Sie sind auf Seite 1von 190

On the Material Properties and Constitutive Equations of

Piezoelectric Poly Vinylidene Fluoride (PVDF)

A Thesis

Submitted to the Faculty

of

Drexel University

by

Mitchell L. Thompson

in partial fulfillment of the

requirements for the degree of

Doctor of Philosophy

April 2002
ii

Acknowledgments

I would like to express my gratitude to Professor Horacio Sosa, my academic

advisor, for his kind assistance, invaluable guidance, and friendship over many years. I

am grateful to Professors Mahmoud El-Sherif, Wei-Heng Shih, Sorin Siegler, and Dr.

Naum Khutoryansky for serving on my Ph.D. committee and for their support,

constructive criticism, and valuable feedback.

I would like to thank my children, Corinne, Marie, and Lee, for their understanding

all the times when I was busy, and I would like to thank my wife, Barbara Nicholson, for

her unfailing support and eternal patience.


iii

Table of Contents

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

List of figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Constitutive Equations and Material Properties . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Scope and Objectives of this Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2. PIEZOELECTRIC MATERIALS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Piezoelectric Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3. CONSTITUTIVE EQUATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1 Constitutive Equations for Piezoelectric Materials . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Constitutive Equations for Viscoelastic Materials . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Constitutive Equations Coupling Piezoelectricity and Viscoelasticity . . . . . . . . 33
3.4 Time - Temperature Superposition (TTS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4. MEASUREMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.1 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2 Equipment and Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 System Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.4 Calibration/Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.5 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.6 Measurement Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

5. DATA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.1 1 Axis (Stretch Direction) Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2 2 Axis (Transverse Direction) Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.3 Effect of Electrode Metal on Material Property Measurements . . . . . . . . . . . . . 76
5.4 Dielectric Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6. ELASTIC AND PIEZOELECTRIC DATA ANALYSIS . . . . . . . . . . . . . . 82


6.1 TTS master Curves for all Elastic and Piezo Coefficient Data Sets . . . . . . . . . . 82
6.2 Summary TTS Master Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.3 Discussion of Summary Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.4 Effects of Stress Level on 1 Axis Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 96
iv

6.5 Comparison and Normalization of Elastic and Piezoelectric Data . . . . . . . . . . . 98


6.6 Novel TTS expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

7. DIELECTRIC DATA ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109


7.1 TTS Expansion of Dielectric Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
7.2 The Cole-Cole Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.3 The Vogel-Fulcher Modified Cole-Cole Equation . . . . . . . . . . . . . . . . . . . . . . 113

8. CONSTITUTIVE EQUATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119


8.1 Constitutive Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
8.2 Useful Frequency and Temperature Bounds . . . . . . . . . . . . . . . . . . . . . . . . . . 122

9. EXPERIMENTAL VERIFICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125


9.1 Experimental Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2 Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
9.3 Experimental Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

10. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135


10.1 List of Key Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
10.2 Key Contributions of this Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

LIST OF REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

APPENDIX A: NOMENCLATURE AND SYMBOLS . . . . . . . . . . . . . . . 147

APPENDIX B: FERROELECTRICITY, ELECTROSTATICS, AND


FIELD EQUATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

APPENDIX C: MEASUREMENT EQUIPMENT . . . . . . . . . . . . . . . . . . 156

APPENDIX D: DERIVATION OF STANDARD LINEAR MATERIAL


BEHAVIOR AS A FUNCTION OF FREQUENCY . . . . . . . . . . . . . . . . . 157

APPENDIX E: ADDITIONAL TTS MASTER CURVE DATA . . . . . . . 160

VITA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
v

List of Tables

2.1 Typical Properties of Commercially Significant Piezo Materials . . . . . . 17

4.1 Sample Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6.1 Effects of Applied Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97


vi

List of Figures

2.1.1 Crystal symmetry diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.1.2 Ferroelectric (left) and Paraelectric crystal phases of a perovskite. Note


the asymmetry in the ferroelectric form . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2.1 Space filling model of the polar $ crystal form (top) and the non polar
" crystal form, bottom. Black = F, white = H, from [1] . . . . . . . . . . . . 16

2.2.2 Crystallographic axes. In PVDF, 1 corresponds to the draw direction,


2 to the transverse direction, and 3 to the thickness (also the poling
axis) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2.3 Thickness (a) vs. length (b) modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2.4 Comparison of d31 over temperature (10 Hz) . . . . . . . . . . . . . . . . . . . . 18

2.2.5 Comparison of permittivity over temperature (1 kHz) . . . . . . . . . . . . . . 19

2.2.6 Comparison of dielectric loss tangent over temperature (1kHz) . . . . . . . 19

2.2.7 Comparison of elastic modulus over temperature (10 Hz) . . . . . . . . . . . 20

2.2.8 Comparison of mechanical loss tangent over temperature (10 Hz) . . . . . 20

3.2.1 Classical viscoelastic material elements. Components are springs and


dashpots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.2.2 Normalized DeBye relaxation of an SLM element . . . . . . . . . . . . . . . . . 28

3.2.3 Example of multiple Maxwell element relaxation spectra (from [74]) . . 30

3.4.1 Elastic compliance data as taken . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.4.2 Elastic data TTS expanded about 50°C . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.4.3 ‘A’ curves are Arrhenius fit to measured data, expanded about 50°C . . 40

4.2.1 Piezotron test rail showing exciter, thermal chamber, force and
displacement gages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.2.2 Overall view of piezotron, showing DSA and other instrumentation. . . . 45


vii

4.2.3 Front view of piezotron test area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

4.2.4 Close up of test cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

4.2.5 Piezotron schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.2.6 Piezotron spring - mass - dashpot schematic . . . . . . . . . . . . . . . . . . . . . 49

4.2.7 Displacement/force spectra from piezotron with metal sample. Small


noise is caused by extremely small displacements. . . . . . . . . . . . . . . . . . 51

4.2.8 Sample as held in clamps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.2.9 Rod support diaphragm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.3.1 Typical signal analyzer configuration state (HP3562A) . . . . . . . . . . . . . 55

4.3.2 Typical example of uncompensated data from the DSA. dB is re


(displacement/load), phase angle is between displacement and load . . . . 56

4.4.1 Piezotron instrumentation electrical noise equivalents . . . . . . . . . . . . . . 57

4.4.2 Electrometer compensation measured driving a voltage into a capacitor


to mimic the PVDF sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.5.1 Picture of a complete PVDF film sample . . . . . . . . . . . . . . . . . . . . . . . . 61

4.6.1 Effects of heating PVDF above its annealing temperature, showing


permanent increase in compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5.1.1 Typical elastic storage compliance, 1 axis . . . . . . . . . . . . . . . . . . . . . . . 64

5.1.2 Typical elastic loss compliance phase angle, 1 axis . . . . . . . . . . . . . . . . 66

5.1.3 Typical piezo storage compliance data, 1 axis . . . . . . . . . . . . . . . . . . . . 66

5.1.4 Typical piezo loss compliance phase angle, 1 axis . . . . . . . . . . . . . . . . . 68

5.2.1 Typical elastic storage compliance, 2 axis . . . . . . . . . . . . . . . . . . . . . . . 69

5.2.2 Typical elastic loss compliance phase angle, 2 axis . . . . . . . . . . . . . . . . 69

5.2.3 Typical piezo storage compliance data, 2 axis . . . . . . . . . . . . . . . . . . . . 71

5.2.4 Typical piezo loss compliance phase angle, 2 axis . . . . . . . . . . . . . . . . . 72


viii

5.2.5 Schematic of electrically driven test circuit. Current is measured by


voltage drop across 1 Gohm 610C resistor. . . . . . . . . . . . . . . . . . . . . . 74

5.2.6 Dielectric behavior of PVDF film at 80°C compared to a low loss


capacitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2.7 Dielectric behavior of PVDF film at 80°C compared to a low loss
capacitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

5.2.8 Electrical models used in trying to fit anomalous d32 data . . . . . . . . . . . 76

5.3.1 Effect of metal electrodes on elastic compliance, 2 axis . . . . . . . . . . . . . 77

5.3.2 Effect of electrodes on elastic compliance phase angle, 2 axis . . . . . . . . 78

5.4.1 Permittivity of PVDF vs temperature and frequency . . . . . . . . . . . . . . . 82

5.4.2 Temperature projection of Fig 5.4.1 data . . . . . . . . . . . . . . . . . . . . . . . 80

5.4.3 Frequency projection of Fig 5.4.1 data . . . . . . . . . . . . . . . . . . . . . . . . . 80

5.4.4 PVDF dielectric loss tangent vs temperature and frequency. Note that
the frequency axis is a linear scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

6.1.1 TTS expansion of data from Figure 5.1.1. 1 axis elastic storage
compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

6.1.2 TTS expansion of data from Figure 5.1.2. 1 axis elastic loss
compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

6.1.3 TTS expansion of data from Figure 5.1.3. 1 axis piezo storage
compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

6.1.4 Comparison of elastic and piezo storage compliances normalized at


room temperature and 1 Hz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

6.1.5 TTS expansion of data from Figure 5.1.4. 1 axis piezo loss
compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

6.2.1a TTS expansion of 1 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . . 87

6.2.1b TTS expansion of 1 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 88

6.2.1c TTS expansion of 1 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . . 88

6.2.1d TTS expansion of 1 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . . 89


ix

6.2.2a TTS expansion of 2 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . . 89

6.2.2b TTS expansion of 2 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 90

6.2.2c TTS expansion of 2 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . . 90

6.2.2d TTS expansion of 2 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . . 91

6.2.3 All TTS expansion curves for 1 axis elastic compliance, 50°C . . . . . . . . 91

6.2.4 All TTS expansion curves for 1 axis elastic loss, 50°C . . . . . . . . . . . . . 92

6.2.5 All TTS expansion curves for 1 axis piezo compliance, 50°C . . . . . . . . 92

6.2.6 All TTS expansion curves for 1 axis piezo loss, 50°C . . . . . . . . . . . . . . 93

6.2.7 All TTS expansion curves for 2 axis elastic compliance, 50°C . . . . . . . . 93

6.2.8 All TTS expansion curves for 2 axis elastic loss, 50°C . . . . . . . . . . . . . 94

6.2.9 All TTS expansion curves for 2 axis piezo compliance, 50°C . . . . . . . . 94

62.10 All TTS expansion curves for 2 axis piezo loss, 50°C . . . . . . . . . . . . . . 95

6.4.1 Comparison of 1 axis elastic compliance at different strain levels . . . . . . 97

6.4.2 Comparison of 1 axis piezo compliance at different strain levels . . . . . . 98

6.5.1 All 1 axis elastic compliance curves normalized at 1Hz, 50°C . . . . . . . . 99

6.5.2 All 1 axis piezo compliance curves normalized at 1Hz, 50°C . . . . . . . . . 99

6.6.1 Elastic compliance 1 axis master TTS curves for all reference
temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

6.6.2 Data in Fig 6.6.1 normalized at 1 Hz and a master curve fit . . . . . . . . . 102

6.6.3 Piezo compliance 1 axis master TTS curves for all reference
temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

6.6.4 Data in Fig 6.6.3 normalized at 1 Hz and a master curve fit . . . . . . . . . 103

6.6.5 Elastic compliance 1 axis measured data compared to predictions


using modified Arrhenius equation (A suffix) . . . . . . . . . . . . . . . . . . . 105
x

6.6.6 Piezo compliance 1 axis measured data compared to predictions using


modified Arrhenius equation (A suffix) . . . . . . . . . . . . . . . . . . . . . . . . 106

6.6.7 Elastic loss TTS master curves for all reference temperatures . . . . . . . 106

6.6.8 Data of Fig 6.6.7 normalized at 1 Hz and a master curve fit . . . . . . . . 107

6.6.9 Piezo loss TTS master curves for all reference temperatures . . . . . . . . 107

6.6.10 Data of Fig 6.6.9 normalized at 1 Hz and a master curve fit . . . . . . . . 108

7.1.1 WLF TTS expansion of dielectric permittivity of PVDF . . . . . . . . . . . 111

7.3.1 Curve fit to permittivity data. First set of temperatures listed are data
curves, second set are curve fits using (7.4) - (7.8) . . . . . . . . . . . . . . . 116

7.3.2 Curve fit to dielectric loss data. First set of plots are data, second set
are curve fits using (7.12) - (7.14) . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

9.1.1 Layout and nomenclature for verification test. . . . . . . . . . . . . . . . . . . 126

9.2.1 Schematic for verification test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

9.3.1 Voltage ratio comparison, measured vs. calculated . . . . . . . . . . . . . . . 134

B.1.1 Ferroelectric poling event, applied field cycled from 0 to maximum,


then back to 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

B.1.2 Displacement/electric field curve for a ferroelectric material showing


typical switching hysteresis and linear dielectric regions . . . . . . . . . . . 151

B.2.1 Displace charged, Q, resulting from applied voltage, V, in a linear


dielectric material. Parallel plate capacitor shown . . . . . . . . . . . . . . . . 154

D.1 Standard linear material model showing two springs and a dashpot . . . 157

E.1a TTS expansion of 1 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . 161

E.1b TTS expansion of 1 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . 161

E.1c TTS expansion of 1 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . 162

E.1d TTS expansion of 1 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 162

E.2a TTS expansion of 1 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . 163


xi

E.2b TTS expansion of 1 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . 163

E.2c TTS expansion of 1 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . 164

E.2d TTS expansion of 1 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 164

E.3a TTS expansion of 1 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . 165

E.3b TTS expansion of 1 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . 165

E.3c TTS expansion of 1 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . 166

E.3d TTS expansion of 1 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 166

E.4a TTS expansion of 1 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . 167

E.4b TTS expansion of 1 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . 167

E.4c TTS expansion of 1 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . 168

E.4d TTS expansion of 1 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 168

E.5a TTS expansion of 1 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . 169

E.5b TTS expansion of 1 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . 169

E.5c TTS expansion of 1 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . 170

E.5d TTS expansion of 1 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 170

E.6a TTS expansion of 2 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . 171

E.6b TTS expansion of 2 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . 171

E.6c TTS expansion of 2 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . 172

E.6d TTS expansion of 2 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 172

E.7a TTS expansion of 2 axis elastic compliance, 50°C . . . . . . . . . . . . . . . . 173

E.7b TTS expansion of 2 axis elastic loss, 50°C . . . . . . . . . . . . . . . . . . . . . 173

E.7c TTS expansion of 2 axis piezo compliance, 50°C . . . . . . . . . . . . . . . . 174

E.7d TTS expansion of 2 axis piezo loss, 50°C . . . . . . . . . . . . . . . . . . . . . . 174


xii

ABSTRACT
On the Material Properties and Constitutive Equations of
Piezoelectric Poly Vinylidene Fluoride (PVDF)
Mitchell L. Thompson
Horacio A. Sosa

The complex elastic and piezoelectric material properties of uniaxially oriented

PVDF were measured over a temperature range of 20°C to 80°C and a frequency range of

0.01 Hz to 100 Hz using specially developed equipment. A WLF Time-Temperature

Superposition (TTS) expansion of the resulting data confirmed previously reported

thermorheologically simple (TRS) behavior for the elastic properties and established that

the piezoelectric behavior is also TRS. The form of the piezoelectric TTS expansion was

identical to that of the elastic expansion, implying that the mechanisms responsible for

viscoelastic behavior are also responsible for visco-piezoelectric behavior. A modified

Arrhenius equation was developed to allow a broader TTS expansion. Empirical

equations describing both viscoelastic and piezoelectric properties over temperature and

frequency were constructed using these expansions and compared to measured data.

The dielectric properties of PVDF were measured over the range of -60°C to 80°C

and from 20 Hz to 1MHz. Cole-Cole and Vogel-Fulcher equations were combined and

the resulting dielectric equations were confirmed by comparison to measured data.

All material property relationships were assembled into a constitutive framework

for piezoelectric materials in the frequency domain. A novel technique was employed to

test the ability of the elastic, piezoelectric, and dielectric property equations to predict

behavior outside of the measured temperature/frequency range with excellent results.


1

CHAPTER 1 INTRODUCTION

1.1 Background

Piezoelectric materials are used to transduce electrical and mechanical energy.

Piezoelectric material technology has enabled a wide variety of commercially successful

sensors and actuators. These devices range in complexity and sophistication from the

ultrasound arrays used in biomedical imaging to the tonal buzzers used in automobile

horns, from high intensity focused ultrasound (HIFU) arrays which can thermally ablate

tumors to the speaker elements in talking greeting cards [1, 2].

The vast majority of piezoelectric materials found in the marketplace today are

inorganic ceramics such as lead titanate (PbTiO3), lead zirconium titanate (PbZrTiO3),

lithium tantalate (LiTaO3), and barium titanate (BaTiO3). Most of these perovskite

materials were pioneered in the late 1940's and 1950's [3], and are characterized by high

elastic moduli, high dielectric constant, low elastic and dielectric lossiness, and high

electro-mechanical coupling factors. Although modern piezoelectric ceramic materials

have proven successful in many applications, they have a number of inherent limitations:

i) Low yield strains, on the order of 0.2%, eliminate ceramics from high strain

sensing applications such as flexure in helicopter rotor blades and in fishing rods.

Biomedical applications such as foot strike force or bite pressure are also impractical for

ceramic materials.

ii) Brittleness makes these materials prone to fracture and crack propagation, and

makes the use of ceramic materials in impact, shock, and ordinance applications

impractical. Brittle piezoelectric materials used in situations undergoing positive an


2

negative stresses must often be preloaded into compression to ensure mechanical stability.

iii) The density of ceramic materials is high, creating problems for weight sensitive

applications such as naval hull mounted and geophysical towed array sonar systems.

iv) The acoustic impedance of ceramic materials (a function of density and stiffness) is

high, providing poor acoustic coupling to lower impedance materials like water or human

tissue.

v) The raw materials and basic processing costs of piezoelectric ceramic and single

crystal materials are relatively high per unit volume, and batch to batch variations in

material properties can be significant.

These limitations may be overcome in specific applications through the use of

polymeric piezoelectric materials. Poly vinylidene fluoride, PVDF, is the foremost of this

class of piezoelectric materials and as a polymer it compares favorably against ceramics in

robustness, toughness, yield and useful operating strain, density, acoustic matching to

tissue and water, chemical inertness, relative ease of manufacture, low material and

processing costs, and maximum sensor/actuator size.

1.2 Constitutive Equations and Material Properties

To understand and to optimize the performance of piezoelectric transducers the

designer must have both constitutive equations and the necessary material property data.

Linear electro-elastic constitutive equations are commonly used to express the coupling of

dielectric, elastic, and piezoelectric properties in piezoelectric materials [4]. The material

properties used in these equations are treated as single valued constants in practical use,

even though they may be functions of temperature, frequency, field, and/or stress.
3

Manufacturers and users of piezoelectric ceramic materials often measure material

property values as a function of temperature [5], and occasionally as a function of field or

stress [6], allowing designers to graphically select material properties for conditions of

interest. Piezo ceramic manufacturers do not normally provide material property data as a

function of frequency [5-7] because most properties are relatively time invariant.

Equations yielding material properties as functions of temperature and frequency are

generally not available.

Material property data reported graphically does not allow closed form numerical

solutions to the constitutive equations over a range of temperature or frequency. Further,

it is evident that material properties provided in this manner cover a discrete range of

temperatures or frequencies, and therefore the constitutive equations cannot be solved for

conditions outside of this range.

Piezoelectric PVDF is a polymer, and as such its material properties are affected

by temperature and frequency. The leading commercial supplier of this material publishes

only limited isochronal or isothermal data [62]. There are other studies of PVDF material

properties in the literature but there is no single work which provides a large range of

viscoelastic, piezoelectric, and dielectric behavior. Again, no functions describing PVDF

material property behavior are available.

Since functions have not been developed to fit the material properties of inorganic

piezoelectric ceramics because of their complexity, functions for polymeric PVDF may

seem unrealistic. The problem is less difficult than may be expected, however, because

with certain polymers it is possible to extrapolate elastic properties over a very broad

range of temperature and frequency from a limited set of measurements. The Time-
4

Temperature Superposition (TTS) method used in this extrapolation formalizes the

empirical equivalence of decreasing temperature and increasing frequency (or increasing

temperature and decreasing frequency) found in the elastic properties of viscoelastic

polymers. A considerable body of work exists regarding the use of this method, as

presented in Ferry’s definitive book [12]. The WLF TTS method described in [12] is

empirical but is derived from detailed studies of the mechanisms of viscoelastic behavior in

different types of polymers and accommodates the significant morphological variations

among polymer classes. This work will use the WLF TTS approach to extrapolate the

elastic properties of PVDF, and will also attempt to use this technique to extrapolate

piezoelectric and dielectric properties.

1.3 Previous Work

Although there is an exhaustive body of general research into piezoelectric PVDF,

as discussed in chapter 2, a search for references which specifically explore the viscoelastic

properties of piezoelectric PVDF returned only four articles. All of these were authored

by research groups at Montana State University [8-11]. Schmidt [8] lays out a general

method for studying the viscoelastic properties of PVDF and suggests an extension of this

effort to include piezoelectric properties as future work. Vinogradov and Holloway [9]

show experimental results and extrapolation of elastic properties across a broad frequency

range through the use of a Time-Temperature Superposition technique, but include no

piezoelectric or dielectric data. As with PVDF material property data reported by other

researchers, no functions describing viscoelastic and piezoelectric material properties with

temperature or frequency are discussed.


5

In a continuation of [9], Vinogradov and Holloway introduced some electrical

measurements in an attempt to determine piezoelectric coefficients [10]. These relatively

unsophisticated measurements were taken using an oscilloscope to capture PVDF output

voltage and phase angle relative to the applied strain under known conditions of

temperature and frequency. They suggested in their conclusions that the piezoelectric

properties of PVDF were time and temperature invariant. A review of their approach led

this author to conclude that a more detailed empirical investigation was warranted.

Vinogradov [11] followed these reports with a discussion of the relaxation

modulus and creep compliance of PVDF which was based upon the elastic data measured

earlier. She constructed equations for curve fitting these time domain functions and

concluded that power law relationships are appropriate in the time domain for these elastic

properties. She did not construct piezoelectric constitutive equations using these results

and did not discuss piezoelectric or dielectric material properties. Frequency domain

material behavior was not discussed.

Other previous works of some importance to this thesis include classical

constitutive equations, derived through thermodynamic formalism or through a

generalization of viscoelasticity, which are discussed in chapter 3. These integral

equations are useful to show the various coupling effects in piezoelectric materials. Also

of some importance are equations developed to describe dielectric behavior, as discussed

in chapter 7. Although these fundamental equations provide some insight, they are of little

practical use to the designer when the relevant material properties are not provided as

functions of time and temperature.


6

1.4 Scope and Objectives of this Thesis

It is the intent of this thesis to formulate more direct and effective constitutive

relationships for piezoelectric PVDF by determining the material property coefficients as

functions of frequency and temperature.

The specific goals of this thesis include:

i) Design and build an apparatus suitable for testing the elastic and piezoelectric

material behavior of thin film PVDF as a function of temperature and frequency.

ii) Measure the dielectric and the in-plane elastic and piezoelectric properties of

PVDF over a wide range of temperatures and frequencies. Present a complete set of data.

iii) Using this data and the WLF Time-Temperature Superposition principal, establish

useful functions to approximate these material properties over a much broader frequency

and temperature range. Further, a significant objective is to determine whether or not the

piezoelectric behavior can be predicted using the TTS approach in a manner identical to

that used to predict elastic properties. This has not been addressed in the literature.

iv) Measure dielectric properties over temperature and frequency, and investigate the

use of the TTS approach to create functions for this complex behavior.

v) Incorporate these elastic, piezoelectric, and dielectric functions into temperature

and frequency dependent constitutive equations.

vi) Validate the material property functions empirically by measuring PVDF behavior

in a frequency and/or temperature region outside of the measured data.


7

1.5 Organization

This work is organized as follows: in chapter 2 the fundamentals of piezoelectricity

and a discussion of piezoelectric polymers are presented as background information.

Chapter 3 contains a summary of the relevant constitutive equations for piezoelectricity

and classical viscoelasticity as well as an introduction to the WLF Time-Temperature

Superposition principle. Chapter 4 details the apparatus and experimental procedures

used to make material property measurements. Elastic, piezoelectric, and dielectric data

are presented in chapter 5. The elastic and piezoelectric data is expanded using TTS, and

the resulting master curves are shown in chapter 6. Functions are also developed to match

this expanded data over a broad temperature and frequency range. The dielectric data is

analyzed in chapter 7 and the dielectric properties are given as functions of frequency and

temperature. The functions developed in chapters 6 and 7 are incorporated into

appropriate constitutive equations in chapter 8. Chapter 9 describes the apparatus and

procedure by which the material property functions are verified outside of the measured

frequency range. Summary and conclusions are presented in chapter 10. Appendix A

contains a list of symbols and nomenclature used in this work.


8

CHAPTER 2 PIEZOELECTRIC MATERIALS

2.1 Background

All crystal structures can be classified into one of 32 possible forms of crystal

symmetry [13]. Eleven of these forms are centrosymmetric. Of the remaining 21 non -

centrosymmetric groups, 20 are known to be piezoelectric, meaning these materials

produce an electric surface charge in response to applied mechanical stress. In 10 of these

crystal groups there is a permanent electric dipole, and the equilibrium of the electrostatic

potential caused by this dipole is distorted by mechanical stress (piezoelectricity) or

temperature change (pyroelectricity). Certain pyroelectric materials can be further

classified as ferroelectric materials, as shown in the summary diagram in Figure 2.1.1.

Examples of the non centrosymmetric (ferroelectric) and the centrosymmetric (para-

electric) crystal structures typical of common ceramic ferroelectric perovskites (ABO3) are

shown in Figure 2.1.2. An introduction to ferroelectricity is provided in Appendix B.

Rochelle salt is the earliest reported piezoelectric material [14, 15]. It was first

produced in France in 1665 for medicinal use by pharmacist Elie Seignette. Brewster

discussed the pyroelectric properties of this material in 1824. Pioneering work on the

direct piezoelectric effect (stress Y charge) in this material was presented by Jacques and

Pierre Curie in 1880. In 1881, Lippmann proposed the converse piezoelectric effect based

upon thermodynamic principals and this outcome was verified by the Curie brothers,

whose early efforts included natural crystals such as tourmaline, topaz, quartz, in addition

to Rochelle salt. Their work resulted in the development of the first scientific
9

Figure 2.1.1 Crystal symmetry diagram.

Figure 2.1.2 Ferroelectric (left) and Paraelectric crystal phases of a


perovskite. Note the asymmetry in the ferroelectric form.
10

instruments using the piezoelectric effect to measure force in the direct mode and high

voltages in the converse mode.

Research on these piezoelectric materials continued throughout the early twentieth

century, in part stimulated by military interests arising from World War I. A major

outcome of this research was the use of piezoelectric quartz in acoustic projectors and

receivers. Ferroelectricity, described by hysteresis behavior, was suggested by

Schrödinger in 1912, and was more firmly established by Valasek in the 1920's primarily

working with triglycine sulphate (TGS) and Rochelle salt [15].

World War II brought increased attention to the study of piezoelectric materials,

especially for use in radio communications and underwater acoustic applications The

discovery of BaTiO3 and other highly piezoelectric materials with a perovskite structure in

the latter half of the 1940's by researchers in the US, England, Russia, and Japan led to a

renewed interest in the study of ferroelectric materials. In the early 1950's ferroelectric

lead zirconium titanate ceramics (PZT) were introduced with even higher performance.

This precipitated further work and led to a greater understanding of the structure and

behavior of polycrystalline ferroelectrics. Extensive research and development throughout

the last four decades has lead to the use of piezoelectric ceramics in a wide variety of

commercial, consumer, and military applications [2] ranging from inexpensive tonal

buzzers used in toys to sophisticated ultrasound imaging transducers used in the latest

medical procedures, from consumer fish finders to highly complex submarine sonar

systems, and from simple grill ignitors to all manner of force and acceleration sensors.
11

2.2 Piezoelectric Polymers

Very weak piezoelectric effects have been reported in biological tissue and other

organic materials since the 1950's, but it was not until 1969 that useful levels of

piezoelectricity were discovered in non biological polymers. While conducting an

investigation into the electret properties of various synthetic polymers in that year, Kawai

[18] discovered that drawn and electrically poled poly vinylidene fluoride showed

exceptional piezoelectricity. A number of subsequent researchers have studied the

piezoelectric, ferroelectric, dielectric, pyroelectric, electrocaloric, photovoltaic,

photoelastic, and optical properties of PVDF [19-50]. Researchers have reported

piezoelectric behavior in a number of other polymers as well [1, 51-59]: Poly vinylidene

cyanide and its copolymers; aromatic and aliphatic polyureas; poly vinyl chloride; aromatic

polyamides (odd nylons); PVDF copolymers with trifluoroethylene (P[VDF-TrFE]),

tetrafluoroethylene (P[VDF-TFE]) , and hexafluoropropylene (P[VDF-HFP]); PVDF

blends with poly methyl methacrylate (PMMA); poly vinyl fluoride; poly vinyl acetate; and

ferroelectric liquid crystal polymers. Lang [60] has published a number of bibliographies

on the piezoelectricity and pyroelectricity of polymers listing well over a thousand

references. In addition, there has been considerable study of piezoelectric polymer-

ceramic composite materials. To the author’s knowledge, with the exception of those

specific references identified in chapter 1, none of this work has addressed the viscoelastic

behavior of PVDF or the effect of this behavior on the piezoelectric performance of PVDF

in a significant or systematic manner.

In the -40°C to +100°C temperature range, PVDF and the P(VDF-TrFE)

copolymers show the most significant piezoelectric behavior by far. The VDF/TrFE
12

copolymer composition and synthesis processes have a large effect on the elastic,

dielectric, and piezoelectric properties of the resulting material. Also, these copolymers

are expensive (>$250/lb resin), extremely difficult to produce, and account for a very

small percentage of the commercially offered piezoelectric polymers. In contrast, PVDF

is inexpensive ($8/lb resin), accounts for the virtually all of the commercially significant

piezoelectric polymer applications, and has properties which are relatively unaffected by

synthesis conditions. For these reasons this work will focus on PVDF.

Poly vinylidene fluoride, or PVDF, has a chemical composition of (CH2-CF2)n and

is more formally known as 1,1 difluoro-ethylene. It is a semi crystalline polymer with a

crystal volume fraction of about 50% after melt extrusion. The crystal structure after melt

processing is a classically spherulitic, non-polar " phase with a helical (TGTG’)

configuration. After melt extrusion, strong mechanical orientation of the polymer is

required to induce the solid state phase transition from the " phase to a highly polar $

phase with a planar zig zag, all trans (TTTT) chain configuration. Figure 2.2.1 shows

space models of both the " and $ crystal phases. Draw ratios of 3.5 to 5 are needed with

accurate temperature and draw rate control. Subsequent poling of the resulting film rolls

is done under controlled temperatures at fields in excess of 100 V/µm [61]. The films are

then thermally annealed to allow the controlled mechanical relaxation necessary to provide

commercially stable material. The material properties of piezoelectric PVDF are made

more anisotropic because of this process history, as the large degree of microscopic order

resulting from the orientation reduces the in plane 1 axis randomness and makes the

macroscopic piezoelectric behavior more consistent with that of the microscopic 2mm

point group symmetry. The detailed morphology and fundamental basis for piezoelectric
13

behavior in PVDF is discussed extensively in [18-20, 22, 23].

The particular crystal symmetry of a piezoelectric material determines which

components of the permittivity, piezoelectric, and stiffness tensors are non-zero and

unique. Piezoelectric coefficients are defined by the relationships between the elastic (T,

S) and dielectric (E, D) properties of piezoelectric materials, discussed in detail in the

following chapter. For example, for PVDF and other materials having 2mm symmetry, we

have in matrix form [3, 4]

*,11 0 0 *
,= * 0 ,22 0 * (2.1a)
* 0 0 ,33 *

*0 0 0 0 d15 0 *
d= *0 0 0 d24 0 0 * (2.1b)
*d31 d32 d33 0 0 0 *

*s11 s12 s13 0 0 0 *


*s12 s22 s23 0 0 0 *
s= *s13 s23 s33 0 0 0 * (2.1c)
*0 0 0 s44 0 0 *
*0 0 0 0 s55 0 *
*0 0 0 0 0 s66 *

The typical crystallographic coordinate axes used in (2.1) are shown in Figure 2.2.2.

Writing indices 11, 22, 33, 23, 31, 12 as 1, 2, 3, 4, 5, and 6 allows the third order tensor d

to be written as shown [13].

Due to the difficulties associated with the orientation and polarization of PVDF it

is only produced commercially in films, typically between 6 and 150 µm thick. Although
14

biaxial orientation (stretching in both in-plane axes) is possible, it produces film with lower

but laterally isotropic piezoelectric properties. The vast majority of commercially available

PVDF is uniaxially drawn as this provides a much higher level of piezoelectric response in

the axis of orientation (1 axis). Uniaxially oriented films are studied in this work.

PVDF is used in the length extensional mode (1 axis) in well over 90% of practical

applications [62] as a result of the in plane anisotropy. The motivation for stressing the 1

axis as opposed to the 2 axis can be easily understood by comparing the piezoelectric

coefficients in Table 2.1, where d32 . d31/10 (d32 is the 2 axis coefficient, d31 is the 1 axis

coefficient, as will be discussed in detail in chapter 3). The motivation for utilizing the 1

axis as opposed to the 3 axis is not so apparent, as their piezoelectric coefficients are

similar. To highlight the differences, it is helpful to compare the voltage output from

applying a force load to the flat face of a piece of film (in the 3 axis) laying on a rigid

surface to the voltage obtained from applying the same load to the edge of the film in the 1

direction (see Figure 2.2.3). In each case the simple form of the equation for open circuit

output voltage is [62]

(2.2)

Because d31 .d33, this reduces to

(2.3)

The thickness of the film is typically in the range of 6 to 150 µm, and the length of even
15

the smallest sensors is generally at least 2 to 3 orders of magnitude larger. Clearly

whenever possible designers will utilize the 1 axis, and the major effort in this work will

focused on the material properties of this axis.

It is useful to contrast the properties of PVDF with several other commercially

significant ferroelectrics in order to emphasize the distinct differences between the

polymeric and ceramic materials. Table 2.1 contains published material property data, and

large differences are apparent in density, dielectric constant, and elastic compliance.

However large the differences in this data, measured at a single frequency and

temperature, the differences throughout the normal range of operating temperatures and

frequencies are more severe. To emphasize this, Figures 2.2.4 through 2.2.8 compare

additional published properties of PVDF [1] and PZT5A [5] in the temperature range

from 0°C to 80°C. It is important to note that these curves represent behavior measured

at the frequencies indicated. The viscoelastic nature of PVDF can be clearly seen in the

elastic storage and loss moduli, whereas the PZT5A modulus is effectively constant in this

temperature range.
16

Figure 2.2.1 Space filling model of the polar $ crystal form (top) and the
non polar " crystal form, bottom. Black = F, white = H (from [1]).

Figure 2.2.2 Crystallographic axes. In PVDF, 1 corresponds to the draw


direction, 2 to the transverse direction, and 3 to thickness (also the poling
axis).
17

Table 2.1 Typical Properties of Commercially Significant Piezo Materials

Parameter PZT4 [5] PZT5A [5] BaTiO3 [6] PVDF [62]


d31 pC/N -122 -171 -78 25
d32 pC/N -122 -171 -78 2
d33 pC/N 285 374 190 -33
d15 pC/N 495 585 260 27
KT33 1300 1700 1700 12
KT11 1475 1730 1450 -
D kg/m3 7600 7700 5700 1880
tan *m .002 .013 .003 .06
tan *e .004 .020 .01 .02
Curie Temp (°C) 325 365 115 125*
L1 m/s (acoustic) 3300 2800 4400 1450
L3 m/s (acoustic) 4000 3760 5040 2250
s11 pm2/N 12.3 16.4 9.1 365
s12 pm2/N -3.8 -5.1 -2.8 -110
s13 pm2/N -4.8 -5.8 -2.9 -209
s21 pm2/N -3.8 -5.1 -2.8 -110
s22 pm2/N 12.3 16.4 9.1 424
s23 pm2/N -4.8 -5.8 -2.9 -192
s33 pm2/N 15.5 18.8 9.5 472
s44 pm2/N 39.0 47.5 22.8 -
s66 pm2/N 32.7 44.3 23.6 -
kt .58 .60 .36 .12
* Often considered a useful upper working temperature. The actual Curie temperature of
PVDF is above the melt temperature (.170°C) but the transition is diffuse.
18

Figure 2.2.3 Thickness (a) versus length (b) modes.

d31 versus Temperature

40%
30%
% Deviation from

20%
Room Temp

10%
0%
-10%
-20%
0 20 40 60 80
Temperature (C)

PZT5A PVDF

Figure 2.2.4 Comparison of d31 over temperature (10 Hz).


19

Permittivity versus Temperature

40%

30%
% Deviation from
Room Temp
20%

10%

0%

-10%
0 20 40 60 80
Temperature (C)

PZT PVDF

Figure 2.2.5 Comparison of permittivity over temperature (1kHz).

Dielectric Loss Tangent versus Temperature

50%

40%
% Deviation from

30%
Room Temp

20%

10%

0%
0 20 40 60 80

Temperature (C)

PZT PVDF

Figure 2.2.6 Comparison of dielectric loss tangent over temperature


(1kHz).
20

Elastic Modulus versus Temperature

40%

20%
% Deviation from Room
Temp 0%

-20%

-40%

-60%
0 20 40 60 80
Temperature (C)
PZT PVDF

Figure 2.2.7 Comparison of elastic modulus over temperature (10 Hz).

Mechanical Loss Tangent versus Temperature

80%

60%
% Deviation from
Room Temp

40%

20%

0%
0 20 40 60 80

Temperature (C)

PZT PVDF

Figure 2.2.8 Comparison of mechanical loss tangent over temperature (10


Hz).
21

CHAPTER 3 CONSTITUTIVE EQUATIONS

This chapter introduces the classical constitutive equations developed for inorganic

piezoelectric materials, discusses the constitutive equations of uncoupled viscoelastic

materials, and summarizes relevant work regarding constitutive relations combining

viscoelasticity and piezoelectricity.

The Time - Temperature Superposition (TTS) method is then presented as a

procedure for predicting material properties outside of the range of measured data. It is a

major focus of later chapters to apply this technique to both the viscoelastic and

piezoelectric properties of PVDF. The information in this chapter is intended to

familiarize the reader with the fundamentals upon which the remainder of this work is

based.

3.1 Constitutive Equations for Piezoelectric Materials

The elastic and dielectric properties of a piezoelectric material are coupled.

Appendix B contains a summary of the field equations and electrostatics which underlie

the constitutive relationships for these materials.

The general constitutive equations commonly used to describe the linear behavior

of piezoelectric materials - away from coercive fields, crystal form phase transitions, or

other anomalies - are derived in a straight forward manner from basic thermodynamic

principals [4, 16]. Starting with the Gibbs free energy

(3.1)
22

where U is the internal energy, differentiation yields

(3.2)

which leads to

(3.3)

The dielectric, piezoelectric, thermoelastic, elastic, and pyroelectric material

properties are defined by the second derivatives of the Gibbs free energy with respect to

the appropriate variables (parameters subscripted to the right of the bar or superscripted

below are assumed constant during the derivation). Then from (3.3):

(3.4a)

(3.4b)

(3.4c)

(3.4d)

(3.4e)
23

Other forms of the piezoelectric coupling coefficients are commonly used, and they

are defined in both their direct and converse forms:

(3.5a)

(3.5b)

(3.5c)

(3.5d)

Further, the piezoelectric coefficients are related by the material properties , and s. For

example:

(3.6)

The linear results of this approach provide the basic constitutive relationships of interest

[3],

(3.7)

(3.8)
24

(3.9)

These constitutive relations may also be derived to provide T, E, and 2 as the

dependent variables. Helmholtz free energy functions can be applied as well [64],

resulting in equivalent set of constitutive relationships. Constitutive equations near

polarization reversal are extremely complex [65] and will not be addressed here.

It is important to note that the coefficients of the material property tensors used in

these classic equations, s, d, ", ,, and p, are defined with the constraint that some

combination of D, E, T, or S is held constant over the infinitesimal ranges accommodated

by the linear nature of these equations. All of these parameters are also defined under

isothermal conditions. The implication of these constraints is significant - in order to use

these equations to predict behavior, both the operating temperature and the material

properties at that temperature must be known for the given boundary conditions. In

practical terms, this derivation implies that temperature variations resulting from heat

generation due to electrical, piezoelectric, and mechanical losses will have an insignificant

effect on the material properties.

Note that the material properties are not time dependent in the derivation shown

above. Adapting the thermodynamic approach to include time dependent material

properties is prohibitively complex and generally does not produce useful results [12].

3.2 Constitutive Equations for Viscoelastic Materials

The origin of viscoelastic behavior in polymers is not completely understood, but it

is generally thought to be molecular in nature. Viscoelastic effects have been attributed to


25

both crystalline and amorphous regions in semi-crystalline polymers such as PVDF.

Polymers are composed of long chain organic molecules - in PVDF, for example, the

typical chain length may be 100,000 monomer CH2-CF2 units. The interaction of adjacent

chains is both physical (stearic) and ionic in nature, and impediments to chain motion are

thought to contribute to lossy viscoelastic behavior. Because microscopic viscoelasticity

is so complex, many simpler mechanisms have been proposed in order to explain specific

behaviors observed in various classes of polymers. This allows the construction of

predictive equations. For example, Drosdov [71] suggests a mathematical model based

upon adaptive links, which are effectively elastic links between polymer chains that stretch,

break, and reform during viscoelastic strain. Dafalis [70] discusses changes in the

conformation of polymer chains without slipping, meaning for example that the chains may

wind or unwind from helical or spring like shapes. Ferry [12] based the WLF approach on

the progressive disentangling of long chain polymer molecules under stress. Matsuoka

[88] among others has proposed a model relating viscoelastic behavior to polymer

morphology, conformation, and intermolecular cooperativity.

As these mechanisms are proposed in order to allow the construction of predictive

mathematical models for specific behaviors, they are often expanded to classes of

polymers which as a group may exhibit these behaviors. Polymers in the class generally

have similarities in density, viscosity, molecular weight, crystal structure, amorphous

content, degree of cross linking, additives, etc. Also, different mechanisms may be used to

predict a specific behavior of the same polymer depending upon whether the material

experiences, for example, viscoelastic creep or creep influenced by thermal or chemical

ageing [89].
26

This subject has been extensively explored [30, 64, 66-83], and there are a number

of mathematical approaches to modeling the behavior of linear and nonlinear viscoelastic

materials which are considered classical. Spring and dashpot mechanical models are

probably most often used to construct ‘ideal’ viscoelastic material behavior, with the

Maxwell (spring and dashpot in series) and Voigt (spring and dashpot in parallel)

configurations most common. The Standard Linear Material model (SLM) is a more

useful construction which incorporates elements of both types. Although it is widely

acknowledged that these simplistic single and dual element models are not adequate for

most real viscoelastic materials, it is instructive to explore SLM behavior [12, 84, 85].

See Figure 3.2.1.

Harmonic excitation of the SLM leads to a first order DeBye type relaxation [86]

in the frequency domain, with the equivalent dynamic modulus given by

(3.10)

(3.11)

(3.12)

where Y ! and Y!! are the modulus storage and loss components, and Y0 and Y1 are

constants. Y* is written in complex notation and known as the complex modulus. See

Figure 3.2.2 for time and frequency domain normalized plots. Details of the derivation of

(3.10) and (3.11) can be found in Appendix D.


27

Figure 3.2.1 Classical viscoelastic material elements. Components are


springs and dashpots.

In the time domain, the one dimensional SLM model subjected to a step function in

strain exhibits an instantaneous elastic stress response combined with a first order stress

decay. It is convenient to think of this response from the perspective of a modulus that is

relaxing, or a relaxation modulus, Y(t),

(3.13)

with Jr as the relaxation time constant for the decaying component. Expanding (3.13)

yields an expression for stress,

(3.14)
28

DeBye Modulus Relaxation for SLM Element

Note DF not centered


Modulus, Arb Scale 2

0
0.01 0.1 1 10 100

Frequency (Hz)

Storage Modulus Loss Modulus Loss Tangent

Figure 3.2.2 Normalized DeBye relaxation of an SLM element.

This expression is useful for step function strain only, but it can be expanded to

include a time varying strain history by breaking this history into an infinite series of small

step functions. Each of the individual step functions is then allowed to produce a resultant

decaying stress function, and all of these are summed. The resulting form

(3.15)

is further resolved in to the classic Boltzmann superposition integral [84]

(3.16)

Here t is time and J is the integration variable. Although relationships between stress,
29

strain, and temperature can be deduced from the thermodynamics of irreversible

processes, in isothermal cases these relationships can in fact be represented over limited

frequencies by multiple Maxwell or Voigt elements in proper sequence [64]. Maxwell or

Voigt elements with carefully chosen parameter values exhibit relaxations at different

frequencies which may be summed together to provide a spectra of first order relaxations

which matches empirical data for certain materials, adapting (3.14) to

(3.17)

The use of this type of spectra to fit data is discussed by many authors [12, 84, 85,

74]. See Figure 3.2.3 for an example of this approach. Typically it is necessary to have at

least one relaxation per frequency decade [85] for well behaved data, and higher densities

may be required for ill behaved data. The assignment of relaxation frequencies is arbitrary

and done to fit the data. (3.17) can be expanded to produce an equation similar to (3.16).

In a similar manner the response of the SLM to a step function stress can be

characterized by creeping, wherein the strain exhibits an initial and immediate response,

S0, followed by an additional time dependent first order strain with an asymptote of S1.

(3.18)

Recalling that for elastic materials


30

Figure 3.2.3 Example of a multiple Maxwell element relaxation spectra


(from [74]).

(3.19)

(3.18) can be rewritten, assuming constant stress, as

(3.20)

Here s(t) is the creep compliance. Note that the creep time constant, Jc is not generally

equal to the relaxation time constant Jr for even the ideal SLM model. Again, this

expression for creep can be expanded using a spectrum of time constants in order to match

empirical data,.

(3.21)
31

and the general Boltzmann superposition integral form becomes

(3.22)

Rewriting this expression for the material stiffness, cijkl, yields

(3.23)

This equation is expanded to include time dependent stiffness and arbitrary strain histories,

leading again to a Boltzmann superposition integral [84, 87, 74]

(3.24)

The time dependent nature of the elastic stiffness is conveyed in the c(t-J) term.

Without the assumption of spacial homogeneity this term would include a positional

dependency c(x,t-J) - all discussions in this work assume homogeneity.

(3.24) reduces to

(3.25)

when the strain function is continuously differentiable in time. Again, this equation may be

easily visualized as a summation of the material’s reaction to a number of infinitesimal

incremental strains in the time domain. The form of (3.25) may be used to predict
32

performance of materials with time varying properties subjected to arbitrary strain

histories.

Rewriting (3.22) to include time dependent thermal expansion yields an equation

described by several authors as useful for viscoelastic materials [32, 74],

(3.26)

Although (3.26) describes time and temperature dependent behavior, it is

important to note that the time dependent elastic compliance in this equation is not

temperature dependent. (3.26) alone is therefore not adequate for predicting the behavior

of a viscoelastic polymer over a range of temperatures without time and temperature

dependent compliance data.

The fact that polymers are viscoelastic in nature does not preclude the use of

purely elastic constitutive equations. If a solution for the behavior of a purely elastic

material under specific loads is known, often times equivalent solutions for viscoelastic

materials in the same circumstances can be formulated using what is known as the

correspondence principle [12, 63, 84, 90]. In this approach, equations such as (3.26) are

not useful, as the elastic modulus of the material must be replaced by the complex

modulus in the governing equations for the problem. The resulting complex solution is

frequency dependent and provides both magnitude and phase angle components. Success

with this approach requires knowledge of the complex material properties in the frequency

domain and over the temperature range of interest.

This use of the correspondence principle is fundamental to the results presented in


33

this work, wherein the compliances in the classical elastic constitutive equations are

modified to include viscoelastic behavior. In addition the TTS technique, used in later

chapters to increase the useful range of material property data, yields a frequency domain

form that is compatible with the correspondence principle. Further, this work will expand

the correspondence principle to include piezoelectric behavior. The body of empirical

work and material property extrapolation contained in later chapters is intended to provide

the data necessary to allow the constitutive equations to be useful over wide frequency

and temperature ranges.

3.3 Constitutive Equations Coupling Piezoelectricity and Viscoelasticity

This author found only a small number of references discussing either general

equations involving both piezoelectricity and viscoelasticity, or constitutive equations

specific to PVDF. These are presented here briefly as background. Kamlah and Jiang

[91] provided constitutive equations for history dependent piezoelectric PZT behavior, but

the focus of their work was on ferroelectric switching hysteresis and not viscoelastic

behavior. Padilla et.al. [92] have recently published interesting work dealing with the

constitutive relations of piezoelectric materials in terms of invariants, but presented only a

cursory discussion of viscoelastic behavior. Lynch and McMeeking [93] provide a

constitutive structure for ferroelectric PVDF that accounts for large strains, but they do

not specifically address time dependent piezoelectric behavior.

Li and Dunn explore the behavior of visco-electro-elastic piezoelectric materials

using the correspondence between quasistatic visco-electro-elastic and static

piezoelectricity [75]. Their work shows a derivation of the complex elastic, dielectric, and
34

piezoelectric compliances as a function of % PZT composition in a 1D, series

PZT/polymer composite model. Coefficients used in their analysis are complex but

independent of temperature or frequency, as they are focused specifically on the

compositional effects, and therefore their results are not directly useful in this work.

Lynch [32] has framed the constitutive relations of viscoelastic PVDF in integral

time domain terms similar to those of Lake [84]. Although this form is elegant, neither

author provided any useful characterization of the time dependent material properties

required by these equations, and neither addressed the issue of temperature dependence.

These researchers have proposed the inclusion of time dependent piezoelectric terms into

(3.32), yielding the following equations for strain and electric displacement:

(3.27)

(3.28)

These forms are constructed around the Boltzmann superposition principal

assuming linear behavior and superposition. In order to use this approach, step function

reactions of charge displacement and strain need to be measured against applied stress

over a large range of times. These measurements need to be repeated a number of times

at each test case in order to reduce noise through averaging, and then repeated at different

temperatures. Equations to represent this material behavior can then be constructed

through curve fitting and data analysis. Although this approach is feasible, the resulting
35

constitutive equations are limited to the range of frequencies and temperatures measured.

The alternative to the time domain integral constitutive form shown in (3.27) and

(3.28) are the elastic equations (3.14) and (3.15), applied through the correspondence

principle in the frequency domain. In this case material properties are measured as a

function of frequency over a number of temperatures. Aaveraging and curve fitting are

then required to produce equations representing these properties as functions of frequency

and temperature. The most important reason for casting constitutive relationships in the

frequency domain for PVDF is the possibility of using the Time - Temperature

Superposition (TTS) principal to extend material property data over a range of

temperatures and frequency beyond those actually measured - and therefore create

constitutive equations valid over a larger frequency-temperature region. As discussed

below, the frequency domain TTS approach has been developed over many years and is a

well known technique for extrapolating viscoelastic material properties.

3.4 Time Temperature Superposition (TTS)

As temperature increases, polymers stressed at a constant frequency tend to

become softer and lossier. As the frequency of excitation increases at constant

temperature, polymers tend to become harder and less lossy. The influence of frequency

and temperature on the elastic properties of polymers has been studied extensively [81, 82,

for example] and is the subject of a classic text by Ferry [12]. A methodology for using

material properties measured over a limited range of temperatures and frequencies to

predict these properties over a range much wider than measured was formalized by

Williams, Landel, and Ferry (WLF). This method uses a reduced variable approach to
36

separate time and temperature, allowing material properties to be expressed in terms of a

single function of each variable.

The use of a frequency ‘shift factor’ fundamental to the WLF technique and allows

correlation of the effects of increasing temperature with decreasing frequency, and

decreasing temperature with increasing frequency. This shift factor, a(2), is a function of

temperature and has the form

(3.29)

where C1 and C2 are empirical constants, 20 is the reference temperature of interest, 2 is

the temperature at which material properties are measured, and 1 = 2 - 20. The shift

factor is used to map measured material properties from the temperature and frequency of

the measurement to a frequency at the reference temperature of interest,

or (3.30)

(3.31)

where Q is a specific value of a thermorheologically simple (TRS) material property, f is

the frequency at which data was measured at temperature 2, and f0 is the frequency at

which the same material behavior can be expected at the desired reference temperature,

20. TRS materials are, by definition, materials for which this approach is valid. The

constants C1 and C2 are unchanging for a given TRS material assuming no phase change,
37

and can be used to map both the storage and the loss terms of various material properties.

To use the WLF approach, a number of isothermal measurements are made at

different temperatures over a frequency range. To simplify the explanation, assume that

the data is measured at a number of frequencies, n, and a number of temperatures, N. Let

i= 1 to n, j = 1 to N, and let Qij represent a material property measured at frequency i and

temperature j. The raw data can be plotted as Qij versus fi, with different curves resulting

for each value of temperature, j, together on the same plot. The TTS mapping is

accomplished graphically by plotting Qij against f* for all i and j, where f* = (fi)[a(1j)].

Recall that there is an assumed reference temperature, 20, in the a(1) term, so that the

entire mapped curve shows data reduced to this temperature. This means that at the

reference temperature, the value of Q may be selected from the curve as a function of the

expanded frequency, f*. After the mapping, or expansion, there is effectively a

dimensional contraction and the resulting curve yields parameter values plotted against

frequency at an assumed reference temperature.

Figure 3.4.1 shows an example plot for PVDF elastic compliance. Graphically

stated, a reference temperature is chosen and the material property data at that

temperature is not altered on the frequency axis. A shift factor is calculated for each

remaining set of isothermal data, and data taken at temperatures above the reference

temperature is shifted to the left towards lower frequencies while data taken at lower

temperatures is shifted to the right towards higher frequencies. Ideally, the data curves

can be shifted such that they join and a single “master curve” emerges, representing the

material property of interest at the reference temperature but extrapolated over a much

larger range of frequencies as shown in Figure 3.4.2.


38

This shifting of the data to higher or lower frequencies is controlled by the

selection of the constants in the WLF equation, which determine the magnitude of the shift

factor. These constants are chosen to provide the best possible master curve across a

range of reference temperatures. In this work, ‘best’ is defined as highest possible r2

regression value for a curve fit to the data.

Elastic Data vs Temperature 0622-2


Stretch Direction (1 axis)

0.9
0.8 Increasing Temperature
Compliance, 1/GPa

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-2 -1 0 1 2
Log Frequency (Hz)

25 35 50 65 80

Figure 3.4.1 Elastic compliance data as taken.

A limitation in the use of the WLF approach is that, although the expanded

frequency range (f0 in (3.30)) may be many orders of magnitude larger than the measured

data, the desired or reference temperature to which the measured data is reduced, 20,

must generally be within the temperature range of the measured data. Choosing 20

outside of the measured temperature range results in a shifted frequency range that does

not include some or all of the measured frequency range - an unacceptable trade off
39

typically. It may seem that a form of WLF expansion using a temperature shift factor

dependent upon frequency might provide extrapolation of the measured temperature

range, but this approach is extremely complex and does not generally yield useful

correlation to experimental data [12]. Clearly the utility of the WLF equation would be

enhanced by the ability to predict material properties as a function of temperature before

frequency extrapolation.

M aster Curve Com pliance vs Freq 0622-2 @ 323K

1.0
Compliance (Gpa^-1)

0.8
0.6
0.4
0.2
0.0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 3.4.2 Elastic data expanded about 50°C.

For PVDF, it appears initially that the empirical elastic and piezoelectric

compliance data follows a general Arrhenius relationship, described by the equation

(3.32)
40

at least over the temperature range measured in this study. Here U is the activation

energy and k is the Boltzmann constant (1.38 @10-23 J/°K). The accuracy of the initial

Arrhenius expansion about 50°C can be seen in Figure 3.4.3 where U = 1.054@10-20 J, or

about 6.32 kJ/mole. The net effect of this expansion is to map material property data

from one temperature to another at a constant frequency to provide additional parameter

estimation over the same range of frequency. Note that although the form of (3.32) is

similar to that of (3.31), the Arrhenius equation (3.32) is effectively a temperature

expansion, not a frequency expansion.

Arrhenius Fit to Elastic Compliance/Temperature


0622-2 Stretch Direction (1 axis)
0.9

0.8
Compliance, Gpa^-1

Increasing Temperature
0.7

0.6

0.5

0.4

0.3
-2 -1 0 1 2
Log Frequency (Hz)
25 35 50 65 80
25A 35A 65A 80A

Figure 3.4.3 ‘A’ curves are Arrhenius fit to measured data, expanded about
50°C.

This is quite a significant finding in that it opens up the possibility that the

measured data can be extended over a wider temperature range prior to the construction
41

of the WLF master curve. This in turn produces master curves with a much wider

frequency range of predicted behavior. Stated graphically, if there is a broader range of

isothermal material property data curves plotted against frequency (larger vertical

separation), when curves above and below the reference curve are shifted to lower and

higher frequencies respectively, the magnitude of the shift is more significant. A review of

elastic compliance versus temperature data for PVDF from other sources [1, 20] which

show isochronal data over a broader temperature range than measured in this work

indicates that the Arrhenius expansion of material properties may be valid from about -20

°C to about 85 °C. If this enhanced temperature range is used the resulting master curves

will contain several orders of magnitude more than the present curves, which are based on

the 20° to 85°C data. This proposed expansion will be discussed in depth in Chapter 6.

Ferry [12] raised the possibility that the WLF technique may be useful for

modeling the dielectric properties of polymers as well as their viscoelastic properties. This

approach will be investigated for PVDF and is discussed in Chapter 7.


42

CHAPTER 4 MEASUREMENTS

4.1 Previous Work

As discussed in the introduction, the only references identified by this author which

were specifically focused on the viscoelastic properties of piezoelectric PVDF were

published by the ferroelectrics group at Montana State University [8-11]. They produced

viscoelastic and piezoelectric measurements of PVDF in the 1 and 2 axes from 1 to 50 Hz

and from 25°C to 81°C. They used both creep techniques (time domain) and frequency

techniques to measure elastic property data. Their dynamic elastic data collection

technique was not specifically detailed but they reported measuring stiffness magnitude

and phase angle. Their piezoelectric data collection was done by measuring oscilloscope

voltage traces in the frequency domain. This data required compensation due to the

interface between the frequency dependent source impedance of the PVDF sample and the

instrument impedance. They did not report specific piezo coefficient values, only peak to

peak output voltages versus strain as a function of frequency and temperature.

Extrapolation of d31 and d32 information from their data is difficult because they did not

measure dielectric properties.

4.2 Equipment and Setup

In this work, elastic and piezoelectric measurements were taken at 20 (25 in some

cases), 35, 50, 65, and 80°C. Measurements at each temperature were made at 80

different frequencies in logarithmic progression with 20 points per decade over 4 decades

from 0.01 to 100 Hz. Data collection was fully automated. See Appendix C for detailed
43

instrumentation specifications.

A unique electromechanical testing apparatus was constructed specifically for this

work. Pictures of the test base and associated test instrumentation are shown in Figures

4.2.1 - 4. A schematic of the system is shown in Figure 4.2.5. In operation, a constant

tension preload and a smaller sinusoidal load were applied to one end of a sample under

test. The displacement of the sample was measured at the point where the load was

applied and the load was measured at the opposite end of the sample, which was clamped

to the load cell. A single displacement transducer provided sufficient accuracy in this

measurement system because the load cell, the load cell mounting structure, the drive rods

applying load to the sample, and the support structure for the exciter had an effective

spring constant much higher than the equivalent spring constant of the sample under test.

Displacement of any part of the mechanical structure other than the sample under test was

considered noise, and significant effort was applied to reduce this by making the system as

stiff as possible. A spring - mass - dashpot schematic of the system is shown in Figure

4.2.6. Actual measurements of system displacement magnitude and phase are shown in

Figure 4.2.7. These were made with a metal sample having a spring constant roughly

three orders of magnitude higher than the PVDF samples tested. This data was measured

at higher frequencies in order to detect system resonances, a major source of errors in

frequency domain elastic measurements. A subtle resonance can be seen at about 600 Hz,

most likely due to the compliance of the load cell, and this limits the safe upper test

frequency to 100 Hz. Calculations from the data in Figure 4.2.7 indicate a peak system

displacement on the order of 0.05 µm at 100 Hz, showing that the apparatus was

sufficiently stiff below this frequency - typical sample displacements were on the order of
44

35 µm. The -20° phase shift shown in Figure 4.2.7 at 100 Hz is not meaningful given the

extremely small displacement of the measurement system.

It must be noted that although the stiffness of the measurement system is quite

high and the resulting high frequency resonance with a stiff sample is not significant, there

are other inertial issues which become important as the displacement increases when

measuring polymer samples. There is a mass associated with the exciter, as shown in

Figure 4.2.6, and this mass forms another resonance with the compliance of the PVDF

samples. In order to make sure that this resonance event is not close to the frequency

range being measured, the length of the samples is limited to about 20 mm. The focus on

keeping any resonance behavior away from the frequency range tested is required in order

to use static equations to deduce material properties - far simpler than using dynamic

equations of motion.

The system had a rigid cast steel base and the framework of a small thermal

chamber from an older Toyo Rheovibron originally used to test material elastic properties.

A T bar slot in the center of the base allowed placement and alignment of the major

system components: displacement transducer, exciter, thermal chamber, load cell

mounting assembly, and tensioning assembly.

The pyroelectric response from PVDF is significant (on the order of 8 V/°C open

circuit), and very small changes in temperature can produce charge equivalent to or larger

than the charge produced when stressing the film to measure piezoelectricity . It was

therefore very important that the chamber used for testing the piezoelectric properties of

PVDF not allow the film temperature to change during isothermal frequency sweeps.
45

Figure 4.2.1 Piezotron test rail showing exciter, thermal chamber, force
and displacement gages.

Figure 4.2.2 Overall view of piezotron, showing DSA and other


instrumentation.
46

Figure 4.2.3 Front view of piezotron test area.

Figure 4.2.4 Close up of test cell.


47

These temperature changes were most critical at lower frequencies, as the thermal

mass of the film sample minimized high frequency pyroelectric noise. Also, thermally

induced pyroelectric charge outside the measurement bandwidth had a reduced effect on

the piezoelectric measurements because the DSA uses very strong digital filtering

functions to exclude all frequencies but the frequency being measured. It is interesting to

note that the minor temperature changes which may seriously affect the piezoelectric

measurements have no effect whatsoever on the elastic measurements.

The temperature chamber was retrofitted with a modern temperature controller

and new insulation, signal wiring, and gasket assemblies. Special isolated interconnection

springs were fabricated and installed on the inside of the chamber and were wired to BNC

connectors on the outside of the chamber housing with fully shielded high temperature

coax cable. Thermal insulation was packed around the cable path to eliminate heat loss.

The thermal mass of the chamber was made quite high by using a large aluminum heat

exchanger block in order to increase the time constant of temperature fluctuations in the

chamber. The temperature controller had both heating and cooling functions. The heating

function applied 120VAC to a 100W resistive heater coil embedded in the large aluminum

mass. The cooling function switched on a solenoid valve allowing cold water to flow

through a copper cooling coil embedded in the same mass in the chamber. The PID

(proportional, integral, and derivative) controller parameters were set to initial conditions

and then allowed to auto-tune for the highest temperature stability prior to testing

samples. The RTD temperature sensing probe used by the controller was attached to the

large thermal mass using a thermally conductive silver filled epoxy in order to eliminate

rapid temperature fluctuations in the probe.


48

Additionally, parameters such as the controller cycling frequency had to be

carefully set to further reduce coherent low frequency thermal noise. For example, when

the system was first assembled the controller applied heat and then switched to a cooling

cycle in order to maintain an exact set point. The cooling cycle was initiated every 45

seconds or so, and even though it only lasted for several seconds this caused a severe

thermal noise problem at a frequency having a period of 45 seconds - about 0.02 Hz. A

small aluminum baffle was also constructed and placed around the sample inside the

thermal chamber to eliminate any thermally induced air currents. The outer shell of the

thermal chamber was heavily insulated with high temperature fiberglass material to prevent

changes in room temperature or the operation of the laboratory HVAC system from

affecting measurements.

The mechanical clamps used at each end of the sample under test were developed

for this work and were specifically designed for compliant thin film materials. The clamps

and their operation are shown in Figure 4.2.8. Special tooling was built to allow accurate

orientation of the sample films in the clamping blocks. Spring loaded clamps were not

used due to the increased mass required for the spring and its housing. Likewise screw

together clamps were avoided in an attempt to reduce mass. The final T slot solution used

the minimum amount clamping material and yet provided maximum stiffness. The major

drawback to this approach is the amount of time required to carefully mount and glue each

sample.

The size of the existing thermal chamber and the requirement to use relatively

short sample lengths required the use of long connecting rods on both ends of the sample.

Dampening of lateral vibrations in these clamp support rods was accomplished using
49

Figure 4.2.5 Piezotron schematic.

Figure 4.2.6 Piezotron spring - mass - dashpot schematic.


50

donut shaped silicone rubber diaphragms. The diaphragms fit snugly around the

connecting rods and the outer diameter was attached rigidly to the test chamber wall (see

Figure 4.2.9). These diaphragms offered high compliance in the direction of sample

motion (rod axis) but were relatively stiff in lateral directions (rod radius). Dampening of

lateral rod vibrations was essential to eliminate phase and magnitude artifacts induced by

lateral rod vibration resonances at higher frequencies. In fact, the upper frequency limit of

the test system was determined by these unwanted resonances.

The force generator was a 3.8 ohm voice coil exciter manufactured by Toyo

Corporation. The exciter was driven by a 180 watt Bruel and Kjaer model 2712 power

amplifier capable of operating from DC to 100 kHz. The power amplifier input signal was

supplied by the HP3562A Dynamic Signal Analyzer (DSA).

The HP3562A is a dual channel, self calibrating multifunction DSA rated from

64µHz to 100kHz. Over the measurement bandwidth of .01 Hz to 100 Hz, the DSA has

an input noise equivalent to 36 µNrms force, 6.1 Drms displacement, and 2 10-15 Coulombs

of charge. Phase resolution is ± 0.5° referenced channel to channel, without

compensation. Although this phase accuracy is sufficient for most measurements taken

with a DSA, it was borderline for the kind of resolution required to provide accurate

material loss parameters.

The force on the sample under test was measured by a semiconductor load cell

rigidly attached to a machined steel mount. The mount was fixed rigidly to an adjustable

carriage assembly which allowed manual pre-tensioning of the sample. The load cell was

calibrated at the factory and the calibration was verified after mounting through the use of

small weights.
51

Figure 4.2.7 Displacement/force spectra from piezotron with metal sample.


Small noise is caused by extremely small displacements.

Figure 4.2.8 Sample as held in clamps.


52

The displacement of the driven end of the sample (left side in Figure 4.2.1) was

measured by a commercially available non-contact eddy current sensor. The sensor output

voltage versus displacement transfer function is nonlinear, so it was calibrated in place

over the displacement range of interest using a precision dial indicator attached to the

adjustable carriage assembly and a rigid 10 mil BeCu test sample. During actual testing

the displacements undergone by the film samples were so small that the eddy current

transfer function was considered linear about the initial measurement point.

Charge was measured using a Kiethley 610C Electrometer, typically set to the 10-9

Coulombs scale. Output from the electrometer was set at 1 Volt per nanoCoulomb. The

interconnection wiring was a guarded four wire arrangement up to the BNC connectors on

the thermal chamber in order to provide maximum accuracy and noise suppression. The

Kiethley electrometer was calibrated over the frequency range of interest using a precision

capacitor and the DSA voltage drive source, and the data recorded for compensation.

A Compaq DeskPro 590 running Windows 95 was used as an HPIB system

controller to collect data from the DSA after every scan. All other system data was

manually entered into the system computer.

The dielectric data was measured using an HPIB (Hewlett Packard Interface Buss)

controlled HP4284A precision LCR bridge with a range of 20 Hz to 1 MHz. A special

sample holder was built to minimize inductive compensation and standard 4 wire dielectric

measurements were taken inside a separate thermal chamber. The chamber was capable of

maintaining temperature from -60°C to 200°C under remote computer control on an HPIB

bus. Custom software was written for the control computer to automatically step the

chamber through a desired temperature range at 10° per step. At least 30 minutes were
53

allowed at each temperature for settling, and the control software ensured that the

temperature was stable before taking data. The refrigeration units and heater fans in the

chamber were automatically turned off by the computer during the relatively quick

dielectric measurements in order to minimize noise. The LCR bridge was calibrated using

precision air capacitors, and the sample holder was auto-compensated before each

measurement run. Data was read directly from the LCR bridge over the HPIB connection

and automatically stored on a hard drive. Dielectric measurements were taken separately

from the elastic and piezoelectric measurements using samples from the same film sheet

from which all samples were taken.

Figure 4.2.9 Rod support diaphragm.


54

4.3 System Operation

The typical procedure for elastic and piezoelectric measurements was to mount a

film sample into the clamps, set a pre-tension on the film, and connect the electrode wires

to the internal spring connectors. The chamber was checked to make sure all gaskets and

insulation were in place, the thermal baffle was inserted, and the chamber was closed.

A set of operating parameters was then programmed into the DSA for each test

sequence, either manually or via the control computer. An example of the programmed

state is shown in Figure 4.3.1.

Once initialized, the DSA supplied the source drive signal to the power amplifier.

The power amplifier converted this signal into a large current drive to power the inductive

coil in the exciter. The exciter applied a force to one end of the film sample. For elastic

property measurements the displacement of this end of the sample was measured and the

output of the displacement sensor was fed back to channel 2 of the DSA. For

piezoelectric measurements, the output from the electrometer charge amplifier was fed

back to channel 2 of the DSA instead of the displacement signal. The other side of the

film was held stationary, and the reaction force exerted by the test fixture on the film

sample was measured by the load cell. In all measurements the output of the load cell was

fed back to channel 1 of the DSA.

The DSA measured two channels of data real time. Twenty data points per decade

were measured over four decades, with 20 to 30 averages per measurement point. A large

number of averages was used to ensure repeatable accuracy, especially at low frequencies.

The frequency response data was collected in a phase angle and dB signal versus

log(frequency) format for scale compression, allowing detection of small anomalies in the
55

presence of larger signals:

Figure 4.3.1 Typical signal analyzer configuration state (HP35562A).

Figure 4.3.2 shows a typical DSA frequency response plot. In addition, the power

spectra of both measured channels were collected versus frequency in order to provide

absolute magnitude data. The power spectra were recorded such that they provided the

peak values of the voltages from both input channels as a function of frequency.

The temperature controller was programmed manually for each temperature set

point then allowed to auto-tune for temperature stability prior to testing samples.
56

12 0.0
11 -0.5
10 -1.0
9 -1.5

Phase
dB

8 -2.0
7 -2.5
6 -3.0
5 -3.5
4 -4.0
0.01 0.1 1 10 100
Log Frequency

Figure 4.3.2 Typical example of uncompensated data from the DSA. dB is re


(displacement/load), phase angle is between displacement and load.

4.4 Calibration/compensation

Compensation spectra were run for the sensors over the bandwidth of interest in

order to determine sensor noise floors. This testing was done with all conditions as much

as possible identical to those in a normal test run. The HP3562A DSA was used to

capture the data, allowing the electrical noise contribution from the DSA to be included in

the compensation. The DSA was configured as in a normal test run, collecting power

spectra for two input channels.

The Kiethley electrometer was evaluated with a 330 picofarad capacitor in place of

a piezoelectric film sample in order to remove any vibrational input noise. The Entran

load cell was tested with no film sample in the clamps and with the output power from the

exciter power amplifier turned off to avoid extraneous vibrational noise. The ECS

displacement transducer was measured at the same time and under identical conditions.
57

The results of these electrical noise floor measurements are shown in Figure 4.4.1. Note

that in all cases the sensor noise floors are significantly less than 1% of the signals of

interest when measuring film samples.

In order to understand the frequency response characteristics of the electrometer,

compensation baselines were recorded using equivalent capacitors driven through

precision voltage sources in order to produce the charge nominally equivalent to that seen

in a normal run. Both magnitude and phase errors were recorded as shown in Figure

4.4.2. These errors were then used to compensate the raw charge versus frequency data

collected for each test run.

No compensation was applied to the load cell or the displacement sensor spectra

because their measured noise floors were extremely small in both phase and magnitude.

Noise Floor

18 90
16 80
Load Noise (mN)

Disp Noise (nm)

14 70
12 60
10 50
8 40
6 30
4 20
2 10
0 0
-2 -1 0 1 2
Log Frequency (Hz)

Figure 4.4.1 Piezotron instrumentation electrical noise equivalents.


58

Ele c t r o m e t e r C o m p e n sa t i o n t h r o u g h C a p a c i t o r

27.5 4
3
27.0 2

Phase Angle, Deg


1
dB

26.5 0
-1
26.0 -2
-3
25.5 -4
-2 -1 0 1 2
Log Frequency Hz

Figure 4.4.2 Electrometer compensation measured driving a voltage into a


capacitor to mimic the PVDF sample.

4.5 Sample preparation

The film samples were prepared from a single film sheet of commercial grade,

polarized PVDF, roughly 12" by 12" with a nominal thickness of 30.5 µm. Samples were

plasma etched under vacuum (argon, 400W, 15 minutes) to enhance surface adhesion and

sputtered with 150 D Ni followed by 300 D of Al on both sides using an MRC 973

sputtering system. Film strips were razor cut and pieces of film were bonded onto the

special clamping blocks while held in the bonding fixture. A rubber toughened cyano-

acrylate adhesive (Loctite 401) was chosen for bonding the film samples into the metal end

blocks because of its superior strength, ability to form thin layers, high stiffness, and

handling ease. After bonding the sample to the clamp blocks, the surfaces of the exposed

sample films were patterned with a resinoid etch mask to provide clearly delineated active

regions. The positive film poling polarity was patterned with a nominal 5 mm by 15 mm
59

rectangle, and the negative side a 6mm by 18 mm rectangle. The exposed metal surfaces

were etched in a dilute solution of FeCl, and the masking resin was removed with

isopropyl alcohol. This left an electrode on the positive side that was somewhat smaller

then that on the negative side, ensuring that there would be little variation in the total

active area (region of overlapping electrodes) due to masking pattern misalignments.

Small 36 gauge solid copper wires were attached to the positive and negative

patterned metal electrodes using silver filled epoxy (TRA-CON 2902 BiPax). Only the

tips of the wires were bonded, using the smallest epoxy pad size possible – typically ½ mm

in diameter. Although the silver epoxy is stiffer than the PVDF film sample, much less

than 1% of the surface area was covered so the effects on the elastic behavior of the test

films was ignored. Figure 4.5.1 shows a prepared sample.

The dimensions of the film sample captured between the clamping blocks were

measured optically to high precision, as were the active electrode areas. The film

dimensions were required to calculate stress, the electrode area was required to calculate

the piezoelectric coefficients. The thickness of the sample strips was measured using a 0.1

µm low force drop gauge (Mitutoyo LVDT style). Dimensional data for all samples is

contained in Table 4.1.


60

Table 4.1 Sample Dimensions

Sample ID Electrode Electrode Mechanical Mechanical Thickness


Length mm Width mm Length mm Width mm :m
1205-1 15.87 5.68 19.86 6.99 31.0
1205-2 15.48 5.70 19.69 6.78 31.3
1205-3 15.82 5.50 19.92 6.25 31.3
1205-4 15.61 5.18 20.02 6.42 31.4
1205-6 16.23 5.37 20.01 6.42 31.5
1206-1 16.22 5.00 20.03 6.22 31.3
1206-2 16.10 5.22 19.79 6.53 31.3
1206-3 16.35 5.17 19.55 5.99 31.4
1206-4 16.76 5.23 20.02 6.55 31.3
1206-5 16.60 4.84 19.94 6.40 31.4
0622-1 16.23 5.11 19.72 6.09 31.2
0622-2 16.54 4.81 19.84 5.92 31.3
0622-3 16.61 4.93 19.74 6.31 31.3
0622-4 15.93 5.14 19.67 6.28 31.4

4.6 Measurement Procedure

Because the DSA is a 2 channel device, two runs were required to measure both

displacement and charge with respect to the applied load on a single sample: one run

measuring load and displacement simultaneously, a second run measuring load and charge

output simultaneously. Test conditions were kept identical for each of the two runs for a

given sample at a given temperature. After each temperature change, more than 4 hours

elapsing before the next measurements were taken. All of the samples were
61

Figure 4.5.1 Picture of a complete PVDF film sample.

tested from lowest temperature to highest temperature.

Several samples were run without resetting the stress preload between

temperatures, and the remainder were reset to a determined value prior to starting the run

at each new temperature. A period of 2 to 4 hours was allowed for the sample to settle

after preload adjustments were made. The results do not indicate any significant

difference between the two procedures.

Typical test time for each sample, after the chamber was set and sealed, was about

8 hours for each of the elastic and piezoelectric measurements at each of the five

temperatures, or about 80 hours total per sample. This allowed about 4 hours for the

sample test chamber to achieve thermal stability, the film sample to completely relax, and

the pretension on the film to be stable before taking data. Actual measurement time was

about 4 hours per run.


62

The most common commercial piezoelectric PVDF films are annealed to 85°C in

order to provide thermal stability and repeatability to 80°C. When films are exposed to

temperatures exceeding their annealed temperature, they will undergo irreversible

mechanical, dielectric, and piezoelectric relaxation. To investigate this effect, several

85°C annealed samples were run to higher temperatures (>100°C), and then retested at

room temperature after a relaxation period. Data from these runs is given in Figure 4.6.1.

Note the distinct change in properties after returning to room temperature. In this work,

the upper temperature limit was 80°C.

Elastic Compliance 1205-6

14.0 Increasing temperature 24


13.0
dB re (disp/load)

12.0 35
11.0 46
10.0 57
9.0 68
8.0 79
7.0
Change after 99C 91
6.0
99
-2 -1 0 1 2
24
Log Frequency Hz

Figure 4.6.1 Effects of heating PVDF above its annealing temperature,


showing permanent increase in compliance.
63

CHAPTER 5 DATA

One of the major objectives of this work is to present a comprehensive range of

elastic, piezoelectric, and dielectric properties for piezoelectric PVDF. Fourteen samples

were tested using the apparatus and procedure outlined in chapter 4, although data sets

from four were rejected as they contained artifacts attributed to poor sample mounting.

Only typical data is presented in this chapter, as a complete set of data for all samples is

difficult to convey in raw form because of the number of plots (4 per sample) and because

of the density of information on each plot (5 curves per plot). A summary of the full data

set is shown in chapter 6 after the raw data was subjected to TTS extrapolation. This

process produces master curves which are less dense and simpler to interpret than the raw

data. All samples tested had properties very similar to the typical data presented in this

chapter.

All samples tested in this work had sputtered electrode metal on both surfaces.

The data presented in sections 5.1, 5.2, and 5.4 is for samples with metal. The effects of

this metal are discussed in section 5.3.

5.1 1 Axis (Stretch Direction) Data

Figure 5.1.1 shows magnitude data for the elastic compliance in units of GPa-1. All

of the data in this section is recorded in terms of the peak sensor voltages measured. The

true magnitude of the elastic storage compliance should be calculated by multiplying the

measured magnitude by the cosine of the phase angle between the applied load and the

resulting strain. The magnitudes shown in Figure 5.1.1 are actually the measured
64

magnitudes, but because the angles are very small the error will be within 0.2% of the true

value. The storage compliance in the 1 axis is monotonic, well behaved, and without

significant inflection. The increase in compliance with increasing temperature is

reasonable, as is the increase with decreasing frequency. These are two features

characteristic of viscoelastic materials.

Elastic Data vs Temperature 0622-2


Stretch Direction (1 axis)

0.9
0.8 Increasing Temperature
Compliance, 1/GPa

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.01 0.1 1 10 100
Frequency (Hz)

25 35 50 65 80

Figure 5.1.1 Typical elastic storage compliance, 1 axis.

Figure 5.1.2 shows the elastic loss tangent associated with the data in Figure 5.1.1.

The angle shown represents the phase lag between the applied load and the resulting

displacement in degrees (360 degrees per period). The increasingly negative phase angle

with increasing temperature at all frequencies seems quite natural as it indicates increasing

loss. The actual elastic loss compliance, s", is calculated by taking the product of the

compliance magnitude and the sine of the phase angle, but again the angles are so small
65

that the ratio of loss to storage elasticity is equivalent to the actual measured phase angle

in radians to within less than 0.5% error. In the range of this data the loss ratio is about

1.75 % per degree of phase angle.

Figure 5.1.2 shows relatively flat curves above approximately 1 Hz, with

increasingly negative phase angle with decreasing frequency. The region above 1 Hz

appears to follow the basic Arrhenius relationship with temperature (see (3.32)). Behavior

at frequencies below this region is not easily explained because the losses continue to

increase with decreasing frequency. The phase measurements are inherently far more

sensitive than the magnitude measurements, as the scale used - 1 degree equals 1.75% loss

- is sensitive to changes in loss compliance on the order of 0.1%. It is not apparent from

the loss phase response that there is a relaxation at lower frequencies; if there were the

phase response would show an inflection and the storage compliance would show a

relatively abrupt shift. A reasonable extrapolation from this data appears to be a straight

line on the phase angle versus log frequency plot below 1 Hz. A concern with this type of

estimate might be that the TTS expansion will produce estimates over frequencies which

are many orders of magnitude below the lowest frequency measured, so small errors can

be multiplied. However, an expansion down to 10-6 Hz would produce a loss angle of

about 5 degrees at 80°C, indicating that the loss compliance was 8.75% of the storage

compliance. This seems reasonable for a crystalline polymer [1,12], and appears to be

supported by more limited isochronal PVDF data in the temperature range of interest [20].

Figure 5.1.3 contains an example of piezoelectric storage compliance data

collected from the same sample. Note that these curves are very similar in shape to the
66

Elastic Phase Data 0622-2


Stretch Direction (1 Axis)

0.0
-0.5
-1.0
Phase Angle

-1.5
-2.0
-2.5
-3.0
Increasing Temperature
-3.5
-4.0
0.01 0.1 1 10 100
Frequency (Hz)

25 35 50 65 80

Figure 5.1.2 Typical elastic loss compliance phase angle, 1 axis.

Pie zo Data 0622-2


Stretch Direction (1 Axis)

30

25
Increasing Temperature
20
d31, pC/N

15

10

0
0.01 0.1 1 10 100
Freq

25 35 50 65 80

Figure 5.1.3 Typical piezo storage compliance, 1 axis.


67

elastic compliance data shown in Figure 5.1.1. In fact, the normalized curves overlay

remarkably well at each temperature, indicating that the e31 coefficient is overwhelmingly

real (recall that e = d/s). This will be shown more clearly in the next chapter.

Figure 5.1.4 shows the associated piezoelectric loss compliance term. The general

shape of the loss curves is similar to the elastic loss in Figure 5.1.2. In this plot the

separation between the temperature curves becomes slightly more pronounced at lower

frequencies than is evident in the elastic loss. This separation will become even more

obvious in the loss curves for the 2 axis data shown in the following section.

There are two regions containing noise shown in Figure 5.1.4: at about 60 Hz

there is some minor feedthrough due to electrical mains power; below 1 Hz there is some

pyroelectric contribution from the PVDF due to the period of the temperature controller.

The 60 Hz noise has been significantly reduced by shielding. The pyro noise was more

difficult to deal with, as discussed in Chapter 4. It was reduced as much as possible by

adding baffles in the thermal chamber to reduce small temperature fluctuations caused by

air currents and by adjusting the cooling water flow rates. The pyroelectric response of

the film is about 30µCoul/(m 2 °K), or roughly 2.3 nCoul/°C for this size sample, and with

the charge values measured during the experiment on the order of several hundred pCoul,

temperature changes on the order of a 0.01 °C are detectable as noise even through a

rigorous spectral filter.


68

5.2 2 Axis (Transverse Direction) Data

Complex elastic and piezoelectric data from a sample cut in the transverse 2 axis is

presented in this section. Figure 5.2.1 shows the elastic storage compliance, Figure 5.2.2

the elastic loss compliance. Figures 5.2.3 and 5.2.4 show transverse piezoelectric data.

The 2 axis elastic data is identical in form to the 1 axis data. The storage

compliance appears to be a little higher, indicating that the 2 axis modulus is lower. The 2

axis phase angle is somewhat lower, suggesting that although the material is softer it is

less lossy. The numeric differences between the elastic data for the two axes are small.

Since the 1 axis clearly shows that the piezo compliance and the elastic compliance

have nearly identical behavior over temperature, it is reasonable to expect that the 2 axis

will behave similarly. However, it is quite obvious that the transverse piezoelectric

response is not similar in form to the 1 axis response. Figure 5.2.3 shows unusual

piezoelectric compliance behavior at lower frequencies. The source of this behavior is not

easily understood. The transition in the 2 axis storage compliance and the peak in the

associated loss angle suggest that there is a relaxation occurring, but Figure 5.2.4 shows

that the relaxation peak is not a significant function of temperature.


69

Elastic Data 1206-4


2 Axis (Trans Direction)

1.0

Increasing Temperature
Compliance, 1/GPa
0.8

0.6

0.4

0.2

0.0
0.01 0.1 1 10 100
Frequency (Hz)
25 35 50 65 80

Figure 5.2.1 Typical elastic storage compliance, 2 axis.

Elastic Phase Angle 1206-4


2 Axis, (Trans Direction)

0.0
-0.5
-1.0
Phase Angle

-1.5
-2.0
-2.5
Increasing Temperature
-3.0
-3.5
-4.0
0.01 0.1 1 10 100
Fre q u e n c y ( Hz )

20 35 50 65 80

Figure 5.2.2 Typical elastic loss compliance phase angle, 2 axis.


70

This temperature independence is not consistent with the viscoelastic behavior

shown clearly in the 2 axis elastic material properties in Figures 5.2.1 and 5.2.2. A more

convincing argument against viscoelastic relaxation is that the polarity of the phase peak in

Figure 5.2.4 is not consistent with viscoelastic loss. The strain in a viscoelastic material

lags behind the stress (a fundamental principal of causality), and the charge produced by

this strain should also lag behind the stress in time. Since the sample measurements are

made relative to the stress, the strain and charge should be expected to have a negative

phase angle, as is seen in the 1 axis data.

In fact, the positive phase angle suggests an electrically resistive material - PVDF

with abnormally high conduction. The charge conducted through a resistive material is the

integration of current flow through the material over time. If we assume that the

piezoelectric charge production is in phase with and proportional to the applied strain on

the piezo material, then

(5.1)

where DS is the strain induced displacement (charge/area). Now assuming a resistive

material,

(5.2)

with DR the displacement from resistive conduction and D the bulk resistivity. This yields
71

(5.3)

Pie zo data 1206-4


2 Axis (Trans direction)

2.5
Increasing Temperature
2
d32, pC/N

1.5

0.5

0
0.01 0.1 1 10 100
Frequency (Hz)

25 35 50 65 80

Figure 5.2.3 Typical piezo storage compliance, 2 axis.

Clearly the resistive component produces a positive phase shift in this type of

measurement. Increased conduction would seem to be supported by the dielectric loss

term in Figure 5.4.3, where the lossiness increases with decreasing frequency and

increasing temperature. Unfortunately, the hypothesis of purely resistive loss is not

supported by results from the 1 axis piezoelectric data. If all samples become appreciably

conductive at higher temperatures then similar loss curves would be present in the 1 axis

data. Pure resistance is a linear parameter and is independent of the magnitude of the

signal produced by the PVDF sample, so the fact that the 1 axis samples produced much

larger signals should not have suppressed linear resistive behavior.


72

Another interesting point is that the electrometer used to measure the charge

produced by the film actually measures the voltage across a large parallel capacitor. The

actual voltage on the PVDF film sample was not exactly zero, but something less than one

volt. The impedance of the voltage measurement is greater than 1013 ohms and there is no

appreciable charge loss through the instrumentation.

The attenuation of measured charge output and the positive phase shift at lower

frequencies cannot be attributed to purely resistive losses, and yet they appear to be

resistive in nature. The actual cause is not known, although this behavior may be the

result of a kind of non linear conduction where the amount of charge conducted is fairly

constant, independent of applied field. In the 1 axis case, this charge quantity is relatively

small compared to the charge measured from the film and is therefore not noticeable in the

Piezo Phase Angle 1206-4


2 Axis, (Trans Direction)
12.0
10.0
8.0
Phase Angle

6.0
Increasing Temperature
4.0
2.0
0.0
-2.0
-4.0
0.01 0.1 1 10 100
Frequency (Hz)

25 35 50 65 80

Figure 5.2.4 Typical piezo loss compliance phase angle, 2 axis.


73

range of data observed. In the case of the 2 axis measurements, with much reduced

output, the amount of this charge is significant.

In order to test this hypothesis, the same sample was driven electrically using the

circuit in Figure 5.2.5. A low loss polyethylene capacitor having a capacitance nearly

equal to the PVDF sample was also tested using the same circuit. The results in Figures

5.2.6 and 5.2.7 clearly show that the phenomenon in Figure 5.2.4 is real and dielectric in

nature, suggesting that the cause is likely not related to unusual elastic behavior or even

possibly to the d32 mechanism itself. Attempts were made to model the PVDF sample as a

series of parallel resistors and capacitors, and capacitors with series resistance added, as

shown in Figure 5.2.8. None of the electrical models were capable of producing the

complex shape of Figure 5.2.4. If a non linear conduction mechanism is assumed and the

amount of charge conducted is constant, Figure 5.2.4 is less confusing. Again, it is

possible that the 1 axis parameters are also affected by this phenomenon, but the value of

the charge is too small to be significant.

This approach can be tested somewhat by close scrutiny of the 1 axis elastic data

from sample 1205-3, taken over a wide range of applied stress. The data may show slight

evidence of this behavior, but at the low stress levels necessary to get small displaced

charge in the 1 axis, it does not appear to be nearly as evident as in the 2 axis. Thermal

noise makes it difficult to interpret this piezo phase data precisely. Sample 1205-3 is

discussed in more detail in section 6.4. The 2 axis data is presented as measured, and a

more detailed investigation of this interesting behavior will be the subject of future work.

The author is aware of a single reference in the literature showing this unusual

behavior. Furakawa and Fukada [20] measured properties of piezoelectric PVDF in 1981.
74

In their data they show d32! and d32" measurements taken at 10 Hz across a wide

temperature range. It is interesting to note that this data clearly shows the d32" term to be

negative at temperatures greater than -40 °C. This has the effect of producing a positive

phase angle when translated to the kind of measurement used in this work. Although the

data is shown in their work, the authors did not discuss this peculiarity. Parts of this data

were included in other reference works [21, 37] but the anomalous d32" behavior was

again not discussed. Personal communication with these researchers has not provided a

satisfactory explanation.

Figure 5.2.5 Schematic of electrically driven test circuit. Current is


measured by voltage drop across 1 Gohm 610C resistor.
75

Driven Current , Film vs Capacitor


Sample 1206-1a, 80C, 328 pF cap
0
-5
-10
dB, re 1 nA

-15
Film
-20
cap
-25
-30
-35
-2 -1 0 1 2

Log Frequency (Hz)

Figure 5.2.6 Dielectric behavior of PVDF film at 80 °C compared to a low


loss capacitor.

Current Phase Angle, Film vs Capacitor


Sample 1206-1a, 80C, 328 pF
90
80
Phase Angle, degrees

70
60
50
Film
40
cap
30
20
10
0
-2 -1 0 1 2

Log Frequency (Hz)

Figure 5.2.7 Dielectric behavior of PVDF film at 80 °C compared to a low


loss capacitor.
76

Figure 5.2.8 Electrical models used in trying to fit anomalous d32 data.

5.3 Effect of Electrode Metal on Material Property Measurements

All PVDF samples studied in this work were sputtered with a tie coat of 150 D of

Ni followed by 300 D of Al on both sides. In order to establish the effect of these

electrodes on the material property measurements, the elastic properties of one sample

were tested with and without electrode metal. The exact same sample was used in both

tests, without removing the PVDF from the clamps. The metal was removed by careful

FeCl etching and a deionized water rinse. As can be seen in Figures 5.3.1 and 5.3.2, the

effects of the metal electrodes are not insignificant. The difficulties associated with

measuring displaced electric charge from a PVDF sample without electrodes are

formidable and beyond the scope if this work. All of the samples in this work were

prepared using the metalization process described above, and the piezoelectric, elastic, and

dielectric data presented herein include the effects of these electrodes. This approach
77

seems reasonable as the electrodes do not to change the general shape of the curves.

Piezoelectric PVDF is rarely used without electrodes. The data in Figure 5.3.1

suggests that it is important that researchers and manufacturers clearly identify the

electrode system used when reporting PVDF data.

Compliance difference, no metal - metal


35C, Sample 1206-6

1 0.14

0.9 0.12
Compliance, 1/GPa

Delta compliance,
0.8 0.10

1/GPa
0.7 0.08

0.6 0.06

0.5 0.04

0.4 0.02
0.01 0.1 1 10 100
Frequency (Hz)

No Metal Metal Delta avg Delta

Figure 5.3.1 Effects of metal electrodes on elastic compliance, 2 axis.

5.4 Dielectric Data

Dielectric data was collected from -60°C to 125°C using an HP 4284A precision

LCR bridge and a thermal chamber, as described in chapter 4. Data taken for clamped and

unclamped samples was nearly identical at temperatures below 85°C. Above 85°C the

unclamped samples exhibited an inflection and decreasing permittivity, while the clamped

samples showed increasing permittivity. Interestingly enough, when the unclamped

sample data is compensated for the increasing thickness of the PVDF as it


78

Phase Angle difference, no metal - metal


35C, Sample 1206-6

0.00

Phase Ange, degrees


-0.50

-1.00

-1.50

-2.00
0.01 0.1 1 10 100
Frequency (Hz)

No Metal Metal Delta Approx

Figure 5.3.2 Effect of electrodes on elastic compliance phase angle, 2 axis.

relaxes above its anneal temperature, this inflection and downward trend disappear. Data

reported in this chapter is for unclamped samples and exceeds the 85°C annealing

temperature by 5°C.

Figure 5.4.1 shows a surface plot of the dielectric permittivity of an unclamped

PVDF sample versus temperature and frequency. The relaxations are more clearly visible

in the projections of this data shown in Figures 5.4.2 and 5.4.3 for clarity. As frequency

increases, the relaxation event moves to higher temperatures. This is expected and normal

for polymer materials, and it leads into the question of whether or not the permittivity can

be modeled as a thermorheologically simple material. This topic will be discussed in

chapter 6.

Figure 5.4.4 shows a surface plot of the dissipation factor (or loss tangent) for

PVDF versus temperature and frequency. The shifting and broadening of the loss peaks
79

can be clearly seen in this figure. Recall that PVDF is a semi crystalline polymer, roughly

50% crystalline, 50% amorphous. Each of these conformations makes a contribution to

the bulk dielectric behavior of PVDF film. The mechanism responsible for the low

temperature dielectric loss maximum in this frequency range is associated with the glass

transition of PVDF. As the polymer is cooled through the glass transition temperature the

amorphous regions undergo chain backbone rearrangements which hinder rotations

around main chain bonds, causing a lossiness in dipole mobility and a subsequent

reduction in permittivity and increase in loss [12, 27, 94]. Conduction (ionic transport) in

polymers is understood to increase with increasing temperature [94]. At lower

frequencies this has the effect of increasing the dielectric lossiness. The high temperature

low frequency corner of Figure 5.4.4 shows a trend towards such conduction. The

dielectric bridge equipment used in this testing is not capable of measuring down to the

very low frequencies to which the elastic and piezoelectric properties were tested.

In chapter 7 the dielectric data will be curve fit and an appropriate equation

derived to allow prediction over wide temperature and frequency ranges.


80

Sample 8504

1.40E-10

1.20E-10

1.00E-10

8.00E-11
Permittivity
6.00E-11

4.00E-11

20 2.00E-11

251
0.00E+00

3981

60
Frequency Hz

20
63096

-20
-60
1000000
T e m p e r a ture
C

Figure 5.4.1 Permittivity of PVDF vs temperature and frequency.

PVDF Permittivity S8504

1.6E-10
1.4E-10
1.2E-10
Permittivity, F/m

100
1.0E-10 1000
8.0E-11 10000
6.0E-11 100000
4.0E-11 1000000

2.0E-11
0.0E+00
-60 -40 -20 0 20 40 60 80
Temperature C

Figure 5.4.2 Temperature projection of Fig 5.4.1 data.


81

Permittivity S8504

1.6E-10

1.4E-10

1.2E-10
-60
Permittivity, F/m

1.0E-10 -30
0
8.0E-11
30
6.0E-11 60
90
4.0E-11

2.0E-11

0.0E+00
10 100 1000 10000 100000 1000000
Frequency

Figure 5.4.3 Frequency projection of Fig 5.4.1 data.

S8504

0.30

0.25

0.20

Loss Tangent 0.15

0.10 1000000
25119
0.05
631 Freq Hz
0.00
90
60

20
30
0
-30
-60

Te m p C

Figure 5.4.4 PVDF dielectric loss tangent vs temperature and frequency.


Note that the frequency axis is a linear scale.
82

CHAPTER 6 ELASTIC AND PIEZOELECTRIC DATA ANALYSIS

In this chapter, the Time-Temperature Superposition principle discussed in chapter

3 is applied to the elastic and piezoelectric data from the samples discussed in chapter 5,

and master curves are constructed for all storage and loss compliances. Summary plots

are made for each parameter and a curve fit is made to each combined data set. Selection

of the appropriate WLF constant values is done empirically by minimizing the r2 value of a

regression fit to the expanded data set of each sample. The constants were shared across

the data files of all samples in order to get an adequate fit. Note that all TTS master curve

data presented in this chapter is normalized to a reference temperature of 50°C (323°K)

unless otherwise specified in order to reduce the density of the information on each plot.

This topic is discussed in length in the final section of this chapter, where equations are

proposed to estimate elastic and piezoelectric material properties over a wide frequency

range for an arbitrary reference temperature.

6.1 TTS Master Curves for all Elastic and Piezo Coefficient Data Sets

In the first part of this section, data shown in the previous chapter is expanded

through the use of the TTS method as described above. TTS curves from the remaining

population of tested samples are presented in Appendix E.

The WLF constants C1 and C2 (see (3.29)) were chosen empirically after creating

the master curves of all four parameters for all samples tested. The fit to most of the

curves using these values was excellent with the exception of piezoelectric loss

compliance, as will be shown in a later section of this chapter. The values used in this
83

work were

(6.1)

Figure 6.1.1 shows the PVDF elastic compliance data of Figure 5.1.1 mapped onto

a single reference temperature. Note that the frequency range of the extrapolated master

curve spans about 12 orders of magnitude as opposed to the four orders of magnitude for

the original data.

After careful inspection the fit to this data was done using an inverse linear curve.

The r2 value of the data relative to the curve fit approaches .998 for the selected WLF

coefficients, indicating a very reasonable fit. The data fitting equation for storage

compliance is written in terms of the log of the frequency:

(6.2)

This equation will accurately predict the storage compliance at the reference temperature,

2o. Recall that 2o is the temperature used to generate the shift factor in (3.29).

Figure 6.1.2 shows the elastic loss compliance for the same sample expanded with

the same shift factor, along with a curve fit. Changing the WLF coefficients for the loss

term would provide a slightly better fit to the data, but a number of researchers have

pointed out the need to keep the constants identical for storage and loss terms [12, 85].

The reduced curve fit accuracy seen in Figure 6.1.2 when compared to Figure 6.1.1 has a

number of sources, but in this author’s opinion the primary cause is the difficulty the DSA
84

has in resolving phase at extremely low frequencies - made more difficult in this case by

the relatively small magnitude of the loss term.

M a ste r Curve C o m p l i a n c e v s Freq 0622-2

1.0

0.8
s' (Gpa^-1)

0.6

0.4

0.2

0.0
-10 -5 0 5 10
L o g F r e q u e n c y (Hz)

Figure 6.1.1 TTS expansion of data from Figure 5.1.1. 1 axis elastic
storage compliance.

Figure 6.1.3 shows the TTS expansion of the storage piezo compliance data of

Figure 5.1.3. The high r2 value in the inverse linear curve fit to this data shows excellent

correlation to the elastic data of Figure 5.1.3. This can be seen by normalizing both

curves to their 1 Hz value and putting them on the same plot, as shown in Figure 6.1.4.

The strength of this correlation between the elastic and piezoelectric compliances is

surprising, clearly indicating that the mechanisms responsible for viscoelastic behavior are

also responsible for ‘visco-piezoelectric’ behavior.

The data spread in the TTS expansion shown in Figure 6.1.5 makes a good curve

fit difficult. The tails from each subset of data tend to increase as frequency decreases.
85

Master Curve Elastic Loss vs Freq 0622-2


7%

% Loss Compliance
6%
5%
4%
3%
2%
1%
0%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.1.2 TTS expansion of data from Figure 5.1.2 1 axis elastic loss
compliance.

This is caused by the increasingly negative phase angle with decreasing frequency

prominent in the raw data. Conduction at elevated temperature and low frequency most

likely account for this behavior, as discussed in later sections. Clearly the piezo loss

response is the most difficult parameter to characterize.

6.2 Summary TTS Master Curves

Figures 6.2.1 (a-d) show typical expanded TTS curves for 1 axis samples tested in

this work. Figures 6.2.2 (a-d) show typical expanded data for 2 axis samples. Figures

6.2.3 to 6.2.10 show all curves for each parameter combined and a best fit of the

expansion data curves. See Appendix E for additional data plots in both the 1 and 2 axes.
86

M a ster Curve d31 vs Freq 0622-2

30

25

20
d31 pC/N

15

10

0
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.1.3 TTS expansion of data from Figure 5.1.3. 1 axis piezo
storage compliance.

Elastic and Piezo compliances normalized


at RT and 1 Hz
1.8
1.6
Compliance

1.4
Normalized

1.2 Elastic
1 Piezo
0.8
0.6
0.4
0.2
0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.1.4 Comparison of elastic and piezo storage compliances


normalized at room temperature and 1 Hz.
87

Master Curve Piezo Loss vs Freq 0622-2

7%
% Loss Compliance 6%
5%
4%
3%
2%
1%
0%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.1.5 TTS expansion of data from Figure 5.1.4. 1 axis piezo loss
compliance.

Master Curve Compliance vs Freq 0622-1


1
0.9
0.8
0.7
s' (Gpa^-1)

0.6
0.5
0.4
0.3
0.2
0.1
0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.2.1a TTS expansion of 1 axis elastic compliance, 50° C.


88

Master Curve Elastic Loss vs Freq 0622-1


7%

6%
% Elastic Loss
5%

4%

3%

2%

1%

0%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.1b TTS expansion of 1 axis elastic loss , 50° C.

Master Curve d31 vs Freq 0622-1


30

25
d31 pC/N

20
15
10
5

0
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.1c TTS expansion of 1 axis piezo compliance, 50° C.


89

Master Curve Piezo Loss vs Freq 0622-1

7%
% Piezo Loss 6%
5%
4%
3%
2%
1%
0%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.1d TTS expansion of 1 axis piezo loss, 50° C.

Master Curve Compliance vs Freq 1206-1a


1.0

0.8
s' (Gpa^-1)

0.6

0.4

0.2

0.0
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.2a TTS expansion of 2 axis elastic compliance, 50° C.


90

Maste r Curve Elastic Loss vs Freq 1206-1a

7%

6%
% Elastic Loss
5%

4%

3%

2%

1%

0%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.2b TTS expansion of 2 axis elastic loss, 50° C.

M a ste r Curve d32 vs Freq 1206-1a


2.0
1.8
1.6
1.4
d32 pC/N

1.2
1.0
0.8
0.6
0.4
0.2
0.0
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.2c TTS expansion of 2 axis piezo compliance, 50° C.


91

Maste r Curve Piezo Loss vs Freq 1206-1a


5%

0%

-5%
% Piezo Loss

-10%

-15%

-20%

-25%

-30%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.2d TTS expansion of 2 axis piezo loss, 50° C.

Elastic Compliance 1 Axis Average

1.0

0.8
s' (GPa^-1)

0.6

0.4

0.2

0.0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.2.3 All TTS expansion curves for 1 axis elastic compliance, 50° C.
92

Elastic Compliance Loss 1 Axis Average


7%
6%
% Elastic Loss
5%
4%
3%
2%
1%
0%
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.2.4 All TTS expansion curves for 1 axis elastic loss, 50° C.

Piezo Compliance 1 Axis Average


30

25

20
d31 pC/N

15

10

0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.2.5 All TTS expansion curves for 1 axis piezo compliance, 50° C.
93

Piezo Compliance Loss 1 Axis Average

7%
6%
% Piezo Loss
5%
4%
3%
2%
1%
0%
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.2.6 All TTS expansion curves for 1 axis piezo loss, 50° C.

Elastic Compliance 2 Axis Average

1.2

1.0
s' (GPa^-1)

0.8

0.6

0.4

0.2

0.0
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.7 All TTS expansion curves for 2 axis elastic compliance, 50° C.
94

Elastic Loss 2 Axis Average

7%
6%
% Elastic Loss 5%
4%
3%
2%
1%
0%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.8 All TTS expansion curves for 2 axis elastic loss, 50° C.

Piezo Compliance 2 Axis Average

2.5

2
d32 pC/N

1.5

0.5

0
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.9 All TTS expansion curves for 2 axis piezo compliance, 50° C.
95

Piezo Loss 2 Axis Average

5%
0%

% Piezo Loss
-5%
-10%
-15%
-20%
-25%
-30%
-10 -5 0 5 10

Log Frequency (Hz)

Figure 6.2.10 All TTS expansion curves for 2 axis piezo loss, 50° C.

6.3 Discussion of Summary Plots

All of the individual curves in the 1 axis elastic and piezo compliances, Figures

6.2.3 and 6.2.5, have a consistent shape. The curves on each plot appear to have minor

offsets in the compliance axis which cause a lower than expected r2 value to the curve fit

for the entire data set. Since all of the samples were taken from the same film section, it is

unlikely that the variations of this magnitude are caused by actual differences in material

properties. The sources of error can most likely be attributed to the sample mounting

process, which can induce small irregularities in the film boundary conditions through lack

of parallelism between end clamp plated, uneven tension in the prestress on the sample

prior to bonding, etc. It is not unreasonable that these kinds of mounting variations could

produce small vertical curve shifts as seen in the plots, and therefore improving the

clamping process will be the subject of future work.


96

The 1 axis elastic and piezoelectric losses are shown in Figures 6.2.4 and 6.2.6.

The sensitivity of this phase measurement coupled with the relatively low loss magnitude

makes this data inherently noisy. However, the elastic loss curve fit is reasonable given

the small magnitude of the loss term overall. The piezoelectric loss data has a wider

spread, limiting the value of this information in the constitutive equations and in predicting

performance. Again, however, the magnitude of the piezo losses is so small that the fitted

curve slope seems reasonable. It may also be simpler to use a constant value for the piezo

loss in situations where this term is not important to the designer.

The curves in the 2 axis elastic compliance plot shown in Figure 6.2.7 have a

shape very near those found in the 1 axis, with the slope slightly sharper and the overall

magnitude of the transverse compliance slightly higher. The 2 axis elastic loss data is

noisy but it is similar in shape and magnitude to the 1 axis loss data.

The 2 axis piezo data represents a challenge as it does not easily suggest

thermorheologically simple behavior. The curves in Figure 6.2.9 clearly show evidence of

the odd conduction mechanism discussed in detail in Chapter 5. Likewise the piezo loss

data in the 2 axis shown in Figure 6.2.10 is unreasonable for the same reasons.

Fortunately the d32 coefficient is of little use to practical sensor designers. For this reason,

in the remainder of this work the components of the d32 coefficient will be given a constant

value of 1.5 pC/n and 2.5% loss, regardless of the frequency or temperature.

6.4 Effects of Stress Level on 1 Axis Parameters

Sample 1205-3 was tested using the normal test procedure with increasing applied

load levels. The sample data sets were labeled 1205-3, 1205-3a, 1205-3b, and 1205-3c
97

with letters for subsequent load levels. Table 6.1 shows the data extracted at 10 Hz for

20°C only in order to scale the effects.

Table 6.1 Effects of Applied Load

Data Set Force N Elongation Strain s11 d31


3 0.1040 3.87E-06 0.019% 3.63E-10 11.06
3a 0.3664 1.36E-05 0.068% 3.61E-10 11.37
3b 1.2316 4.57E-05 0.229% 3.62E-10 11.57
3c 4.6986 1.91E-04 0.956% 3.97E-10 12.03

Figures 6.4.1 and 6.4.2 show the elastic and piezoelectric storage compliances from this

test after the usual TTS expansion. It is clear from this data that a factor of 50 change in

the level of applied load does not significantly affect the compliance values.

1 Axis Ela stic Storage Compliance


Strains from .019% to .956%
1.0

0.8

3a
s' (GPa^-1)

0.6
3b
3c
0.4
3

0.2

0.0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.4.1 Comparison of 1 axis elastic compliance at different strain


levels.
98

1 Axis Piezo Storage Compliance


Strains from .019% to .956%

25

20

3a
d31 pC/N

15
3b
10 3c
3
5

0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.4.2 Comparison of 1 axis piezo compliance at different strain


levels.

6.5 Comparison and Normalization of Elastic and Piezoelectric Data

It is clear from Figures 6.2.6, 8 that the elastic and piezo compliance summary data

shows a series of similar curves offset on the compliance axis. Although this offset does

imply that there is variability in the measured data, it detracts from any relationship that

may be observed. Therefore a normalized data set is proposed in order to clarify the

behavior of individual samples over frequency. A reference frequency of 1 Hz was chosen

in the center of the measured data, and all other data was normalized to the 1Hz value for

each sample. The results are shown in Figures 6.5.1, 2. Use of the normalized form

allows each sample to scale from compliance measured at 1 Hz. Note also that the

similarity between the elastic compliance and piezoelectric compliance is so strong that the

same curve fit parameters yield very good results on both data sets. This clearly
99

demo nstrat
Elastic Compliance 1 Axis Average
Normalized @ 1Hz, 50C
es that the
1.6
piezoe 1.4 lectric
1.2
effect can be
s' re 1Hz
1.0
expan 0.8 ded
0.6
using 0.4 elastic
0.2
TTS techni
0.0
ques. -10 -5 0 5 10

Log Frequency (Hz)

Figure 6.5.1 All 1 axis elastic compliance curves normalized at 1Hz, 50° C.

Piezo Compliance 1 Axis Average


Normalized @ 1Hz, 50C
1.6
1.4
1.2
d31 re 1Hz

1.0
0.8
0.6
0.4
0.2
0.0
-10 -5 0 5 10
Log Frequency (Hz)

Figure 6.5.2 All 1 axis piezo compliance curves normalized at 1 Hz, 50° C.
100

6.6 Novel TTS Expansions

The best curve fit to the elastic and piezoelectric storage compliance TTS master

curves, averaged over the samples, yields the following inverse linear equations at a

reference temperature of 50°C (323°K):

Gpa -1 (6.3)

pC/N (6.4)

where the s!11(1 Hz, 323K) and d!31(1 Hz, 323K) refer to actual material property values

measured at 1 Hz at a temperature of 323K (50°C). The regression r2 values for both of

these curves exceed 0.99. The curve equations are extremely close and will be combined

hereafter using constant values of 0.055 and 1.0 which were found to yield best results.

The WLF TTS method used above provides a reasonable basis for expanding the

frequency over which material parameter predictions can be made. The typical output from

a normal TTS expansion are master curves showing the parameter values expanded over

frequency for a reference temperature. One of the drawbacks of the equations above is that

although the curve fit equations are a function of frequency, they must be recalculated for

the specific estimation temperature. The WLF equation for the temperature dependent

frequency shift factor, introduced in chapter 3, is

(3.29)
101

Expanded Master Curves


Elastic Compliance
0.9
0.8
0.7
Increasing Temperature
293
s' (GPa^-1)
0.6
308
0.5
323
0.4
0.3 338
0.2 353
0.1
0.0
-10 -5 0 5 10 15

Log Frequency Hz

Figure 6.6.1 Elastic compliance 1 axis master TTS curves for all reference
temperatures.

which easily allows the data sets to be recast for each reference temperature, 20. The

frequency term is not explicit by the very nature of the frequency shift factor a(2), meaning

that a material parameter prediction cannot be calculated for a specific temperature and

frequency. Typically a graph is created with the TTS method showing the material

parameter against the expanded frequency, and the desired result must be chosen from the

operating frequency.

In order to allow explicit prediction of material properties for a specific temperature

and frequency, the following technique is proposed. The reference temperature in the WLF

equation was altered in turn to match each temperature at which material property data was

collected. The data was then expanded in the normal manner, and at each temperature a

new inverse linear equation was calculated for the resulting master curve. This series of

master curves is plotted in Figures 6.6.1 & 3 and then they


102

Normalized s' Master Curves


2.0
1.8
1.6 293
308
Nomalized s'
1.4
1.2 323
1.0 338
0.8
353
0.6
0.4 Fit
0.2
0.0
-10 -5 0 5 10 15

Log Frequency (Hz)

Figure 6.6.2 Data in Fig 6.6.1 normalized at 1 Hz and a master curve fit.

were all normalized to their 1Hz value as shown in Figures 6.6.2 & 4. Figures 6.6.2 & 4

also contain the identical curve fit. This curve fit on the normalized data has the form

. (6.5)

Consider now that many temperature dependent physio-chemical processes are

reasonably well represented by the Arrhenius equation [12, 63, 94]. This appears to be the

case for the elastic and piezoelectric compliance of PVDF at constant frequency, as is

shown in Figures 6.5.3 and 6.5.4 below. The general Arrhenius equation has the form

(6.6)
103

Expanded Master Curves


Piezo Compliance
30

25
Increasing Temperature 293
20
d31 pC/N

308
15 323

10 338
353
5

0
-10 -5 0 5 10 15

Log Frequency (Hz)

Figure 6.6.3 Piezo compliance 1 axis master TTS curves for all reference
temperatures.

Normalized Piezo d31' Master Curves


2.0
1.8
Nomalized Compliance

1.6 293
1.4 308
1.2 323
1.0 338
0.8
353
0.6
Fit
0.4
0.2
0.0
-10 -5 0 5 10 15

Log Frequency (Hz)

Figure 6.6.4 Data in Fig 6.6.3 normalized at 1 Hz and a master curve fit.
104

where again 1 = ( 2 - 20), Q is some material property, and Q0 is a known value of Q at

the initial temperature 20. k is the Boltzmann constant and U is the effective activation

energy for PVDF. There is a minor modification to this form proposed by other

researchers [12] which allow an additional 1 term, as shown in (6.7) by the constant $.

Further, the constant " has been added by this author to account for a small shift with

frequency, yielding

(6.7)

The Arrhenius expansion using this equation correlates well to measured data in

Figures 6.6.5 and 6. This equation is useful in both the elastic and piezoelectric compliance

of PVDF. (6.7) indicates that the expansion of a storage compliance, Q, with temperature

change occurs at constant frequency and that data measured at any frequency at an initial

temperature 20 can be mapped to that same frequency at another temperature.

Because the form of the normalized master curves is nearly identical regardless of

the reference temperature, as seen in Figures 6.6.1 and 3 by the very close match to the

single fit equation, it is possible to suggest the following condensed equation for elastic and

piezoelectric compliance:

(6.8)

Loss compliance approximations are made in the same manner, with


105

(6.9)

The frequency and temperature at which the initial material property value is

measured are f0 and 20, and f represents the frequency at which the material property value

is extrapolated. Figures 6.6.7-8 and 6.6.9-10 show the normalized loss master curves with

the curve fit approximation given in the equation above. Note that the Arrhenius

expansions are about 50° C (353° K), and best results are obtained when 20 is near this

temperature.

Arrhenius Fit to Elastic Compliance/Temperature 0622-


2 Stretch Direction (1 axis)
0.9

0.8
Increasing Temperature
s' ( Gpa^-1)

0.7

0.6

0.5

0.4

0.3
-2 -1 0 1 2
Log Frequency (Hz)
25 35 50 65 80
25A 35A 65A 80A

Figure 6.6.5 Elastic compliance 1 axis measured data compared to


predictions using modified Arrhenius equation (A suffix).
106

Arrhenius Fit to Piezo Compliance/Temperature


0622-2 Stretch Direction (1 Axis)

24

22 Increasing Temperature

20
d31 pC/N

18

16

14

12
-2 -1 0 1 2
Log Frequency (Hz)
25 35 50 65 80
25A 35A 65a 80A

Figure 6.6.6 Piezo compliance 1 axis measured data compared to


predictions using modified Arrhenius equation (A suffix).

Expanded Master Curves


Elastic Loss Compliance
7%
6%
% Elastic Loss

Increasing Temperature 293


5%
308
4%
323
3%
338
2%
353
1%
0%
-10 -5 0 5 10 15

Log Frequency (Hz)

Figure 6.6.7 Elastic loss TTS master curves for all reference temperatures.
107

Normalized Elastic s11'' Master Curves


4.0

Nomalized Elastic Loss


3.5
293
3.0
308
2.5
323
2.0 338
1.5 353
1.0 Fit
0.5
0.0
-10 -5 0 5 10 15

Log Frequency (Hz)

Figure 6.6.8 Data of Fig 6.6.7 normalized at 1 Hz and a master curve fit.

Expanded Master Curves


Piezo Loss Compliance
7%
Increasing Temperature
6%
% Piezo Loss

293
5%
308
4%
323
3%
338
2%
353
1%
0%
-10 -5 0 5 10 15

Log Frequency ( Hz)


Figure 6.6.9 Piezo loss master TTS curves for all reference temperatures.
108

Normalized Piezo d31'' Master Curves


4.0

Nomalized Piezo Loss


3.5 293
3.0 308
2.5 323
2.0
338
1.5
353
1.0
Fit
0.5
0.0
-10 -5 0 5 10 15

Log Frequency (Hz)

Figure 6.6.10 Data of Fig 6.6.9 normalized at 1 Hz and a master curve fit.
109

CHAPTER 7 DIELECTRIC DATA ANALYSIS

The general study of the dielectric behavior of polymers is exhaustive, with

hundreds of articles and many texts written on the subject. The mechanisms commonly

associated with dielectric relaxations in polymers are described in detail by Böttcher [86]

and McCrum [84] among others. As an example, the classical and simplistic DeBye

theory of dielectric relaxation is based upon a model of a rigid spherical molecule with

polarity rotating with viscous friction under the influence of an electric field. Other

modifications to this approach by researchers such as Eyring have attempted to include the

effects of dipole rotation and translations in terms of energy barriers. Further

modifications have been suggested to include the relaxation broadening effects of the

distribution of molecular weight (polydispersity), and the effects of the fractional free

volume. Empirical modifications to these kinds of theories have produced well known

models useful for particular materials, such as the Cole-Cole, Fuoss-Kirkwood, and

Davidson-Cole [84] equations. These approaches are useful in conveying a basic

understanding of polymer dielectric behavior, but only in a very general manner - they are

not directly useful in this study of PVDF.

Specific studies of the dielectric properties of piezoelectric PVDF are of course

more appropriate to this work [1, 20, 22, 51, 95-98]. The physical mechanisms for the

particular dielectric behavior of PVDF are discussed by [22, 94, 96-99] among others and

are thought to result primarily from the large dipole moment inherent in the PVDF " and

$ crystal structures, the semi-crystalline nature of PVDF, and the morphology of the

amorphous-crystalline interface. PVDF has one of the highest permittivities of any


110

engineering polymer. This is a result of rotation and displacement of the dipoles under the

stress of an electric field, which is accomplished through backbone chain motion, sub

chain motion, conformational rearrangements (degrees of order/disorder in the crystal-

amorphous boundary), and molecular diffusion. Other theories, such as those forming the

basis of the WLF equation, are based on free volume calculations [100, 101]. These

theories generally include ‘hindrance’ mechanisms, such as viscous and Coulomb friction

between adjacent chains, in order to explain the dielectric loss behavior.

Despite the amount of empirical data reported on PVDF, there are few research

efforts which attempt to provide equations describing dielectric behavior as a function of

temperature and frequency. Several of these approaches are discussed below.

7.1 TTS Expansion of Dielectric Data

The success of the TTS expansion applied to the elastic and piezoelectric data in

chapter 6, and the strong indication that the relaxation mechanisms in both of these

parameters are identical, leads one to consider the possibility of using the same technique

to expand the dielectric properties. Malecki [96] and others have proposed using the TTS

expansion technique for the dielectric behavior of PVDF with the caveat that the WLF

TTS approach is not reasonable in the temperature and frequency ranges of the dielectric

relaxation. Figure 7.1.1 shows an attempt to use the TTS approach to expand the

permittivity data from chapter 5. This expansion was done with the same WLF constants

used to determine the shift factor for the elastic and piezoelectric properties in chapter 6.

The temperature range of the dielectric data used in the expansion was -20°C to +90°C

instead of the original -60°C to +125°C temperature range over which the data was
111

measured in order to more closely match the elastic and piezoelectric data temperature

range. Even over these reduced temperatures, however, the resulting WLF TTS

expansion is disappointing. Clearly this temperature and frequency range contains

significant dielectric relaxation. The frequency shift factor constants were altered in an

attempt to provide a closer fit to the data without significant success. This implies that the

elastic/piezoelectric and dielectric relaxation mechanisms are not the same, and therefore

an alternative approach is necessary.

WLF TTS Expansion of Permittivity


C1 & C2 same as elastic case, To=298K, -20 to 90C data

1.6E-10
1.4E-10
Permittivity F/m

1.2E-10
1E-10
8E-11
6E-11
4E-11
2E-11
0
0 2 4 6 8 10 12 14
Log Frequency (Hz)

Figure 7.1.1 WLF TTS expansion of dielectric permittivity of PVDF.


112

7.2 The Cole-Cole Equation

The classical dielectric relaxation equation offered by DeBye [84] is often used to

model single dielectric relaxation events as a function of frequency:

(7.1a)

(7.1b)

(7.1c)

In these equations, ,r and ,u represent the permittivity at very low frequency and at very

high frequency, respectively. JE is the time constant for the relaxation, being equal to the

reciprocal of the radian frequency at which the relaxation occurs. This approach is also

used in modified form to describe relaxation distributions, with each unique relaxation in

the distribution having a different time constant.

The major drawback with this equation is that it generally does not fit well to

empirical data [86, 94, 96]. The relaxation described by (7.1) occurs in about a single

order of magnitude in the frequency domain, while dielectric relaxations in polymers are

typically much broader. In order to accommodate this broadening, Cole and Cole [93]

modified the DeBye expressions with an exponent, ( (0 < ( # 1), such that

(7.2a)
113

This is known as the Cole-Cole equation. From (7.2a) we have

(7.2b)

(7.2c)

Lowering the value of ( flattens the permittivity versus log(T) relaxation curve,

and setting ( = 0 has the effect of making the permittivity independent of frequency.

Conversely, setting ( = 1 produces the sharp transition of the DeBye equations.

7.3 Vogel-Fulcher Modified Cole-Cole Equation

The Cole-Cole equations are able to reproduce the dielectric relaxations found in

PVDF [96] at a given temperature. However, (7.2) is clearly not a function of

temperature, and therefore this equation is of very limited value when attempting to

predict dielectric behavior over a range of temperatures.

Capellades [95] suggested the well known Vogel-Fulcher equation as one method

to predict the relaxation frequency as a function of temperature in PVDF. Equation

coefficients given by [95] produced a reasonable fit to the PVDF data he presented. This

equation was given in the following form:

(7.3)
114

Capellades et.al. determined the coefficient values to be: 2M = 187 °K, (U/k) = 446, 2 =

the temperature of interest in °K, and J0 = 10 -10.98 seconds. Recasting this equation in

terms of relaxation frequency as a function of temperature gives

(7.4)

where Tr is the frequency at which the relaxation peak occurs.

A study of the dielectric data presented in chapter 5 led this author to consider

integrating the Vogel-Fulcher equation into the Cole-Cole equation in order to provide a

method for specifically predicting permittivity as a function of temperature and frequency.

In addition, the data clearly suggests that there is a power law relationship between

permittivity and frequency in regions well away from the relaxation event, as seen by the

distinct linear slope in the permittivity versus log(T) curves in Figure 5.4.3 away from the

relaxations. Further, a number of authors have suggested that power relations are

reasonable in these frequency regions [95- 97]. Therefore, various slopes were empirically

curve fit to both the high frequency and low frequency regions of the measured

permittivity data. It was determined in this work that a slope on the high frequency part of

the spectrum was not necessary. Also, the Vogel-Fulcher coefficients provided in [95]

were adjusted slightly to give the best performance for the data in chapter 5. One reason

for the difference may be that Cappellades et. al. used piezoelectric PVDF films produced

by Solvay, based on the Solef series of PVDF resins. The PVDF films in this work were

produced by MSI using ATOFINA’s Kynar PVDF resins, which are produced using a

slightly different emulsion synthesis process.


115

The resulting Vogel-Fulcher modified Cole-Cole equation describes complex

permittivity over temperature and frequency:

(7.5)

where J(2) is the relaxation time constant as a function of temperature. With the

inclusion of the low frequency ,r slope, (7.5) becomes

(7.6)

(7.7)

(7.8)

(7.9)

(7.10)

The temperature dependent relaxation frequency described by this equation

matches the dielectric data reported in chapter 5 quite well, as seen in Figure 7.3.1.
116

Permittivity S8504
1.6E-10
-60
1.4E-10
Permittivity, F/m -30
1.2E-10 0
1.0E-10 30
8.0E-11 60
6.0E-11 90

4.0E-11 -60
-30
2.0E-11
0
0.0E+00
10 100 1000 10000 100000 1000000 30
60
Frequency (Hz)
90

Figure 7.3.1 Curve fit to permitivity data. First set of temperatures listed
are data curve, second set are curve fits using (7.4) - (7.8).

Turning now to the dielectric loss term, note that (7.2c) approaches zero as the

frequency approaches zero. This expression is accurate for the displacement loss, typically

assumed to be the result of molecular friction, but it clearly does not accurately model the

empirical data. An additional term is required to account for the losses due to conduction

through the dielectric material [44, 102]. This conduction is in phase with the voltage on

the dielectric, and cannot be distinguished from the displacement losses by the measuring

equipment. Also, it cannot contribute to the storage permittivity, ,!, and cannot be

detected by observing that parameter. McCrum [84] discussed this problem and

suggested that the complete dielectric loss term be composed of two parts:

(7.11)
117

Here ,"D is determined by expanding (7.2c) in the manner of (7.6) to produce

(7.12)

Equation (7.12) did not produce a satisfactory fit to the measured data, which

showed too rapid a rise in conductivity with decreasing frequency and increasing loss with

increasing frequency due to electrode resistance. Improved results are found by modifying

(7.11) as follows:

(7.13)

(7.14)

Here A is the electrode area of the test sample, a is the sample thickness, and RS is the

electrode equivalent series resistance. 7 is an arbitrary curve fit constant of unit value

with dimensions of [sec3/4]. The third part of (7.13) is derived from simply adding a

resistance in series with the parallel resistor and capacitor film model. This resistance

could not be compensated in the equipment and was calculated to be about 6 ohms, which

seems quite reasonable for a metalization layer with a sheet resistivity of about 2 S/9.

This term is only really important at frequencies above 100 kHz for the samples tested.

Although the dielectric loss tangent is by far the most difficult material parameter to
118

cureve fit, (7.14) produces an acceptable match to the data as seen in Figure 7.3.2. The

value of D (resistivity) shown in (7.14) was also determined by curve fitting. The need for

an exponent on the radian frequency in the second term of (7.13) and the fact that the bulk

resistivity numbers produced by curve fitting are smaller than typical values given in the

literature [103] are not ideal. There is an indication that there is some other mechanism at

work in the dielectric loss tangent, one not familiar to this author or others that have

reported in this field. Nonetheless, the primary intent of this work is to provide useful

cure fit equations such as (7.13).

Dielectric Loss vs Temp vs Freq


0.30 -30
0.25 0
Dielectric Loss

30
0.20
60
0.15 90
0.10 -30
0
0.05
30
0.00
60
10 100 1000 10000 100000 1000000
90
Frequency (Hz)

Figure 7.3.2 Curve fit to dielectric loss data. First set of plots are data,
second set are curve fits using (7.12) - (7.14).
119

CHAPTER 8 CONSTITUTIVE EQUATIONS

A number of empirical equations have been developed in previous chapters in

order to describe and predict the elastic, piezoelectric, and dielectric material properties of

PVDF over frequency and temperature. The purpose of this chapter is to summarize these

equations and to frame them in terms of constitutive relations.

8.1 Constitutive Equations

Recall from chapter 6 we have

(6.8)

(6.9)

Further, from chapter 6 these constants are drawn:

(8.1)

(8.2)
120

It appears convenient to rewrite (6.8) and (6.9) into a shorter form. To this end let

(8.3)

(8.4)

Equations (6.8) and (6.9) then become

(8.5)

(8.6)

The dielectric equations developed in chapter 7 are presented here as an integral

part of the constitutive equations:

(7.6)
121

with (7.12) combined with (7.13) to produce

(8.7)

Recall the strain and displacement constitutive equations from chapter 3,

(3.7)

(3.8)

Substituting (8.1) - (8.7) and (7.6) into these equations for the isothermal, uniaxial strain

cases and isothermal displacement case yields the following equations in complex notation:

(8.8)

(8.9)
122

(8.10)

Contrary to current piezoelectric constitutive equations, (8.8) through (8.10) are

constructed in complex form to allow designers to predict complex (lossy) behavior in the

frequency domain. Curve fits between these material property equations and measured

data are good and were shown in chapters 6 and 7.

8.2 Useful Frequency and Temperature Bounds

The expanded form of the piezoelectric and elastic terms, discussed in chapter 6

and contained in the equations above, have no intrinsic bounds in temperature or

frequency. However, PVDF is a polymer and as such has some fundamental limits. Recall

that commercially available PVDF is annealed to a specific temperature during a post

poling process in order to provide thermal stability. Heating the material to temperatures

much above this annealing temperature causes permanent mechanical relaxation

(shrinkage) in PVDF that is not clamped. This relaxation causes irreversible losses in

piezoactivity, primarily due to dipole relaxation. It is therefore not advisable to use the

constitutive equations when projecting material properties at temperatures above the

annealing temperature of the film being used. Currently the majority of commercial PVDF

is annealed to 85°C, but films stable to 125°C are now being introduced. These new films

are expected to grow to as much as 50% of the market, driven by the widely accepted

industrial maximum sensor temperature range of 125°. The constitutive relations


123

developed in this work are broad and are based on a technique of relative scaling: material

property values measured at one frequency and temperature are expanded to predict

material properties over a much wider range of temperatures and frequencies. It is the

author’s expectation, therefore, that the general material property framework constructed

in this work will be applicable to PVDF films annealed to 125°C. Verification of this will

be the subject of future work.

Because PVDF is a semi-crystalline polymer, it has a significant fraction of

amorphous material. When the temperature to a critical point the amorphous material

‘freezes’ and transitions to a glassy phase, suffering a radical change in material properties.

The temperature at which this occurs is the glass transition temperature, or Ig. It is no

surprise that the glass transition temperature is a function of frequency [1, 20, 22],

increasing as frequency increases. This phenomenon may be observed clearly in the ridge

which occurs in the dielectric loss term in Figure 5.4.4, as the temperature of the peak of

this ridge is the glass transition temperature. Based on this data, and on curves of d31

versus temperature at a single frequency of 10 Hz [1, 20], a reasonable low temperature

limit at low frequencies is -15°C.

The operating bandwidth of PVDF transducers is quite broad. Some low

frequency applications may extend to 0.005 Hz, including force detection, biomedical

pulse pressure sensors, and geophones. The natural limit to the use of any ferroelectric

material at low frequencies in the voltage mode is the equivalent internal resistance of the

material itself. A convenient figure of merit for this conduction term is the product of the

bulk resistivity and the dielectric permittivity, in units of seconds, representing the time
124

constant associated with a first order voltage decay through the material (the time it takes

for a voltage on the film to decay to 1/e of its original value). For PVDF this term is

about 2,500 seconds, or about 60 µHz. This is the intrinsic low frequency limit for the

extrapolation used in these constitutive equations. In charge mode applications it is

difficult to achieve frequencies much below 0.005 Hz without serious drift.

High frequency applications range from ultrasound transducers, which operate

from hundreds of kilohertz to several megahertz, up to bulk acoustic wave transducers

used in RFID (radio frequency identification) tags, which operate at about 900 MHz.

Practical use of PVDF in its length extensional mode is limited to low megahertz

frequencies due to its acoustic velocity. Thin film PVDF half wave or quarter wave

resonators for ultrasonic IDT (interdigitated electrode transducer) filters have a minimum

useful length of perhaps 50 µm, resonating at about 15 MHz.

In summary, the material property equations developed in this work are expected

to be useful in a bandwidth from 0.005 Hz to 15 MHz., and from about -15°C up to the

annealing temperature, which may extend to 125°C.


125

CHAPTER 9 EXPERIMENTAL VERIFICATION

The data expansion techniques developed in chapter 6 led to empirical equations

describing PVDF material properties over a wide range of temperatures and frequencies.

These material property equations were incorporated into constitutive equations, as

discussed in chapter 8. These constitutive equations are therefore capable of predicting

piezoelectric, elastic, and dielectric performance over the frequency and temperature range

of the material property equations. This chapter discusses testing done to verify the

accuracy of these predictive equations. Specifically, the piezoelectric and viscoelastic

properties were measured at high frequency and the results compared with calculations.

9.1 Experimental Technique

The technique used for verification testing is entirely electrical and does not

involve the apparatus previously used in chapter 4. The technique, suggested by Linville

in [105], was adapted by the author for this work and is described briefly below. Figure

9.1.1 contains a schematic outline of a sample in this test configuration. In this approach,

a sample of PVDF film prepared with two electrode patterns on the top side and a

continuous ground electrode on the other is clamped at the ends under tension. A drive

voltage is applied to one electrode of the sample, piezoelectrically inducing stress into the

material bounded by this electrode. This stress induces a piezoelectric response in the

remaining material and a sense voltage is measured from a second electrode covering this

region. Using the nomenclature in Figure 9.1.1, the relationship between the piezoelectric

and elastic material properties can be predicted by the ratio of the measured (sensed)
126

voltage to the applied (drive) voltage.

The equivalent spring constants for the driven (subscript a) and the sensed

(subscript b) film segments are [104]

(9.1)

Figure 9.1.1 Layout and nomenclature for verification test.

Notice that the compliances of the two segments are written separately, denoted

by the “a” and “b” subscripts. Using the simple force/spring constant/displacement

relationship F = k ), and knowing that the force, F, is induced by the piezoelectric effect

(9.2)
127

the displacement ) of the a-b junction can be written as

(9.3)

It is important to note that the open circuit voltage developed by the sensing side

of the test sample has a polarity that causes the compliance of the material under the

sensing electrode to decrease. This has the net effect of working against the stress

induced by the driving voltage and decreasing the displacement, ), and therefore this

electromechanical coupling is considered in the analysis. This compliance is written as

s11D, recalling the superscript convention from chapter 3. Because the electronic driving

circuit applying a voltage to side “a” of the sample has a very low source impedance, the

compliance of this driven side of the film is written as s11E . The electro-mechanical

coupling coefficient of a piezoelectric material is defined by a simple relationship of stored

electrical energy that results from applied mechanical energy (or vice versa),

(9.4)

and the relationship between the open circuit (s11D) and short circuit (s11E) compliance of a

piezoelectric material is [4]

(9.5)

Rewriting (9.1) and substituting into (9.3) for ) yields


128

(9.6)

Knowing that the open circuit voltage of a piezo material can be simply written as

(9.7)

and substituting S1 = )/b with (9.6) into (9.7) produces, after simplification,

(9.8)

Simplification of (9.4) and insertion of actual material properties leads to the

calculated value of k312 = 0.0045. Since this term is quite small (n1), we can ignore it.

Setting the two electrode regions equal, a = b, gives the following simplified equation

which will be used in this chapter to relate piezoelectric, elastic, and dielectric material

properties to the ratio of measured voltage to driven voltage:

(9.9)

It can be seen from (6.8) that the frequency and temperature dependence of d31 and s11 are

identical. This implies that the voltage ratio in (9.9) will behave in a manner similar to

(6.8) over temperature/frequency when (9.9) is used with the initial d31 and s11 values.
129

9.2 Apparatus

Samples were prepared by slitting metalized PVDF film to a width of 1 cm and

cutting pieces to a length of 3 cm. Electrode regions were reticulated using an extremely

sharp razor tip (Exacto type) to skive the electrode metal from the PVDF, using a straight

edge and a micrometer under a stereoscopic microscope. A small bead of conductive

silver epoxy was used to bond a 30 AWG wire to the surface of the sensing electrode.

The test sample was then carefully installed in the fixture and the clamp on one side was

tightened. Tension was applied to the other side of the sample as the second clamp was

tightened, leaving the film taut between the clamps. The sample was mounted such that

the driven electrode occupied exactly one half of the gap between the clamps. Dimensions

were measured using an optical inspection station accurate to within 0.05 mm, and actual

dimensions were recorded.

The drive electrode (side “a”) was connected to the source output of an HP3562A

Dynamic Signal Analyzer (see Appendix C for specifications). This output was also

connected to the DSA channel 1 input. The sense electrode on the other side of the film

sample (side “b”) was connected to the input of a pre-amplifier. This amplifier had a 50

Gohm input impedance and was built using a Burr Brown OP124A instrumentation opamp

designed to measure very low level voltages with high transducer source impedance. The

amplifier system had a gain of 1.0 and a gain-bandwidth product of over 1 MHz. It was

powered by batteries and mounted in a metal enclosure in order to minimize electrical

noise. The output of this pre-amplifier was connected to channel 2 of the DSA. The DSA

source was programmed to supply 5 Vp to the driven electrode and also to channel 1,
130

over a frequency range of 10 Hz to 100 kHz. The temperature of the test apparatus and

film sample was not controlled directly, although the apparatus was insulated and the

temperature was measured at 24°C.

The use of a pre-amplifier was essential in order to achieve the desired

measurement bandwidth. A high input impedance was necessary in order to lower the first

order high pass filter pole created by the film capacitance combined with the amplifier

input impedance. A 50 Gohm input impedance and 50 pF of film capacitance gave a low

end roll off frequency of about 0.1 Hz. This frequency was low enough that the phase

deviation caused by the filter above 10 - 20 Hz was below 0.5 degrees and the phase

information was therefore useful above that point. In contrast, the DSA input impedance

is 1 Mohm, and without a preamplifier the lowest useful frequency would have been about

90 kHz. Likewise, the low output impedance of the amplifier was necessary to overcome

the shunt capacitance of the coax cable between the amplifier and the DSA (about 70 pF)

and the losses due to the complex DSA input impedance (1 Mohm in parallel with 100 pF)

at high frequencies.

A number of unusual problems were encountered during this test. One issue

requiring serious consideration was the length of the sample between the clamps. The

acoustic velocity in this axis of PVDF is about 1,500 m/s, and in order to push the first

mode acoustic resonance above 100 kHz the sample had to be below 5 mm or so in

length. The first mode, or half wave resonance of the sample can be approximated by

(9.10)
131

where 8 is the separation between the clamps. In order to make sure that the resonance

was well above 100kHz the clamp separation was set to 3.0 mm, giving a resonance at

roughly 250 kHz. Even with the resonance that high, piezoelectric loss data could not be

measured accurately at frequencies above about 5kHz due to the resonance phase shift.

Recall that the loss terms of interest are quite small, on the order of a few degrees, and

that there is a 180 degree phase shift through the acoustic resonance.

It was also discovered that the mass of the tiny epoxy bead and the small attached

wire were sufficient to cause out of plane resonances with film displacements normal to

the flat surface. Although the magnitude of these resonances were quite small, they

caused a subtle phase shift around 10 - 15 kHz which made it even more difficult to

accurately measure the piezoelectric loss term. In order to reduce this effect, a special

cutout feature was added to one of the clamps allowing the region of the sample with the

epoxy bead to be supported underneath by one clamp, thereby blocking out of plane

motion. The active electrode region on the measurement side of the sample was created

such that the region around the epoxy bead had no ground electrode. Even with the

modifications, a slight resonance persisted.

The need to keep the sample small to avoid acoustic resonances had the

unfortunate effect of reducing the sense electrode area to a typical source capacitance

value of 50 pF. This low capacitance value, coupled with the very small Vb/Va ratio of

about 1/500, created another condition requiring careful experimental attention. Figure

9.2.1 shows an electrical schematic of the test apparatus. Note that in addition to the

capacitance of the drive film, Ca, and the sense film, Cb, there is a coupling capacitance
132

depicted, Cc. Weak capacitive coupling is a natural occurrence in any physical apparatus,

but in this case even very weak coupling is significant. Because Vb . Va/500, the effect of

charge injected by the coupling capacitor on the measured output voltage can be

calculated by the simple equation

(9.11)

where Vb! is the measured voltage and Vb is the actual open circuit voltage generated by

the PVDF in section b. The exact source of the coupling is not known - it may have been

due to fringing effects between the adjacent drive and sense electrodes. Extremely careful

measurement of all three capacitance values was necessary to properly compensate the

measured voltage ratio. Cs is the shunt capacitance of the preamp input.

Substituting actual values leads to the conclusion that the charge injected by the

measured coupling capacitor value of about 0.080 pF is roughly equal to the generated

signal from section b. The data presented in the next section has been compensated by

removing the effects of this coupling. All capacitances were measured


on an HP4284A

precision LCR bridge with a resolution of 0.1 fF (0.0001 pF). Using 1,024 averages

produced steady and consistent capacitance baselines. The coupling capacitance was

deduced by comparing the expected series capacitance equivalent of the sense and drive

capacitances, Cb and Ca, to the actual series capacitance value, Cab, measured from the top

of Ca to the top of Cb with the ground connection removed:


133

(9.12)

Figure 9.2.1 Schematic of verification test.

9.3 Experimental Data

Figure 9.3.1. shows the results of verification testing at a temperature of 24°C.

The measurements discussed in the last section were used to remove the artifact of

injected charge due to coupling capacitance from the raw data. This compensation

baseline was adjusted slightly from about 80 fF to about 67 fF to allow the 10 Hz Vb/Va

ratios to match those calculated from actual d31 data, implying that the Cc measurement is

slightly in error. Because this coupling capacitance is so small compared to the


134

capacitance of the sample (about 1/1000 Cb), this seems a reasonable correction. The

calculated values shown in Figure 9.3.1 were obtained using (6.8) for elastic and

piezoelectric compliances, and (7.6) for the dielectric permittivity at 24°C.

The fit of the measured data to the calculated values is remarkably good.. Even

though the fundamental resonance occurs at 250 kHz with a fairly low Q, there is a slight

but measurable impact at 100 kHz.

The phase response of the film was quite reasonable at frequencies below 1 kHz.

The phase measurement is so sensitive to the onset of acoustic resonance that above this

frequency the phase data is not representative of material loss properties, and therefore no

comparison could be made to calculated values for elastic loss.

Voltage Ratio Verification

2.0

1.9
Va/Vb/1000

1.8
Meas

1.7 Calc

1.6

1.5
10 100 1000 10000 100000

Frequency (Hz)

Figure 9.3.1 Voltage ratio comparison, measured vs. calculated.


135

CHAPTER 10 CONCLUSIONS

In this work the elastic and piezoelectric material properties of PVDF were

measured over a temperature range of 20°C to 80°C and a frequency range of 0.01 Hz to

100 Hz. Special equipment and techniques were developed specifically for these

measurements. A Time-Temperature Superposition (TTS) expansion of the resulting

data confirmed thermorheologically simple behavior for these parameters. A novel TTS

expansion technique was employed to allow an even broader expansion of the original

data. Equations describing the viscoelastic and piezoelectric properties over a wide range

of temperature and frequency were constructed using these expansions. The dielectric

properties of PVDF were measured over the range of -60°C to 80°C and from 20 Hz to

1MHz. Modifications were proposed to classical dielectric equations and confirmed for

PVDF by comparison to measured data. The resulting material property relationships

were put into the framework of piezoelectric constitutive equations. Finally, a special

technique was employed to test the ability of the elastic, piezoelectric, and dielectric

property equations to predict behavior outside of the measured temperature/frequency

range.

The primarily focus of this work was to provide useful constitutive relationships

containing functions describing the elastic, piezoelectric, and dielectric properties of

piezoelectric PVDF film over temperature and frequency. It is expected that these

equations will be used by transducer engineers and others designing and evaluating PVDF

sensors for new applications. This work will also enable a broader understanding of the

viscoelastic and visco-piezoelectric behavior of this piezo material.


136

10.1 List of Key Results

C An apparatus was designed and built to measure complex piezoelectric and elastic

compliances from 0.01 Hz to 100 Hz over a temperature range of 20 to 80°C. Complex

dielectric behavior was reported from 20 Hz to 1 MHz and from -60 to +80°C.

C Vinogradov et. al. [9] showed that the viscoelastic behavior of both in plane axes

of PVDF were thermorheologically simple (TRS), meaning that it can be predicted over

temperature and frequency using the WLF equation. Their measurements covered a

limited range of frequencies from 1 to 50 Hz. Their findings were confirmed in this work

over the much broader frequency range of 0.01 Hz to 100 Hz.

C Vinogradov et. al. [10-11] suggested that the piezoelectric properties of PVDF

were unaffected by either temperature or frequency and could be considered constant.

Clear evidence was presented in this work showing that the piezoelectric properties are a

function of temperature and frequency. Further, these functions were shown to be TRS.

C The TTS expansions of both the elastic and piezoelectric data in this work were

accomplished using the identical WLF constants, and it was clearly shown that the form of

the piezoelectric expansion was exactly that of the elastic case. This holds for both the

storage and loss terms associated with the complex variables d31*, s11*, and s22*. The

similarity between the behavior of these piezoelectric and elastic compliances over

temperature and frequency was unanticipated. It appears that the same mechanisms are

responsible for both viscoelastic and visco-piezoelectric behavior.

C The storage and loss terms of the d32 coefficient were found to vary little over the

measured range, and given the relative unimportance of this term constant values were

assigned.
137

C The unusual sign of the d32" term suggested by Furukawa [20] was confirmed in

this work. The value of this compliance loss parameter was expected to be negative on

theoretical grounds but was clearly found to be positive. No reasonable explanation for

such behavior has been determined yet. This is a topic for further work.

C A frequency modified Arrhenius equation was proposed to allow the expansion of

the raw elastic and piezoelectric data along the temperature axis prior to the typical TTS

frequency axis expansion. This technique allowed prediction of material properties over a

much broader range of temperature and frequency compared to the normal TTS approach.

The equation was compared to data with excellent results.

C The similarity in elastic and piezoelectric behavior over temperature and frequency,

combined with the Arrhenius expansion in the temperature axis, led to the construction of

powerful equations for predicting material properties. These equations allow

extrapolation over a wide range of temperatures and frequencies from a single value of

each parameter measured at a known temperature and frequency. For example,

piezoelectric and elastic properties measured at 1 Hz and 50°C can be used to predict the

piezoelectric and elastic material properties over the useful operating range of PVDF:

#0.001 Hz to low Mhz and from -15°C to 85°C. Further, it is expected that these

equations will also be applicable to new PVDF films with a maximum operating

temperature of 125°C.

C The classical Cole-Cole equation for complex permittivity was proposed and

modified with a temperature dependent Vogel-Fulcher term in order to accurately predict

dielectric behavior. The loss equation was further modified to account for the conductive

losses through PVDF at low frequencies and increased temperature. The resulting
138

equations for ,! and ," were compared to measured data with good results.

C The equations developed in this work predicting the behavior of material

properties over temperature and frequency were incorporated into a set of constitutive

relations for electric displacement and elastic strain. These equations were written in

complex form to facilitate solutions in the frequency domain.

C The effect of the thin, sputtered metal electrodes on elastic material property

measurements was investigated. Data indicated that even thin metal electrodes have an

impact on the results and must be considered. A similar impact on the piezoelectric

property measurements is expected.

10.2 Key Contributions of this Work

C A broad range of frequency and temperature data were presented in a single work

for a commercial grade of piezoelectric PVDF.

C Thermorhelologically simple (TRS) behavior was confirmed for the elastic

properties of PVDF over a frequency range larger than reported to date.

C The d31 piezoelectric behavior was shown to be TRS for the first time.

C The TRS behavior of the elastic and piezoelectric material properties were shown

to be identical, implying that the physical mechanism responsible for visco elastic behavior

is responsible for visco-piezoelectric behavior.

C Material property equations were given as functions of temperature and frequency.

C Material property functions were provided in a set of frequency domain

constitutive equations.
139

List of References

1 Wang TT, Herbert JM ,Glass AM. The Applications of Ferroelectric Polymers.


Blackie, Glascow and London, 1988

2 Trolier-McKinstry S, Newnham RE. Sensors, Actuators, and Smart Materials.


MRS Bulletin, April 1993

3 Berlincourt D. Piezoelectric Crystals and Ceramics, in Ultrasonic Transducer


Materials. 1971; 64-124, Edit by O. Mattiat, Plenum Press, New York-London

4 Ikeda T. Fundamentals of Piezoelectricity. 1996; Oxford Science Publications

5 Piezoelectric Ceramics Data Book for Designers. Published by Morgan Matroc


Inc; Milwaukee, WI

6 Piezoelectric ceramics, product catalog, applications notes. Published by Sensor


Technology Ltd; Collingwood, Ontario, Canada; 1995

7 Users guide to ultrasound & optical products. Published by Valpey-Fisher


Corporation; Hopkinton, MA; 1996

8 Schmidt VH, Brandt D., Holloway F., Vinogradov A., Rosenberg D. Piezoelectric
Polymer Actuator and Material Properties. Proc. Tenth IEEE Intl. Symp. on Appl.
of Ferroelectrics Aug 1996; East Brunswick, NJ

9 Vinogradov AM, Holloway F. Mechanical Testing and Characterization of PVDF,


a Thin Film Piezoelectric Polymer. J Adv Materials 1997:29 (1)

10 Vinogradov AM, Holloway F. Electro-mechanical properties of the piezoelectric


polymer PVDF. Ferroelectrics 1999: 226;169-181

11 Vinogradov A. A constitutive model of piezoelectric polymer PVDF. Proc SPIE


Mar 1999:3667;711-718

12 Ferry JD. Viscoelastic Properties of Polymers. 1980; John Wiley and Sons

13 Taylor GW, et. al. Editors. Piezoelectricity. 1985 Gordon and Breach, NY, NY

14 Busch G. Early History of Ferroelectricity. Ferroelectrics 1987; 74; 267-284

15 Kanzig W. History of Ferroelectricity 1938-1955. Ferroelectrics 1987:74; 285-


291
140

16 Misui T, Tatsuzaki I, Nakamura E. An Introduction to the Physics of


Ferroelectrics, 1976: Gordon and Breach, NY, NY

17 Halliday D, Resnick R. Physics, Parts I and II. John Wiley & Sons 1966; 963-
964

18 Kawai H. The Piezoelectricity of PVDF. Jap. J. Appl. Phys 1969: Vol 8; 975

19 Murayama N, Nakamura R, Obara H, Segawa M. The Strong Piezoelectricity in


PVDF. Ultrasonics 1976:14; 15

20 Fukada E, Furakawa T. Piezoelectricity and Ferroelectricity in Polyvinylidene


Fluoride. Ultrasonics Jan 1981; 31-39

21 Wang H, Zhang QM, Cross LE. Piezoelectric, dielectric, and elastic properties of
poly(vinylidene fluoride/trifluoroethylene). J Appl Phys 1993:74(5);3394-3398

22 Kepler RG, Anderson RA. Ferroelectric polymers. Advances in Phys 1992:41(1);


1-57

23 Broadhurst MG, Davis GT, McKinney JE. Piezoelectricity and pyroelectricity in


polyvinylidene fluoride - A model. J. Appl. Phys. Oct 1978:49 (10)

24 Furakawa T, Aiba J, Fukada E. Piezoelectric relaxation in polyvinylidene


fluoride. J Appl Phys May 1979:50(5);3615-3621

25 Ewing RL. The electrocaloric effect on the pyroelectric charge in polyvinylidene


fluoride. Ferroelectric 1993:141;243-256

26 Sasabe H, Nakayama T, Kumazawa K, Miyata S, Fukada E. Photovoltaic effect


in Poly(vinylidene fluoride). Polymer Journal 1981:13(10);967-973

27 Bune AV, Fridkin VM, Verkhovskaya KA. Some investigations of photoelectric


and optic properties in the ferroelectric PVDF. Ferroelectrics 1988:81;361-364

28 Nanayakkara SJ, Orchard GAJ, Davies GR, Ward IM. Thermal expansion of
Poly(vinylidene fluoride). J Polymer Science: Part B: Polymer Physics
1987:25;1113-1128

29 Wang TT. Stress Relaxation and its Influence on Piezoelectric Retention


Characteristics of Uniaxially Stretched Poly(Vinylidene Fluoride) Films. J. Appl.
Phys Mar 1982: 53(3)

30 Yamada K, Oie M, Takayanagi M. Piezoelectricity and viscoelastic crystalline


dispersion in highly oriented poly(vinylidene fluoride). J Polymer Science:
141

Polymer Physics Edition 1984:22; 245 - 254

31 Toda M. Elastic Properties if Piezoelectric PVF2. J Appl. Phys Sep


1980:51(9);4673-4677

32 Lynch CS. Polyvinylidene Fluoride (PVDF) Elastic, Piezoelectric, Pyroelectric,


and Dielectric Coefficients and Their Non-Linearities. Ferroelectrics 1993:150;
331-342

33 Bretz PE, Hertzberg RW, Manson JA. Mechanisms of fatigue damage and
fracture in semi-crystalline polymers. Polymer Sep 1981:22;1272-1278

34 Bur AJ, Barnes JD, Wahlstrand KJ. A study of thermal depolariztion of


polyvinylidene fluoride using x-ray pole figure observations. J Appl Phys Apr
1986:59(7);2345-2354

35 Bauer F. PVF2 polymers: ferroelectric polarization and piezoelectric properties


under dynamic pressure and shock wave action. Ferroelectrics 1983:49;231-240

36 Choy CL, Leung WY, Chen FC. Heat capacity of poly(vinylidene fluoride) and
polytetrafluoroethylene between 5 and 200(K. J Polymer Science: Polymer
Physics Edition 1979:17;87-94

37 Das-Gupta. Ferroelectric polymers and ceramic-polymer composites. Key


Engineering Materials 1994:92-93;15-30;Trans Tech Publications

38 Fukada E. History and recent progress in piezoelectric polymer research. Proc


IEEE Ultrasonics Symposium 1998; 597-605

39 Gafurov AK, Yakubov NI. A study of mechanical heat stability and resistance to
thermal degradation in copolymers of vinyl fluoride with fluorothylenes. Polymer
Science USSR 1987:29(8);1895-1901

40 Ibos L, et. al. Correlation between pyroelectric properties and dielectric behavior
in ferroelectric polymers. Ferroelectrics 2000: 238;163-170

41 Kokorowski SA. Analysis of adaptive optical elements made from piezoelectric


bimorphs. J Opt Soc Amer Jan 1979:69;181-187

42 Swei GS, Rickert SE, Lando JB. Structural aspects of pyroelectric and
piezoelectric polymers: Poly(vinylidene fluoride) and its derivatives. 1988

43 Sao GD, Tiwary HV. Thermal expansion of poly(vinylidene fluoride) films. J


Appl Phys Apr 1982:53(4);3040-3043
142

44 Singh R, Tandon RP, Chandra Sinha, Chandra Subhas. Low frequency AC


conduction and dielectric relaxation in PVDF films. Proc Ninth IEEE Int Symp
on Applications of Ferroelectrics Aug 1994; Univ Park, PA

45 White CC. A novel method to determine the mechanical properties of ultra-thin


films. Low Dielectric Constant Materials IV Symp Apr 1998:MRS; San
Francisco, CA; 177-182

46 Gao Z, Tsou AH. Mechanical properties of polymers containing fillers. J Polymer


Science: Part B: Polymer Physics 1999:37; 155-172

47 Broadhurst MG, Davis GT. Physical basis for piezoelectricity in PVDF.


Ferroelectric 1984:60;3-13

48 Furukawa T. Piezoelectricity and pyroelectricity in polymers. IEEE Trans Elec


Ins June 1989:24(3);375-394

49 Wada Y, Hayakawa R. Piezoelectricity and pyroelectricity of polymers. Jpn J.


Appl Phys Nov 1976:15(11);2041-2057

50 Date M, Fukada E, Wendorff JH. Nonlinear piezoelectriciy in oriented films of


PVDF and its copolymers. IEEE Trans Elec Ins June 1989:24(3);457-460

51 Nalwa HS . Ferroelectric polymers, chemistry, physics, and applications. 1995;


Marcel Dekker, Inc., New York

52 Brown LF, Scheinbeim JI, Newman BA. High frequency dielectric and
electromechanical properties of ferroelectric nylons. Proc Nonth Int Symp
Applications of Ferroelectrics Aug 1994; Univ Park, PA

53 Ibos L, et. al. Thermal behavior of ferroelectric polyamide 11 in relation to


pyroelectric properties. J Polymer Science: Part B: Polymer Physics 1999:37;715-
723

54 Green JS, Rabe JP, Rabolt JF. Studies of chain conformation above the curie
point in a vinylidene fluoride/trifluoroethylene random copolymer.
Macromolecules 1986:19;1725-1728

55 Ikeda S, Suzuki H. Correlation between ferroelectric phase transition and


polarization reversal phenomena observed through nonlinear dielectric anomaly in
ferroelectric fluorocopolymers. Polymer J 1992:24(11);1223-1227
143

56 Kim KJ, Kim GB, Vanlencia CL, Rabolt JF. Curie transition, ferroelectric crystal
structure, and ferroelectricity of a VDF/TrFE (73/25) copolymer 1. The effect of
the consecutive annealing in the ferroelectric state on curie transition and
ferroelectric crystal structure. J Polymer Science: Part B: Polymer Physics1994:
32;2435-2444

57 LeGrand JF. Structure and ferroelectric properties of P(VDF-TrFE) copolymers.


Ferroelectric 1989:91;301-317

58 Nalwa HS. Recent developments in ferroelectric polymers. JMS - Rev Macromol


Chem Phys 1991:C31(4);341-432

59 Pelrine RE, Kornbluh RD, Joseph JP. Electrostriction of polymer dielectrics with
compliant electrodes as a means of actuation. Sensors and Actuators
1998:A64;77-85

60 Lang S, I. Bibliography on Piezoelectricity and Pyroelectricity of Polymers 1961-


1980. Ferroelectrics 1981:32; pp191-245

61 Bloomfield PE. Production of ferroelectric oriented PVDF films. J Plastic Film


and Sheeting Apr 1988:4(2);123-129

62 Technical Manual. Published by Measurement Specialties, Inc; Norristown, PA.

63 Chandrasekharaiah DS, Debnath L. Continuum Mechanics. Academic Press


1995; 325-351

64 Schapery RA. Application of thermodynamics to thermomechanical, fracture,


birefringrent phenomena in viscoelastic media. J Appl Phys May 1964:35 (5):
1451 - 1465

65 Lynch CS, Chen W, Liu T. Multiaxial constitutive behavior of ferroelectric


materials. Smart structures and Materials: Behavior and Mechanics 2000; Proc
SPIE: 3992

66 Christensen RM. A nonlinear theory of viscoelasticity for application to


elastomers. Trans ASME Dec 1980:47; 762-768

67 Connor P, Jones JP, Llewellyn JP, Lewis TJ. Electric field induced viscoelastic
changes in insulating polymer films. Annual Report Conf on Electrical Insulation
and Dielectric Phenomenon Oct 1998;IEEE;Atlanta, GA;27-30

68 Castagnet S, Gacougnolle JL, Dang P. Correlation between macroscopical


viscoelastic behaviour and micromechanism in strained α polyvinylidene fluoride
(PVDF). Materials Science and Engineering 2000:A276;152-159
144

69 Dooling PJ, Buckley CP. The onset of nonlinear viscoelasticity in multiaxial


creep of glassy polymers: A constitutive model and its application to PMMA.
Polymer Engineering and Science Jun 1998:38(6);892-904

70 Dafalias YF. Constitutive model for large viscoelastic deformations of


elastomeric materials. Mechanics Research Comm 1991:18 (1); 61-66

71 Drozdov AD. A model for he nonlinear viscoelastic response in polymers at finite


strains. Int J Solids Structures 1998:35 (18); 2315-2347

72 Drozdov AD. A constitutive model in viscoelastoplasticity of glassy polymers.


Polymer 1999:40; 3711-3727

73 DeFilippo V, Saigal A, Zimmermann MA. Characterization and modeling of the


creep behavior of Xydar G930 Liquid Crystalline Polymer. Proc ImechE
Structural Analysis in Microelectronics and Fiber Optics (ASME), Atlanta, GA,
Nov 1996; 37- 44

74 Kaliske M, Rothert H. Formulation and implementation of three dimensional


viscoelasticity at small and finite strains. Computational Mechanics 1997:19;
228-239

75 Li JY, Dunn ML. Viscoelastic behavior of heterogeneous piezoelectric solids.


Active Materials: Behavior and Mechanics Mar 2000; Proceedings of SPIE
conference:3992; 331-340

76 Moginger B, Creep of polymers - a nonlinear viscoelastic extension of the time


dependent creep compliance and its comparison to experimental results. Intl Conf
on Mech of Time Dependent Materials Mar 1998;SEM; Pasadena, CA

77 Qian Z, Liu S. Unified constitutive modeling from viscoelasticity to


viscoplasticity of polymer matrix composites. Proc Amer Soc for Composites 12
Tech Conf Oct 1997; Dearborn, MI; 165-174

78 Reese S, Govindjee S. A theory of finite viscoelasticity and numerical aspects.


Int J Solids Structures 1998:35 (26/27); 3455-3482

79 Schapery RA. Nonlinear viscoelastic solids. Intl J Solids and Structures 2000:37;
359 - 366

80 Spathis G, Kontou E. An experimental and analytical study of the large strain


response of glassy polymers with a non contact laser extensometer. J Appl
Polymer Science Mar 1999:71 (12); 2007 - 2015
145

81 Syed Asif SA, Pethica JB. Nano scale viscoelastic properties of polymer
materials. Proc Mat Res Soc Symp 1998 Thin-Films - Stresses and Mechanical
Properties VII :505 103-108

82 Simon PP, Ploehn HJ, Investigating time-temperature superposition in crosslinked


polymers using the tube-junction model. J Polymer Science: Part B: Polymer
Physics 1999:37; 127 - 142

83 Vigny M, Aubert A, Hiver JM, Aboulfaral, G’sell C. Constitutive viscoplastic


behavior of amorphous PET during plane strain tensile stretching. Boucherville,
Qc, Sep 1998: 8 (11)32;358-384

84 Lakes R. Viscoelastic Solids. CRC Press 1999

85 Christensen RM. Theory of Viscoelasticity: An Introduction. Academic Press


1982

86 Böttcher CJF. Theory of electric polarisation. Elsevier Publishing Company.


1952

87 Biot MA. Theory of stress-strain relations in anisotropic viscoelasticity and


relaxation phenomena. J Appl Physics Nov 1954:25 (11); 1385-1391

88 Matsuoka S, Viscoelasticity and thermal analysis. J Thermal Analysis and


Calorimetry:59; 131-141, 2000

89 Le Huy M, Evrard G. Methodologies for lifetime predictions of rubber using


Arrhenius and WLF models. Die Angwandte Makromolekulare Chemie,
1998:261/262 (4624); 135-142

90 Golden JM, Graham GAC. Boundary value problems in linear viscoelasticity.


Springer-Verlag 1988

91 Kamlah M, Jiang Q. A model for PZT ceramics under uni-axial loadings. Report
No. FZKA-6211; Forchungszentrum Karlsruhe GmbH, Karlsruhe: Dec 1998

92 Padilla P, et. al. Constitutive relations for piezoelectric materials in terms of


invariants. Mechanics Research Communications Mar-Apr 2001:28 (2): 179-186

93 Lynch CS, McMeeking RM. Finite Strain Ferroelectric Constitutive Laws.


Ferroelectrics 1994: 160; 177-184

94 McCrum NG, Read BE, Williams G. Anelastic and dielectric effects in polymeric
solids. John Wiley & Sons Ltd. 1967
146

95 Capellades MA, et. al. Effect of the crystallinity on the dielectric properties of
polyvinylidene fluoride. Conf Rec IEEE Intl Symp on Electrical Insulation 1994;
Pittsburgh, PA; 554-558

96 Malecki J, Hilczer B. Dielectric behaviour of polymers and composites. Key


Engineering Materials 1994:92;181-216

97 Nakagawa K, Ishida Y. Dielectric relaxations and molecular motions in


poly(vinylidene fluoride) with crystal form II. J Polymer Science: Polymer
Physics Edition 1973:11;1503-1533

98 Takahashi T, Date M, Fukada E. Dielectric hysterysis and rotation of dipoles in


polyvinylidene fluoride. Appl. Phys Lett Nov 1980:37(9);791-793

99 Cowie JMG. Polymers: chemistry & physics of modern materials. Blackie


Academic & Professional; 1991

100 Abe Y, Kakizaki M, Hideshima T. Effect of the distribution of free volume on


the β relaxation in poly(vinylidene fluoride). Jpn J Appl Phys Aug
1984:24(8);1074-1077

101 Yoshio a, Kakizaki M, Hideshima T. Effect of the distribution of free volume on


the β relaxation in poly(vinylidene fluoride). Jap J Appl Phys Aug
1985:24(8);1074-1077

102 Hikita M, Nagao M, Sawa G, Ieda M. Dielectric breakdown and electrical


conduction of poly(vinylidene-fluoride) in high temperature region. J Phys D:
Appl Phys 1980:13;661-666

103 Lide DR, Frederikse HPR. CRC handbook of chemistry and physics. 74th Ed
1994: CRC Press

104 Rao SS. Mechanical Vibrations. Addison-Wesley 1990. Second Edition

105 Rezvani B, Linvill JG. Measurement of piezoelectric parameters vs bias field


strength in piezoelectric polyvinylidene fluoride (PVF2). Appl Phys Lett June
1979:34(12);828-830
147

APPENDIX A

Nomenclature and Symbols

The Einstein summation convention is used throughout except where noted.

Tensors of order one or higher are shown in bold.

dA Surface area element, magnitude dA, direction normal to dA


b Body Force
c Stiffness [N/m2]
C1, C2 WLF equation constants
d, e, g, h piezoelectric coefficients
D Electric displacement [Coulombs/m2]
E Electric Field [Volts/m]
f Frequency
f0 Frequency of interest, frequency at which measurements are known
i, j Indices, 1through 6
J Viscoelastic Creep Compliance (m2/N)
j Current density, ampere/m2
j Imaginary operator
kt Planar electromechanical coupling factor
k Boltzman constant (1.38 10-23 J(K)
K Dielectric constant, ε/ε0
L Length, m
m, n Cartesian axes for PVDF:
1 - stretch direction (in film plane),
2 - transverse to stretch direction (in film plane),
3 - normal to surface of the film
m Meter (length)
p Pyroelectric charge coefficient [Coulombs/(m2 · (C]
q Charge density, coulombs/m3
Q Charge, coulombs
s Compliance [m2/N]
S Entropy
S Strain [m/m]
T Stress [N/m2]
u Displacement
148

v Velocity
V Voltage (volts)
W Width, m
x Position
Yi Young’s modulus of elasticity (1/sij , i = j)

α Thermal coefficient of expansion, constant field condition, [m/m/(C]


α Arrhenius correction coefficient
β Arrhenius correction coefficient
/ Divergence symbol (Nabla operator)
∆ Displacement
ε Dielectric permittivity [Farad/m]
ε0 Dielectric permittivity of free space [8.85 E-12 Farad/m]
εr Low frequency dielectric permittivity
εu High frequency dielectric permittivity
γ Exponent in Cole-Cole equation
Λ Curve fitting constant for dielectric loss (1 sec3/4)
ω Radian frequency, 2πf (Hz)
φ Electric field potential, Volts
ρ Bulk material resistivity, (ohm ·m)
θ0 Reference temperature, (K
θ Temperature of interest
Θ Change in temperature [(K], θ - θ0
149

APPENDIX B

Ferroelectricity, Electrostatics, and Field Equations

B.1 Ferroelectricity

Piezoelectric materials having the following properties are generally classified as

ferroelectric [16]: a permanent dipole, the ability to switch the dipole orientation with

the application of a sufficiently high electric field, and the existence of what is known as

a Curie temperature above which the material enters a non-polar, non-ferroelectric,

paraelectric crystal phase. Ferroelectric materials tend to behave as linear dielectrics

under low electric fields and often have extremely high permittivities. The permittivity

of all ferroelectric materials is dependent upon frequency, field, elastic stress, and

temperature, although this dependence is generally more severe for polymeric

ferroelectrics.

Linear reversible dipole rotation under electric field occurs in all dielectric

materials, including ferroelectrics. However, when ferroelectric materials are subjected

to high fields the energy in the material is sufficient to overcome internal energy barriers

and cause permanent dipole rotation. This process is known as poling and the resulting

permanent dielectric displacement is a phenomenon unique to ferroelectric materials.

Poling tends to align the randomly oriented dipoles in an unpoled ferroelectric and thus

produces a net non zero polarization in the bulk material. Poling is the mechanism

through which ferroelectric materials become macroscopically piezoelectric.


150

As seen in Figure B.1.1, poling induced dipole reorientation results in a large

charge displacement which remains after the applied field returns to zero. Application of

a high field of the opposite polarity will result in dipole reversal and a large charge

displacement of the opposite polarity. When subjected to high AC fields, ferroelectric

materials exhibit a classical D/E hysteresis, as shown in Figure B.1.2. The hysteretic D/E

relationship is the most fundamental description of ferroelectric behavior, with two key

parameters: coercive field, Ec, and remnant polarization, Pr. Ec is defined as the field at

which the total displacement is zero and is approximately equal to the field at which half

of the dipole reversal has occurred. Pr is defined as the displaced charge remaining at

zero field.

It is important to note that linear piezoelectricity occurs only in the linear regions

shown in Figures B.1.1, 2. An unpoled ferroelectric material has a linear dielectric slope

in the low field region of the D/E curve, but since Pr = 0 for this condition, all of the

piezoelectric coefficients are  0, as they are intrinsically a function of Pr. In this work it

is assumed that all fields on piezoelectric materials remain below the point at which

polarization reversal occurs and in the resulting constitutive equations piezoelectricity is a

linear phenomena.
151

Figure B.1.1 Ferroelectric poling event, applied field cycled from 0 to


maximum, then back to 0.

Figure B.1.2 Displacement/Electric field curve for a ferroelectric material


showing typical switching hysteresis and linear dielectric regions.
152
153

B.2 Electrostatics

The elastic and dielectric properties of piezoelectric materials are coupled, as

discussed in chapter 3, and therefore a brief review of pertinent electrostatics is

appropriate.

Electric flux is defined to originate on positive charge and terminate on negative

charge such that one coulomb of charge produces one coulomb of electric flux [3].

Gauss’s law states that the total electric flux out of closed surface is equal to the total

charge on the surface:

D # dA  Q . (B.1)
3

D is clearly a vector quantity. Comparing (B.1) to Gauss’ law for electricity (B2) [17]

yields (B.3).

ε E # dA  Q , (B.2)
3

D  εE. (B.3)

Also, using (B.1) with charge density q = /·D, Gauss’ divergence theorem yields

( /# D ) d v  Q . (B.4)
2

φ, the electric potential between two points A and B, is a scalar quantity defined as the
154

work required to move a positive charge from point A to point B. E, the electric field

intensity (or simply the electric field), can be described as

E   /φ . (B.5)

From (B.5) E is a vector quantity, revealing permittivity ε to be a second order

tensor from (B.3). This property can be simply described as the propensity of a material

to displace electric charge in response to an applied electric. In a linear dielectric the

charge displacement is a linear function of the applied field, and all discussions in this

work will assume this behavior - see Figure B.2.1.

All real (non ideal) dielectric materials have energy loss mechanisms associated

with dipole rotation and charge displacement in a time varying electric field. It is

therefore convenient to represent the permittivity tensor in complex notation, εij’ - jεij”,

where εij” represents the loss term and εij’ the reversible storage term, in a manner

analogous to elastic compliance. Likewise, real dielectric materials are not perfect

insulators and therefore suffer DC conduction losses. These dielectric loss components

can be ignored in many applications when using low loss ceramic materials, but become

significant in some lossy ceramics and in polymers. This is especially true at elevated

temperatures.

B.3 Field Equations

The following field equations apply to all materials regardless of their internal

structure [63]. The equations are based upon the laws of conservation of mass, the
155

Figure B.2.1 Displaced charge, Q, resulting from applied voltage, V, in a


linear dielectric material. Parallel plate capacitor shown.

balance of linear and angular momentum, and the balance of energy.

0ρ Continuity (B.6)
 div( ρ v )  0
0t

div T  ρ b  ρ ü Motion (B.7)

T  TT Motion (superscript is transpose) (B.8)

Other equations also valid for all materials:

1
S  ( / u  / u T)
2
Strain-displacement (B.9)
156

curl curl S  0 Compatibility (B.10)

Although these equations represent fundamental laws holding for all points in any

media, there are more unknown field functions than there are equations. Constitutive

equations are therefore necessary in order to fully determine the state of the system.
157

APPENDIX C

Measurement Equipment

1. Dynamic Signal Analyser


Manufacturer: Hewlett Packard
Model: 3562A DSA
Frequency Range: 64 )Hz to 100 kHz
Source output distortion: -60 dB max
CMRR: 80 dB
Channel cross talk: -140 dB
Voltage accuracy: ±0.15dB ± 0.015% input range
Noise (20 to 1kHz): -134 dBV/Hz (0.20 )V Hz)

2. Load Cell
Manufacturer: Entran, Inc. (Fairfield, NJ)
Model: ELF-T3-20L
Load range 20.0 pounds force (~90N)
Temperature Drift: ± 2.5 mV/100(F
Sensitivity: 12.5 mV pound
Amplifier: Entran PS-30A
Gain: 20.0 (250 mV/pound, 56.2 mV/N)

3. Displacement Sensor
Manufacturer: Magnetic Moments, LLC (Goleta, CA)
Model: ECS998s
Sensitivity (in range of interest): 3.288 V/mm

4. Power Amplifier
Manufacturer: Bruel & Kjaer
Model: 2712
Frequency response: DC to 100 kHz.
Output: 180VA max
Distortion: -80 dB

5. Digital Multimeter (DMM)


Manufacturer: Hewlett Packard
Model: 8842A
Resolution: 5 ½ digits
158

APPENDIX D

Derivation of Standard Linear Material Behavior


as a Function of Frequency

This appendix shows the derivation of the complex modulus of the Standard

Linear Material (SLM) and shown in Figure 3.2.1.

k2

k1 F(t)
u(t)
Figure D.1 Standard Linear Material model showing two springs
and a dashpot.

The approach used is based on the kind of first order impedance calculations

commonly encountered in electrical circuit analysis. Using a simple definition of

mechanical impedance [104],

F (t)  u (t) Z M(t) , (D.1)


159

where F(t) is an applied force, u(t) is displacement, and ZM(t) is the mechanical

impedance. Looking at the Maxwell component of the SLM model, the displacement can

be written as the sum of the displacements from the spring and the dashpot. Here the

dampening coefficient is shown as ζ, the equivalent spring constant of the material as k1,

and j is the complex operator.

F (t) 1
u (t)  u spring (t)  udash (t)   F (t) dt . (D.2)
k1 ζ2

Letting F(t) = Fo e jωt,

F0 e jωt F0 k1  jωζ
u (t)   e jωt jω dt  F0 e jωt [ ]. (D.3)
k1 jωζ 2 jωζk1

From (D.1), assuming that u(t) has the form Uo e jωt,

2
jωζk1 (jωζk1)(k1  jωζ) k1 ω2 ζ2  jωk1 ζ (D.4)
ZMAX   
k1  jωζ 2
k1  ω2ζ2
2
k1  ω2ζ2

Substituting a constant, τ, leads to the final Maxwell impedance,

ζ
τ (D.5)
k1

k1 ω2 τ2 k1 ω τ
Z MAX  j .
1  ω2τ2 1  ω2τ2 (D.6)
160

The SLM has another spring element mechanically in parallel with the Maxwell

element. Letting k2 be the spring constant of this second spring, the SLM impedance can

be written as

k1 ω2 τ2 k1 ω τ
ZS L M  ( k2  )  j . (D.7)
1  ω2τ2 1  ω2τ2

The final conversion to the form shown in equations (3.10) and (3.11) can be

made by realizing that the spring constant of a sample is a function of material properties

and geometry,

Y AC kL
k , Y , (D.8)
L AC

where AC is the cross sectional area and L is the sample length. Y’ is the real part

of (D.7), and Y” is the imaginary part.

The loss tangent it defined as Y”/Y’. The location of the peak of this loss term is

not generally coincident with the peak of the Y” curve, which occurs at ω = τ. The peak

loss frequency can be determined by setting the derivative of the loss tangent equal to 0

and solving for a non trivial value of ω, which yields:

1 k2
ωloss peak  ( )1/2 . (D.9)
τ k2  k1
161

APPENDIX E

Additional TTS Master Curve Data

Figures E.1 (a-d) through E.5 (a-d) show expanded TTS curves for 1 axis

samples tested in this work. Figures E.6 (a-d) and E.7 (a-d) show expanded data for 2

axis samples. This data is presented in addition to the data shown in chapter 6, and

represents data taken from a larger population of samples.

Each plot contains a curve fit obtained through the use of the WLF TTS equations

discussed in chapter 6. All fit curves are produced using the identical WLF constants.
162

M a s t e r C u r v e C o m p lia n c e v s Fr e q 0 6 2 2 - 2

1 .0

Co mp lian ce (G p a^-1) 0 .8

0 .6

0 .4

0 .2

0 .0
-1 0 -5 0 5 10
Lo g F r e q u e n c y ( H z )

Figure E.1a TTS expansion of 1 axis elastic compliance, 50( C.

M aster Curve Elas tic Loss vs Freq 0 622 -2


7%

6%
% L os s Com plian ce

5%

4%

3%

2%

1%

0%
-10 -5 0 5 10

Log Fr e que ncy (Hz )

Figure E.1b TTS expansion of 1 axis elastic loss, 50( C.


163

M a s te r C u r ve d 3 1 vs Fr e q 0 6 2 2 -2

30

25

20
d 31 p C/N

15

10

0
-10 -5 0 5 10

L o g Fr e q u e n c y ( