THE THERMOSONIMETRY OF POLYMERSbyKEVIN ANDREW SOULSBURYB.Sc. (Hons.), The University of Surrey, UK, 1987.A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIES(DEPARTMENT OF CHEMISTRY)We accept this thesis as conformingto the required standard/.THE UNIVERSITY OF BRITISH COLUMBIAAPRIL 1994© Kevin A. Soulsbury, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(SignaDepartment of____________________The University of British ColumbiaVancouver, CanadaDate 1\iDE-6 (2/88)IIAbstractThermosonimetry is the measurement of sound (acoustic emission) emitted froma substance as a function of temperature, while the substance is heated in a controlledmanner. In this work the thermo-oxidative decompositions of poly(vinyl chloride),polyethylene, polypropylene, poly(ethylene terephthalate) and ethylene-vinyl acetatehave been studied using thermosonimetry. The amount of acoustic emission producedwas dependent on the polymer, the heating rate and the sample mass but not themolecular mass of the polymer. Thermosonimetry was also used to obtain apparentactivation energies and the reaction orders for the thermo-oxidative decompositionprocesses under a variety of experimental conditions. Typically, the activation energiesobtained were higher than values reported in the literature obtained using thermalanalysis methods such as thermogravimetry.An apparatus has been developed that permits simultaneous thermogravimetricanalysis (TG) and thermosonimetry (TS). The acoustic emission commenced at themaximum rate of mass loss, and confirmed that thermosonimetry was not suitable formeasuring the relative thermal stability of polymers. Extensive chemical analysis of thepolymer residues was used to study the decomposition mechanisms but did not revealany direct link between the chemical structure of the polymers and the acousticemission. Microscopic analysis of the residues, using scanning electron microscopy,suggested that the physical nature of the residues and the physical processesoccurring during thermo-oxidative decomposition determined whether acousticemission was produced. For PVC and PET, gas evolution and fracture processes werelinked to the production of acoustic emission. For PE and PP, which did not producesignificant amounts of acoustic emission, these processes were not observed to thesame degree.IIIFurther studies sought to investigate whether acoustic emission was producedduring other polymer processes. Polymerization and phase transitions did not produceacoustic emission. The crystallization of isotactic polypropylene was found to produceacoustic emission. The acoustic emission had been linked to the formation of cavitiesbetween the polymer spherulites. Thermosonimetry was also used to measure therates of crystallization from the isothermal crystallization of polypropylene.ivTable of ContentsAbstract iiTable of Contents ivList of Tables xiiiList of Figures xviiGlossary xxvAcknowledgments xxixCHAPTER 1. INTRODUCTION I1.1 Polymer science 21.1.1 Polymerization processes 21.1.1.1 Step-growth polymerization 31.1.1.2 Chain-growth polymerization 31.1.2 Copolymerization 51.1.3 Chemical properties of polymers 61.1.4 Polymer microstructure 61.1.5 Physical states of polymers 81.1.6 Thermal transitions 91.1.7 Crystallization 101.1.7.1 Crystallinity 121.1.7.2 Morphology 121.1.8 Amorphous polymers 131.1.8.1 Plasticization 141.2 Analytical methods used in polymer science 151.2.1 Determination of molecular mass 151.2.1.1 Experimental methods to determineM 16V1.2.1.2 Experimental methods to determineM 171.2.1.3 Experimental methods to determineM 171.2.2 Characterization of polymer structures 191.2.2.1 Nuclear magnetic resonance spectroscopy (NMR) 191.2.2.2 Infra-red and Raman spectroscopy 191.2.2.3 Liquid chromatography (LC) 211.2.2.4 Mass spectrometry (MS) 211.2.2.5 Electron spin resonance spectroscopy (ESR) 221.2.2.6 Thermal analysis (TA) 231.2.2.7 Pyrolysis methods 241.2.2.8 Gas chromatography (GC) 251.3 Thermal deterioration of polymers 251.3.1 Thermo-oxidative decomposition of PE and PP 261.3.2 Thermal deterioration of PVC 291.3.3 Thermal deterioration of PET 321.3.4 Thermal decomposition of ethylene-vinyl acetate copolymers 351.4 Acoustic emission 361.4.1 Introduction 361.4.2 Chemical acoustic emission 361.4.3 Acoustic emission and polymers 381.4.4 Sound wave propagation 391.4.5 Instrumentation for acoustic emission 421.4.5.1 Continuous monitoring of the acoustic emission 421.4.5.2 Capture of individual acoustic emission signals 431.4.5.2.1 Cumulative acoustic spectral analysis 441.4.5.2.2 Waveform analysis 46vi1.5 Thermosonimetry 491.5.1 Instrumentation 501.5.2 Signal analysis 521.5.3 Applications 521.5.3.1 Minerals 531.5.3.2 Glasses, ceramics and refractories 531.5.3.3 Metals and alloys 541.5.3.4 Inorganic materials 551.5.3.5 Organic and polymeric materials 571.6 Scope of this thesis 58CHAPTER 2. DEVELOPMENT OF THERMOGRAVISONIMETRY 592.1 Experimental 602.2 Heat sources 612.2.1 Basic oven apparatus 612.2.2 Computer-controlled furnace apparatus 632.2.3 Incorporation of feedback control routines 682.2.3.1 Feedback control using a difference algorithm 692.2.3.2 Feedback control using a PID controller 702.2.4 Temperature calibration 762.2.5 Optimized furnace apparatus 842.3 Waveguide apparatus 842.4 Balance apparatus 872.5 Data acquisition techniques 892.5.1 Continuous acoustic level monitoring 892.5.2 Individual acoustic signal capture apparatus 89vii2.6 Transducer and waveguide calibration 902.6.1 Calibration procedure 902.6.2 Calibration results 922.7 Development of TGS 962.7.1 Design and construction of transmitter and receiver circuits 962.7.2 Evaluation of circuit 972.7.2.1 Calibration of IR transmitter/receiver system 972.7.2.2 Calibration results for IR transmitter/receiver circuit 982.8 Overall Conclusions 100CHAPTER 3. EXPERIMENTAL STUDIES 1023.1 Introduction 1023.2 Chemicals 1023.2.1 Polymers 1023.2.1.1 Chemical data 1033.2.2 Comparative materials 1053.2.3 Reagents 1053.3 Thermosonimetry apparatus 1063.4 Preliminary studies 1063.4.1 Thermo-oxidative degradation of poly(vinyl chloride) 1063.4.1 .1 Acoustic emission studies using the oven 1063.4.1.2 Initial acoustic emission studies using the furnace 1073.4.1.3 Comparison of acoustic emission from various PVCresins 1073.4.1.4 Comparison of AE and mass loss measurements 1083.4.1.5 Visual studies 109VIII3.4.1.6 Comparison of the acoustic emission produced by thehydrolysis of aluminum chloride and the acousticemission produced during the thermal degradation ofPVCinair. 1103.4.1.7 Acoustic emission from the isothermal degradation ofPVCinair 1113.4.2 Glass transition temperature (Tg) 1123.4.3 The crystalline melting point (Tm) 1133.4.4 Crystallization 1133.4.5 Polymerization 1153.4.6 Plasticization of PVC 1163.5 Polymer studies using thermosonimetry 1173.5.1 General experimental details 1173.5.2 Polymer studies 1183.5.3 Effect of sample mass 1183.5.4 Molecular mass studies 1183.5.5 Effectof heating rate 1193.5.6 Copolymer studies 1193.6 Polymer studies using thermogravimetry 1203.6.1 Effect of molecular mass 1203.6.2 Effect of heating rate 1203.6.3 Effect of copolymer composition 1203.7 Thermogravisonimetry studies 1213.7.1 Polymer studies using thermogravisonimetry 1213.7.2 Other thermogravisonimetry studies 122ix3.8 Other polymer studies 1233.8.1 Reference thermogravimetry studies 1233.8.2 Thermal decomposition studies using residue analysis. 1243.8.3 Thermal stability of PVC using a modified ASTM method 1253.8.4 Other thermal stability studies 1263.8.5 Optical microscopy studies 1263.9 Data analysis methods 1283.9.1 Introduction 1283.9.2 Continuous acoustic level monitoring 1283.9.3 Individual acoustic signal capture 1303.9.3.1 Cumulative acoustic spectral analysis 1303.9.3.2 Waveform analysis 1313.9.4 Software 134CHAPTER 4. RESULTS AND DISCUSSION 1354.1 Introduction 1354.2 Preliminary studies 1354.2.1 Thermo-oxidative degradation of poly(vinyl chloride) 1354.2.1 .1 Acoustic emission studies using the oven 1354.2.1.2 Initial acoustic emission studies using the furnace 1394.2.1.3 Comparison of acoustic emission from various PVCresins 1424.2.1.4 Comparison of AE and mass loss measurements 1494.2.1.5 Visual studies 151x4.2.1.6 Comparison of the acoustic emission produced by thehydrolysis of aluminum chloride and the acousticemission produced during the thermal degradation ofPVC in air 1524.2.1.7 Acoustic emission from the isothermal degradation ofPVC in air 1554.2.1.7.1 Determination of kinetic parameters for theacoustic emission from the isothermaldegradation of PVC in air 1594.2.2 The glass transition temperature (Tg) 1624.2.3 The crystalline melting point (Tm) 1624.2.4 Crystallization 1634.2.4.1 Non-isothermal crystallization 1634.2.4.2 Isothermal crystallization of PP 1704.2.4.2.1 Determination of kinetic parameters for theacoustic emission from the isothermalcrystallization of PP 1734.2.4.3 Comparison of crystallization results using theacoustic probe and the acoustic waveguide. 1784.2.5 Polymerization 1814.2.6 Plasticization of PVC 1824.3 Polymer studies using thermosonimetry 1834.3.2 Polymer studies 1834.3.2.1 Temperature-dependent behavior of PVC 1884.3.2.2 Determination of kinetic parameters from thethermo-oxidative decomposition of PVC and PET 192xi4.3.3 Effect of sample mass 2004.3.4 Molecular mass studies 2104.3.5 Effect ofheating rate 2184.3.5.1 Determination of kinetic parameters using multipleheating rate methods 2244.3.6 Copolymer studies 2294.4 Polymer studies using thermogravimetry (TG) 2364.4.1 Effect of molecular mass 2364.4.2 Effect of heating rate 2384.4.3 Effect of copolymer composition 2414.5 Thermogravisonimetry studies 2434.5.1 Polymer studies using thermogravisonimetry 2434.5.2 Other thermogravisonimetry studies 2474.6 Other polymer studies 2484.6.1 Reference thermogravimetry studies 2484.6.2 Thermal decomposition studies using residue analysis 2524.6.2.1 Physical form of the residues 2534.6.2.2 IR absorption spectra 2544.6.2.3 Elemental analysis 2624.6.2.4 Mass spectrometry 2644.6.2.5 Scanning electron microscopy (SEM) 2704.6.3 Thermal stability of PVC using a modified ASTM method 2814.6.4 Other thermal stability studies 2814.6.5 Optical microscopy studies 2834.6.6 Conclusions 284CHAPTER 5. CONCLUSIONS 285CHAPTER 6. FUTURE WORK 288BIBLIOGRAPHY 291APPENDIX I CIRCUIT SCHEMATICS 313APPENDIX 2 SOFTWARE ROUTINES 317xiiXIIIList of Tables1. Characteristics of step-growth and chain-growth polymerization 22. Chemical structures of some common polymers 73. Reported thermal transition temperatures for common polymers 104. Average weighted sum of the squares of the error as a function of 765. Empirical functions used for the conversion of furnace temperatureto sample temperature 816. Melting points determined for benzoic acid and PET using the furnaceapparatus 837. Regression analysis of the calibration results using the direct connectionmethod and using a waveguide 958. Inherent viscosity and viscosity-average molecular mass of thePVC samples 1039. Particle size distribution of the PVC resins 10410. Resin characterization data for the PE samples 10511. The descriptors used to mathematically characterize theacoustic emission signals 13212. Number of AE signals collected for the three PVC resins during thethermal degradation of PVC in air 14413. Regression analysis of the average power spectra from the thermaldegradation of the four PVC resins in air 14614. Regression analysis of the average power spectra from theacoustic emission produced during the hydrolysis of AICI3 and the thermaldegradation of PVC in air 15315. Summary of acoustic emission results from the isothermal degradationof PVC in air at 483, 468 and 453 K 156xiv16. Kinetic parameters calculated for the acoustic emission data from theisothermal crystallization of PP 17517. The number of AE signals acquired and the total cumulativeacoustic RMS as a function of the polymer 18418. Regression analysis of the average power spectra obtained usingthe four polymers 18619. Regression analysis of the average power spectra from thetwo reactions of the thermo-oxidative decomposition of PVC 19020. Kinetic parameters for acoustic emission from the thermo-oxidativedecomposition of PET using the Coats-Redfern method 19621. Kinetic parameters for acoustic emission from the thermo-oxidativedecomposition of PVC using the Coats-Redfern method 19822. Characteristic temperatures for the thermo-oxidative decompositionof PVC using thermosonimetry as a function of sample mass 20523. Regression analysis of the average power spectra obtained usingvarious masses of PVC 20724. Kinetic parameters for acoustic emission from the thermo-oxidativedecomposition of PVC using the Coats-Redfern method.Effect of sample mass 20925. Regression analysis of the average power spectra obtained from thethermo-oxidative decomposition of the six PVC resins 21326. Kinetic parameters for acoustic emission from the thermo-oxidativedecomposition of PVC using the Coats-Redfern method.Effect of molecular mass 21727. Coefficients of determination obtained by regression analysis of theaverage power spectra using the five different heating rates 221xv28. Kinetic parameters for acoustic emission from the thermo-oxidativedecomposition of PVC using the Coats-Redfern method.Effect of heating rate 22329. Calculated activation energies for the acoustic emission datausing the Ozawa method 22830. Regression analysis of the average power spectra from thetwo reactions of the thermo-oxidative decomposition of PVAc 23631. Kinetic parameters for the dehydrochiorination of PVC in air obtainedfrom TG using the Coats-Redfern method. Effect of molecular mass 23832. Kinetic parameters for the dehydrochiorination of PVC in air obtainedfrom TG using the Coats-Redfern method. Effect of heating rate 24033. Activation energies for the dehydrochlorination of PVC in air obtainedfrom the TG results using the Ozawa method 24134. A comparison of the acoustic emission onset temperatures andthe temperature at which the maximum rate of mass loss was observedfor the thermo-oxidative decomposition of the four polymers 24535. Kinetic parameters for the thermo-oxidative decomposition of PVC,PE, PP and PET measured using TGS 24636. Kinetic parameters for thermal decomposition of PVC in heliummeasured using TG with heating rates of 5, 10 and 20 K/mm 24937. Kinetic parameters for the thermo-oxidative decompositionof PVC and PE measured using TG 25238. Elemental analysis results for the residues produced by heating thepolymers, PET, PP, PE and PVC to various temperatures in air 26239. Thermal stability of the PVC resins as measured by the modifiedASTM method 281xvi40. Relative thermal stability of the six PVC resins, measuredusing a melting point apparatus 282xviiList of FiguresFig. 1. Various stereoregular forms of polymers. (A) Isotactic,(B) Syndiotactic and (C) Atactic 8Fig. 2. A schematic of the steps involved in principal componentsanalysis (Adapted from [185]) 48Fig. 3. The thermosonimetry apparatus designed by Lønvik(Adapted from [148]) 51Fig. 4. A schematic of the thermosonimetry apparatus developed by Clarkand Garlick [149] 59Fig. 5. A schematic of the thermogravisonimetry apparatus 60Fig. 6. A schematic of the oven apparatus used for some of the preliminaryexperiments 62Fig. 7. A typical heating profile for the oven apparatus 63Fig. 8. The furnace equipment used during the thermosonimetry studies 64Fig. 9. Typical heating profiles for the furnace apparatus 65Fig. 10. The voltage-temperature calibration for the thermocoupleamplifier circuit 67Fig. 11. A schematic of the oven feedback control 68Fig. 12. Typical heating profiles as a function of the gain constant Kg 71Fig. 13. Typical heating profiles as a function of the proportional gainconstant (Kr) 74Fig. 14. A comparison of recorded temperatures using the furnace andsample thermocouples 79Fig. 15. A comparison of the furnace, sample and reference temperatures forheating rates of 1, 5 and 10 K/mm 80xviiiFig. 16. The waveguides used during the thermosonimetry experiments 85Fig. 17. The holder used to support the waveguide and transducer during thethermosonimetry experiments 86Fig. 18. A schematic of the apparatus used to calibrate the transducer 91Fig. 19. A reproduction of the calibration supplied for the transducer usedthroughout the experimental work 93Fig. 20. A comparison of the transducer calibrations obtained using a directconnection between the source and the transducer and usinga waveguide between the source and the transducer 94Fig. 21. Transducer calibrations using (A) a direct connection between theamplifier and the transducer, and (B) the infra-redtransmitter/receiver circuit 99Fig. 22. Three-dimensional plots showing the frequency response of(A) using a direct connection between the amplifier and thetransducer, and (B) using the IR transmitter/receiver circuit 101Fig. 23. The apparatus used to visually study the thermaldegradation of PVC 109Fig. 24. The apparatus used to assess the thermal stability of thePVC resins 125Fig. 25. Occurrence of AE signals as a function of time during the thermaldegradation of PVC in air using the oven 136Fig. 26. Typical AE signals from the thermal degradation of PVC in airusing the oven. Signals are shown for each of the three trials,together with their corresponding frequency spectra 138Fig. 27. Occurrence of AE signals as a function of time during the thermaldegradation of PVC in air using the furnace 139xixFig. 28. Average power spectra of the AE signals collected during thethermal degradation of PVC in air using the furnace 141Fig. 29. Occurrence of AE signals as a function of time during the thermaldegradation of PVC in air for three different PVC resins 142Fig. 30. The onset temperature of AE as a function of the viscosity-averagemolecular mass for the four PVC resins 143Fig. 31. Average power spectra of the AE signals collected for the PVCresins #2, #4, and #5 during the thermal degradation of PVC in air 145Fig. 32. Median AE descriptor values as a function of viscosity-averagemolecular mass for the PVC resins 147Fig. 33. Comparison of thermosonimetry (TS) and thermogravimetry (TG)results for the thermal degradation of PVC in air. (A) TS curves.(B) TG curves. (C) TS and derivative thermogravimetric(DTG) curves 150Fig. 34. Average power spectra from (A) the hydrolysis of AlCl3, and(B) the thermal degradation of PVC in air 152Fig. 35. Principal components analysis of the acoustic emission signalsacquired during the hydrolysis of AlCl3 and during thethermal degradation of PVC in air 154Fig. 36. Cumulative acoustic RMS curves as a function of time for theisothermal degradation of PVC in air at temperatures of(A) 483 K, (B) 468 Kand (C) 453K 155Fig. 37. Total number of individual acoustic waveforms collected as afunction of reaction temperature for the isothermal degradationof PVC inair 157xxFig. 38. Examples of the general plots obtained for the median descriptorvalues as a function of reaction temperature 158Fig. 39. A typical cooling profile for the non-isothermal crystallizationexperiments together with the fitted curve using theempirical function given in the text 164Fig. 40. The sample and oil bath temperature as a function of time for oneof the non-isothermal crystallization experiments 165Fig. 41. TS curves for the non-isothermal crystallization of PP 166Fig. 42. AE signals from the non-isothermal crystallization of PP.Typical signals are shown from all three trials together withthe corresponding frequency spectra 167Fig. 43. The occurrence of AE signals as a function of sample temperatureduring the non-isothermal crystallization of benzoic acid 169Fig. 44. Occurrence of acoustic emission as a function of time for PPcrystallized at temperatures of 373, 383, 393 and 403 K 170Fig. 45. Temperature-time calibration plots for the isothermal crystallizationofPP 172Fig. 46. Determination of the kinetic parameters for the isothermalcrystallization of PP, using the method described by Pracella et al. 174Fig. 47. Comparison of the observed and calculated data for the isothermalcrystallization of PP. Crystallization temperatures are nominally(A) 373, (B) 383, (C) 393 and (D) 403 K 176Fig. 48. Determination of the relationship between crystallization rate andcrystallization temperature for the isothermal crystallization of PP 177xxiFig. 49. Comparison of the average power spectra obtained from thenon-isothermal crystallization of benzoic acid using (A) the acoustic probeand (B) a waveguide 179Fig. 50. Principal components analysis of the acoustic emission signalsacquired during the non-isothermal crystallization of benzoic acid usingeither the waveguide or the acoustic probe 181Fig. 51. TS curves for the thermo-oxidative decomposition of the polymers,PVC, PE, PP and PET 183Fig. 52. The average power spectra for the thermo-oxidative decompositionof PVC, PE, PP and PET in air 185Fig. 53. Median descriptor values for the four polymers used 187Fig. 54. Average power spectra from the two regions of AE activity duringthe thermo-oxidative decomposition of PVC 189Fig. 55. Principal components analysis of the acoustic emission signalsacquired during the two reactions found from the thermo-oxidativedecomposition of PVC 191Fig. 56. Comparison of experimental and fitted curves for thethermo-oxidative decomposition of PET 195Fig. 57. Effect of sample mass on the TS curves for the thermo-oxidativedecomposition of PVC in air. Sample masses were 200, 100, 75,50, 25and 10mg 200Fig. 58. The total cumulative acoustic RMS as a function of the samplemass for the thermal oxidative decomposition of PVC in air 201Fig. 59. The number of acoustic emission signals collected as afunction of the PVC sample mass 203xxiiFig. 60. Median ‘RMS’ descriptor values as a function of the sample massfor the thermo-oxidative decomposition of PVC in air 206Fig. 61. The number of acoustic emission signals as a function of theviscosity-average molecular mass for the six PVC resins 211Fig. 62. The onset temperature of acoustic emission as a function ofthe molecular mass of the PVC resin for the thermo-oxidativedecomposition of PVC 212Fig. 63. The average power spectra of the acoustic emission from thesix PVC resins used 214Fig. 64. Median AE descriptor values as a function of theviscosity-average molecular mass for the six PVC resins 215Fig. 65. Effect of heating rate on the TS curves for the thermo-oxidativedecomposition of PVC. Heating rates used were 10, 7, 5, 3and I K/mm 219Fig. 66. The number of acoustic emission signals collected as a functionof the heating rate for the thermo-oxidative decomposition of PVC 220Fig. 67. Application of Kissinger’s method to the analysis of acousticemission data from the thermo-oxidative decomposition of PVC 226Fig. 68. Application of Ozawa’s method to the analysis of acousticemission data from the thermo-oxidative decomposition of PVC 229Fig. 69. Effect of vinyl acetate content on the TS curves for thethermo-oxidative decomposition of EVA copolymers 230Fig. 70. The amount of acoustic emission collected as a function of the vinylacetate content for the thermo-oxidative decompositionof EVA copolymers 232xxiiiFig. 71. Average power spectra from the two regions of AE activityduring the thermo-oxidative decomposition of PVAc 235Fig. 72. Effect of molecular mass on the TG curve for thethermo-oxidative decomposition of PVC 237Fig. 73. Effect of heating rate on the TG curve for the thermo-oxidativedecomposition of PVC 239Fig. 74. Effect of copolymer composition on the TG curve for thethermo-oxidative decomposition of EVA 242Fig. 75. Thermogravisonimetry results for the thermo-oxidativedecomposition of PVC, PE, PP and PET 244Fig. 76. TGS curves for the dehydration ofCuSO4.5H20usinga heating rate of 5 K/mm 247Fig. 77. TG curves for the thermal decomposition of PVC in heliumusing heating rates of 5, 10 and 20 K/mm 248Fig. 78. TGcurves for the thermo-oxidative decomposition of PVCand PE in air using a heating rate of 10 K/mm 251Fig. 79. IR absorption spectra of PE.(A) PE. (B) PE heated to 672 K 258Fig. 80. IR absorption spectra of PVC.(A) PVC. (B) PVC heated to 559 K 259Fig. 81. IR absorption spectra of PP.(A) PP. (B) PP heated to 559 K 260Fig. 82. lR absorption spectra of PET 261Fig. 83. The dependence of the atomic ratio H/C on the decompositiontemperature of PVC and PET 263xxivFig. 84. Mass spectra obtained from the thermal desorptionmass spectrometry of the PVC resin and the PVC residues 265Fig. 85. SEM photographs of PP before and after thermo-oxidativedecomposition 271Fig. 86. SEM photographs of PE before and after thermo-oxidativedecomposition 272Fig. 87. SEM photographs of PVC before and after thermo-oxidativedecomposition 276Fig. 88. SEM photographs of PET before and after thermo-oxidativedecomposition. 278APPENDIXFig. A. Schematic of the computer-controlled zero voltageswitching circuit 313Fig. B. Schematic of the thermocouple amplifier circuit 314Fig. C. Schematic of circuit used to amplify the electrical signal from the‘True RMS’ meter prior to acquisition by the computer 314Fig. D. Schematic of the circuit used to provide a regulated power supplyfor the circuits shown in Figure B and Figure C 315Fig. E. Schematic of the IR receiver circuit 315Fig. F. Schematic of the IR transmitter circuit 316xxvGlossaryA pre-exponential factora.c. alternating currentAE acoustic emissionAIBN AzobisisobutyronitrileASTM American Society for Testing and MaterialsAvg. average value (mean)dB decibelsDOP di-n-octyl phthalateDSC differential scanning calorimetryDTA differential thermal analysisEa activation energyESR electron spin resonanceEVA ethylene-vinyl acetate copolymerFFT fast Fourier transformFTIR Fourier transform infra-redGC gas chromatographyHDPE high density polyethyleneIR infra-redk rate constantxxviIc crystallization rate constantkHz kilohertzconstant for Huggins equationLC liquid chromatographyLDPE low density polyethyleneMHz megahertznumber-average molecular massMS mass spectrometryM viscosity-average molecular massmV millivoltsweight-average molecular massmlz mass to charge ratioN number of data pointsn degree of polymerization (or number of moles in stoichiometry)Av Avrami exponentnumber of pairs used for regression analysisreaction orderOCV oven control valuePID proportional integral derivativePCA principal components analysisPET poly(ethylene terephthalate)xxviiPE polyethylenePP polypropylenePVC poly(vinyl chloride)PVAc poly(vinyl acetate)Py pyrolysiscoefficient of determinationRMS root mean squareSEC size exclusion chromatographySEM scanning electron microscopyt timesampling timeTA thermal analysisT crystallization temperatureTG thermogravimetryTg glass transition pointTGS thermogravisonimetryTj temperature at which the thermal event commencedTf temperature at which the thermal event ceasedTm crystalline melting temperaturetmed time of median signalT temperature at which the maximum rate is foundxxviiiTreq required temperatureTS thermosonimetryz Huggins exponenta. degree of conversionheating rate11 viscosity[i] limiting viscosity numberinherent viscosityxxixAcknowledgmentsI wish to thank all the past and present members of my research group.Although Steve, Peter, Eric, Julie, Dave, Terrance, Pat, Bru, Timmy, Paul, Ade,Megan, Larry and Thanh have all gone on to new pastures, their support and advicewas much appreciated, in most cases. Both Ivan (“Want to see my chick magnet”)Brock and (“Dr. ?“) Oliver Lee had to “suffer” my company throughout my degreeand their patience should be admired. Mike (“Nice hair”) Kester was luckier but stillhad to suffer a little. ‘Good Luck’ to the graduate students I have met whilst at UBC.Dr. Don (“I still hate St. Louis”) Yapp deserves a special mention for his bitchinessand Mark (“Liverpool ?“) Aston should be mentioned if for nothing else, then for hisperformance at my wedding.Big thanks go to Dr. Adrian Wade who provided a constant source of helpand support. His encouragement was especially welcome during those dark monthswhen the end of the tunnel was not visible. I wish to acknowledge financial supportfrom the following agencies; the Institute for Chemical Science and Technology, theCanadian National Network of Centres of Excellence, NSERC, and UBC.I wish to thank my guidance committee, especially Dr. R.C. Thompson whogave me many useful suggestions. Dr. Y. Koga should be thanked for hisenthusiasm and for reading an early draft of one chapter. I also wish to thankDr. Geoffrey Herring for his help and encouragement.Frequently I required the help of the Dept. of Chemistry staff and I wish tothank everyone there for their technical help. A special acknowledgment goes toBrian Snapkauskas for his efforts. My gratitude goes also to Peter Borda though myleaving will probably be thanks enough. Thanks also go to Spiros Pergantis fordoing the MS analysis, Dr. Peter Wassell for help with the IRs and to Mary Magerfor her help with the SEM photographs.On a more personal level, the full support of my family, both here inVancouver and back in England, cannot be acknowledged enough. I hope the endresult is worth the wait. I also want to express my gratitiude to my father whounfortunately did not live to see this day.Finally the most important person. Without Helen’s support, you would notbe reading this. Countless times her encouragement was all that kept me going.Words cannot begin to express my gratitude. To Helen, this dissertation isdedicated.ICHAPTER 1. INTRODUCTIONThe introduction presents a survey of the areas covered by the experimentalstudies. It is split into two sections, polymer science and the acoustic emissiontechnique. For the polymer science section, a review is presented of the variety ofareas of importance including polymerization processes, polymer chemistry andpolymer properties. In addition, a brief survey is given of the analytical methodscurrently used in polymer science and a few typical applications.As the technique of acoustic emission is relatively new and little understood,a concise background is given for the phenomenon. Although the technique hasnot been greatly applied to polymer chemistry, a few important applications havebeen reported and these are discussed in detail. For acoustic emission studies, theexperimental apparatus and analysis methods can have a great effect on the natureof the results obtained and thus a brief discussion of the types of apparatus anddata analysis methods is given.The later work in this dissertation concentrated on the use ofthermosonimetry to study polymer decomposition. A review is given of thistechnique, with a particular emphasis on the information available fromthermosonimetry and the areas of science where the technique has been applied.1.1 Polymer scienceThe field of polymer science is vast, encompassing many important areas.For reasons of brevity only a short introduction is provided to these areas.ttl Polymerization processesThe two major polymerization processes are step-growth and chain-growth.These terms were introduced by Flory in 1953 [1] and are based on the reactionmechanisms of the polymerizations. They correspond approximately to the olderterms of condensation and addition reactions, though there are exceptions. Themain characteristics of the step-growth and chain-growth processes are shown inTable 1.Step-growth polymerization Chain-growth polymerizationAny two molecular species can react. Only the growing chain mayreact.Monomer is used up early in the reaction. Monomer concentrationgradually decreases duringreaction.Polymer molecular mass increases High molecular mass polymer isgradually throughout the reaction. formed at once. There is littlechange in molecular massduring reaction.Long reaction times are needed to obtain Long reaction times give higherhigh molecular mass polymers. yield.At any instant, there is a smooth, At any instant, there is only highcalculable distribution of molecular mass. molecular mass polymer,unreacted monomer, and smallamount of growing chains.2Table 1. Characteristics of step-growth and chain-growth polymerization.31.1.1.1 Step-growth polymerizationThe formation of Nylons are examples of step-growth polymerizaflon.Nylon-6, 10 is formed from the Schotten-Baumann reaction between decanedioicchloride and I ,2-diaminohexane [2], e.g.,0(n) Cl—b—(CH28—C l + (n)H2N—(CH6—N[inccl4] [inH2O]0II IIC—(CH2)8—C NH—(CH6-- H + (2n-1)HCIn1.1.1.2 Chain-growth polymerizationUnsaturated monomers such as alkenes undergo chain-growthpolymerization. Three different stages of the polymerization process arerecognized; initiation, propagation and termination.Initiation leads to the chemical generation of reactive centres on selectedmonomer molecules. Usually the reactive centres are free radicals generated by asmall amount of initiator such as azobisisobutyronitrile (AIBN) or benzoyl peroxide.However anionic initiators (e.g., KNH2 n-butyl lithium) and cationic initiators(e.g., Lewis acids, HCI) may also be used, to produce ionic reactive centres.The initiator AIBN undergoes photolysis when irradiated with UV light of360 nm, e.g.,CH3 CH3 CH3C—C—N=N—C—CH 1w 2 CH3—c—CN + N2CN CN4Once the free radical (R.) is produced, it reacts rapidly with a molecule of themonomer (e.g., a vinyl halide), to produce the reactive centre, e.g.,HR. + CH2’C X RCH2—CxPropagation occurs through a series of reactions in which the reactive centreof the growing molecule reacts with monomer to still further increase the length ofthe chain, e.g.,r 1 H r 1RtCH2--CHXj-CH_ .+ CH2HX RtCH2_CHXI-CH.fl ) n+1 XThe two most common termination reactions are combination anddisproportionation, e.g.,I HH+ ...—CH2— ...—CH2—C C—CH...x x xxH H H H+ ...—CH2— ...—CH2—C— + ...—CH=Cx x xChain transfer reactions may occur to terminate the growing chain,e.g., reaction with solvent.H I.;1—CH2---C + CCI4 ...—CH2C-Cl + CCI3x x5Another distinct type of chain-growth polymerization is via co-ordination.This field came into existence with the pioneering work of Ziegler [3] and Natta [4]who received the Nobel Prize for Chemistry in 1963 for their efforts. TheZiegler-Natta catalysts are complexes of alkyl compounds of metals of group I, II orIll, combined with the halide or other derivatives of a transition metal(e.g., (CH3C2)Aland TiCl4). These complexes contain a vacant co-ordinationsite on the transition metal to which a monomer may bond. Reaction occursbetween two bonded monomers and because of the spatial requirements of thecomplex, the polymerization proceeds along highly stereospecific lines. Thisproduces a polymer in which all of the monomer units have identicalstereochemistry.1.1 2 CopolymerizationChain-growth polymerization between two or more different monomers leadsto the formation of a copolymer. Strictly speaking, many step-growthpolymerizations are also ‘copolymerizations’ as they involve reaction between twomonomers (e.g., Nylon-6,1O). Different arrangements of the monomers in theproduct may arise. Random copolymers have an irregular propagation of the twomonomers, while block copolymers have a long sequences of one monomer joinedto long sequences of the other monomer. This arrangement is important as blockcopolymers tend to possess the properties of the two homopolymers whereasrandom copolymers are a compromise with respect to the properties of the twohomopolymers.61.1.3 Chemical properties of polymersTable 2 indicates the chemical structure of the polymers used during thecourse of these studies.Commercial polymers are very chemically inert. For example, polyethylenehas good resistance to attack by acids and bases. Polypropylene does not dissolvein any common solvents at room temperature but at high temperature aromatic andchlorinated hydrocarbons are solvents [5]. Many polymers, in their raw state, aresusceptible to photochemical and oxidative degradation. As a result mostcommercial polymers contain stabilizers to inhibit such degradation.Polymers may be classified as either thermoplastics or thermosets withrespect to their response to thermal treatment. Thermoplastics melt when heatedand resolidify when cooled, while thermosets do not melt when heated and willdecompose if heated to a sufficiently high temperature. Generally, thermosets arecross-linked polymers while thermoplastics are linear polymer molecules. However,all polymers will decompose if heated to a sufficiently high temperature.The degree of polymerization n varies widely. For example, commercial PEpolymers have values between 400 and 50000 [6], while for PVC, n lies between500 and 1500 [7].1.1.4 Polymer microstructureThe properties of a polymer depend not only on the chemical composition ofthe polymer but also on the subtle differences in the microstructure of the polymer.Polymers may be linear or branched or may form a network leading to widelydifferent properties. Orientational differences in microstructure may arise as twomodes of addition are possible for chain-growth polymerization of asymmetric vinyl7Polymer Repeat unitPolyethylene(PE) —OH2]--...nPolypropylene r ?H3l(PP) ...fcH2—cH—1...LPoIy(vinyl chloride)(PVC) ... [CH2CHC...nPoly(ethylene terephthalate)(PET)Ethylene-vinyl acetate(EVA) ... -FCH2—]-CH—CH20c=0CH3nnTable 2. Chemical structures of some common polymers.monomers. This leads to two different configurations, e.g., head-to-tail(i.e., HTHTHTHT); head-to-head, tail-to-tail (i.e., THHTTHHTTH). Generally, thehead-to-tail configuration is found.Many polymers contain tertiary carbon atoms in the chain that can beeffectively regarded as chiral centres. Isotactic polymers have all asymmetriccarbon atoms in an identical configuration. Syndiotactic polymers have a regularalternating arrangement of asymmetric carbon atoms. Atactic polymers have no8regularity in the arrangement of configuration of asymmetric carbon atoms presentin the polymer chain.H H HH: H H: H(A)kH kH kH kx x x xH H HH: x: H: x(B)kH kH H bX H X HH H HH: H x: H(C)H kH HX X H XFigure 1. Various stereoregular forms of polymers.(A) Isotactic, (B) Syndiotactic and (C) Atactic.Geometric isomerism occurs when conjugated dienes are polymerized, e.g.,cis- and trans- I ,4-polybutadiene.1.1.5 Physical states of polymersThere are three extreme physical states found in polymers; the polymer melt,the polymer glass and the polymer crystal. In the polymer melt, there is completelyfree rotation around the single bonds in the molecules. The molecules are incontinuous motion and can slip past other molecules easily. In the polymer glass,rotation around the single bonds ceases due to intermolecular repulsions betweenthe chains. The molecules become trapped in a chaotic, disordered, entangledstate. In polymer crystals, the molecules fit together in a regular lattice which isstabilized by intermolecular attractions. Rotation around the single bonds is stillpossible.91.1.6 Thermal transitionsThe two major thermal transitions associated with polymers are the glasstransition point (Tg) and the crystalline melting temperature (Tm).The glass transition point is the temperature below which the free rotationaround the single bonds in the molecule ceases. Above this temperature, thepolymer ceases to be brittle and glassy, becoming less rigid and more rubbery. Theglass transition temperature is influenced by many structural factors including chainflexibility, polarity, symmetry, molecular mass, and tacticity. Any structural featurewhich provides less restriction to the free rotation of the single bonds in themolecule lowers its Tg.The crystalline melting temperature is the highest temperature at which apolymer crystal lattice is stable. Crystalline polymers do not have a well-definedmelting temperature because they are best regarded as mixtures, comprised ofcrystallites of a range of relative molecular masses. The temperature over whichmelting occurs is also an indication of polymer crystallinity. As the molecular massor crystal size of a component decreases, the crystalline melting temperature getslower. The crystallite size and crystal perfection are determined by the rate ofcrystallization. Table 3 indicates reported transition temperatures for polymers usedin this work [8]. For the crystalline melting point, the value usually quoted is thetemperature of melting of the highest melting crystallites.10Polymer Tg (K) T (K)Polyethylene (linear) 193 410Polyethylene (branched) 193 378 - 398Polypropylene 254 449Poly(vinyl chloride) 356 493 - 513Poly(ethylene terephthalate) 342 540Table 3. Reported thermal transition temperatures for common polymers.1.1.7 CrystallizationThe crystallization of polymers from the melt is a complex process thatresults in the ordering of the long randomly coiled polymer chains in the melt intothree-dimensional chain-folded arrangements in the crystalline state. The formationof the crystalline material from the polymer melt is a two stage process. In the firststage, the polymer chains, stimulated by intermolecular interactions, becomeordered in a parallel array thus creating the stable nucleus. Stabilization of longrange order by secondary valence forces then occurs leading to the packing of themolecules into a three-dimensional ordered structure. The first stage is known asspontaneous nucleation. In the second stage, growth of the crystalline regionoccurs. This growth is counteracted by the thermal redispersion of the chains at thecrystal-melt interface. The final size of the crystalline region is determined by thecompetition between these two processes. At temperatures above Tm, the thermalmotion is sufficient to halt the growth of the crystalline region and at temperaturesbelow Tg the polymer chains are too entangled for ordering to occur. As a result,crystallization can only occur between Tm and Tg.11Turnbull and Fisher described the growth rate of polymer spherulites,G (length/time), as a function of temperature according to the equation [9],G = Go.expAEdJ.exp:M)(Equation I)where Go is a pre-exponential factor, AEd, the activation free energy oftransport of a crystallizing segment across the melt-crystal interface, A4, theactivation free energy required to form a nucleus of critical size, T, thecrystallization temperature (K) and k, Boltzmann’s constant (J/K).In reality, measurable crystallization rates occur between (Tm - 10 K) and(Tg + 30 K) with a maximum rate being found at a temperature which is dependenton the molecular mass of the sample. At temperatures near Tm, the rate ofcrystallization is nucleation-controlled and at temperatures near Tg, the rate ofcrystallization becomes diffusion-controlled.Kinetic approaches to determine the overall crystallization rate are based onthe Avrami equation [10], i.e.,Wiexp(-k t (flAv) (Equation 2)where lc is the rate constant, Wo and W1 are the masses of the melt at zerotime and that left at time t. The exponent Av is the Avrami exponent, which can beindicative of the geometric form of the growth. For example, if sporadic nucleationis assumed to be a first-order process and the growth unit is a disc(i.e., two-dimensional) then Av is 3. Experimentally, the rate of crystallization ismeasured by recording the density changes which occur using the technique ofdilatometry [11]. Deviations from the Avrami equation towards the end of thecrystallization process are attributed to secondary crystallization. This involves the12development of more perfectly formed crystallites from the crystallinestructures [12].1.1.7.1 CrystallinityThe crystallinity in polymers is never perfect; amorphous regions are alwayspresent. Linear symmetrical polymeric molecules generally produce highlycrystalline polymers as a regular three-dimensional close-packing arrangement ofthe chains is possible. Branching of the polymer decreases crystallinity.Intermolecular bonding helps to promote crystallinity. The crystallinity of a polymervaries greatly with tacticity. lsotactic polymers tend to form helices to provide themost favorable steric interactions and syndiotactic polymers are also sufficientlyregular to crystallize. Atactic polymers are more amorphous [13].1.1.7.2 MorphologyA variety of distinct morphological features have been identified during thecrystallization of polymers from the melt. Generally single crystals are onlyobserved if the polymer is crystallized from dilute solution.Crystallites are formed in the melt by diffusion of molecules into close-packedarrays which then crystallize. Different portions of one chain may be associatedwith different crystallites as the size of a crystallite rarely exceeds 100 nm. Thisleads to a strain on the polymer with retardation of crystallite formation and theintroduction of imperfections in the crystallites. Eventually, as a result of the strainsimposed by surrounding crystallites, further growth ceases. As a result, the polymercontains ordered regions with non-discrete disordered interfacial areas. Currentexplanations for the structure of the crystallites involve the concept of chain folding,where the polymer chain is folded many times in the ordered regions. For example,in PE the molecules can fold so that only 5 chain carbon atoms are involved in the13fold itself [13]. The crystallites are not arranged randomly, rather they form regularbirefringent structures of circular symmetry known as spherulites. The spherulitegrows radially from a nucleus, which may be a foreign particle or may arise due to adensity fluctuation in the system. Growth occurs due to the formation of small fibres(fibrils) which spread out from the nucleus into the amorphous material. Themolecular chains lie perpendicular to the fibril axis. Branching and twisting alsooccurs, producing bundles of diverging and spreading fibrils and leading to theobserved spherical shape of the spherulite [14]. As the spherulite grows, eventuallythe growing front will interfere with other spherulites, leading to the formation of anirregular matrix. Typically the size of the spherulites varies from greater than thesize of the crystallites to millimetres in diameter, depending greatly on thetemperature of crystallization.1.1.8 Amorphous polymersFive distinct states exist for a linear amorphous polymer. These states havebeen identified on the basis of mechanical behavior and are related to the molecularmotion in the polymer. As the molecular motion of the amorphous polymer isdetermined implicitly by the chain flexibility of the polymer and the temperature ofthe system, these states are a function of the temperature of the system.In the glassy state, the temperature of the polymer system is below Tg andmolecular motion is frozen. As the temperature of the system is raised through Tg,molecular motion increases. In the region just above Tg, typically T to Tg + 30 K,molecular motion is still slow and the material possesses leathery properties. Asthe temperature is increased, the amorphous polymer passes through a rubberystate to a rubbery flow state. If the temperature is increased sufficiently the viscous14flow state is attained. In this state, molecular motion is increasing and the polymerphysically resembles a viscous liquid.1.1.8.1 PlasticizationIf two polymers of different Tg are mixed then the expected Tg will be aweighted average of the two Tg values. In this way, the expected values of Tg canbe determined for any copolymer composition, provided that the individual valuesare known. This is known as internal plasticization. More commonly, the polymer ismixed with a non-volatile solvent (plasticizer) and the same principle is applied.This is known as external plasticization as the solvent and polymer are notchemically bonded to each other.Poly(vinyl chloride) is compatible with many plasticizers such as di-n-octylphthalate. By the addition of such plasticizers, the glass transition temperature maybe reduced from 356 K (83 °C) to 270 K (-3 °C) [15]. In this way, addition of theplasticizer converts the PVC from a rigid solid at room temperature into a rubberymaterial with more uses. Water acts as a plasticizer in fibres, such as Nylon-6,6, todepress Tg below room temperature. Thus during the steam ironing of clothes, Tg islowered and a lower ironing temperature may be used.Physically, the plasticizer acts as a lubricant where the small non-volatilesolvent molecules space the polymer chains further apart and thus molecularmotion is possible at lower temperatures as there is less restriction to free rotationabout the single bonds.151.2 Analytical methods used in polymer scienceBecause polymeric materials are so complex, it is no surprise that the scopeof analytical methods used for polymer analysis is vast. Numerous books havebeen written on the subject. Publications by Bark and Allen [5], and Campbell andWhite [16] can be recommended. For more recent information, the journal“Analytical Chemistry” produces an analytical review issue every two years whichcontains an article on the analysis of synthetic polymers, e.g., [17-19]. In the mostrecent review [19], Ca. 900 references were cited covering articles published from1990 to 1992. Approximately one-third of the articles reported applications ofnuclear magnetic resonance (NMR) spectroscopy. Other techniques listed, in orderof popularity, were infra-red and raman spectroscopy, liquid chromatography, massspectrometry, electron spin resonance spectroscopy, thermal analysis, pyrolysistechniques and gas chromatography.Cowie [20] divided the analysis of polymers into two classes; the first wasconcerned with the determination of molecular masses and the second with thecharacterization of chain dimensions and structures in the polymer. Analyticalmethods are also used to probe polymer crystallization, polymer degradation andother polymer processes.1.2.1 Determination of molecular massThe determination of the molecular mass of a polymer is of fundamentalimportance as many polymer properties (e.g., Tm) are influenced by the molecularmass. As both step-growth and chain-growth polymerization produce a polymerwith molecules having different chain lengths (and thus molecular mass),experimental measurement of the molecular mass of the polymer yields an averagevalue only. Several different measurements of average molecular mass have16developed, largely as a result of the variety of physical properties used to determinethem. The most common are the number-average molecular mass (Ms), theweight-average molecular mass (Mw) and the viscosity-average molecularmass (My);1/V(3) M=11(4) M= (5)1=1 1=1 i=1where M is the molecular mass of species 1, A’ is the number of moles ofspecies I and v is a constant for the given polymer and solvent system. Typically vranges from 0.5 to I and is either found from tabulated values or by using suitablecalibration standards. In general, M andM are greater thanM, withM generally10-20% less thanM.Although polymer properties are often directly related to molecular mass, thepolymer molecular mass distribution (IvilviD) is also important. The heterogeneityindex (polydispersity) (M IM) may be used as measure of the molecular massdistribution. Typically, values of 1.5 to 2.0 are common, though values of 20 to 50are sometimes found.1.2.1.1 Experimental methods to determine MThe methods used to determine the number-average molecular mass havebeen classified as end-group assay, thermodynamic methods and transportmethods.In end-group assay, an analytical method is used to determine the number ofend-groups present in a known mass of polymer. Thermodynamic methods depend17on the colligative properties of dilute solutions. They include lowering of vaporpressure, ebulliometry, cryoscopy and osmotic pressure measurements. The maintransport technique is based on the lowering of the vapor pressure and is known asvapor pressure osmometry [21].1.2.1.2 Experimental methods to determineExperimental methods used to determineM are often based on lightscattering. The scattering of light occurs when a light beam strikes matter. Theincident beam induces vibration in the nuclei and excitation of electrons. When theexcited nuclei and electrons return to lower energy states, they re-emit light thoughunlike the incident beam, the emitted light is propagated in all directions. The valueofM may be determined from the relative intensity of the scattered light [22].The weight-average molecular mass may also be measured using methodsbased on ultracentrifugation [23].1.2.1.3 Experimental methods to determineMCapillary viscometry is used to determineM by measuring the time taken fora series of concentrations of the polymer solution to pass between two set points inthe bulb of a viscometer. These efflux times can then be related to the viscosity.For a polymer solution of viscosity i and a pure solvent of viscosity q, the relativeviscosity rir is equal to the ratio of the two viscosities (iiIii) at low concentrations ofthe solute provided that the densities of the solution and solvent are equal. Thisleads to the Huggins equation [24], i.e.,= [i-i] + K’2 c (Equation 6)18where flsp is the specific viscosity (i.e., ir -1), c is the concentration (gIdl),[11] is the limiting viscosity number and K’ is a shape dependent factor known as theHuggins constant.For solutions of linear polymers, viscosity measurements may then beempirically related to the molecular mass (My). For a given polymer and solventsystem at a specified temperature and pressure, the Mark-Houwink equation [25]may be used, i.e.,[ii] = KvMvZ (Equation 7)where K and z are constants which may be established by calibration withpolymer fractions of a known molecular mass. Alternately, tabulated values of theconstants for a given polymer and solvent system may be used [25]. For arandomly coiled polymer in an ideal solvent both K and z take a value of 0.5.The American Society for Testing and Materials (ASTM) has published astandard test method for the dilute solution viscosity of vinyl chloride polymers [26].In this method, the inherent viscosity of vinyl chloride polymers in cyclohexanone isdetermined. The inherent viscosity (1jpJ) is equal to the natural logarithm of therelative viscosity divided by the solute concentration (0.2 g/dl). For the purposes ofdetermining M, the relative viscosity can then be related to the limiting viscositynumber, i.e.,1 (Tir__1 3 (Tir[ii = + In (Equation 8)191.2.2 Characterization of polymer structures1.2.2.1 Nuclear magnetic resonance spectroscopy (NMR)Ebdon reviewed the use of NMR spectroscopy for polymer analysis [27]. Thetechnique is used for the identification of polymers, the determination of copolymercomposition and microstructural analysis.In 1960, Bovey and Tiers [28] first reported the use of NMR spectroscopy toobtain configurational information from polymers. Using poly(methyl methacrylate),they demonstrated that the observed splitting of the resonance of the x-methylprotons could be attributed to different configurations. Now NMR spectroscopy isroutinely used for the determination of polymer configurations. The identification oflinkage isomers or geometric isomers using NMR spectroscopy has also beendescribed [29].Due to the non-invasive and non-destructive nature of NMR spectroscopy,the technique is suitable to study polymer reactions in situ. Lukas et al. [30]discussed the use of 13C NMR spectroscopy to monitor the chlorination of PVC.NMR spectroscopy has also been used to study the thermal decomposition ofpolymers. Suebsaeng et a!. [31] discussed the use of 13C NMR and 31 NMR tocharacterize solid products from the thermal decomposition of PET in the presenceof flame retardants such as red phosphorus.1.2.2.2 Infra-red and Raman spectroscopylnfra-red (IR) and Raman spectroscopic methods rank among the mostimportant techniques for polymer identification. The use of IR spectroscopy forpolymer identification far outweighs that of the Raman method. However, Ramanspectroscopy can be used to distinguish certain polymers (e.g., polyamides) thatare indistinguishable from their IR spectra.20IR and Raman spectroscopy provide extensive information on polymermicrostructure. Chain conformations can be determined. For example, the twoconformational isomers of poly(ethylene terephthalate) can be distinguished fromtheir IR spectra due to the presence of bands specific to one or other isomer [32].Identification of configurational isomers using IR and Raman methods has alsobeen reported. For example, the three different tactic forms of polypropylene arereadily distinguished from their JR spectra [33]. Here, the differences in the lRspectra of the polypropylenes may also be attributed to differences in thecrystallinity of the three forms.In addition to its use in fundamental polymer analysis, JR spectroscopy hasbeen extensively used in many areas of applied polymer science. For example,quantitative analysis by JR spectroscopy can enable the determination of copolymeror blend composition. Chemical reactions (e.g., oxidation or degradation) can befollowed by monitoring the appearance or disappearance of a specific functionalgroup. Sevecek and Stuzka [34] used lR spectroscopy to analyze solid residuesfrom the thermal degradation of a number of common polymers including PVC,polystyrene and polyamide at a variety of temperatures.Fourier transform-IR spectroscopy (FTIR) is often used, in combination withthermal analysis methods such as thermogravimetry (TG), to detect the gaseousproducts evolved during the thermal degradation of polymers. Wieboldt et a!. [35]identified hydrogen chloride, benzene, an unknown long-chain aliphatic ester,carbon monoxide and carbon dioxide as products of the thermal degradation of PVCusing TG-FTIR. Raman spectroscopy has also been used in polymer degradationstudies. Bowden et a!. [36] discussed the use of a Raman method to investigateboth the chemical and thermal degradation of PVC.211.2.2.3 Liquid chromatography (LC)The majority of recent publications involving the use of LC for the analysisand characterization of polymers have centred on the use of size exclusionchromatography (SEC) or gel permeation chromatography (GPC) to determinemolecular mass and molecular mass distributions [37].1.2.2.4 Mass spectrometry (MS)The popularity of MS methods for polymer analysis arises mainly due to thesmall sample masses required for analysis and the flexibility of sample type (solid orliquid). Kyranos and Vouros [38] reviewed a number of the mass spectrometricmethods used for polymer analysis.Mass spectrometry is frequently used to study the thermal degradation ofpolymers. Combined pyrolysis and GC-MS systems (Py-GC-MS) are very popular.Adams [39] used a pyrolysis-mass spectrometry (Py-MS) method with negativeionization to study the degradation mechanisms of terephthalate esters. Risby andco-workers discussed the use of linear programmed thermal degradation massspectrometry (LPTD-MS) for studying the thermal degradation of polystyrene andPVC [40]. The LPTD-MS method enabled a series of scans to be obtained duringthe programmed heating of the polymer. Using this instrument, an activation energyof 138 kJ/mol for the dehydrochlorination of PVC was obtained. Systems combiningthermal analysis (TA) and mass spectrometry for polymer analysis have also beenreported [41,. 