?“ :1 I 2 ‘° 0. .. . ”\A [LA —~.~ .15.; s. 00 10. 00 1s 00 20 00 29i$3 . Time (min) $25. 03 35. 00 Figure 5.22 TlC chromatogram of sample PBTB from DH-TD/GCD analysis Potential Identities of Volatile Compounds Based on the chromatograms of the DH-TDIGCD analyses for sample CONT, PATA, PBTA, PBTB, and PCTA, identifications of detected peaks were made by matching the spectrum of each peak with the build-in mass spectra library. Several groups of aromatic and non-aromatic compounds, including aldehydes, alcohols, aromatics, and hydrocarbons were found for each sample. The results are listed in Table 5.9. 173 Table 5.9 Potential Identities of volatile compounds from the DH-TD/GCD analyses Sample CONT PATA PBTA PBTB PCTA Aldehydes Heptanal V V V V V Butenal x x x V x Octanal V V V V V Nonanal V V V V V Decanal V V V V V Octadecanal V x x x x Alcohols Octanol V V V V V Aromatics Benzene, 1,3-bis(1,1- x x x x V dimethylethyl) Phenol, 2,4-bis(1 ,1 - V V x x V dimethylethyl) Hydrocarbons . Dodecane V V V V V Tetradecane V V V V V Hexadecane V V x V x Heptadecane x V V x V The compound benzene, 1,3-bis(1,1-dimethylethyl)benzene was only detected in sample PCTA. Earlier sensory tests found sample PCTA had a stronger odor compared to the'other four samples. A hypothesis that this particular compound was responsible for the stronger odor of sample PCTA could thus be made. However, caution must be taken when making such a hypothesis considering the low repeatability of the DH-TD/GCD analysis. Therefore, further investigation was necessary. 174 CHAPTER 6 ANALYSIS OF VOLATILES USING SPME COUPLED WITH GC-MS INTRODUCTION AND OBJECTIVES In Chapter 5, we used dynamic headspace and thermal desorption techniques to concentrate volatile compounds generated from the HDPE film samples. The parameters were optimized but the low repeatability problem could not be solved due to the connection between the thermal desorption unit and the GC—MS, and the split-less injection mode on the GC. However, one particular compound, 1,3-bis(1,1-dimethylethyl)-benzene, was found in sample PCTA only, which was then hypothesized to be responsible for the stronger odor of the sample. This chapter reports on utilization of another sample preparation technique, solid-phase micro-extraction, or SPME, to concentrate volatile compounds and improve the repeatability of the results. Moreover, the human nose was to be utilized to detect and describe the odor profile generated from adhesive-coated HDPE film samples by using an ODO ll sniffing unit. The ODO ll unit works by splitting the eluted flow from the capillary column of a GC to two parts, one of which goes to a sniffing port, where 175 a human nose helps detect problematic odor, and another goes to mass spectrometry for identification. The goal of this component of this research was to break down the odor system of the HDPE film sample into individual components with GC. and then investigate each component’s contribution to the odor system with the CDC ll sniffing unit, making it easier to determine the objectionable volatile compound(s). MATERIALS AND METHODS Three SPME fibers were considered: 100 um PDMS, 65 um PDMS/DVB, and 75 um CAR/PDMS(Supelco, Bellefonte, PA). PDMS fibers are non-polar fibers and preferred for adsorbing non-polar compounds, while both PDMS/DVB and CAR/PDMS fibers are bi-polar fibers and suitable for polar volatiles. Kanavouras (2003) compared the effectiveness of the 100 pm PDMS fiber and the 65 um PDMS/DVB fiber in his study of Volatiles originating from virgin and oxidized olive oil samples. The PDMS/DVB fiber was determined to be able to adsorb more volatile compounds in his study. Chung (2004) compared the 65 um PDMS/DVB and the 75 um CAR/PDMS fibers in her study and concluded that the 75 pm CAR/PDMS fiber was a better candidate for her applications because it adsorbed more ,target volatiles at her chosen temperature. Our preliminary study showed the effectiveness of the PDMS/DVB and the CAR/PDMS fibers were comparable to each other, but the CAR/PDMS fiber 176 tended to have volatile carryover problems. Thus the 65 um PDMS/DVB SPME fiber was chosen in our study. Optimization of SPME Parameters Two strips of HDPE film PCTA (10” x 1") were enclosed in a 20 ml screw top glass vial (Supelco, Bellefonte, PA), which was capped with a' Mininert Valve (Supelco, Bellefonte, PA). The red-green pin on the valve can be easily pushed to close the valve during headspace preparation or to open the valve during the sampling process with SPME. The enclosed glass vial with HDPE film samples was heated at 100°C in an oven for 20 minutes. These conditions were consistent with the conditions used in the E-nose analysis to generate the headspace. After 20 minutes of heating, the glass vial was taken out of the oven and transferred to and kept in an air-conditioned room (25°C) or in a temperature- controlled water bath for 10 minutes, so that the temperature of the headspace was equilibrated with the environmental temperature. After 10 minutes, the red-green pin on the Mininert Valve was pushed open and the SPME assembly was inserted into the glass vial and the fiber was exposed to the headspace for a pre-specified period of time. 177 Table 6.1 lists the parameters that were considered in optimizing the SPME method. Table 6.1 Parameters tested in optimizing SPME method Parameters Settings Headspace generation 100°C for 20 min Temperature equilibrium time before SPME sampling 10 min SPME sampling temperature (°C) 25, 35, 45, 60, 75 SPME sampling timflmin) 5, 15, 30 Upon completion of the SPME sampling, the SPME assembly was transferred and inserted into the GC injection port, where the SPME fiber was exposed to the high temperature in the injection port to release the adsorbed compounds for3 minutes. The released volatiles were carried by the helium gas to the top of the 60 column, where liquid nitrogen was used to cryo-focus the volatiles before the temperature program of the 60 was started. The GC-MS equipment used in the study was located in the Postharvest Laboratory in the Department of Horticulture at Michigan State University. The gas chromatograph was Agilent 6980 equipped with a HP-5MS fused silica capillary column (30 m in length x 0.25 mm ID, with a coating of 0.25 pm in thickness). The mass spectrometer was Leco Peasus ll ChromaTOF (Time of Flight). 178 Table 6.2 lists the parameters used in the SPME GC—MS. Table 6.2 Parameters used in the SPME/GC-MS with constant column gas flow GC Injection mode Splitless Injector temperature 220°C Transfer line 240°C temperature Temperature program 40°C for 0 min, increased to 240°C at 40°/min, stayed at 240°C for 3 min. Column flow Constant flow at 1 ml/min Purge flow 10 ml/min Purge time 10 seconds MS . . _ . Solvent delay 100 seconds Mass range 29 — 270 u Acquisition rate 8 spectra/second Comparisons of Gas Chromatogram Profiles The gas chromatogram profiles of sample CONT, PATA, PBTA, PBTB, and PCTA were compared with one another to help identify the unique compound(s) which could potentially make sample PCTA have a stronger odor compared to other HDPE film samples (based on the results obtained in the sensory evaluation tests reported in Chapter 3). The HDPE film samples were enclosed in the 20 ml glass vial capped with Supelco Mininert Valve and heated at 100°C in an oven for 20 minutes to generate the volatiles. The enclosed glass vial was then transferred to and kept 179 in a water bath set up at 60°C for 10 minutes so that the headspace reached the desired temperature. The SPME sampling time was 30 minutes. The SPME GC-MS parameters are listed in Table 6.2. Description of Odor Profile with Sniffing Test An ODO ll sniffing system (SGE USA, Austin, TX) was connected to the GC/MS system. Figure 6.1 displays how the system was connected to the gas chromatograph. Humidified air Carrier gas tank provides flow for both - ~ ;: , ODO II and GC column (ARRIrIt ms ‘ ' ”AIR To MS (or other detector) Helium gas to sniffing port Auxiliary Helium gas in Flow splitter Modified from (SGE, 2003) Figure 6.1 ODO ll module controls carrier gas to flow splitter and humidified air to sniffing nose cone 180 As shown in Figure 6.1, the ODO ll module controls the auxiliary carrier gas flow from the module to the flow splitter, which provides make-up gas to the sniffing port. Volatile compounds elute out of the GC column and the flow is divided into two paths: one goes to the sniffing port and another goes to the mass spectrometer for identification. The ODO ll module also regulates the flow of air, which is humidified before going to the heated transfer line that is connected to the glass sniffing nose cone. Humidified air is used to avoid drying of the nasal mucous membranes if an analysis runs for a long period of time. Figure 6.2 shows a picture of a sniffing port that is connected to a gas chromatograph, as well as a detailed drawing of the heated transfer line that is connected to the glass sniffing nose cone. 181 Flexible outer stainless steel Flexible robust heater casing coil that closely hugs capillary column Electrical _ connector * . Reprinted from (SGE, 2002) Figure 6.2 Heated transfer line connected to sniffing nose cone in ODO ll module Two adhesive-coated HDPE films were studied. One was sample PCTA and another was PATA. Procedures described in the previous section 182 “comparison of gas chromatogram profiles” were used to prepare the odor system with 65 pm PDMS/DVB SPME. The ODO ll sniffing unit was turned on by turning on the heat and humidified air for the transfer line. A timer was started simultaneously with the start of the GC/MS. The investigator then immediately put his nose by the glass sniffing nose cone. When an odor was perceived at the sniffing port, the investigator recorded the time from the timer and a description of the sensed odor. RESULTS AND DISCUSSION Optimization of SPME Parameters Preliminary studies identified two of the peaks in the gas chromatogram of the odor system of sample PCTA as acetone and 1,3-bis(1,1-dimethylethyl)— benzene (data not shown, identified by the built-in library of the GC/MS). Acetone is a solvent commonly used in plastic processing and it carries a strong and unique “fruity” odor. On the other hand, 1,3-bis(1,1-dimethylethyl)—benzene (also called 1,3-Di-tert-butylbenzene) was found in the DH-TD/GC-MS analysis only from sample PCTA, which had been found to have stronger odor by a 183 sensory panel. Therefore, these two compounds were used as anchoring compounds in optimizing SPME parameters. As explained earlier, it is more important to keep sampling parameters consistent than to reach a full equilibrium before taking samples in SPME analysis (see Figure 2.8). Extreme caution was taken in preparing the odor system of the HDPE film sample and using SPME to adsorb volatile compounds for the GC/MS analysis. The purpose of optimizing SPME parameters was to adsorb as much of the chosen anchoring compounds as possible from the headspace to the SPME fiben Effect of SPME Sampling Time Three SPME sampling times, 5 min, 15 min and 30 min, under a controlled sampling temperature of 35°C were chosen to investigate the effect of time on the quantities of anchoring compounds adsorbed by the SPME fiber. Table 6.3 lists the areas of the peaks identified as acetone and 1,3-di-tert- butylbenzene in the selected ion current chromatograms. An m/z ratio of 58 was chosen for quantifying the peak for acetone and an mlz ratio of 175 was used for 1 ,3-d i-tert-butyl benzene. 184 Table 6.3 Response areas for specific ions of volatile compounds in analyzing sample PCTA with SPME/GC—MS under different sampling times (sampling temperature 35°C) Acetone (mlz 58) . 1,3—Di-tert-butylbenz&e (mlz 175) No. 5 min 7 15 min 30 min 5 min , 15 min 30 min 1 1063757 696802 864362 205372 31 1 186 921 158 2 992068 509095 59361 8 32548 299098 6441 51 3 824543 498742 677928 1 99864 287495 4861 90 Ave. 9601 23 56821 3 71 1969 145928 299260 683833 ratio of 1,3-di-tert-butylbenzene to acetone. Area Response Figure 6.3 shows the peak responses with SPME sampling time and the 1200000 T —L L T 800000 600000 200000 4 .I- DAcetone as 1,3-Di-tert-butylbenzene -°- Ratio of 1,3-Di-Tert-butylbenzene to Acetone 15 0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 TIme (min) Figure 6.3 Average response areas of acetone and 1,3-di-tert-butylbenzene determined by SPME/GC-MS for sample PCTA versus SPME sampling time 185 Ratio As shown in Figure 6.3, a longer sampling time favored the extraction of 1,3—di-tert-butylbenzene by the SPMEfiber but not necessarily for acetone. Even after 30 min of sampling, equilibrium of 1,3-di-tert-butylbenzene had not been established between the SPME fiber and the headspace inside the sampling vial. The data showed 5 min favored the extraction of acetone, probably because other volatile compounds were competing with acetone for the adsorbing sites on the SPME fiber with longer sampling times. The longer the sampling time, the more higher-boiling-point volatile compounds such as 1,3-di-tert-butylbenzene started to compete with lower-boiling-point volatiles compounds such as acetone. As a result, a slight decrease of peak response was observed with acetone from 5 min to 15 min. The increased quantities of adsorbed acetone between 15 min and 30 min along with 1,3-di-tert-butylbenzene might have been due to physical changes of the SPME fiber such as swelling due to the adsorbed volatile compounds. Nevertheless, ratios of the two compounds kept increasing from the sampling time of 5 min to the sampling time of 30 min, indicating a longer sampling time was more favorable to the extraction of 1,3-di-tert-butylbenzne with the SPME fiber. Considering 1,3-di-tert-butylbenzene was of more interest in the study, along with the need for an efficient SPME sampling process, 30 min was chosen as the sampling time for the subsequent SPME analysis. 186 Effect of SPME Sampling Temperature After choosing the SPME sampling time (30 min), the effect of sampling temperature was investigated. Odor systems of sample PCTA were prepared with the 65 pm PDMS/DVB SPME fiber at five different sampling temperatures: 25°C, 35°C, 45°C, 60°C, and 75°C. Tables 6.4 and 6.5 list the areas of peaks identified as acetone and 1,3—Di- tart-butylbenzene, respectively, in the selected ion current chromatograms. Again, an mlz ratio of 58 was chosen for quantifying the peak for acetone and an mlz ratio of 175 was used for 1,3-di-tert-butylbenzene. Table 6.4 Peak response areas for acetone peak in sample PCTA with SPME/GC-MS at different sampling temperatures (sampling time 30 min) Duplicate 25°C 35°C 45°C 60°C 75°C 1 1336234 864362 397145 358949 179310 2 992043 59361 8 385682 2681 36 1 12296 3 1 105876 677928 540259 220958 1 13746 Average 1 144718 71 1969 . 441029 282681 1351 17 Table 6.5 Peak response areas for 1,3-di-tert-butylbenzene peak in sample PCTA with SPME/GC-MS at different sampling temperatures (sampling time 30 min) Duplicate 25°C 35°C 45°C 60°C . 75°C I 1 61 7073 921 1 58 168371 1 3084395 1934817 2 392890 ‘ 644151 1518118 1496782 1451668 3 597791 4861 90 81 7489 2467235 1387441 Average 53591 8 683833 1 339773 2349471 1 591309 187 Figure 6.4 is the average peak response area versus sampling temperature for both acetone and 1,3-Di-tert-butylbenzene in sample PCTA with SPME/GC-MS. sampling temperature between 25°C and 75°C while the peak area for 1,3-Di- ten-butylbenzene increased with the sampling temperature from 25°C to 60°C, As shown in Figure 6.4, the peak area for acetone decreased with SPME and then started to decrease after 60°C. Area Response 3500000 T 12.00 I DAcetone I E331,3-Di-ten~butylbenzene "'00 3000000 "7 .... Ratio of 1 3.01 tert b be t , - - utyl nzene to Acetone 10.00 I 9 00 2500000 I O i 8.00 I 2000000 + 7'00 I 6.00 I 1500000 1 5-00 4.00 1000000 I 3.00 500000 200 1.00 o 0.00 25 35 45 60 75 Temperature (C) Figure 6.4 Average response areas of acetone and 1,3-di-tert-butylbenzene determined by SPME/GC-MS for sample PCTA versus SPME sampling temperature 188 Ratio The different trends observed for the two compounds in Figure 6.4 can be explained by the partition coefficient. The SPME sampling process is essentially a partition process of analytes between the coating on the fiber and the headspace. The amount of a particular compound that can be extracted by a SPME fiber is therefore proportional to its partition coefficient, which is affected by the sampling temperature (Yang & Peppard, 1994; Zhang & Pawliszyn, 1993; Zhang et al., 1994). Partition coefficients for compounds usually increase with temperature at first, and then start to decrease after reaching an optimum temperature. Ratios of the two compounds kept increasing from the sampling temperature of 25°C to 75°C, indicating a higher sampling temperature was more favorable to the extraction of 1,3—di-tert-butylbenzne. However, Figure 6.4 also shows that 1,3-di-tert-butylbenzene started to escape from the SPME fiber after the optimum temperature of 60°C was passed. The higher ratio at 75°C compared to that at 60°C was probably because more acetone escaped from the SPME fiber than 1,3-Di-tert—butylbenzene did at the sampling temperature of 75°C. Because the compound 1,3—di-tert—butylbenzene was found in sample PCTA only, based on the DH-TD/GC-MS study in chapter 5, and sample PCTA was perceived as having a stronger odorcompared to other adhesive-coated HDPE film samples, more attention was given to this particular compound. A 189 SPME sampling temperature of 60°C was thus chosen for the subsequent investigations. Comparisons of Gas Chromatogram Profiles Odor systems of five different HDPE films were studied and compared with each other (samples CONT, PATA, PBTA, PBTB, and PCTA) using SPME/GC-MS analysis with a SPME sampling temperature of 60°C for 30 min. Confirmation of 1,3-Di-tert-Butylbenzene Studies in chapter 5 indicated that the compound 1,3-di-tert-butylbenzene was found in the odor profile of sample PCTA only. Figure 6.5 shows the mass spectrum of the standard compound (NIST, 2005). 1000+ 175 000 - ‘ 57 «I - 3 41 zoo-I . as 91 190 i 29 i 15 51 77 115 125 147 1 20 ‘ID 60 BO 11131201401601002003240260 Figure 6.5 Mass spectrum of standard compound 1,3-Di-teIt-butylbenzene 190 Figure 6.5 shows the mlz ratio of the base peak of 1,3-Di-tert- butylbenzene was 175, which was thus used as the fingerprint peak for the compound. The mlz 175 ion for 1,3-Di-tert-butylbenzne for sample PCTA was observed in the chromatogram at an RT of 271-272 seconds, which was used as the anchoring retention time to confirm whether the m/z 175 ion was observed in the chromatograms for the other samples (CONT, PATA, PBTA, and PBTB). Table 6.6 lists response areas of the mlz 175 ion at the RT of 271-272 seconds if it was detected in the odor system of the HDPE film. Three duplicates were tested for each film and the results are displayed in Figure 6.6. Table 6.6 Areas of m/z 175 ion at retention time 271 - 272 seconds Duplicate CONT PATA PBTA PBTB PCTA 1 9440 2227 2138 2269 3084395 2 5374 3494 2189 3759 1496782 3 4407 3395 2419 2420 2467235 Average 6407 3039 2249 2816 2349471 Obviously, compound 1,3-di-tert-butylbenzene exists in the odor systems of all HDPE film samples, but their quantities in samples CONT, PATA, PBTA, and PBTB were much lower than that of sample PCTA. Among the first four samples, the quantity of the compound in sample CONT was significantly higher than those of the other three samples. 191 The results proved the importance of using the fingerprint peak and the retention time of a particular compound in the chromatogram to detect the compound, which could otherwise not be detected due to its extremely low quantity. 10000 — 9000 4 8000 -1 7000 1 6407 6000 1 5000 - 4000 L U 3039 - 2816 3000 2249 2000 ~’ 1000 CONT PATA PBTA PBTB Figure 6.6 Average area of mlz 175 ion detected in the headspaces of HDPE films except sample PCTA using SPME/GC-MS Origin of 1,3-Di-tert-Butylbenzene Some researchers have found the compound 1,3-di-tert-butylbenzene in polyolefins including LDPE, HDPE and PP after they were exposed to gamma- irradiation (Demertzis et al., 1999; Jean at al., 2004; Krzymien et al., 2001; Lee et al., 2004). 192 Their studies concluded that the additive lrgafos 168, tris(2,4-Di-tert- butylphenyl)phosphite, which is commonly used as a stabilizer in polyolefin processing, decomposed during the gamma irradiation and resulted in volatile compounds including 1,3-di-tert-butylbenzene and 2,4-di-tert-butylphenol. Lee (2004) attributed the differentiation of odor profiles of red pepper powder by the E-nose system after different levels of gamma irradiation (0, 3, 5, and 7 kGy) to the formation of 1,3-di-stert-butylbenzene. Based on their work, it is reasonable to believe the additive lrgafos 168 was used in preparing our HDPE film samples, which decomposed during the odor preparation process. We hypothesized that the odor profile of sample PCTA was perceived as stronger by the sensory panelists because it contained a much higher concentration of 1,3—di-tert-butylbenzene. However, this hypothesis needed to be tested with the sniffing test that is discussed in a subsequent section. Potential Identities of Volatile Compounds The identity of a particular compound can be confirmed by comparing its RT and mass spectrum with those of the standard. As one example, Figure 6.7 shows the mass spectrum of the compound detected at RT 271-272 seconds in 193 analyzing the odor profile of sample PCTA with SPME/GC-MS, which was almost identical to the mass spectrum of its standard (Figure 6.5). 175 57 400‘: 41 200 29 55 91 190 115 51 77 126 147 159 20 40 6O 60 1a: 120 1‘10 160 180 200 220 240 260 Figure 6.7 Mass spectrum of volatile compound detected at RT 271-272 seconds in SPME/GC-MS analysis Table 6.7 lists the retention time of several compounds detected in the odor profiles of the HDPE film samples. Table 6.7 Confirmation of volatile compounds by comparing their retention times with those of standards Acetone‘ Hexanal1 Benzaldehyde2 1014-60-413 Std“ Run5 Std‘ Run'5 Std“ Run5 Std“ Run5 1 1 19.64 121.52 164.97 165.77 205.02 206.02 272.02 271.27 2 1 18.52 120.89 165.39 165.64 205.89 206.02 272.77 271 .89 3 1 17.59 122.14 166.39 166.39 204.39 206.64 271.89 272.02 Aver 1 18.58 121 .52 165.58 165.93 205.10 206.23 272.23 271.72 Diff° 2.48 0.21 0.55 -0.-‘l8 Note: All retention times were in seconds and three duplicates were tested; 1 - The standard compound was tested individually; 2 — The compound was tested in a standard mixture including acetone, ethyl acetate, propyl acetate, butyl acetate, hexyl acetate, hexanal, benzaldehyde, and 1 ,3-Di-tert-butylbenzene; 194 3 — CAS registration number for 1,3-Di-tert-butylbenzene; 4 — Retention time of the standard compound; 5 - Retention time of the compound detected in the odor profile of sample PCTA in the SPME/GC-MS analysis (sampling temperature/time were 60°C for 30 min); 6 - % Difference was calculated as (Run RT - Std RT)/Std RT * 100%. Table 6.8 lists volatile compounds detected in the odor profiles originated from different HDPE film samples. The TIC chromatograms with the tentative identities of compounds are listed in Appendix 8. Table 6.8 Volatile compounds tentatively identified from the HDPE film samples using SPME/GC-MS analysis Sample . CONT PATA PBTA PBTB PCTA Ketones . _ _ , H w ‘ Acetone V V V V V Aldehydes Hexanal V V V V V Heptanal V V V V V Octanal V V V V V Nonanal V V V V V Decanal V V V V V Aromatics . . Benzaldehyde V V V V V Benzene, 1 ,3-bis(1,1- V V V V V dimethylethyl) Phenol, 2,4-bis(1,1- V V V V V dimethylethyl) Hydrocarbons , Decane V V V V V Dodecane V V V V V Tetradecane V V V V V Hexadecane V V V V V Nonadecane V V V V V 195 Data in Table 6.8 indicates that the headspaces of the five different HDPE films had similar major components but differed from each other in terms of their concentrations. The result again confirmed the conclusion in the previous section “Confirmation of 1,3—Di-tert-Butylbenzene”, that this compound existed in the odor profiles of all the HDPE films. Description of Odor Profile with Sniffing Test Comparison of Two Retention Times Several standard compounds weretested using the SPME/GC-MS analysis with the sniffing port attached. The purpose of the study was to confirm that the time the odor of one compound was detected at the sniffing port was the same as the time the compound was detected by the mass spectrometer. Preliminary studies found that the flow of the auxiliary carrier gas to the sniffing port affected the amount of compounds going to the mass spectrometer after they eluted out of the GC column and split into two portions at the flow splitter of the ODO ll module. The higher the flow rate of the auxiliary carrier gas, the lower was the amount of the eluted compounds that went to the MS detector. Therefore, the flow control for the auxiliary carrier gas on the ODO ll module was adjusted to a scale reading “5” so that the majority of the eluted volatile 196 compounds went to the sniffing port for the investigator to describe detected odors, but sufficient material went to the MS so that they could be easily detected. Moreover, the GC capillary column flow program had to be adjusted as well due to the modified flow scheme at the flow splitter (see Figure 6.1). Preliminary studies found the GC could not maintain a constant carrier gas flow in the capillary column and would shut itself down after the inlet reached a certain pressure. As the result, a ramped pressure program was used to replace the constant flow column mode (see Table 6.9). Table 6.10 lists the times the standard compounds were detected by the GC/MS and at the sniffing port of the ODO ll module. Table 6.9 New parameters used in the SPME/GC-MS with ramped pressure program GC . Injection mode . Splitless Injector temperature 220°C Transfer line 240°C temperature Temperature program 40°C for 0 min, increased to 240°C at 40°lmin, stayed at 240°C for 3 min. Column flow Ramped pressure program (initial pressure 21.4 psi, increased to 30.4 psi at 1.8 psi/min, and stayed at 30.4 psi for 3 min). Purge flow 10 ml/min Purge time 10 seconds MS , _ . . - . Solvent delay 0 seconds Mass range 29 — 270 u Acquisition rate 8 spectra/second 197 Table 6.10 Comparison of retention times of volatiles determined by the two detectors using SPME/GC-MS and ODO ll sniffing port . Duplicate 1 2 3 Ethyl acetate GC RT (second) 1 15.41 1 15.53 1 15.03 ODO ll RT (second) 120 115 115 Butyl acetate 60 RT (second) 149.66 149.91 149.03 ODO ll RT (second) 149 147 148 Hexanal GC RT (second) 146.78 150.28 150.28 ODO ll RT (second) 146 144 146 Benzaldehyde GC RT (second) 1 190.78 190.91 190.53 ODO ll RT (second) 191 191 193 Hexyl acetate GC RT (second) 198.91 199.16 198.78 ODO ll RT (second) 196 196 197 1 ,3-Di-tert- 60 RT (second) 258.53 258.78 258.41 butylbenzene ODO ll RT (second) 260 258 257 Data in Table 6.10 confirmed the claim made by the manufacturer of the ODO ll sniffing module, SGE Analytical Science, that it took a similar time for the compounds to reach both the sniffing port and the MS by introducing the make- up gas at the exact point where the column flow is split between the two detectors so as to ensure the flow to the olfactory detector traveled at the same speed as the flow to the MS (SGE, 2002). The matching of the two retention times to each other. is considered crucial in investigating the contribution by each component compound to the odor profile, 198 because othenivise the identity of the compound cannot be determined when its associated odor is perceived at the sniffing port. Descriptions of Odor Profiles Odor profiles from sample PATA and PCTA were studied. The investigator was trained first with the standard 1,3-di-tert-butylbenzene to get familiar with its characteristic odor, after which the two odor profiles were presented in a randomly chosen order. Three sets of samples were tested by three investigators and each set of samples (standard, sample PATA, and sample PCTA) was tested by each investigator on the same day to ensure a consistent olfactory sensitivity. The side-by-side comparison of odor profiles of samples PATA and PCTA described by each investigator is listed in Appendix 9. Because the majority of the eluted volatile compounds was split to the sniffing port, the area respOnses detected by the MS were relatively low. To construct the profiles, the summation of selected ions was used, including mlz ratios of 41, 43, 57, 58, 60, 84, 85, 89, 70, 71, 73, 99, 147, 175, and 191. The resultant chromatograms are shown in Figures 6.8 and 6.9. 199 1.60904 I Area I 1.40904 I 1.20904 I 1.00904 I 8.00E+03 I 6.00E+03 I I 4.00903 I I I 2.00E+03 - WWII 0.00E+OO .2 *7 1* 1.00 1.50 2.0 Caramel. buttery Caramel, Overcooked ml Buming plastic Balloon. plastic odor Plastic. stinky 0 2.50 3.00 3.50 4.00 4.50 5.00 5.50 6.00 6.50 7.00 7.50 8.00 Time (min) Stinky. plastic Cardboard Paperboard. balloon. rubber smell '55— a— Plastic. metallic odor 5, 4— Plastic, balloon I33- ‘5. ’ I i _g fie 7.# ———\‘w—~—_.V~regfi Figure 6.8 Odor profile of sample PATA detected in the sniffing test 2.50E+04 ] Area 2.00E+04 - 1.50E+04 I 1.00E+04 < 5.00E+03 - Caramel, buttery Balloon. plastic odor Burning. manure Caramel. Overcooked coca Stinky. plastic Plastic. off odor Stinky plastic - 7:? g 3 .D in a a 2’ f» E 7a_ Rubber band. ,3 g plastic-like odor E 3 0.00E+00 1.00 1.5 O 2.00 2.50 3.00 3.50 r T r 4.00 4.50 l 1 r r 1 v l 5.00 5.50 6.00 6.50 7.00 7.50 8.00 Time (min) Figure 6.9 Odor profile of sample PCTA detected in the sniffing test 200 As shown in Figures 6.8 and 6.9, the odor profiles of sample PATA and PCTA were similar to each other. However, it must be mentioned that different terms might have been used by the investigators to describe the odor of the same volatile compound because they were neither pre-screened nor trained to improve their olfactory sensing sensitivity and judging precision. As a result, the same odor might be described one way by one person but differently by another, depending on the individual’s preference and sensitivity to the particular compound. The intensity of the odor of compound 1,3-di-tert-butylbenzene was perceived as “very weak" even though it did carry a plastic or a burning fabric odor, which was confirmed in the training process with the standard. However, a strong “balloon-like, plastic-like, and stinky” odor was perceived at nearly the same retention time as that 1,3-di—tert-butylbenzene (see Table 6.10) in the odor profile of sample PATA, whose quantity in the odor profile of sample PATA had been determined to be very low compared to that of sample PCTA (see Table 6.6). The strong “balloon-like, plastic-like, and stinky” odor perceived at the retention time 260 seconds (4 min 20 seconds) was more likely from some other volatile compound that was not detected by the GC/MS while the ODO II was attached. Even though its concentration was low, its threshold value was 201 probably very low as well, which made it a significant contributor to the odor profiles of the HDPE film samples. These observations indicated that this particular compound, 1,3-di-tert- butylbenzene, was not the crucial compound that made the perceived odor of sample PCTA by the sensory panelists stronger than those of the other HDPE film samples. Instead, it is more likely the odor profile of the HDPE film was the result of the interactions of two or more odorous compounds existing in the system. Each odorous compound contributed to the odor profile, depending on its concentration and threshold value. In the other words, sample PCTA had a stronger odor probably because its odor profile contained higher concentrations of two or more of these yet unidentified odorous compounds. 202 CHAPTER 7 CORRELATION OF ANALYSES OF E- NOSE, SENSORY EVALUATION, AND GC-MS INTRODUCTION AND OBJECTIVES As mentioned earlier, the objective of the study was to investigate the potential of utilizing the electronic nose system as a quality control tool. One way to reach that goal is to use the E-nose to predict the sensory scores of odor profiles of HDPE film samples, or to predict the quantities of crucial volatile compound(s). This chapter is devoted to discussion of the strengths and the weakness of three analytical techniques, along with the correlation among them (see Figure 2.14). E-NOSE AND SENSORY EVALUATION Instrument Sensitivity versus Human Nose Sensitivity The PCA module of the E-nose system was capable of differentiating the headspaces of the five HDPE films (see Figure 3.9). On the other hand, sensory panels said there was no significant difference in terms of “Acceptability” and 203 “Intensity” of the perceived odor among samples CONT, PATA, PBTA, and PBTB (see Table 4.7, Figure 4.4, and Figure 4.5). The E-nose system seemed more discriminating than the noses of the untrained panelists in our application. However, whether the higher discriminating capability of the E-nose system was significant from a practical standpoint requires further scrutiny because the odor profiles of sample CONT, PATA, PBTA, and PBTB were all grouped as “acceptable” by the sensory panelists even though the E-nose system can differentiate the headspaces. Subjective versus Objective Judgment In analyzing the headspaces of the HDPE film samples with the electronic nose system, it was concluded that the PCA (Principle Component Analysis) module of the system was capable of differentiating those profiles from each other. Moreover, the DFA (Discriminative Function Analysis) module of the E- nose system proved capable of identifying correctly an unknown sample as one of the training groups. However, both modules only gave objective judgments. In order to get subjective information such as acceptability, preference, and good or bad, sensory evaluation was still necessary. 204 Sensory panels in both the pair-wise ranking test and the quantitative affective consumer test found sample PCTA had a stronger perceived odor profile. However, when the panelists in the consumer test were asked to evaluate the intensity and the acceptability of the odor profile of sample PCTA, they indicated a neutral response to both questions. The information obtained from the sensory evaluation could not be offered from the E-nose analysis, which proved the importance of both analyses and showcased how they can complement each other to help the investigator to study the odor profiles of HDPE film samples. One more interesting observation that was made in the study and is worthy of mentioning was the comparison of the sensor responses of the E-nose system and the responses from the sensory panel. As shown in Figure 7.1, it was found that the odor profile of sample PBTA, rather than sample PCTA, generated the highest sensor responses in the E-nose analysis, even though the latter was perceived by the sensory panels as having a stronger odor profile than any other tested samples. 205 Figure 7.1 Side-by-side comparison of E-nose sensor responses to headspaces of sample PBTA and PCTA A very likely reason could be the complexity of the volatile system, which was composed of hundreds of volatile compounds at different concentrations. In the E-nose analysis, all volatile compounds existing in the odor profile interacted with the sensors and thus influenced the sensor responses. On the other hand, only those that were odorous and whose concentrations were above their threshold values interacted with human noses in the sensory evaluation. Even though the odor system of sample PBTA generated the highest sensor responses in the E-nose system, some of the volatile compounds in the system might not carry any odors, or their concentrations may have been below their thresholds. On the other hand, the odor system of sample PCTA might have more odorous compounds that were perceived by the sensory panelists. 206 The comparison again proved the importance of looking at the data from both the E-nose analysis and the sensory evaluation, in order to get an accurate and thorough understanding of the odor profiles. Partial Least Squares Using Sensory Scores The potential of the E-nose system as a quality control tool in a manufacturing environment can be investigated by the PLS.(Partial Least Square) module. If a correlation between sensory scores and E-nose analyses could be developed, a few obstacles associated with sensory evaluation could be avoided. These obstacles include the high cost, the time-consuming training process, and the susceptibility of results to panelists’ health and mood. PLS Based on Five Groups of Samples Figure 7.2 is the PLS plot of the “Acceptability” scores of the odor profiles based on the results listed in Table 4.7. Five groups of samples and 7 duplicates in each group were used to build the PLS model. The x-axis of the plot represents the experimental (or actual) values while the y-axis is the predicted values. The predicted sensory scores were calculated from the PLS model built 207 by the E-nose software, which correlated sensor responses of the E-nose with the experimental (or actual) sensory scores. Each solid dot on the plot represents one duplicate of the analyzed samples. The straight line is the line on which the predicted values equal the experimental values. The closer a dot is to the straight line, the closer its predicted value was to its experimental value. Correlation ‘ ll 6119274 10.400 10.200- 10.000— 9.800- 9.8004 9.4004 9.200~ 9.000- 8.800- 8.800— 8.400— 8.200- 8.000- 7.800— 7.600- 7.400- 7.200- 7-m- l I I I i i 7. 30 7.50 8.00 ‘ .350 9.00 9.50 10.00 10.50 . (ExpodmontaiorAetuai) (Original data were generated on 10/18/06) Predicted Figure 7.2 PLS plot of "Acceptability" scores of odor profiles of 5 groups of samples with training data for the E-nose shown only 208 Ideally, each cluster of dots should be next to or on the straight line when the PLS model fits the data perfectly. However, variations from one duplicate to another always induce some spread of data along the Y-axis. As shown in Figure 7.2, a bigger variation among .the duplicates of sample PCTA was observed than in the other samples. The PLS model under-estimated the “Acceptability” scores of odor profiles of sample CONT, PATA, and PBTA but over-estimated significantly that of sample PCTA. A correlation coefficient of 0.61 shown in Figure 7.2 also indicated the PLS was not robust or effective, which was further confirmed by projecting validation samples to the PLS model to estimate their “Acceptability” scores (see Table 7.1). Table 7.1 Using 5 groups of validation samples to evaluate the PLS model Samples Actual Panel . Predicted Average Opinion from scores Opinion scores predicted PLS model CONT_39 1 0.22 Acceptable 9.80 9.69 Acceptable CONT_40 10.22 Acceptable 9.60 9.69 Acceptable CONT_41 10.22 Acceptable 9.67 9.69 Acceptable PATA_10 9.89 Acceptable 9.41 9.36 Acceptable PATA_1 1 9.89 Acceptable 9.28 9.36 Acceptable PATA_9 9.89 Acceptable 9.39 9.36 Acceptable PBTA_49 9.07 Acceptable 8.57 8.59 Neutral PBTA_50 9.07 Acceptable 8.62 8.59 Neutral PBTA_51 9.07 Acceptable 8.58 8.59 Neutral PBTB_29 9.62 Acceptable 9.89 9.49 Acceptable PBTB_30 9.62 Acceptable 9.40 9.49 , Acceptable PBTB_31 9.62 Acceptable 9.19 9.49 Acceptable PCTA_19 7.25 Neutral 9.08 9.06 Acceptable PCTA_20 7.25 Neutral 9.03 9.06 Acceptable PCTA_21 7.25 Neutral 9.07 9.06 Acceptable 209 In Chapter 4, the HSD (Honestly Significant Difference) for “Acceptability” was determined as 1.4 and the neutral point of the unstructured scale carried a value of 7.5. As the result, any odor with an “Acceptability” score between 6.1 and 8.9 would be considered as neutral. In the sensory evaluation, the panel perceived the odor of sample PCTA as neutral and that of sample PBTA as acceptable. However, the PLS model over-estimated the scores for sample PCTA, based on which an “acceptable” judgment would have been made, and under-estimated the scores for sample PBTA, based on which the odor would have been considered as “neutral”. PLS Based on Three Groups of Samples Next, three groups of samples (CONT, PBTA, and PCTA) were selected to build the PLS model in the E-nose (see Figure 7.3). 210 10.000- . . 10.600- , ‘ ‘ ' 10.400— ' 10.200- 10.000— 9.800- 9.500- Predicted (D | 7. 400- PBTA 7.00 7.50 8'00 8'50 9.00 9.50 10'00 10.50 (Experimental or Actual) (Original data were generated on 10/18/06) Figure 7.3 PLS plot of "Acceptability" scores of odor profiles of 3 groups of samples with training data for the E-nose shown only The correlation coefficient increased to 0.83. Again, validation data was used to evaluate the robustness of the model (see Table 7.2). The data showed that the PLS model under-estimated the sensory scores for the odor profile of sample PBTA, which lead to an incorrect “neutral” evaluation of the odor. 211 Table 7.2 Using 3 groups of validation samples to evaluate the PLS model Samples Actual Panel Predicted Average Opinion from ' scores Opinion scores predicted PLS model . CONT_39 10.22 Acceptable 1 1 .29 10.20 Acceptable CONT_40 10.22 Acceptable 9.94 10.20 Acceptable CONT_41 10.22 Acceptable 9.37 10.20 Acceptable PBTA_49 9.07 Acceptable 8.03 8.08 Neutral PBTA_50 9.07 Acceptable 8.05 8.08 Neutral PBTA_51 9.07 Acceptable 8.16 8.08 Neutral PCTA_19 7.25 Neutral 9.79 8.74 Neutral PCTA_20 7.25 Neutral , 8.46 8.74 Neutral PCTA_21 7.25 Neutral 7.96 8.74 Neutral PLS Based on Pairs of Samples Table 7.3 lists the correlation coefficients of PLS models when different pairs of samples were used to build the PLS model in the E-nose system, which showed improvements compared to the PLS model based on three groups of samples. Table 7.3 Correlation coefficients of PLS models based on different pairs of samples Samples CONT PATA PBTA PBTB PCTA CONT / 0.996773 0.996001 0.979335 0.921467 PATA / 0.987000 0.998566 0.984701 PBTA / 0.997945 0.996099 PBTB I 0.991600 PCTA / 212 The correlation coefficient was considered important in evaluating the robustness and effectiveness of the PLS model. However, it was equally important to validate the model with test data (validation data), as shown in Table 7.1 and 7.2. As one example, Figure 7.4 is the PLS plot based on samples PCTA and PBTA. Each cluster of solid dots (in total 7 points) represents one group of training data. The empty dots are the validation data (3 duplicates). — 9.2% 8.000— Validation data BBEIJ- for PBTA 3.5:!)- 8'm_ Training data Bazw— for PCTA Training data § for PBTA 5: 8.030- 7.800— 7.600- 7-400‘ ‘ Validation data 7.2113— for PCTA 0 ‘ , 1m | i l I I l I I l 7.20 7.40 7.50 7.00 8.00 8.20 8.40 3.50 8.80 9.00 9.20 (ExportmentalorActual) (Original data were generated on 10/18/06) Figure 7.4 PLS plot of "Acceptability" scores of odor profiles of sample PCTA and PBTA with both training data and validation data 213 Table 7.4 lists the predicted “Acceptability” scores based on the PLS model shown in Figure 7.4. More PLS models based on pairs of samples inVolving sample PCTA and their validation results are listed in Appendix 10. Table 7.4 Validate the PLS model based on sample PCTA and PBTA Samples Actual Panel Predicted Average Opinion from scores Opinion » scores predicted PLS model PBTA_49 9.07 Acceptable 8.93 9.03 Acceptable PBTA_50 9.07 Acceptable 9.09 9.03 Acceptable PBTA_51 9.07 Acceptable 9.06 9.03 Acceptable PCTA_19 7.25 Neutral 7.30 7.32 Neutral PCTA_20 7.25 Neutral 7.28 . 7.32 Neutral PCTA_21 7.25 Neutral 7.39 7.32 Neutral So far every pair of samples used to test the effectiveness of the PLS module of the E-nose system always had sample PCTA because it was the only one whose odor was found significantly stronger than those of others in the sensory evaluation (see Table 4.7). To further investigate the effectiveness of the PLS module, samples PATA and PBTB were paired to build and validate the PLS model because their “Acceptability” scores were the closest to each other among all the pairs of scores (see Table 4.7). The results are shown in Figure 7.5 and Table 7.5. 214 9.920 9900— . ~ 0 9880— ' 9.980- 9.840- 9.820- 9.800- 9,790— 9.790— 9.740— 9.720— 9.700— 9.880— 9.680- 9.840— Validation data for PATA Training data for PBTB Training data for PATA 9.1320- Validation data . for PBTB 9.600 9.60 9.85 9'70 9'75 900 9.85 9.90 (Experimental or Actual) (Original data were generated on 10/18/06) Predicted °0 Figure 7.5 PLS plot of "Acceptability" scores of odor profiles of sample PATA and PBTB with both training data and validation data Table 7.5 Validate the PLS model based on sample PATA and PBTB Samples Actual Panel Predicted Average Opinion from scores Opinion scores predicted PLS model PATA_10 9.89 Acceptable 9.88 9.88 Acceptable PATA_1 1 9.89 Acceptable 9.86 9.88 Acceptable PATA_9 9.89 Acceptable 9.89 9.88 Acceptable PBTB_29 9.62 Acceptable 9.62 9.62 Acceptable PBTB_30 9.62 Acceptable 9.62 9.62 Acceptable PBTB 31 9.62 Acceptable 9.62 9.62 Acceptable 215 Correct predictions were made to evaluate the acceptability of the odor profile of the HDPE film sample based on the PLS model built by the E-nose system. This showcases the possibility of using the E-nose system to make correct subjective judgments which were initially only possible with costly sensory evaluation. However, it would be dangerous to generalize the statement by saying the E-nose system could be used in any application because the results in Figure 7.2 and 7.3 proved the complexity of data tended to jeopardize the robustness and effectiveness of the PLS model in predicting the sensory scores. E-NOSE AND GC-Ms ANALYSIS In Chapter 6, major potential component volatiles in the odor profiles of the HDPE film samples were identified, among which acetone and nonanal had previously been mentioned as important contributors to the off odor of HDPE- based packaging materials (Maneesin, 2001). Table 7.6 lists the response areas of acetone and nonanal detected in the odor profiles of five HDPE film samples with the SPME/GC-MS analysis. Appendix 11 explains how the original data from GC-M-S was processed to get the data listed in Table 7.6. 216 Table 7.6 Average response areas of acetone and nonanal detected in the odor profiles of different HDPE film samples in SPME/GC-MS analysis Sample * CONT PATA PBTA * PBTB PCTA Acetone 72929 837784 1980094 97014 1784602 Nonanal 6017466 1851 147 1394606 1080479 1 3931678 Data in Table 7.6 was then used to build PLS models to predict the response areas of acetone and nonanal in the odor profiles. PLS Based on Five Groups of Samples All five groups of samples, CONT, PATA, PBTA, PBTB, and PCTA were used to build the PLS model, which was then used to predict the response areas of acetone and nonanal in the odor profiles of validation samples based on their E-nose sensor responses. Table 7.7 lists the predicted response areas and the percentages of difference of the predictions from the actual areas when all five groups of samples were used to build the PLS model in the E-nose system. Correlation coefficients of 0.96 and 0.42 were obtained in the PLS model for acetone and nonanal, respectively. However, the models proved ineffective, as can be seen in Table 7.7, in predicting the response areas for either compound in the odor profiles of the HDPE film samples. 217 Table 7.7 Predicted response areas of acetone and nonanal based on the PLS model of five groups of samples Acetone . Nonanal . Sample Prediction Ave. % Diff1 Prediction Ave. %Diff1 CONT_32 1 18371 1558786 CONT_33 68283 85305 17 4578414 4126995 -31 CONT_34 69261 6243785 PATA_2 1 530306 3226480 PATA_3 1 389259 1427785 70 3272042 31 87805 72 PATA_4 1 363788 3064894 PBTA_42 2030657 2544512 PBTA_43 1 772624 1 826275 -8 2192277 2283498 64 PBTA_44 1 675545 21 1 3705 PBTB_22 1 33502 1 920658 PBTB_23 1 1 1549 120412 24 2168431 2136555 98 PBTB_24 1 1 6186 ‘ 2320577 PCTA_12 3160451 27168847 PCTA_13 746504 1 547853 1 3 1888663 10408647 -25 PCTA_14 736603 21 68433 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. PLS Based on Three Groups of Samples Three groups of samples, CONT, PBTA, and PCTA, were used to build the PLS model to predict the response areas of acetone and nonanal in their odor profiles. Correlation coefficients of the models were improved to 0.99 and 0.75 for acetone and nonanal, respectively. Table 7.8 lists the predicted response areas 218 of acetone and nonanal in the odor profiles of validation samples based on their E-nose sensor responses and the PLS model. Table 7.8 Predicted response areas of acetone and nonanal based on the PLS model of three groups of samples Acetone Nonanal Sample Prediction Ave. % Diff‘ Prediction Ave. %Diff‘ CONT_32 88244 516817 CONT_33 80153 87947 21 8843125 10441686 74 CONT_34 95444 21965115 PBTA_42 2726726 3614981 PBTA_43 2521001 2527948 28 3543651 3425828 146 PBTA_44 2336117 3118853 PCTA_12 1092304 1513611 PCTA_13 1135062 1192056 -33 3980545 4524301 -68 PCTA_14 1348803 8078748 1. The percentage of the difference between the prediction and the actual value (see Table'7.6) divided by the actual value. Obviously, the PLS model was still not effective. PLS Based on Pairs of Samples Tables 7.9 to Table 7.12 list the predicted response areas of acetone and nonanal when pairs of samples were used to build the PLS model, which was then used to predict the areas for the validation samples based on their E-nose responses. 219 Table 7.9 Predicted response areas of acetone and nonanal based on the PLS model of samples CONT and PCTA Acetone Nonanal Sample Prediction Ave. % Diff1 Prediction Ave. %Diff1 CONT_39 1 18512 6835595 CONT_40 59451 77069 6 57031 56 6026371 0.15 CONT_41 53242 5540361 PCTA_1 9 938422 1 1 768314 PCTA_20 1 2631 52 1 1 82694 -34 1 2723262 1 2476705 ‘I 0 PCTA_21 1 346509 12938539 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. Table 7.10 Predicted response areas of acetone and nonanal based on the PLS model of samples PATA and PCTA Acetone Nonanal Sample Prediction Ave. % 0111‘ Prediction Ave. %Diff‘ PATA_10 856445 1963260 PATA_11 1056522 907127 8 3438313 2361511 28 PATA_9 ~ 308414 1682961 PCTA_19 1724812 12720433 PCTA_20 1557396 161230 -10 9686047 10679749 23 PCTA_21 1 5541 81 9632768 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. Table 7.11 Predicted response areas of acetone and nonanal based on the PLS model of samples PBTA and PCTA Acetone Nonanal . Sample Prediction Ave. % Diff1 Prediction Ave. %Diff1 PBTA_49 1964731 16571 15 PBTA_50 1 981 943 1969365 -0.5 1 366089 14761 15 6 PBTA_51 1 979422 1 405141 PCTA_1 9 1789727 1 3074536 PCTA_20 1787198 1 791 832 0.4 1 3490362 12762543 -8 PCTA_21 1798571 1 1 722730 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. 220 Table 7.12 Predicted response areas of acetone and nonanal based on the PLS model of samples PBTB and PCTA Acetone Nonanal ’ Sample Prediction Ave. % Diff1 Prediction Ave. %Diff1 PBTB_29 74869 860631 PBTB_30 1661 1 1 162090 67 1 732529 1677560 55 PBTB_31 245289 243951 9 PCTA_19 1 57251 5 12466997 PCTA_20 2000872 1840920 3 1 5403524 14308466 3 PCTA L21 1949372 1 5054878 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. Data in Tables 7.9 to Table 7.12 Show the percentages of difference varied from 0.15% to 67%, indicating the PLS models were not robust. The effectiveness of the model was affected by which pair of films and which volatile compound were studied. For example, the PLS model based on sample PBTA and PCTA predicted the response'areas of both acetone and nonanal satisfactorily (see Table 7.11). The model in Table 7.9 was more effective in predicting the response areas of nonanal, while the model in Table 7.11 was better in predicting the areas of acetone. PLS Based on Transformed Data A close observation of the response areas of acetone and nonanal in Table 7.6 indicated a wide range among different samples (10E4 to 10E6 for acetone and 10E6 to 10E7 for nonanal). A L0910 transformation was therefore 221 made to the data to make them more uniform to see if it would improve the robustness and effectiveness of the PLS models. However. no significant improvements were obtained from the data transformation (see Appendix 12). Yuzay (2004) proposed using PCA and sensor responses in the E-nose system to detect outliers and eliminate them before building PLS models. Figure 7.6 is the PCA plot of all the training and validation data, based on which CONT_32, PATA_6, PATA_11. PBTA_48, PBTB_30, PBTB_31. PCTA_12. and PCTA_18 were eliminated from the model training and validation process. 0.200 - ' 1*. 0.150- 0.100- 0.050- 0.000- (32:1.782 -0.050-‘ 0.100- 0.150- m I l I I l l I I I I I -1.40 -1.20 4.03 -0.80 0.60 -0.40 0.20 0.00 0.20 0.40 0.50 0.80 1.00 CT : 97.372 (Original data were generated on 10/03/06) Figure 7.6 Using PCA to detect outliers before building PLS models 222 Tables 7.13 to 7.16 list the predicted response areas of acetone and nonanal based on the PLS models without the outliers. Table 7.13 Predicted response areas of acetone and nonanal of samples CONT and PCTA based on the PLS model without outliers , Acetone Nonanal - Sample Prediction Ave. % 0111‘ Prediction Ave. %Diff‘ CONT_39 79590 6157144 CONT_40 97414 86662 19 6492655 6291594 5 CONT_41 82982 6224982 PCTA_19 1792154 13947142 PCTA_20 1761395 1848053 4 13883892 14056036 0.9 PCTA_21 1990611 14337073 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. Table 7.14 Predicted response areas of acetone and nonanal of samples PATA and PCTA based on the PLS model without outliers Acetone Nonanal Sample Prediction .Ave. % Diff1 Prediction Ave. %Diff1 PATA_1 0 81 7074 1731 523 PATA_8 76691 2 792834 -5 14621 24 1 6001 64 -‘l 4 PATA_9 79451 5 1 606845 ' PCTA_1 9 1 597429 1 0364863 PCTA_20 1 600201 1 6261 13 -9 1 041 2890 1 0882783 -22 PCTA_21 1 680708 1 1870598 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. 223 Table 7.15 Predicted response areas of acetone and nonanal of samples PBTA and PCTA based on the PLS model without outliers Acetone Nonanal Sample Prediction Ave. % Diff‘ Prediction Ave. %Diff‘ PBTA_49 1960186 1744306 PBTA_50 1974515 1973515 03 1484527 1512235 8 PBTA_51 1985846 1307872 PCTA_19 1796569 12015296 PCTA_20 1817475 1798605 0.8 9300128 11915184 -14 PCTA_21 1781771 14430127 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. Table 7.16 Predicted response areas of acetone and nonanal of samples PBTB and PCTA based on the PLS model without outliers Acetone Nonanal Sample Prediction Ave. ~ % Diff1 Prediction Ave. %Diff1 PBTB_27 94416 1054854 PBTB_28 101191 89140 -8 1121001 1001821 7 PBTB_29 71 814 829608 PCTA_1 9 1 512940 12054699 PCTA_20 19041 34 1 766242 -1 14752592 1 38021 31 -0.9 PCTA_21 1881651 14599103 1. The percentage of the difference between the prediction and the actual value (see Table 7.6) divided by the actual value. 224 Table 7.17 compares the percentages of difference before and after the outliers were eliminated. Table 7.17 Comparison of percentages of difference before and after outliers were eliminated from PLS models Acetone CONT/PCTA PATA/PCTA PBTA/PCTA PBTB/PCTA Before 6, -34 8, -10 -0.5, 0.4 67, 3 After 19, 4 -5, -9 -0.3, 0.8 - -8, -1 Nonanal CONT/PCTA PATA/PCTA PBTA/PCTA P BTBIPCTA Before 0.15, 10 28. 23 6, -8 55. 3 After 5, 0.9 -14, -22 8, -14 7, -0.9 Eliminating outliers did improve the effectiveness of the PLS model based on the pair of samples PBTB and PCTA. However, the percentage of difference increased in predicting the area of acetone in the pair of samples CONT and PCTA. The results further confirmed that the robustness and effectiveness of the PLS models were affected by which pair of samples and which volatile compound were investigated. Moreover, improvement by eliminating outliers was not obtained in all the PLS models. 225 CHAPTER 8 SUMMARY AND CONCLUSIONS ELECTRONIC NOSE Settings of major data acquisition parameters, including incubation time, temperature, acquisition time, delay, and sample Size, in the Alpha MOS Fox 3000 E-nose analysis for HDPE films were investigated and optimized. The headspace of the HDPE film was generated from 2 strips of 10" x 1" HDPE film samples crimp-closed in a 10 mil glass vial, which was heated at 100°C for 20 minutes. Sensor responses were recorded during the ten minutes (600 seconds) of the acquisition period and a fifteen minute (900 seconds) delay was used so that the sensor responses could go back to their baselines. The E-nose system was proved capable of differentiating the headspaces of the five different HDPE film samples (sample CONT, PATA, PBTA, PBTB, and PCTA) with a discrimination index of 87 in the PCA (Principle Component Analysis) module. A DFA (Discriminative Function Analysis) model was built during the training process, which had a percentage of recognition of 94. The model was then validated successfully by correctly identifying unknown samples as one of the training samples. 226 SENSORY EVALUATIONS The painlvise ranking test was used to differentiate the odor profiles of the HDPE film samples by an untrained panel with 50 panelists. Friedman analysis was used to analyze the data and calculate rank sums as well as the HSD (Honestly Significant Difference) value. The odor profile of sample PCTA was perceived significantly stronger than those of other samples. No Significant difference was found among the odor profiles of sample CONT, PATA, PBTA, and PBTB. In the quantitative affective consumer test, the “Acceptability" and “Intensity” of the odor profiles were evaluated by 100 panelists using unstructured scales. The RCBD (Random Complete Block Design) was used in the experimental design, which proved effective and efficient. Sensory scores of both “Acceptability" and “Intensity" were determined and their HSD values were calculated. Painlvise comparison of the sensory scores indicated a “neutral” opinion by the panel for the odor profile of sample PCTA and “acceptable" opinions for the odor profiles of all other samples. In terms of the intensity, the odor profile of sample PCTA was found Significantly stronger than those of others, which confirmed the result of the painlvise ranking test. However, the position of its intensity score on the unstructured intensity 227 scale indicated the odor profile of sample PCTA was perceived by the panel as “weak” to “neutral". GC-MS ANALYSIS DIP (Direct Insertion Probe) analysis was used to pre-study the headspaces of sample CONT and PCTA. Mass spectra indicated the existence of hydrocarbons in the headspaces of both samples. Moreover, results indicated a more complex volatile system from sample PCTA and the presence of compounds other than hydrocarbons. Two techniques, DH-TD (Dynamic Headspace and Thermal Desorption) and SPME (Solid-Phase Microextraction) were used to collect and concentrate volatile compounds in the headspace generated from the HDPE film samples before analysis with GC-MS (Gas Chromatography — Mass Spectrometry). In DH-TD, a Carbotrap 300 mUIti-bed thermal desorption tube was selected to collect and concentrate volatile compounds. However, the modified connection between the transfer line and the GC column in the DH-TD/GC-MS analysis proved not effective and thus a low repeatability was observed. Data from the DH-TD/GC-MS indicated compound 1,3-di-tert-butylbenzene only existed in the headspace of sample PCTA, which was then assumed to be the 228 objectionable volatile compound that was responsible for the stronger odor of the sample. The 65 pm PDMS/DVB SPME fiber was used in the SPME/GC-MS analysis. The SPME sampling temperature and time were optimized as 60°C and 30 minutes. Comparison of gas chromatograms indicated same groups of volatile compounds existed in the odor profiles of all the HDPE film samples, including ketones, aldhydes, aromatics, and hydrocarbons, among which acetone, hexanal, benzaldehyde, and 1,3-di-tert-butylbenzene were confirmed by comparing their retention times and mass spectra with those of standard compounds. More volatile compounds mighthave been collected in the DH-TD overall because the extraction of volatile compounds by the SPME fiber was affected by the SPME sampling temperature and time but the extraction with the DH-TD was not. However, the DH-TD analysis required much more complicated instrument setup and a longer time to recover the thermal desorption tube (60 minutes). On the other hand, SPME analysis proved effective and efficient in collecting volatile compounds of interest. A much Simpler instrument setup and fewer steps were involved in the SPME procedure. Moreover, no recovery time was needed in the SPME analysis because the fiber was cleaned at the same time as the extracted volatiles were desorbed to the GC column. 229 An ODO ll sniffing unit was connected to the GC/MS instrument to split the eluted flow from the 60 column so that odors Of the volatile compounds could be detected at the same time as they were detected by the mass spectrometer. The efficiency Of the sniffing unit was confirmed by comparing the two retention times (RT at the sniffing port and RT at the mass spectrometer) Of several standard compounds, including ethyl acetate, butyl acetate, hexanal, benzaldehyde, hexyl acetate, and 1,3-di-tert-butylbenzene. The odor profiles Of samples PATA and PCTA were described in the sniffing test. Compound 1,3-di-tert-butylbenzene was found existing in the odor profiles Of all the HDPE film samples, contradicting the results in the DH-TD/GC- MS analysis. Its odor was determined “weak” in the sniffing test. It was concluded that the compound was not responsible for the stronger odor profile Of sample PCTA. Instead, it was more likely various odorous compounds affected the perceived odor profiles Of the HDPE film samples, depending on their characteristic odors and respective threshold values. CORRELATION OF E-NOSE, SENSORY EVALUATION AND GC-MS ANALYSIS The E-nose system was capable Of making Objective judgments about the Odor profiles. In addition, the E-nose system seemed more discriminating than the noses Of the untrained panelists in detecting the difference among the Odor 230 profiles Of the HDPE films. However, sensory evaluation was necessary when subjective judgments were wanted. Correlation coefficients in the range Of 0.92 to 0.99 were Obtained in the PLS (Partial Least Square) models in the E-nose system when the sensory scores Of pairs of samples were used. The models were proved effective and robust in predicting the “Acceptability” sensory scores, based on which correct subjective judgments were made. The successful correlation between the E-nose system and the sensory evaluation proved the potential Of the E-nose system as a quick and accurate tool to replace costly-and time-consuming sensory evaluations to predict the acceptability Of Odor profiles Of the HDPE film samples, if the E-nose system was trained first with the results from the sensory evaluation. Response areas Of acetone and nonanal from the SPME/GC-MS analysis were used to correlate the E-nose system and the GC/MS by building PLS models. Unlike those for predicting sensory scores, the PLS models for response areas were not effective or robust. Logm transformation Of the original data did not improve the effectiveness or the robustness Of the PLS models Significantly. However, eliminating outliers by using the PCA module in the E-nose seemed to improve the effectiveness Of 231 some PLS models, depending on which pairs of samples and which volatile compound were studied in the correlation between the E-nose and GC/MS analysis. The correlation between the E-nose system, sensory evaluation, and GC- MS analysis also proved it was important to complement the three techniques with one another. RECOMMENDATIONS FOR FUTURE WORK Using a trained panel in the quantitative descriptive analysis might help improve the accuracy Of the “ACOeptability” and “Intensity" sensory scores when the unstructured scale is used. A training process would help the panelists reach consensus on the correlation Of a sensory score and its appropriate position on the scale, which in turn would help avoid the skewness Of the resultant data. An alternative to using the unstructured scale would be using a structured scale. It would also be interesting to investigate the correlation between the opinion Of the trained panel and the opinions Of untrained consumers. Moreover, descriptive tests by a pre-screened and trained sensory panel could be used to describe the characteristic Odors Of the HDPE film samples, which might help identify the major volatile contributors to the Odor system. 232 More work in the SPME/GC-MS analysis might be worthwhile in order to get a more accurate characterization of the volatile profile. A slower temperature program would help better separate volatile compounds from each other, which in turn would help the detection Of the Odors at the sniffing port. If the investigators in the sniffing test were pre-screened and trained, improved consistency and accuracy would be Obtained. AS a result, similar sensitivity would be maintained among the inveStigators and the same terminology would be used in describing the sensed Odors. The correlation between the E-nose and sensory evaluation was successful but that between the E-nose and the GC-MS was not. One possible reason might be that the sensory scores represented the overall‘evaluation Of the perceived Odor profiles by the sensory panel while the PLS models were built to predict the amount Of one particular compound in the very complex Odor system which contained hundreds Of volatile compounds. It would thus make more sense to get a value to represent the overall amount Of the major odorous compounds in the Odor system, such as a weighted average Of ratios Of their concentrations tO their corresponding threshold values, which would then be used tO build the PLS model tO correlate the E-nose system and the GC/MS analysis. 233 APPENDICES APPENDIX 1 WORKSHEET USED IN THE PAIRWISE RANKING TEST 5 1 16 7 10 18 17 5 20 18 9 16 11 15 17 7 20 1O 11 20 6 2 19 '3 1 11 10 3 8 5 1 10 12 8 18 14 18 15 15 8 4 5 2 16 11 5 2 18 4 11 16 11 20 17 4 9 12 6 16 19 3 12 7 6 13 3 15 8 3 15 13 7 3 '1 14 17 14 4 12 2 20 13 5 11 8 4 6 14 10 4 12 4 8 4 9 2 1 6 6 19 9 1 12 3 16 9 7 5 12 2 14 6 13 7 13 5 17 7- 10 17 3 19 10 5 12 8 11 14 16 11 8 13 14 16 15 1 12 2 .1 15 6 15 9 19 1 4 16 17 14 1O 19 17 '18 2 10 13 2 13 9 18 15 10 20 12 6 3 19 9 3 8 20 4 ‘5 6 19 16 9 18 7 14 18 14 2' 20 8 19 20 1 17 7 17 9 19 15 20 11 13 19 13 234 APPENDIX 2 WORKSHEET AND DATA IN THE QUANTITATIVE AFFECTIVE CONSUMER TEST Judge Sample’ Accept Intensity Judge Sample‘ Accept Intensity * 1 - CONT, 2 — PATA, 3 — PBTA, 4 -— PBTB, 5 — PCTA. 235 Judge Sample‘ Accept Intensity Judge Sample‘ Accept Intensity * 1 — CONT, 2 — PATA, 3 - PBTA, 4 — PBTB. 5 - PCTA. 236 Judge Sample" Accept Intensity Judge Sample”? ‘31 39 * 1 — CONT, 2 — PATA, 3 — PBTA, 4 — PBTB. 5 — PCTA. 237 Judge Sample‘ Accept Intensity Judge Sample’ ~=Accept Intensity * 1 — CONT. 2 — PATA, 3 - PBTA, 4 -'PBTB, 5 — PCTA. 238 Judge Sample' Accept Intensity t-Judge Sample“ Accept intensity) 63 * 1 — CONT. 2 - PATA, 3 — PBTA. 4 - PBTB, 5 — PCTA. 239 Judge Sample‘ Accept Intensity Judge Sample" Accept. "Intensity 79 * 1 — CONT, 2 — PATA. 3 - PBTA, 4 - PBTB, 5 — PCTA. 240 Judge Sample“ Accept Intensity Judge Sample* Accept Intensity 95 " 1 - CONT. 2 - PATA, 3 -— PBTA. 4 - PBTB, 5 - PCTA. 241 APPENDIX 3 SCORESHEET USED IN THE PAIRWISE RANKING TEST MULTIPLE PAIRED COMPARISONS TEST Name: Date: Sample: Plastic films Difference: _Qdor instructions: I. Receive the sample tray and note each sample code below according to its position on the tray. I J . Open and smell the odor in each sample pair from left to right and note which sample has a stronger odor (stronger smell). Indicate by placing an “X" next to the code. Samples should be smelled in order and r‘c-smell is not allowed. '.t) . Continue until all 4 pairs have been evaluated. DO not fatigue your olfactory system by sniffing too long or too often. Take a break as IICCL‘SSHI'y. Pair number Left sample Right sample 4 3 1 A FINE—1F] LIDLILJ l If you perceive no difference, please make a best guess. (“NO difference" response is not permitted.) 242 APPENDIX 4 SCORESHEET USED IN THE QUANTITATIVE AFFECTIVE CONSUMER TEST QUESTIONNAIRE FOR UNSTRUCTURED SCALING TEST Name: Date: Sample: Plastic films Instructions: 1. There are FIVE vials of samples presented in this test. We are asking you to evaluate the acceptability and intensity of the odor of the sample. by marking a line at the appropriate point. 2. Please follow the sample ID shown below to evaluate your samples. 3. Open one vial each time and sniff the headspace immediately. Make a vertical line on the horizontal line to indicate your rating of the acceptability, then a vertical line on a 2nd horizontal line to indicate your rating of the intensity of odor. Rest as much as you want before you go ahead to the next sample. 4. Please try to evaluate each sample individually. Do not compare among them when you make your evaluation. 5. please do not go back to sniff evaluated samples for a 2'“ time because the odor profile might have changed since your first snufing. Please sniff the samples in the following order: Overall opinion of the odor l l I I Not at all Very much acceptable acceptable Intensity of the odor I I Extremely Extremely weak strong 243 APPENDIX 5 SAS PROGRAM AND ITS OUTPUT SAS Program data sensory; input panelists samples acceptability intensity @@; cards; /*data set emitted in the program*/ Title 'Produce Residuals Dataset in Analyzing Acceptability Using GLM'; proc glm data=sensory; class panelist sample; model acceptability = panelist sample; random panelist; output out=predict r=resid p=pred; run; proc univariate normal plot data=predict; var resid; run; symbol value=dot interpol=none color=blue height=l; Title 'Residuals Diagnostic'; proc gplot data=predict; Title 'Figure 4.2: Residual versus sample ID in analyzing acceptability'; plot resid*sample/vref=0; run; Title 'Analysis of Acceptability Using Mixed'; proc mixed data=sensory; class panelist sample; model acceptability = sample; random panelist; lsmeans sample/adjust=tueky diffs; run; quit; 244 Output for Analyzing Acceptability Produce Residuals Dataset in Analyzing Acceptability Using GLM The GLM Procedure Class Level Information Class Levels values . panelist 100 1 10 100 11 12 13 14 15 16 17 18 19 2 20 21 22 23 24 25 26 27 28 29 3 30 31 32 33 34 35 36 37 38 39 4 40 41 42 43 44 45 46 47 48 49 5 50 51 52 S3 :2 :: S6 57 58 59 6 60 61 62 63 64 65 66 67 68 69 7 70 71 72 73 74 75 76 77 8 80 81 82 83 84 85 86 87 88 89 9 9O 91 92 93 94 95 96 97 98 99 sample 5 1 2 3 4 5 Number of observations Read 500 Number Of Observations Used 500 Produce Residuals Dataset in Analyzing Acceptability Using GLM The GLM Procedure Dependent Variable: acceptability Sum Of Source DF Squares Mean Square F value Pr > F Model 103 5506.06004 53.45689 4.04 <.0001 Error 396 5237.74468 13.22663 Corrected Total 499 10743.80472 R-Square Coeff var Root MSE acceptability Mean 0.512487 39.49484 3.636843 9.208400 Source DF Type I 55 Mean Square F value Pr > F panelist 99 4954.880720 50.049300 3.78 <.0001 sample 4 551.179320 137.794830 10.42 <.0001 Source DF Type III SS mean Square F value Pr > F panelist 99 4954.880720 50.049300 3.78 <.0001 sample 4 551.179320 137.794830 10.42 <.0001 Produce Residuals Dataset in Analyzing Acceptability Using GLM The GLM Procedure Source Type III Expected Mean Square panelist varCError) + S vaGCanelist) sample var(error) + Q(sample) 245 Produce Residuals Da T N Mean Std Deviation 3. skewness -0 Uncorrected SS 52 Coeff variation Ba Location Mean 0.0000 Median 0.1979 Mode -0.8116 NOTE: The mode displayed Te Test Student's t Sign Signed Rank Test Shapiro-wilk Kolmogorov-Smirnov Cramer-von Mises Anderson—Darling Q taset in Analyzing Acceptability Using GLM he UNIVARIATE Procedure variable: resid Moments 500 Sum weights 500 0 Sum observations 0 23982751 variance 10.4964823 .5228333 Kurtosis 0.64147781 37.74468 Corrected SS 5237.74468 . Std Error Mean 0.14488949 sic Statistical Measures variability 0 Std Deviation 3.23983 0 variance 10.49648 0 Range 19.22700 Interquartile Range 3.71800 is the smallest of 4 modes with a count of 2. sts for Location: Mu0=0 -Statistic- ----- p Value ------ t 0 Pr > Itl 1.0000 M 20 Pr >= |M| 0.0810 S 3456 Pr >= ISI 0.2854 Tests for normality --Statistic--- ----- p value ------ w 0.98054 Pr < w <0.0001 D 0.06855 Pr > D <0.0100 w-sq 0.431538 Pr > w—Sq <0.0050 A—Sq 2.514278 Pr > A-Sq <0.00SO uantiles (Definition 5) Quantile Estimate 100% Max 8.4394 99% 6.8039 95% 4.9624 90% 3.9124 246 Produce Residuals Dataset in Analyzing Acceptability Using GLM The UNIVARIATE Procedure variable: reSTd Quantiles (Definition 5) Quantile Estimate 75% 03 2.0444 50% Median 0.1979 25% Q1 -1.6736 10% -4.3276 5% -S.S906 1% -9.1721 0% Min -10.7876 Extreme observations ------ Lowest---—- -----Highest---- value Obs value Obs -10.7876 75 6.8794 259 —10.3316 145 6.9834 207 -10.2716 398 7.1194 468 -10.1366 340 7.4924 181 —9.4436 477 8.4394 331 Histogram # Boxplot 8.S+* 1 .* 2 I .#**** 9 I .******* 13 I .k*********# 22 I .***#*********fi**** 35 I .*************flfik**k**** 45 + ..... + .*i***************************** 62 I I .********k*#*#**************************** 81 *__+__* '*******t******************#********* 72 I I .***fi*********t******** 43 + _____ + .*#**#*t*******#*t 33 I _*******w***fi* 25 I .teeeeeete-A’t 21 I .******** 16 I .** 3 I ** 3 0 ****fi 9 o .* 1 0 -10.5+** 4 0 ----+----+----+----+----+----+----+----+- * may represent up to 2 counts 247 Produce Residuals Dataset in Analyzing Acceptability Using GLM The UNIVARIATE Procedure variable: reSTd Normal Probability Plot 8.5+ * +++* +***** +**** **** **** **** ***** ***** ***** ***+ **** *** **** +*** +++* +++ ** + ***** * -10.5+** +----+-—--+----+----+----+----+----+----+----+----+ -2 -1 0 +1 +2 248 Analysis Of Acceptability Using Mixed The Mixed Procedure Model Information Data set WORK . SENSORY Dependent Variable acceptability Covariance Structure variance Components Estimation Method REML Residual variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment Class Level Information Class Levels values panelist 100 1 10 100 11 12 13 14 15 16 17 18 19 2 20 21 22 23 24 25 26 27 28 29 3 30 31 32 33 34 35 36 37 38 39 4 40 41 42 43 44 45 46 47 48 49 5 50 51 52 53 54 SS 56 57 58 59 6 60 61 62 63 64 65 66 67 68 69 7 70 71 72 73 74 75 76 77 78 79 8 80 81 82 83 84 85 86 87 88 89 9 90 91 92 93 94 95 96 97 98 99 sample 5 1 2 3 4 5 Dimensions Covariance Parameters 2 Columns in x 6 Columns in Z 100 Subjects 1 Max Obs Per Subject 500 Number of Observations Number of Observations Read 500 Number of Observations Used 500 Number of observations Not Used 0 Iteration History Iteration Evaluations -2 Res Log Like Criterion 0 l 2925.08167653 249 Analysis of Acceptability Using Mixed The Mixed Procedure Iteration History Iteration Evaluations —2 Res Log Like Criterion 1 1 2837.72674190 0.00000000 Convergence criteria met. Covariance Parameter Estimates Cov Parm Estimate panelist 7.3645 Residual 13.2266 Fit Statistics -2 Res Log Likelihood 2837.7 AIC (smal er is better) 2841.7 AICC (smaller is better) 2841.8 BIC (smaller is better) 2846.9 Type 3 Tests of Fixed Effects Num Den Effect DF DF F value Pr > F sample 4 396 10.42 <.0001 Least Squares Means Standard Effect sample Estimate Error DF t value Pr > Itl sample 1 10.2160 0.4538 396 22.51 <.0001 sample 2 9.8920 0.4538 396 21.80 <.0001 sample 3 9.0650 0.4538 396 19.98 <.0001 sample 4 9.6200 0.4538 396 21.20 <.0001 sample 5 7.2490 0.4538 396 15.97 <.0001 Analysis of Acceptability Using Mixed The Mixed Procedure Differences of Least Squares Means Standard Effect sample _sample Estimate Error DF t value Pr > Itl Adjustment Adj P sample 1 2 0.3240 0.5143 396 0.63 0.5291 Tukey-Kramer 0.9702 sample 1 3 1.1510 0.5143 396 2.24 0.0258 Tukey-Kramer 0.1680 sample 1 4 0.5960 0.5143 396 1.16 0.2472 Tukey-Kramer 0.7748 sample 1 5 2.9670 0.5143 396 5.77 <.0001 Tukey-Kramer <.0001 sample 2 3 0.8270 0.5143 396 1.61 0.1086 Tukey-Kramer 0.4932 sample 2 4 0.2720 0.5143 396 0.53 0.5972 Tukey-Kramer 0.9844 sample 2 5 2.6430 0.5143 396 5.14 <.0001 Tukey-Kramer <.0001 sample 3 4 -0.5550 0.5143 396 -1.08 0.2812 Tukey-Kramer 0.8173 sample 3 5 1.8160 0.5143 396 3.53 0.0005 Tukey-Kramer 0.0042 sample 4 5 2.3710 0.5143 396 4.61 <.0001 Tukey-Kramer <.0001 250 Output for Analyzing Intensity Produce Residuals Dataset in Analyzing intensity Using GLM The GLM Procedure Class Level Information Class Levels Values panelist 100 1 10 100 11 12 13 14 15 16 17 18 19 2 20 21 22 23 24 25 26 27 28 29 3 30 31 32 33 34 35 36 37 38 39 4 40 41 42 43 44 45 46 47 48 49 S 50 51 52 53 54 SS 56 57 58 59 6 60 61 62 63 64 65 66 67 68 69 7 70 71 72 73 74 75 76 77 78 79 8 80 81 82 83 84 85 86 87 88 89 9 90 91 92 93 94 95 96 97 98 99 sample 5 1 2 3 4 5 Number of Observations Read Number of observations Used 500 500 Produce Residuals Dataset in Analyzing intensity Using GLM Dependent Variable: intensity Source Model Error Corrected Total R—Square 0.464707 Source panelist sample Source panelist sample The GLM Procedure Sum Of DF Squares 103 3484.933160 396 4014.266120 499 7499.199280 Coeff Var 83.75963 DF Type I SS 99 2611.415280 4 873.517880 DF Type III 55 99 2611.415280 4 873.517880 251 Root MSE 3.183871 Mean Square F value 33.834303 3.34 10.137036 intensity Mean 3.801200 Mean Square F value 26.377932 2.60 218.379470 21.54 Mean Square F Value 26.377932 2.60 218.379470 21.54 Pr > F <.0001 Pr > F <.0001 <.0001 Pr > F <.0001 <.0001 Produce Residual Source panelist sample Produce Residual N Mean Std Deviation Skewness Uncorrected SS Coeff variation 5 Dataset in Analyzing intensity Using GLM The GLM Procedure Type III Expected Mean Square Var(Error) + S Var(panelist) Var(Error) + Q(sample) 5 Dataset in Analyzing intensity Using GLM The UNIVARIATE Procedure Variable: resid Moments 500 Sum weights 500 0 Sum observations 0 2.83630419 variance 8.04462148 0.47144594 Kurtosis 0.78466027 4014.26612 Corrected SS 4014.26612 . Std Error Mean 0.12684338 Basic Statistical Measures Location Variability Mean 0.00000 Std Deviation 2.83630 Median -0.18130 variance 8.04462 Mode -3.58480 Range 18.67300 Interquartile Range 3.56300 Tests for Location: Mu0=0 NOTE: The mode displayed is the smallest Of 7 modes with a count of 2. Test -Statistic- ----- p value ------ Student's t t 0 Pr > Itl 1.0000 Sign M -9 Pr >= |Ml 0.4471 Signed Rank S -2858 Pr >= ISI 0.3771 Tests for Normality Test --Statistic--- ----- p Value ------ Shapiro—wilk w 0.983963 Pr < w <0.0001 Kolmogorov-Smirnov 0 0.056283 Pr > D <0.0100 Cramer-von Mises w-Sq 0.301602 Pr > w-Sq <0.0050 Anderson-Darling A-Sq 1.990163 Pr > A-Sq <0.0050 Quantiles (Definition 5) Quantile Estimate 100% Max 10.2932 99% 7.6677 95% 5.0522 90% 3.6012 252 Produce Residuals Dataset in Analyzing intensity Using GLM The UNIVARIATE Procedure variable: reSTd Quantiles (Definition 5) Quantile Estimate 75% Q3 1.5342 50% Median -0.1813 25% Q1 -2.0288 10% -3.2743 5% -4.2033 1% —6.3913 0% Min -8.3798 Extreme Observations ----- Lowest---—- -----Highest----- value Obs Value Obs -8.3798 331 7.7352 250 -7.4758 207 8.5442 340 -7.0528 231 9.4402 369 —6.6758 348 9.7732 335 —6.5868 206 10.2932 75 Histogram # Boxplot 10.5+* 1 0 .* 2 0 .* 1 0 .*** 6 0 '*fit* 7 0 .«kes-t 8 I .ae******** 19 I oewannwnaaenn 23 I .**#*********** 27 I .***************#***fi************** 67 + _____ + .*********************************fi****** 80 I + I .****************************fl******fl*** 78 * _____ * '**************fl**#*******k** 56 I I .********#*k*****************k*t* 63 + ..... + .*********#****# 30 I .********fi 18 I .eaaa 8 I .‘k'h 3 I .* 2 0 -8.5+* 1 0 ----+----+----+----+----+----+-—--+----+ * may represent up to 2 counts 253 Produce Residuals Dataset in Analyzing intensity Using GLM The UNIVARIATE Procedure Variable: reSTd Normal Probability Plot 10.5+ * * * **fi*+ *** +++ **+++ ***+ ***+ +*** +***** ****** *#*** **** ***** ****+ ***** ****+ *i+ * -8.5+* +----+----+----+—---+----+----+----+----+—---+-—-—+ -2 -1 0 +1 +2 254 Analysis of intensity Using Mixed The Mixed Procedure Model Information Data Set WORK.SENSORY Dependent Variable intensity Covariance Structure variance Components Estimation Method REML Residual Variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment Class Level Information Class Levels values panelist 100 1 10 100 11 12 13 14 15 16 17 18 19 2 20 21 22 23 24 25 26 27 28 29 3 30 31 32 33 34 35 36 37 38 39 4 40 41 42 43 44 45 46 47 48 49 5 50 51 52 53 54 55 56 57 58 59 6 60 61 62 63 64 65 66 67 68 69 7 70 71 72 73 74 75 76 77 78 79 8 80 81 82 83 84 85 86 87 88 89 9 90 91 92 93 94 95 96 97 98 99 sample 5 1 2 3 4 5 Dimensions Covariance Parameters 2 Columns in x 6 Columns in Z 100 Subjects 1 Max Obs Per Subject 500 Number of Observations Number of Observations Read 500 Number of Observations Used 500 Number of Observations Not used 0 Iteration History Iteration Evaluations -2 Res Log Like Criterion 0 1 2711.87961304 255 Effect p sample sample sample sample sample sample sample sample sample sample sample #WWNNNI—‘I—‘I—‘H Effect sample sample sample sample sample _S Analysis of intensity Using Mixed The Mixed Procedure Iteration History Iteration Evaluations -2 Res Log Like Criterion 1 1 2668.96870974 0.00000000 Convergence criteria met. Covariance Parameter Estimates COV Parm Estimate panelist 3.2482 Residual 10.1370 Fit Statistics —2 Res Log Likelihood 2669.0 AIC (smal er is better) 2673.0 AICC (smaller is better) 2673.0 BIC (smaller is better) 2678.2 Type 3 Tests of Fixed Effects Num Den Effect DF DF F value Pr > F sample 4 396 21.54 <.0001 Least Squares Means Standard sample Estimate Error DF t value Pr > Itl 1 3.0080 0.3659 396 8.22 <.0001 2 2.9060 0.3659 396 7.94 <.0001 3 3.7970 0.3659 396 10.38 <.0001 4 2.9340 0.3659 396 8.02 <.0001 5 6.3610 0.3659 396 17.39 <.0001 Analysis of intensity Using Mixed The Mixed Procedure Differences of Least Squares Means Standard ample Estimate Error DF t value Pr > Itl Adjustment 2 0.1020 0.4503 396 0.23 0.8209 Tukey-Kramer 3 -0.7890 0.4503 396 -1.75 0.0805 Tukey-Kramer 4 0.0740 0.4503 396 0.16 0.8695 Tukey-kramer 5 -3.3530 0.4503 396 -7.45 <.0001 Tukey-Kramer 3 -0.8910 0.4503 396 -1.98 0.0485 Tukey-Kramer 4 -0.0280 0.4503 396 -0.06 0.9504 Tukey-Kramer 5 -3.4550 0.4503 396 -7.67 <.0001 Tukey-Kramer 4 0.8630 0.4503 396 1.92 0.0560 Tukey-kramer 5 -2.5640 0.4503 396 -5.69 <.0001 Tukey-Kramer 5 —3.4270 0.4503 396 -7.61 <.0001 Tukey-Kramer 256 A A CA HOA COO Adj .9994 .4032 .9998 .0001 .2783 .0000 .0001 .3101 .0001 .0001 APPENDIX 6 IIC CHROMATOGRAPH OF SAMPLE CONT IN DIP ANALYSIS (2“ SET) Max 15.0985 - _ . RT Mag. Abund. L 1‘3 A :5 22 I TIC2N{.\ .___/ *1. D 3432 81 ”55" WW ywao. 1 e. seas A ‘ ___-._..J. L... A—A l L‘L‘. - '— — vv 7‘ '7" :x_u—.w ~ .. '42. 2 P. 87Gb ’9 I AJAH-A ' w‘ "- bhe ._.A_ ‘ “ A ww—VATY‘ vr‘V‘ V a+l**“'7" " v2.3 7.6613 Abundance 9" uJ t *16.8 0.9318 $11.1 1.3625 *1.0 15.0985 i”1.3 11.7968 - . ‘ . . . 44.7 3.2355 6 air éoe see 803 ieee xzee Scan APPENDIX 7 IIC CHROMATOGRAPH OF SAMPLE PCTA IN DIP ANALYSIS (2ND SET) hflax:34n3CTl7 IQI‘ AnaSLlAtRHWd. 5 13 L5 as - ’N r TI- /I/ \k a) -" ’ «1.3 35‘7.4a g 55.9 - .. :1“. “Mi 3 33 534 A A I L Lfll.-L_ ;;9 -- *3.8 8.9??? g “A _ 1 A - 1_._ w,_ _ ‘2 623 .11. ‘ 1““ ‘_ "me ALML '5-3 6.4893 375 7 Iv , ‘f*3.9 8.7493 .1. I: ' -‘ ‘ “*> " "”' ' ' If» x3.1 10.9314 357 366 '” x3.s 6.9738 539 “ x3.1 11.153; 173 +81.o 34.3217 6;: .. "MET; ' 5.. ~ . - , “1.2 29.4743,- ~U~ 40» see sea zoae 1202 257 APPENDIX 8 TIC CHROMATOGRAMS 0F ODOR PROFILES FROM HDPE FILMS ANALYZED WITH SPME/GC-MS 3.00E+06 — 2.50E+06 ~- 2.00E+06 ~ 1 .50E+O6 -- 1 .OOE+06 ~ 5.00E+05 0.00E+OO ~ 1 .50 2.50 2.50E+06 2.00E+06 ~ 1 .50E+06 1 .OOE+06 ,, 5.00E+05 0.00E+OO -> 1 .50 Area 2.50E+06 - Nonanal 2.00E+06 - 1.50E+06 ~ Octanal 1 .OOE+O6 ,. 1 \ Dodecane Benzaldehyde l. 50054.05 i Acetone Heptanal W _L_.LIL l -J _ -- 6.50 \..L 5.50 4.50 7.50 1 .3-Di-tert-butylbenzene 2,4«bis(1 .1-dimethylethyl)-phenol Decanal Tetradecane Hexadecane Nonadecane l Li W A / K .i‘“ A. A‘ v .5 0.00E+OO 1 1:1: Hexanal 2.50 3.50 6.50 7.50 Time (min) 5.50 Figure A8.1 Total ion current chromatograms of odor profiles of triplicates of sample CONT 258 1.80E+06 ~ 1.6oE+os ~J L4OE+O6-< 1.20E+06 ~- 1.ooe+oe 8.00E+05 - 6.00E+05 —— 4.00E-I-05 « 2.00E+O5 kA __-. J‘A A._‘A_ .- A A A o.ooE+oo 4 1 .50 2.50 3.50 4.50 1 .60E+06 ‘ 1 .40E+06 » 1 ZOE-+06 ' 1 .OOE+06 8.00E+05 6.00E+05 4.00E+O5 2.00E+05 o.ooE+oo -« 4-4 LA 5.50 6.50 7.50 k AMAL-‘ M Ah“*4—-— 1 .50 2.50 3.50 4.50 IVea 1 .60E+06 » 1 .4OE+06 Nonanal ne 1 .20E+06 - Octanal 1 .DOE+06 - Decanal r 8.00E+O5 egos-+05 . Benzaldehyde Acetone Heptanal 4.ooe+os ~ 2.00E+05 ~— Dodecane \L Hexanal \ W 2.50 0.00E+OO * 1 .50 3.50 5.50 6.50 7.50 1 ,3—Di-tert-butylbenzene 2.4-bi (1 .1-dimethylethyl)-phenol Nonadecane l Hexadecane I \ 5.50 6.50 7.50 Time (min) Figure A8.2 Total ion current chromatograms of odor profiles of triplicates of sample PATA 259 2.00E+06 - 1.8OE+06 -‘ 1.60E+06 1.4OE+06 - 1.20E+06 ,. 1.00E+O6 . 8.00E+05 6.00E+05 4.00E+05 2.ooE+os a L L _ , _4- - LL- 1 ._ U o 0.00E+OO ~ 1 .50 2.50 3.50 4.50 5.5 6.50 7.50 1 .80E+06 1 .60E+06 1 .4OE+06 1 ZOE-+06 1 .OOE+06 8.00E+05 6.00E+05 . 4.00E+05 2.00E+05 l I I 0.00E+OO . , lLlLlJ ‘LA4_ . _ . _ 1 .50 2.50 3.50 4.50 5.50 6.50 7.50 Area 1 ~8°E+°6 1 .3-Dl-tert—butylbenzene 2,4-bis(1 ,1-dimethylethyl)-phenol 1 .eoe+oe 1 .4oe+oe Nonanal Octanal Decanal 1 .20E+06 1.OOE+06 ' Benzaldehyde 8.OOE+O5 ‘ Acetone Heptanal 6.00E+05 -' Tetradecane Hexadecane Nonadecane V , V l L I L o.ooe+oo 4 m-Lu 4A.- ”A.“ T... _ use. 1 .50 2.50 3.50 4.50 5.50 6.50 7.50 Time (min) Hexanal \ Dodecane 4.00E+05 - 2.00E+05 d Figure A8.3 Total ion current chromatograms of odor profiles of triplicates of sample PBTA 260 1.eoE+oe ~. 1.4OE+06 i 1.20E+06 ---‘ 1.oos+os a. 8.00E+05 -- 6.00E+05 l 4.00E+05 --‘ 2.00E+05 9 L.“ o.ooe+oo % 1 .50 2.50 3.50 4.50 1.8OE+O6 -- 1.60E+06 - 1.40E-I-06 - 1.20E+06 1.00E+06 8.00E+05 6.00E+05 ~ 4.00E+05 - 2.00E+05 1 7.50 LL1YL 0.00E+OO ? 1 .50 2.50 3.50 4.50 Area 1 .60IE+06 -- . 1 ,3—DI-teIt-butylbenzene 1 ACE-+06 1 1 .2OE+06 : Nonanal Octanal Decanal 1 .OOE+06 J 8.00E+05 «1 Benzaldehyde 6.00E+05 1 ‘ Heptanal i v 4-°°E+°5 “I Acetone l Dodecane Hexanal W 2.00E+05 v1 Tetradecane 5.50 6.50 7.50 2,4—bis(1 ,1-dimethylethyl)-phenol Hexadecane Nonadecane l 0.00E+OO 1 .50 3.50 2.50 6.50 7.50 Time (min) Figure A8.4 Total ion current chromatograms of odor profiles of triplicates of sample PBTB 261 2.SOE+06 1 2.00E+06 ~, 1.50E+06 — 1.00E+06 . 5.00E+05 — 0.00E+00 «Cw-14 . , kl.‘ ‘ ._ L A; 1 .50 2.50 3.50 4.50 5.50 6.50 7.50 2.sos+os - 2.00E+06 1,505+os . 1.00E+06 , 5.oos+05 a O_OOE+OO -_W J 1 k1 A A k - A—AL— A A 1.50 2.50 3.50 4.50 5.50 6.50 7.50 Area . 1.8OE+06 1,3-DI-tert-butylbenzene 1 .60E+06 J Nonanal ‘1’ “c’ 2,4-bis(1,1-dimethylethyl).phenol 1.4OE+06 - 0 § \1/ § 13 1.2oe+os — «I o ‘ Octanal § 3 c 1.ooe+oa - o m '- § . c 'c 8.00E+05 3 ‘3 Benzaldehyde o m 6.ooE+05 ; I _Nonadecane - Acetone \L 4.00E+05 ~ Hexanal / ‘1, 1 He tanal 2.005405 —' ‘l/ p : V o.ooE+oo - L“ - A- . -- 1 .50 2.50 3.50 4.50 5.50 6.50 7.50 Time (min) Figure A8.5 Total ion current chromatograms of odor profiles of triplicates of sample PCTA 262 APPENDIX 9 DESCRIPTION OF ODOR PROFILES USING GCIMS WITH ODO II SNIFFING PORT Investigator 1 PATA PCTA RT Description R.T. Description 1:38 Something minty, apple-like 1:35 Plastic 1:56 Stinky 2:20 Plastic - 2:25 Something odorous 2:25 Stinky 2:33 Sweet 2:30 Apple smell 2:37 Stinky 2:40 Stinky 2:44 Sweet 2:48 Sweet 2:54 Stinky 2:55 Something odorous 3:04 Plastic 3:02 Something odorous 3:12 Stinky 3:17 Sweet 3:20 Something odorous 3:26 Plastic 3:30 Sweet 3:34 Plastic 3:34 Sweet 3:39 Stinky, off Odor 3:45 Caramel 3:45 Caramel 3:52 Plastic 3:56 Something odorous 4:06 Something odorous 4:04 Caramel 4:14 Plastic 4:21 Sweet 4:21 Plastic 4:30 Something odorous 4:29 Something odorous 4:40 Plastic 4:44 Something odorous 4:54 Caramel 4:50 Candy, sweet 5:01 Somethingsweet 5:04 Detergent smell 5:10 Something sweet 5:12 Stinky 5:20 Plastic 5:26 Detergent smell, fresh 5:48 Something odorous 5:56 Detergent smell 6:13 Something odorous 6:17 Sweet 263 Investigator 2 PATA PCTA RT Description R.T. Description 1:33 gght odor 1:44 Light odor 1:53 Caramel (lightly) 1:54 Acid Odor 2:04 Buttery 2:10 Plastic-like odor 2: 1 6* Plastic-like Odor 2:1 7 Solvent, plastic 2:20 Plastic-like Odor 2:25 Solvent-like Odor 2:28 Stinky Odor 2:35* Strong Odor 2:33 Solvent-like Odor 2:40 Buttery, caramel 2:38* Strong stinky Odor 2:49 Buttery, caramel 2:45 Plastic, electric-like 2:54 Sweet Odor 2:54 Plastic-like 3:04 Stinky 3:08 Metallic 3:10 Mushroom-like Odor 3:15 Manure 3:15 Mushroom-like Odor 3:20 Metallic 3:20 Something burning 3:30 Buttery 3:30 Sweet 3235* Strong Odor 3:33 metallic 3:44 Buttery, caramel 3:41 * Stinky (strong), followed by caramel 3:54 Plastic 3:54 Stinky (strong), followed by caramel 3:58 Strong manure Odor 3:59 Manure 4:06 Plastic 4:05 Plastic 4:16 Plastic-like odor 4:17 Plastic 4:21 Rubbery 4:20 Balloon 4:31 Plastic 4:37 Off odor 4:43 Cardboard 4:44 Stinky 4:46 Cardboard 4250* Strong caramel 4:49 Stinky 4:55 Buttery 5:08 Metallic 5:04 Slightly off Odor 5:13 Balloon-like Odor 5:1 1 Stinky, plastic-like 5:25 Slightly Off odor 5:21 Off Odor 5:34 Slightly off Odor 5:47 Slightly Off Odor 5:48 Light odor 6:09 Balloon, rubbery Odor 6:10 Rubber band 7:27 Cardboard 7:20 Balloon 7:46 Plastic-like Odor * The perceived Odor was strong. 264 Investigator 3 PATA PCTA RT Description R.T. Description 0:56 Sweet 1 :28 Sweet 1 :22 Light smell 1:37 Butter coca 1:41 SomethinLOdorous 1:54 Stinky 1:50 Something sweet 2:02 Stinky 2:11 Plastic-like Odor 2:09 Plastic-like odor 2:15* Strong off Odor 2:17 Plastic-like Odor 2:21 * Strong stinky odor 2221* Perfume-like, plastic, stinky 2225* Strong plastic Odor 2:30 Fruity odor 2:30 Ripen fruit 2:37* Stinky, strong 2:37* Stinky, strong 2242 Plastic-like odor 2243 Perfume 2:49 Stinky 2252* Strong plastic 2:56 Ripen fruit 3:O4* Burning plastic 3202 Something burning 3:08 Burning plastic 3: 1 2 Plastic 3:18* Perfume 3:18 Perfume, fruity 3:29 Stinky 3:29 Burning wire 3234 Stale 3:36* Burning wire 3:40 Perfume 3:45 Overcooked caramel 3:45 Burning coca butter 3253* Strong perfume 3255 Something burning, sweet 4:06* Cologne 4209 Perfume, burning 4:17* Balloon 4:15 Perfume, burning 4226 Something odorous 4:20* Plastic, strong 4229* Strong, stinky 4:30 Burning rubber 4:40 Plastic 4239 Something odorous 4:45 Burning wire 4245 Dry paper sheet, burning 4:51* Perfume, followed by plastic 4256 Something burning 5:10 plastic 5:05 Burning, slightly sweet 5:21 Cardboard 5216 Light plastic scent 5226 Sweet, cardboard 5229 Plastic 5235 Light smell 5:42 Plastic, smoky 5:50 Somethm odorous 5259 Something odorous 6:00 Paperboard 6:12 Plastic, slightly sweet 6248 Sweet 6:22 Plastic 7:01 Coca butter 7:10 Plastic 7:21 Plastic 7:19 Something odorous 7258 Something burning, stinky 7257 Something odorous * The perceived Odor was strong. 265 APPENDIX 10 PLS MODELS BASED ON PAIRS OF SAMPLES AND THEIR VALIDATION DATA Validation data for CONT Training data for PCTA Pledictod CD a Training data for CONT Validation data for PCTA 7.00 7.110 0.00 950 9.00 9.50 10.00 10:50 11.00 (ExpoflmntalorActual) Figure A10.1 PLS plot of “Acceptability" scores of odor profiles of sample PCTA and CONT with both training data and validation data Table A10.1 Validate the PLS model based on sample PCTA and CONT Samples Actual Panel Predicted Average Opinion from scores Opinion scores predicted PLS model CONT_39 10.22 Acceptable 9.77 10.23 Acceptable CONT_40 10.22 Acceptable 10.41 10.23 Acceptable CONT_41 10.22 Acceptable 10.51 10.23 Acceptable PCTA_19 7.25 Neutral 7.85 7.64 Neutral PCTA_20 7.25 Neutral 7.57 7.64 Neutral PCTA 21 7.25 Neutral 7.51 7.64 Neutral 266 10.2w 111an 9.900: 9.500- 9.400- 9.200- 3.01]- 8.811]- 8.511]- 8.4m4 8.200- 8.000- 7.31]- 7.811]- 7.400— 7.200— HID- 8.313- Predicted Training data for PCTA Validation data for PATA Validation data for PCTA Training data for PATA Figure A10.2 PLS plot of “Acceptability” scores of odor profiles of sample PCTA and PATA with both training data and validation data Table A10.2 Validate the PLS model based on sample PCTA and PATA Samples Actual Panel Predicted Average Opinion from scores Opinion scores predicted PLS model PATA_1O 9.89 Acceptable 9.81 9.63 Acceptable PATA_1 1 9.89 Acceptable 9.08 9.63 Acceptable PATA_9 9.89 Acceptable 10.01 9.63 Acceptable PCTA_19 7.25 Neutral 7.37 7.61 Neutral PCTA_20 7.25 Neutral 7.73 7.61 Neutral PCTA 21 7.25 Neutral 7.73 7.61 Neutral 267 10E!!! 9500- Validation data 9400.». for PBTB 9.200-»—- ‘ ' 0000-9— ~ -- ; "”I 8.800” Trairfi'n data 3.5004 for POTgA Predicted oo '8: I Training data for PBTB .— __. _ - l I ' 1 Validation data 7'2"“ ‘9 \Lfor PCTA 7.000- ° s.uI-,—-———--50 ~ — -—1- - - 7.00 0501000 (Experimental or Actual) O Figure A103 PLS plot Of “Acceptability” scores of Odor profiles Of sample PCTA and PBTB with both training data and validation data Table A103 Validate the PLS model based on sample PCTA and PBTB Samples Actual Panel Predicted Average Opinion from scores Opinion scores predicted PLS model PBTB_29 9.62 Acceptable 9.83 9.29 Acceptable PBTB_3O 9.62 Acceptable 9.18 9.29 Acceptable PBTB_31 9.62 Acceptable 8.87 9.29 Acceptable PCTA_19 7.25 Neutral 7.35 7.23 Neutral PCTA_20 7.25 Neutral 7.16 7.23 Neutral PCTA_21 7.25 Neutral 7.18 7.23 Neutral 268 APPENDIX 11 RESPONSE AREAS OF ACETONE AND NONANAL BASED ON THE DATA FROM SPME/GC-MS ANALYSIS Table A11.1 lists the response areas of the selected ion for acetone (mlz 58) and nonanal (mlz 57) in the SPME/GC-MS analysis. Table A11.1 Response areas of ion mlz 58 for acetone and ion mlz 57 for nonanal detected in the Odor profiles Of different HDPE film samples in SPME/GC-MS analysis Acetone (mlz 58) Duplicate CONT PATA PBTA PBTB PCTA 1 12201 157529 293257 10832 358949 2 8210 137601 313017 18934 268136 3 14246 102985 334667 16335 220958 Average 11552 132705 313647 15367 282681 Nonanal (mlz 57) Duplicate CONT PATA PBTA PBTB PCTA 1 895002 124454 156424 93162 1675261 2 689027 403363 132386 144804 1912029 3 524490 120826 199861 140635 1294369 Average 702840 216214 162890 126200 1627220 Figures A11.1 and A112 are the standard mass spectra of acetone and nonanal, respectively (NIST, 2005), based on which all the mlz ions and their intensities can be determined (NIST, 2005) and the results were listed in Table A112. 269 43 15 '58 27 20 4O 60 BO 100120140150130200220240260 Figure A11.1 Standard mass spectmm of acetone (NIST, 2005) IDDLI-1 9’ 1 4 800- 1 1 600 3 43 98 4m — 29 . 70 20 40 60 so 100 120 110 160 183 200 220 240 260 Figure A11.2 Standard mass spectrum of nonanal (NIST, 2005) 270 Table A11.2 Ions and their intensities in the mass spectra of acetone and nonanal (NIST, 2005) Acetone mlz Intensity :mlz Intensity mlz Intensity mlz Intensity mlz Intensity 12 5 25 10 31 5 44 24 59 13 13 5 26 - 49 39 34 54 1 60 1 14 59 27 75 40 7 55 3 15 305 28 18 41 19 56 3 16 6 29 37 42 68 57 9 24 2 30 2 43 999 58 331 Nonanal mlz Intensity mlz Intensity mlz Intensity mlz Intensity *m/z . Intensity 27 212 43 584 67 204 84 9 113 18 28 44 44 478 68 292 85 53 114 168 29 381 45 142 69 327 86 27 124 115 31 27 54 124 70 381 95 292 141 9 39 142 55 469 71 159 96 230 412 9 40 18 56 575 81 248 97 9 41 549 57 999 82 319 98 593 42 177 58 27 83 88 99 53 Calculations were then made to calculate the percentage of the intensities of mlz 58 ion and mlz 57 in the total intensities Of all the ions of acetone and nonanal, respectively. Acetone 331 (m/258)°/o = x100% = 15.84% 2090 Nonanal 999 / 57 / = (m z )o" 8551 x100% = 11.68% 271 Based on the calculations shown above and the data in Table A11.1, the response areas of compound acetone and nonanal were determined (see Table A11.3). Table A11.3 Average response areas of acetone and nonanal detected in the odor profiles of different HDPE film samples in SPME/GC-MS analysis Sample CONT PATA PBTA PBTB PCTA Acetone 72929 837784 1980094 97014 1784602 Nonanal 6017466 1851 147 1 394606 1080479 13931678 272 APPENDIX 12 PLS MODELS BASED ON TRANSFORMED DATA OF PAIRS OF SAMPLES Data in Table 7.6 was Log10 transformed first (see Table A12.1), which was then used to build the PLS models. The predicted values Of the validation samples based on the models were then reverse-converted to the predicted response areas (see Tables A122 to A125). Table A12.1 Log10 transformed average response areas of acetone and nonanal Sample ‘ CONT PATA PBTA PBTB PCTA Acetone 4.86 5.92 6.30 4.99 6.25 Nonanal 6.78 6.27 6.14 6.03 7.14 Table A12.2 Predicted response areas Of acetone and nonanal Of sample CONT and PCTA by PLS model based on transformed data Acetone Nonanal Sample Prediction1 Ave.2 % Diff Prediction1 Ave.2 %Diff CONT_39 5.049226 6.833474 CONT_40 4.782503 72051 -1 6.757622 6010206 -.01 CONT_41 4.741 183 ' 6.745572 PCTA_19 5.941827 7.066130 PCTA_20 6.082370 1 1 1 1069 38 7.1 001 87 12339096 -1 1 PCTA_21 6.1 13026 7.1 07533 1. Predicted values Of the Log10 transformed data. 2. Average response area after reverse-converting the average of predicted values listed in the previous column. 273 Table A12.3 Predicted response areas Of acetone and nonanal of sample PATA and PCTA by PLS model based on transformed data Acetone Nonanal Sample Prediction‘ Ave? % Diff Prediction1 Ave? %Diff PATA_1 0 5.929363 6.293779 PATA_1 1 6.01 9338 893693 7 6.524975 2239450 21 PATA_9 5.904864 6.231 670 PCTA_19 6.234741 7.098314 PCTA_20 6.1 89243 1 599937 -10 6.974865 1 0355935 26 PCTA_21 6.188325 6.972389 1. Predicted values of the Log10 transformed data. 2. Average response area after reverse-converting the average of predicted values listed in the previous column. Table A12.4 Predicted response areas of acetone and nonanal of sample PBTA and PCTA by PLS model based on transformed data Acetone Nonanal Sample Prediction‘ Ave? % Diff . Prediction‘ Ave? %Diff PBTA_49 6.296240 6.209817 PBTA_50 6.300451 1989951 0.5 6.131689 1450481 4 PBTA_51 6.299836 6.143030 PCTA_19 6.251375 7.1 10338 PCTA_20 6.250697 1 786226 0.1 7.124948 12540379 -1 0 PCTA_21 6.253737 7.059646 1. Predicted values of the Log10 transformed data. 2. Average response area after reverse-converting the average of predicted values listed in the previous column. ' Table A12.5 Predicted response areas of acetone and nonanal of sample PBTB and PCTA by PLS model based on transformed data Acetone Nonanal Sample Prediction1 Ave? % Diff Prediction1 Ave? %Diff PBTB_29 4.891031 5.940023 PBTB_30 5.201855 141676 46 6.221 129 1495753 38 PBTB_31 5.361 001 6.363428 PCTA_1 9 6.1 89163 7.087778 PCTA_20 6.305519 1830209 3 7.187546 141471 19 2 PCTA_21 6.292820 7.176680 1. Predicted values of the Log10 transformed data. 2. Average response area after reverse-convening the average of predicted values listed in the previous column. 274 BIBLIOGRAPHY Acree, T. E., & Teranishi, R. (Eds.). (1993). Flavor Science: Sensible Principles and Techniques. Washington DC, USA: American Chemical Society. Aishima, T. (19913). Aroma Discrimination by Pattern Recognition Analysis of Responses from Semiconductor Gas Sensor Array. Journal of Agricultural and Food Chemistry, 39(4), 752-756. Aishima, T. (1991b). Discrimination of Varieties and Roasting Levels in Coffee Beans by Pattern Recognition Analysis of Responses from a Semiconductor Sensor Array. Paper presented at the Association Scientifique lntemationale du Cafe, 14e Colloque.1991.San Francisco, CA, USA. Alpha-MOS. (1996). Fox 3000 Electronic Nose Operating Manual. Toulouse, France: Alpha M.O.S. Alpha-MOS. (1997). Technical Note 1: Electronic Nose - Technology Overview. Toulouse, France: Alpha M.O.S. Alpha-MOS. (1999). Fox 3000 Electronic Nose Software Manual. Toulouse, France: Alpha M.O.S. Alpha-M.O.S. (2001). Technical Note 20: Quantifications with Sensor Array Systems. Toulouse, France: Alpha M.O.S. Alpha-M.O.S. (2002). Alpha MOS Operating Manual: AlphaFox/AlphaKronos/AlphaPrometheus. Toulouse, France: Alpha M.O.S. Anderson, L. D., Hernandez, R. J., Gray, |., 8. Harte, B. (1988). Release of Components from a Plastic Container during Microwave Heating. Packaging Technology and Science, 1(3), 1 17-121. Anon. (2006). Sensory Overload: How Scents are Adding a New Dimension to Packaging. Active Intelligent Packaging News, 4(2), 5-10. Arthur, C. L., & Pawliszyn, J. (1990). Solid Phase Microextraction with Thermal Desorption Using Fused Silica Optical Fibers. Analytical Chemistry, 62(19), 2145-2148. Arthur, C. L., Killam, L. M., Buchholz, K., Potter, D. W., Chai, M., Zhang, Z. Y., et al. (1992/1993). Solid Phase Microextraction: An Attractive Alternative. Environmental Laboratory(Dec/Jan), 10-14. 275 Arthur, C. L., Potter, D. W., Buchholz, K., Motlagh, S., & Pawliszyn, J. (1992a). Solid Phase Microextraction: for the Direct Analysis of Water: Theory and Practice. LC.GC, 10(9), 656-661. Arthur, C. L., Killam, L. M., Motlagh, S., Lim, M., Potter, D. W., & Pawliszyn, J. (1992b). Analysis of Substituted Benzene Compounds in Groundwater Using Solid-Phase Microextraction. Journal of Environmental Science and Technology, 26(5), 979-983. Arthur, C. L., Pratt, K., Motlagh, S., & Pawliszyn, J. (19920). Enviromental Analysis of Organic Compounds in Water Using Solid Phase Micro Extraction. Joumal of High Resolution Chromatography, 15(11), 741-744. Baigrie, B. (Ed .). (2003). Taints and Off-Flavours in Food. Boca Raton, FL, USA: CRC Press LLC. Bartlett, P. N., Elliott, J. M., & Gardner, J. W. (1997). Electronic Noses and Their Application in the Food Industry. Food Technology, 51(12), 44-48. Booker, J. L., 8 Friese, M. A. (1989). Safety of Microwave-Interactive Paperboard Packaging Materials. Food Technology, 43(5), 110-117. Boudreau, J. C. (1983). Chemical Senses and Food Flavor: An Overview. Paper presented at the Instrumental Analysis of Foods: Recent Progress, Proceedings of a Symposium of the 3rd International Flavor Conference, July 27-30, 1983 Corfu, Greece. Brody, A. L., 8 Marsh, K. S. (Eds.). (1997). The Wiley Encyclopedia of Packaging Technology (2nd ed.). New York, NY, USA: John Wiley 8 Sons, Inc. BSI. (1992). Glossary of Terms Relating to Sensory Analysis. London, UK: British Standard BS 5098; ISO 8589. Chai, M., Arthur, C. L., & Pawliszyn, J. (1993). Determination of Volatile Chlorinated Hydrocarbons in Air and Water with Solid-Phase Microextraction. Analyst, 118(12), 1501-1505. Chang, Y.-Y. (1996). Organic Vapor Permeability of High Barrier Polymer Membranes by a Dynamic Purge and Trap/Thermal Desorption Procedure. Michigan State University, East Lansing, MI, Master of Science. Chen, L. S., 8 Ou, A. S. M. (2001). Aroma Quality Analysis of Taiwan Local Tea by Sensory Evaluation and Electronic Nose. Paper presented at the IFT Annual Meeting.2001.New Orleans, Louisiana, USA. Chung, H.-Y. (2004). Evaluation of Light-Oxidized Off-Flavors in Reduced Fat Milk and Cheddar Cheese Using Sensory Evaluation and the Electronic Nose. Michigan State University, East Lansing, MI, Doctor of Philosophy. 276 Clark, T. J., 8 Bunch, J. E. (1997). Qualitative and Quantitative Analysis of Flavor Additives on Tobacco Products Using SPME-GC-Mass Spectrometry. Journal of Agricultural and Food Chemistry, 45(3), 844-849. Coello—Perez, M. S., Sanz, J., 8 Cabezudo, M. D. (1997). Analysis of Volatile Components of Oak Wood by Solvent Extraction and Direct Thermal Desorption - Gas Chromatography - Mass Spectrometry. Journal of Chromatography A, 778, 427-434. Culter, J. D. (1997). 'Electronic Nose' Helps Engineers Solve Packaging Material Problems. Part 1. Sensor Technology. Food, Cosmetics and Drug Packaging, 20(12), 231-235. Culter, J. D. (1998). 'Electronic nose' helps engineers solve packaging material problems. Part 2. Packaging material examples and experiment guidelines. Food, Cosmetics and Drug Packaging, 21(1), 15-19. Culter, J. D. (1999). The Control of Product and Packaging Quality with the Electroni Nose. TAPPI Journal, 82(6), 194-200. Czamecki, R. J. (1997). Solvent Retention and Odor in Flexo Packaging. Flexo, 22(6), 50-53. Das, R. (2003). Development of Electronic Nose Method for Evaluation of HDPE Data, Correlated with Organeleptic Testing. Michigan State University, East Lansing, MI, Master of Science. Delmonte, J. (1947). The Technology of Adhesives. New York, NY, USA: Reinhold. Delpha, C., Siadat, M., 8 Lumbreras, M. (2001). An Electronic Nose for Discrimination of Forane 1343 and Carbon,Dioxide in Humidity Control Atmosphere. Sensor and Actuators B, 78, 49-56. Demertzis, P. G., Franz, R., 8 Welle, F. (1999). The Effects of Gamma-lrridiation on Compositional Changes in Plastic Packaging Films. Packaging Technology and Science, 12(3), 119-130. Deventer, D. V., 8 Mallikarjunan, P. (2002a). Comparative Performance Analysis of Three Electronic Nose Systems Using Different Sensor Technologies in Odor Analysis of Retained Solvents on Printed Packaging. Journal of Food Science, 67(8), 3170-3183. Deventer, D. V., 8 Mallikarjunan, P. (2002b). Optimizing an Electronic Nose for Analysis of Volatiles from Printing Inks on Assorted Plastic Films. Innovative Food Science 8 Emerging Technologies, 3(1), 93-99. 277 Devos, M., Patte, F., Rouault, J., Laffort, P., 8 Van-Gemert, L. J. (1990). Standardized Human Olfactory Thresholds. New York, NY, USA: Oxford University Press. Durst, G. L., 8 Laperle, E. A. (1990). Styrene Monomer Migration as Monitored by Purge and Trap Gas Chromatography and Sensory Analysis for Polystyrene Containers. Journal of Food Science, 55(2), 522-524. Dynatherm. (1989). Thermal Desorption Unit Model 890/891: Installation and Operation Manual. New Jersey, USA: CDS Analytical, Inc. Esteban, J. L., Martinez-Castro, |., 8 Sanz, J. (1993). Evaluation and Optimization of the Automatic Thermal Desorption Method in the Gas Chromatographic Determination of Plant Volatile Compounds. Journal of Chromatography A., 657, 155-164. Ettre, L. S. (2001). Headspace - Gas Chromatography: An Ideal Technique for Sampling Volatiles Present in Non-Volatile Matrices. In R. L. Rouseff 8 K. R. Cadwallader (Eds.), Headspace Analysis of Foods and Flavors: Theory and Practice (pp. 9-32). New York, NY, USA: Kluwer Academic/Plenum Publishers. Fazzalari, F. A. (1978). Compliation of Odour and Taste Threshold Values Data. Philadelphia, PA, USA: American Society for Testing and Materials ASTM Data Series DS 48A. Ferreira, V., Lopez, R., 8 Aznar, M. (2002). .Olfactometry and Aroma Extract Dilution Analysis of Wines. In J. F. Jackson 8 H. F. Linskens (Eds.), Analysis of Taste and Aroma (pp. 89-122). New York, NY, USA: Springer-Verlag. Forsgren, G., 8 Sjostrom, J. (1997). Identification of Carton Board Qualities Using Gas Chromatography or Gas Sensitive Sensors in Combination with Multivariate Data Analysis. Nordic Pulp & Paper Research Journal, 12(4), 276- 281. Forsgren, G., H., F., 8 Ericsson, B. (1999). Taint and Odour Related Quality Monitoring of Two Food Packaging Board Products Using Gas Chromatography, Gas Sensors and Sensory Analysis. Nordic Pulp 8 Paper Research Journal, 14(1), 5-16. Freire, M. T. d. A., Castle, L., Reyes, F. G. R., 8 Damant, A. P. (1998). Thermal Stability of Polyethylene Terephthalate Food Contact Materials: Formation of Volatiles from Retail Samples and Implications for Recycling. Food Additives and Contaminants, 15(4), 473-480. Freire, M. T. d. A., Damant, A. P., Castle, L., 8 Reyes, F. G. R. (1999). Thermal Stability of Polyethylene Terephthalate (PET): Oligomer Distribution and Formation of Volatiles. Packaging Technology and Science, 12(1), 29-36. 278 Galotto, M., 8 Guarda, A. (1999). Comparison between Thermal and Microwave Treatment on the Overall Migration of Plastic Materials Intended to be in Contact with Foods. Packaging Technology and Science, 12(6), 277-281. Gardner, J. W., 8 Persaud, K. C. (Eds). (2000). Electronic Noses and Olfaction 2000: Proceedings of the 7th lntemational Symposium on Olfaction and Electronic Noses, Brighton, UK, July 2000. Philadelphia, PA, USA: Institute of Physics Publishing. Gameau, F.-X., RiedI, B., Hobbs, s, Pichette, A., & Gagnon, H. (2004). The Use of Sensor Array Technology for Rapid Differentiation of the Sapwood and Heartwood of Eastern Canadian Spruce, Fir and Pine. HoIz Roh. Werkst, 62(6), 470-473. Goulding, T. M. (1994). Pressure-Sensitive Adhesives. In A. Pizzi 8 K. L. Mittal (Eds), Handbook of Adhesive Technology (1st ed., pp. 549-564). New York, NY, USA: Marcel Dekker, Inc. Griffith, N. M. (1974). Sensory Properties of the ChIoroanisoles Chemical Senses and Flavour, 1, 187-195. Grimm, C. C., Lloyd, S. W., Miller, J., 8 Spanier, A. M. (2002). The Analysis of Food Volatiles Using Direct Thermal Desorption. In R. Marsili (Ed.), Flavour, Fragrance, and Odor Analysis (pp. 55—73). New York, NY, USA: Marcel Dekker, Inc. A Gruner, A., 8 Piringer, O. (1999). Component Migration from Adhesives Used in Paper and Paperboard Packaging for Foodstuffs. Packaging Technology and Science, 12(1), 19-28. Gutcho, M. (Ed.). (1983). Adhesives Technology: Developments Since 1979 (1st ed.). Park Ridge, NJ, USA: Noyes Data Corporation. Harper, W. J., 8 Kleinhenz, J. P. (1998). Factors Affecting Sensory and Electronic Nose Threshold Values for Food Aroma Compounds. In W. J. Hurst (Ed.), Electronic Noses 8 Sensor Ana y Based Systems : Design 8 Applications : Proceedings of the 5th lntemational Symposium on Olfaction and the Electronic Nose (pp. 308-317). Lancaster, PA, USA: Technomic Publishing Co. Harper, W. J. (2001). The Strengths and Weaknesses of the Electronic Nose. In R. L. Rouseff 8 K. R. Cadwallader (Eds), Headspace Analysis of Foods and Flavors: Theory and Practice (pp. 59-71). New York, NY, USA: Kluwer Academic/Plenum Publishers. Hartman, T. G., Lech, J., 8 Rosen, R. T. (1991a). Determination of Off-Odors and Other Volatile Organics in Food Packaging Films by Direct Thermal Analysis GC-MS. Paper presented at the ASMS 39th Conference on Mass Spectrometry and Allied Topics, May 19-24, 1991 Nashville, TN, USA. 279 Hartman, T. G., Lech, J., 8 Rosen, R. T. (1991b). Breath Analysis of Garlic-Dome Phytochemicals in Human Subjects by Combined Adsorbent Trapping, Short Path Thermal Desorption GC-MS. Paper presented at the ASMS 39th Conference on Mass Spectrometry and Allied Topics, May 19-24, 1991 Nashville, TN, USA. Hartman, T. G., Overton, S. V., Manura, J. J., Baker, C. W., 8 Manos, J. N. (1991c). Short Path Thermal Desorption: Food Science Applications. Journal of Food Technology, 45(7), 104-105. Haugen, J.-E. (2001). Electronic Noses in Food Analysis. In R. L. Rouseff 8 K. R. Cadwallader (Eds), Headspace Analysis of Foods and Flavors: Theory and Practice (pp. 43-57). New York, NY, USA: Kluwer Academic/Plenum Publishers. Heinio, R. L., 8 Ahvenainen, R. (2002). Monitoring of Taints Related to Printed Solid Boards with an Electronic Nose. Food Additives and Contaminants, 19(Supplement), 209-220. Helsper, J. P. F. G., Davies, J. A., 8 Verstappen, F. W. A. (2002). Analysis of Rhythmic Emission of Volatile Compounds of Rose Flowers. In J. F. Jackson 8 H. F. Linskens (Eds), Analysis of Taste and Aroma (pp. 199-223). New York, NY, USA: Springer-Verlag. Heydanek, M. G. (1978). How to Spot a Stinker - Predicting Flavour Effects of Packaging Materials. Food Product Development, 12(5), 17-18. Hill, P. G., 8 Smith, R. M. (2000). Determination of Sulphur Compounds in Beer Using Headspace Solid-Phase Microextraction and Gas Chromatographic Analysis with Pulsed Flame Photometric Detection. Journal of Chromatography A, 872(2), 203-213. Ho, Y., Yam, K., Young, S., 8 Zambetti, P. (1994). Comparison of Vitamin E, lrganox 1010 and BHT as Antioxidants on Release of Off-Flavor from HDPE Bottles. Journal of Plastic Film 8 Sheeting, 10(3), 194-212. Hodgins, D. (1997). The Electronic Nose: Sensory Array-Based Instruments that Emulate the Human Nose. In R. Marsili (Ed.), Techniques for Analyzing Food Aroma (pp. 331-371). New York, NY, USA: Marcel Derek, Inc. Hollifleld, H. C., Breder, C., Dennison, J., Roach, J., 8 Adams, W. (1980). Container-Derived Contamination of Maple Sirup with Methyl Methacrylate, Toluene, and Styrene as Determined by Headspace Gas-Liquid Chromatography. J. Assoc. Off. Anal. Chem, 63(2), 173-177. Horwood, J. F., T., L. G., Rawshaw, E. H., 8 Stark, W. (1981). An Off-Flavour Associated with the Use of Sorbic Acid during Feta Cheese Maturation. Australian JoumaI of Dairy Technology. 36(1), 38-40. 280 Hurst, W. J. (1998). Electronic Noses 8 Sensor Array Based Systems : Design 8 Applications Paper presented at the The 5th lntemational Symposium on Olfaction and the Electronic Nose, September 27-30, 1998 Baltimore, MA, USA. IFT. (1975). Minutes of Sensory Evaluation Division. Paper presented at the Annual Meeting of Institute of Food TechnologistsJune 10, 1975.Chicago, IL, USA. IOPP. (1995). Adhesives in Packaging: Principles, Properties, and Glossary. Hemdon, VA, USA: Institute of Packaging Professionals. IOPP. (2002). Adhesives in Packaging: analysis of the types of adhesives, their specifications and properties. Naperville, IL, USA: Institute of Packaging Professionals. Jackson, J. F ., 8 Linskens, H. F. (Eds). (2002). Analysis of Taste and Aroma. New York, NY, USA: Springer-Verlag Berlin Heidelberg. Jeon, D.-H., Park, G. Y., Kwak, I.-S., Lee, C.-W., Park, H.-J., 8 Lee, K.-H. (2004). Effects of Irridiation on Residues and Migration Levels of Antioxidants and Their Degradation Products of LDPE Packaging Film. Paper presented at the 2004 IFI' Annual Meeting.JuIy 12-16, 2004.Las Vegas, NV, USA. Kanavouras, A. (2003). Quality Issues and Shelf Life Evaluation of Packaged Olive Oil in Glass and Plastic Containers. Michigan State University, East Lansing, MI, Doctor of Philosophy. Kilcast, D. (1996). Sensory Evaluation of Taints and Off-Flavours. In M. J. Saxby (Ed. ), Food Taints and Off-Flavours (pp. 1-40). Glasgow, UK: Blackie Academic 8 Professional. Kilcast, D. (2003). Sensory Analytical Methods in Detecting Taints and Off- Flavours in Food. In B. Baigrie (Ed.), Taints and Off-Flavours in Food (pp. 5-30). Boca Raton, FL, USA: CRC Press LLC. Kim, H., Gilbert, S. G., 8 Hartman, T. G. (1988). Characterization of Undesirable Volatile Compounds in PVC Films. Paper presented at the Frontiers of Flavor: 5th International Flavor Conference.July 1-3, 1987.Porto Karras, Chalkidiki, Greece. Koziel, J., Y., J. M., 8 Pawliszyn, J. (2000). Air Sampling with Porous Solid- Phase Microextraction Fibers. Journal of Analytical Chemistry, 72(21), 5178- 5186. Krzanowski, W. J. (1988). Principles of Multivariate Analysis. New York, NY, USA: Oxford University Press. 281 Krzymien, M. E., Carlsson, D. J., Deschenes, L., 8 Mercier, M. (2001). Analyses of Volatile Transformation Products from Additives in Gamma-lrridiated Polyethylene Packaging. Joumal of Additives and Contaminations, 18(8), 739- 749. Kwo, H.-M. (1991). Measurement of Volatiles from a Susceptor Material Using a Thermal Desorption Method. Michigan State University, East Lansing, MI, Master of Science. Laoharavee, T. (1998). The Permeability of Binary Organic Vapor Mixtures through Polymer Membranes by a Dynamic Purge and Trap/Thermal Desorption Procedure: Theoretical and Practical Considerations. Michigan State University, East Lansing, MI, Master of Science. Larson, M. (1991). Taste and Value Drive Microwave Foods. Packaging 36(2), 33-35. Lawless, H. T., 8 Heymann, H. (1998). Sensory Evaluation of Food: Principles and Practices. New York, NY, USA: Chapman 8 Hall. Lee, J. H., Sung, T. H., Lee, K. T., 8 Kim, M. R. (2004). Effect of Gamma- lrridiation on Color, Pungency, and Volatiles of Korean Red Pepper Powder. Joumal of Food Chemistry and Toxicology, 69(8), C585-C592. Lord, T. (2003). Packaging Materials as a Source of Taints. In B. Baigrie (Ed.), Taints and Off-Flavours in Food (pp. 64-111). Boca Raton, FL, USA: CRC Press LLC. Louch, D., Motlagh, S., 8 Pawliszyn, J. (1992). Danymics of Organic Compound Extraction from Water Using Liquid Coated Fused Silica fibers. Journal of Analytical Chemistry, 64(10), 1 187-1 199. MacRae, A. W., 8 Falahee, M. (1995). Theoretical Note on a Practical Problem: Effective Screening of Drinking Water for Taints. Food Quality and Preference, 6(2), 69-74. Maehashi, K., 8 Arai, S. (2002). Taste Evaluation for Peptides in Protein Hydrolysates from Soybean and Other Plants. In J. F. Jackson 8 H. F. Linskens (Eds), Analysis of Taste and Aroma (pp. 46-68). New York, NY, USA: Springer- Verlag. Maneesin, P. (2001). GC-MS and Electronic Nose Analysis on Off-Flavor Components in HDPE Containers and Correlation with Sensory Evaluation. Michigan State University, East Lansing, MI, Doctor of Philosophy. Manura, J. J., 8 Hartman, T. G. (1992). Applications of a Short-Path Thermal Desorption GC Accessory. American Laboratory, 24(8), 46-53. 282 Marsili, R. (Ed.). (1997). Techniques for Analyzing Food Aroma (Vol. 79). New York, NY, USA: Marcel Dekker, Inc. Marsili, R. (Ed.). (2002). Flavor, Fragrance, and Odor Analysis. New York, NY, USA: Marcel Dekker, Inc. Martos, P., 8 Pawliszyn, J. (1997). Calibration of Solid Phase Microextraction for Air Analyses Based on Physical Chemical Properties of the Coating. Journal of Analytical Chemistry, 69(2), 206-215. Mazza, G., 8 Pietzak, E. M. (1990). Headspace Volatiles and Sensory Characteristics of Earthy, Musty Flavoured Potatoes. Food Chemistry, 36(2), 97- 112. McNeaI, T. P., 8 Hollifield, H. C. (1993). Determination of Volatile Chemicals Released from Microwave-Heat-Susceptor Food Packaging. Journal of AOAC lntemational, 76(6), 1268-1275. Meilgaard, M. C. (1982). Prediction of Flavor Differences Between Beers from Their Chemical Composition. Journal of Agricultural and Food Chemistry, 30(6), 1009-1017. Meilgaard, M. C., CiviIIe, G. V., 8 Carr, B. T. (1999). Sensory Evaluation Techniques. Boca Raton, FL, USA: CRC Press LLC. Mermelstein, N. H. (1997). Electronic Nose. Journal of FOod Technology. 51(12), 46-48. Mielle, P. (1996). 'Electronic Noses': Towards the Objective Instrumental Characterization of Food Aroma. Trends in Food Science 8 Technology, 7(12), 432-438. Morales, M. T., Rios, J. J., 8 Aparicio, R. (1997). Changes in the Volatile Composition of Virgin Olive on during Oxidation: Flavors and Off-Flavors. Journal of Agricultural and Food Chemistry, 45(7), 2666-2673. Motion, R., Franich, R., 8 Song, G. (2004). Evaluation of a Portable Electronic Nose for Identification of Wood Chips. Paper presented at the 58th Appita Annual Conference and Exhibition: Incorporating the Pan Pacific Conference.April 19-21, 2004.Canberra, Australia. NIST. (2005). NIST Chemistry WebBook. NIST Standard Reference Database Number 69, June 2005 Release, from National Institute of Standards and Technology: http://webbook.nistgov/chemistry. Nunez, A., 8 Gonzalez, L. F. (1984). Pre-Concentration of Headspace Volatiles for Trace Organic Analysis by Gas Chromatography. Journal of Chromatography, 300, 127-162. 283 Ott, R. L., 8 Longnecker, M. (2001). An Introduction to Statistical Methods and Data Analysis (5th ed.). Pacific Grove, CA: Duxbury. Parliment, T. H. (1997). Solvent Extraction and Distillation Techniques. In R. Marsili (Ed.), Techniques for Analyzing Food Aroma (pp. 1-26). New York, NY, USA: Marcel Dekker, Inc. Pawliszyn, J. (1993). Kinetic Model of Supercritical Fluid Extraction. Journal of Chromatographic Science, 31(1), 31-37. Peled, R., 8 Mannheim, C. H. (1977). Off-Flavors from Packaging Materials. Modern Packaging, 50(1), 45-48. Pizzi, A., 8 Mittal, K. L. (Eds). (1994). Handbook of Adhesive Technology (1 st ed.). New York, NY, USA: Marcel Dekker, Inc. Poste, L. M., Mackie, D. A-., Bulter, G., 8 Larrnond, E. (1991). Laboratory Methods for Sensory Analysis of Foods (1 st ed.). Ottawa, Canada: Canada Communication Group PublishingCenter. Potter, D. W., 8 Pawliszyn, J. (1992). Detection of Substituted Bezenes in Water at the pg/ml Level Using Solid-Phase Microextraction and Gas Chromatography- Ion Trap Mass Spectrometry. Journal of Chromatography, 625, 247-255. Potter, D. W., 8 Pawliszyn, J. (1994). Rapid Determination of Polyaromatic Hydrocarbons and PonchIorinated Biphenyls in Water Using Solid-Phase Microextraction and GC/MS. Journal of Environmental Science and Technology, 28(2), 298-305. Reid, W. J. (2003). Instrumental Methods in Detecting Taints and Off-Flavours. In B. Baigrie (Ed.), Taints and Off-Flavours in Food (pp. 31-64). Boca Raton, FL, USA: CRC Press. Rencher, A. C. (1995). Methods of Multivariate Analysis (1 st ed.). New York, NY, USA: John Wiley 8 Sons Inc. Rosenberg, L. (1985). Pressure Sensitive Adhesives: FDA Regulations. Paper presented at the 1985 Polymers, Laminations and Coatings Conference.September 9 -11, 1985.Chicago, IL, USA. Ruberto, G. (2002). Analysis of Volatile Components of Citrus Fruit Essential Oils. In J. F. Jackson 8 H. F. Linskens (Eds), Analysis of Taste and Aroma (pp. 123- 157). New York, NY, USA: Springer-Verlag. Saxby, M. J. (1992). Index of Chemical Taints: Leatherhead Food Research Association. 284 Saxby, M. J. (1996a). Food Taints and Off-Flavours (2nd ed.). Bishopbriggs, Glasgow, UK: Blackie Academic 8 Professional. Saxby, M. J. (1996b). A Survey of Chemicals Causing Taints and Off-flavours in Food. In M. J. Saxby (Ed.), Food Taints and Off-Flavors (pp. 41-72). Glasgow, UK: Blackie Academic 8 Professional. Schiffman, H. R. (2001). Sensation and Perception: An Integrated Approach. New York, NY, USA: John Wiley and Sons Inc. SGE. (2002). Olfactory Detector Outlet Sniffer Unit - "ODO II". San Antonio, TX, USA: SGE Analytical Science. SGE. (2003). ODO II Installation and Operating Instructions (12 ed.). San Antonio, TX, USA: SGE Analytical Science. Siripatrawan, U. (2002). Electronic Sensor Array Incorporating Artificial Neural Network Algorithms for Rapid Identification and Quantification of Escherichia CoIi and Salmonella Enterica Serovar Typhimurium and Their Volatile Metabolites. Michigan State University, East Lansing, MI, Doctor of Philosophy. Siripatrawan, U., Linz, J. E., 8 Harte, B. (2004a). Solid-Phase Microextraction, Gas Chromatography, and Mass Spectrometry Coupled with Discriminant Factor Analysis and Multilayer Perception Neural Network for Detection of Escherichia coli. Journal of Food Protection, 67(8), 1597-1603. Siripatrawan, U., Linz, J. E., 8 Harte, B. (2004b). Rapid Method for Prediction of Escherichia coli Numbers Using an Electronic Sensor Array and an Artificial Neural Network. Journal of Food Protection, 67(8), 1604-1609. Siripatrawan, U., Linz, J. E., 8 Harte, B. (20063). Electronic Sensor Array Coupled with Artificial Neural Network for Detection of Salmonella Typhimurium. Sensor and Actuators B, 1 19, 64-69. Siripatrawan, U., Linz, J. E., 8 Harte, B. (2006b). Detection of Escherichia coli in Packaged Alfalfa Sprouts with an Electronic Nose and an Artificial Neural Network. Journal of Food Protection, 69(8), 1844-1850. Soroka, W. (2002). Fundamentals of Packaging Technology (3rd ed.). Naperville, IL, USA: Institute of Packaging Professionals. Sunesson, A. L., Nilsson, C. A., Andersson, B., 8 Carlson, R. (1992). Thermal Desorption Cold Trap-Injection in High-Resolution Gas Chromatography: Multivariate Optimization of Experimental Conditions. Joumal of Chromatography 623, 93-103. Supelco. (1998a). Supelco Chromatography Products: 1998 Catalog. Bellefonte, PA, USA: Sigma-Aldrich Co. 285 Supelco. (1998b). Solid Phase Microextraction: Theory and Optimazation of Conditions, Bulletin 923. Bellfonte, PA, USA: Sigma-Aldrich Co. Supelco. (19980). Solid Phase Microextraction of Volatile Compounds, Bulletin 11. Bellfonte, PA, USA: Sigma-Aldrich Co. Supelco. (2000). Solid Phase Microextraction: Solvent/ess Sample Preparation for Monitoring Flavor Compounds by Capillary Gas Chromatography, Bulletin 869C. Bellfonte, PA, USA: Sigma-Aldrich Co. Supelco. (2001a). A Practical Guide to Ouantitation with Solid Phase Microextraction, Bulletin 929. Bellfonte, PA, USA: Sigma-Aldrich Co. Supelco. (2001 b). Solid Phase Microextraction Troubleshooting Guide, Bulletin 928. Bellfonte, PA, USA: Sigma-Aldrich Co. Supelco. (2001c). Solid Phase Microextraction Applications Guide, Bulletin 925. Bellfonte, PA, USA: Sigma-Aldrich Co. Supelco. (2004). Air Sampling of VOCs by SPME for Analysis by Capillary GC, Application Note 141. Bellfonte, PA, USA: Sigma-Aldrich Co. Supelco. (2005). Supelco Chromatography Products for Analysis and Purification: 2005 - 2006 Catalog. Bellefonte, PA, USA: Sigma-Aldrich Co. Tan, T., Lucas, 0., Moy, L., Gardner, J. W., 8 Bartlett, P. N. (1995). The Electronic Nose - A New Instrument for Sensing Vapors. LC-GC lnt., 8(4), 218- 225. Tombesi, N. B., 8 Freije, H. (2002). Application of Solid-Phase Microextraction Combined with Gas Chromatography-Mass Spectrometry to the Determination of Butylated Hydroxytoluene in Bottled Drinking Water. Journal of Chromatography A, 963(1), 179-183. Vetenskapsakademien, K. (2000). The Nobel Price in Chemistry 2000 - Information for the Public. Sweden: The Royal Swedish Academy of Sciences. Wellnitz-Ruen, W., Reineccius, G. A., 8 Thomas, E. L. (1982). Analysis of the Fruity Off-Flavour in Milk Using Headspace Concentration Capillary Column Gas Chromatography. Joumal of Agricultural and Food Chemistry, 30(3), 512-514. Werkhoff, P., 8 Bretschneider, W. (1987a). Dynamic Headspace Gas Chromatography: Concentration of Volatile Components After Thermal Desorption by lnterrnediate Cryofocusing in a Cold Trap, I. Principle and Applications. Journal of Chromatography, 405, 87-98. Werkhoff, P., 8 Bretschneider, W. (1987b). Dynamic Headspace Gas Chromatography: Concentration of Volatile Components After Thermal 286 Desorption by Intermediate Cryofocusing in a Cold Trap, II. Effect of Sampling and Desorption Parameters on Recovery. Journal of Chromatography, 405, 99- 106. Whitfield, F. B., Freeman, D. J., 8 Shaw, K. J. (1983). Trimethylarsine: An Important Off-Flavour Component in Some Prawn Species. Chemistry Industry, 20, 786-787. Wieczodek, J. (1990). Microwave Product Packaging: Choosing the Right Adhesives. Joumal of Packaging Technology, 4(4), 24-26. Willing, B.-I. L., Brundin, A., 8 Lundstrom, I. (1998). Odour Analysis of Paperboard, the Correlation between Human Senses and Electronic Sensors Using Multivariate Analysis. Packaging Technology and Science, 11(2), 59-67. Yang, X. G., 8 Peppard, T. (1994). Solid-Phase Microextraction for Flavor Analysis. Journal of Agricultural and Food Chemistry, 42(9), 1925-1930. Yuzay, I. E. (2004). Development of Electronic Nose Method for Evaluation of Residual Solvents in Low Density Polyethylene. Michigan State University, East Lansing, MI, Master of Science. Zhang, Z. Y., 8 Pawliszyn, J. (1993). Headspace Solid-Phase Microextraction. Analytical Chemistry, 65(14), 1843-1852. Zhang, Z. Y., Yang, M. J., 8 Pawliszyn, J. (1994). Solid-Phase Microextraction. Analytical Chemistry, 66(17), 884A-853A. Ziegleder, G. (1998). Volatile and Odorous Compounds in Unprinted Paperboard. Packaging Technology 'and Science, 11(5), 231-239. 287 lIlflllllllfllllfllllfllllMilli