42].MS methods may be used to determine polymer molecular masses andmolecular mass distributions. Vincenti and co-workers [43] used desorptionchemical ionization-MS to determine the molecular mass of polystyrene and22poly(ethylene glycols) polymers. Results obtained were consistent with thosedetermined using a SEC method.The use of secondary ion mass spectrometry (SIMS) for polymercharacterization has grown. Typically, SIMS is used to study polymer-polymerinterfaces but other applications, consistent with the use of SIMS as a surface probehave been reported. Feld and co-workers [44] discussed a combination oftime-of-flight SIMS and plasma desorption MS, which was successfully used todetect contaminants on the surface of polymers.1.2.2.5 Electron spin resonance spectroscopy (ESR)Hill and co-workers gave an extensive review of the application of ESRspectroscopy to polymer characterization [45]. The main interest in the use of ESRspectroscopy for polymer characterization is in the study of free-radical initiatedpolymerization reactions where ESR can be used to provide detailed information onthe kinetics and mechanisms of such processes. Yamada, Kageoka and Ostu [46]used an ESR technique to obtain rate constants and apparent activation energiesfor the propagation and termination processes occurring during the free-radicalinitiated polymerization of styrene.ESR spectroscopy has been used to study polymer degradation. Liebmanand co-workers [47, 48] discussed the use of ESR spectroscopy, in conjunction withother techniques of organic analysis to study the thermal decomposition of PVC andchlorinated PVC. The rate of free-radical generation was found to be more rapid inair and resulted in increased bond scission [48].The mechanical stressing of polymers has also been studied usingESR spectroscopy. Betteridge et al. [49] reported that crack formation in polymersduring stressing led to the production of free-radicals.231.2.2.6 Thermal analysis (TA)Thermal analysis is defined as “a group of techniques in which aphysicalproperty ofa substance and/or its reactionproducts is measured as afunction oftemperature whilst the substance is subjected to a controlled temperature program” [50].The most important thermal analysis techniques for polymer analysis are differentialthermal analysis (DTA), differential scanning calorimetry (DSC) andthermogravimetry. The techniques of DTA and DSC are both concerned directlywith the measurement of the thermal properties of the material under examination.In DTA, the difference in temperature between a substance and a referencematerial is recorded while both specimens are subjected to identical temperatureprograms. DSC is identical except that the energy supplied to maintain thesubstance and reference material at the same temperature throughout thetemperature program is recorded. In TG, the mass of a substance is recordedwhilst the substance is subjected to a controlled heating program.The applications of TA methods are vast [51-54]. In the field of polymeranalysis, thermogravimetry is mainly used for investigating the thermal degradationof polymers. Here, TG is preferred to DTA and DSC techniques as the complexreactions taking place may not be clearly endothermic or exothermic in nature,making results from the latter two techniques more difficult to interpret [55]. Usingthermogravimetry, Liebman et a!. [47] investigated the thermal decomposition ofPVC and chlorinated-PVC in inert and oxidative atmospheres. The use of TG toexplore the chemistry of thermal degradations has also been reported. Forexample, Ahmad and Manzoor described the use of TG to assess the catalyticeffect of ZnCl2 on the dehydrochlorination of PVC [56]. As discussed earlier, TG isfrequently used in conjunction with other methods (e.g. FTIR, MS) as the techniquedoes not give any indication of the nature of the products. The determination of24copolymer composition or polymer formulation is also routinely undertaken usingTG. For example, Cassel and Gray discussed a method to determine the amount ofplasticizer present in a PVC formulation [57].Wendlandt [58] discussed a number of the applications of DTA and DSC topolymer characterization. These two methods are used to study the thermaltransitions occurring in polymers. The glass transition temperature is routinelymeasured using the DSC and DTA techniques [59]. The degree of crystallinity in apolymer may be also determined [60]. The study of crystallization processes is animportant application of DSC or DTA methods [61]. In a practical application of this,DSC has been used to identify the components of plastic waste based on thecharacteristic values of Tm [62].1.2.2.7 Pyrolysis methodsLiebman and Levy [63] covered the use of pyrolysis and GC methods for theanalysis of polymers in a 1985 publication while a full list of general pyrolysisapplications was given by Wampler in 1989 [64]. Usually the pyrolysis oven islinked to GC and/or MS. FTIR detectors have also been used.One of the major uses of pyrolysis methods for polymer characterization is incompositional analysis. Using time-resolved pyrolysis-field ionization MS, Plageand Shulten [65] were able to correlate the structural units present in numerouspolyesters with the observed mass signals. The spectra obtained with polyesterswhich differed only by a single methylene group in the repeat unit weredistinguished.Pyrolysis methods can also provide information regarding the sequences ofmonomers in a copolymer. Using pyrolysis at 823 and 873 K, Rao et a!. [66] were25able to confirm sequence distributions for a butadiene-acrylonitrile-methacrylic acidcopolymer.The thermal degradation of polymers has been extensively studied usingpyrolysis methods. Shimizu and Munsen [67] discussed a Py-MS method whichwas used to study degradation processes in polystyrene, poly(methyl methacrylate)and PVC.1.2.2.8 Gas chromatography (GC)Currently, the major use of GC for polymer analysis and characterization is inconjunction with MS and/or pyrolysis techniques. However, applications involvingthe use of GC alone have also been reported. Typically, GC analysis is used toidentify and quantify the lower molecular mass additives present in the polymer.Supercritical fluid extraction (SEE) and supercritical fluid chromatography (SFC) arenow extensively used for this purpose. Ryan et a!. [68] discussed the extraction ofantioxidants from PE and PP using SFE and subsequent separation using SFC.1.3 Thermal deterioration of polymersThree types of thermal deterioration are generally recognized: thermaldegradation, thermal decomposition and pyrolysis, though frequently the terms areinterchanged. Thermal degradation applies to a thermal process leading to a lowdegree of polymer deterioration [69]. From a practical viewpoint, this process is ofmost interest to the polymer scientist since the physical properties of a polymer willchange with a low degree of degradation generally rendering the materialunworkable. Thus thermal degradation studies intend to gain an understanding ofthe degradation mechanism(s) and often develop suitable thermal stabilizers.Thermal decomposition refers to a thermal process by which the polymer isheated to a final temperature in excess of that required for extensive breakdown of26the polymer [69]. The main interest in this process is to investigate reactionproducts and assess toxicological hazards.Strictly speaking, the term ‘pyrolysis’ should refer to a transformation as theresult of heat alone [70], i.e., thermal decomposition in a non-oxidative atmosphere.A major interest in this process is to investigate aspects of the structure of theoriginal starting materials.The research detailed in this dissertation centred on the thermaldecomposition of polymers in air and therefore the following sections concentrateon this area. The polymers used are discussed individually, except for PE and PPwhich behave similarly.1.3.1 Thermo-oxidative decomposition of PE and PPWhen a polymer is oxidized, the usual result is a decreased molecular massand discoloration of the polymer. Under mild conditions of thermal oxidation,degradation of hydrocarbon polymers proceeds as shown [71].I. Initiation:RH+02 -> R+HO2HO2 + RH —* R +H20II. Radical conversion:R+O2 -* R02Ill. Chain propagation:RO2H RO + HOR02+RH -* R + RO2HRO2+RH —* R + products27IV. Degenerate chain branching:RO+RH -* R+ROHHO+RH R+HOV. Termination:R, RO2, productsIn practice the rate of oxidative degradation is often monitored using IRspectroscopic analysis of the non-volatile products [72]. Although alcohols andhydroperoxides (at 3550 cm-1)may be detected, the most important region of the IRspectrum is between 1710 cm-1 and 1735 cm-1. This region is a compositeabsorbance due to aldehydes and ketones during the early stages ofthermo-oxidative degradation and with carboxylic acids also being found duringlater stages. The presence of carboxylic acids is said to be indicative of chainscission processes.The rate of oxidation is affected by the chemical structure of the polymer [73].As branching leads to a lowering of the dissociation energy of the adjacentC-H bonds then linear PE is more stable to oxidation than branched PE and PP.However the effect of branching becomes less important at higher temperatures.The morphology of the polymer also has an effect. For PE, crystallinity influencesthe rate of oxidation as oxygen is unable to penetrate into the crystalline regionsand thus high density PE (HDPE) has greater resistance to oxidation than lowdensity PE (LDPE) [74].Studies of the thermo-oxidative decomposition of PE and PP are moreinfrequent but are of more interest here. Mitera et a!. [75] studied the thermaloxidation of PE and PP at 623 K. The major products identified for PE were ketonesand aldehydes (C212)together with olefins and paraffins. For PP, the main28oxygenous products were methyl alkyl ketones, though pyrolysis products includingthe hydrocarbonsC6H12,C9H18 and other oligomers of PP were found.Classification of the thermal oxidation products of PP and PE into three types wasdiscussed. The first type resulted from the continuing surface reaction of thepolymer with oxygen. The second type resulted from the reaction of primarybreakdown products with oxygen whilst the final type resulted from the reaction oftwo secondary products with each other or the reaction of secondary products withoxygen.The nature of products obtained from the thermo-oxidative decomposition ofPE and PP is affected by the heating rate. Wampler and Levy reported aprogression from oxygenated products to pyrolysis products as the heating rate wasraised from 10 K/mm to 20 K/ms for both PP and PE [76]. At heating rates below30 K/mm oxidation was the favored reaction with little or no pyrolysis products beingseen.Thermogravimetry has been used to study the thermal oxidation of PE andPP. Vasile, Costea and Odochian [77] identified three mass-loss stages during thethermo-oxidative decomposition of LDPE. These mass-loss stages, which could notbe separated, occurred in the temperature range between 493 K and 863 K, theonset and completion temperatures depending on the heating rate. Kinetic analysisof the data suggested a continuous modification of the reaction mechanism due tocompetition between the different processes involved (i.e., various decompositionpaths, diffusion and volatilization).Khorami et a!. [78] reported the use of FTIR spectroscopy to analyze theevolved gas from thermogravimetric analysis of several polymers included PE andPP. Using thermogravimetry, PP was found to be more stable in an inertatmosphere whilst PE was more stable in air. From the FTIR spectra of the evolved29products, methanol, carbon dioxide, water and carbon monoxide were identifiedtogether with the presence of aliphatic hydrocarbons and carbonyl compounds.1.3.2 Thermal deterioration of PVCThe low stability of PVC is such that if PVC was discovered today, it would beunlikely that the polymer would make it out of the research laboratory [79].Although advances have been made in elucidating the mechanisms of thedecomposition of PVC and the effects of polymer structure on decomposition, manycontroversies still remain, e.g., the role of HCI in the decomposition mechanism.Chirinos-Padron and von Schoettler [80] suggested that these discrepancies can beoften attributed to the difficulty of preparing interlaboratory reproducible samples.In an inert atmosphere, the thermal degradation of PVC begins in the regionof 423 K and is observed at an even lower temperature in air [81]. Two mainregions of mass loss are found during the thermal decomposition of PVC. The firstmass loss is mainly due to dehydrochlorination whilst the second mass loss is dueto carbonization of the dehydrochlorinated residue [82]. The dehydrochiorination ofPVC may be represented as shown below;H H HH H H H H H H H H H H H H11111 II A I II I...—C—C—C—C—C—C—C—C-—... ..—CC--C—C—C—C—C—C—... + HCII III II II IICIHCIHCHCIH CIHCIHCIHH HHHH HHH HHHHH HHH11111 I A 11111111...—C=C—CC—C=C—C—C—... .4 ...—C=C—C=C----C---C—C—C—... + HCII I I ICI H CI H CI H+ HCIetc....30Guyot and Benevise [83] suggested that dehydrochlorination was a chainreaction initiated by the thermal instability of irregular structures present in the PVC.These irregular structures include allylic chlorine atoms or branch points withtertiary chlorine atoms [84]. Bataille and Van [85] compared the rate ofdehydrochlorination of PVC under four kinds of dynamic atmosphere: helium,oxygen, air and hydrogen chloride. The rate of dehydrochiorination was highest inoxygen and lowest in helium. No catalytic effect due to the presence of HCI gas inthe heating atmosphere was reported. This is sharp contrast to other studies wherea catalytic effect due to HCI was observed. For example, Ahmed and Manzoorreported that the catalytic effect of HCI was due to the secondary bond associationof the polar HCI molecule with the double bond of the polyene making any allylicchlorine atoms more labile [56].The polyene sequences formed during dehydrochlorination causeunacceptable levels of discoloration even at 0.1% dehydrochlorination [86]. Thepolymer discolors from yellow to dark brown as degradation proceeds. Acomparison of the UV-visible absorption spectra of PVC degraded in air and PVCdegraded in an inert atmosphere shows that the polyene sequences are shorterwhen oxygen is present leading to a less intense color [87]. The reaction betweenoxygen and the polyenes is considered to proceed via a radical mechanism, te.,OCR+ ROCOCR+ ROOH31The peroxy radicals are formed as shown earlier for PE and PP. They canalso initiate further HCI loss leading to the autoacceleration observed when oxygenis present, i.e.,ROO + ...—CH2CH I—CH-- I ... ...—CH2—CHCI—CH—CHCI—...+ ROOH- ci.T 1 —nHCI...—CH2---JCH=CHj—... 4 ...—CH2---CH=CH— HCI—...L Jn+1In 1960, Talamini and Pezzin investigated the thermal dehydrochlorination ofPVC in nitrogen and in air at temperatures between 423 K and 523 K usingconductivity measurement of the evolved HCI [88]. Activation energies equivalentto 125 kJ/mol and 100 kJ/mol were determined for the reactions under nitrogen andair respectively. Similar results, confirming the difference in activation energy as afunction of atmosphere, were reported by Guyot and Benevise [83]. In more recentyears, these activation energy differences have been disputed. Abbas and Sørvikcompared the thermal dehydrochlorination of PVC in nitrogen, air and oxygen [87].Although the reported activation energies were well within the range of earlier datano significant differences due to atmosphere were reported. Differences in thepolyene reactions were reported with chain scission being observed in oxygenwhilst cross-linking was observed in nitrogen. Both reactions were observed in air.Boettner, Ball and Weiss [89] investigated the thermal degradation of PVCusing TG, DTA, IR spectroscopy and GC. Five separate stages of mass loss were32postulated though four of these occur during the carbonization region and aredifficult to separate visually. Qualitative analysis identified 59 volatile productsincluding HCI, CC, C02, benzene, paraffins and olefins. An increase in reactiontemperature led to an increase in CO and CC2, with only trace quantities of HCI andbenzene being found once the temperature was raised above 573 K. A higherheating rate (e.g., 50 K/mm) led to increased production of hydrocarbons, especiallyunsaturated compounds, with a corresponding decrease in CO2 and CO. Thereforepyrolysis dominates at higher heating rates whilst oxidation is the majordecomposition mechanism at lower heating rates. This concurs with the studiesusing PP and PE by Wampler and Levy [76].The effect of polymer structure on the thermal stability of PVC has also beendisputed. Chirinos-Padron and von Schoettler reported that for the three PVCresins used, the relative thermal stability did not correlate with polymerizationtemperature, molecular mass or syndiotactic content of the resin [80]. Theyindicated that the presence of additives used during the polymerization processappeared to be a significant parameter. This is in sharp contrast to the studiesreported by Kamal, Adam and Al Getta who indicated a molecular massdependence of the thermal stability of PVC [90].1.3.3 Thermal deterioration of PETThe thermal deterioration of condensation polymers such aspoly(ethylene terephthalate) has been studied less than that of the polyolefins orvinyl polymers. Ahlstrom [91] attributed this in some part due to the fact thatcondensation polymers may be chemically degraded and thus are more readilystudied by conventional analytical techniques.33Based on the identification of the small molecular mass fragments obtainedfrom the thermal decomposition of poly(ethylene terephthalate), threedecomposition pathways have been proposed [31, 39, 92];1. Random scission of the ester links involving a cyclic transition state thatproduces a carboxylic acid and a vinyl ester. This reaction is expected to beinitiated by free radicals when oxygen is present [93].OCC\CHOCCO..V...—O—CC—OH +H2CCH—O——O—...2. Transfer of a hydrogen with bond breakage to form a hydroxyl group andan unstable radical....OCC\CHOCCO..—QCC + HO—CH2CH— ——O—...343. Decarboxylation.—OCC—0H _O_C + CO2Bednas et al. [92] reported that the major volatile products of decompositionwere C02, CO and acetaldehyde. It was proposed that these products were formedfrom subsequent reactions of the reaction products shown above. For example,acetaldehyde would be formed by the reaction of the carboxylic acid and vinyl esterproducts of reaction 1.Day, Parfenov and Wiles [94] suggested that although the thermal oxidationof PET leads to the formation of numerous products, from a mechanistic point ofview only 10 major products were important. These products were identified as CC,CC2, CH4, acetylene, acetaldehyde, benzene, vinyl benzoate, benzoic acid, divinylterephthalate and ethylene dibenzoate.Carlsson et al. [95] reported the characterization of involatile residues fromthe thermal oxidation and pyrolysis of PET. IR analysis using the KBr disctechnique was reported to be unsatisfactory due to the optical problems caused bythe tough black residues. An alternate method based on diffuse reflectance IRspectroscopy was more successful with carboxylic acid end groups and linearanhydrides being identified. The spectra were reported to be similar forexperiments in air and in nitrogen with the dominant features being identical andonly small kinetic differences being found. Thus it was suggested that the thermaldecomposition of PET occurred in an 02-deficient local environment regardless ofthe external environment. This conclusion disagreed with the earlier work ofBirladeanu, Vasile and Schneider [96] who reported different activation energies for35the thermal decomposition of PET in air and in nitrogen. For all the heating ratesused, the activation energy in air was significantly lower than that in nitrogen.Cooney, Day and Wiles [97, 98] studied the thermal decomposition of PETusing TG. At least three stages of degradation were identified. Comparison of theresults obtained using a number of different numerical methods led to activationenergies of 122, 201 and 142 kJ/mol being reported for the three main stages ofdecomposition in air.1.3.4 Thermal decomposition of ethylene-vinyl acetate copolymersStudies of the thermal decomposition of ethylene-vinyl acetate (EVAc)copolymers have concentrated on quantifying the amount of vinyl acetate present.Maurin et a!. [99] reported that during the thermal decomposition of EVAc innitrogen two stages of decomposition were observed. The lower temperaturedecomposition was attributed to the vinyl acetate component, with the highertemperature decomposition being due to decomposition of the ethylene component.In a related study, a TG-FTIR method was used to identify the products from thethermal decomposition of an EVAc copolymer using a heating rate of 10 K/mm in anitrogen atmosphere [100]. Acetic acid was identified as the main product from thethermal decomposition of the vinyl acetate component which occurred from 633 to723 K. The thermal decomposition of the ethylene component was observed overthe temperature range of 723 to 823 K and products identified included I -butene,ethylene, methane and carbon monoxide.361.4 Acoustic emission1.4.1 IntroductionAn acoustic emission (AE) is a “transient elastic wave generated by the rapidrelease ofenergy within a material” [101]. Acoustic emissions are in the form of burstsof mechanical energy. The occurrence rate of these bursts varies enormously fordifferent systems. For rates above 50 kHz, the length of the bursts exceeds thetime interval between bursts and “continuous emission” is observed [102]. Acousticemission occurs over a wide frequency range including the audible region (10 Hz to16-20 kHz) and up to frequencies in excess of 1 MHz. For example, metallurgistsare aware of the “cry of tin”, which are the numerous small creaking soundsproduced when a sample of tin is deformed [103].Historically, AE has been the subject of research in materials science,seismology and mechanical engineering [104-107]. Newer areas of interest aremore widespread. Wade et a!. reviewed the occurrence of AE in a variety of fieldsincluding chemistry, entomology, wood science and botany [108].1.4.2 Chemical acoustic emissionIn 1978, van Ooijen et a!. [109] reported hearing a strong cracking soundduring the formation of dichloro(pyrazine)zinc(ll). The original work attributed theacoustic emission to either a change in the metal co-ordination or a rapidpolymerization process. More detailed studies reported by Munro [110] in 1991,attributed the acoustic emission to crystal fracture.Betteridge and co-workers studied many common chemical reactions to see ifAE was widely found in chemistry [111,112]. Of the forty-three chemical systemsinvestigated, thirty-two were found to emit detectable acoustic emission. Systemsgiving rise to acoustic emission included the oxidation of unsaturated hydrocarbons37by KMnO4, reaction of luminol with peroxide, acid-base reactions and therecrystallization of KCI.In an analogous study, Navratil investigated the incidence of acousticemission from a number of chemical reactions of particular relevance to separationprocesses in the chemical and mineral industries [113]. Reactions giving rise toacoustic emission including ion exchange reactions, thermal drying and thermaldecomposition. In one experiment, the thermal decomposition of ammoniumbicarbonate could be detected even with as little as 0.1 mg ofNH4HCO3!Sawada et al. detected AE from the gelation of sodium carbonate andcalcium chloride [101]. In a separate work, the same workers found that acousticemission was generated as a result of the dissolution and the precipitation ofsodium thiosulfate [114]. Signals generated during the precipitation process wereCa. 10 times bigger in amplitude than those generated during the dissolutionprocess.Belchamber and co-workers investigated the hydration of silica gel usingacoustic emission detection [115]. Two distinct types of acoustic emissions weredetected; the first due to fracture of the silica gel granules and the second due toformation and bursting of gas bubbles. The amount of acoustic emission detectedduring the experiment was linked to the mass of silica gel used, its particle size andto its initial water content.Evaluation of the AE phenomenon as a method of quantitative chemicalanalysis was also reported by Wentzell and co-workers [116,117]. Effervescence ofoxygen produced from the catalytic decomposition ofH20by the enzyme catalasewas monitored acoustically as an indirect method of determining peroxide. Goodcorrelation between the number of acoustic emission events and the concentrationofH20was shown for concentrations over the range of 5 to 100 mMH20.38Recent chemical acoustic emission research has included the electrolysis ofwater [118,119], osciNating reactions [120], and crystallization studies [121-123].1.4.3 Acoustic emission and polymersThe acoustic emission technique has been widely used during themechanical testing of polymers. Qian and Zhu gave an extensive review of thesubject [105]. During the stressing of the polymer specimen, a number ofdeformation processes such as crack growth and necking occur to dissipate thestrain energy. These deformation processes cause the acoustic emissions. As aresult, the acoustic emission detected is inherently related to the prefractureprocesses going on in the polymer, which vary with the type and form of thepolymer.Betteridge and co-workers reported several studies regarding the acousticemission from polymers under stress. In the earliest work, stressing of the polymerswas also carried out in the cavity of an ESR spectrometer. The results indicatedcorrelation between stress and both acoustic emission and free radicalgeneration [49]. Later studies concentrated on the statistical analysis of theacoustic emission signals from the stressing of polymers [124-127]. In particular,the use of mathematical methods led to classification of the acoustic emissionsignals resulting from the stressing of a variety of polymers.In 1953, Kaiser first observed that during the cyclic loading of metals, AE isnot detected in repeated loading cycles up to the maximum strain of previousruns [128]. For polymers the “Kaiser effect” may or may not be observed, againdepending on the physical properties of the polymer. For example, Qian and Wangreported that the “Kaiser effect” was observed for an acrylonitrile-styrenecopolymer (ANS) in the glassy state but was not observed for a cis-polybutadiene39rubber [129]. For the rubber, any crack formed during loading will be self-healingand thus repeated loading will have little effect on the specimen. For the morebrittle ANS polymer, any microcrack formed will be permanent and thus of effect inlater loading cycles.Of more relevance to the work reported here, are studies concerning acousticemission during the crystallization of polymers. Galeski and co-workers observedacoustic emission during the isothermal crystallization of polymers includingpoly(methylene oxide), high density PE and isotactic PP [130,131]. Acousticemission was found to occur only for the PP and poly(methylene oxide) polymers.Optical microscopy revealed cavities located between the spherulites formed duringcrystallization for those polymers giving acoustic emission. For the PE sample,which gave no significant amounts of acoustic emission, only small spherulites andno cavities were found. It was therefore suggested that the acoustic emission wasproduced by the abrupt stress release in these cavities. The stress build-up in thecavities was attributed to density changes during crystallization. Comparison of theAE technique and DSC showed that most of the acoustic emission signals weregenerated near to the completion of the crystallization process. Although it was notdiscussed in their paper, figures indicated that the onset of the acoustic emissioncoincided with the peak of the DSC thermogram. Similar results were found fromstudies investigating AE from the non-isothermal crystallization of PP [132].1.4.4 Sound wave propagationSound waves are longitudinal waves, i.e., the particles of the transmittingmedium (such as air molecules, collections of atoms in a solid) move in the directionof the wave motion. Propagation of the sound waves is due to the transfer ofenergy through the material. The propagation is in the form of successive40compressions and rarefractions of the particles. The propagation results in theparticles being displaced from their equilibrium positions. Beattie [102] describedan acoustic wave as an oscillating strain (or stress) moving through a material. Theresistance of the particles to being displaced from their equilibrium position,together with the density of the material, determines the velocity of the wavethrough the material.The concept of sound as purely a longitudinal compression wave is anoversimplification. Two types of acoustic waves occur in bulk materials. These arelongitudinal (compressional) waves, where the particle motion is parallel to thedirect of propagation and transverse (shear) waves where the particle motion isperpendicular to the direction of propagation. The velocity of the transverse wave istypically 50 % of that of the longitudinal wave and thus in acoustic emission studies,the longitudinal wave is of main interest. Other types of acoustic waves such asRayleigh waves or Love waves may arise due to surface effects [102].As the acoustic wave propagates through the material, a number ofprocesses occur which change the characteristics of the wave. For example, thevelocity dependency on frequency leads to a change in the shape of the wave as itpropagates through a material. Attenuation effects lower the energy of the wavedue to scattering or dispersion, and these may also be frequency dependent.When the acoustic wave reaches an interface between two surfaces, part ofthe wave will be reflected and part of it will be transmitted. The intensities of theresulting waves are determined by the acoustic impedances of the two materials.The acoustic impedance of a material is given by the product of the wave velocity inthe material and the density of the material. For good transmittance of the acousticwave, the acoustic impedances of the two materials should be identical. For a wave41entering perpendicular to the interface, the intensity of the transmitted component isgiven by,(4Z1Z2)2 (Equation 9)(Z1 2)where 1 is the fraction of the original intensity transmitted and the Z ‘s are theacoustic impedances of the materials (KgI(m2s)).An additional effect arises due to the non-ideal nature of the surfaces. Formost solids, the surface is rough when considered at the microscopic level.Therefore only a fraction of the two surfaces is actually in contact. This leads to alarge decrease in the transmittance of the acoustic wave. Thus in acoustic emissionstudies, a thin layer of a fluid or a setting solid known as the couplant is placedbetween the two surfaces to fill the microscopic gaps in the contact. Typicallysilicone grease is placed on the surface on the transducer to act as a couplant,though Beattie [133] indicated that a variety of other couplants including water,petroleum grease, dental cement, butter and honey had found some use in acousticemission studies.In summary, the characteristics of the acoustic waves are greatly influencedby transmission through various materials and across the interfaces betweenmaterials. Where possible, the number of materials that the acoustic wavepropagates through between the source and detector should be minimized.Although a quantitative description of the transmission of acoustic waves throughmaterials is still not available, previous acoustic emission studies have indicated theacoustic emission is indicative of the chemical and physical processes occurring insamples and can be related to the nature and location of the acoustic emission(e.g., [111]).421.4.5 Instrumentation for acoustic emissionThe acoustic emission generated by the source (within the sample) isconverted to an analog electrical impulse by a broad-band piezoelectrictransducer [102]. The acoustic emission signal1 is then amplified and filtered toremove the majority of low frequency ambient noise and any high frequencyinterferences. Typically, a band-pass filter of 50 kHz to 2 MHz is used.Numerous methods have been used to monitor the acoustic emission signals.These may be related to continuous monitoring of the acoustic emission level oracquisition of individual acoustic emission signals.1.4.5.1 Continuous monitoring of the acoustic emissionContinuous monitoring of the acoustic emission level is used to investigatethe kinetics of the process(es) producing the acoustic emission. For example,VanSlyke showed that a model that represented two simultaneous first-orderprocesses could be used to fit cumulative AE curves for the hydration of alkali metalhydroxides [134].The earliest method of continuously monitoring the acoustic emission levelwas to use a chart recorder to give a simple measure of the intensity andoccurrence of the acoustic emission as a function of time. Obviously, this was farfrom ideal as post-data collection processing was very limited.The most frequently described method in the literature for the collection ofacoustic emission data is the “ring-down counting” method [105]. In this method, allthe components of the acoustic emission signal whose amplitudes are greater thana pre-set threshold level are counted. These counts are given as a rate or a totalFor the sake of clarity, the term ‘acoustic emission signal’ will be used to refer to theelectrical signal produced by the transducer whilst the term ‘acoustic emission’ refers to theacoustic wave generated by the material [102].43count. The main disadvantage with this method is its lack of theoreticalsignificance [135].A more recent method has involved the use of integrators [136]. This methodconsiders the total energy of acoustic events rather than just a measure of the peakamplitude of the acoustic event and thus is especially suitable for quantitativeapplications where the AE signals differ widely in both peak amplitude and totalenergy.1.4.5.2 Capture of individual acoustic emission signalsThe characteristics of the individual acoustic emission signals may provideextensive information about the mechanisms producing the acoustic emission.Belchamber and co-workers classified five types of acoustic emission producedfrom the hydration of silica gel based on the characteristics of the individualacoustic emission signals [115]. Two of the types represented acoustic emissionsignals produced by gas bubbles and two others represented signals produced byfracture of the silica gel granules. The fifth class (one signal) was excluded.In early chemical AE studies reported by Betteridge et al. [111], individualacoustic emission signals were acquired using minicomputers with dedicated dataacquisition systems. More recently, transient digitizers such as digital storageoscilloscopes have been extensively used [120]. The acoustic emission signals aretypically captured, each as a 1024-point signal with 8-bit resolution at rates up to5 MHz, prior to transfer to a microcomputer for later processing. Some I 2-bitdevices are available but the increased resolution is achieved at slower acquisitionrates.44A wide range of methods have been used to analyze the individualAE signals. Wade et a!. classified these as either cumulative acoustic spectralanalysis or individual waveform analysis [108].1.4.5.2.1 Cumulative acoustic spectral analysisThe acoustic emission signals are acquired as time-domain signals, i.e., thesignal amplitude as a function of time. For spectral analysis of the acousticemission signals, a fast Fourier transform (FFT) [137] is applied to the time-domainsignal to give the frequency spectrum. Mathematically [138], the Fourier transformmay be represented as+c,DI(o)= fft).exp(-it) dt (Equation 10)-where ft) is the time-domain response, I() is the intensity at a frequency ofU) and I is used here to denote the square root of -1. As it is impossible tosample an infinite number of data points at an infinite sampling frequency then thediscrete Fourier transform is used rather than the continuous Fourier transformgiven in Eqn. 10, i.e.,NI(nöf) = f(köt)exp(-ik&nöJ) öt (Equation 11)k=1where n is an integer and N is the number of data points in the time-domainsignal, which were acquired at regular intervals of öt seconds. This transformationproduces a frequency domain signal consisting of N12 real points and N12 imaginarypoints with a frequency interval of öfs1. To combine the real and imaginary points,the power spectrum is calculated. The power spectrum P(U)) may be defined as45P(cr) = PL2() +IM2() (Equation 12)where the real spectrum is RL(co) and the imaginary spectrum is IMQx). Ofmore importance here is the magnitude spectrum, M(co), given byM(o) = g P(c) (Equation 13)Unfortunately, this is also known as the power spectrum. The definition givenin Eqn. 13 is used throughout the work discussed here.The individual power spectrum from an acoustic emission signal is of limiteduse because of the inherent variability found due to the nature of the signals. As aresult, average power spectra are preferred. These are generally reproducible overa period of many months for particular chemical systems and transducercombinations [139].Acoustic spectral analysis using average power spectra has proven to be ofuse in distinguishing the mechanisms producing the acoustic emission. Wentzelland Wade compared the acoustic emission from eight chemical systems on thebasis of the average power spectra [139]. Differences in the physical processesoccurring were linked to differences in the frequencies predominating in theacoustic emission with low-frequencies (e.g. <250 kHz) resulting from bubblerelease and higher frequencies resulting from crystal fracture.In other work, the intensity of frequencies across the 100-800 kHz range inthe average power spectra obtained from the electrolysis of water was found toincrease when the applied potential was raised beyond that required for gasevolution [118].461.4.5.2.2 Waveform analysisThe individual acoustic emission signals contain potentially, an enormousamount of information. The previous section dealt with the use of spectral analysisto distinguish acoustic emission signals based on their frequency spectra but thereis no reason why the acoustic emission signals may not differ in other ways too(e.g., duration). For example, the two classes of acoustic emission signalsproduced from the hydration of silica gel may also be distinguished on the basis oftheir intensities with granule fracture signals being more intense than bubblerelease signals [1831.In 1981, Betteridge and co-workers reported the use of descriptive statisticalfactors (descriptors) to characterize the acoustic emission signals [124].Descriptors including the mean and median frequencies of an individual acousticemission signal were used to mathematically compare acoustic emission signalsarising from various polymers under stress. In subsequent papers [1 25,1261, theset of descriptors was expanded to quantify signal characteristics including thepeak-to-peak amplitude, half-life and bandwidth. In 1991, Wentzell et al. [140]discussed the use of about 50 descriptors for chemical acoustic emission analysis.These descriptors were classified into four non-exclusive groups: (1) descriptorsassociated with the magnitude of the signal, (2) descriptors associated with theduration of the signal, (3) descriptors measuring the central tendency of the powerspectrum, and (4) those descriptors which characterized the dispersion of the powerspectrum.The use of descriptors to characterize the acoustic emission signals requiresthe application of mathematical methods to aid the discrimination process.Sometimes the required discrimination can be obtained by simply comparing thedescriptor values calculated for the acoustic emission signals of interest.47Wade et a!. [121] found that acoustic emission signals from the crystallization ofKNO3 could be detected acoustically based on differences in the ‘RMS’ of the realacoustic emission and background signals even though the peak amplitudes of thetwo types of signals were similar. However, in the majority of cases a more complexapproach is required.Empirical statistical methods including pattern recognition are ideal for use inthe analysis of acoustic emission waveforms. Methods such as principalcomponents analysis (PCA) [141] and cluster analysis [142] have been used inacoustic emission analysis. For example, Cook reported the use of PCA tovisualize differences between the acoustic emission signals arising from thecrystallization of KBr and artificially produced signals for the impact and fracture ofKBr crystals [123].The method of data visualization and analysis most commonly used in thisresearch group has been PCA. Figure 2 indicates a schematic of the mathematicalsteps involved in PCA. Prior to PCA, the multivariate data are expressed in matrixform where the number of rows is equivalent to the number of objects (e.g., acousticsignals) and the number of columns is equivalent to the number of variables (e.g.,descriptors). A data set ofp variables can be exactly represented byp principalcomponents, due to the fact that PCA is a rotation of axes of the data set. Abstractfactor analysis leads to the formation of two matrices, the scores and loadingsmatrices. The scores matrix contains information about the relationships betweenthe objects while the loadings matrix contains the principal components as well asinformation about the interactions between the variables. The principal componentsare linear combinations of the original variables such that each successive principalcomponent explains the maximum amount of variance possible in the data notaccounted for by previous principal components. Most of the variance of the data is48accounted for in the first few principal components and thus the dimensionality ofthe data (Le., number of variables used) may be significantly reduced.correlationr 1 analysis r -i correlation[ J LI matrixraw abstract factor analysisdata[]loadings[ ] matrixscoresmatrix Relationships/ among variables///Relationshipsamong samplesFigure 2. A schematic of the steps involved in principal components analysis(Adapted from [185]).Overall, Meglen [143] suggested that PCA may be viewed as a means toproject the original data from its multidimensional representation down to two orsometimes three dimensions for graphical display. Practically, the experimentalistviews the fewest number of principal components that are required to representperhaps 80-90 % of the variance.In principal component analysis, any grouping (or clustering) of objects isusually determined visually from plots of the principal components. In clusteranalysis, when there is no a priori information about the group structure of the data,a variety of methods may be used to divide a set of objects into a number ofclusters. It is important to note that cluster analysis methods do not compute thenumber of clusters present in a data set. Rather, they assign the members of eachcluster given the number of clusters in the data. In this respect the use in tandem of49PCA and cluster analysis was envisaged. PCA would allow visualization of thenumber of clusters present in a data set and cluster analysis would categoricallyassign the objects that belong to the cluster based on mathematical means ratherthan a personal (possibly biased) basis. Numerous types of clustering techniquesare available. Please see reference [142] for more details.In a recent publication, this author reported a rules-based approach to theclassification of chemical acoustic emission signals [144]. The approach was basedon calculating the likelihood that the descriptor values of a candidate acousticemission signal lay within the descriptor distributions determined for the group ofinterest. Using this method, the signals belonging to the two processes producingacoustic emission during the hydration of silica gel (i.e., fracture and bubbleevolution) were distinguished. It was also applied to ‘intelligently’ distinguishbetween low level signals and background noise.t5 ThermosonimetryThermosonimetry (TS) has been defined as “that technique in which the soundemitted by a substance is measured as afunction of temperature whilst the substance issubjected to a controlled temperature program” [50]. Therefore TS is concerned withthe detection of thermally-induced acoustic emission. The acoustic emission maycome either from purely mechanical sources or from physicochemical sources [145].In the former, mechanical strain within a substance is induced due to the applicationof a thermal stress. Often this is the result of a thermal shock treatment.Mechanical strain may also be brought about by a thermal event, e.g., phasetransitions, sublimation, decomposition, oxidation or combustion. The mechanicalstrain is released with the propagation of microcracks and may be accompanied byacoustic emission depending on the level of thermal stress.50Thermosonimetry is concerned with the detection and meaning of the variousacoustic emissions occurring prior to, during and after thermal events. As a result,TS investigations contribute to an understanding of the thermal behavior of solidmaterials and the dynamic processes of the solid state.1.5.1 InstrumentationThe earliest thermosonimetry apparatus was reported by Scott who studiedthe decrepitation of minerals [146]. The method consisted of heating a sample ofthe mineral being studied in a glass tube suspended in a furnace and listening tothe decrepitation by means of a rubber tube or a microphone-amplifier-headphonelink. Smith and Peach [147] developed an apparatus in which the mineral samplewas placed in a PyrexTM tube having a thermocouple built into it to enable accuratetemperature measurement. The Pyrex tube was then mounted so that thesample was suspended in an electric furnace. A crystal microphone was fixed tothe other end of the tube as the detector.Modern designs of thermosonimetry apparatus are based on the apparatusfirdst described by Lønvik in 1978 [148] (Fig. 3). In this apparatus, the sample isheld in the sample head of a quartz stethoscope. Sonic activity in the sample ispicked up and transmitted by the stethoscope to a piezoelectric cell, which convertsthe vibrations to electrical signals. To prevent interference from external noise, thepiezoelectric cell is fixed on a heavy recoil foundation and a seismic mount. Use ofan acoustic couplant such as a silicone oil improves signal transfer between thestethoscope and the piezoelectric cell. Direct insertion of a thermocouple in thesample caused severe damping of the acoustic signals and thus the thermocouplewas situated.as close as possible to the sample without touching it.51Protection tube of aluminaSampleStethoscopePt/i 0% Rh thermocoupleFurnace thermocoupleHeating element of silicon carbideRadiation shieldsTemperature control systemAmplifier systemFigure 3. The thermosonimetry apparatus designed by Lønvik.(adapted from [1481)Clark [145] also discussed the construction of an apparatus forthermosonimetry. Improvements over Lønvik’s design included a waveguideassembly that allowed accurate positioning of the waveguide and improvedreproducibility. Subsequent modification of this apparatus by the addition of a DTAsample holder enabled the measurement of concurrent TS-DTA data [149].In 1990, Shimada and Furuichi [150] described an apparatus which permittedsimultaneous TS-DTA. Although the thermocouple required for the DTAmeasurement was inserted into the sample no damping effects were reported.FurnaceSeismic mount ofthe pick-up systemInlet foratmospheric controlStethoscope mountingand cell basementPiezoelectric cellPre-amplifierVacuum sealed housing ofthe pick-up system5215.2 Signal analysisThe electrical signal generated by the piezoelectric transducer is a complex(unknown) function of the mechanical distortions or perturbations taking place in thesample during the acoustically active event. These distortions create elastic wavesin the waveguide. These waves appear as damped ringing oscillations of thenatural resonance frequencies of the waveguide assembly [135]. Additionalmodification of the wave occurs due to transmission through any material or acrossany interface. Lønvik summarized the output from the waveguide to thepiezoelectric transducer as being a complex cascade of overlapping ‘ring-down’pulses of various amplitudes, frequencies and rates of decay [151].Usually, TS data are recorded based on counting the number of events per°C or sometimes the number of events per second. Almost all studies have used‘ring-down’ counting methods though the use of frequency analysis has also beenreported [152].1.5.3 ApplicationsThermosonimetry is often used in combination with other thermal analysistechniques in order to more fully characterize materials. In 1989, Clark gave anextensive review of areas in which thermosonimetry had been applied [135]. It wassuggested that hard and brittle materials give more acoustic emission as theyrelease the thermal (or mechanical) stresses by microcrack generation andpropagation [153]. Soft and ductile materials release the thermal stress by plasticflow and thus are acoustically less active. As a result, materials such as minerals,glasses, ceramics, refractories, metals and alloys have been widely studied by TSwhereas organics and polymers have received little attention. The following53sections detail some of the applications of thermosonimetry. For a more extensivereview, Clark’s work [135] should be consulted.1.5.3.1 MineralsThe use of thermosonimetry to study mineral decrepitation was establishedby Scott [146], Smith and Peach [147]. Scott studied a number of minerals usingthe technique including quartz, calcite, siderite and pyrite [146].Many minerals have been investigated using thermosonimetry. Studiesindicated that acoustic emission from the dehydration of brucite, Mg(OH)2 at673-723 K occurs in stepwise bursts, 3 K apart due to the formation of temporaryinterfaces [148]. The mineral kaolinite,A12(OH)4.Si05has been studied byseveral workers [154-156]. Hindar et a!. [154] used thermosonimetry, together withother thermal analysis techniques, to investigate the decomposition processes.Two regions of TS activity were observed. The lower temperature regioncorresponded to dehydroxylation and the higher temperature region was attributedto a solid-state phase transition.Thermosonimetry has also been used to study organic minerals. Lønvik andco-workers investigated the thermal behavior of Green River oil shales using anumber of complementary techniques including DSC, electrical conductivity,dielectric spectroscopy and thermosonimetry [157]. Four distinct chemical orphysical processes were identified. By comparison of the data obtained using thevarious techniques, possible origins of the processes were discussed.1.5.3.2 Glasses, ceramics and refractoriesClark investigated the growth of microcracks in a soda-lime-silica glass andin a borosilicate glass using thermosonimetry [158]. The number of AE eventsduring quench cooling was proportional to the calculated applied thermal stress.54Based on the thermosonimetry results, microcrack lengths and microcrack densitieswere determined.Thermosonimetry has been used to assess thermal shock resistance andthermal shock degradation in many ceramics and refractories. Clark indicated thatthermosonimetry could replace the existing thermal cycling and mechanical testingmethods for assessing thermal shock resistance [135].Determination of the glass transition temperature (Tg) and the softeningtemperature (T) of inorganic glasses using TS has been reported [159]. A goodcorrelation was found between TS and dilatometry results.1.5.3.3 Metals and alloysStagni and Congiu Castellano reported the detection of an ct-B transition inSn single crystals using thermosonimetry [160]. Acoustic emission was onlydetected if the Sn crystals were first cooled below a characteristic temperature(Ca. 287 K) prior to heating to 403 K. Increasing the time the single crystals spentbelow the characteristic temperature prior to heating led to an increase in number ofa-nuclei and an increase in the TS activity on heating.Numerous studies using thermosonimetry to study martensitictransformations in alloys have been reported. These transformations are oftencharacterized by the co-operative shearing movements of groups of atoms that leadto shape deformation and often audible “cIicks. Picornell et a!. discussed the useof a combination of optical microscopy and thermosonimetry to study the martensitictransformation in a Cu-Zn-Al alloy [161]. They concluded that the acoustic emissionwas related to accelerations in the movement of the martensite microplates.The crystallization kinetics of metallic glasses have been studied usingTS [162, 163]. Hunderi and Lønvik [162] reported studies for metallic glasses55having the nominal compositions ofFe78Mo2B0andFe40Ni40P14B6.Comparisonof TS and DSC results indicated that the TS peak temperature was higher than theDSC peak temperature at heating rates greater than 2 K/mm. On extrapolation to azero heating rate (i.e., isothermal conditions), the TS peak temperature coincidedwith the DSC onset temperature. An excellent agreement between activationenergies obtained using the two techniques was found.1.5.3.4 Inorganic materialsThe phase transitions of many inorganic materials have been studied usingthermosonimetry. Clark and Garlick reported TS curves for the National Bureau ofStandards-International Confederation of Thermal Analysis (NBS-ICTA)recommended DTA standards, including KCIO4 and S102 [149]. Two distincttemperature regions of activity were identified. The lower temperature region wassuggested to be associated with fluid inclusion release and microcrack propagationwhilst the higher temperature region was associated with polymorphictransformations. Comparison of TS and DTA results obtained during concurrentexperiments indicated that the TS peak temperature coincided with the onset of theDTA peak. It was concluded that the acoustic emission released is a measure ofthe strain release prior to the thermal event. The influence of heating rate andparticle size on the TS results was also reported. Generally, increasing the heatingrate and the particle size led to an increase in the TS activity.Shimada and co-workers described extensive investigations on thesimultaneous thermosonimetry and DTA of inorganic salts including the alkali metalperchlorates [150, 164-168], NaN3 [150], KNO3 [150] and MgCl2.6H20 [150].Thermosonimetry studies involving KCIO4 confirmed the earlier work of Clark andGarlick [149]. Supplementary results using scanning electron microscopy (SEM)56indicated that the low temperature region of TS activity for KCIO4 resulted from thebreak up of large particles overlapped with the orthorhombic to cubic transition ofKCIO4. The high temperature region of TS activity was identified as being due tothe melting and decomposition of KCIO4, followed by the solidification of the KCIproduct [1661. The influence of particle size, sample form, sample thickness,sample holder and sample mass on the TS results was also reported. No change inthe TS activity as a function of sample thickness or sample holder design wasfound. For the other parameters, the influence was less where the acousticemission was due to a thermal event as opposed to where the acoustic emissionwas due to a mechanical event. For KCIO4 and NaClO4.H20,linearity in thevariation of the AE count rate in the low temperature region as a function of samplemass (Ca. 500 mg to I mg) was found [165,167].The alkali metal dichromates have also been studied usingthermosonimetry [169-171]. Clark et a!. used thermosonimetry in conjunction withhigh temperature X-ray diffraction, DTA and hot stage microscopy to study thethermal behavior ofK2CrO7[169]. Other studies indicated that the influence ofparticle size depended on the transition.Sakiev and co-workers reported the use of thermosonimetry to determine thedecomposition temperature of inorganic salts including CuSO4.5H20[172],(NH4)2C03andNH4CIO powders [173]. Using thermosonimetry, the temperatureof onset of dehydration of CuSO4.5H20was determined to be Ca. 319-321 K, whichis significantly lower than that determined by TG or DTA (i.e., 343-345 K). Thisdiscrepancy was explained by the formation of defects including microcracks inlocalized regions of the crystals prior to significant decomposition which led to thegeneration of acoustic emission.57Clark investigated the frequency distributions during the thermosonimetry ofa number of inorganic materials includingK2S04,KCIO4,CuSO4.5H20andK2CrO7 [152]. Different frequency distributions were observed for differentsubstances and this was related to the nature of the processes occurring duringheating.1.53.5 Organic and polymeric materialsThe thermosonimetry of organic materials has been rarely studied incomparison with the other areas discussed.Clark indicated that the crystallizations of benzoic acid and adipic acid couldbe monitored using thermosonimetry [174]. Frequency spectra of the acousticemission detected for the two compounds were found to be distinguishable. Theeffect of doping adipic acid crystals with various unsaturated fatty acids on thethermal behavior of adipic acid was assessed using thermosonimetry [175]. The TSactivity was found to decrease with increasing impurity content according to aninverse square law.Phase transitions of organic compounds have been found to be acousticallyactive. Lee et al. studied the phase Il/Ill polymorphic transition of hexachloroethaneat 317 K using thermosonimetry [176]. The results were found to correlate well withdilatometry measurements. Other systems studied using thermosonimetry include2,2-dimethyl-1,3-dipropanol [177], p-cresol and liquid crystals [114].As discussed earlier in section 1.4.3, thermosonimetry has been used toinvestigate the crystallization of organic polymers but no other areas of syntheticpolymer chemistry had been studied using the technique.581.6 Scope of this thesisThe initial goal of the present work was to investigate the acoustic emissionphenomenon with regard to synthetic polymer processes. During the preliminaryinvestigations, the phase transitions of polymers, the thermal degradation ofpolymers and polymerization processes were examined. Initially each of these werestudied to see if acoustic emission was produced during the process. The use ofacoustic emission to monitor the process was then assessed, e.g., whether acousticemission could provide useful information about the kinetics of each process.Subsequent studies centred on the thermo-oxidative decomposition ofpolymers using thermosonimetry. The thermo-oxidative decomposition of PVCserved as the primary system for study. Investigations centred on the effect ofchanging known experimental parameters, e.g., sample mass, type of polymer,copolymer composition. Where possible, experiments undertaken using TG servedas a basis for comparison. Other experiments used thermogravisonimetry (TGS), apurposely developed combination of TG and TS to facilitate a better understandingof the acoustic emission phenomenon. To permit a better understanding of thesignificance of the TS results, extensive use was made of other thermal analysismaterial, both in terms of analysis methods and the general understanding ofthermal processes.Other experiments were aimed at facilitating a better understanding of thedegradation mechanisms and the mechanisms producing the acoustic emission.Techniques such as IR spectroscopy, elemental analysis, mass spectrometry andscanning electron microscopy were all employed to assess the residues producedas a result of polymer decomposition.Prior to the TS and TGS studies, a suitable experimental apparatus wasdeveloped. Chapter 2 details the evolution of this apparatus.59CHAPTER 2. DEVELOPMENT OF THERMOGRAVISONIMETRYThermal analysis methods tend to provide less specific information on thebulk sample than many other analytical methods, e.g., IR spectroscopy.Nevertheless they are widely used. As with many analytical methods, the use of asingle thermal analysis technique does not usually provide sufficient informationabout a given system. This is especially true for the thermosonimetry techniquewhere the origins of the phenomenon are “uncertain” [108]. The components of athermosonimetry apparatus were first discussed in detail by Lønvik [148]. His initialdesign did not provide for the use of thermal methods besides thermosonimetry.Thus instruments that combined differential thermal analysis (DTA) andthermosonimetry (TS) were developed [149,150]. A schematic for the TS-DTAsystem developed by Clark and Garlick [149] is shown in Figure 4.Figure 4. A schematic of the thermosonimetry apparatus developed by Clarkand Garlick [149].60This is not simultaneous TS-DTA but DTA and TS run concurrently onseparate samples in the same furnace. In the TA literature, it has been usual toterm this method, a “combined technique” [178]. Shimada and Furuichi reporteddetails of studies using a simultaneous TS-DTA system in 1990 [150].2.1 ExperimentalThe present work evaluates a combination of thermogravimetry (TG) and TS.This technique was termed thermogravisonimetry (TGS) and a schematic of theapparatus developed is shown in Figure 5.Figure 5. A schematic of the thermogravisonimetry apparatus.The TGS and TS-DTA instruments have some features in common,especially the design of the thermosonimetry components. The main differencebetween the TGS and the TS-DTA apparatus shown in Fig. 4 is that the TGSapparatus permits a truly simultaneous experiment (i.e., TS and TG).FurnaHeater61An apparatus for thermosonimetry was developed before incorporating thecomponents for thermogravimetric analysis. As the thermosonimetry apparatusdeveloped consisted of a number of important components, it is pertinent to dealwith each component individually.2.2 Heat sourcesDuring the time covered by the work in this thesis, several heat sources wereused in the thermosonimetry apparatus. Experiments involving the thermo-oxidativedecomposition of polymers used the arrangements discussed in the followingsections.2.11 Basic oven apparatusIn the initial investigations, heat was provided by a laboratory oven (CENCOmodel LR-29744,CAT #95470-1 6; Central Scientific Company of Canada Ltd,Vancouver, BC). A hole of 25 mm diameter was drilled through the bottom of theoven to allow passage of a waveguide such that the piezoelectric transducer couldbe kept external to the heat, whilst the sample was inside the oven. A hole was alsodrilled through the top of the oven. This enabled the approximate temperature ofthe sample to be monitored by placing a thermometer through this hole. Theposition of the thermometer was adjusted so that the temperature bulb was adjacentto the sample held in the waveguide. The heating rate and the final temperaturereached were established only by adjustment of a dial on the oven. No externalcontrol or temperature programming was available. This apparatus is shown inFigure 6.62Chart recorderDigital storageoscilloscopeFigure 6. A schematic of the oven apparatus used for some of the preliminaryexperiments.Preliminary experiments using this apparatus indicated that the primitivetemperature control noted above was unsatisfactory. The heating profile was notsmooth and the final temperature reached was unstable. Further studies implied alow degree of reproducibility both in the heating rate and the final temperaturereached. Figure 7 indicates a typical heating profile using the oven. The oven alsocaused periodic electrical noise spikes which were collected by the acousticemission data acquisition instrumentation as spurious signals. Two 75 dB highfrequency mains noise filters and suppressors (ISOBAR Model IBAR-4-GS; TrippeManufacturing, Chicago, IL), placed in series, were used to minimize themains-borne interferences but did not eliminate the problem. These electricalspikes were caused by the oven being switched off whilst near the peak voltageOvenThermometerWaveguide located inside ovenin contact with transducerConditioning amplifierTransducerComputerDigital signal63during the a.c. cycle, leading to a very large momentary change in the voltage andgenerating the electrical noise.An alternative heating apparatus was then designed which eliminated theseproblems.550 I I500450- JDt400jcL) 350 -300- -250 I I I0 3000 6000 9000 12000Time (seconds)Figure 7. A typical heating profile for the oven apparatus.22.2 Computer-controlled furnace apparatusA vertical tubular furnace (Model F21120, Thermolyne, Dubuque, IA, USA)was purchased. The 2-inch quartz liner required for the furnace was constructed byUBC Dept. of Chemistry Glass Shop. The apparatus used is shown in Figure 8.Initial testing of the furnace indicated that its natural heating curve was morerepeatable (and of much greater range) than the previous oven. However, it stilllacked adequate control and electrical noise spikes due to the switching of the ovenwere still a problem. Discussion with staff of the UBC Dept. of ElectricalEngineering led to the design and construction of a computer-controlled64zero-voltage switching device. A schematic of this circuit is shown in Figure A ofAppendix 1.Waveguide locatedinside tubular furnaceWaveguide IHolder AE acquisitionI apparatusBalance LFigure 8. The furnace equipment used during the thermosonimetry studies.Use of this circuit led to the elimination of the electrical noise spikes; now theoven was being switched at the zero voltage of the a.c. cycle rather than at randompoints between zero and peak voltage. The power supply incorporated into thecommercial furnace was removed as the power to the furnace was provided via thecomputer-controlled zero-voltage switching circuit. The actual power supplied tothe furnace was determined by an 8-bit digital value supplied via the parallel port ofa PC/AT computer (Type 286-12, Campus Computers, Vancouver, BC). Thefraction of the a.c. mains voltage supplied to the furnace was determined by thevalue received, with a value of 255 corresponding to the maximum supply. Softwareto enable the user to input the heating value was written in BASIC (MicrosoftQuickBasic version 4.5, Microsoft Inc., Mississauga, ON). The code used is listedin Appendix 2. Figure 9 indicates three trials of the heating curve obtained when a65constant value of 125 was supplied to the computer controfled zero-voltageswitching device. The improved repeatability of both the final temperature and theheating curve, were apparent (compare with Fig. 7).1800 3600 5400 7200Time (seconds)900800— 7000D60000E05004003000 9000 10800Figure 9. Typical heating profiles for the furnace apparatus.During the preliminary experiments with this apparatus, the need to recordthe temperature at which certain acoustic events occurred became apparent. Thetemperature of the furnace was indicated by an analog meter, ‘ocated on the controlunit of the furnace. Unless the temperature given by this meter was recordedmanually, no record of the temperature at any time during the experiment would bekept. Automatic logging of temperature values was implemented by interfacing thecomputer to the thermocouple circuitry of the furnace. The ‘Type K’(chromel-alumel) thermocouple used for measuring the temperature of the furnacewas disconnected from the analog meter of the furnace and connected to athermocouple amplifier circuit [179]. A schematic of this circuit is shown in Figure B66of Appendix 1. The circuit was constructed by Dr. L.E. Bowmanof this researchgroup. The circuit was designed so that its output voltage is linearly related to thetemperature of the thermocouple. For each I K rise in temperature, the voltageoutput from the thermocouple amplifier circuit increases by about 10 mV. Theoutput was connected to the second analog-to-digital channel(AD#1) of a dataacquisition card (IBM Data Acquisition and Control Adapter; IBM, Boca Raton, FL)installed in the PC-AT computer. The input range for all the channels was set from0 to 10 volts. Channel AD#0 was used for acquisition of acoustic emission data.The voltage readings from the thermocouple interface were calibrated so thattemperature was recorded by the computer, instead of voltage.Figure 10 indicatesthe correlation between temperature and the output from the operational amplifiercircuit.Calibration data were taken from the information sheet provided with thethermocouple amplifier chip [179] though only a range of 20 °C (293 K) to 1000 °C(1273 K) was used here. Linear regression of the data indicateda coefficient ofdetermination (r2) of 0.9999, a slope of 10.30 ± 0.02 mV/°C, andan intercept of-31 ± 13 mV, i.e.,10.3To -31(Equation 14)where V0 is the output voltage from the thermocouple amplifier circuit andToc is the temperature in degrees Celsius. The givenerrors refer to the 95 %confidence limits.The precision of the intercept was not as high as might be expected and wasequivalent to an uncertainty of 1.3 °C in the readings. This was not consideredsubstantial enough on a relative scale for concern, considering that most67experiments would be done at temperatures between 180 and 600 °C. Thesevalues were used for the calibration of the interface.1000080006000 -c04000 -20000 I0 250 500 750 1000Temperature (°C)Figure 10. The voltage-temperature calibration for the thermocouple amplifiercircuit.As the analog-to-digital converter represents the input voltage (0-10 V) as aintegral fraction of 4095 (for 12-bit resolution) then the overall equation to convertthe value acquired by the computer to the temperature measured by thethermocouple may be written as:TK = 273.2+ j_bO0OOAD-3 (Equation 15)where TK is the temperature in Kelvin and AD is the digital representation ofthe input to the computer (0 to 4095). The factor of 10000 arises due to theconversion of the input range (0-10 V) to millivolts. The software code written torecord the thermocouple reading is listed in Appendix 2.682.2.3 Incorporation of feedback control routinesThe temperature control for the furnace apparatus was improved further bydevelopment of a feedback system. To facilitate this, the computer programs foroven control and thermocouple measurement were merged. The schematic inFigure 11 indicates how the feedback system operated.Zero-voltageswitching circuit/FurnaceIermocoupleThermocouple Amplifier( Required reference_______________) temperature T(t)cacuIatedNew power output calculatedusing feedback algorithmCOMPUTERFigure 11. A schematic of the oven feedback control.LPT1C IBM Data Acquisition Card\/C Comparison of currentand reference temperatures HDAt the start of the experiment, an oven control value (OCV) between 0 and255, was sent by the computer to the zero-voltage switching device. On receipt ofthis value, it set the output power to the oven. This led to a change in thetemperature, which was recorded by the thermocouple in the furnace.69The thermocouple output was then sampled by the computer through thethermocouple measuring circuit. A feedback algorithm was then used to generate anew oven control value using the temperature value acquired such that changes inthe oven control value minimized differences between the actual temperature andthe required temperature. Obviously, this required temperature varied with timewhen a temperature ramp is used. This procedure was repeated continuously every4 seconds throughout the experiment.2.2.3.1 Feedback control using a difference algorithmMethods used in feedback control systems vary greatly in complexity. In theinitial work, the following feedback controller was used.= CINT(OCV + Kg ts(Treq -1)) (Equation 16)where OCVnew and OCVcur are the new and current oven control values(0255), K is a gain constant, is the time between measurements, 1req is therequired temperature (K) and T is the measured temperature (K). The BASICfunction ‘CINT’ makes OCVnew an integer with fractional values less than and equalto 0.5 rounded downwards and values greater than 0.5 rounded upwards. Inaddition, the value of OCVnew is restricted to a range of 0 to 255, the 8-bit inputrange of the controller.This ‘difference’ controller is based on a very simple method: the controlleroutput is related to the current oven control value and the difference between themeasured temperature and the required temperature (i.e., the temperature error).No allowance is made for the rate of increase or decrease of temperature. Inpreliminary experiments the effect of constant K was established, with the followingvalues being used; 0.1, 2.5, 20 and 50. It was decided to aim to increase the70temperature linearly from room temperature to 773 K (500 °C) at a rate ofapproximately 15 K/mm and then seek to keep the temperature constant at therequired temperature for 150 mm. In this way, it was possible to assess thesuitability of the difference algorithm for both isothermal and non-isothermalconditions.The heating rate was set to a higher value than normal (e.g., 1-10 K/mm).Differences in the performance of the algorithm when using the various Kg valueswould be more apparent than at a lower heating rate. Figure 12 shows typicalresults obtained. For all experiments, there was an excellent degree ofreproducibility and therefore, for clarity, only one trial is shown for each value Of Kg.When Kg = 0.1, the algorithm responded too slowly to changes in the oventemperature. This was evident from the high amplitude of the temperatureoscillations. For the other values of Kg, a better response of the algorithm to thechanging temperatures concerning overshoot correction and undershoot correctionwas seen. There was still a noticeable oscillation present within the observedtemperature profile. It was decided to consult manufacturers of thermal analysisinstrumentation to find possible solutions to this problem.2.2.3.2 Feedback control using a PID controllerAfter discussion with Mr. B. Crowe of TA Instruments (TA Instruments,Wilmington, DE), a more involved method based on a Proportional IntegralDerivative (PID) controller was employed [180]. This algorithm may be regarded asthe combination of proportional, integral and derivative terms. The proportionalterm is similar to the previously used ‘difference’ algorithm in that a linearrelationship exists between the controller output and the error.718007006000500EH-4003008007006000500EH-400300Figure 12. Typical heating profiles as a function of the gain constant Kg.In this case, the error is expressed as a fraction of the operating temperaturerange. The contribution due to the proportional term may be expressed as:PP= Kp.TTrT+p0 (Equation 17)max mmwhere PP is the controller output due to the proportional term, K is theproportional gain constant, set by the user, and P0 is the controller output when themeasured temperature is equal to the required temperature. The maximum andminimum allowed operating temperatures of the furnace are Tmax andrespectively.0 3600 7200 10800 0 3600 7200 10800Time (seconds) Time (seconds)72The proportional term (PP) can be used alone to program a furnace.However due to the inherent offset problem associated with proportionalcontrollers [180], additional terms are generally required. The integral term is usedto help reduce the offset error. In practice, it depends on the history of the error andmay be expressed mathematically as:PIK dt+P0 (Equation 18)max mm0where P1 is the controller output due to the integral term and K is the integralgain constant.The derivative term is used to reduce the overshoot and oscillation of thecontroller response. It also helps to compensate for the natural thermal lag of thefurnace. This is the time it takes for the furnace temperature to respond to a changein input energy. The derivative term derives a controller output that depends on theinstantaneous rate of change of the error, i.e.,PD=K.(Tre,T(Equation 19)max mmwhere PD is the controller output due to the derivative term, Kd is thederivative gain constant and the derivative in the equation measures the rate ofchange of error.Overall, the controller operates as a combination of the three terms(i.e., OCVnew = PP + P1 + PD) though the P0 terms are not included. When using acomputer based control system, as proposed here, the overall controller algorithmmay be expressed as shown in Eqns. 21 to 26. For the purposes of clarity, the termgiven in Eqns. 17 to 19 to express the difference between the required and73measured temperatures as a function of the operating temperature range wasdenoted as the relative error in temperature (Er), i.e.,rep -Er= T T (Equation 20)\. max miii)dEr Er - Er..rjld (Equation 21)Er.jjld Er (Equation 22)SCMSU?Vf+Er (Equation 23)PIKp K1• SUM (Equation 24)dErPD=K •Kd —i-— (Equation 25)Er+FI+PD) R (Equation 26)where Er..old is the previous error, Er is the current error, SUM is the runningsum of errors, P01 is the overall controller output, andR0 is the maximum output(Le., 255).It was decided to employ only an isothermal or a non-isothermal heatingprofile for the experiments as the coefficients for this controller become moredifficult to select if the two types of heating profiles are combined during oneexperiment. The literature for the thermal analysis of polymers contains a majorityof non-isothermal studies and so it was pertinent to continue mostly in the directionof using a non-isothermal heating profile.In preliminary experiments using this PID controller, the heating rate was setto 5 K per minute up to a final temperature of 873 K. Initial studies indicated thatchanging the value of the constant K affected the temperature profile obtainedmore than changing the constants K and Kd. To minimize the number ofexperiments required to obtain best values for these, the values of the constants Kand Kd were set to approximately zero (i.e., 0.01). This allowed a rapid assessmentD0S.1)0FD0ci)0Eci)I—Figure 13. Typical heating profiles as a function of the proportional gain constant(ICr).74of the effect of changing the constant K. The following values ofK were used; 20,40, 60, and 80. Figure 13 indicates typical results obtained. There was a highdegree of reproducibility when using the same K value. Therefore only oneexperiment for each value ofK is shown.9008007006005004003009008007006005004003000 1800 3600 5400Time (seconds)0 1800 3600 5400Time (seconds)75When K was set to 20 the measured temperature was less than the requiredtemperature during the majority of the experiment, especially at the highertemperatures. For the other values of there was a good overall agreementbetween the required temperature and the measured temperature throughout theexperiment. The undershoot-overshoot problem still remained significant attemperatures lower than 450 K. In all experiments, the difference between themeasured and the required temperature became less as the temperature wasincreased. To determine the optimal value ofK from those tested, a calculation ofthe weighted sum of the squares of the error (WSSE) was made, i.e.,WSSE = TTreq)2 (Equation 27)reqi=0The number of points N may be different for each experiment and thereforeWSSEthe average value was calculated, i.e., NThe acoustically emissive reactions studied were found at temperatureshigher than 550 K and thus it was important to select the value ofK that led to atemperature profile with highest agreement at the temperature range where theacoustic emission was found. Eqn. 27 shows that the calculation was weighted witha preference towards accuracy at higher temperatures by using a weighting factor of1TreqTable 4 shows the WSSE/N values for the three values that were to beconsidered for use. When = 40 the minimum values were found. As a result ofthe good linearity above 500 K this value of was used in the remainder of the TS76studies. K and Kd were not optimized and a value of 0.01 was used for thesethroughout.K WSSEN440 4.5, 5.5, 7.360 5.7, 7.8, 9.180 10,11,13Table 4. Average weighted sum of the squares of the error as a function of K.2.2.4 Temperature calibrationCalibration of the furnace was undertaken to ensure that the temperatureprofile experienced by the sample in the waveguide was comparable to thatmeasured by the thermocouple. Two methods of temperature calibration arecommonly found in thermogravimetry [181]. One is based on use of ferromagneticmaterials that exhibit known Curie point transitions. Small pieces of theferromagnetic material are placed in the sample pan. A small permanent magnet isthen placed either above or below the sample pan. The vertical component of themagnetic field causes the balance to read either more or less depending on theposition of the magnet. If the magnet is placed below the sample pan, the magneticforce on the sample acts as an equivalent magnetic mass on the balance beam toincrease the apparent sample mass. On heating, the magnetic domains in theferromagnetic materials become disorientated and transform to the paramagneticstate at characteristic temperatures (i.e., the Curie Point). At these temperaturesthe thermogravimetry curve will show an apparent mass loss due to the loss ofmagnetic mass. These transitions are reversible and thus the procedure may be77repeated many times. The main disadvantage with this technique is that it is onlyreally suited for thermogravimetry instruments that are constructed with amicrofurnace inside the furnace tube, since a small horseshoe magnet may beplaced around the narrow microfurnace tube [181]. It is more difficult to do with abigger furnace, especially of the type used here as we need a bigger magnet andthe turnings of the heating element affect the magnetic field. For this reason thismethod was not practical here.The other method involves the use of a fusible link of a metal with a knownmelting point [181]. The link is heated until it melts and drops away from the samplepan, thus a loss of mass is experienced. For the apparatus developed, this is notreally possible. The height of the sample compartment of the waveguide used is60 mm; the metal link would not be at the same position in the furnace as thesample. Thus any calibration would still be incorrect. Furthermore, it is not reallyadvisable for molten pieces of metal to be falling out of the furnace onto the surfaceunderneath.It wasdecided to use a more approximate method based on comparing thetemperatures of two thermocouples; one functioning as the furnace thermocoupleand one touching the bottom of the sample compartment of the waveguide. In theinitial experiments, an analog pyrometer probe (Model 8396, Cole-ParmerInstrument Company, Chicago, IL) was set up such that the tip of the pyrometerprobe was just touching the bottom of the sample compartment of the waveguide inthe furnace. The heating rate for this experiment was set to 5 K per minute up to afinal temperature of 773 K. Results indicated that there was a very largediscrepancy in the temperatures measured with the thermocouple measuredtemperature being higher than the pyrometer measured temperature.78A temperature discrepancy of Ca. 200 K was observed at the final temperature.After checking that this discrepancy was not caused by the pyrometer, changeswere made to the apparatus to reduce this error.A hole of Ca. 10 mm diameter, was drilled through the quartz tube lining thefurnace at the position where the thermocouple touched the tube. The positioningof the thermocouple was adjusted so that the tip of the thermocouple was nowcentered inside the hollow quartz tube. The calibration procedure described wasthen repeated using the newly constructed waveguides (“Type C”, Fig. 16). Duringthis procedure the waveguide apparatus was positioned in the furnace so that thetip of the furnace thermocouple was touching the side of the sample compartment ofthe waveguide. Figure 14 shows that there was a better agreement between thetwo sets of temperature measurements though there was still a small discrepancy ofCa. 25 K in the final temperatures reached.Use of the pyrometer probe presented problems. The meter could only beread to an accuracy of ± 5°C and there was no easy way to automate the reading ofthe temperature values. In addition, the meter which displayed the temperaturereadings had a slow response. This would prove to be a problem during theplanned studies involving a higher heating rate of 10 K/mm. As a result, a second‘Type K’ thermocouple in the form of a probe was constructed by the UBC Dept. ofChemistry Electrical Shop to replace the pyrometer. This thermocouple wasinterfaced to the computer in a similar manner to that described previously exceptthat the output from the associated thermocouple amplifier circuit was connected tothe third analog-to-digital channel (AD#2) of the data acquisition card. It was nowpossible to read the sample temperature and the furnace temperaturesimultaneously.79800 I7006000) //D •1o .10) r:iE00 -——Furnace temperature------Pyrometer temperature300 —Reference temperature -200 I I0 1800 3600 5400Time (seconds)Figure 14. A comparison of recorded temperatures using the furnace and samplethermocouples.The effect of heating rate on the relationship between the furnace andsample temperatures was established. Heating rates of 1, 3, 5, 7, and 10 K/mm upto a final temperature of 873 K (600 °C) were used. Figure 15 indicates the resultsobtained for heating rates of 1, 5 and 10 K/mm. For all heating rates, an excellentagreement was found between the reference temperature and the furnacetemperature. This confirmed that the PID algorithm is successful in controlling theheating rate of the oven. The sample temperature lagged behind the furnacetemperature with the greatest lag being found at lower temperatures and at thehigher heating rates. This thermal lag was equivalent to a maximum of Ca. 50 K inthe temperature region of interest, i.e., above 550 K.9008007006005004003009008007006005004003009008007006005004003000 1000 2000 3000 4000 5000 6000 70000 500 1000 1500 2000 2500 3000 3500Time (seconds)Furnace temperature Sample temperatureReference temperature.Figure 15. A comparison of the furnace, sample and reference temperatures forheating rates of 1, 5 and 10 K/mm. For a heating rate of 1 K/mm, thefurnace and reference temperatures overlap for the majority of theplot.800 6000 12000 18000 24000 30000ci)ci)Sci)As a result of the discrepancy between the furnace and sampletemperatures, an empirical function was used to convert the furnace temperaturesrecorded during the experiment to the sample temperatures to get a more accurate81representation of the temperatures at which the acoustic events were beingdetected. Table 5 displays the empirical functions used for each heating rate.Heatinci rate Empirical Function Fitted parameters(K/mm) (y- sample temperature)(x- furnace temperature)(a, b, c, d, e, f- coefficients)c dlnx e a=1O10,b=O.61,— a + bx +—+—4- —xJ x2 c-8.7x107,d5 9e = -1.9x109r2 = 0.99983 a680,b0.70,c dlnx ey = a + bx + +j_cz5.5x107,d3.6e = -1.2x109r2 = 0.99995 C dlnx e a=1040,b= 0.56,a + bx +—+—4-—x[ x2 X c-8.4x107,d5 6e -1.8x109r2 0.9999‘- exy a + bx + cx’s[x +dInx.jx + — a -42000, b = -2472,ilixc=2.7, d=-5800e=25500r2 = 0.9999e10 ya+bx+cx2J+dx3 +— a=-2220,b=3.4,‘J C = -8.3x10,d= 1.5x106e = 27200r2 0.9998Table 5. Empirical functions used for the conversion of furnace temperature tosample temperature.82To measure of the accuracy of the conversion from furnace temperature tosample temperature, the melting temperatures of some compounds weredetermined using the furnace apparatus. Please see section 3.2 for sources of thechemicals used.McGhie had discussed the use of tin metal for the calibration of furnacesusing a fusible link method [182]. The melting point was reported as 505 K. Asample of tin metal in the form of filings (1 g) was placed in a waveguide set upinside the furnace. Samples were illuminated with an optic fibre light source(#1177, Cambridge Instruments Inc., Buffalo, NY) placed at the bottom of thefurnace. The sample was then heated at a rate of 5 K/mm to a temperature of600 K. The sample was monitored using a strategically placed mirror to visuallydetect the melting of the tin metal. During the three experiments carried out usingthis configuration, no melting was detected and closer observation of the sample oncooling indicated that no melting had occurred. It was apparent however that aslight change in color of the tin metal had taken place. This was attributed tooxidation of the surface of the metal. This conclusion is confirmed by Harrison whoreported that surface oxidation of tin occurs at temperatures greater than 473 Kresulting in significant tarnishing of the surface [183]. Analogous studies to detectthe melting of the tin metal using a melting point apparatus (Gallenkamp meltingpoint apparatus MF-370, #889339, England) gave similar results over the sametemperature range.Although the previous procedure had not given the intended results, theproblem was obviously due to the material chosen. From crystallization studiesusing benzoic acid and PET, it was known that the melting of both compounds couldbe visually detected. The procedure used for the determination of the melting pointof tin using the furnace apparatus was repeated for benzoic acid and PET. The83furnace temperature at which the melting was detected was converted to the sampletemperature using the appropriate function in Table 5. Repeated measurementsusing the melting point apparatus were also undertaken. Table 6 indicates theresults together with reported literature values. The literature value given for PETwas determined using a DSC technique for the PET sample at Kodak (EastmanChemical Company, Kingsport, TN).Compound Literature Furnace Melting pointvalue (K) apparatus (K) apparatus (K)Benzoic acid 395.28 [184] 398, 400, 395 394, 390, 391PET 520.47 535, 530, 538 519, 515, 514Table 6. Melting points determined for benzoic acid and PET using the furnaceapparatus. Error in readings ± 5 K.The melting temperatures determined using the furnace apparatus wereslightly high in comparison with those determined using the melting point apparatusand the literature values but overall there was an excellent agreement among thethree sets of values.This mode of temperature calibration could be used to assess the sampletemperature from the furnace temperature. However, this temperature calibrationwould not account for differences in the sample temperature due to the thermalconductivity of the sample or any changes in the sample temperature as a result ofan exothermic or endothermic event occurring in the sample. In DTA and DSCtechniques, this problem is overcome by inserting an additional thermocouple intothe sample itself [185]. This method leads to the measurement of the actualtemperature profile experienced by the sample However, this solution is often notvalid for thermogravimetry as the thermocouple leads can interfere with the balance84mechanism [186]. Furthermore, Clark reported that the insertion of a thermocoupleinto the sample causes severe mechanical damping during thermosonimetry [149].For any experiments where a more accurate temperature measurement wasrequired, it was proposed that another experiment be carried out with athermocouple placed in the sample and without acoustic emission measurement.2.2.5 Optimized furnace apparatusPrevious sections have detailed the development of a heat source suitablefor thermosonimetry. During the majority of the experimental studies discussed inthis dissertation, the control of the furnace using the PID controller had beenoptimized and the temperature calibration procedures and modifications wereutilized.2.3 Waveguide apparatusIn TS, a piezoelectric transducer is used to detect the acoustic waves. Thetransducers only operate over a limited temperature range. Therefore it isnecessary to use a sample holder that enables the sample to be heated, whilst thetransducer is kept external to the heat, but still in mechanical contact with thesample. In the studies reported in the literature, a quartz waveguide is used.Although using a quartz sample holder may promote additional thermal reactions,quartz provides a number of advantages for thermosonimetry. In particular, quartzhas a high sound transmission quality, is thermally stable over a wide temperatureand is able to be shaped easily thus enabling a waveguide to be made from a singlepiece of material. Figure 16 indicates the dimensions of the three waveguidesused. All these designs were used during the development of the TS apparatus.However, in the final apparatus, only ‘Type C was used.85(b) Diameter of holder(e) Holder thickness (2 mm)(a) Depth of holder‘Type A’(a) = 45 mm(b) = 40 mm(c) 140 mm‘Type B’(a) = 60 mm(b) = 20 mm(c)=lOOmm (c) Heightofstem‘Type C’(a) = 60 mm(b) = 20 mm(c) = 260 mm(d) Diameter of stem (10 mm)Figure 16. The waveguides used during the thermosonimetry experiments.Sterling had reported acoustic emission studies using waveguides of differentdimensions and materials [187], and had found the propagation of acoustic wavesthrough the different waveguides to be different. Therefore no attempts were madehere to directly compare experimental results using the three designs ofwaveguides.All experiments used a broad bandwidth piezoelectric-based acousticemission transducer (Model 8312, serial number 1381596; Bruel & Kjaer Canada,Richmond, BC), which incorporated an internal 34 dB preamplifier. Figure 17 showsthe holder used to ensure repeatable positioning of the waveguide on the activesurface of the transducer. The mass of the waveguide holder was made as low aspossible, as it was to be positioned on top of a balance during experiments.86Therefore the cylinder was constructed out of styrofoam, though for a few of theearly experiments an aluminum cylinder was used. Repeated use of the styrofoamcylinder led to an increase in the diameter of the waveguide hole and the possibilityof spurious signals due to movement of the waveguide. Therefore, the cylinder wasreplaced at regular intervals.FF—I10- F-IF -I40F —I70Figure 17. The holder used to support the waveguide and transducer during thethermosonimetry experiments. All measurements are in mm.The waveguide holder was situated on top of the balance, as shown inFigure 8. To protect the balance and transducer from the effects of radiated heat, a25-I-20-r501CylinderTransducerAluminum support4087stainless steel plate was used as a heat shield. The plate, which had dimensions of150 mm by 100 mm with a hole of diameter 15 mm in the centre was held, usingclamps, just below the bottom of the furnace. The quartz waveguide was slottedthrough the holes in the plate and the waveguide holder so that the base of thewaveguide was in contact with the active area of the transducer. The waveguidewas also positioned so that it did not touch the heat shield. Good acoustictransmission. between the waveguide and the transducer was ensured by placing asmear of Apiezon grease (Type L or Type H, Apiezon Products, London, UK) on theactive area of the transducer prior to contact with the waveguide.The transducer converts the acoustic pressure waves to an electrical signal.This signal was fed to the ‘pre-amp’ input of a wideband conditioning amplifier(Model 2638, serial #1283218; Bruel & Kjaer Canada, Richmond, BC) where it wasfurther amplified and filtered. The input impedance was set to 50 2. Theamplification setting used was variable with a user selectable range of 0 dB to60 dB. Typically a gain of 30 dB was used. A frequency range of 50 kHz to 2 MHzwas used in all experiments.2.4 Balance apparatusAn electronic balance (Model FX-400, A&D Company Ltd, A&D EngineeringInc., Milpitas, CA) was connected to the PC/AT control computer via acommunication port (COMI) using a RS232 interface. Communication routinesused were based on those indicated in the manufacturer’s instruction manual [188]and are listed in Appendix 2. These routines were merged with the feedback looproutines to enable simultaneous data collection from the balance and the control ofthe furnace.88The balance was positioned beneath the furnace for all experiments> eventhose where it was not being used to record data. In this way> more consistency inthe positioning of the sample holder was obtained. To minimize drafts> the plasticbreeze break supplied with the balance was used when mass measurements werebeing made. To enable this> the transparent plastic window on the top of the breezebreak was removed so that the sample holder could protrude above the breezebreak.Preliminary experiments indicated that it was not possible to acquire acousticemission and mass data simultaneously with adequate accuracy. The balancemeasurements indicated a large degree of instability> which were caused by theacoustic cable connected to the acoustic transducer. This form of mechanicalproblem had been reported in the thermal analysis literature, though in theinstances given> the instability is caused by thermocouple leads when athermocouple is placed inside the sample compartment [186]. Initially> this problemwas tackled by stripping the outside shielding of the acoustic cable and removingthe casing of the connector. This helped to reduce the mass of the cable andincrease its flexibility. Experiments indicated that this did help to improve thestability of the balance measurements. The improvement was still not enough toenable accurate simultaneous TS and TG measurements to be made when usinglow sample masses (e.g., less than I g).An lR transmitter/receiver apparatus which involved no physical attachment>was then developed as a potential solution to the problem. This is discussed insection 2.7.892.5 Data acquisition techniquesTwo methods were utilized to acquire acoustic emission data during theinvestigations: continuous acoustic level monitoring and individual signal capture.2.5.1 Continuous acoustic level monitoringContinuous acoustic level monitoring was achieved by sampling the a.c.output from the conditioning amplifier using a ‘true RMS’ voltmeter (Model 8922A;John Fluke Manufacturing Co., Inc., Everett, WA, USA). This had a 10 MHzbandwidth, which was much more than the 2 MHz available through the amplifier.The output voltage from the linear output of the voltmeter was linear with respect tothe acoustic RMS magnitude. This output was amplified with a gain of four using anoperational amplifier circuit constructed in this laboratory. A schematic of theoperational amplifier circuit is shown in Figure C of Appendix 1. The output from theoperational amplifier circuit was connected to the first analog-to-digital channel(A/D#0) of the data acquisition card installed in the IBM PC/AT compatiblecomputer. Software to control the data acquisition was written by Dr. 0. Lee.2.5.2 Individual acoustic signal capture apparatusIndividual signal capture was achieved by sampling the a.c. output of theconditioning amplifier using a digital storage oscilloscope (Model 2430A; Tektronix,Beaverton, OR). In all experiments, a record length of 1024 points was used.Digitization was with 8-bit precision, and the conditioning amplifier gain wasadjusted to maximize the use of the resolution provided. Experiments used atimebase of 20 iJS per 50 point division (equivalent to a 2.5 MHz samplingfrequency), this being approximately twice the upper response frequency limit of thetransducer. Generally, an amplitude scale setting of 200 mV per division,equivalent to Ca. ± 1.0 V vertical full scale, was used. An acquisition trigger level of90100 mV was set, except where stated otherwise, to eliminate collection ofbackground noise. The oscilloscope communicated with a 12 MHz IBM PC/ATcompatible computer (Type Doppler System II, Doppler Computers, Vancouver, BC)across a IEEE-488 instrumentation interface. Control of the oscilloscope was via anindustry-standard adapter card (Model PC-HA; National Instruments, Austin, TX)and IEEE-488 driver software (National Instruments).For all experiments, mains-borne interference was minimized by powering allunits through a pair of 75 dB high frequency mains noise filters and suppressors(ISOBAR Model IBAR-4-GS; Trippe Manufacturing, Chicago, IL).2.6 Transducer and waveguide calibrationThe acoustic emission transducer used throughout the experiments did nothave a flat response for the frequency range over which it is calibrated (100 kHz to1 MHz). The transducers are calibrated by the manufacturer, based on a free-fieldRayleigh-wave reciprocity method [189]. It was not possible to accuratelyreproduce the method of calibration used by the manufacturer so a comparativemethod was used.2.6.1 Calibration procedureA schematic of the apparatus used for the calibration procedures is shown inFigure 18. The transducer to be calibrated was placed on a piece of foam matting(25 mm depth) to reduce unwanted laboratory vibrations. This transducer was heldin contact with another transducer of known flat response (Model FAC500, serial#132591, AET Corp., Sacramento, CA) using clamps, with a smear of Apiezongrease (Type L) being placed between the two surfaces to ensure good acoustictransmission. The electrical signal from the transducer being calibrated was fed tothe conditioning amplifier described earlier. The a.c. output from the amplifier was91sampled by the ‘true RMS voltmeter. For the purposes of thiscalibration the valuesdisplayed by the voltmeter were recorded manually. The output from the amplifierwas also sampled by the oscilloscope to enable a visual display of the waveformdetected.Figure 18.Felt padA schematic of the apparatus used to calibrate the transducer.A function generator board (Model PCIP-SST, Keithley MetraByte Corp.,Taunton, MA), installed in the PC/AT computer, was connected to the AETtransducer, which acted as the sound source. A sine wave signal was generatedusing software provided by the manufacturer. During the calibration procedure, thefrequency of this signal was varied from 10 kHz to 1.25 MHz inintervals of 10 kHz.The frequency generated was checked for accuracy at 100 kHzand at 500 kHzusing a routine provided by themanufacturer. In the initial transducer calibrationthe amplitude of the generated signal was set to 500 mV and anamplification of0 dB was used. In subsequent studies, the amplitude of the generated signal was(\jSourceTransducerTransducerbeingcalibratedif92set to 50 mV, 20 mV and 10 mV. For these studies, the amplification was set to30 dB.Calibration of the waveguide apparatus was also attempted to enable acomparison of the two arrangements. A waveguide, of the same dimensions as the‘Type B’ waveguide was used. However, the walls of the sample compartment wereremoved so that the transducer providing the signal was in direct contact with thebottom of the sample compartment. The waveguide holder, described insection 2.3, was used to support the waveguide. Apiezon grease (Type L) wasused to provide additional acoustic coupling between the surfaces in contact. Thecontact was adjusted to provide the maximum RMS reading when a sine wave of100 kHz was being generated. The procedure described above for calibrating thetransducer was repeated using the same signal frequencies, amplitudes andamplifications.2.6.2 Calibration resultsFigure 19 shows a reproduction of the calibration supplied with thetransducer used here. The response of the transducer is broadband (rather thanresonance) but cannot be considered independent of frequency. The ordinate scaleshows the decibels of amplification required to give a signal of I V per m/s.Therefore, those frequencies with the lowest values are the most sensitivefrequencies (or resonant frequencies) of the transducer. Thus the main resonancefrequencies of the transducer used here were 500 kHz and from 800 kHz to 1 MHz.Typical calibration results using a direct connection between the twotransducers and using a waveguide between the two transducers are shown inFigure 20. Calibration data are shown for signal amplitudes of 500 mV and 20 mV.93858075E7060551000Figure 19. A reproduction of the calibration supplied for the transducer usedthroughout the experimental work.At the higher signal amplitudes, i.e., 500 and 50 mV, the main differencesbetween the results obtained using a direct connection and using the waveguidewere related to the intensity of the resonance peak at 800 kHz. As seen frompanels (A) and (B) in the Figure 20, the intensity of the resonance frequency wasdouble when a direct connection was used. The intensities of the other frequencieswere similar though it may be argued that the results showed more of a bandstructure in the top panel as opposed to more of a ‘spiked’ appearance in panel (B).The variation in response as a function of frequency was thus smaller when a directconnection was used.At the lower signal amplitudes, i.e., 20 and 10 mV, the resonance peak at800 kHz was more prominent with an intensity approximately seven times greaterobserved when using a direct connection. The intensities of other frequencies areapproximately double when using a direct connection. When the signal amplitude0 200 400 600 800Frequency (kHz)94was 10 mV, there was evidence that many of the intensities were at the backgroundlevel as the signals were difficult to detect using the oscilloscope.Cl)01 206000 500Frequency (kHz)1000Figure 20. A comparison of the transducer calibrations obtained using a directconnection between the source and the transducer and using awaveguide between the source and the transducer.(A) 500 mV input voltage, direct connection.(B) 500 mV input voltage, waveguide connection.(C) 20 mV input voltage, direct connection.(D) 20 mV input voltage, waveguide connection.40030020010002001 000L0 500 1000Frequency (kHz)800600400200(D)‘95Surprisingly, the manufacturer’s calibration (Fig. 19) did not indicate theintense resonance peak at around 800 kHz. Throughout the work reported here,this resonance was always found.Regression analysis was used to investigate the correlation between the setsof data (Table 7). The intensities recorded for each frequency formed the x, y pairs.Coefficient of Slopedetermination (r2)500 mV 0.618 1.49±0.0950 mV 0.572 2.7 ± 0.220 mV 0.543 2.5 ± 0.410 mV 0.645 3.2 ± 0.2Table 7. Regression analysis of the calibration results using the directconnection method and using a waveguide. X-values were thewaveguide data and y-values were the direct connection data. Thereported errors refer to the 95 % confidence limits.There was an acceptable agreement between the results obtained using thetwo methods. The slopes indicated that on average, using a direct connection ledto approximately twice the intensity being recorded, i.e., use of a waveguideresulted in attenuation by a factor of 2 to 3.In conclusion, these results, together with those shown previously inFigure 20, indicate that the use of a waveguide was feasible. Although, differenceswere found between the two methods, these results also suggested that there weremany similarities between the results obtained.962.7 Development of TGSAs the balance instability problems were being caused by the presence of acable connected to the acoustic transducer, the obvious solution to the problem wasto remove the cable, It was therefore necessary to find another way of transmittingthe electrical signal from the transducer to the wideband conditioning amplifier thatinvolved no mechanical contact between the amplifier and the transducer. It wasdecided to investigate the possibility of using a battery-powered IRtransmitter/receiver system to solve this problem.2.7.1 Design and construction of transmitter and receiver circuitsWhen considering the design of a suitable IR transmitter/receiver system, anumber of important requirements were necessary;1. Acoustic emission signals are transient signals, lasting generally not longerthan a few milliseconds. As a result, circuits used to transfer the electricalsignal from the transducer to the amplifier must be able to respond fastenough to these transients.2. The amplitudes of the electrical signals from the transducer are typically lessthan 1 V. The circuits used to transfer the electrical signal must be sensitiveenough to respond to small changes in the signal amplitude.3. The upper response frequency limit of the transducer was approximately1 MHz. Therefore the sampling frequency of the circuits used to transfer theelectrical signal from the transducer to the amplifier should be greater than2 MHz to ensure that no aliasing occurs. Aliasing results when the samplingrate is too low and leads to a recorded signal possessing frequencies thatare not present in the actual signal [190].974. The transmitter circuit was to be positioned on the balance and thus it wasimportant that the circuit be developed with this in mind; the mass of thecircuit had to be minimized, the lifetime of the power supply (batteries)maximized, and a stable geometry chosen which was not subject to vibrationfrom local air currents.The IR transmitter/receiver circuit was designed and constructed inconjunction with Dr. L.E. Bowman. Schematics of the transmitter and receivercircuits are shown in Figures E and F of Appendix 1. The transmitter section wasdesigned in a manner which allowed its easy connection to the acoustic transducerwhilst positioned on the balance.2.7.2 Evaluation of circuitSeveral methods were used to evaluate the circuit and to assess its potentialfor use. The circuit was initially calibrated using artificially generated signals.Later, chemical systems known to give acoustic emission were used. Details of thechemical studies are given later in section 3.7.27.2.1 Calibration of IR transmitterlreceiver systemThe IR transmitter/receiver system was initially evaluated using thecalibration method discussed in section 2.6.1, i.e., using a flat response transducerand the “True RMS” meter. The frequencies of the signals produced by the functiongenerator were varied from 100 kHz to 820 kHz in intervals of 10 kHz and theamplitude was set to I V. Preliminary studies had suggested that the response ofthe IR transmitter/receiver system was negligible above 800 kHz. Therefore, theupper frequency limit was set lower than for previous studies. For all experiments,the amplifier gain was set to 10 dB.98Initially, this calibration process was undertaken with a cable connected fromthe amplifier to the acoustic transducer. Then the IR transmitter/receiver circuit wasused. The distance between the transmitter and receiver was less than 5 mm.When changing the connections, extreme care was taken to ensure that the contactbetween the source transducer and the calibration transducer was not altered.When using the lR circuit, the receiver part of the circuit was connected to the directinput of the wideband conditioning amplifier. The input impedance of the amplifierwas switched to 2 M2. This method assessed the sensitivity of the IR systemrelative to a mechanical connection between the acoustic transducer and theamplifier.The frequency response of the IR transmitter/receiver system was alsoevaluated with a view to using this system with the individual signal captureapparatus discussed in section 2.5.2. The procedure described above wasrepeated, except that the a.c. output of the conditioning amplifier was sampled byan oscilloscope card (Model PCIP-SCOPE, Keithley MetraByte Corp., Taunton, MA)installed in the same PC as the function generator board. This calibration processwas automated using software developed in this laboratory by Mr. l.H. Brock. Eachsignal acquired was checked before storage to ensure that it was not over-range.Over-range signals were defined here as signals where more than 50 % of the 1024points acquired had values at the lower or upper digitization limits. Results werecalculated based on the average of four signals acquired at each frequency.2.7.2.2 Calibration results for IR transmitterlreceiver circuitThe RMS calibration data for a direct connection and using the IR circuit areshown in Figure 21. When comparing the intensities at individual frequencies,many differences are observed though the maximum intensities observed are99comparable and indicate a similar sensitivity to the signals. At low frequencies, theintensities were higher using the IR transmitter/receiver circuit but the response ofthis circuit greatly diminished at frequencies greater than 400 kHz. The responseusing a direct connection between the amplifier and the transducer was lessdependent on frequency though several resonances were apparent.1000900800700600E (A)50040030020010000 200 400 600 800700600500400U,3002001000 I I I0 200 800Figure 21. Transducer calibrations using (A) a direct connection between theamplifier and the transducer, and (B) the IR transmitter/receiver circuit.The frequency calibration results are shown in Figure 22. At each inputfrequency, the FFT of the time-domain signal acquired was calculated in order toget the resulting frequency spectrum. Although, the frequency interval used was(B)400 600Frequency (kHz)10010 kHz, the figure shows the resulting frequency spectra every 50 kHz for reasonsof clarity. The frequency spectra not shown followed the same trends as observedin the figure. Using a direct connection led to a very clean frequency spectrum; theinput frequency was detected at an intensity level many times the other frequencycomponents in the spectrum. The relative intensities of the most intensefrequencies gave the same trends as shown above for the RMS data, i.e., no realdifference in the intensities at high and low frequencies, except for the presence ofspecific resonances. The results using the IR transmitter/receiver circuit were verydifferent. Again, the highest intensities were at low frequencies (<400 kHz).However, the detected frequency spectra are very noisy with many intensefrequency components at other than the input frequency. Closer examination of theresults concluded that the majority of the ‘extra’ frequencies were harmonicsindicating that the IR transmitter/receiver is unsuitable for frequency related studies.This was further enforced by regression analysis of the results obtained in a similarmanner to that described previously. The coefficient of determination (r2) was0.0132 with a slope of 0.08 ± 0.15 (95 % confidence limit).2.8 Overall ConclusionsThis work has detailed the development of a thermogravisonimetry apparatusfor use in thermo-oxidative decomposition studies. Although the lRtransmitter/receiver circuit was unsuitable for frequency characterization of thesignals, continuous monitoring of the acoustic emission was viable. This facilitatedcombined TS/TG experiments as had been hoped. Almost all aspects of theapparatus underwent revisions during this development process until a suitableapparatus was achieved. A schematic of the developed thermogravisonimetryapparatus was shown in Figure 5.Figure 22.>--4--,(1)ci,>Cci,>-U,Cci)-4-,Cci)>-4-,Cci)Three-dimensional plots showing the frequency response of (A) usinga direct connection between the amplifier and the transducer, and(B) using the IR transmitter/receiver circuit. Every fifth spectrum isshown.1011 50010005001 000750500750600450 “—300150Q)Q250Detectfreq.(kI-i)750600450s’—300 /150q)102CHAPTER 3. EXPERIMENTAL STUDIES3.1 IntroductionThe area of investigation covered by this dissertation, i.e., acoustic emissionfrom polymers, is largely unreported in the scientific literature. Thus the initialexperiments aimed to survey a wide scope of the phenomenon in the general fieldof polymer chemistry. After these initial experiments, further experimentsconcentrating on the thermal decomposition of polymers were undertaken. Thesestudies are detailed from section 3.5 onwards.As the theoretical origins of the AE phenomenon were uncertain the majorityof the experimental investigations were of a comparative nature. It was felt that bycomparing the results of AE studies with those obtained using established analyticaltechniques a greater insight into the sources of the AE should be possible. Theeffect of chemical parameters on the nature of the AE was also assessed with thisaim in mind.3.2 Chemicals3.2.1 PolymersSamples of poly(vinyl chloride) (PVC) were supplied by Esso (Esso ResearchCentre, Sarnia, ON) and were white resins containing no additives or thermalstabilizers. Polypropylene (PP) was supplied by Hercules (Hercules Canada Inc.,Iberville, PQ) and was in the form of fibres of Ca. 0.1 mm diameter. Pellets of lowdensity polyethylene (LDPE) and high density polyethylene (HDPE) were suppliedby Nova Husky (Nova Husky Research Corp., Calgary, AB). The pellets haddiameters between 1.5 and 5 mm. The poly(ethylene terephthalate) (PET) powderwas supplied by Kodak (Eastman Chemical Company, Kingsport, TN).103Ethylene-vinyl acetate (EVA) copolymer beads were obtained from Aldrich(Cat #s 18,106-4 & 34,050-2, Aldrich Chemical Company Inc., Milwaukee, WI).Vinyl acetate polymer (PVAc) was obtained from the same source (Cat # 18,948-0)3.2.1.1 Chemical dataTable 8 shows the inherent viscosities and viscosity-average molecularmasses (My) of the six PVC resins. The inherent viscosities were determined at thesource using the ASTM-D1243 standard test method [26]. The intrinsic viscositywas then calculated from the inherent viscosity. Use of the Mark-Houwink equation(Eqn. 9) led to an estimation of the viscosity-average molecular mass of thepolymer. Values of the constants K and z used were 0.0163 ml/g and 0.77respectively [25].PVC Inherent Viscosity EstimatedM (g/mol)resin (ASTM D1243)(d hg)#1 0.756 59200#2 0.745 58100#3 1.02 88200#4 0.982 83900#5 0.948 80000#6 1.79 188000Table 8. Inherent viscosity and viscosity-average molecular mass of the PVCsamples.104The particle size distributions of the PVC resins were estimated by vigorouslyshaking a portion of each resin (2 g) for a period of 2 minutes through a series ofsieves (Canadian Standard Sieve Series, W.S. Tyler Company of Canada Ltd.,St. Catherines, ON). The mass of resin retained by each sieve was determined as apercentage of the original sample mass (± 0.5 %) (Table 9). The fifth PVC resin(‘#5’) was more heterogeneous than the other resins. However repeated analysis ofthis sample gave similar results.PVC resin Mesh size Mesh size Mesh size Mesh size0.595 mm 0.354 mm 0.074 mm 0.043 mm#1 99% 1%#2 99% 1%#3 99.5% 0.5%#4 99% 1%#5 2% 65% 33%#6 100%Table 9. Particle size distributions of the PVC resins.A similar method was undertaken to estimate the particle size distribution ofthe PET polymer. Over 99.5 % of the sample was retained by the 0.595 mm sieve,with the rest being retained by the 0.074 mm sieve.Size exclusion chromatography of the PET polymer undertaken at Kodakindicated thatM = 26 632 g/mol andM = 48 035 g/mol.105Table 10 indicates the resin characterization data supplied with the PEsamples.Resin M (glmol) M (glmol) Density (glcc)HDPE 19000 129000 0.946LDPE 4900 37600 0.922Table 10. Resin characterization data for the PE samples.3.2.2 Comparative materialsAmmonium chloride (AnalaR), benzoic acid (AnalaR), copper sulfatepentahydrate (AnalaR), calcium oxalate (AnalaR) and silica gel (6-20 mesh) wereobtained from BDH (BDH Canada Ltd, Vancouver, BC). Anhydrous aluminumchloride was as supplied by Aldrich. Tin metal pieces used for temperaturecalibration of the furnace were obtained from Sigma (Sigma Chemical Company,St. Louis, MO).3.2.3 ReagentsDi-n-octyl phthalate (DOP) used for the plasticization experiments wassupplied by Esso (Esso Research Centre, Sarnia, ON). Styrene used for thepreparation of polystyrene was obtained from BDH. The Azobisisobutyronitrile(AIBN) initiator was supplied by Nova Husky. For the preparation of Nylon-6, 10,decanedioic chloride was obtained from Eastman (Eastman Kodak Ltd, Rochester,NY). 1 ,6-Hexanediamine was as supplied by Aldrich. Sodium hydroxide (AnalaR)was obtained from BDH.1063.3 Thermosonimetry apparatusDetails of the thermosonimetry apparatus were discussed in Chapter 2. Inthe following sections, the apparatus used is stated with the experimental procedureand further reference should be made to the appropriate paragraphs of that chapter.3.4 Preliminary studiesThe areas of polymer chemistry initially assessed included thermaldegradation, phase transitions and polymerization reactions. All experiments wererepeated in triplicate except where otherwise stated.3.4.1 Thermo-oxidative degradation of poly(vinyl chloride)Experiments were undertaken to assess if acoustic emission was producedby the thermal degradation of poly(vinyl chloride) in air. These studies aided in thedevelopment of the thermogravisonimetry apparatus described in Chapter 2. Onlythe individual acoustic signal capture apparatus was used.3.4.1.1 Acoustic emission studies using the ovenA sample of the PVC resin (10 g, ‘#6’), was placed in the waveguide(‘Type A’, Fig. 16). The oven apparatus was set up, as in Fig. 6. An amplifier gainof 40 dB was used. The vertical scale of the oscilloscope was set to 200 mV perdivision and an acquisition trigger level of 100 mV was used. The sample washeated at an average rate of Ca. 5 K per minute to a final temperature of about500 K. Acoustic data were collected throughout the period of heating. The samplewas weighed before and after the experiment so that the total mass loss could befound. In a further experiment, the procedure was repeated except that no samplewas placed in the waveguide. This served as a blank experiment.1073.4.1.2 Initial acoustic emission studies using the furnaceThe use of the oven was discontinued and instead the furnace was used. Asample of the same PVC resin (5 g) was placed in the smaller waveguide(‘Type B’, Fig. 16) and the waveguide apparatus positioned in the centre of thefurnace. For these experiments, an input value of 125 was supplied to the heatercontroller throughout and therefore the heating profile approximated to that shownin Figure 9. This corresponded to the furnace being heated at an average rate ofCa. 8 K per minute up to about 800 K followed by approximately isothermalconditions. The amplifier gain was set to 30 dB. The vertical scale of theoscilloscope was set to 100 mV per division and the acquisition trigger level was setat 110 mV. Acoustic data were acquired throughout the course of the experimentwhich was of 2 hours duration.3.4.1.3 Comparison of acoustic emission from various PVC resinsA series of experiments was undertaken to see if there were discernibledifferences in the AE produced by the thermal oxidation of different PVC resins.Resins used were ‘#2’, ‘#4’ and ‘#5’. These resins, together with “#6”, formed a setof resins having a wide range of molecular masses. For each resin, the proceduregiven above was repeated. A background experiment was also undertaken. Herethe procedure was repeated except that no sample was used. Background signalswere collected by setting the acquisition trigger level at 0 mV (instead of 110 mV).In order to limit the number of signals, an acquisition delay of 12 seconds betweensignals was used.1083.4.1.4 Comparison of AE and mass loss measurementsIt was not possible to accurately record the mass of the sample as a functionof time (and temperature) while the acoustic transducer cable was connected. Forpurposes of comparison, two sets of experiments using portions of the same PVCresin were undertaken, one set in which the acoustic data were acquired and oneset in which the mass of the sample was monitored. In these experiments the sameapparatus was used but the cable was disconnected from the transducer for thoseexperiments where the mass was being recorded. This method was not as reliableas recording both acoustic and mass data simultaneously but it did enable anapproximate comparison to be made between the acoustic data and the progress ofthe reaction as monitored by the change in sample mass.Development of the feedback control using the difference algorithm(section 2.2.3.1), led to these experiments being undertaken with a programmedlinear heating rate. However no isothermal region was used. As the features of athermogram are more apparent at low heating rates [1911, then the programmedheating rate was set to 5 K per minute to a final temperature of 773 K. In theexperiments, a sample mass of 500 mg was used. For experiments measuring themass loss, the sample was introduced into the waveguide once the TS apparatushad been set up. This was accomplished by pouring the sample down a hollowglass tube into the waveguide, which was already situated in the centre of thefurnace. In this way, the balance could be set to measure the mass of the sampleonly. For the acoustic experiments, only the continuous acoustic level monitoringapparatus was used and the amplifier gain was set to 30 dB.1093.4.1.5 Visual studiesIt was not possible to visually monitor the sample undergoing degradationinside the furnace other than by using a mirror placed over the furnace which wasfar from satisfactory. As a result, an experiment which enabled better visualobservation was undertaken.A sample of PVC was thermally degraded to determine the onset temperatureof degradation and the onset temperature of AE. The apparatus used is shown inFigure 23.TransducerAcoustic probeTest tube— Sample—— Heating bathFigure 23. The apparatus used to visually study the thermal degradation of PVC.The heating bath (100 mm x 50 mm PyrexTM crystallizing dish) containedhigh temperature silicone oil (Cat # 17563.3, Aldrich). The oil was stirred to ensurea good uniformity of temperature. The speed of stirring was adjusted to ensure that110acoustic signals were not acquired solely as a result of stirring. Temperature wasmeasured using a thermometer (range 0 to 400°C, Fisher Scientific, Vancouver, BC)placed in the bath. The sample of PVC (250 mg) was placed in a boiling tube(150 mm x 18 mm PyrexTM) and the acoustic probe affixed such that the probe tipwas imbedded in the sample. The quartz acoustic probe used had a stem length of250 mm and a stem diameter of 10 mm. The flat surface in contact with thetransducer had a thickness of 2 mm and a diameter of 25 mm. This design ofwaveguide was necessary as the conventional waveguide apparatus could not beused. A piece of ‘Congo Red’ paper (Coleman and Bell Co., Norwood, OH) fixednear the top of the boiling tube was used to detect the onset of thermal degradationas indicated by the evolution of hydrogen chloride. The boiling tube was thensuspended in the bath, adjacent to the thermometer ensuring that the sample wasbelow the surface of the oil. The bath was then heated slowly at a rate ofCa. 60 K/hr until extensive thermal oxidation of the sample had occurred. Acousticdata were acquired during the heating period.3.4t6 Comparison of the acoustic emission produced by the hydrolysis ofaluminum chloride and the acoustic emission produced during thethermal degradation of PVC in air.The hydrolysis of aluminum chloride produces HCI gas. Sibbald reportedthat acoustic emission was detected during the addition of water to aluminumchloride [177]. This reaction was studied to see if the acoustic emission producedwas similar to the acoustic emission produced during the thermal degradation ofPVC in air. Although aluminum chloride will undergo hydrolysis in air due to thepresence of water vapor, preliminary experiments indicated that this could not bevisually or acoustically detected if a sample of aluminum chloride (1 g) was left to111stand for a period of 20 minutes. Therefore in these experiments, the hydrolysiswas initiated by the addition of a small volume of water. To approximately matchthe media through which the acoustic emission propagated during the thermaldegradation of PVC in air, water was added to the aluminum chloride, rather thanthe other way round. This ensured that for the two sets of experiments beingcompared, the acoustic waves were propagating through a solid though obviouslymatching the exact characteristics of the solids was very difficult.Aluminum chloride (1 g) was placed in a waveguide (‘Type B’, Fig. 16). Thewaveguide was placed in the waveguide holder and the transducer connected. Theacquisition settings were as described in section 3.4.1.2, to enable a truercomparison of the results obtained. Drops of water were added down the side ofthe sample holder to react with the aluminum chloride until Ca. 200 acousticemission signals had been collected. Earlier experiments had shown that addingwater dropwise down the sides of the empty vessel did not produce acousticemission.3.4.1.7 Acoustic emission from the isothermal degradation of PVC in airThe isothermal degradation of PVC was investigated to see if kineticinformation such as reaction rates and energies of activation could be readilydetermined for the acoustic process(es).As a result of earlier observations, it was known that AE from the thermaldegradation of PVC in air commenced in the region of about 450 to 490 K whenusing the oil bath apparatus. In the studies of the thermal oxidation of PVC underisothermal conditions, the same apparatus as used for the visual studies wasemployed. The oil bath was heated to either 453, 468 or 483 K. The oil bath washeld at this temperature for a minimum of 10 minutes prior to the boiling tube112containing the PVC being suspended in the bath. AE data was collected for aperiod of 5 hours from this time. Both the continuous acoustic level monitoring andthe individual acoustic signal capture apparatus were used. The amplifier gain wasset to 30 dB, the acquisition trigger level set to 150 mV and the vertical scale on theoscilloscope was set to 200 mV per division. A I second delay was used to limit thenumber of individual waveforms captured.Overall, these results had indicated that there was a wide range ofpossibilities for studying the thermal degradation of PVC in air using AE techniques.The stepwise improvement of the furnace apparatus necessitated that a number ofthe studies be repeated due to inaccuracies found when using the earlythermosonimetry equipment. Section 3.5 details only the experiments carried outwith the final version of the thermosonimetry apparatus.3.4.2 Glass transition temperature (Tg)The glass transition point (Tg) of an amorphous polymer is the temperature atwhich the bulk polymer ceases to be brittle and glassy in character and becomesless rigid and more rubbery. It is difficult to measure this transition using manyconventional thermal techniques so if AE was found to detect this transition then thebenefits would be high. The glass transition point varies greatly with polymer asshown in Table 3. As the glass transition point of PVC is at an easily obtainabletemperature, then PVC was used for this study.The sample mass was 250 mg. The optimized furnace apparatus was used.A linear heating rate of 1 K/mm was employed and the experiment was stoppedonce the temperature had reached 373 K. Acoustic data were acquired using thecontinuous acoustic level monitoring apparatus. Acquisition settings were as givenin section 2.5.1.1133.4.3 The crystalline melting point (Tm)A sample of PP (200 mg) was used. The apparatus shown in Figure 23 wasused. The sample was placed in the boiling tube and the oil bath heated up toCa. 460 K at a rate of Ca. 10 K/mm. Acoustic data were collected throughout theheating period. The continuous acoustic level monitoring apparatus was used andthe amplifier gain was set to 30 dB.3.4.4 CrystallizationAE during the crystallization of polymers, under isothermal andnon-isothermal conditions, had been reported in the literature [130-2]. Thesestudies used signal counting methods and no attempt was made to characterize theindividual acoustic signals.In the initial experiments, the non-isothermal crystallization of PP, HDPE andLDPE was studied. The procedure used was based on that described byShen et a!. [132]. For those experiments where a polyethylene polymer was used,the sample mass was 200 mg, whereas for the PP experiments a mass of 100 mgwas used. For comparison, the non-isothermal crystallization of benzoic acid wasstudied using the same procedure. It had been previously reported that thisprocess produced acoustic emission [174]. For the benzoic acid, a sample mass of100 mg was used.The apparatus shown in Figure 23 was used. The oil bath was heated to460 K and then kept at this temperature for 20 minutes. The hot plate was thenturned off and the acquisition of acoustic data commenced. The temperature of theoil bath was recorded every minute for 40 minutes after which the data acquisitionwas halted. For comparative purposes an additional experiment was run for eachpolymer in which the acoustic probe was replaced by a thermometer. This enabled114the actual temperature of the polymer material to be recorded thus giving a truermeasure of the cooling profile of the polymer. As discussed earlier in section 1.5.1,the direct insertion of a thermocouple in a sample while undertaking acousticemission studies was reported to cause severe damping of the signals [148].Isothermal crystallization studies concentrated on PP. The procedure usedwas based on that described by Galeski et al. [1311. The following modifications tothe experimental apparatus and procedure described for the non-isothermalcrystallization of PP were made. After the sample had been heated to 460 K andheld at this temperature for 10 minutes, the boiling tube containing the sample wastransferred to a second oil bath which had been heated to a temperature close tothe melting temperature of the polymer under investigation (i.e., Tm). Temperaturesof 373, 383, 393 and 403 K were used. Once the boiling tube was in place,acquisition of acoustic data was commenced. The temperature of the oil bath waschecked every 2 minutes and heat supplied to the oil bath adjusted to ensure thatthe temperature was within ±2 K of the reaction temperature. Again, an additionalexperiment at each temperature was undertaken during which the acoustic probewas replaced by a thermometer so the temperature of the polymer could bemeasured.The non-isothermal crystallization studies for benzoic acid were repeatedusing the optimized furnace apparatus. A sample mass of 150 mg was used. Thefurnace was heated to a temperature of Ca. 460 K at a rate of 5 KIm in and then heldat this temperature for 20 minutes. At the end of this period, the heating wasstopped and acoustic data collection commenced for a period of 2 hours. Theindividual signal capture apparatus was used for this study. A comparison wasundertaken of the acoustic emission signals using the waveguide and using theacoustic probe.1153.4.5 PolymerizationIn preliminary experiments, simple chain-growth and step-growthpolymerization reactions were used to investigate whether AE was produced duringthese reactions.The chain polymerization of styrene was undertaken based on the proceduredescribed by Fessenden and Fessenden [192]. Styrene (5 g) was weighed into abeaker (30 mL PyrexTM). Azobisisobutyronitrile (AIBN) (100 mg) was added asinitiator and the mixture stirred. The beaker was covered with a watchglass andthen the mixture was heated on a hot water bath for about thirty minutes. Aftercooling to room temperature, the bottom of the beaker was smeared with apiezongrease and the beaker placed on top of the acoustic emission transducer. Dataacquisition was then commenced for a period in excess of 12 hours using thesettings given in section 2.5.1.The formation of Nylon 6-10 was used as an example of a step-growthpolymerization reaction. The procedure used was based on the method describedby Pavia et al. [193]. The bottom of the beaker (50 mL, PyrexTM) to be used as thereaction vessel was greased and placed in contact in with the acoustic transducer.The acoustic data acquisition settings used were as above. Data collection wascommenced at the start of the addition of the chemicals to the reaction vessel. Anaqueous solution of I ,6-hexanediamine (5% (v/v)1 10 mL) was poured into thebeaker and 10 drops of aqueous sodium hydroxide solution (20% (w/v)) added. Atetrachloromethane solution of decanedioic chloride (5% (vlv), 10 mL) was thenslowly poured into the beaker, forming two immiscible layers with polymer formationat the liquid-liquid interface. A copper-wire hook (a 6-inch piece of wire bent at oneend) was used to firstly free the walls of the beaker from polymer strings. The mass116of polymer at the centre of the interface was hooked and the wire was raised slowlyenabling continuous formation of the polymer, producing a rope.34.6 Plasticization of PVCThe plasticization of PVC was investigated using the AE technique. In thisprocess, a plasticizer, such as di-n-octyl phthalate, is mixed with the polymer. Thisleads to the PVC becoming more pliable as a result of the lowering of the Tg of thepolymer.After discussions with Mr. J.R. Wallace of Esso (Esso Research Centre,Sarnia, ON), the following procedure was utilized. The optimized furnace apparatuswas set up. Di-n-octyl phthalate (DOP) (5 g) was added to the PVC (2.5 g). Thefurnace was heated at a rate of about I Kfmin from room temperature to about400 K. AE data were collected throughout the heating period, using the continuousacoustic level monitoring apparatus. The acquisition settings were as given insection 2.5.1 except that the amplifier gain was set to 50 dB to ensure maximumsensitivity.The interaction of the plasticizer and the PVC polymer on mixing was alsoinvestigated using the AE technique. PVC (500 mg) was placed in a beaker, whichwas positioned on top of the acoustic transducer. A layer of apiezon grease hadbeen previously applied to the active surface of the transducer to ensure a goodacoustic contact. DOP (200 mg) was added to the PVC and the continuousacoustic emission monitoring apparatus was again used to detect any acousticemission produced during the mixing. Again, the amplifier gain was set to 50 dB.1173.5 Polymer studies using thermosonimetryThe preceding sections dealt with preliminary investigations into theapplication of the acoustic emission technique to various areas of polymerchemistry. On the basis of these preliminary experiments, the thermal degradationand the crystallization of polymers were found to be the two areas of polymerchemistry which showed promise with regard to use of the AE technique.AE from the isothermal and non-isothermal crystallization of polymers hadalready been reported in the literature [130-132] and an adequate explanation forthe sources of the acoustic emission given. Later studies concentrated on thethermal decomposition of polymers in air.The thermosonimetry experiments detailed in sections 3.5.1 to 3.5.6 areprimarily concerned with investigating the effect of change of chemical parameters,such as molecular mass, on the acoustic emission. Other experiments that soughtto develop a fundamental explanation of the source(s) of the acoustic emissionarising from the thermo-oxidative decomposition of polymers are detailed insubsequent paragraphs.3.5.1 General experimental detailsThe furnace apparatus used for the majority of these studies was theoptimized furnace apparatus described in section 2.5. For each of the experimentalstudies, the following procedure and acquisition parameters were used, exceptwhere otherwise stated. All experiments were carried out in triplicate.The polymer sample was placed in the waveguide (‘Type C’, Fig. 16) and thethermosonimetry apparatus set up, as in Fig. 8. The furnace was heated using theprogrammed heating rate to a final temperature of 873 K. Except where indicated,the heating rate was set to 5 K per minute. Both the continuous acoustic level118monitoring and the individual acoustic signal capture apparatus were used. Thesignal capture acquisition trigger level was set to 100 mV and the vertical scale onthe oscilloscope was set to 200 mV per division. An acquisition delay of 1 secondwas used to limit the number of signals collected to a manageable size. Theamplifier gain was set to 25 dB.3.5.2 Polymer studiesThermosonimetry of four different polymers was carried out to see if theacoustic emission produced was a function of the polymer being decomposed.The polymers used were PVC, HDPE, PP and PET and the sample massused in each trial was 200 mg.3.5.3 Effect of sample massExperiments were undertaken to see if the acoustic emission producedduring the thermosonimetry of PVC was influenced by the sample mass used.Sample masses were 200, 100, 75, 50, 25 and 10 mg. In addition, experimentswere undertaken in which no sample was used. These acted as blank experiments.To assess the effect of changing the acquisition delay on the amount ofacoustic emission produced, a thermosonimetry experiment was undertaken using asample mass of 200 mg but no acquisition delay. Only one such experiment wasundertaken.3.5.4 Molecular mass studiesThermosonimetry of the six PVC resins was undertaken to investigate theinfluence of polymer molecular mass on the acoustic emission produced. Theseresins were prepared according to the same batch process and as a result, the only119recorded differences in the polymers were in their viscosity as a result of molecularmass differences. The sample mass used in each case was 50 mg.3.5.5 Effect of heating rateThermosonimetry of PVC was undertaken using five different heating rates toassess whether thermosonimetry was subject to the same heating rate effects asother thermal analysis techniques (e.g. [194]).For each experiment, a portion of the same PVC resin (“#2”, 50 mg) wasused. Experiments were undertaken using heating rates of 10, 7, 5, 3 and I K/mm.The final temperature used for all experiments was 873 K.3.5.6 Copolymer studiesThermosonimetry of different ethylene-vinyl acetate copolymers wasundertaken to see if the technique could detect changes in copolymer composition.Copolymers with a vinyl acetate content of 18 and 40 % were used. It was notpossible to obtain copolymers with a higher vinyl acetate content. The difficulty inhandling ethylene-vinyl acetate copolymers with a high vinyl acetate content meantthat these materials were unavailable commercially.Low density polyethylene and vinyl acetate homopolymers were also used.These polymers may be regarded as ethylene-vinyl acetate copolymers containing0 % and 100 % vinyl acetate respectively. The sample mass used was 200 mg.1203.6 Polymer studies using thermogravimetryThermogravimetry was used as a comparative method in order to assess theacoustic emission results. Three separate studies were undertaken; the effect ofheating rate, the effect of molecular mass and the effect of copolymer composition.The thermosonimetry apparatus described in section 3.5.1 was used. However forthese experiments the acoustic cable was disconnected from the transducer to allowstable mass readings.3.6.1 Effect of molecular massThermogravimetry of three PVC resins (“2”, “5” and “6”) was undertaken toassess if the polymer molecular mass influenced the results. A sample mass of200 mg and a heating rate of 5 K/mm were used. Again the results obtained werecompared to the analogous thermosonimetry experiments (section 3.5.4).3.6.2 Effect of heating rateThermogravimetry of one of the PVC resins (“#2”) was undertaken usingheating rates of 10, 7, 5, 3 and I KImin. These results were to be compared tothose undertaken earlier using thermosonimetry. However, due to the lack ofsensitivity of the thermogravimetry apparatus, the sample mass used was 200 mgwhereas the sample mass used for the analogous thermosonimetry experimentswas 50 mg.3.6.3 Effect of copolymer compositionThermogravimetry of the four copolymers was undertaken. Again a samplemass of 200 mg and a heating rate of 5 K/mm were used.1213.7 Thermogravisonimetry studies3.7.1 Polymer studies using thermogravisonimetryThe thermal decomposition of the polymers PVC, PET, HDPE and PP in airwas investigated using the newly-developed TGS technique. The optimized furnacewas used though a number of modifications were made so that the TGSexperiments could be carried out.The IR transmitter circuit was attached to the transducer and positioned onthe balance. The lR receiver circuit was positioned so that the light emitting diodeof the transmitter circuit and the photodiode of the receiver circuit were adjacent toeach other but a distance of 5 mm apart. To improve the stability of the massreadings, a two-piece wind break was constructed by UBC Dept. of ChemistryMechanical shop. The cylindrical vertical piece, constructed from a commercialcomposite material (CeleronTM, Universal Plastics, Burnaby, BC) had a depth of180 mm, a thickness of 5 mm and a diameter of 480 mm. It was placed so that itencompassed the balance and the base plate on which the furnace was mounted.A hole of diameter Ca. 30 mm was cut in the side to permit the connection of cablesto the apparatus. The horizontal piece, constructed from a transparentpolycarbonate material (LexanThi, Universal Plastics Burnaby), was Ca. 560 mmsquare. A hole, of diameter 40 mm, was cut in the centre of this piece to permit thewaveguide to extend into the furnace. To enable the piece to fit around the bracketholding the furnace, a notch of depth 130 mm and length 80 mm was cut out fromone of the sides at Ca. 240 mm from the end. The horizontal piece was placed overthe balance ensuring that the waveguide could be positioned in the waveguideholder. The stainless steel plate used previously as a heat shield was positionedon top of the horizontal piece of the wind break. To prevent transmission of the122heat from the stainless steel to the wind break, the stainless steel shield was raisedusing two pencils placed at both sides of the shield.In each experiment, 200 mg of the polymer was used. The heating rate was5 K/mm up to a final temperature of 873 K. Only the continuous acoustic levelmonitoring apparatus was used. Studies reported earlier (section 2.7), had disputedthe use of the TGS apparatus for the collection of the individual acoustic emissionsignals due the impure frequency response of the circuit. The input impedance ofthe main amplifier was switched to 2 M2 as the direct connection input to theamplifier was being used. This led to the background noise level being significantlylower even though the main amplifier gain was still 25 dB.3.7.2 Other thermogravisonimetry studiesFor several inorganic systems, it had been reported that acoustic emissionwas detected prior to the onset of the corresponding thermal transition. Examplesof systems studied include the decomposition of KCIO4 [151, 167] and thedehydration of CuSO4.5H20[174]. For the latter, acoustic emission was detectedat Ca. 320 K, while TG measurements, made under identical conditions on aseparate instrument suggested the onset of the thermal event was Ca. 345 K.Preliminary work had indicated that for the thermal degradation of severalpolymers in air, acoustic emission was detected after the thermal event hadcommenced. To eliminate the possibility that this was in error due to the apparatusemployed, the dehydration ofCuSO45H20was studied using the TGS technique.Copper sulfate pentahydrate (500 mg) was placed in the waveguide(‘Type C’, Fig. 16) and the TGS apparatus set up. The heating rate was set to5 K/mm and the sample was heated until a temperature in excess of 473 K hadbeen reached. The acquisition settings were as given above. As it was found123necessary to determine if the signals collected were real or noise, then theindividual acoustic signal capture acquisition equipment was used. Use of the lRtransmitter/receiver circuit led to a lower background noise level. Therefore thetrigger was set to 40 mV and the vertical scale of the oscilloscope was set to100 mV/div.3.8 Other polymer studiesA number of other thermal decomposition studies were undertaken in order toenable the acoustic emission results to be directly related to the chemical reactionsoccurring.3.8.1 Reference thermogravimetry studiesTwo sets of TG studies were undertaken at industrial chemical laboratoriesusing commercial thermal analysis instruments. In the first one, thermaldegradation of one of the PVC resins (“#2”) was carried out in a helium atmosphereby Mr. R. Kwasny at Dow (Dow Chemical Canada Inc., Fort Saskatchewan, Alta.).The PVC sample (Ca. 10-20 mg) contained in a large quartz pan was heated at 5,10 and 20 K/mm.More detailed information was provided by using TG coupled withFTIR-spectroscopy to study the thermo-oxidative decomposition of the HDPEpolymer and the same PVC resin. These experiments were undertaken byDr. J.W. Mason at Seal Laboratories (Seal Laboratories, El Segundo, CA). Aportion of each polymer (28.053 mg HDPE or 13.385 mg PVC) was heated in air ata rate of 10 K/mm.1243.8.2 Thermal decomposition studies using residue analysis.The types of studies described previously, i.e. TS, TG and TGS did not giveextensive chemical information regarding the products and residues arising from thethermo-oxidative decomposition. A series of studies using PVC, PET, PP andHDPE were undertaken to determine the chemical identity of some of these species.The thermosonimetry apparatus described in section 3.5.1 was used. Theheating rate was 5 K/mm and a sample mass of 200 mg was used. The finaltemperature was varied in order to obtain products at various stages of thermaloxidation. Each polymer was heated to each of the following nominal temperatures;498, 548, 598, 648, 698, 748 and 798 K. In order to try to quench any reactionsonce the required temperature was reached, the waveguide was carefully removedfrom the apparatus while the furnace was still hot. On cooling, the solid residueswere analyzed by suitable, available analytical techniques. Prior to chemicalanalysis, the residues were heated for 2 hours at 393 K in a laboratory oven toremove moisture.Samples for IR spectroscopic analysis were prepared as KBr pellets.Typically a sample mass of 1-2 mg in 100 mg KBr (BDH Canada Ltd, Vancouver,BC) was used. The spectra were recorded using a FTIR spectrometer (ModelMBI 00, Bomem, St. Jean-Baptiste, PQ). Both reflectance and transmission spectrawere recorded.Elemental analysis for carbon and hydrogen was carried out by Mr. P. Borda,UBC Dept. of Chemistry. Mass spectrometric analysis was undertaken for aselection of the PVC residues at the UBC Dept. of Chemistry by Mr. S. Pergantis. Athermal desorption method was employed. Apparatus limitations prevented the useof a linear heating rate. The final temperature was Ca. 900 K.125Scanning electron microscopy (SEM) was also used to view the residues. ASEM system (S2300, Hitachi) situated in the UBC Dept. of Metals and MaterialsEngineering was used. An accelerating voltage of 20 kV was used with the detectorat a working distance of 15 mm. The samples were Au-coated prior to analysis toprovide an electrically conductive surface. The Au coat was deposited under areduced pressure Ar atmosphere (<200 mtorr), with a direct current of 20 mAapplied for a period of 4 minutes, resulting in a 100 A thick coating.3.8.3 Thermal stability of PVC using a modified ASTM methodA measurement of the thermal stability of the PVC samples was undertakenbased on the ASTM method D4202-82 [195]. The modified apparatus used forthese experiments is shown below (Fig. 24).StopperGlass TubeCongo RedIndicator PaperTest TubeSampleOil BathFigure 24. The apparatus used to assess the thermal stability of the PVC resins.126A heating bath containing high temperature silicone oil was heated to atemperature of 393 ± 2 K. The oil was continuously stirred using a magnetic stirrer.Poly(vinyl chloride) resin was placed in the test tube (18 mm x 150 mm) to a depthof 25 ± 2 mm. The test tube was then immersed in the oil bath to the level of theupper surface of the resin. The thermal stability of the sample was assessed as thetime elapsed from the test tube being placed in the oil bath to the appearance of thefirst definite blue colour on the ‘Congo red’ indicator paper. In order to aid thevisual detection of the colour change, a magnifying glass was used to view the‘Congo red’ paper.Two samples of different poly(vinyl chloride) resins were run concurrently.The procedure was repeated until two successive replicates were no more than± 10% apart [195].3.8.4 Other thermal stability studiesA qualitative comparison of the thermal stability of the six PVC resins wascarried out using a melting point apparatus (Gallenkamp). Three samples werecompared simultaneously. A small portion of each PVC resin (1-2 mg) was placedin a capillary, which was then put into the apparatus. A thermometer measured thetemperature of the apparatus. The heating rate was set at a low level (Ca. 5 K/mm)and the thermal degradation of each sample was monitored visually by detecting theonset of the discoloration of the PVC sample with respect to its neighbours.3.8.5 Optical microscopy studiesVisual monitoring of the thermal oxidation of PVC was investigated moreclosely using optical microscopy.A low resolution optical microscopy system assembled by Mr. A.P. Cook, wasused to view the samples. A continuously focusable microscope (CFM) (Infinity127Photo-Optical Company, Boulder, CC), with a camera (VK-C150, Hitachi) attachedwas used. The degree of magnification was estimated by measuring thedimensions of the image of an object of known size. In addition, the real-time imagewas displayed using a colour monitor (KV-2064R, Trinitron, Sony Corp.) andrecorded by a standard VHS video cassette recorder. It was also possible to storeselected images using a frame grabber card (PCVisionplus, Imaging Technol. Inc.,Woburn, MA) installed in a PC/AT computer. Further details of this system areprovided elsewhere [122].A sample of the PVC resin (“#2”, 200 mg) was placed in a waveguide of thesame dimensions as that of the ‘Type C’ waveguides (Fig. 16), except that walls ofthe sample compartment were removed. This waveguide was situated on top of thetransducer. The individual acoustic waveform apparatus was used and theacquisition settings were as discussed previously, i.e., trigger level of 100 mV, a1 second acquisition delay and a vertical scale of 200 mV/div.The camera of the optical microscope system was positioned above thewaveguide to give the optimum visual display of the PVC resin. It was necessary toilluminate the PVC resin using an optic fibre light source to provide sufficient lightfor the camera to detect the image. The sample was then gently heated using an airgun (#6966K, 120 V, 2.2 A, Ungar Ltd, Vancouver, BC) until extensive acousticemission had occurred. Throughout the heating period the camera images wererecorded using the VHS cassette recorder.1283.9 Data analysis methods3.9.1 IntroductionThe analysis of the data collected was an exhausting process. As nocomparable studies were available, the scope of possible analysis methods wasinfinite. The temptation was avoided to repeatedly try different analysis methodsuntil a relationship was found regardless of the usefulness of the relationship.Reference was made to other areas of thermal analysis and thus a number of themethodologies used had their origins in TG, DTA or DSC. Details are given belowfor the general methods used for the analysis of the acoustic emission data.Two types of acquisition methods were employed. The following sectionsoutline these methods.3.9.2 Continuous acoustic level monitoringRoutines used to continuously monitor the acoustic level acquired one datapoint every 0.8 seconds. As experiments were often for a duration of many hoursthis gave rise to a large unmanageable volume of data. The data reductiontechnique of BoxcarAveraging [196] was employed. Firstly the data were dividedinto separate groups (‘boxcars’) of equal size. If it is assumed that the randomnoise in each boxcar has an average value of zero then each group of data can bereplaced by the average of the group. In preliminary work, the effect of changingthe number of points in each boxcar was assessed. Increasing the number of pointsled to a greater degree of smoothing of the curve but led to an overall decrease inthe total intensity calculated. Decreasing the number of points in each boxcar led toa better representation of the actual curve observed but a large increase in the sizeof the data set. As a compromise, for all the experiments discussed here, the datawere averaged every 5 points, giving an averaged data point every 4 seconds.129Two types of continuous acoustic level monitoring results were obtained.The first type, termed the ‘acoustic RMS’, gave the individual intensities at eachpoint while the second type, which was termed the ‘cumulative acoustic RMS’, wasobtained by simple addition of all the values up to and including the point of interest.Both types had their uses, with the latter being used extensively for the comparisonof TS and TG results.For a technique to be regarded as thermoanalytical, the measurement has tobe expressed as a function of temperature rather than time [50]. Where possible,the data were expressed as a function of the reference furnace temperature usingthe programmed temperature profile (e.g., 5 K/mm). It was known that the sampleand furnace temperatures were not identical and so in the later experiments usingthe optimized furnace apparatus, the furnace temperatures were converted toapproximate sample temperatures as discussed earlier (section 2.2.4).Characteristic temperatures for the acoustic emission, such as the onsettemperature or the temperature of maximum rate of emission, were assessed by thecombination of two methods. Firstly a close inspection was made of the acousticRMS data and approximate values were obtained from the graphical plots ofacoustic RMS vs. temperature (or time). In the second method, the derivative of thecumulative acoustic RMS was obtained using a 13-point derivative quadraticSavitzky-Golay filter [197,198]. In this method, the point of interest and the pointslying 6 units or less on either side are multiplied by given coefficients, the totalcalculated and then normalized according to a given value. By repeating thisprocedure for all points, the derivative curve can be constructed. Smoothed curvesor the second derivative curves are obtained in a similar way, though using differentcoefficients. Obviously increasing the number of points used for the filter increasesthe degree of smoothing though this is not always favorable. For those cases130where the TS curves were used for subsequent kinetic analysis, a 31 -pointderivative quadratic filter was used to improve the quality of the curves.The TG data were acquired using a similar method as discussed for the dataobtained using continuous acoustic monitoring data and thus similar procedureswere employed for the data analysis. To enable TG experiments using differentsample masses to be compared, the mass readings were scaled as a percentage ofthe original mass reading.3.9.3 Individual acoustic signal captureTwo types of ‘unwanted signals were removed prior to detailed signalanalysis. Under-trigger signals arise due to a trigger level instability of the digitalstorage oscilloscope such that sometimes a waveform with a peak level less thanthe desired trigger level will be captured. Over-range signals are waveforms wherea user-defined percentage of the 1024 points acquired have values at the lower orupper digitization limits. Although, these waveforms are regarded as real signals,and thus are included in any signal counts, their analysis can lead to misleadingresults [120]. The over-range percentage was set to 2 %.As discussed earlier, the methods used to analyze the individual acousticemission signals had been classified in two ways; cumulative acoustic spectralanalysis methods and waveform analysis methods [145].3.9.3.1 Cumulative acoustic spectral analysisPower spectra were calculated using a fast Fourier transform method. As thepower spectra obtained from individual waveforms were variable, much use wasmade of the average power spectra, i.e., the power spectrum obtained by averagingthe power spectra of all the individual waveforms collected.131The main methods used to compare average power spectra were eithervisual comparison of sets of average power spectra obtained under differentconditions or the use of the regression analysis. In the latter method, the intensitiesat a particular frequency for two power spectra were used as x,y pairs in regressionanalysis. If the frequency spectra were similar then a high correlation ofdetermination (r2) was found. This method was first reported by Wentzell andWade who assessed the reproducibility of average power spectra obtained usingdifferent transducers and the same transducer over a period of months [139]. Asmany trials were undertaken for each experimental condition, then many r2 valueswere obtained. By comparing the mean r2 values for each set of conditions, trendsin the average power spectra were assessed.3.9.3.2 Waveform analysisThe acoustic emission signals were characterized using descriptive statisticalfactors (‘descriptors’). In this way, the properties of each signal, e.g., frequencycontent, amplitude, half-life etc., could be represented mathematically allowingcomparison of the signals. As there is usually no prior knowledge of whichdescriptors are likely to be the most informative for a system then typically manydescriptors are used. The use of 50 descriptors has been reported [140] but here amaximum of 32 descriptors were used (Table 11). Further discussion regarding thecalculation of descriptors is given elsewhere [177].As the individual acoustic emission signals are variable, then averagedescriptor values for the trial are frequently used. The most typical signals for a trialwere found by comparing the descriptor values for an individual signal with theaverage descriptor values for the trial.132Descriptor DefinitionTime I Amplitude Domain: 18 descriptors generated from raw signalsPeak Largest amplitude in waveform (in positive or negativedirection).Rlw’S RMS level of waveform.Area Area of waveform, about mean.Crest Ratio of Peak amplitude to RMS level.50-Cross Number of crossings of 50% maximum possible amplitude.25-Cross Number of crossings of 25% maximum possible amplitude.10-Cross Number of crossings of 10% maximum possible amplitude.0-Cross Number of crossings of signal mean voltage.Kurtosis Fourth statistical moment.t area/2 “Half-I ife” of signal i.e. time to half area.1/8 t - 8/8 t Normalized RMS of signal in each 1/8th of the waveform.Frequency Domain: 14 descriptors generated from the power spectrumof signals.Frq-?vIax Frequency in power spectrum with maximum intensity.Frq-Med Median frequency in power spectrum.Frq-Mean Mean frequency.F-Crest Power spectrum crest factor.FBW>15% Power spectrum bandwidth, using only frequencies withpowers> 15% of power at Frq-?vIaxF-QrtLBW lnterquartile bandwidth.DFBJ - Normalized area under power spectrum in DefinedDFB8 Frequency Bands I - 8. These bands are normallydiscrete 1/8th’s of the power spectrum, but are fully userdefinable.Table 11. The descriptors used to mathematically characterize the acousticemission signals.133For many acoustic emission systems including those reported here, normalitytesting such as use of the Kolmogorov-Smirnoff test [199] indicates that thedescriptor distributions do not obey a normal distribution. As a result, medianvalues were used to characterize the average values of the descriptors.Apart from a direct comparison of the median values, the other method usedhere to compare the results obtained using different trials involved the use ofmedian and quartile values [121]. In this method, known here as the ‘d-quarter’method, the resolution of a descriptor to distinguish between two trials is given by :-(xmed(1) - xmed(2)R(J,2) l\d(])..1/4-d(2)+114J (Equation 28)where xmed(1), the median value for trial one, is greater than xmed(2), themedian value for trial two. The parameters d(1)114 and d(2)+114 refer to the lowerquartile from trial one and the upper quartile from trial two respectively. If theresolution, R(1,2), is less than unity then the descriptor cannot discriminate betweenthe two trials.The methods discussed in the previous paragraph were used to compare thedescriptor distributions. Principal components analysis (PCA) was used to comparethe individual signals directly. Only brief details are given here. The work byBrock eta!. [120] provides more details of the procedures used here. Prior to PCA,correlation analysis of the descriptors was undertaken. One descriptor from eachpair of descriptors found to be highly correlated (>90 %) was excluded from PCA.The descriptors were autoscaled and then link scaled to correct for variancedifferences in those descriptors which are directly related to each other,e.g., ‘DFBJ -DFB8’ [177].In a few isolated cases, cluster analysis was also used. In the ‘k-means’method [142], the number of classes expected is set for the acoustic emission134signals of interest defined by N descriptors and then the algorithm decides themembers of each class based on a measurement of the distance apart of theobjects in N—dimensional space. This method was used to confirm classidentification but was not used to pre-suppose class differences.3.9.4 SoftwareMany commercial software packages were used during the analysis of thedata. Spreadsheets such as QuattroPro (Special Edn., Borland International Inc.,Scotts Valley, CA) were used for trivial data analysis tasks such as scaling. Graphswere constructed in SigmaPlot (ver. 5.0, Jandel Scientific, Corte Madera, CA) orExcel (ver. 4.0a, Microsoft Inc., Mississauga, ON). Regression analysis and curvefitting were undertaken using TableCurve (ver. 3.10, Jandel Scientific, CarteMadera, CA). Basic statistical analysis was undertaken using SigmaStat (ver.1.01,Jandel Scientific, Corte Madera, CA).The general data analysis software was written in-house using BASIC(Microsoft QuickBASIC ver. 4.5 and Professional BASIC, ver. 7.0, Microsoft Inc.,Mississauga, ON) by several members of Dr. Wade’s research group including theauthor. Assembly language routines (PROBAS library ver. 4.0, HammerleyComputer Services, Laurel, MD) were used to speed up the operation andfunctionality of many of the programs. For many critical operations including forexample, the calculation of fast Fourier transforms, use was made of add-onlibraries (“Science & Engineering Tools for Microsoft QuickBASlC”, ver. 6.1,Quinn-Curtis, Needham, MA). In other cases, routines were adapted from suitablereference books (e.g. [200]). Further details of these programs are given inreference [120]. All programs were tested with suitable simulated data prior to use.135CHAPTER 4. RESULTS AND DISCUSSION4.1 IntroductionPresentation of the results starts with the preliminary studies where a generalsurvey of polymer chemistry was undertaken in order to discover processes whichgive rise to acoustic emission. Subsequently, studies investigating thethermosonimetry and thermogravisonimetry of polymers in air are reported togetherwith related studies using non-polymeric materials. In the discussion, particularemphasis is made with respect to the sources of the acoustic emission and theusefulness of AE for studying polymer chemistry.4.2 Preliminary studiesThe experimental details for the preliminary experiments discussed in thissection may be found in section 3.4 (i.e., results discussed in section 4.2.j refer toexperiments described in section 3.4.1.3 etc.).4.2.1 Thermo-oxidative degradation of poly(vinyl chloride)4.2.1.1 Acoustic emission studies using the ovenFigure 25 indicates the occurrence of AE signals as a function of time foreach of the three experiments. The calculated RMS voltage of each signalwaveform is given as the ordinate, and indicates the relative intensity of the signals.The onset time of the AE was reproducible. The agreement was surprisinglygood considering the crudeness of the experimental apparatus. The intensity andnumber of signals for the third trial was greater than for the first two. After theunder-trigger and electrical spike signals had been removed, 285 AE signals hadbeen acquired for the third trial whereas 223 and 172 signals were acquired for136other two. In the blank experiment, no AE signals were seen although 20 signals,readily identifiable from their waveforms as electrical spike signals, were collected.Therefore the AE was linked to the presence of the sample.E0>0C-)C,,00C)00Cr,Figure 25. Occurrence of AE signals as a function of time during the thermaldegradation of PVC in air using the oven.Using the temperature-time calibration for the oven (Fig. 7), the AE wasfound to commence at around 500 K. This value agrees with the range oftemperatures reported for the dehydrochlorination of PVC in air, i.e., 450 to525 K [88]. The mass losses found during the three experiments were 12, 11 and10 % respectively which corresponds to partial dehydrochiorination of the PVC ascomplete dehydrochlorination of the polymer is equivalent to a weight loss of around58 %. The residues from the experiments were dark brown in colour and not black2020210: F, I-i I iii0 600 1200 1800 2400Time (seconds)3000 3600137as would be expected if complete dehydrochiorination had occurred. It wastherefore surprising that a region of intense acoustic activity followed by a decreasein activity was observed. Although the temperature was not changing greatly, it wassuspected that the dehydrochlorination of PVC would be the prominent reaction andtherefore continuation of the intense AE would be expected if there was a directrelationship between the AE and the dehydrochlorination process.Individual waveform analysis of the AE signals indicated similarities betweenthe signals acquired for the three experiments (Fig. 26). The AE signals displayedwere all acquired during the most intense region of acoustic activity. The signalsare very similar in appearance both in the time domain and frequency domain.Additional confirmation of the similarity of the signals is provided by regressionanalysis of the average power spectra for the three experiments, which gavecoefficient of determination values (i.e., r2) of around 0.81 (number of correlationpairs, n0 = 3). All three signals have major frequency components between 100 kHzand 400 kHz. It has been reported that gas evolution produces acoustic emissionwith a frequency range of 50 kHz to 250 kHz [139]. Frequency componentsbetween 250 kHz and 400 kHz may be the result of fracture processes [139]. Theobserved signals are from individual events, yet possess frequencies that could beattributable to either mechanism.(i-I0>ci)D0EUUC(1)1381 .00.50.0—0.5— 1.00.120.090.060.030.000.120.090.060.030.00CDaCDDCDD(I)0.00 0.05 0.10 0.15 0.20 0 250 500 750 1000 1250Time (milliseconds) Frequency (kHz)Figure 26. Typical AE signals from the thermal degradation of PVC in air usingthe oven. Signals are shown for each of the three trials, together withtheir corresponding frequency spectra.4.2.1.2 Initial acoustic emission studies using the furnaceE0ci)>0C)C,)0C-)00Figure 27. Occurrence of AE signals as a function of time during the thermaldegradation of PVC in air using the furnace.139The residues of these experiments were black in colour and resembledcarbon black. This would suggest that complete dehydrochlorination had occurred.The bulk of the acoustic activity occurred over the same general time period foreach of the experiments with AE commencing at around 1400 seconds (Fig. 27).The temperature-time calibration for the heating rate used here (Fig. 9), indicatedthat the temperature at which the AE commenced (27j) was around 690 K. Thisvalue was higher than that for the previous experiments. However, at this stage ofthe investigation, there was no way of knowing which temperature readings werecorrect. Furthermore, the sources of the acoustic emission with respect to the>0.20.10.00.20.10.00.20.10.00 1200 2400 3600 4800Time (seconds)6000 7200140thermal degradation of PVC were still unknown. It had been reported that the massof sample and heating rate affect the temperature at which a thermal event isdetected [201]. However, this effect was unlikely to account for the largediscrepancy observed here.The results showed that the acoustic activity rises to a maximum and thendecreases somewhat slowly. This was in agreement with the previous experiments.It indicated that the mechanism producing the acoustic emission is ceasing by theend of the experiment. Comparison of the number of individual waveforms collectedduring the experiments indicated a good degree of precision. 781, 648 and 641acoustic signals were collected for the three experiments. The individual waveformscollected during each experiment were found to be more variable. Regressionanalysis of the average power spectra for the three experiments gave r2 values of0.54 to 0.74 (n = 3). Inspection of the average power spectra for the threeexperiments (Fig. 28), indicated that the first two trials produced average powerspectra with higher intensities of frequencies between 200 kHz and 350 kHz.1410.0100.0050.000>0.0100,0050.0000.0100.0050.0000Frequency (kHz)Figure 28. Average power spectra of the AE signals collected during the thermaldegradation of PVC in air using the furnace.500 10001424.2.13 Comparison of acoustic emission from various PVC resinsThe acoustic emission signals were concentrated within a short time intervalfollowed by infrequent signals throughout the rest of the experiment (Fig. 29). Thiswas analogous to the results shown earlier (Fig. 25). However, the acousticemission commenced at an earlier time (and lower temperature) for the resultsshown here. Resins used in these two sets of experiments were known to differ inmolecular mass. The effect of the polymer molecular mass on the onsettemperature of acoustic emission was investigated. For these experiments, thetemperature at which the first acoustic emission signal was captured (Ti) was- (a). .k‘ii LI>E0V>0C)(I,130C)00cn0.20.10.00.20.10.00.20.10.0- (b)I ‘‘[ UIJ(C)IIiiIl iii0 1200 2400 3600 4800 6000 7200Time (seconds)Figure 29. Occurrence of AE signals as a function of time during the thermaldegradation of PVC in air for three different PVC resins.(a) Resin #2, (b) Resin #4, and (c) Resin #5.143determined. This value was plotted as a function of the viscosity-average molecularmass (My) of the polymer for all four PVC resins used (Fig. 30).750725 -700 -Va675650 :625 -5) —(I) —cCD 600 -575 -C550 I40 60 80 100 120 140 160 180 200Viscosity—overage molecular moss (Mv/i 000)Figure 30. The onset temperature of AE as a function of the viscosity-averagemolecular mass for the four PVC resins. The points shown astriangles on the plot correspond to two experimental values. The pointshown as a light square is an outlier and is discussed in the text.The results suggest an increase in the onset temperature for acousticemission (i.e., Ti) with higher molecular mass of the PVC polymer. The scatter invalues obtained for the PVC polymers of a similar molecular mass is obvious. TheTint value shown on the plot as a light square was not caused by a single spuriousacoustic emission signal. Examination of the occurrence of acoustic emission as afunction of time for all the experiments, indicated that in this case, the acousticemission started at a lower temperature. At this stage of the investigation, thetemperature control of the furnace had not been optimized. The observed scatterand the outlying value may be due to the lack of adequate control.144The total number of acoustic emission signals collected during each of theexperiments were compared (Table 12). There is a high degree of variability in theresults especially when compared with the results obtained using the other PVCresin (i.e., #6). For these preliminary experiments, a high sample mass was used(i.e., 5 g). This was necessary to ensure that if acoustic emission was producedduring the thermal degradation then it would be detected. This sample mass wasfar in excess of the sample mass typically used for thermal analysis i.e., a few mg toless than 0.5 g. As a result, additional sources of irreproducibility may be largethermal gradients throughout the sample or diffusion problems caused by the bulk ofthe sample [202]. Later experiments used lower sample masses.Trial 1. Trial 2. Trial 3.PVC Resin#2. 1825 4104 781PVC Resin #4 854 3399 859PVC Resin #5 1289 2180 886Table 12. Number of AE signals collected for the three PVC resins during thethermal degradation of PVC in air.Analysis of the average power spectra indicated similar results to previousexperiments. The most intense frequencies in the spectra were between 100 kHzand 250 kHz with a variable content in the region between 250 kHz and 400 kHz.The average power spectra of the blank experiment indicated that the signalscollected contained low frequency components with the characterized transducerresonance at around 800 kHz also being observed.145Figure 31. Average power spectra of the AE signals collected for the PVC resins#2 (A), #4 (B), and #5 (C) during the thermal degradation of PVC inair. The intensities of the average power spectra shown in (D) for theblank experiment have been multiplied by a factor of 10, to enablecomparative intensity scales to be used.cijL)ci)ci)0.0 150.0100.0050.0000.0150.0100.0050.0000.0150.0100.0050.0000.0150.0100.00500000 500 1000Prequency ([ 0.05- : 0.00017 -.- 0.00016-0.04 - : .C.) 0.00015 -CI)ci)0.03 - I 0.00014 - I I40 80 120 160 200 40 80 120 160 200Viscosity—average molecular mass (Mv/1000)Figure 32. Median AE descriptor values as a function of viscosity-averagemolecular mass for the PVC resins. (A) ‘RlvfS’, (B) ‘t @area/2’descriptors.The left panel indicates results typical for the majority of the descriptors.Changes in the viscosity-average molecular mass of the PVC resin did not effect themedian descriptor values. The right panel of Figure 32 indicates results typical ofthose seen for six of the descriptors; ‘t @area/2’, ‘2/8 t’, ‘3/8 t’, ‘6/8 I’, ‘7/8 t’ and ‘8/8 t’.In each case, the median descriptor values for the PVC resin of lowest molecular148mass (‘#2’) differed from the median descriptor values for the PVC resin of highestmolecular mass (‘#6’) though there was a close agreement in the median descriptorvalues for the intermediate molecular mass PVC resins. For the ‘2/8 t’ and ‘3/8 t’descriptors, the median values for the highest molecular mass resin were lower thanthe median values for the lowest molecular mass resin whilst for the otherdescriptors the median values were greater for the highest molecular mass resin.The descriptors listed are all time-domain descriptors indicating that changes in themolecular mass of the PVC resin influenced the time-domain characteristics of theAE signal but did not appear to affect the frequency content of the signals. As theintensity related descriptors (i.e. ‘RAilS’, Teak’) were not affected by the change inmolecular mass it appeared that the duration of the AE signal was influenced by thechange in molecular mass. If the duration of the acoustic emission signal increasesthen the value of the ‘t @area/2’ descriptor increases. An increase in the duration ofthe signal leads to an increase in the higher time octiles. As the time octiles arenormalized based on the total RMS of the signal then the lower time octilesdecrease in value correspondingly. This argument is consistent with the resultsindicated from the figure and thus it can be proposed that the duration of theacoustic emission signals was greater for the highest molecular mass resin. Thevariability in the results obtained suggests that further trials using each resin werenecessary. This would permit a better assessment of any trends as a function of theviscosity-average molecular mass.1494.2.1.4 Comparison of AE and mass loss measurementsThe acoustic emission commenced at temperatures of 630, 615 and 641 K,for the three experiments (Fig. 33A). Good reproducibility in the cumulativeacoustic RMS curves is seen. The mass loss commenced at between 600 and610 K, i.e., at a lower temperature than the start of the acoustic emission (Fig. 33B).The acoustic emission continued towards the end of the degradation process whenthe rate of mass loss was low. Thus the differences in the AE and mass lossinitiation times cannot be explained by any lack of sensitivity of thethermosonimetry. Closer examination of the results indicated that the acousticemission commenced close to the temperature at which the maximum rate of massloss was seen (Fig. 33C). For the thermogravimetry experiments, the maximum rateof mass loss was found at 630, 641 and 635 K, respectively. Comparison with thenumbers given earlier for the onset temperature of the acoustic emission (i.e., 630,615 and 641 K) indicated reasonable agreement. This empirical relationshipbetween the thermosonimetry and thermogravimetry results is interesting given thatin previous TS-TA experiments, the temperature of the maximum rate of acousticemission was found to correspond to the onset temperature of the DTA peak [149].This discrepancy may be due to the different types of materials being studied as thereferenced study involved the thermosonimetry of inorganic materials includingKCIO4 and Si02 powders. Alternatively, this discrepancy may be a result of the TGand TS experiments not being undertaken at the same time leading to possibleerrors due to the irreproducibility of the experiments.(A)150C)U)D0C)0ci)>0DED0-oci)0C)U)(I)(B)(C)NO‘4-.- 0oci)U)0ci) —C) 0Lci)0U)U)U)0U)U)0E‘4-0ci)I.0Figure 33.1 .00.80.60.40.20.01008060402000.050.00—0.05—0.10—0.15—0.20300 800Temperature (K)Comparison of thermosonimetry (TS) and thermogravimetry (TG)results for the thermal degradation of PVC in air.(A) TS curves. (B) TG curves.(C) TS and derivative thermogravimetric (DTG) curves.1 .0C-)C0.8L/) 0‘-‘4L1.L)U) CDC-)04 QCD0..— C(I)C-)0.2400 500 600 7000.01514.2.1.5 Visual studiesWhen monitoring the thermal degradation of PVC using a combination ofvisual observation, acoustic emission and detection of HCI evolution, there was adifference in the time/temperature of onset of thermal degradation as measured bythe three methods. Using the ‘Congo Red’ indicator paper, hydrogen chloride wasfirst detected at a temperature of 423 K whilst the first real indication of discolorationof the lower regions of the PVC resin was observed at 413 K. The acousticemission commenced further into the degradation process with the first acousticemission signals being detected at Ca. 473 K.The small difference in the onset temperature of thermal degradation of thePVC as determined visually and using ‘Congo Red’ indicator paper, was probablydue to the slow diffusion of the HCI from the bottom of the PVC resin where thetemperature is highest, to the surface and subsequent detection. As a result,significant HCI may not be detected until a higher temperature even though some‘browning’ of the sample was observed. The significant difference in the onsettemperature of thermal degradation (as determined using these two methods)compared to the onset temperature of acoustic emission further enforced the beliefthat the acoustic emission was indirectly related to the degradation process ratherthan being a direct measure of the degradation process. It was noticeable that priorto the start of the acoustic emission, the surface of the polymer had blackened andappeared brittle. This was the first evidence that the acoustic emission possiblyresults from gas escaping through the surface.U)ctG)1524.2.1.6 Comparison of the acoustic emission produced by the hydrolysis ofaluminum chloride and the acoustic emission produced during thethermal degradation of PVC in airDifferences between the average power spectra from the two chemicalsystems are evident (Fig. 34). The hydrolysis of AICI3 produced acoustic emissionsignals with the most intense frequencies at Ca. 100 kHz while the thermaldegradation of PVC in air produced acoustic emission with a higher frequencycontent although the intensities of frequencies around 100 kHz were also high.0.020.010.000.0100.0050.000Figure 34. Average power spectra from (A) the hydrolysis of AICI3, and (B) thethermal degradation of PVC in air.0 500 1000Frequency (kHz)153Mean coefficient of determinationAid3 and AICI3 trials 0.6 ± 0.4 (n = 3)PVC and PVC trials 0.6 ± 0.3 (ne, = 3)AICI3 and PVC trials 0.1 ± 0.1 (n = 9)Table 14. Regression analysis of the average power spectra from the acousticemission produced during the hydrolysis of AlCl3 and the thermaldegradation of PVC in air.The results obtained by regression analysis of the average power spectrawere analyzed using the well-known ‘Students t-test’ [203]. Significant differencesexisted between the results obtained using two trials from the same system andwhen using one trial from each system (Table 14). Thus the average power spectrafrom each system may be regarded as different.Calculation of the median descriptor values indicated a number of differencesin the descriptor values. Not surprisingly, the acoustic emission signals producedby the two chemical systems could be discriminated using the median values formany of the frequency descriptors. Principal components analysis was undertakenusing only those descriptors for which a separation was evident. Due to differencesin the number of signals collected, 200 signals were randomly selected from eachsystem to test whether the two systems could be discriminated. An acceptableseparation of the two classes of acoustic emission signals was found using principalcomponents analysis but this was only seen if prior classification of the signals wasemployed (Fig. 35). The use of cluster analysis gave unsatisfactory results as theclassifications obtained agreed with the known classes to a very low degree.Therefore this method of analysis was not pursued further.cz0ciE00ci00ciFigure 35. Principal components analysis of the acoustic emission signalsacquired during the hydrolysis of AlCl3 and during the thermaldegradation of PVC in air.These results provided further evidence that the acoustic emission producedduring the thermal degradation of PVC in air was not directly, and specificallyrelated to the evolution of HCI.1544-2-0-—2 -I I I I0 0 000 00 0 00 000000V000V 0 0V 0V y wV 0 0 00 00 0V0 V VVV00%00•VT VV V V•0 • O’VVV 0 VV0o00cg800—4- I I I I I—6 —6 —4 —2 0 2 4Principa componentAluminum chloride°PVC64.2.1.7 Acoustic emission from the isothermal degradation of PVC in air>U)C)CI]0C)ci)zzFigure 36. Cumulative acoustic RMS curves as a function of time for theisothermal degradation of PVC in air at temperatures of (A) 483 K,(B)468 Kand (C)453 K.1552.42.01 .61 .20.80.40.02.42.01.61 .20.80.40.02.42.01.61 .20.80.40.0(C) -0 50 100 150 200 250 300Time (mins)156There were large discrepancies between the results obtained for the acousticemission during the isothermal degradation of PVC in air at 483, 468 and 453 K(Fig. 36). The acoustic emission increased as the temperature was raised butreplicate experiments at a given temperature still gave a high degree of variability.The acoustic emission results were assessed numerically using three specificparameters: onset time for the acoustic emission (t1), total cumulative acoustic RMSand the number of waveforms collected (Table 15).T (K) t1 (s) Total cumulative Number ofacoustic RMS (V) individual signals483 558 2.259 31231094 1.796 233526 1.581 2605Avg.=560 Avg.=1.879 Avg.=2688468 518 0.935 22992618 0.927 20431806 1.694 2062Avg.= 1650 Avg.= 1.18 Avg.=2135453 5086 0.102 2713118 0.371 7791990 0.842 1704Avg.= 3398 Avg.=0.438 Avg.= 918Table 15. Summary of acoustic emission results from the isothermal degradationof PVC in air at 483, 468 and 453 K.As expected the t1 values for the acoustic emission decreased as thetemperature was raised. However, the precision of the measurements was low.Therefore it was not possible to distinguish the three experiments at one157temperature from the three experiments at the next temperature based on thet1 values, even though the average t values for the three temperatures were quitedifferent. The total cumulative acoustic RMS increased as the temperature wasraised but again a low precision was obtained leading to similar problems indistinguishing the experiments run at different temperatures. The total number ofindividual acoustic waveforms collected also increased as the temperature wasraised. Although the values at the lowest temperature were variable, the overallhigher precision of these measurements led to the experiments run at differenttemperatures being distinguished based on the total number of individualwaveforms collected (Fig. 37).35003000 -cn 2500 -CC2000-1500 -1000 -500 -450 455 460 465 470 475 460 485 490Temperature (K)Figure 37. Total number of individual acoustic waveforms collected as a functionof reaction temperature for the isothermal degradation of PVC in air.158Analysis of the average power spectra obtained for the nine experimentsyielded a range of r2 values from 0.68 to 0.95 (n0 = 36) with no differences beingfound between the average power spectra obtained using the three reactiontemperatures.Median descriptor values were calculated for the three trials at each of thethree temperatures. These were plotted as a function of the reaction temperature todetect any differences between the acoustic emission produced at differenttemperatures. Four types of plot were found and examples are shown (Fig. 38).0.10 0.00014 -_________________A • B0.00013 -0.09.• 0.00012 - •• . V0.08Cl] 0.00011 -(V0.07 I 0.00010 - I I I450 460 470 480 490 450 460 470 480 49001.3 350-C)C’] C: 300- •1.2.• • 250 ••1.1200 •• •1.0 150—450 460 470 480 490 450 460 470 480 490Temperature (K)Figure 38. Examples of the general plots obtained for the median descriptorvalues as a function of reaction temperature. (A) ‘RA’IS’, (B) ‘t area/2’,(C) ‘3/8 t’ and (D) ‘FrqMed’. Duplicate points are shown as a lighttriangle.159In Panel A, the median descriptor values appeared to increase as thetemperature was raised but the low precision of the measurements meant that thevalues obtained at different temperatures could not be completely distinguished.This result was seen for some of the descriptors related to intensity, e.g., ‘RUS’. InPanel B, no variation in the median descriptor values was found as the reactiontemperature was changed. In Panel C, the median descriptor value appeared todecrease as the reaction temperature was raised. Unlike the results in PanelA, themedian descriptor values obtained at the highest and lowest temperatures could becompletely distinguished indicating that there was a significant change in themedian descriptor values. Although this result was found for a number ofdescriptors, no real pattern in the types of descriptors was apparent. Panel Dindicates an identical result to Panel C except that now the median valuesincreased as the reaction temperature was raised. The ‘FrqMed’ descriptor gavethis result, suggesting a change in the frequency content of the acoustic emissionsignals as the reaction temperature was raised. Again, additional trials wouldpermit these trends to be further substantiated.4.2.17.1 Determination of kinetic parameters for the acoustic emission fromthe isothermal degradation of PVC in airThe general form of the kinetic expression used in both dynamic andisothermal thermogravimetry may be written asA.exp(jJ.f(cL) (Equation 29)where c is the fractional reaction or degree of conversion, A is thepre-exponential factor, Ea the activation energy, R is the gas constant, flx) is afunction of the degree of conversion at a time t and T is the absolute temperature.160Typically for polymer degradation studies, f(a) = (1a)Q1r) [97], where r is thereaction order. For thermogravimetry, c is given by(mo-rn)cc. = (Equation 30)(m0 - rnf)where m0 is the initial mass, m is the mass at a particular time and mf is themass of the sample when the reaction is complete.Many different integral and differential forms of this general equation areused in the analysis of isothermograms. Basan and Guven compared the use ofvarious isothermal TG methods applied to the degradation of PVC [204]. Of themethods they discussed, two methods were applied to the analysis of the acousticemission data to see if a similar relationship existed here.In the general rate method, the rate constant is found from the intercept andthe reaction order from the slope of the line for a plot of ln(daldt) vs. ln(1-cc.) [205].The activation energy is then found by determining the rate constant at differenttemperatures.In Flynn’s method, the following rearrangements are made to the generalform of the kinetic expression (i.e., Eqn. 29).From the Arrhenius equation,k_A.expRTJ (Equation 31)and from general kinetic theory= kf(cc.) (Equation 32)161Rearrangement of Eqn. 32 givesp tidaJ f(a) = fk dt (Equation 33)0and then integration of Eqn. 33 givesF(a) = kt (Equation 34)where F(a) is the integrated form of the left hand side of Eqn. 33. This is followedby substituting kfrom the Arrhenius equation (Eqn. 29) and taking the naturallogarithm of both sides. Further rearrangement gives the final expression,ln(t) = ln(F(a)) - ln(A) + (Equation 35)The activation energy is obtained directly from the slopes of the parallel linesobtained by plotting lnQ) vs. lIT for a series of values of a [206].For the analysis of the AE curves, the fractional reaction (a) was calculatedwith the point of zero conversion (i.e., a = 0, t = 0) being taken as the value of thecumulative acoustic RMS at the time determined previously as the start of thereaction. The end of the reaction was set as four hours from the defined start of thereaction. Values of a and corresponding times were then calculated for each trial ateach temperature. A range of 0.05 to 0.95 in 0.01 intervals was used for a tominimize errors at the start and the end of the reaction. These values of a and twere then processed according to the two methods of analysis.It was not possible to analyze the cumulative acoustic RMS data using thetwo methods described. For the constructed plots, no linear relationships werefound. This failure may be in part due to aspects of the methods of analysis used.162Both analysis methods gave rise to potential difficulties when dealing with theAE data. Basan and Guven [204] among others, reported that the success of thegeneral rate equation method lies in the ability to determine the derivative (daldt)accurately, something that was a problem due to the stepwise nature of thecumulative acoustic RMS curve. Flynn’s method involved the comparison of valuesobtained from isothermograms at different temperatures. As the total cumulativeacoustic RMS at 453 K was low, determining the corresponding time for each valueof a. would be subject to additional errors.Later work concentrated on dynamic thermal studies due to the limitations ofisothermal methods, e.g., the difficulty of repeating reaction conditions. Isothermalmethods were not pursued further, although the results indicated that the methodwas worthy of future study if the experimental procedure could be improved.4.2.2 The glass transition temperature (Tg)Although Lønvik [159] reported that acoustic emission could be detectedduring the softening process of common glasses, no acoustic emission wasdetected during the set of experiments described in section 3.4.2. This result is nottoo surprising as the glass transition of polymers is difficult to detect using manyconventional methods of thermal analysis [207].4.2.3 The crystalline melting point (Tm)Acoustic emission was not detected during the melting of polymers. Severalworkers including Clark and Garlick [149], Lønvik [163], Shimada and Furuichi [150]reported the detection of acoustic emission during the melting of inorganic saltssuch as KCIO4 and Cal2. The acoustic emission detected was attributed to theformation of cracks and dislocations [149]. The melting of polymers is in many waysdifferent to the melting of inorganic salts. For example, the polymers do not have a163well-defined melting point but a melting range. It appears from a comparison of theresults that crack formation and dislocations are not found in the polymers prior tomelting in a sufficient degree to lead to acoustic emission.4.2.4 CrystallizationThese studies were split into two sections, isothermal and non-isothermalexperiments. Results are presented for studies using low density polyethylene(LDPE), high density polyethylene (HDPE), polypropylene (PP) and benzoic acid.4.2.4.1 Non-isothermal crystallizationFor the polyethylene polymers, few acoustic emission signals were capturedduring the experiments and no apparent relationship between the occurrence ofacoustic emission and time (or temperature) was observed. For example, the threetrial experiments using HDPE gave 6, 5 and 3 signals whilst the three trials usingLDPE gave 4, 3 and 4 signals respectively. During the experiments withpolypropylene, more acoustic emission was detected with 126, 53 and 46 signalsbeing collected during the 40 minute acquisition period. These results concur withthose reported by Galeski et a!. [131] who observed acoustic emission during theisothermal crystallization of PP but very little acoustic emission during theisothermal crystallization of linear polyethylene.Since the rate of cooling of the oil bath was non-linear, empirical functionswere used to transform the time of occurrence of each acoustic emission signal tothe equivalent oil bath temperature. Figure 39 indicates a typical cooling profile forthe polypropylene experiments, together with the curve calculated using theempirical function shown in Eqn. 36. A good approximation of the cooling curvewas possible using this function.164I:::j— 3202400Time (seconds)Figure 39. A typical cooling profile for the non-isothermal crystallizationexperiments together with the fitted curve using the empirical functiongiven in the text.y = a + bx +cx2ln(x) + dx3 (Equation 36)where y is temperature (Kelvin), x is the time (s) and a to dare empirical coefficients.Figure 40 indicates the sample and oil bath temperatures as a function oftime. These results were determined in the experiment in which the thermometerreplaced the acoustic probe in the polypropylene sample. During the region ofinterest the oil bath temperature was close to the measured sample temperatureand therefore using the oil bath temperature was a good approximation to thesample temperature for experiments where the sample temperature was notrecorded. Based on these results, a cooling rate of ca. 4 K/mm during thetemperature range of interest was inferred.0 600 1200 1800450425400E 375ci)F—350Using these cooling profiles, the acoustic emission data were expressed as afunction of sample temperature (Fig. 41). The majority of the acoustic emissionsignals were detected within a limited temperature range between 388 and 373 K.These results are in close agreement with the work by Shen et al. [132] whoobserved copious acoustic emission in the temperature range between 383 and363 K during the non-isothermal crystallization of PP. A few isolated AE signals attemperatures below 343 K were also reported though this was only seen during oneof the three trials reported here. As discussed in section 1.4.3, the acousticemission was reported to be due to stress release from regions occluded by thespherul ites formed during crystallization.Visually, it was not possible to detect the onset of crystallization during theexperiments and so no relationship between the crystallization of PP and theacoustic emission could be found.1650 400 800 1200 1600 2000Time (seconds)Figure 40. The sample and oil bath temperature as a function of time for one ofthe non-isothermal crystallization experiments.166>E0ci)>0CoE50C-)00Cr)0.30.20.10.00.30.20.10.00.30.20.10.0Trial 2—ITrial 3:II Ii I ii I450 425 400 375 350Sompe tempercture (K)Figure 41. TS curves for the non-isothermal crystallization of PP.Shen et a!. [132] also found that the majority of the acoustic emissionoccurred towards the end of crystallization based on a comparison of the acousticemission and differential scanning calorimetry (DSC) results so the samerelationship should apply here. This is in agreement with studies reported involvingthe crystallization of inorganic salts such as KBr [122]. The DSC results alsoindicated that the onset of acoustic emission is close to the peak temperature of theDSC curve although this was not identified by the authors. Similar conclusions areevident from the work reported by Galeski et a!. [131]. A comparison of these ideasand those discussed earlier by Clark and Garlick [149] leads to the followingproposal. For a heating process, if acoustic emission is produced then the onset ofthe DTA peak coincides with the peak acoustic emission activity while for a coolingI.0>a)-ociE000.00.20 0 250 500 750 1000 1250Frequency (kHz)CD0 0Cl).Cl)Q)IC)ci)01000 1250Figure 49. Comparison of the average power spectra obtained from thenon-isothermal crystallization of benzoic acid using(A) the acousticprobe and (B) a waveguide.The results shown confirmthat the frequency content of the two sets ofexperiments are dissimilar.The average power spectra of those experimentsundertaken using the waveguide had higher intensities of frequencies above250 kHz and below Ca. 400kHz. It was also noticeable that the frequency spectrawere similar below 250 kHz suggesting that the acoustic probe attenuatesfrequencies above 250 kHzin comparison with the waveguide.Comparison of the mediandescriptor values from thetwo sets of experimentsalso indicated a number ofdifferences. For the majority of the descriptors,separation of data from thetwo sets of experiments based on the median valueswas possible. The only descriptors that did not give aseparation were ‘FBW>15%’,0 250 500750Freqcency (kHz)180‘FQrtLBW’, ‘DFB7’, ‘DFB8’, ‘50-Cross’, ‘25-Cross ‘0-Cross’, ‘FCrest’ and ‘Peak’. For themajority of the other descriptors, the median descriptor valueswere higher for thewaveguide experiments. This trend was not isolated to time or frequency domaindescriptors with higher values being found for both types. The increase in thevalues of frequency descriptors agreed with the average powerspectra results.Therefore it was concluded that the type of apparatus does have a significant effecton the characteristics of the acoustic emission signals collected. As a result, themajority of the subsequent experiments were undertaken using a waveguide ratherthan using the acoustic probe. As an additional indication of the differencesbetween the acoustic emission acquired using the two experimental setups,principal component analysis was undertaken. A random selection of 100 signalsfrom each of the six experiments were classified according to either the acousticprobe or the waveguide being used. Only those descriptors for which a trend wasobserved were used, making a total of 23 variables. Scaling ofthe data and thesettings for principal components analysis were as given previously. Figure 50indicates a plot of the first principal component vs. the second principal componentfor the results.The plot indicates that it is possible to separate the data according toapparatus to a very high degree. These results confirm that useof the mediandescriptor values is a suitable way to assess the nature of the dataas hereindividual waveforms are being compared.(NI.Cci)C0E0U00C)C0181..• ••• .•8-6-4-2-0-2-—4 -—66.6.2.6.6.6.6.2. 6.6.6. •6. 6.6.6.•. .•.II—8 —6 —4 —20 2 4 68PrncipaI component #1Acoustic probe6. WaveguideFigure 50. Principal components analysis of the acoustic emission signalsacquired during the non-isothermal crystallization ofbenzoic acidusing either the waveguideor the acoustic probe.4.2.5 PolymerizationThe results indicated that acoustic emission was notproduced during thechain polymerization reactionundertaken even thoughpolystyrene was formed.For the Nylon-6, 10 experiments, a few signals were detected but these werecaused by liquid drops falling off the nylon rope and back into the beaker with the182signals corresponding tothe drops hitting the surface of the liquid. It was concludedthat this polymerization process did not produce acoustic emission.These results were in agreement with work subsequently carried out at UBCby Shiundu, studying theformation of polyurethanes [210].4.2.6 Plasticizationof PVCThe results indicated that acoustic emission was not detected duringplasticization. As the PVCwas observed to swell on the addition of the plasticizerthen it was expected thatacoustic emission wouldbe produced due to thelargevolume change. Navratilreported the detection of acoustic emission during swellingexperiments using ion exchange resins and molecular sieves. This was attributedto gas release [113]. Thus there was no evidence to contradict the results givenhere.1834.3 Polymer studies using thermosonimetry4.3.2 Polymer studiesThe amount of acoustic emission collected was dependent on the polymerused. PVC and PET gave the most acoustic emission during thermo-oxidativedecomposition while PE andPP gave very little acousticemission under the sameexperimental conditions (Fig. 51).200 mg PVC-dI200 mgPE>II200 mg PP0• IU)o iiIo200 mg PET300 400 500 600 700 800Sample temperature (K)TS curves for the thermo-oxidative decomposition of the polymers,PVC, PE, PP and PET.Figure 51.184The total number of acoustic emission signals andthe total cumulativeacoustic RMS collected during the experiments were reproducible for eachpolymerand further substantiatedthis trend (Table 17).Number of individual acoustic Total cumulativeemission signals.acoustic RMS (V)PVC 1233, 1605, 1326 1.88,2.23, 1.88Avg.= 1388Avg.= 2.00PE 64, 87, 1050.0583, 0.0757, 0.0997Avg.= 85Avg.=0.0779PP 36, 34, 270.0338, 0.0424, 0.0502Avg.= 32Avg.=0.0421PET 892, 817, 7001.23, 1.06, 0.837Avg.= 803Avg.= 1.04Table 17. The number of AE signals acquired and the total cumulative acousticRMS as a function of thepolymer.The frequency content ofthe acoustic emission signals was similar for allfour polymers (Fig. 52). The relative intensities of the most prominent frequenciesfor the acoustic emissionsignals from the PP experiments were greater thantheintensities of the most prominent frequencies for theother experiments thoughfewerindividual signals were collected during the PP experiments.The results of regressionanalysis of the average power spectra obtained forthe four polymers are shown in Table 18. In general, the mean coefficient ofdetermination between trials using the same polymer was low, which precluded anyreal discrimination of the average power spectra fordifferent polymers.185Figure 52.>‘4‘-4U)ci)ci)• 4-4--4ci)The average power spectra for the thermo-oxidative decomposition ofPVC, PE, PP and PETin air.0.020.010.000.020.010.000.020.010.000.020.010.00-PET0500Frequency ([0C)C,.]Polymer2— PE, 3— PET,Median descriptor values for the four polymers used.(A) ‘RMS’, (B) ‘Frq @Max’,(C) ‘F-Crest’, and (D)‘DFB5’.The abscissa is not anumerical scale; the given numbers are labelsfor the polymers. The figure serves to show the reproducibilityofdescriptor values, andno trends are to be expected.A-aaa-a Ca Co 1 2 3 40.090.080.070.060.050.044.03.53.02252001751 501251001 .21.00.80.60.40.2• B aaeCaaa0 1 2 34CDaaaBCaCaCB aaaa CCC0 1 2 34II0 1 2 341— Pvc,Figure 53.4—PP188Using median descriptor values itwas possible to distinguish the acousticemission signals when different polymers were decomposed. However,comparisons of the individual descriptor values did not give similar separations.Principal components analysis didnot indicate any differences between the acousticemission signals obtained using the different polymers. Thus it must be concludedthat the differences observed areonly found when using the medianvalues.432.1 Temperature-dependentbehavior of PVCTwo separate reactions producingsignificant amounts of acoustic emissionwere detected during the thermo-oxidative decomposition of PVC (Fig. 51). Thesewere associated with dehydrochlorination of the PVC and the subsequentcarbonization of the dehydrochiorinated residue. The characteristics of theindividual acoustic emission signals from the two processes were determined to seeif differences were apparent since the processes giving rise to the two regions ofacoustic emission were so physically and chemically different.Individual acoustic emission signals from each of the three PVC trialsweredivided into two groups according totheir time of acquisition. This ledto 240, 537and 285 signals being produced during the first reaction, equivalent to19, 33 and22 % of the total number of acoustic emission signals. Closer examination ofFigure 51 suggested that the intensity of the acoustic emission signals(measuredas acoustic RMS) was generally lower for the first reaction.The average power spectra of the two groups of signals for the one of thetrials is shown in Figure 54. A number of differences are apparent. Theaveragepower spectrum for the first reactionstage contained high intensities at frequenciesaround 100 kHz (Fig. 54A). The second power spectrum showed a decrease in the-,—C,)a.)a)-a)Figure 54. Average power spectrafrom the two regions of AE activity during thethermo-oxidative decomposition of PVC.(A) Lower temperature reaction stage.(B) Higher temperature reaction stage.The results obtained from regressionanalysis using the six average powerspectra (i.e., two for each trial) are shown in Table 19. Significant differencesexisted between the results for (1) andthose for either (2) and (3) below. Thus theaverage power spectra of each of the two reactions identified were notcharacteristic of the process giving riseto the acoustic emission as the meanr2 value between the first and secondreactions of the same trial was higherthan themean r2 values between two average power spectra corresponding to the samereaction but from different trials.189intensities of these frequencies suchthat they were now comparable to theintensities of the higher frequency components between 150 and 300 kHz.0.030.020.010.000.0100.0050.0000 5001000Frequency (kHz)190mean r2 value(1) 1st and 2nd reactions, same trial0.66 ± 0.07 (nc = 3)(2) 1st reaction, 2 different trials0.5 ± 0.2 = 3)(3) 2nd reactions, 2 different trials0.5 ± 0.1 (n = 3)(4) 1st and 2nd reactions, 2 differenttrials 0.4 ± 0.2 (n = 6)Table 19. Regression analysis of theaverage power spectra from the tworeactions of the thermo-oxidative decomposition of PVC.Median descriptor values were reproducibly different for the majority of thedescriptors when comparing signals acquired during the two reactions. Principalcomponents analysis was undertaken, using only those descriptors for which adifference had been found. Two-hundredsignals were selected at random fromeach reaction. A partial separation of thetwo classes of signals was found withsignals acquired during the second reaction having lower 1st principal componentvalues (Fig. 55). Similar results were obtained for the other trials and also for anidentical analysis using all the signals collected during the trial. Here, theseparation was more difficult to detect.Factor analysis indicated that the major contributors to the I st principalcomponent were the ‘Frq-Med’, ‘FqrILBW’ and ‘DFBJ’ descriptors. Higher values ofthe first two descriptors were found for signals collected during the second reactionwhile ‘DFBJ’ had lower values for the second reaction. On the basis of theseresults, it was concluded that higher frequency components are more prominent inthe signals from the second reaction. Thisis in agreement with the average powerspectra.I.Ca)C0E0U0(-)C0191I I I0Tv0V06-4-2-0-—2 -—4 -—6 -00 0000,0c8 00V0V 0 •‘VT,. Q700—6 —4 —2 0 24 6PrincpaI component #1First reaction stage0 Second reaction stageFigure 55. Principal components analysis ofthe acoustic emission signalsacquired during the two reactions found from the thermo-oxidativedecomposition of PVC.Two methods were used to assess whether the signals produced by the tworeactions could be discriminated using signal classification methods. Previously the‘d-quartile’ method had been used to successfully classify acoustic emission signalsoccurring during the thermo-oxidative degradation of PVC as either noise or realacoustic emission signals [144]. Using this method, inadequate discrimination wasfound between the two types of signals as thecalculated resolution values werealways less than unity. Cluster analysis also didnot give a satisfactory192discrimination between the two classes of signals postulated here. A ‘fuzzyclusteranalysis’ method [142] is better suited for this type of analysis. In this type ofmethod, the probability of the signal belonging to each class is used rather thansetclassifications. However, no further attempts were made to distinguish the acousticemission signals from the two reactions, though these results indicated someoveralldifferences.4.3.2.2 Determination of kinetic parameters from the thermo-oxidativedecomposition of PVC and PETIn Eqn. 29, the general form of the kinetic expression used in both dynamicand isothermal thermogravimetry was given as :-dc (Ea’A. exp Wja)If the heating rate 3 (dTldr) is constant, then mathematicallydct (‘1’idct= ij) --(Equation 38)Combining the above equations and integrating we getTdot A I (EaF(ot)=J=jexp1,RTJ dT(Equation 39)Many approaches have been used to simplify this expression. Although theright-hand side of Eqn. 39 has no exact integral, Coats and Redfern [211] usedf(ot) = (lot)Q2r) and appropriate substitutions to get the following equation :-u1(1ot)(1ri AR ( 2RT EalogT2(lflr) J = log f3E Ea J - 2.3RT(Equation 40)193For r = 1, the following form was used,C 1-(1-x) AR C 2RTEalog - logT2 J = log I31a1\J 2.3RT(Equation 41)As (1 2RT/Ea) was expectedto be constant for the temperature range overwhich the reaction occurs, the left-hand side of Eqn. 40 orEqn. 41 can be plottedagainst lITto find the activation energy (and the pre-exponential factor). Thisapproach is known as the Coats-Redfern method.Several workers including MacCullum and Munro [212] have discussed theso-called ‘kinetic compensation effecf with respect to thethermal decomposition ofpolymers. This effect was first seen in catalysis studies where it was observed thatdiffering treatments of a catalyst led to a change in the calculated activation energybut not in the rate of the reaction. A relationship between the activation energy andthe pie-exponential factor was proposed, i.e.,log A = a Ea + b(Equation 42)where a and b are compensation parameters, specific forthe system ofinterest. For example, calculated values of a and b for thethermal decomposition ofPET were 0.0854 mol/kJ and-2.22 respectively while for PVC, values of0.0772 mol/kJ and -0.38 were reported [212]. As a resultof this effect, it has beenargued that attaching theoretical significance to experimentally determined valuesof A and Ea is not advised [213]. Throughout this work, thecalculated kineticparameters are used only forcomparative purposes and no theoretical significanceis attached to the values obtained.The Coats-Redfern method was used to analyze the TS data. It was felt thata method in which the kineticparameters were determineddirectly from the194thermosonimetry data without having to calculate the values of daldt ord&dTwouldbe suitable. The applicability of this method was first assessed usingthe three PETtrials and the three PVCtrials. The PE and PPtrials were deemed to haveproduced insufficient acoustic emission for thistype of analysis to be of interest.Values of c from 0.05 to0.95 in steps of 0.05 were used with thecorresponding values of T. ln each case, the value of r was varied from 0.1 to 5using steps of 0.1. The‘correct value’ of r is the value which gives thehighestcoefficient of determination value (r2) [211]. Toensure that the noise associatedwith the data did not lead to erroneously high values of r’ the change in the valueof r2(2)was monitored as was increased.If tr2was less than 0.0005 for achange in n of 0.1 thenit was assumed that the ‘correct value’ of n was theprevious value ofTo check the validity of the results obtained, two different methods wereemployed. Dollimore and co-workers [214] discussed a method for generatingTG curves based on rearranging Eqn. 29 to give(Eadc =A.expRTJf1x) dt(Equation 43)If Ea and A are known, together with the starting temperature T0 and theincrement time de, then dcL can be calculated. This value is then addedto thecurrent value of cx. Knowing the heating rate 13 (KIs) permits the next temperature,T, to be calculated fromT= T0 + f3dt and the whole TG curve can be constructed.In the first method, a curve was generated using the following parameters,f(a)=(1cx)(h2r),Ea 120.0 KJImol, 13=5 Klmin,A = 1x100s- and nr=2.0.Atemperature range of 298 K to 900 K was employed. Using the same proceduresubsequently employed for analysis of the thermosonimetry data, an excellentFigure 56.0rJ0C.)0Comparison of experimental andfitted curves for the thermo-oxidativedecomposition of PET.Ea = 349 kJ/mol, A = 7x1021 s1 and r = 2.8195agreement was found between these ‘theoretical’ values and the calculatedparameter values, i.e., Ea = 119.9 ± 3KJ/mol, A = 9.87x10 S1n = 2.0andr2 = 0.9998. In the second method, a curve based on the calculated parametervalues was directly compared with the TS data for one of the PETtrials. Closeexamination of Figure 51 had indicated that a few acoustic emission signals wereobserved prior to the main reaction during the thermo-oxidative decomposition ofPET. This pre-stage was excluded from the data analysis as theamount of acousticemission was so small.tO0.80.60.40.20.0650 700750 800850Temperature (K)196The residuals obtained from fitting the experimental data were non-randomindicating that an incomplete model for thedata was being used. This may becaused by a second reaction. However, it was felt that the fit obtained wasacceptable to permit comparisons betweenthe results obtained under differentconditions. Table 20 indicates the numerical results obtained for the three PETtrials. On the suggestion of Dr. D.P. Chong, the average nrvalue was calculated forthese results. The values of Ea and A recalculated using this value of r’ (Le., 2.6)are also shown. It was felt that this approach might provide more useful resultswhen comparing Ea values obtained using different experimental conditions.Trial r Ea (kJ/mol) A(-1) mean r2 value1 2.1 256±21 4x1014 0.9772.6 304 ±26 9x1017 0.9732 2.8 349± 11 7x1020.9962.6 327 ± 12 2x1 020 0.9953 2.8 327 ±32 3x1019 0.9652.6 307 ±31 1x108 0.963Table 20. Kinetic parameters for acousticemission from the thermo-oxidativedecomposition of PET using the Coats-Redfern method. The secondrow of values for each trial lists the values obtained when using theaverage reaction order. The given errors refer to the 95 % confidencelimit.Although thermosonimetry could not be used as a comparative measure ofthe mass loss occurring during decomposition, it was felt that by comparing kineticparameter values obtained using both thermosonimetry and thermogravimetry, agreater understanding of the mechanism(s) producing acoustic emission might beobtained. In section 4.5.1, thermogravisonimetry results for the thermo-oxidative197decomposition of PET are given. Using the TG results, Ea values between 214 and264 kJ/mol, an average reaction order of 1.1 and pre-exponential factors between4xlOl3to4x1016s1 were calculated (Table 35). Comparison of the kineticparameter values is complicated, given that all three kinetic parameters have aninfluence on the thermal analysis curves obtained. At a constant heating rate theoverall shape of the thermal analysis curve is determined by rwhile the position ofthe thermal analysis curve on the temperature axis is determined by both Ea andA [215]. An increase in n, leads to a gentler slope of the thermal analysis curve andthus a longer time during which the reaction is occurring at a significant level. Thusthese numerical results confirm the experimental observationthat the acousticemission occurs over a larger temperature range than the mass loss. Comparisonof the Ea and A values is further complicated by the ‘kinetic compensation effect’which was seen in the form of an inverse relationship between Ea and A. Theprocedure used to calculate the pre-exponential factor depended on a knowledge ofEa while the reverse was not true. It was felt that a comparison of the values of Eawould be subject to less error and perhaps be more enlightening with respect to theprocesses occurring. The Ea values using thermogravimetry were lower than theEa values using thermosonimetry.Cooney et a!. [97] postulated that three stages of reaction were associatedwith the thermo-oxidative decomposition of PET in air. Their experiments usedthermogravimetry and a heating rate of 5 K/mm. For the main reaction, use of theCoats-Redfern method gave an activation energy of 218.1 ± 4.8 kJ/mol and areaction order of 3/2. These values differ from those given abovebut notsurprisingly the level of agreement is higher between the two sets ofthermogravimetry values. It must be stressed that the sample mass used here198(200 mg) was 10 times greater than that in the quotedwork and this may accountforsome of the differences observed.The two reactions producing acoustic emission duringthe thermo-oxidativedecomposition of PVC weretreated separately. Typical results obtained using theCoats-Redfern method areshown in Table 21. Again,the average reaction orderwas calculated and the kinetic parameters recalculatedusing this value ofTrial Ea (kJ/mol) A (s-i) Mean r2value1. A 2.5 216±16 4x1017 0.9812.8 239± 17 6x1019 0.980B 2 256± 133x1014 0.9901.9 247 ± 13 6x1013 0.9902. A 2.3 127 ±8 1x109 0.9842.8 151±11 2x101 0.981B 1.6 189 ±91x100 0.9911.9 211±12 5x101 0.9893. A 3.5 270 ±21 3x1 024 0.9772.8 217 ± 19 1x109 0.971B 2 199± 77x1010 0.9951.9 199 ±7 7x101 0.995Table 21. Kinetic parameters for acoustic emissionfrom the thermo-oxidativedecomposition of PVC using the Coats-Redfern method.Process A. - Lower temperature reaction stage.Process B. - Higher temperature reaction stage.The kinetic parameter valuesdetermined for the thermo-oxidativedecomposition of PVC varysomewhat, especially for thefirst reaction. This is nottoo surprising since the amount of acoustic emission collected during the firstreaction is far less than that collected during the secondreaction. The first reaction199is less defined and variations from ideality are more likely. The result obtained forthe first reaction in the thirdtrial where the calculated values of r and Ea weresignificantly higher than the other trials is probably a result of this variation.A comparison was made ofthese results and the thermogravimetry resultsobtained in section 4.4. Forthe latter, it was not possible to obtain values of thekinetic parameters for the second reaction observed. For the dehydrochlorinationreaction, the average reaction order was 2.7, with an activation energy between 344and 361 kJ/mol and a pre-exponential factor between 5x1031 and 2x1 Q33 s. Herea somewhat surprising trendwas observed. In almost acomplete reversal of theresults obtained using PET,the reaction orders were very similar while the values ofboth Ea and A were higher when calculated using the thermogravimetry data. Noexplanation of these trends was apparent.Activation energies for the dehydrochiorination of PVC inair are widelyreported in the literature. A general survey of the relevant papers referenced in thisdissertation found that typicalvalues of Ea ranged from 60to 200 kJ/mol. Values ofthe pre-exponential factor were given less frequently but arange of lxi oto 3x1 s1 was observed.Reported values of the reaction order ranged from 1.0to 3.5. Comparing the kineticparameter values determined here usingthermogravimetry and the literature values for the sameprocess suggests that thevalues determined in the former case appear erroneous.A further discussion of thisdiscrepancy is given in section 4.4. Typically, the activation energies determinedusing thermosonimetry were higher than the values reported in the literatureobtained by thermogravimetry. This is not surprising giventhat acoustic emissionwas not produced until a significant amount of mass loss had been observed.The use of an average reaction order for each set of trialsseemed to bebeneficial when comparing theresults obtained under different experimental200conditions. In subsequent analysis usingthe Coats-Redfern method, this procedurewas employed exclusively.4.3.3 Effect of sample mass10mg Pvc300 400 500 600700 800Sample temperature (K)Figure 57. Effect of sample mass on theTS curves for the thermo-oxidativedecomposition of PVC in air. Sample masses were 200, 100, 75, 50,25 and 10mg.200mg Pvci I II——100mg pvcIII75mg pvcC)Cl)C)Cl)CC)‘fijihdIL. hi_IIIry50mg pvcI I .iii25mg vciiULIIlII1II ti 1 I_______________201The amount of acoustic emission was dependent on the sample mass beingused (Fig. 57). However, a definite quantitative relationship between the samplemass and the amount of acoustic emission was not apparent.The amount of AE activity decreased as thesample mass was loweredthough this was more apparent for the firstregion of AE activity than for thesubsequent region of activity. Figure 58 indicates the total cumulative acousticRMS as a function of the sample mass. The three blank experiments gave noacoustic RMS and are included in the plot.2.5 I I Ia 02.0 0-CA 00(1) 1.5-D0C-)08>o 1.00E000.5 8C—0.0I II I0 25 50 75 100 125150 175 200Sample moss (mg)Figure 58. The total cumulative acousticRMS as a function of the sample massfor the thermal oxidative decomposition ofPVC in air. The darktriangle represents two overlapping values.The reproducibility of the results was low. Itwas surprising that the totalcumulative acoustic RMS values were similar for sample masses of 10 and 25 mgwith the sample mass of 10 mg giving a slightly higher value. One of the trials using202a sample mass of 50 mg also gave a similar value, further illustrating the lowreproducibility of the results. Forthe trials using a sample mass of 10 mg, thesample did not completely coverthe bottom of the waveguide. The additionalacoustic emission detected mayhave been due to an increase in the surfacearea-to-mass ratio. There was anincrease in the total cumulativeacoustic RMS asthe sample mass was raised between 25 and 75 mg. Above 75mg, no real changewas observed. Exact reasons for this attenuation are unknown.An increase insample thickness may account for the attenuation due to absorption of the acousticemission by the sample. Use of the ‘t-test’ indicated that there were significantdifferences in the total cumulative acoustic RMS at the 95 % confidence levelbetween the three highest masses and the other three masses.However the lowreproducibility of the results obtained using a mass of 50 mg meant that the resultsusing 50, 25 and 10 mg could not be discriminated.Analysis of the individual acoustic emission waveforms indicateda similartrend in the total number of individual waveforms acquired as a function of thesample mass. In this case, the values obtained using masses of 25 and 10 mg werenot significantly different. Figure59 shows the results. Again, the three blankexperiments are included.Attenuation at the higher samplemasses may again be due to anincrease insample thickness. However, useof a I second acquisition delayalso affected thenumber of individual acoustic emission signals collected during theexperiment. Inthe experiment in which no acquisition delay was set, a total of 2163 acousticemission signals were acquired. This was significantly higher thanthe number ofsignals collected using a I seconddelay for the same mass (i.e., 1233,1600 and1310) and thus part of the non-linearity observed may be explainedby the use of anacquisition delay.2031800-C1600acC01400 aa aC1200aC)>o 1000S B800-aa60 B II I I I0 25 50 75 100 125 150 175 200Sample mass (mg)Figure 59. The number of acoustic emission signals collected as a function ofthePVC sample mass.Overall, these results were a little disappointing. Previous workers includingBelchamber [115] and Shimada [165] have demonstrated the utility of acousticemission for quantitative analysis but inthis case, the same conclusion was notforthcoming. The linear range for a plot of the amount of acoustic emission as afunction of the sample mass was very low and even then the reproducibility oftheresults was low. In the case of the acoustic RMS measurements, this lack ofreproducibility was such that the valuesfound at 25 and 50 mg were notsignificantly different. Furthermore, thetrials using a sample mass of 10 mgindicated that the surface area was probably a factor in determining the amountofacoustic emission detected, seriously questioning the use of thermosonimetryforquantitative analysis using the apparatusdeveloped.Coats and Redfern [216] reported that three physical effects may beobserved when the sample mass is increased in thermal analysis. As the sample204mass is increased, the deviation of the sample temperature from alineartemperature change will increasedue to the endothermic or exothermic reactionsoccurring in the sample. In addition, a larger sample mass may lead to thermalgradients in the sample and hinder the diffusion of any gaseous products from thesample. Wendlandt [217] summarized the practical implications ofthese effects onTG curves. No changes in the onset temperature with an increase in sample masswere expected. For endothermicreactions, the reaction was expected to continueto a higher temperature when a larger sample mass is used. For exothermicreactions the heat produced by the reaction and the subsequent increase inreaction rate was expected to compensate for the increase in sample mass and thusno effect was indicated. For DTA,it has been reported that the peak minimum of anendothermic reaction is shifted toa higher temperature when a larger sample massis used but there was only a minimal effect observed for the peak maximum of anexothermic reaction [218]. Furthermore for DTA, the area under the transition peakis directly proportional to the sample mass.As the thermo-oxidative decomposition of PVC is an exothermic reaction nochanges in either the temperatureof the maximum rate of reaction (T) or thetemperatures at which the acoustic emission commenced (Ii) or ceased (Tf) wereexpected. Table 22 indicates these characteristic temperatures asa function of thesample mass. Values are given to±5 K. As the sample mass wasdecreased from200 to 10 mg, changes were observed. For the first acoustic emission reaction, theTi values were similar for sample masses between 50 and 200 mg. For the samplemasses of 25 and 10 mg, the T1 value increased. Due to the low reproducibility ofsome of the results, the same trendcould not be presumed for the values of T andTf. For the second reaction, analogous results were observed for the 7 values. Forthe T and Tf values, no apparent change was found as a result of changing the205sample mass. The changes in the 1 values at lower sample masses may resultfrom a decrease in the contact between the walls of the sample holder and thesample leading to a decrease in the sample temperature compared to the expectedsample temperature.Mass(mg) — T1(K) T(K) Tf(K)200 A 500-520 520-555 600-670100 A 530-540 585-605 625-64075 A 520-540 540-560 620-63050 A 520-540 560-565 620-63025 A 555-575 585-620 650-66010 A 570-590 600-625 645-700200 B 685-715 770-800 850-860100 B 690-710 780-800 850-86075 B 700-710 800-815 850-86050 B 690-710 760-790 840-86025 B 720-730 770-775 825-84510 B 730-755 775-805 840-860Table 22. Characteristic temperatures for the thermo-oxidative decomposition ofPVC using thermosonimetry as a function of sample mass.Process A. - Lower temperature reaction stage.Process B. - Higher temperature reaction stage.206The intensities of the individual acoustic emission signals were affected by achange in sample mass. Close examination of Figure 57 indicated that similaracoustic RMS values were found for sample masses of 75, 100 and 200 mg. Adecrease in the average acoustic RMS values was evident when the sample masswas 50 mg and a further decrease was found when sample masses of 25 and 10 mgwere used. These changes were assessed further by looking at those descriptorsrelated to the intensity of the individual acoustic emission signals (i.e., ‘RAilS’, ‘Peak’,and ‘Area’). However, the median descriptor values for the intensity descriptors didnot indicate any differences in the acoustic emission signals as a function of thesample masses (e.g., Fig. 60).0.060 I I I I I I I00.058 080.055> 000U) 0 0c o.os.0.2 00ILl 000.050 000.0480.045 I II0 25 50 75 100 125 150 175 200 225Sample mass (mg)Figure 60. Median ‘RMS’ descriptor values as a function of the sample mass forthe thermo-oxidative decomposition of PVC in air.207The average power spectra from the experiments usingvarious samplemasses were similar. Table 23 displays the correlation coefficients obtained whencomparing the average power spectra from each experiment.200&200 0.6±0.1100&100 0.8±0.375&75 0.93±0.0250&50 0.93±0.0525&25 0.88±0.0510&10 0.8±0.2200& 100 0.52±0.09200 &75 0.55±0.08200&50 0.65±0.07200 &25 0.60±0.08200&10 0.3±0.1100&75 0.84 ±0.07100&50 0.70±0.05100&25 0.66±0.07100&10 0.71±0.0875&50 0.80±0.0575&25 0.73±0.0575&10 0.76±0.0450&25 0.85±0.0250&10 0.63±0.0525&10 0.59±0.06The extensive results presented did not suggest an abrupt change in theaverage power spectra when the sample mass was decreased. However, closerexamination of the results indicated that the highest mean r2 values were generallyfound either when comparing the spectra of two trials using thesame mass or whenMean r2 value Mean r2 valueTable 23. Regression analysis of the average power spectra obtained usingvarious masses of PVC.208using similar sample masses. For sample masses of 50, 25 and 10 mg, this trendwas followed completely while for the other sample masses, typically one value didnot fit the trend. Only for a sample mass of200 mg was this trend not apparent. Nodetailed explanation can be given for this conclusion though again the samplethickness may be contributing to the nature of the acoustic emission detected, allother factors being equal. A sample massof 50 mg was used, where possible, forthe remainder of the experiments as the results reported here indicated that usingthis mass gave a high number of acousticemission signals.The kinetic parameters for the acoustic emission were calculated using theCoats-Redfern method to assess whetherthere was any significant changes as thesample mass was decreased. The rangeof results obtained using sample massesof 100, 75, 50, 25 and 10 mg are shown inTable 24. The individual results using asample mass of 200 mg were shown earlier(Table 21). As discussed earlier, theaverage reaction order was found for each set of replicates and then the Ea and Avalues were calculated using this r value.As the sample mass is decreased, changesoccurred to the activationenergies calculated for the second acousticemission reaction. For masses of 200,100 and 75 mg, the results obtained are very similar with the value of Ea being in therange of 172 to 247 kJ/mol. For sample masses of 50, 25 and 10 mg, the calculatedvalues of Ea are generally higher though no relationship between the sample massand an increase in activation energy was found. The average reaction orderscalculated for the second acoustic emissionreaction were very consistent with arange of 1.3 to 1.8, encompassing all the calculated values.209Mass (mg) Avg. n Ea(kJImol) A (s1)200 A 2.8151-239 2x101-6x19 >0.971100 A 1.6146-188 3x1010-2x14 >0.97175 A 3.0 222-271 2x1018-3x12>0.94650 A 2.2 190- 233 5x1 014 - 2x1Q19 > 0.95525 A 2.1176-264 8x1012-3x120 >0.97110 A 2.4198-237 2x1014-4x17 >0.973200 B 1.6199-247 5x101-6x13 >0.989100 B 1.3172 - 207 6x18- 2x1 011 > 0.96075 B 1.5 175-193 1x109-2x10>0.98650 B 1.7 252- 342 8x1 012 - 2x1021 > 0.97125 B 1.7360 - 436 5x1 021 - 2x1027 > 0.99610 B 1.8280-389 8x1015-1x123 >0.993Table 24. Kinetic parameters for acoustic emission from the thermo-oxidativedecomposition of PVC using the Coats-Redfern method.Effect of sample mass.Process A. - Lower temperature reaction stage.Process B. - Higher temperature reaction stage.For the first acoustic emission reaction, the variation in the results obtainedat any one sample mass precluded the observationof any trends in the reactionorder or activation energyas a function of sample mass. Based on the discussionof the effect of changing the sample mass on the characteristic temperatures,common sense would expect that the activation energy for an exothermic reaction210would not be a function of thesample mass. At low samplemasses, the observedincrease in the activation energy for the second acousticemission reaction mayresult from the aforementioned decrease in actual sampletemperature.In thermosonimetry the TS curve is defined directly by the number of signalsacquired as the curve is constructed by the summation ofthe intensities of all theindividual signals. In TG andmany other TA techniques the recorded measurementis a direct indication of the extent of reaction and thus thenumber of measurementsis less critical. As the sample mass was decreased the number of acousticemission signals acquired decreased. As a result the curve became lesswell-defined, and the accuracy of the model used to calculate the kinetic parameterswould also be poorer at lower masses. However, the level of agreement betweenthe model and the experimental data was sufficient to permit a detailed analysis ofthe data. This explanation may account for some of the variability observed in theparameters obtained for the first acoustic emission reaction where the number ofsignals acquired was low regardless of the sample mass.4.3.4 Molecular mass studiesThe TS curves for six PVC resins used here were similarto the results shownin Figure 51 for a sample mass of 50 mg. No differences were observed in theintensities of the signals whenthe resin was changed. Preliminary resultsdiscussed earlier were inconclusive as to whether the characteristics of the PVCresin influenced the acousticemission detected. Figure 61indicates the number ofindividual acoustic emission signals collected as a function of the viscosity-averagemolecular mass of the PVC resin.211/_7 /1300-AC o1200C0.20U) o1100 0(J0U) 1000Co0(_)0 0C 0900 -S.-0VE 800°Cz700 I //40 60 80100 180 200Vscosity—overoge moIecuor moss (Mv/i 000)Figure 61. The numberof acoustic emission signals as a function of theviscosity-average molecular mass for the six PVC resins. The darktriangle represents two overlapping values.Minor differences were observed between the number of individual acousticemission signals collectedfor different PVC resins but there was no relationshipfound between the numberof signals and the molecular mass of the polymer.These results agree with those shown in Table 12.In Figure 30, some differences in the onset temperature of the acousticemission were observed forthe four different PVC resins used at that stage of theinvestigations. Using six PVC resins and the optimized thermosonimetry apparatusno differences were apparent (Fig. 62).212_______////0600 -0000 00A00ai3000550 0‘V0F-0500 I//—40 60 80 100 180 200Viscosity—average moecuar mass (Mv/i 000)Figure 62. The onsettemperature of acousticemission as a functionof themolecular mass of the PVC resin for the thermo-oxidativedecomposition of PVC.The dark triangle represents two overlappingvalues.Thermal stability results, given later in sections4.6.3 and 4.6.4, indicated thatthe onset of thermal dehydrochiorination was not a function ofM. However, nodirect relationship between the thermal stability of PVC and the onset temperatureof acoustic emission wasfound using either of these measures of thermal stability.This was not surprising given the inherent variability in the onset temperatures foracoustic emission (Fig. 62). This lack of reproducibility precluded the use ofthermosonimetry as a suitable indicator of the thermal stability of PVC.Analysis of the averagepower spectra indicateda number of differencesbetween some of the spectra obtained when different resins were used (Table 25).Statistical analysis of theresults using the ‘t-test’,indicated that the average power213spectra from PVC resins ‘#1’ and‘#3’ were different to the other power spectra butcould not be distinguished from each other.The average power spectra for the one trial of each of the six PVCresins areshown in Figure 63. No major differences between the average power spectra ofPVC resins ‘#1 • and ‘#3’ and the average power spectra for the remaining four PVCresins were seen. It was concluded that the average power spectra did not indicatedifferences in the resins.#1  0.7 ±0.3#1 0.4±0.1#1  0.83 ±0.05#1 0.4±0.1#1 0.4±0.1#1  0.36 ±0.05#2 0.9±0.3#2 0.37±0.05#2 0.83 ±0.05#2 0.84±0.03#2 0.7±0.1#3 0.88±0.07#3 0.39±0.07#3 0.43 ±0.05#3 0.44±0.05#4 0.8±0.3#4 0.84±0.05#4 0.6±0.1#5 0.87±0.05#5 0.69 ±0.08#6 0.6±0.3Mean r2 value Mean r2 valueTable 25. Regression analysis ofthe average power spectra obtained from thethermo-oxidative decomposition of the six PVC resins.214Figure 63.U)4)000C’)ci)ct5The average power spectraof the acoustic emission from the six PVC0 5001 000Frequency (kHz)resins used.215Analysis of the descriptor values indicated that for some oftheintensity-related descriptors achange in the median value was observed as themolecular mass increased. Results are shown below for the‘RAilS’ and ‘1/8 t’descriptors. The ‘Peak’, ‘Area’and ‘10-Cross’ descriptors gave similar results to thoseseen for the ‘RAilS’ descriptor. No trend was observed for the other descriptors.AB0.065-1.25-a) 1.20- 0ct 0.060 ->0 1.15-s-I000.055 - 0 001.10 -511000 081.05- 0C) 0•050-Cl)1.00-a)0.045 - I I0.95 -__________________40 80 120 160 20040 80 120 160 200Viscosity—average molecular mass (Mv/i 000)Figure 64. Median AE descriptor values as a function of the viscosity-averagemolecular mass for the six PVCresins.(A) ‘RAilS’ and (B) ‘1/8 t’ descriptors.These results did not agree with those shown in Figure 32 but it was felt thatof the two sets of data, theseshould be the more accurate.Principal components analysis was undertaken using only those descriptorswhich gave a trend, i.e., ‘RAilS’,‘Peak’, ‘Area’, ‘10-Cross’ and ‘1/8t’. One-hundredacoustic emission signals wererandomly selected for each PVC resin. The resultsindicated no differences between the PVC resins and thus it was concluded that any216differences seen between the median descriptor values were not applicable whencomparing the individual acoustic emission signals.Kinetic parameter values were calculated using the Coats-Redfernmethod (Table 26). No apparent differences were found in the calculated reactionorders for either of the two acoustic emission reactions. The values obtained didshow a good degree of reproducibility especially for the second reaction.For the activation energies, the values obtained for the first acousticemission reaction were variable and no real trends were observed. For the secondacoustic emission reaction, the PVC resin of lowest molecular mass had the highestEa values. The Ea values obtained using the other five resins weresimilar.Comparison of the TS results with TG results (see section 4.4.1) for the first processindicated that the reaction orders and apparent activation energies were similarthough in many cases higher for the latter results.Changes in the particle size of a sample can affect the TA curve to a largeextent [217]. In general, small particles reach equilibrium to a greater extent duringthermal decomposition and thus at a given temperature the degree ofdecomposition will be greater [216]. Although no accurate particle sizedeterminations were undertaken on the PVC resins used here, it was known fromsemi-quantitative measurements that variations in the particle sizes existed(Table 9). In particular, PVC resin “#5” was known to be more heterogeneous thanthe other resins. However, the low reproducibility of the onset temperatures foracoustic emission (Fig. 62) prevented any discrimination of the various particlesizes on the basis of onset temperature. Similar results were obtained when usingthe T and Tf values for the two processes.217M Avg.n Ea(kJ/mOl) A(s1)(g/mol)58 100 A 2.2 190-233 5x1 014 - 2x1 019 > 0.95559 200 A 1.8 173- 250 lxi 013 2x1 019 > 0.97680000 A* 2.1 3102x105 0.97783900 A 1.4 231 -318 4xl018-2x16 >0.96088 200 A 2.1 199-320 2x1 015 - lxi 026 > 0.968188 000 A 1.7 203-288 3x1015 -2x102 > 0.99758 100 B 1.7 252-342 8x1 012 - 2x1Q21 > 0.97159 200 B 1.3 183-243 3x109 - 7x1 014 > 0.93180000 B 1.3 140 -162 1x107-4x8 >0.96983900 B 1.1 135-183 3x106-8x19 >0.95188 200 B 1.1 166- 182 2x109 - 3x1 017 > 0.938188000 B 1.0 174-225 2x108-7x1 >0.931Table 26. Kinetic parametersfor acoustic emission from the thermo-oxidativedecomposition of PVC using theCoats-Redfern method. Effect ofmolecular mass.* For PVC resin #5, the first reaction was of a lowintensity in two of the three trialsand the Coats-Redfern methodcouldnot be used.Process A. - Lower temperaturereaction stage.Process B. - Higher temperaturereaction stage.2184.3.5 Effect of heating rateThe acoustic emissionprofile was highly dependent on the heatingrate. Atthe faster heating rates, more than two regions of activity were found while at ratesof I K/mm and especially 3 K/mm only two regions of activity were present (Fig. 65).Wendlandt [194] reported that, for a single-stage reaction, the 7j value (or initialprocedural decomposition temperature) is greater when a higher heating rate isused. Examination of the figure did not revealthis trend and this lackof correlationwas confirmed by numerically determining theonset temperatures for the acousticemission in each case.The temperature rangeover which the reactionis foundwas reported to be higher for a faster heatingrate [194]. This trendwas seen forthe acoustic emissionresults, if a comparisonis made of the resultsusing thefastest and slowest heating rates.In DTA and DSC, an increase in the heating rate leads to a greaterpeakheight for the transition[219]. Applying this analogy to thermosonimetry wouldsuggest that at the faster heating rates the acoustic emission would be of greaterintensity than that detected using slow heatingrates. Close examination ofFigure 65 suggested that the average intensityof the acoustic emission signals wasgreater when a faster heating rate was used, in agreement with theproposed idea.‘a)>SC-)• -4(1)0C-)300 400 500 600700 800Sample temperature (K)219Figure 65. Effect of heating rate onthe TS curves for the thermo-oxidativedecomposition of PVC. Heating rates used were 10, 7, 5, 3 andI K/mm.220300000(02 2500 --(00CO2000-°-3 1500-0 0o 00 VE 1000 --0805000 24 6 810HeaUng rate (K/mm.)Figure 66. The number ofacoustic emission signals collected as a function of theheating rate for the thermo-oxidative decomposition of PVC.The number of acoustic emission signals collected during the experimentsusing different heating rates varied with the heating rate (Fig. 66). The greatestnumber of acoustic emission signals was found at the slowest heating rate. Therewas an exponential relationship between the number of signals collected as afunction of the heating rate, i.e.,Avg. N 881 + 3297.exp(12)(r2 = 0.963) (Equation 44)where Avg. N is the average number of acoustic emission signals acquiredusing a heating rate f3 (K/mm). Results obtained from two of the three trials using aheating rate of 7 K/mm were much higher than expected andthus restricted thepractical use of this relationship. Overall, the increased acquisition time at slow221heating rates compensates for the decreased temperature range for acousticemission using these heating rates. Similar results were found for the totalcumulative acoustic RMS.Comparison of the average power spectra obtained using the five heatingrates indicated few differences. The mean r2values are shown in Table 27.Statistical analysis using the ‘t-test’ indicated no significant differences between thevalues obtained at the 95 % confidence level.Thus there was no change in theoverall frequency content of the acoustic emission signals when the heating ratewas changed.10&10 0.8±0.110&7 0.72±0.0810&5 0.78±0.0410&3 0.6±0.110&1 0.86±0.037&7 0.6±0.37&5 0.7±0.17&3 0.6±0.17&1 0.7±0.15&5 0.93±0.055&3 0.63±0.095&1 0.77±0.023&3 0.6±0.33&1 0.6±0.21&1 0.9±0.1Analysis of the individual acoustic emission signals indicated a number ofsurprising results. Median values for many of the intensity-related descriptorsMean r2 value Mean r2 valueTable 27. Coefficients of determination obtained by regression analysis of theaverage power spectra using the five different heating rates.222including ‘RMS’ and ‘10-Cross’ were found to be greater for the three trials usingaheating rate of I K/mm than those values for the three trails using a heating rateof10 K/mm. The discrepancy between these results and those observed earlier forthe results using the cumulative acoustic RMS may be explained by considering thedifferences between the two modes of acquisition. The individual acoustic emissionwaveforms were acquiredusing a I second delay while the acoustic RMS data is acontinuous measure. Thusthere is the possibility thatthe individual waveformsacquired are not a completerepresentation of the nature of the acoustic emissionsignals being produced.The kinetic parameters calculated using the Coats-Redfern methodsuggested a number of differences as the heating rate was changed (Table 28).For heating rates of 10, 5 and 3 K/mm, the average reaction orders for the firstprocess are higher than theother average reaction orders calculated for the sameprocess. No discernible trends were observed for the range of calculated apparentactivation energies as a function of heating rate, though this range was large. Forthe second process, the reaction orders were similarregardless of the heating rate.At a heating rate of 10 Klmin, the Ea values calculatedwere lower than the valuescalculated using the other heating rates.In section 4.4.2, kinetic parameters were determinedfor the thermo-oxidativedecomposition of PVC using the same heating rates but employing TG. In general,the average reaction ordersand apparent activation energies were lower for the firstprocess using TS. For example, using a heating rate of 7 K/mm, the apparentactivation energies were between 148 to 151 kJ/mol for TS and between 274 and294 kJ/mol using TG. Forboth experimental techniques, the apparent activationenergies were typically lower using faster heating rates although no differenceshould be expected.223Vasile and co-workers [77] reported that the apparent activationenergy forthe thermo-oxidative decomposition of LDPE increased as theheating rate wasraised. This was attributed tochanges in the decomposition mechanism. A similarexplanation may be found to account for the variability observedhere.Heating rate Avg. r Ea(kJ/mol) A (s-i)r2(Kimin)10 A 2.5 181-207 4x1014-1x7 >0.9417 A 1.6 140-153 2x1010-3x101 > 0.9535 A 2.2 190 -233 5x1 014 - 2x1 019 > 0.9553 A 2.0 239 -250 3x1 019 - 5x1 020 > 0.980I A 1.1 120-166 2x107-3x12 >0.94010 B 1.2 132-179 1x106-2x19 >0.9117 B 1.8 222-297 4x1012-9x17 >0.9685 B 1.7 252- 342 8x1 012 - 2x1 021 > 0.9713 B 2.0 230 - 280 2x1 014 - 7x1 016 > 0.9941 B 1.7 249 -268 3x1 015 - 3x1 016 > 0.996Table 28. Kinetic parametersfor acoustic emission from the thermo-oxidativedecomposition of PVC using theCoats-Redfern method. Effect ofheating rate.Process A. - Lower temperaturereaction stage.Process B. - Higher temperaturereaction stage.Multiple heating rate methods have been widely applied to obtainkineticinformation from TG curves. Suchmethods require that the processis independent224of the heating rate, e.g., activation energies and reaction orders are largelyunaffected by a change in heating rate. The results in Table 28 suggested that forthe two acoustic emission processes observed during the thermo-oxidativedecomposition of PVC, this may not be true,it was felt however, that the use ofmultiple heating rate methods for the determination of kinetic parameters might stillbe of use here.4.3.5.1 Determination of kinetic parameters using multiple heating ratemethodsPreviously the use of an integral method enabledapparent activationenergies to be calculated for the acoustic emission data based on Eqn. 29.Differential methods have been extensively usedfor the same purpose though suchmethods were not tried here due to the aforementioned problem in determining thetrue values of dcx!dt or da/dT for the acoustic emission data.The maximum rate observed for the reaction occurs when d(dcddt)/dI is zero.Differentiating Eqn. 29 with respect to time and assuming a constant heating rate f3leads to the following expression after further rearrangement.Eai3 (Ea2 expl RT A g(c,)(Equation 45)1’’mr mnwhere cmn refers to the value at the maximum rate. If g(cL), thedifferentiated form of f(cmn), is assumed to beindependent of f3 then the followingexpression is valid,d[1n(/Tmn2)J Ead [11 Tmn2(Equation 46)225This is the basis of Kissinger’s method [220]. Day and Budgell [221] amongothers, have criticized this procedure due to the inherent problems in determiningT accurately. However, Hunderi and Lønvik [162] reported activation energies forthe crystallization of metallic glasses using thermosonimetry and this method.Therefore, the Kissinger method was used to determine activation energies for thetwo processes producing acoustic emission during the thermo-oxidativedecomposition of PVC (Fig. 67).Kissinger’s method was partly successful. For the lower temperatureprocess, there was a lack of reproducibility of the results and no trend wasobserved. For the higher temperature process, regression analysis of the meanvalues at each heating rate gave an apparent activation energy of 100 ± 46 kJ/mol(r2 = 0.926) for the process. To measure the validity of the results, simulated curveswere generated as discussed earlier in section 4.3.2.2. The same heating rateswere used as during the experimental studies, i.e., 10, 7, 5, 3 and 1 K/mm while thekinetic parameters were as given previously, i.e. an activation energy of 120 kJ/mol.Using the same procedure as employed for the acoustic emission data, theactivation energy was determined as 124 ± 5 kJ/mol with r2> 0.999. Thus a goodagreement was seen between the ‘theoretical’ and ‘calculated’ activation energies.Figure 67. Application of Kissinger’s method to the analysis of acoustic emissiondata from the thermo-oxidativedecomposition of PVC.(A) Lower temperature reaction stage.(B) Higher temperature reaction stage. The triangle indicates twooverlapping values.In light of the aforementioned criticism of the Kissinger method, alternative226(A)000000$6-038-.0—10.0—10.50—11.0—11.5—12.0—12.5—13.00.00168—11.0—11.5—12.0—12.5—13.0—13.5—14.00.001200.00172 0.001760.001801/T (K)(B)0.00130 0.001401/T1)methods were sought which didnot depend solely on one characteristic value227determined from the thermal analysis curve. Ozawa [222] suggested that if theparameters f(c), A and Ea are independent of Tand the latter two parameters arealso independent of c, then Eqn. 39 may be separated and integrated to give;(AEa (Ea”ilogF(cx)=log R J-log 3+logpJ (Equation 47)wherep(u)= (exP(-u)) + J) du (Equation 48)Doyle [223] suggested for (EaIRT) > 20;(Ea” (Ea’1ogp9 -2.315 - 0.4567(Equation 49)This leads to the following equation;(AEa” O.45ó7Ealog F(a) log R J - log - RT (Equation 50)Thus the apparent activation energy can be obtained from a plot of log 13 vs.l/T for a fixed degree of conversion ct. This method isgenerally known as Ozawa’smethod. Using the simulated curves, good results wereobtained from this method.Apparent activation energies ranged from 122 ±3 to 128±6 kJ/mol with a meanvalue of 125 kJ/mol, i.e., slightly higher than the “theoretical value” of 120 kJ/mol.The second process producing acoustic emission datawas analyzed usingOzawa’s method (Table 29). For the first process, resultsare not shown as theinherent variability observed led to almost random values. The activation energy228ct Ea(kJ/mOl) r20.05 121 ±61 0.9320.10 120±68 0.9150.15 122±73 0.9010.20 121 ±77 0.8930.25 119 ±79 0.8850.30 116±78 0.8810.35 112±77 0.8770.40 108 ±77 0.8710.45 103 ±74 0.8640.50 96±72 0.856X Ea(kJImOl) r20.55 92 ±78 0.8460.60 90 ±72 0.8390.65 89 ± 73 0.8320.70 89 ±84 0.8270.75 89 ±76 0.8290.80 89±76 0.8160.85 90±80 0.8090.90 90±82 0.8000.95 92±88 0.782Table 29. Calculated activation energies for the acoustic emission data using theOzawa method. The given errors are equivalent to the 95 %confidence limits.The activation energies calculated using the Ozawa method decrease from120 to 90 kJ/mol as the degree of conversion is increased. The r2 values alsodecreased as the degree of conversion is increased. Regression lines, togetherwith the mean values for each heating rateare shown in Figure 68 for a values of0.05, 0.40 and 0.80. There is an increasing inadequacy of the model used to fit theexperimental data.was obtained from regression analysisof the mean values obtained at eachtemperature.2291 .21 .00.80.600.40.20.00.0011 0.0012 0.0013 0.00151/T (K1)Figure 68. Application of Ozawa’s method to the analysis of acoustic emissiondata from the thermo-oxidativedecomposition of PVC.(Results are from the higher temperature reaction, i.e., Process B).Overall, the kinetic parameterscalculated using the multiple heating ratemethods of analysis appeared tobe an improvement on the values obtained usingthe Coats-Redfern method.4.3.6 Copolymer studiesThe TS curves for the PE and PVAc (Fig. 69) were related to theTG curvesof these polymers under the same conditions (see section 4.4.3). For PE, theacoustic emission commenced ata temperature close to the endof the singledecomposition step observed inthe TG curve. For PVAc, the onset of the firstregion of acoustic emission at Ca.570 K was close to the maximumrate of mass0.0014230loss for the corresponding decomposition step observed. The subsequent acousticemission was split into three or possibly four overlapping regions with noobviousrelationship to the remaining gradualdecomposition observed in the TG curve.COCl)C)Cl)0C)300 400 500 600700 800Sample temperature (K)Figure 69. Effect of vinyl acetate content on the TS curves for thethermo-oxidative decomposition of EVAcopolymers.For both copolymers the main region of acoustic emission was observedbetween ca. 750 and 800 K. Again, the acoustic emission was found to commenceat a temperature close to the maximum rate of mass loss for the correspondingdecomposition step in the TG curves.Closer examination of the TS curves231suggested that the temperature and intensity distributions of the acoustic emissionsignals in this region were similar to those observed for the PE. However, theobserved increase in the amount and relative intensityof the acoustic emission inthis region as the vinyl acetate content was raised suggesting that the presence ofvinyl acetate affected this region of acousticemission. For the lower vinyl acetatecontent copolymer, a small amount of acousticemission was detected betweenCa. 650 and 700 K. This region of acoustic emission was difficult to relate to theother TS curves though the onset temperaturedid occur close to the end of theobserved mass loss for the elimination of acetic acid for this copolymer. For theother copolymer, the remaining region of acoustic emission between Ca. 680 and730 K was also found for the PVAc homopolymer. It was a little surprising that noacoustic emission was found corresponding tothe first decomposition step for thiscopolymer.The use of thermosonimetry to quantify the vinyl acetate content wasassessed by establishing a calibration line for the amount of acoustic emission as afunction of the vinyl acetate content. Both thetotal cumulative acoustic RMS andthe number of individual waveforms collected were used. For the PVAchomopolymer, the amount of acoustic emission acquired was far higher than for theother polymers used here. The average number of individual waveforms collectedwas 1128 and the average total cumulative acoustic RMS was 1.68 V. Theseresults were omitted from the calibration graphsthough unfortunately this left onlythree different copolymer compositions on whichto base the calibration (Fig. 70).232>(.1)C)U)0C)a>aEaUa0F—Ci)a0U)0Ci)U)Ea)L)100The calibration results shown in Figure 70 suggest a linear relationshipbetween the vinyl acetate content in the EVA copolymer and the amount of acousticemission collected in the range of 0 to 40 % vinyl acetate. Regression analysis ofthe mean values gave a slope of Ca. mV per percent vinyl acetate for the acousticRMS data and a slope of Ca. 4 signals per percent vinyl acetate for theindividualwaveform data. These slopes indicate a lack of sensitivity of the method for the0.20.10.00 10 20 30 40Vinyl acetate content (%)2000Figure 70.0 10 20 3040Vinyl acetate content (%)The amount of acoustic emission collected as a function of the vinylacetate content for the thermo-oxidative decomposition of EVAcopolymers.(A) Total cumulative acoustic RMS.(B) Number of individual acoustic waveforms.233determination of the vinyl acetate content as typically the variability of theresultsobtained using three trials of the same copolymer was far in excess of the sensitivitycalculated here. For example, the three trials using the EVA (18 % VA)copolymerresulted in 90, 139 and84 individual waveformsbeing collected. This variabilitywould prevent the routine determination of vinylacetate content using thereportedexperimental procedure,though a number of changes (e.g., use of a slower heatingrate) could be made to improve matters.As the results obtainedusing PVAc were so different to those seen for theother compositions, it seemed unlikely that thecalibration could be extended to thewhole range of EVA compositions even if copolymers with a vinyl acetatecontentgreater than 40 % wereavailable. Closer examination of the TS curve for PVAc(Fig. 69) showed a lot of similarities to the TS curves shown earlier for PVC(e.g. Fig. 51). For the PVC,two regions of acoustic emission were found; the firstbetween 520 and 630 Kand the main region ofactivity commencing at Ca. 720 K.Two corresponding regions were found for the PVAc polymer under thesameconditions. In addition, asignificant amount of acoustic emission was foundbetween Ca. 680 K andthe start of the main region of activity for the PVAc polymer.Closer examination of theTS curves showed that the onset temperature of theacoustic emission for thefirst region of activity was higher for the PVAc polymerwhile the main region ofacoustic emission occurred at a similar temperature.Given that the acoustic emission was not a directmeasure of decompositionbut still was related to decomposition, this change in onset temperature may beexplained according to the difference in the thermal stability of the two polymers.Both polymers degradeaccording to similar mechanisms, initially by theeliminationof HX, where X is Cl for PVC and X is CH3O2for PVAc. Chullis and Hirschlerattributed differences in the thermal stability of vinyl polymers to changes in the234dissociation energy of the R-X bond [224], i.e., the dissociation energy of R-Cl isless than the dissociation energy of R-OC(O)CH3.When the average power spectra for the four polymers were compared, thelow numbers of signals collected for some of the trials led to a high variability of theaverage power spectra for the individual polymers and thus no comparison of theresults was attempted. The median descriptor values did not reveal any discernibletrends in the descriptor values as the vinyl acetate content was increased. Formany of the intensity related descriptors) there was an increase in the mediandescriptor values as the vinyl acetate content was increased from 0 to 40 %. This isin agreement with the acoustic RMS results shown in Figure 70. However, in eachcase the median descriptor values were lower for the PVAc polymer. This wasfurther evidence that something quite different was occurring in the PVAc polymerwith respect to the other compositions used here.For PVC, the separation of the two regions of acoustic emission found in theTS curves had indicated minor but significant differences in the characteristics ofthe acoustic emission signals produced during each region (section 4.3.2.1). Astwo regions of acoustic emission were observed for the vinyl acetate containingpolymers then a similar approach was applied. For the copolymers, low numbers ofsignals were produced during one of the two regions of activity and therefore onlythe PVAc was used. Principal component analysis using the descriptors given inTable 11, did not indicate any significant differences in the characteristics of theacoustic emission signals produced in the two regions of activity. Typical averagepower spectra of the acoustic emission produced in the two regions are shown inFigure 71.2350.030.020.01U)0.000.015Lt0.0100.0050.000Figure 71. Average power spectra from the two regions of AE activity during thethermo-oxidative decomposition of PVAc.(A) Lower temperature reaction stage.(B) Higher temperature reaction stage.Visual comparison of the two average power spectra shown indicatessignificant differences between the acoustic emission produced during the tworegions of activity. Similar results are seen as discussed earlier for PVC. Theaverage power spectra for the acoustic emissionproduced during the lowertemperature reaction contains more intense lowerfrequencies than the acousticemission produced during the other reaction. The results of regression analysis ofthe six average power spectra are shown in Table30. Use of the ‘t-test’ suggestedthat for PVAc, the average power spectra from the two regions of activity could bediscriminated. Significant differences existed between the values obtained when0 500 1000Frequency ([ 0.99780 000 2.3 230 - 274 3x1 20- 3x1 23 > 0.987188 000 2.2 262- 304 3x1 Q22 - 2x1Q26 > 0.994Table 31. Kinetic parameters for the dehydrochiorination of PVC in air obtainedfrom TG using the Coats-Redfern method. Effect of molecular mass.4.4.2 Effect of heating rateFor all of the TG curves, therewas a general decrease in themeasured massprior to the onset of reaction (Fig. 73). For heating rates of10, 7 and 5 K/mm, onlythe dehydrochlorination reaction was prominent. For the fastest heating rate a largeinstability in the mass readings was seen at higher temperatures for all the trialsundertaken. For heating rates of 3 and I K/mm, the secondregion of mass loss,due to carbonization, was present though not prominent. Close examination of theresults indicated that using a faster heating rate led to an increase in the Tj values.The kinetic parameters calculated for the dehydrochlorination process usingthe Coats-Redfern method areshown in Table 32. The apparent activationenergies using the two fastestheating rates were lower thanthose obtained usingthe two slowest heating rates though the high values obtainedusing a heating rateof 5 K/mm disputed this trend.Q)CC\2(ciV(1)U)(ci(ci00VbOcci-jVC-)V300 400 500 600 700 800Sample temperature (K)Figure 73. Effect of heating rate on the TG curve of thethermo-oxidativedecomposition of PVC. Mass of sample = 200 mg.239240Heating rate Avg.n Ea(kJImOI)A(s’)(K/rn in)10 2.1 258 - 270 3x1021 - 5x1 023 >0.9987 2.3 274 - 294 2x1024 - 3x1 027 >0.9975 2.7 344 - 361 5x1031 - 2x1 >0.9973 2.2 301 - 320 lxi27- 3x1 028 >0.9991 1.7 300 -315 5xiQ27- 6x1 028 >0.999Table 32. Kinetic parameters for thedehydrochiorination of PVC in air obtainedfrom TG using the Coats-Redfern method. Effect of heating rate.In section 4.6.1, results are given when acommercial TG-FTIR instrumentwas used to study the thermo-oxidative decomposition of PVC. Using theCoats-Redfern method, a reaction orderwas 2.1 and an apparent activation energywas 224 ±20 kJ/mol were calculated. The heating rate was 10 K/mm and althoughthe sample mass used in the commercialinstrument was lower (i.e., 13.385 mg),comparable results were found. In bothcases, the apparent activation energieswere higher than typical literature values.The kinetic parameters obtained using multiple heating rate methods ofanalysis proved to be more in agreementwith literature values. Using the Kissingermethod gave an apparent activation energy of 132 ±43 kJ/mol (r2 = 0.956) for thedehydrochiorination process. The Ozawa method gave apparent activationenergies from 134 to 106 kJ/mol with a mean value of 124 kJ/mol (Table 33).241a Ea(kJ/mOl) r20.05 134 ±44 0.96870.10 136±38 0.97560.15 136 ±38 0.97540.20 134 ±38 0.97560.25 132 ±37 0.97390.30 132 ±41 0.97270.35 130 ±38 0.97560.40 129±34 0.97690.45 129 ±38 0.97540.50 126±38 0.9733a Ea(kJ/mOI) r20.55 126 ±38 0.97400.60 125±38 0.97510.65 122 ±38 0.97390.70 120 ±34 0.97600.75 118±32 0.97780.80 116 ±32 0.97700.85 110±32 0.97720.90 107 ± 28 0.97790.95 106 ±28 0.9782Table 33. Activation energies for the dehydrochiorination of PVC in air obtainedfrom the TG results using the Ozawa method.4.4.3 Effect of copolymer compositionCharacteristic TG curves were observed for the two EVA copolymers and thetwo constituent homopolymers (Fig. 74). For the PE (i.e., 0 % vinylacetate), asingle decomposition step was observed between Ca. 640 and 790K. For thePVAc, two stages of decomposition were observed. The first reaction, due to theelimination of acetic acid, was observed between Ca. 530 and 625 K. Thesubsequent reaction, due to carbonization of the residue, was seen as a verygradual mass loss from Ca. 650 K. For the EVA copolymers, the mass lossassociated with the elimination of acetic acid from the vinyl acetate component was242observed between Ca. 570 and 660 K. This increased stability of the vinyl acetateunits may be attributed to the presence of the ethylene comonomer [225]. Furtherreaction commenced at a temperature very close to the end of the first stage andcontinued until Ca. 790 K. This reaction was attributed to decomposition of theethylene component. In PVAc, the carbonization reaction occurred at a similartemperature and thus this reaction probably contributed a little to the observedmass loss here.ci.CC.’)C)U)U)(6C’)00C)8flc6C)C-)C)0- 300 400 500 600 700Sample temperature (K)Figure 74. Effect of copolymer composition on the TG curve for thethermo-oxidative decomposition of EVA. Sample mass = 200 mg.800243The overlap of the two steps observed during the thermal decomposition ofEVA has been previously reported [1001 and measures have been taken to improvethe resolution of the two reactions, (e.g., use of a dynamic heating rate algorithmthat changed the heating rate according to the rate of reaction [99]). Due to thecomplexity of these algorithms and the required apparatus, this type of system wasnot employed here.Several workers have reported the use of TG to determine the vinyl acetatecontent in an EVA copolymer [99, 226]. The vinyl acetate content was determinedfrom the product of the mass loss for the first reaction and the ratio of molecularmass of the vinyl acetate monomer to the molecular mass of acetic acid. For theresults shown in Figure 74, vinyl acetate contents of 18.7 and 36.0 % werecalculated for the two copolymers. These compare favorably with the nominalvalues of 18 and 40 %, especially given the crudeness of the TG apparatus.4.5 Thermogravisonimetry studies4.5.1 Polymer studies using thermogravisonimetryThe temperature of the maximum rate of mass loss was close to the onset ofacoustic emission for PVC (Fig. 75). Surprisingly for PET, the temperature ofmaximum rate of mass loss was close to the onset of the pre-stage of acousticemission rather than the main reaction stage. For PP and PE, the small amounts ofacoustic emission produced during the decomposition reactions made the onset ofacoustic emission difficult to detect and also meant that any spurious noisecollected was significant when looking at the results obtained. The results shown inTable 34 confirm these observations.Figure 75. Thermogravisonimetry results for the thermo-oxidative decompositionof PVC, PE, PP and PET.244100755025075()J20C)z1.20.80.40.00.0050.0040.0030.0020.00 10.0000.080.060.040.020.000.160,120.080.040.000300 400 500 600 700 800Sample temperature (K)Acoustic emission Mass75I.100 q.0)50 B0)cI250 B0)755025Mass of sample = 200mg.245Trial T values (TS) (K) T values (TG) (K)PVC 1 536 5352 536 Avg.= 534 534 Avg.= 5333 530 531PE 1 745 7402 659 Avg. = 651 746 Avg.= 7433 550 744PP 1 560 7182 600 Avg. = 597 680 Avg.= 7003 630 703PET 1 689 7022 700 Avg.= 690 712 Avg.= 7033 680 685Table 34. A comparison of the acoustic emission onset temperatures and thetemperature at which the maximum rate of mass loss was observed forthe thermo-oxidative decomposition of the four polymers. The valuesfor PE and PP are estimates due to the small amounts of acousticemission produced.For PET and PVC, the agreement between the onset of acoustic emissionand the maximum rate of mass loss is excellent. For the other polymers any chanceof assessing this relationship was precluded due to the lack of acoustic emissionproduced during the thermo-oxidative decomposition of the polymer. Theseresultsconfirm the preliminary investigations reported in section 4.2.1.4. As simultaneousTG-TS were being undertaken, the reliability of the proposed relationship betweenthe two sets of results should be beyond reproach.Kinetic parameters were calculated for both the TS and TG results using theCoats-Redfern method (Table 35). For PVC, the reaction orders and apparent246activation energies calculated using from TG curves were higher than thosecalculated using thermosonimetry. This was in agreement with the trend reportedwhen the TS and TG results obtained at different timeswere compared. For PET,the reaction orders and apparent activation energies were of similar magnitudes forthe two techniques though both the highest and lowestactivation energies wereobtained using thermosonimetry. This trend was againin general agreement withthe previous results.Polymer Avg. r Ea (kJ/mol) A (s-I)r2PVC A TS 1.6 151 - 176 2x1Oil- 3x10I > 0.990TG 2.7 344- 361 5x1 031 - 2x1 033 > 0.997B TS 2.1 269- 276 2x1 015 - 2x1016 > 0.984PE TG 0.5 128-138 2x106-1x17>0.997PP TG 0.8 71 -97 4x102-3x1>0.990PET TS 1.0 200-350 lxlOII-1x1022 >0.914TG 1.1 214-264 4x10l3-4x16 >0.996Table 35. Kinetic parameters for the thermo-oxidative decomposition of PVC,PE, PP and PET measured using TGS.Process A. - Lower temperature reaction stage for PVC.Process B. - Higher temperature reaction stage for PVC.Comparison of the kinetic parameters obtained for PVC from the TS curvesusing the IR transmitter/receiver circuit (i.e., TGS system) and those from the TScurves using a direct connection between the transducerand the amplifiersuggested a number of differences in the results. For the first region of acousticemission, the calculated reaction orders were lower fromthe TGS curves while theactivation energies were similar. For the second region ofacoustic emission the247Figure 76. TGS curves for the dehydration ofCuSO4.5H20using a heating rateof 5 KImin.In all three trials, similar results were seen and therefore only one trial isshown (Fig. 76). Unfortunately, the amount of acoustic emission collected was lowbut for each trial acoustic emission was detected prior to the onset of the mass lossdue to the step-wise dehydration of the CuSO4.5H20[227], i.e.,reaction orders were very similar whilethe activation energies were generally lowerfor the TS system. No definite explanation is forthcoming for these differences.However, it was suspected that these differences were due to a measurement effectas a result of using the TGS apparatus. It was very unlikely that these differenceswere due to changes in the chemical reactions occurring as the same experimentalconditions (e.g., sample mass, heatingrate etc.) were being used.4.5.2 Other thermogravisonimetry studies4>30.5(J32U(1)0U00C))C))fl A300 330 360390 420Samp’e temperature (K)0.3450CuSO45H20 —> CuSO.H0+4H248For the three trials, the first acoustic emission produced from the sample wasdetected at between 330 and 340 K while a significant mass losswas first detectedat around 350 K. The difference in these temperatures is small,but it may besuggested that the acoustic emission is commencing prior to or close to the start ofthe thermal event. Therefore a significantly different result was seen to thatobserved for the polymer studies.4.6 Other polymer studiesThe purpose of these studies was to determine (if possible) the chemical andphysical changes that occur simultaneously with acoustic emission. As a result ofthese studies, it was hoped that the sources of the acoustic emission phenomenonfrom the thermo-oxidative decomposition of polymers could be positively identified.4.6.1 Reference thermogravimetry studiesCCCEU)U)CE0Ccn000’0.-JCU)0ci)0100806040200300 400 500 600 700Temperature (K)800 900Figure 77. TG curves for the thermal decomposition of PVC in heliumusingheating rates of 5, 10 and 20 K/mm.249The TG curves shown in Figure 77 are reproductions of the results obtainedat the industrial laboratories. The curvesshowed minor differences from theTG curves obtained for the thermo-oxidative decomposition of PVC (e.g., Fig. 72).Unfortunately it was not possible to use these sets of TG curves to compare directlythe effects of changing the reaction atmosphere as different instruments were usedand the sample masses were lower for the TG curves in helium. For theexperiments using helium, the second decomposition stage was well-defined andthus the kinetic parameters were determined for both stages of reaction. Theresults obtained the Coats-Redfern methodare shown in Table 36. Although, thethree TG curves were obtained using different heating rates, no attempt was madeto analyze the data using the multiple heating rate methods as it was felt that theuse of only three points for each plot was unsuitable.Heating rate fl Ea(kJImOl)A(s1)(Kimin)5 A 3.1 224±30 2x1019 0.97910 A 1.7 184± 12 4x1014 0.99220 A 2.3 194± 16 5x1015 0.9805 B 1.6 284 ±25 2x1018 0.99010 B 0.4 118± 13 7x105 0.97920 B 2.5 347 ±30 3x1022 0.986Table 36. Kinetic parameters for the thermal decomposition of PVC in heliummeasured using TG with heating rates of 5, 10and 20 K/mm.250The results suggested that the activation energies were a function of theheating rate for both decompositionprocesses. For both processes, the reactionorders and apparent activation energies were lower using a heating rate of10 K/mm. Knuemann, Schleussner and Bockhorn reported apparent activationenergies of 140 and 226 kJ/mol for the two processes in helium using a TG-MSmethod [228]. Here a multiple heatingrate method, analogous to the Ozawamethod, was used to analyze TG curves produced using heating rates fromI to100 K/mm. In other work, Liebman etal. used a differential method of analysis toobtain activation energies from the dehydrochlorination of PVC in an inertatmosphere (N2) using TG [47]. Typical Ea values were from Ca. 85 to140 kJ/mol.Obviously the Ea values determined here were higher than those reported intheliterature. This agreed with the general trend found throughout the work reported inthis dissertation.The TG curves obtained using the combined TG-FTIR technique are shownin Figure 78. Again, the curves are approximate reproductions of the actual TGcurves obtained. The data recorded using these techniques was interpretedinconjunction with Dr. J.W. Mason (SealLaboratories, El Segundo, CA). A briefsummary of the compiled report is given.During the thermo-oxidative decomposition of the PVC resin, only HCI, CO2and H20were detected. HCI gas wasnot detected until the temperature hadreached in excess of 525 K and it wasfound up to 700 K. The percentage ofHCIlost from the sample during dehydrochiorination was 61.83 %, which was indicativeof a raw PVC resin containing only theadditives derived from the polymerizationprocess [229].During the thermo-oxidative decomposition of the HDPE sample, the mainmass loss was observed at Ca. 700 K. Prior to this, only CO2 and H20wereCc0Eci)Cl)Cr)0E0C00ci)0Cci)C)ci)0Figure 78. TG curves for the thermo-oxidative decomposition of PVC and PE inair using a heating rate of 10 K/mm.For the PE curve, a number of small but significant masslosses wereobserved prior to decomposition of the polymer. These mass losses whichaccounted for 4.13 % of the total sample were attributed to an antioxidant (Ca. 2 %),a lubricant (1 %) and an UV inhibitor (1 %). A residue (1 %) left after decomposition251evolved. During the decomposition, additional compoundswere seen. The mixtureobserved comprised of aromatic and aliphatic hydrocarbons together with someester or ketone compounds. The products found were ingeneral agreement withthose reported earlier in section 1.3.1.100806040200300 400 500 600 700 800Temperature (K)900252of the polymer was attributed to the presence of silica, used as an anti-blockingagent.Kinetic parameters for the two TG curves (Fig. 78) were calculatedusing theCoats-Redfern method (Table 37).Polymer r Ea (kJ/mol) A (s-i)Pvc 2.1 224 ± 19 7x1 018 0.993PE 1.9 450 ± 100 3x1 030 0.983Table 37. Kinetic parameters for the thermo-oxidativedecomposition of PVC,and PE measured using TG.The apparent activation energies calculated are again higher than would beexpected from the literature. For the PE, the extremely highvalue was caused bythe difficulty in accurately defining the TG curve for the decomposition step.Surprisingly, given the previous results, the r and Ea values for the PVC polymerare in a close agreement with the values obtained using theTG apparatusdeveloped here and the same heating rate (i.e., r = 2.1-2.2 andEa = 258 - 270 kJlmol, respectively, from Table 32).4.6.2 Thermal decomposition studies using residueanalysisThe following sections detail the results of analysis undertaken on theresidues produced by heating the four polymers PVC, PE, PPand PET to varioustemperatures. The nominal temperatures used were converted to approximatesample temperatures, as discussed previously, to enable comparison of theseresults and those reported using the thermal analysis techniques. Sample253temperatures of 450, 503, 559, 615, 672, 728 and 783 K were found to correspondto the given nominal temperatures.It was impossible to ‘quench’ any reactions occurring at the temperature ofinterest and it must be realized that the residues obtained were not truerepresentations of the samples at the given temperature. However,it was felt thatcomparing the residues from various temperatures a true picture of the reactionstaking place might be found. Where a direct comparison was made of the analysisresults and those obtained using either TS or TG, Figure 75 was used as the resultswere obtained using the same sample mass and heating rate.4.6.2.1 Physical form of the residuesThe physical form of the polymer residues were very dependent on thetemperature to which the polymer had been heated. Although the results of SEMare given subsequently, visual differences in the residues were alsoof interest. Thefollowing paragraphs aim to detail a few of the general changes observed as thesample temperature was changed for each polymer.For the PVC residues, as the sample temperature was increased to 503 K,some discoloration of the residue was observed. As the sample temperature wasincreased further, the particle size became smaller and the sample color darkeneduntil the residue resembled char. For sample temperatures of 672 Kand above, theresidues resembled soot.For PE, the only visual change observed for sample temperatures below559 K was a gradual melting of the polymer beads. Even for a sampletemperatureof 559 K, the shape of the individual beads was still evident though thebeads werefused together and a small amount of degradation (i.e., discoloration) was observed.By 672 K, the residue was very heterogeneous with regions of black and brown254pieces (i.e., decomposition/carbonization). The residue observed at 728 K was inthe form of black pieces and only small amounts of a similar residue were observedat 783 K.For PP, the residues observed at 450 K were no longer fibres, butratherlumps of the polymer formed from the solidification of the melt.As the temperaturewas increased, the residues became darker in color and a decrease in the particlesizes was observed. No residues were found for 728 and 783 K, i.e., the materialhad been completely volatilized.For PET, a slight browning of the polymer was observed when thesampletemperature was 559 K. By 628 K, the polymer was in the form of alight brownpowder. Extensive degradation was observed at 672 K and the residues resembledthose from the PE. Only small amounts of residue were found at728 and 783 K.4.6.22 IR absorption spectraThe IR absorption spectra obtained for the residues were very difficult tointerpret. As the sample temperature was raised, an increased difficulty inpreparing the KBr pellets from the resulting material was experienced. The intenseabsorption of the residues produced at higher temperatures alsoinhibited theanalysis. The use of a reflectance method for the residues did notimprove thespectra. The assignments given are based on comparing a number of spectra runon the same residue. Books by Bower and Maddams [230], Colthup,Daly andWiberley [231], and a review by Maddams [32] were consulted.The IR absorption spectrum of PE is shown in Figure 79A. The two strongpeaks at 2920 and 2850 cm1 are assigned as CH2 stretching modes(i.e., u(CH2)).The corresponding bending mode (i.e., o(CH2))mode is found at 1469cm.Another strong peak is found at 719 cm-1. This doublet peak is assigned as an out255of plane deformation (i.e., ‘y(CH) mode). When the PE was thermally decomposed,no real changes were observed in the IR absorption spectra until the temperaturewas 672 K (Fig. 79B). At this temperature, the rate of mass loss observed was highindicating that decomposition of the polymer was ongoing (Fig. 75). Here anadditional peak at 1714 cm-1 was found. This may be attributed to a u(C=O) modeof carbonyl groups, probably ketones and provides evidence for oxidation of the PE.The IR absorption spectrum of PVC is shown in Figure 80A. Absorptions inthe region of 2970-2825 cm1 are assigned as u(CH) and u(CH2)modes. Otherstrong absorptions include the ö(CH2)and (CH) modes at 1430 and 1250 cm-1respectively. The presence of chlorine in the molecule is indicated by strongabsorptions at 690 and 620 cm1 which are assigned as due to u(CCI) modes. Theminor peak at 1730 cm1 is probably due to an impurity. The IR absorption spectraobtained when PVC was heated to 450 and 503 K did not differ from the spectra ofthe untreated polymer. For the higher temperature, this was surprising given thatthe residue at this temperature was discolored. At a temperature of 559 K, theabsorptions due to the u(CCI) modes had disappeared, indicating that completedehydrochiorination had occurred (Fig. 80B). The corresponding TGS curveindicated that acoustic emission would be detected at this temperature while thefirst region of mass loss was finished. At this temperature, evidence of unsaturatedstructure or aromatics was indicated by the presence of peaks at 3010 and1615 cm1. The former isa u(CH) mode of unsaturated compounds and the latter isassigned as a u(C=C) mode of unsaturated or aromatic structures. Confirmation ofthe partial aromatic nature of the residue was given by a peak at 745 cm-1, whichmay be assigned as y(CH) of aromatics. Oxidation was indicated by a strong peakat 1729 cm-1, indicative of a u(C=O) mode and a peak at 1142 cm-1, assigned as a256u(C-O) mode. It was not possible to obtain lR absorption spectra that wererepresentative of the residues obtained at higher temperatures.The lR absorption spectrum of PP (Fig. 81A) was very similar to that of PE(Fig. 79A) but here additional peaks were found due to the methyl groups in thestructure. Peaks at 2960 and 2870 cm-1 are assigned as o(CH3)modes and thepeak at 1377 cm-1 is assigned as a o(CH3)mode. The peaks at 1161 and 974 cm1are assigned as u(C-C) modes, the latter being shifted due to the combination of theu(C-C) mode and a ‘y(CH3)mode. Again, a possible impurity was found at1730 cm1. When the PP was thermally decomposed, there appeared to be agradual increase in an absorption at 1720 cm-1 once the temperature reached559 K and above (Fig. 81 B). This proves evidence for oxidation of the polymer asthis peak may be assigned as a o(C=O) mode, probably for ketones.The IR absorption spectrum of poly(ethylene terephthalate) is shown inFigure 82. The intense peaks at 1720 and 1260 cm1 are assigned as due tou(C=O) and u(C-O)0 modes respectively. The corresponding u(C-O) mode isfound at 1105 cm-1. The peak at 725 cm1 has been attributed to a y(CH) mode of ap-substituted aromatic [32]. The use of lR spectroscopy to characterize theresidues produced by the thermal oxidation of PET did not prove successful. Forthe residues produced at temperatures up to 672 K, the IR absorption spectra wereindistinguishable from that of the untreated polymer. This is in agreement with theresults of elemental analysis, which indicated no significant differences in theresidues produced at temperatures lower than 672 K. It was not possible to obtainrepresentative lR absorption spectra of the residues obtained at this temperatureand above. This temperature was close to the start of the decomposition stepobserved in the TG curve.257Overall, although these results were somewhat disappointing, confirmation ofmany of the known facets of the reactions was given. For PE and PP, the increasedintensity of absorptions in the region of 1715 to 1720 cm1 as the temperature wasraised confirmed that oxidation of the polymers had occurred. Although Suebsaengand co-workers [31] reported the use of conventional FTIR spectroscopy tocharacterize the solid products from the thermal decomposition of PET, severalworkers including Carlsson et a!. [95] have questioned the use of transmissionmethods when KBr pellets are employed. Severe optical problems result primarilydue to the poor dispersion of the residue particles in the KBr but also due to the lowtransmission of the residues.The IR absorption spectra of the PVC residues were the most informative.Assignments made from the spectra confirmed many of the known aspects of thethermo-oxidative decomposition of the polymer, including the loss of HCI, theformation of C=C bonds and aromatic compounds. Oxidation was again confirmedby absorptions in the region of 1715 cm1. Sevecek and Stuka [34] reported the useof IR absorption spectroscopy for the study of PVC residues after thermaldecomposition. The band assignments given here were in good agreement withtheir results, though given the difficulty of obtaining good IR absorption spectrausing the residues from higher temperature, their extensive findings seem a littlesurprising.258A.B.Figure 79. IR absorption spectra of PE.(A) PE.(B) PE heated to 672 K.CM-I259A.B.Figure 80. IR absorption spectra of PVC.(A) PVC.(B) PVC heated to 559 K.CM-I260A.B.Figure 81. IR absorption spectra of PP.(A) PP.(B) PP heated to 559 K.too.96 B4-94. 44..26110163-99 24-92.04892300. 1700. 1100.500, CM—I3500. 2900.Figure 82. IR absorption spectra of PET.2624.6.2.3 Elemental analysisThe results for elemental analysis of the residues produced by heatingthepolymers at a heating rate of 5 K/mm to the given temperatures are shown below(Table 38). For PE, a number of the residues obtained were very heterogeneous innature. Elemental analysis was not carried out using the 559 and 615K residuesfor this reason.T(K) %C %H %OPVC theory 38.44 4.84*298 K 38.57 4.88*450K 38.50 4.81*503K 39.45 4.86*559K 83.17 6.93 9.9615K 87.98 6.84 5.18672 K 89.39 6.20 4.41728K 83.98 3.21 12.8783K 83.84 2.86 13.3PE theory 85.63 14.37298 K 84.85 14.21 0.94672K 84.57 14.00 1.43728K 83.16 13.13 3.71Elemental analysis results for the residues produced by heating thepolymers, PET, PP, PE and PVC to various temperatures in air.Theoretical values were based on the chemical structures givenearlier in Table 2. The oxygen content was obtained by difference. Asthe % Cl was unknown for many of the PVC residues, then the oxygencontent was only calculated for those residues where IR spectroscopyhad shown the absence of CI.%C %H %OPET theory 62.50 4.20 33.3298K 62.10 4.21 33.7559 K 59.26 4.06 36.7615K 61.97 3.21 34.8672K 82.22 3.85 13.9728K 83.38 3.64 13.0783 K 76.61 3.21 20.2PP theory 85.63 14.37298 K 85.60 14.39 0.01559K 83.07 13.97 2.96615K 80.87 13.58 5.55672K 80.89 13.26 5.85Table 38.C-)00C-)E0Figure 83. The dependence of the atomic ratio H/C on the decompositiontemperature of PVC and PET.The atomic H/C ratio was a good indicator of the degree of decomposition forthe PVC and PET polymers. During the dehydrochlorination of PVC, the ratiodecreased due to the loss of HCI. The subsequent decrease in the atomic H/C ratioas the temperature was raised further confirmed that products containing a higheratomic percentage of hydrogen than carbon were being found. These productswere probably aliphatic rather than aromatic. For PET, the main change in theI .263The results shown indicate the increasing amount of oxidation occurringduring the polymer decompositions. Sevecek and Stuzka [34] discussed the use ofatomic H/C ratios in evaluating the thermal decomposition of polymers includingPVC. Figure 83 indicates the atomic H/C ratios for the thermo-oxidativedecomposition of PET and PVC.1.6 -1.4 -1.2 -1.0 -0.80.6 -0.4 -0.2 - II I I I . II I I300 400 500 600 700 800 900Sample Temperature (K)0PVC vPET264atomic H/C ratios was between 615 and 672 K. This result illustrated the problem ofquenching the reactions, as the corresponding TG curve showed that the mainregion of the decomposition commenced at Ca. 675 K (Fig. 75). The change in theatomic H/C ratio could not be confirmed as being due to the loss of one compoundonly, as numerous volatiles were found at this stage of the decomposition. Thedecomposition products identified earlier by Day, Parfenov and Wiles [94] thatcontained a higher percentage of hydrogen than carbon included methane andacetaldehyde. For PP and PE, the atomic H/C ratios changed little with thedecomposition temperature. For PP, the ratio was 2.003 for the polymer and 1.953for the 672 K residue. For PE, the theoretical ratio was 2.000 and this haddecreased to 1.882 for the 728 K residue. These results suggested that thedecomposition mechanisms occurring for these polymers involved the loss of unitshaving molecular formulae similar to the empirical formula of the polymer(e.g., aliphatic). For PVC and PET, changes in the atomic H/C ratios as a functionof the decomposition temperature suggested different decomposition mechanisms.For example, in PVC, a chain-stripping mechanism (‘zipper mechanism’) leads tothe elimination of HCI [82].4.6.2.4 Mass spectrometryThe residues obtained by heating the PVC to sample temperatures of 559,672 and 783 K were analyzed by mass spectrometry. The mass spectra obtainedwere compared to the mass spectrum obtained from the PVC resin to assess anychanges due to the thermo-oxidative decomposition (Fig. 84).Due to experimental restrictions, the lowest mlz value acquired was 30 for thePVC resin and 50 for the residues. Although this prevented the detection of anyHCI (m/z = 36, 38) remaining in the residues, the IR absorption spectra indicated265100806040 -20-0a,10080Csq, 6040a,200Figure 84. Mass spectra obtained from the thermal desorption mass spectrometryof a PVC resin and the PVC residues.To simplify the spectra, only peaks with an intensity greater than 10 %of the base peak are included.(A) PVC resin.(B) Residue from heating to 559 K.(C) Residue from heating to 672 K.(D) Residue from heating to 783 K.that complete dehydrochiorination had occurred prior to the temperatures at whichthese residues were obtained. Thus no HCI was expected in the mass spectra ofthe residues. For the residues obtained at 672 and 783 K, two regions of significantion production were observed during the thermal desorption experiments. However,no significant differences were found between the ions produced from the tworegions.(A)Ca,aa)Csa,100806040200100806040 -200(D)J I’ iiIi iiIIIii, iiI iIi 1 1h Ihhh Iiiii50 75 100 125 150 175 200 225 250Mass to charge ratio (m/z)266The mass spectra recorded for the PVC resin were very similar to thosereported elsewhere [40, 67, 228, 232-3]. The base peak in the spectrum was due toHCI (mlz = 36). Confirmation of the presence of chlorine in the base peak was givenby the mlz = 38 fragment, which had roughly one-third the intensity of the base peak.The second most intense peak in the mass spectrum resulted from the molecularion of benzene (mlz = 78). Other aromatic fragments observed include toluene(n2Iz = 91), indene (C9H8)(mlz = 115) and the molecular ion of naphthalene(mlz = 128). Of the remaining peaks, the small peak at mlz = 51 was identified asC4H3,a decomposition product of benzene [228] and the fragment peak atmlz = 105 was identified as an ethylbenzene fragment.The mass spectrum from thermal desorption mass spectrometry of the 559 Kresidue showed a number of similarities to the spectrum of the original material andmore surprisingly to that of the residue obtained at 783 K. In terms of thethermo-oxidative decomposition of PVC, this residue corresponded to a temperatureclose to the end of the dehydrochlorination process. Toluene, indene, naphthaleneand benzene fragments were again found though now the base peak was fromtoluene (mlz = 91) and the amount of benzene found had decreased considerably.Numerous aromatic fragments of higher mlz values were now present and thistogether with the increased complexity of the spectrum, indicated a moreheterogeneous sample than the PVC resin. Overall, a number of groupings ofpeaks were observed with the intensities gradually decreasing at higher mlz values.Among the compounds identified were substituted napthalenes includingmethylnaphthalene (mlz = 141), and ethylnaphthalene (mlz = 155), together withanthracene (m/z = 178) and methylanthracene (mlz = 191). Of the remaining peaks,those at m/z values of 202 and 215 belonged to a similar aromatic series. The peakat mlz = 54 has been identified elsewhere as I ,4-butadiene [233]. The peak at a mlz267value of 165 was not reported in any of the literature sources which were concernedwith the analysis of PVC resin or the products produced from the degradation ofPVC in a non-oxidative atmosphere. Thus this fragment probably resulted from thethermo-oxidative decomposition of the PVC. An extensive search of mass spectraldata in the literature (e.g. [234]) using the other peaks in the spectrum as possiblefragments of this ion, revealed a number of possible identities, It was known thatthe fragment was unlikely to contain nitrogen and as the mlz value was odd then thefragment was not a molecular ion. A possible source of this fragment wasI ,2-dimethylanthraquinone(C16H202). In the reported mass spectrum of thiscompound, the base peak was mlz = 165 and fragments having mlz values of 178,77 and 63 were also found. It must be stated however that this identification is byno means certain. Further oxygenated products were not observed in the massspectra. However, as the lowest mlz value acquired was 50, then any carbondioxide (mlz = 44) produced from the decarboxylation of oxygenated products wouldnot be detected.The mass spectrum recorded for thermal desorption mass spectrometry ofthe 672 K residue was quite different and less complex than the mass spectraobtained for the other residues. In terms of the thermo-oxidative decomposition ofPVC, the temperature at which this residue was obtained was close to the start ofthe second decomposition step and perhaps given the difficulty in quenching thereactions reflected a state during the second decomposition step. The base peakwas at m/z = 54 and other major peaks were at mlz values of 69, 83 and 97. Asthese fragments differing in values by 14 units then it was proposed that they werefrom the same fragmentation pathway. Further evidence of this was indicated bythe presence of a less intense peak at a mlz value of 111. Reference to theliterature indicated that these ions may result from the fragmentation of26811 -cyclopentyluneicosane (C26H52). Again, this was not a positive identification ofa compound, rather a suggestion at the nature of the compounds found. There wasagain evidence for indene (C9H8)(mlz = 115), naphthalene (mlz = 129),methylnaphthalene (mlz = 141), and ethylnaphthalene (mlz = 155). The peaks atmlz = 171 and 185 appeared to be the next members of the naphthalene series, i.e.,propylnaphthalene and butylnapthalene. Of the remaining peaks, the presence of apeak at mlz = 149 was the most surprising. Commonly, in mass spectrometrystudies involving plasticized PVC, a peak at this mlz value is attributed to phthalicanhydride (i.e., asC6H4(CO)2-OHjfrom the phthalate plasticizer [232].The mass spectrum recorded for the 783 K residue was the most complex ofall the four spectra compared. This temperature was close to the end of the seconddecomposition step and thus it was expected that extensive carbonization had takenplace. A series of bands at regular intervals were observed. The most intensepeaks of these bands corresponded to the fragments identified for the 559 Kresidue. In fact, apart from differences in the relative intensities of the peaks, a verysimilar spectrum to that seen for the 559 K residue was observed. This was verysurprising but perhaps it suggests that many of the decomposition products arevolatiles such that the residue does not change greatly.Although the highest mlz value shown in Figure 84 is 250, fragments havinghigher mlz values were observed for the three spectra recorded for the residues.Typically, the fragments at mlz values greater than 250 occurred in the form ofbands. For the 559 K residue low intensity bands were found between mlz values of290 and 400. For the 672 K residue, the band structure was again evident thoughnow the upper limit of the bands was close to a m/z value of 500. Two prominentfragment peaks were also observed; the first at a mlz value of 368 had an intensityof Ca. 10 % of the base peak intensity while the latter at a mlz value of 446 had an269intensity of Ca. 80 % of the base peak intensity. No identification was forthcomingfor these fragments but the fact that the mass difference between the two fragments(i.e., 78) corresponded to the loss of benzene, was perhaps significant. For thehighest temperature residue, the band structure was complex and extended up tothe highest mass scanned (i.e., 700). Above a mlz value of 250, three main regionsof band structure were observed; the first from ca. 310 to 384, the second fromCa. 428 to 518 and the last from Ca. 626 to 700. Within these regions, the mostprominent peaks were typically 13 to 16 units apart though the low intensity of thesebands meant that no real trends in the differences could be established.In conclusion, the mass spectra analyzed confirmed the presence of many ofthe types of compounds reported to be formed during the thermal decomposition ofPVC in air. These included aromatics and oxygen-containing products. In manycases, confirmation of the compounds produced was not possible due to difficultiesin deducing fragmentation pathways but the general nature of the residues waswithout doubt. The mass spectra obtained for the 672 K residue appeared to be inconflict with the other two spectra. Perhaps this was a sampling problem as suchdifferences were not expected. The results of elemental analysis, discussed earlierdid not appear to indicate such a discrepancy.2704.6.2.5 Scanning electron microscopy (SEM)The use of scanning electron microscopy (SEM) revealed considerableinformation about the nature of the residues obtained. Although, the samples weregold-coated prior to their examination, still in some cases the quality of the resultsobtained was affected by charging of the particles.For PP, the fibrous nature of the polymer was very evident (Fig. 85A). As thetemperature was raised, the polymer melted. This was evident for the resultsobtained for the 559 K residue (Fig. 85B). Surprisingly as the temperature wasraised to the region at which thermal decomposition had been first observed(Ca. 615 K), no real changes were observed, except that the particle size becamesmaller. Obviously the material was decomposing without significant changes in itsmicroscopic structure.The PE beads were found to be very smooth though in some cases,microscopic indentations were seen (Fig. 86A). As the temperature was raised, thepolymer melted though, as discussed earlier, the melting process was more of afusing of the beads rather than melting of the individual beads to form a melt in theconventional sense. By a temperature of 615 K, the polymer resembled more of aglassy solid (Fig. 86B). For this polymer, thermal decomposition was observed by672 K. The residue obtained by heating to this temperature was similar to theresidue obtained previously but contained small bubbles embedded in the surfaceof the material (Fig. 86C). As the temperature was increased further a drasticchange started to occur. Parts of the 728 K residue (Fig. 86D) resembled a pumice,having a rough irregular surface while other parts of the residue, seen in the sameSEM photo were more similar to the previous photo. For the residue obtained byheating to 783 K, a graphite-like structure was observed (Fig. 86E).Figure 85. SEM photographs of PP before and after thermo-oxidativedecomposition.(A) PP fibres. Magnification x800.(B) PP heated to 559 K. Magnification x400.271A.B.‘4I2272A.B.Figure 86. SEM photographs of PE before and after thermo-oxidativedecomposition.(A) PE beads. Magnification xl 00.(B) PE heated to 615 K. Magnification x300.273C.D.Figure 86. SEM photographs of PE after thermo-oxidative decomposition(C) PE heated to 672 K. Magnification x800.(D) PE heated to 728 K. Magnification x400.274E.Figure 86E. SEM photograph of PE heated to 783 K. Magnification x400.275For PVC, the powder prior to thermal decomposition was slightlylayered(Fig. 87A). The residue obtained at 503 K was similar in nature though it wasevident that some preliminary melting of the polymer had occurred(Fig. 87B). Theresidue obtained at 559 K indicated that considerable physicalchanges hadoccurred. Reference to the TG curve for this polymer (Fig. 74) indicated that, notsurprisingly, this temperature was close to the end of the first decomposition step.The residue was in the form of small smooth sheets and there was considerableevidence of bubble formation and bursting on the surface of the sheets (Fig. 87C).As the temperature was increased further, the polymer particles underwentconsiderable fracture, getting smaller and smaller (Fig. 87D). Fissures were alsoseen in the material. No evidence for bubble formation was seenfor the 615 Kresidue or the residues obtained at higher temperatures.The PET polymer was in the form of lumps (Fig. 88A). As the temperaturewas raised there was increasing evidence of pieces breaking awayfrom the polymer(Fig. 88B). At 672 K, there was some evidence of craters in thematerial, possiblydue to bubble bursting (Fig. 88C). Another SEM photograph takenof the 728 Kresidue suggested considerable bubble bursting (Fig. 88D) but perhaps some ofthis damage was caused by the electron beam. Overall it was verydifficult to findmicroscopic changes occurring in the polymer as a result of decomposition.276A.B.Figure 87. SEM photographs of PVC before and after thermo-oxidativedecomposition.(A) PVC powder. Magnification x600.(B) PVC heated to 503 K. Magnification x200.Hitu0-I-’Cl)0aS0C.)a)0a).tooxx3cc‘000z=ECUC,_()00),_0)0)>0)01Q-tor-4-lfl(ØOooCl)‘.1-0-c.0aL.0)CDU)0oc)coa)D0)U-F-.F-—04I’Iicibcoc’i1/-sJth>ax90EI-ci)cS0(Y)ci)xC(UV0=co30Saci)4-’F-s;t•LUEo_a)0)‘4-tOo0a0-Dci)o0ci)-Sal—Fa_2WLLJLUCo-accoa)D0)U-.4Figure 88. SEM photographs of PET after thermo-oxidative decomposition.(C) PET heated to 672 K. Magnification x300.(D) PET heated to 728 K. Magnification xl 50.279C.D.280For PVC, considerable acoustic emission was found under the experimentalconditions used here, between 550 and 650 K and from 730 K. Inspection of theSEM photographs obtained here, suggested that the acoustic emission wasproduced as the polymer changed from the melted powder (Fig. 87B) to the sheets(Fig. 87C). The gas evolution, evident from the bubble formation observed in thisresidue may be linked to the acoustic emission. This was confirmed by the resultsobserved for the PET polymer. However, the fact that only a small amount ofacoustic emission was produced when PE was decomposed, suggested thatsomething more complex was involved as bubbles were also evident for PE. Thesecond region of acoustic emission corresponded to the observed fracturing of thePVC during the second decomposition step. This is not surprising given thatfracture processes were known to produce acoustic emission during the stressing ofpolymers.Overall, the SEM photographs, while very useful, did not produce the asdefinitive conclusions regarding the sources of the acoustic emission as had beendesired. Without undertaking the SEM analysis of the sample whilst thermallydecradinc the sample and recordinQ the acoustic emission, it could be impossible tostate what was the definite source of the acoustic emission or what the effect ofquenching the reactions had had on the results obtained. Brown [235] confirmedthat although methods involving the use of thermomicroscopy using scanningelectron microscopy had been attempted, a number of problems remain. The use ofelectron beam leads to possible damage on the sample but more importantlyrequires a vacuum system. Obviously, thermo-oxidative studies would not bepossible. Temperature control would also be a problem with this type of system.In section 3.8.5, an attempt was made to visually follow the decomposition ofPVC in air using microscopy but the results were disappointing (section 4.6.5).28146.3 Thermal stability of PVC using a modified ASTM methodTable 39 indicates the results obtained when using the modified ASTMmethod to assess the thermal stability of PVC.PVC resin M (g/mol) Time (s) (±5 s) Average time (s)“#1” 59 200 700, 605, 550 578“#2” 58100 320,350 335“#3” 88 200 425, 450 438“#4” 83 900 1200, 1000, 915, 865 890“#5” 80 000 655, 700 678“#6” 188 000 240, 240 240Table 39. Thermal stability of the PVC resins as measured by the modifiedASTM method. Averages were based on the last two values given,which differed by less than 10 %.This procedure was not ideal as it was often quite difficult to accuratelyassess if HCI was detected by the indicator paper. The results obtained did notreveal any link between the viscosity-average molecular mass and the thermalstability of the PVC resin. In addition, no link was found previously between theseresults and measurements made using thermosonimetry.46.4 Other thermal stability studiesThe relative stability of the six PVC resins, measured using the melting pointapparatus, is shown in Table 40. Surprisingly, given the previous thermal stabilityresults, a tenuous relationship was found between the relative thermal stability andthe viscosity-average molecular mass. Repeated measurements gave the same282results. Apart from resin “#3”, the relative thermal stability increased with theM value. The main difference between this method and the other methods used toassess the thermal stability of the six PVC resins (e.g., TG or ASTM methods) wasthat this method was concerned with observing the appearance of discoloration ofthe resin. The other methods were more concerned with the evolution of the HCIgas. Maybe these two events cannot be regarded as simultaneous since the firstHCI can be trapped in the resin once produced. Given the disputed evidence for alink between molecular mass and thermal stability for PVC (i.e., [80, 90]), it wasdifficult to state whether the results obtained were truly significant.PVC resin M (g/mol) Relative Stability“#1” 59 200 2“#2” 58100 1“#3” 88200 3“#4” 83900 5“#5” 80000 4“#6” 188000 6Table 40. Relative thermal stability of the six PVC resins, measured using amelting point apparatus. The lowest value indicates the resin oflowest thermal stability.2834.6.5 Optical microscopy studiesCombined acoustic emission-optical microscopy studies proved difficult. Toprotect the camera from the effects of the HCI produced during the thermaldecomposition of the PVC in air, it was only possible to achieve a maximummagnification of 20 times. Cook used this apparatus to monitor the crystallization ofKBr by quantifying the ‘whiteness’ of the images recorded as a function of time[122]. Attempts to use a similar method here to follow the discoloration of thepolymer proved unsuccessful. As the camera is only focused on the surface of thepolymer then any discoloration at the base of the sample was not seen. Typically,the base of the polymer was the first region in which discoloration was observed.As the PVC was heated, the material first browned and then blackened asthe dehydrochiorination continued. It was very noticeable that some regions of thesample were degraded before other regions, giving the residue a heterogeneousappearance. The particle size of the material appeared to decrease at this stage.As the material blackened, it became increasingly difficult to sufficiently illuminatethe sample so that the camera could distinguish the individual particles. As thetemperature increased further, small bubbles appeared to form on a few of theparticles. After a long period of heating (Ca. 30 mins), significant amounts ofacoustic emission (Ca. 15 signals per mm) were acquired but no correspondingchanges were observed using microscopy. This was not too surprising as the SEMresults presented earlier suggested that the majority of the changes were takingplace on a much smaller scale than the range observed here. Typically the smallestsize that could be distinguished using this apparatus was Ca. 0.5 mm.Several workers have reported the use of thermal analysis systems combinedwith microscopy systems (e.g., [236-7]). Due to the problems associated with the284decomposing sample, reflectance methods are often used though transmissionmethods (using normal or polarized light) have also been reported.4.6.6 ConclusionsThe studies detailed above provided a large degree of information on thechemical and physical processes occurring during the thermo-oxidativedecomposition of the four polymers. The acoustic emission was found to be relatedto the physical changes occurring in the polymeric materials as a result ofthermo-oxidative decomposition. Although in some respects the chemical changesduring the decomposition reaction define the physical changes to a large degree, itwould be erroneous to say that the acoustic emission produced was directly a resultof chemical reactions, for example the dehydrochlorination of PVC.2855. CONCLUSIONSAcoustic emission was produced during the crystallization of polymers fromthe melt and during the thermal degradation of polymers. Polymerization processesand phase transitions (e.g. Tg and Tm) did not produce acoustic emission. Theplasticization of PVC also did not produce acoustic emission.The crystallization studies indicated that the acoustic emission wasempirically related to the rate of crystallization though the significance of the rateconstants obtained using acoustic emission was uncertain. In isothermal studies,the TS curves obtained at different temperatures were related in a manneranalogous to results obtained for other measurements of crystallization.The majority of the studies reported herein centred on the thermosonimetry ofpolymers. The amount of acoustic emission produced was very dependent on thenature of the polymer. For polypropylene and polyethylene, the TS curves werereproducible in terms of the temperature range over which the acoustic emissionwas produced and also the amount of acoustic emission produced. However, theamount of acoustic emission produced was small and prevented furtherinvestigation using these two polymers. For poly(ethylene terephthalate) andespecially poly(vinyl chloride), the TS curves were again reproducible and for thesepolymers the amount of acoustic emission produced was large. For theethylene-vinyl acetate copolymer, the amount of acoustic emission was proportionalto the amount of vinyl acetate in the copolymer when the vinyl acetate accounted forup to and inäluding 40 % of the composition. Unfortunately, use of this method forthe compositional analysis of these copolymers was prevented by the lack ofsensitivity of the method and the huge departure from a linear relationship for the286poly(vinyl acetate) homopolymer. The sensitivity may have been improved by achange of the experimental conditions. The TS curves using poly(vinyl acetate)were similar to those obtained using poly(vinyl chloride) and suggested that thenature of the decomposition mechanism influenced the TS curves. The smallamounts of acoustic emission produced by both polyethylene and polypropylenefurther enforced this idea, as both polymer decompose according to a similarmechanism.One of the initial aims of the studies was to assess whether thermosonimetrycould be used to assess the thermal stability of polymers. Many of the studiesreported in the review of thermosonimetry, indicated that acoustic emission wasproduced prior to the onset of the thermal event. Studies indicated that it was notpossible to use TS curves for this means. Unfortunately, for the polymer systems,the acoustic emission was produced after a large amount of decomposition hadoccurred. Further studies involving the use of TG, both in separate experimentsand as the combined thermogravisonimetry apparatus, suggested the onset of theacoustic emission coincided with the maximum rate of mass loss in those caseswhere both of these events could be measured with confidence. This relationshipwas observed for the dehydrochlorination reaction of poly(vinyl chloride), thedecomposition of poly(ethylene terephthalate) and poly(vinyl acetate) but not for thedecomposition of polypropylene and polyethylene, where small amounts of acousticemission were produced.The TS curves obtained from the thermo-oxidative decomposition of PVCunder different experimental conditions, e.g., heating rate, sample size, showedmany of the same trends as predicted for other more common thermal analysistechniques. The use of the TS curves to obtain kinetic information, such as Eavalues, was not as successful especially for single heating rate methods although287the relative success of multiple heating rate methods again confirmed that the TScurves were influenced in similar ways to other thermal analysis methods.Attempts to correlate the thermosonimetry results with the chemistry of thedecomposition processes proved difficult. Although, the amount of acousticemission was dependent on the polymer, the decomposition mechanism and thusthe chemical structure, this link was the result of the nature of the physical formproduced by decomposition rather than the chemistry directly. The results ofscanning electron microscopy suggested differences between the residuesproduced by the four polymers at different temperatures but without a direct visualrecord of the changes occurring on a microscopic level as a result of decomposition,no sources of the acoustic emission could be confirmed. As these studies linkedthe acoustic emission directly to physical changes occurring during decompositionrather than chemical changes, then this work would probably be of more interest tomaterials science now rather than the chemist.This work also resulted in the optimization of a furnace apparatus and itssubsequent modification to enable thermosonimetry experiments to be undertaken.To enable combined TS and TG studies, i.e., the newly-conceived technique ofthermogravisonimetry (TGS), an lR transmitter/receiver circuit was constructed andevaluated. In addition to its future use in this technique, potential applications ofthis type of circuit include other systems where a direct connection between asensor and the ‘outside world’ is not possible.288CHAPTER 6. FUTURE WORKMany areas exist in which the work reported here can be extended. Prior tofurther extensive investigations, the possibility of redesigning the thermosonimetryapparatus using a commercial DTA or TG instrument as the heat source should bere-evaluated. At an early stage of the research detailed in this dissertation,consultation with a commercial thermal analysis manufacturer led to a proposalinvolving the modification and use of a commercial TG instrument for the TSstudies. Unfortunately, no agreement was reached and as a result, thethermosonimetry apparatus was developed from a commercial furnace instead.Using a commercial TG instrument would provide greater sensitivity for any TGexperiments and possible lead to the use of smaller sample masses for allexperiments. With a view to possible future studies involving thermal transitionswhere no mass change is observed, the acquisition of a combined DTA/TGinstrument would be the most favorable.The work reported in this dissertation has proved that the physical form of thematerials determines to a large extent the amount of acoustic emission produced.Future studies requiring the development of a combined thermosonimetrymicroscopy system, possibly together with DTA, would enable a closer investigationof the microscopic changes occurring in the polymers during decomposition and theprobable assignment of the sources of acoustic emission.Thermal degradation studies reported in this dissertation were limited tothermo-oxidative degradation. Despite the sterling efforts of Mr. B. Snapkauskas(UBC Dept. of Chemistry, Mechanical shop), attempts to construct a box suitable forinert atmosphere experiments proved unsuccessful. Future work should focus onhow changing the reaction atmosphere influences the acoustic emission. Studies289could include using different inert atmospheres and alsoquantitatively changing theamount of oxygen present in the reaction atmosphere. Again the use of anapparatus based on a commercial thermal analysis instrument would present anumber of advantages in this case.The acoustic emission acquisition apparatus shouldalso be improved. Inacoustic emission studies, a direct connection between the sample and the activesurface of the transducer is preferred as this maximizes the sensitivity of thesystem. Current transducer designs can only function between 273 K and up toCa. 373 K typically. Obviously, the use of a transducer suitable for high temperaturestudies would be preferable but is currently unlikely. Indeed, a sensor wasdeveloped in this laboratory for this purpose butproved to have a limited lifetime.The storage capacity of the individual waveformacquisition apparatusprevented continuous acquisition of the acousticemission signals produced duringthe decomposition of the polymers. In time, improvements are likely in storagecapacity and data reduction techniques to permita continuous system to be used.Further studies should include more experimentsto assess the effect of changingthe acoustic emission acquisition parameters, e.g., trigger level and gain settings,on the acoustic emission produced.Thermosonimetry does not enable a direct measureof the total acousticemission produced up to a point in time in the same way that the mass recorded atany instance is a direct measure of the extent ofreaction. The use of TS curves toassess thermal reactions directly is probably limited until a method is found to givethis total measurement.From a wider viewpoint, the use of thermosonimetryto study other organicchemistry systems should be explored. The phasetransitions of such systems areknown to be acoustically active (e.g. [176, 177]). Other workers have used TA290instruments to study numerous organic reactions [238] and thermosonimetry mightbe applied to study similar reactions. On a more applied nature, preliminaryexperiments, not reported here, indicated that the burning of flame retardants gavelarge amounts of acoustic emission. Thus thermosonimetry might be used to thesetypes of systems and possibly extended to study fires directly.291BIBLIOGRAPHY1. P.J. Flory, Principles of polymer chemistry, Cornell University Press,Ithaca, NY, (1953).2. L.F. Fieser and K.L.Williamson, Organic Experiments, 6th Edn., pp. 287-288,DC Heath, Lexington, MA, (1987).3. K. Ziegler, E, Holzkamp, H. Briel, and H. Martin, “Polymerisation von Athylenund anderen Olefinen”, Angew. Chem., 67, 426, (1955).4. G. Natta, P. Pino, P. Corradini, F. Danusso, E. Mantica, G. Mazzanti, andG. Moranglio, “Crystalline high polymers of x-olefins,” J. Am. Chem. Soc.,77, 1708-10, (1955).5. L.S. Bark and N.S. Allen, Analysis of polymer systems, Table 1.7, pp. 17-19,App. Sci. Pub., London, (1982).6. H.F. Mark, N.M. Bikales, C.G. Overberger and G. Menges (Eds.),Encyclopedia of polymer science and engineering, , 383,John Wiley & Sons, New York, (1985).7. W.V. Titow, PVC technology, 4th Edn., 13, Appl. Sci. Pub., London, (1984).8. E.A. Collins, J. Barres, and F.W. Bilmeyer Jr., Experiments in polymerscience, Table 9-2, 222, John Wiley & Sons, New York, (1973).9. D. Turnbill and J.C. Fisher, “Rate of nucleation in condensed systems”,J. Chem. Phys., 17, 71-73, (1945).10. M.J. Avrami, “Kinetics of phase change. I. General theory”, J. Chem. Phys.,Z 1103-1112, (1939).11. J.M.G. Cowie, Polymers: chemistry & physics of modern materials, 2nd Edn.,pp. 244-2445, Blackie, Chapman and Hall, New York, (1991).29212. F.C. Perez-Cardenas, L. Felipe delCastillo, and R. Verra-Graziano,“Modified Avrami expression for polymer crystallization kinetics”,J. App!. Polym. Sd., 4, 779-782, (1991).13. F.W. Billmeyer, Jr., Textbook of polymer science, 3rd Edn., 155,John Wiley & Sons,New York, (1984).14. J.M. Marentette and G.R. Brown, “Polymer spherulites, 1.Birefringence andmorphology”, J. Chem. Ed., 70(6), 435-39,(1993).15. J.M.G. Cowie,Polymers: chemistry &physics of modern materials, 2nd Edn.,331, Chapman and Hall, New York, (1991).16. D. Campbell and J.R. White, Polymer characterization, physical techniques,Chapman and Hall,London, (1989).17. C.G. Smith, R.A. Nyquist, S.J. Martin, N.H. Mahie, P.B. Smith andA.J. Pasztor Jr., “Analysis of synthetic polymers”, Anal. Chem., 61(12),214R-238R, (1989).18. C.G. Smith, R.A. Nyquist, P.B. Smith, A.J. Pasztor Jr., andS.J. Martin,“Analysis of syntheticpolymers”, Anal. Chem., 63(12), 11 R-32R, (1991).19. C.G. Smith, P.B. Smith, A.J. PasztorJr., M.L. McKelvy, D.M. Meunier,S.W. Froelicher andA.S. Ellaboudy, “Analysis of synthetic polymers andrubbers”, Anal. Chem., 65(12), 217-243R, (1993).20. J.M.G. Cowie, Polymers: chemistry &physics of modern materials, 2nd Edn.,pp. 187-229, Chapman and Hall, New York, (1991).21. A.H. Wachter and W. Simon, “Molecular weight determination of polystyrenestandards by vapourpressure osmometry”, Anal. Chem., 4i 90-94, (1969).22. F.W. Billmeyer, Jr., Textbook of polymer science, 3rd Edn., pp. 75-84,John Wiley & Sons, New York, (1984).23. J.W. Williams,Ultracentrifugation ofmacromolecules: modern topics,Academic Press, NewYork, (1972).29324. M.L. Huggins, “The viscosity of dilute solutions of long chain molecules IV.Dependence on concentration”, J. Am. Chem. Soc., , 2716-2718, (1942).25. M. Kurata and Y. Tsunashima, “Viscosity-molecular weight relationships amdunperturbed dimensions of linear chain molecules”, in Polymer handbook,3rd Edn., Chapter Vll-1 2, J. Brandrup and E.H. Immergut (Eds.),John Wiley & Sons, New York, (1989).26. American Society for Testing and Materials (ASTM), “Standard test methodfor dilute solution viscosity of vinyl chloride polymers”, ASTM DesignationD 1243-79 (Reapproved 1990), in Plastics 08-01, pp. 412-414,ASTM Committee on Standards, Philadelphia, PA, (1991).27. J.R. Ebdon, “Microstructural analysis of synthetic organic polymers by highresolution nuclear magnetic resonance”, in Analysis ofpolymer systems,Ch. 2, pp. 21-49, L.S. Bark and N.S. Allen (Eds.), AppI. Sci. Pub., London,(1982).28. F.A. Bovey, and G.V.D. Tiers, “Polymer NSR spectroscopy. II. The highresolution spectra of methyl methacrylate polymers prepared with free radicaland anionic initiators”, J. Polym. Sd., 44, 173-1 82, (1960).29. C.W. Wilson Ill, “NMR study of molecular chain structure of polyvinylidenefluoride”, J. Polym. Sci. A, 1, 1305-1310, (1963).30. R. Lukas, M. Kolmsky and D. Doskocilova, “The structure of chlorinatedpoly(vinyl chloride). VI Comparison of poly(vinyl chloride) andf3-f3-d2..poly(vinyl chloride) with1H-NMR and 13C-NMR spectroscopy”,J. Polym. Sci., Polym. Chem. Edn., 17, 2691 -2699, (1979).31. T. Suebsaeng, C.A. Wilkie, V.T. Berger, J. Carter, and C.E. Brown, “Solidproducts from thermal decomposition of polyethylene terephthalate:Investigation by CP/MAS 13C-NMR and Fourier transform-IR spectroscopy”,J. Polym. Sci. Polym. Chem. Edn., 22, 945-957, (1984).32. W.F. Maddams, “Infrared and Raman spectroscopy”, in Analysis of polymersystems, Ch. 3, 69, L.S. Bark and N.S. Allen (Eds.), AppI. Sci. Pub., London,(1982).33. D. Campbell and J.R. White, Polymer characterization, physical techniques,Ch. 5, pp. 72-73, Chapman and Hall, London, (1983).29434. P. Sevecek and V. Stuzka, “Study of solid char residues after thermaldegradation of polystyrene, PVC and polyamide, part II”, Fire and Materials,11, 89-93, (1987).35. R.C. Wieboldt, G.E. Adams, S.R. Lowry and R.J. Rosenthal, “Analysis of avinyl chloride polymer by TGA-FTIR”, Amer. Lab., 71-76, (Jan. 1988).36. M. Bowden, P. Donaldson, D.J. Gardiner, J. Birnie and D.L. Gerrard,“Monitoring polyvinyl chloride degradation using Raman microline focusspectrometry”, Anal. Chem., 63, 291 5-291 8, (1991).37. H.G. Barth and B.E. Boyes, “Size exclusion chromatography”, Anal. Chem.,64, 428R-442R, (1992).38. J.N. Kyranos and P. Vouros, “The role of mass spectrometry in the analysisof polymers”, J. Appl. Polym. Sd. : AppI. Polym. Symp., 43, 211-240, (1989).39. R.E. Adams, “Pyrolysis mass spectrometry of terephthalate polyesters usingnegative ionization”, J. Polym. Sd. Polym. Chem. Edn., 20, 119-129, (1982).40. T.H. Risby, J.A. Yergey, and J.J. Scocca, “Linear programmed thermaldegradation mass spectrometry of polystyrene and poly(vinyl chloride)”,Anal. Chem., 54, 2228-2233, (1982).41. A. Kettrup, K. -H. Ohrbach, G. Matuschek, and W. Klusmeier, “Thermaldegradation of polymers and flame retardants”, J. Therm. Anal., 35, 291-303,(1989).42. K.-H. Ohrbach, G. Radhoff and A. Kettrup, “Investigations of the thermaldecomposition of PVC cable material by means of simultaneousTG-DTG-DTA-MS analysis”, Thermochim. Acta, 85, 403-406, (1985).43. M. Vincenti, E. Pelizzetti, A. Guarini, and S. Costanzi, “Determination ofmolecular weight distributions of polymers by desorption chemical ionizationmass spectrometry”, Anal. Chem., 64, 1879-1 884, (1992).44. H. Feld, A. Leute, R. Zurmuehlen and A. Benninghoven, “Comparative andcomplementary plasma desorption mass spectrometry/secondary ion massspectrometry investigations of polymer materials”, ibid, 63, 903-910, (1991).29545. D. Hill, J. O’Donnell and P. Pomery, “Applications of ESR in polymerchemistry”, Electron Spin Reson., 202-226, (1992).46. B. Yamada, M. Kageoka, T. Ostu, “ESR study of the radical polymerization ofstyrene. 5. Temperature dependence of propagation and termination rateconstants over a wide temperature range”, Polym. Bull. (Berfin), 29, 385-392,(1992).47. S.A. Liebman, J.F. Reuwer, Jr., K.A Gollatz and C.D. Nauman, “Thermaldecomposition of poly(vinyl chloride) and chlorinated poly(vinyl chloride).I. ESR and TGA studies”, J. Polym. Sd. A-I, , 1823-1833, (1971).48. S.A. Liebman, D.H. Ahlstrom, E.J. Quinn, A.G. Giegley, and J.T. Melusky,“Thermal decomposition of poly(vinyl chloride) and chlorinated poly(vinylchloride). II. Organic analysis”, ibid, , 1921-1935, (1971).49. D. Betteridge, J.V. Cridland, T. Lilley, N.R. Shoko, M.E.A. Cudby andD.G.M. Wood, “Acoustic emission and e.s.r. studies of polymers understress”, Polymer, 23, 178-184, (1982).50. R.C. MacKenzie, “Nomenclature in Thermal Analysis: Fourth Report toCouncil of Nomenclature Committee”, in Proc. 5th mt. Conf. Therm. Anal., 1,pp. 559-565, H. Chihara (Ed.), Heyden and Son Ltd., London, (1977).51. R.C. MacKenzie and PG. Laye, “Thermal analysis”, Chem. Br., 22(11),1005-1011, (1986).52. W.W. Wendlandt and P.K. Gallagher, in Thermal characterization ofpolymeric materials, Ch. 1, Tables II, Ill and IV, pp. 8-9, E.A. Turi (Ed.),Academic Press, New York, (1981).53. D. Dollimore, “Thermal analysis”, Anal. Chem., 64, 147R-153R, (1992).54. N.B. Milestone, “The use of thermal analysis in polymer characterization”,Chem. New Zealand, 55-58, (June 1988).55. G. Widmann, “Thermal analysis of plastics”, Amer. Lab., 98-103, (Jan. 1987).56. Z. Ahmad and W. Manzoor, “Thermogravimetric analysis of ZnCl2 catalyzeddegradation of PVC”, J. Therm. Anal., 38, 2349-2357, (1992).,29657. R. B. Cassel and A. P. Gray, “Microprocessor-control led thermogravimetricseparations”, Thermochim. Acta, 36, 265-277, (1980).58. W.W. Wendlandt, Thermal analysis, 3rd Edn., pp. 424-442, Vol. 19 of“Chemical Analysis, a series of monographs on analytical chemistry and itsapplications”, P.J. Elving, J.D. Winefordner and l.M. Kolthoff (Eds.),John Wiley & Sons, New York, (1986).59. S. Nakamura, M. Todoki, K. Nakamura, and H. Kanetsuna, “Thermal analysisof polymer samples by a round robin method. 1. Reproducibility of melting,crystallization and glass transition temperatures”, Thermochim. Acta, 136,163-178, (1978).60. D. Campbell and J.R. White, Polymer characterization, physical techniques,Ch. 12, 313, Chapman and Hall, London, (1983).61. M. Pracella, E. Benedetti and F. Galleschi, “A study of crystallizationbehaviour and compatibility of high-density polyethylene/linear low-densitypolyethylene blends”; Themochim. Acta, I 163-1 77, (1990).62. M.E. Brown, Introduction to thermal analysis: techniques and appllcations,Ch. 4, 46, Chapman and Hall, New York, (1988).63. S.A. Liebman and E.J. Levy (Eds.), Pyrolysis and G.C. in polymer analysis,Marcel Dekker, New York, (1985).64. T.P. Wampler, “A selected bibliography of analytical pyrolysis applications1980-1989”, J. Anal. App!. Pyrol., , 291-322, (1989).65. B. Plage and H-R. Schulten, “Thermal degradation and mass spectrometricfragmentation processes of polyesters studied by time I temperature-resolvedpyrolysis-fie Id ionization mass spectrometry”, Macromolecules, ,2642-2648, (1990).66. M.R. Rao, “Correlation of binder mechanical properties with microstructurefor poly(butadiene-acrylonitrile-methacrylic acid) terpolymer: 13C-NMR andpyrolysis-GC studies”, J. AppI. Polym. Sc, 4753-766, (1991).29767. Y. Shimizu and B. Munsen, “Pyrolysis I chemical ionization massspectrometry of polymers”, J. Polym. Sd., Polym. Chem. Edn., 17,1991-2001, (1979).68. T.W. Ryan, S.G. Yocklovich, J.C. Watkins and E.J. Levy, “Quantitativeanalysis of additives in polymers using coupled supercritical fluidextraction-supercritical fluid chromatography”, J. Chromato., 505, 273-282,(1990).69. C.F. Cullis and M.M. Hirschler, The combustion of organic polymers, 3,Clarendon Press, Oxford, (1981).70. D.M. Considine and G.D Considine (Eds.), Encyclopedia of chemistry,4th Edn., 793, Van Nostrand-Reinhold, New York, (1984).71. T. Kelen, Polymer degradation, Ch. 6, 108, Van Nostrand-Reinhold,New York, (1983).72. N. Grassie and G. Scott, Polymer degradation and stabilization, Ch. 4, 90,Cambridge Univ. Press, New York, (1985).73. W. L. Hawkins, Polymer degradation and stabilization, pp. 14-18,Springer-Verlag, New York, (1984).74. R.H. Hansen, “Thermal and oxidative degradation of polyethylene,polypropylene, and related olefin polymers”, in Thermal Stability of Polymers,1 Ch. 6, pp. 161 -1 65, “Monographs in Macromolecular Chemistry”,R.T. Conley (Ed.), Marcel Dekker, New York, (1970).75. J. Mitera, J. Michal, J. Kubat, and V. Kubelka, “Analysis of thermo-oxidationproducts of polypropylene and polyethylene by gas chromatography/massspectrometry”, Z. Anal. Chem., 23-27, (1976).76. T.P. Wampler and E.J. Levy, “Effect of heating rate on oxidative degradationof polymeric materials”, J. Anal. App!. Pyrolysis, , 153-161, (1985).77. C. Vasile, E. Costea, and L. Odochian, “The thermoxidative decomposition oflow density polyethylene in non-isothermal conditions”, Thermochim. Acta,j4, 305-311, (1991).29878. J. Khorami, A. Lemieux, H. Menard, and D. Nadeau, “Usefulness of Fouriertransform infrared spectroscopy in the analysis of evolved gas from thethermogravimetric technique”, in Compositional analysis bythermogravimetry, ASTM STP 997, pp. 147-1 59, C.M. Earnest (Ed.),American Society for Testing and Materials, Philadelphia, PA, (1988).79. T.R. Manley, “Thermal analysis of polymers”, Pure & App!. Chem., 61(8),1353-1360, (1989).80. A.J. Chirinos-Padron and GA. von Schoettler, “Characterization anddegradation studies of suspension-polymerized poly(vinyl chloride): Part I:Thermal degradation”, Polym. Degrad. Stab., 33, 213-227, (1991).81. T. Kelen, Polymer degradation, Ch. 5, pp. 77-105, Van Nostrand Reinhold,New York, (1983).82. C.F. Cullis and M.M. Hirschler, The combustion of organic polymers,pp. 143-147, Clarendon Press, Oxford, UK, (1981).83. A. Guyot and J.P. Benevise, “Sur Ia degradation thermique du chlorure depolyvinyle. Ill. Etudes cinetiques sous atmosphere d’air”, J. App!. Polym. Sci.,6(22), 489-495, (1962).84. D. Braun, “Thermal degradation of polyvinyl chloride”, Pure & AppI. Chem.,26, 173-192, (1971).85. P. Bataille and B.T. Van, “Mechanism of thermal degradation of poly(vinylchloride)”, J. Polym. Sci. A-I, 10, 1097-1108, (1972).86. T. Hjertberg and E.M. Sørvik, “Thermal degradation of PVC” in Degradationand stabilization of PVC, Ch. 2, pp. 21-79, E.D. Owen (Ed.), Elsevier,New York, (1984).87. K.B. Abbas and E.M. Sørvik, “On the thermal degradation of poly(vinylchloride). II. The effect of atmosphere”, J. App!. Polym. Sd., j, 3577-3594,(1973).88. G. Talamini and G. Pezzin, “Kinetic study on the reaction of polyvinyl chloridethermal dehydrochlorination”, Makrolmo!. Chem., 39, 26-38, (1960).29989. E.A. Boettner, G. Ball and B. Weiss, “Analysis of the volatile combustionproducts of vinyl plastics”, J. App!. Poly. Sd., 13, 377-391, (1969).90. l.M. Kamal, G.A. Adam and H.K. Al Getta, “The effect of molecular weight onthe thermal stability of polyvinyl chloride, Thermochim. Acta, 62, 355-359,(1983).91. D.H. Ahistrom, “Microstructure of synthetic polymers”, in Pyrolysis and G.C.in polymer analysis, Ch. 5, 256, S.A. Liebman and E.J. Levy (Eds.),Marcel Dekker, New York, (1985).92. M.E. Bednas, M. Day, K. Ho, R. Sander, and D.M. Wiles “Combustion andpyrolysis of poly(ethylene terephthate). I. The role of flame retardants onproducts of pyrolysis”, J. App’. Polym. Sci., , 277-289, (1981).93. C.F. Cullis and M.M. Hirschler, The combustion of organic polymers, Ch. 3,151, Clarendon Press, Oxford, UK, (1981).94. M. Day, V. Parfenov and D.M. Wiles, “Combustion and pyrolysis ofpoly(ethyethylene terephthalate). Ill. The effect oftri(2,3-dibromopropyl)phosphate on the products of pyrolysis”,J. App!. Po!ym. Sci., 27, 575-589, (1982).95. D.J. Carisson, M. Day, T. Suprunchuk, and D.M. Wiles, “Pyrolysis ofpoly(ethylene terephthalate) fibers: Characterization of involatile residues”,J. App!. Po!ym. Sd., 28, 715-724, (1983).96. C. Birladeanu, C. Vasile, and l.A. Schneider, “Ober die kinetik desthermischen und thermooxidativen abbaus von polyathyleneterphthalat”,Makromol. Chem., 177, 121-129, (1976).97. J.D. Cooney, M. Day, and D.M. Wiles, “Thermal degradation of poly(ethyleneterephthalate): A kinetic analysis of thermogravimetric data”,J. App!. Po!ym. Sci., , 2887-2892, (1983).98. ibem, “Flame retardancy effects on the thermal degradation of poly(ethyleneterephthalate) fabrics as studied by thermogravimetric analysis”, inProc. 7th mt. Conf. Therm. Anal., 2, pp. 1325-1 331, B. Miller (Ed.),John Wiley & Sons, Chichester, UK, (1982).30099. M.B. Maurin. J. Wai Chun Pang and M. A. Hussain, “Thermogravimetricanalysis of ethylene-vinyl acetate copolymer with dynamic heating rates,Thermochim. Acta, 209, 203-207, (1992).100. M.B. Maurin, L.W. Dittert and A.A. Hussain, “Thermogravimetric analysis ofethylene-vinyl acetate copolymers with Fourier transform infrared analysis ofthe pyrolysis products”, ibid, 186, 97-102, (1991).101. T. Sawada, Y. Gohshi, C. Abe and K. Furuya, “Acoustic emissions arisingfrom the gelation of sodium carbonate and calcium chloride”, Anal. Chem.,57(1), 366-367, (1985).102. A.G. Beattle, “Acoustic emission, principles and instrumentation”,J. Acoustic Emission, 2(112), 95-1 28, (1983).103. M. Arrington, “Tin cries - but most materials talk. An introduction to thetechnology and applications of acoustic emission”, Phys. Technol., 12, 16-23,(1981).104. J. Rodgers, “Acoustic emission in aircraft structural integrity andmaintenance programs”, J. Acoustic Emission, Jrn 113-120, (1982).105. R. Qian and X. Zhu, “Acoustic emission of polymers”, in Applied polymercharacterization, Ch. lI-F, pp. 157-1 74, J. Mitchell (Ed.), Hanser, Munich,ERG, (1987).106. T.J. Holroyd, “Acoustic emission from an industrial viewpoint”,J. Acoustic Emission, j4}, 193-199, (1988).107. P.H. Hutton, “Development and application of acoustic emission methods inthe United States - A status review”, J. Acoustic Emission, §j, 167-176,(1987).108. A.P. Wade, D.B. Sibbald, M.N. Bailey, R.M. Belchamber, S. Bittman,J.A. McLean and P.D. Wentzell, “An analytical perspective on acousticemission”, Anal. Chem., 63(9), 497A-507A, (1991).109. J.A.C. van Ooijen, E. van Tooren and J. Reedijk, “Acoustic emission duringthe preparation of dichloro(pyrazine)zinc(ll)”, J. Am. Chem. Soc., 100(17),5569-5570, (1978).301110. B.F. Munro, “Acoustic emission from the synthesis ofdichloro(pyrazine)zinc( II)”, M. Sc. Thesis, Dept. of Chemistry, UBC,Vancouver, BC, (Aug. 1991).111. D. Betteridge, M.T. Joslin and T. Lilley, “Acoustic emissions from chemicalreactions”, Anal. Chem., 53, 1064-1073, (1981).112. M.T. Joslin, “Acoustic emissions from chemical reactions”, Anal. Proc., 19,330-333, (1982).113. J.D. Navratil, “Acoustic emission monitoring of separation processes”, inProc. 2nd mt. Conf. Sep. Sd. Technol., 1, pp. 223-230, M. H. I. Baird andS. Vijayan (Eds.), Can. Soc. Chem. Eng., Ottawa, ON, (1989).114. T. Sawada, Y. Gohshi, C. Abe and K. Furuya, “Acoustic emissions fromphase transitions of some chemicals”, Anal. Chem., 57(8), 1743-1745,(1985).115. R.M. Beichamber, D. Betteridge, M.P. Collins, T. Lilley, C.Z. Marczewski andA.P. Wade, “Quantitative study of acoustic emission from a model chemicalprocess”, ibid, 58(8), 1873-1877, (1986).116. P.D. Wentzell, S.J. Vanslyke and K.P. Bateman, “Evaluation of acousticemission as a means of quantitative chemical analysis”,Analytica Chim. Acta, 246, 43-55, (1991).117. K.P. Bateman, “Chemical analysis by acoustic emission”, B.Sc. thesis, Dept.of Chemistry, Dalhousie University, Halifax, NS, (March 1990).118. T.G. Crowther, A.P. Wade, PD. Wentzell and R. Gopal, “Characterization ofacoustic emission from an electrolysis cell”, Analytica Chim. Acta, 4,223-234, (1991).119. T.G. Crowther, “Application of chemical acoustic emission to industrialprocesses”, M.Sc. thesis, Dept. of Chemistry, UBC, Vancouver, BC,(April 1991).120. l.H. Brock, 0. Lee, K.A. Soulsbury, P.D. Wentzell, D.B. Sibbald andA.P. Wade, “Toolbox for chemical acoustic emission data acquisition andanalysis”, Chemom. Intell. Lab. Sys., j, 271-290, (1992).302121. A.P. Wade, K.A. Soulsbury, P.Y.T. Chow and l.H. Brock, “Strategies forcharacterization of chemical acoustic emission signals near the conventionaldetection limit”, Analytica Chim. Acta, 246, 23-42, (1991).122. A.P. Cook, “Acoustic emission from the crystallization of potassium bromidedoped with lead”, M.Sc. thesis, Dept. of Chemistry, UBC, Vancouver, BC,(Jan. 1993).123. A.P. Cook, L.E. Bowman and A.P. Wade, “Study of KBr/Pb(N03)2growthand crystal morphologies by acoustic emission and photomicrographictechniques”, J. Crystal Growth, 131, 395-41 2, (1993).124. D. Betteridge, P.A. Connors, T. Lilley, N.R. Shoko, M.E.A Cudby andD.G.M. Wood, “Analysis of acoustic emissions from polymers”, Polymer, 24,1206-1212, (1983).125. R.M. Beichamber, D. Betteridge, Y.T. Chow, T. Lilley, M.E.A. Cudby andD.G.M. Wood, “Looking for patterns in acoustic emissions”, in 1st mt. Symp.on Acoustic Emission from Reinforced Composites, Paper 1, San Francisco,July 19-21 1983, Soc. Plastic Industry, New York, (1983).126. idem, “Evaluation of pattern recognition analysis of acoustic emission fromstressed polymers and composites”, J. Acoustic Emission, 4(4), 71-83,(1985).127. R.M. Belchamber, D. Betteridge, M.P. Collins, T. Lilley, C.Z. Marczewski andAG. Hawkes, “Time series analysis of acoustic emission signals from glassreinforced plastics”, in Acoustic emission monitoring and analysis inmanufacturing, i4 pp. 1-9, D.A. Dornfeld (Ed.), Am. Soc. Mec. Engineering,New York, (1984).128. J. Kaiser, “Erkenntnisse und Folgerungen aus der Messung von Gerãuschenbei Zugbeanspruchung von metal I ischen Werkstoffen”,Archly. ElsenhUttenwessen, 24, 43-45, (1953).129. R. Qian and T. Wang, “Some observations on the acoustic emission ofpolymers”, Eur. Polym J., 19(10111), 947-948, (1983).130. A. Galeski, L. Koenczoel, E. Piorkowska and E. Baer, “Acoustic emissionduring polymer crystallization”, Nature, , 40-41, (1987).303131. idem, “Acoustic emission during crystallization of polymers”,J. Polym. Sd. B., Polym, Phys., , 1171-1186, (1990).132. J. Shen, D. Xu, R. Yao, R. Liu and D. Li, “Study of acoustic emission duringnon-isothermal crystallization of polypropylene”, Chin. J. Polym. Sd., fjj,69-74, (1990).133. A.G. Beattie, “Acoustic emission couplants I: The ASTM survey”,J. Acoustic Emission, 2(112), 67-68, (1983).134. S.J. VanSlyke, “Monitoring acoustic emission from some model chemicalsystems”, B.Sc. Thesis, Dept. of Chemistry, UBC, Vancouver, BC,(April 1989).135. G.M. Clark, “Thermoacoustic emission analysis (Thermosonimetry)”,Therm. Anal. Revs. Abs., 18(6), 1-4, (1989).136. P.D. Wentzell, S.J. VanSlyke and A.P. Wade, “Programming direct memoryaccess data acquisition”, TrAC, Trends in Analytical Chemistry, LIj, 3-8,(1990).137. W.H. Press, B.P. Flannery, S.A. Teukoisky and W.T. Vetterling,Numerical recipes: the art of scientific computing, Ch. 12.2, pp. 390-395,Cambridge Univ. Press, New York, (1988).138. R.G. Brereton, “Fourier transforms: Use, theory and applications tospectroscopic and related data”, Chemom. Intell. Lab. Sys., , 17-31, (1986).139. P.D. Wentzell and A.P. Wade, “Chemical acoustic emission analysis in thefrequency domain”, Anal. Chem., 2638-2642, (1989).140. P.D. Wentzell, 0. Lee and A.P. Wade, “Comparison of pattern recognitiondescriptors for chemical acoustic emission analysis”, J. Chemom., ,389-403, (1991).141. R.E. Aries, D.P. Lidiard and R.A. Spragg, “Principal component analysis”,Chem. Br., 27(9), 821-824, (1991).142. N. Bratchell, “Cluster analysis”, Chemom. Intel!. Lab. Sys., 6, 105-1 25,(1989).304143. R.R. Meglen, “Examining large databases: A chemometric approach usingprincipal components analysis”, J. Chemom., , 163-179, (1991).144. K.A. Soulsbury, A.P. Wade, and D.B. Sibbald, “A rules-based approach toclassification of chemical acoustic emission signals”,Chemom. Intel!. Lab. Sys., 15, 87-105, (1992).145. G.M. Clark, “Instrumentation for thermosonimetry”, Thermochim. Acta, 27,19-25, (1978).146. H.S. Scott, “The decrepitation method applied to minerals with fluidinclusions”, Econ. Geol., 43, 637-654, (1948).147. F.G. Smith and P.A. Peach, “Apparatus for the recording of decrepitation inminerals”, ibid, 44,449-451, (1949).148. K. Lønvik, “An experimental investigation of the thermal decomposition ofbrucite by thermosonimetry”, Thermochim. Acta, , 27-44, (1978).149. G.M. Clark and R. Garlick, “Thermosonimetry of the NBS-ICTA temperaturestandards”, ibid, 34, 365-375, (1979).150. 5. Shimada and R. Furuichi, “A new apparatus for simultaneousmeasurement of acoustic emission and differential thermal analysis and itsapplication to dehydration, phase transition, and decomposition studies ofseveral inorganic salts”, Bull. Chem. Soc. Jpn., 63(9), 2526-2530, (1990).151. K. Lønvik, “Thermosonimetry: A method of material characterization”, inProc. 4th mt. Conf. Therm. Anal., 3, pp. 1089-1105, I. Buzas(Ed.),Heyden and Son, London, (1974).152. G.M. Clark, “Frequency distributions during thermosonimetry”, inProc. 2ndEur. Symp. Therm. Anal., 1, pp. 85-88, D. Dollimore (Ed.),Heyden and Son, London, (1981).153. A.A. Griffith, “VI. The phenomena of rupture and flow in solids”,Phil. Trans. Roy. Soc., 221A, 163-198, (1920).305154. J. Hindar, J.L. HoIm, J. Lindemann, and K. Lønvik, “Investigation of somekaolines by simultaneous DTA/TG and thermosonimetry”, inProc. 6th mt. Conf. Therm. Anal., 2, pp. 1313-1318, W. Hemminger (Ed.),Birkhaeuser Verlag, Basel, Switzerland, (1980).155. V. Lach, “Some applications of the decrepitation technique andthermosonimetry in research on minerals and building materials”,Thermochim. Acta, 110, 265-267, (1987).156. V. Lach and Z. Slavik, “Application of thermosonimetry for the study of firingin ceramics”, ibid, 3, 589-592, (1985).157. K. Lønvik, K. Rajeshwar, and J.B. Dubow, “New observations on chemicaland structural transformations in Green River oil shales”, J. Therm. AnaL, 42,11-19, (1980).158. G.M. Clark, “Thermosonimetry of glass”, in Proc. 7th mt. Conf. Therm. Anal.,1, pp. 300-305, B. Miller (Ed.), John Wiley & Sons, Chichester, UK, (1982).159. K. Lønvik, “Thermosonimetry (TS) of transitions in inorganic glass materials”,Thermochim. Acta, 175, 105-108, (1991).160. L. Stagni and A. Congiu Castellano, “Experiments on acoustic emissiongenerated during rapid heating of tin single crystals”, J. Mater. Sd., ,1431-1437, (1977).161. C. Picornell, C. Segui, V. Torra, F.C. Lovey, and R. Rapacioli, “Systematicstudy of the martenistic transformation in a Cu-Zn-Al alloy. Opticalmicroscopy and simultaneous thermosonimetry of a microplate”,Thermochim. Acta, 113, 171-183, (1987).162. 0. Hunderi and K. Lønvik, “Thermosonimetric investigation of thecrystallization of metallic glasses”, in Proc. 3rd lnt. Conf. Rapidly QuenchedMetals, 1, pp. 375-382, B. Cantor (Ed.), The Metals Soc., London, (1978).163. K. Lønvik, “Thermosonimetry”, Thermochim. Acta, 110, 253-264, (1987).164. S. Shimada and R. Furuichi, “A study on the thermal decomposition of KCIO4and NaCIO4by acoustic emission thermal analysis.”, ibid, 163, 31 3-316,(1990).306165. S. Shimada, Y. Katsuda and R. Furuichi, “A study on the thermaldecomposition of KCIO4 by acoustic emission thermal analysis: influences ofparticle and sample sizes on acoustic emission curve”, ibid, 183, 365-368,(1991).166. S. Shimada, Y. Katsuda, R. Furuichi and M. Inagaki, “Simultaneousmeasurements of acoustic emission and differential thermal analysis fordifferent KCIO4 samples”, ibid, 14 91-98, (1991).167. 5. Shimada, “Acoustic emission in the process of dehydration and thermaldecomposition of NaClO4.H20”, ibid, 196, 237-246, (1992).168. idem, “Thermosonimetry and thermomicroscopy of the phase transition anddecomposition of CsCIO41’, ibid, 200, 31 7-326, (1992).169. G.M. Clark, M. Tonks, and M.J. Tweed, “Thermal properties of potassiumdichromate”, J. Therm. Anal., 12, 23-31, (1977).170. idem, “Application of thermal analysis to phase transition studies ofpyrocompounds”, in Proc. 1st Eur. Symp. Therm. Anal., 1, pp. 448-451,D. Dollimore (Ed.), Heyden and Son, London, (1976).171. K. Lønvik, “Thermosonimetry of some alkali metal dichromates”,Thermochim. Acta, 29, 243-246, (1979).172. S.N. Sakiev, A. Kholov and F. Gulamova, “Determination of thedecomposition temperature of solids by the acoustic emission method”,Russ. J. Phys. Chem., 63(5), 773-774, (1989).173. S.N. Sakiev, “Determination of dissociation temperature by acousticemission”, J. App!. Chem. (USSR), 64(11), 2270-2271, (1992).174. G.M. Clark, “Some applications of thermosonimetry to organic crystals”,Anal. Proc., 23, 393-394, (1986).175. G.M. Clark and J.E. Fairbrother, “Dislocation densities in adipic acidcrystals”, in Proc. 2nd Eur. Symp. Therm. Anal., 1, pp. 255-258,D. Dollimore (Ed.), Heyden and Son, London, (1981).176. 0. Lee, Y. Koga, and A.P. Wade, “Thermosonimetry of the phase Il/Illtransition of hexachloroethane”, Talanta, 37(9), 861-873, (1990).307177. D.B. Sibbald, “Development of data acquisition and analysis methods forchemical acoustic emission”, M.Sc. thesis, Dept. of Chemistry, UBC,Vancouver, BC, (July 1990).178. W.W. Wendlandt and P.K. Gallagher, in Thermal characterization ofpolymeric materials, 1st Edn., Ch. 1, 6, E.A. Turi (Ed.), Academic Press, NewYork, (1981).179. Information sheet for AD594/AD595 monolithic thermocouple amplifier withcold junction compensation, Analog Devices, Norwood, MA.180. C. D. Johnson, Process control instrumentation technology, 3rd Edn.,pp. 359-377, John Wiley and Sons, New York, (1988).181. C.M. Earnest, “Modern thermogravimetry”, Anal. Chem., 56(13), 1471-1486,(1984).182. A.R. McGhie, “Thermometric calibration in thermogravimetric analysis”,Anal. Chem., 55, 987-988, (1983).183. P.G. Harrison, in The chemistry of tin, 1st Edn., Ch. 1, 8, P.G. Harrison (Ed.),Chapman and Hall, New York, (1989)184. CRC handbook of chemistry and physics, 66th Edn., C-I 26, CRC Press,Boca Raton, FL, (1985).185. W.W. Wendlandt, Thermal analysis, 3rd Edn., 214, Vol. 19 of“Chemical Analysis, a series of monographs on analytical chemistry and itsapplications”, P.J. Elving, J.D. Winefordner and l.M. Kolthoff (Eds.),John Wiley & Sons, New York, (1986).186. W.W. Wendlandt and P.K. Gallagher, in Thermal characterization ofpolymeric materials, 1st Edn., Ch. 1, pp. 27, E.A. Turi (Ed.), Academic Press,NewYork, (1981).187. M. Sterling, “Characterization of waveguides for chemical acoustic emission”,B.Sc. Thesis, Dept. of Chemistry, UBC, Vancouver, BC, (April 1991).188. Instruction manual for A&D FXIFY series electronic balance, pp. 18-20, A&DCompany Ltd., A&D Engineering Inc., Milipitas, CA, (1987).308189. Bruel & Kjaer Master Catalogue, 766, Bruel & Kjaer Canada, Richmond, BC,(May 1989).190. C.D. Johnson, Process control instrumentation technology, 3rd Edn., 440,John Wiley and Sons, New York, (1988).191. W.W. Wendlandt, Thermal analysis, 3rd Edn., Fig. 2.5, 16, Vol. 19 of“Chemical Analysis, a series of monographs on analytical chemistry and itsapplications”, P.J. Elving, J.D. Winefordner and I.M. Koithoff (Eds.),John Wiley & Sons, New York, (1986).192. R.J. Fessenden and J.S. Fessenden, Techniques and experiments fororganic chemistry, 1st Edn., pp. 255-257, Willard-Grant, Boston, MA, (1983).193. D.L. Pavia, G.M. Lampman and G.S. Kriz, Introduction to organic laboratorytechniques: A contemporary approach, 3rd Edn., pp. 415-417, SaundersCollege Publishers, Philadelphia, PA, (1988).194. W.W. Wendlandt, Thermal analysis, 3rd Edn., pp. 13-17 & 228-32, Vol. 19 of“Chemical Analysis, a series of monographs on analytical chemistry and itsapplications”, P.J. Elving, J.D. Winefordner and l.M. Kolthoff (Eds.),John Wiley & Sons, New York, (1986).195. American Society for Testing and Materials (ASTM), “Standard test methodfor thermal stability of poly(vinyl chloride) (PVC) resin”, ASTM DesignationD 4202-82 (Reapproved 1987), in Plastics 08-01, pp. 373-374, ASTMCommittee on Standards, Philadelphia, PA, (1991).196. M.A. Sharaf, D.L. Illman and B.R. Kowalski, Chemometrics, Ch. 3,pp. 102-103, Vol. 82 of “Chemical Analysis, A series of monographs onanalytical chemistry and its applications”, P.J. Elving, J.D. Winefordner andl.M. Kolthoff (Eds.), John Wiley & Sons, New York, (1986).197. A. Savitzky and M.J.E. Golay, “Smoothing and differentiation of data bysimplified least squares procedures”, Anal. Chem., 36(8), 1627-1639, (1964).198. J. Steiner, Y. Termonia and J. Deltour, “Comments on smoothing anddifferentiation of data by simplified least squares procedure”, Anal. Chem.,44(11), 1906-1 909, (1972).309199. J. Neter, W. Wasserman and G.A. Whitmore, Applied statistics, pp. 418-420,Allyn and Bacon Inc., Boston, MA, (1978).200. W.H. Press, B.P. Flannery, S.A. Teukoisky and W.T. Vetterling, Numericalrecipes: The art of scientific computing, Cambridge Univ. Press, New York,(1988).201. W.W. Wendlandt, Thermal analysis, 3rd Edn., pp. 12, Vol. 19 of“Chemical Analysis, a series of monographs on analytical chemistry and itsapplications”, P.J. Elving, J.D. Winefordner and l.M. Kolthoff (Eds.),John Wiley & Sons, New York, (1986).202. idem, ibid, 33.203. Students (W.S Gossett), “The probable error of a mean”, Biometrika, 6, 1-25,(1908).204. 5. Basan and 0. Guven, “A comparison of various isothermalthermogravimetric methods applied to the degradation of PVC”,Thermochim. Acta, 106, 169-178, (1986).205. M. Letort, “Definition et determination des deux ordres dune reactionchimie”, J. Chim. Phys., 34, 206-216, (1937).206. J.H. Flynn and L.A. Wall, “A quick direct method for the determination ofactivation energy from thermogravimetric data”, J. Polym. Sd. Polym. Letters,4, 323-328, (1966).207. E.A. Collins, J. Bares, and F.W. Billmeyer Jr., Experiments in polymerscience, 216, John Wiley & Sons, New York, (1973).208. J.N. Hay, “Applications of thermal analysis of polymers”, in Thermal analysis- techniques and applications, Ch. 8, pp. 167-1 72, E.L. Charsley andS.B. Warrington (Eds.), Royal Soc. Chem., Cambridge, England (1992).209. J.M.G. Cowie, Polymers: chemistry & physics ofmodern materials, 1St Edn.,199, Chapman and Hall, New York, (1991).210. P.M. Shiundu, Technical report to Dow Chemicals, Dept. of Chemistry, UBC,Vancouver, BC, (Jan. 1992).310211. A.W. Coats and J.P. Redfern, “Kinetic parameters from thermogravimetricdata”, Nature, 201. 68-69, (1964).212. J.R. MacCullum and M.V. Munro, “The kinetic compensation effect for thethermal decomposition of some polymers”, Thermochim. Acta, 203, 457-463,(1992).213. W.E. Brown, D. Dollimore, and A.K. Gaiwey, Reactions in the solid state,Ch. 5.4, 97, Vol. 22 of “Comprehensive Chemical Kinetics”, C.H. Bamfordand C.F.H. Tipper (Eds.), Elsevier, Amsterdam, Holland, (1980).214. D. Dollimore, T.A. Evans, Y.F. Lee and F.W. Wilburn, “Calculation ofactivation energy and pre-exponential factors from rising temperature dataand the generation of TG and DTG curves from A and E values”,Thermochim. Acta, 188, 77-85, (1991).215. M.E. Brown, Introduction to thermal analysis: techniques and applications,Ch. 13, 149, Chapman and Hall, New York, (1988).216. A.W. Coats and J.P. Redfern, “Thermogravimetric analysis - A review”,Analyst, 906-294, (1963).217. W.W. Wendlandt, Thermal Analysis, 3rd Edn., pp. 34-37, Vol. 19 of“Chemical Analysis, a series of monographs on analytical chemistry and itsapplications”, P.J. Elving, J.D. Winefordner and l.M. Kolthoff (Eds.),John Wiley & Sons, New York, (1986).218. idem, ibid, 258.219. V.J. Griffin and P.G. Laye, “Differential thermal analysis and differentialscanning calorimetry”, in Thermal analysis - techniques and appilcations,Ch. 3, 24, Fig. 5, E.L. Charsley and S.B. Warrington (Eds.), Royal Soc.Chem., Cambridge, UK, (1992).220. H.E. Kissinger, “Reaction kinetics in differential thermal analysis”,Anal. Chem., 29(11), 1702-1706, (1957).221. M. Day and D.R. Budgell, “Kinetics of the thermal degradation ofpoly(phenylene sulfide)”, Thermochim. Acta, 203, 465-474, (1992).311222. T. Ozawa, “A new method of analyzing thermogravimetric data”,Bull. Chem. Soc. Jpn., 38, 1881-1886, (1965).223. C.D. Doyle, “Estimating isothermal life from thermogravimetric data”,J. App!. Polym. Sd., fl, 639-642, (1962).224. C.F. Cullis and M.M. Hirschler, The combustion of organic polymers, Ch. 3,150, Clarendon Press, Oxford, UK, (1981).225. H.F. Mark, N.M. Bikales, C.G. Overberger and G. Menges (Eds.),Encyclopedia of polymer science and engineering, 4, 153,John Wiley & Sons, New York, (1985).226. J. Chiu, “Applications of thermogravimetry to the study of high polymers”, inThermoanalysis of fiber and fiber-forming polymers, pp. 25-43,R.F. Schwenker (Ed.), “Applied Polymer Symposia No. 2”,John Wiley & Sons, New York, (1966).227. A.l. Vogel, A textbook of quantitative inorganic analysis (including elementaryinstrumental analysis), 4th Edn., 850, J. Basset, R.C. Denney, G.H. Jefferyand J. Mendham (Eds.), Longman Group, London, (1981).228. R. Knuemann, M. Schleussner and H. Bockhorn, “TG-MS analysis ofpyrolysis of PVC and other polymers”, in Proc. 22nd Ann. Conf. Inst. Chem.Technol., pp. 36/1-36/12, ICT, Darmstadt, Germany, (1991).229. J.W. Mason, “STA!FTIR analysis of resin samples”, Analysis report, SEALJob No. 7365, SEAL Laboratories, El Segundo, CA, (April 21, 1993).230. D. I. Bower and W. F. Maddams, The vibrational spectroscopy ofpolymers,Cambridge Univ. Press, New York, (1989).231. N.B. Coithup, L.H. Daly and S.E. Wiberley, Introduction to infrared andRaman spectroscopy, 2nd Edn., Academic Press, New York, (1975).232. V.A. Zeman, “Die identifizierung einiger handelsüblicher polymerer durchthermischen abbau im massenspektrometer”, Angew. MakromoL Chem., j.,1-24, (1973).312233. D.H. Ahistrom, “Microstructure of synthetic polymers”, in Pyrolysis and G. C.in polymer analysis, Ch. 5, Table 5,249, S.A. Liebman and E.J. Levy (Eds.),Marcel Dekker, New York, (1985).234. Mass Spectrometry Data Centre and l.C.l. Ltd., Eight peak index ofmassspectra, 2nd Edn., HMSO, Mass Spectrometry Data Centre, Aldermaston,England, (1974).235. M.E. Brown, Introduction to thermal analysis: techniques and applications,Ch. 5, 54, Chapman and Hall, New York, (1988).236. H.E. Wiedemann and G. Bayer, “Advances in thermomicroscopy withsimultaneous DSC”, Thermochim. Acta, 92, 399-402, (1985).237. M.E. Brown, Introduction to thermal analysis: techniques and applications,Ch. 5, 51, Chapman and Hall, New York, (1988).238. S. Sang and J. Fuchs, “Application of thermal analysis to organic chemistry:a review”, Thermochim. Acta., 148, 325-334, (1989).0 ‘ii z >< - C) C) C -I ci) C) m -I C) Cl)C)FigureA.Schematicofthecomputer-controlledzerovoltageswitchingcircuitC)314+12VTherriocouple(Type K)A1rie1 (—) OP—07Vout1k 0 -12VVout = lOmV/CelciusFigure B. Schematic of the thermocouple ampNfier circuit.+1 2VVin0.3 3,u Fl[-<0VoutopPMIAMP is aoP—07Ez1I- 255 or OSCFREQ% <0 THEN askFNow send value to the circuit‘Port address used for LPT1 was hexadecimal 378,‘equivalent to decimal 888‘(Note. May change with machine)OUT 888, OSCFREQ%2. Recording the thermocouple readingThe routine given below was written by Dr. 0. Lee, to acquire thethermocouple reading from the second analog-to-digital channel (AD#1) of anIBM data acquisition and control adapter card (IBM DACA) installed in thePC-AT computer. The input range was set to 0-10 V.This routine serves both to initialize the IBM DACA for analog input andretrieve a value from an analog-to-digital conversion. The variable channel% isset to 1 for the thermocouple readings. The routine returns the digitized integervalue (0 - 4095) or a value of -1 if initialization mode is being used.The program code is written in QuickBASIC3193. Recordinc the balance readincThe following routine details the commands used to access the FX-400balance used for thermogravimetry experiments.Prior to operation, the software parameters on the balance must be at thefollowing settings. Details of how to do this are given on page 20 of the manualsupplied with the balance. In addition, the balance must be switched on, and notbe in standby mode.C3 Data ouput. Set to command mode (#3).C4 Data output baud rate. Set to 4800 baud (#3).C5 Terminator Set to (#0)C6 Time out Set to time limit on (#0)The program code is written in QuickBASlC.‘Open the balance as ouput device #1, connected to the serial port COM1OPEN “comi :4800,,,,LF,0P3000” FOR RANDOM AS #1‘Zero the balance reading prior to adding samplePRINT #1, “R”‘Question the balance for a reading during the experimentPRINT #1, “Q”‘Receive the data and convert to a valueINPUT #1, balanceHdr$, balanceDat$ ‘balanceDat$ contains the data‘balanceHdr$ gives header infobalanceDat? VAL(balanceDat$)The data can then be manipulated as required.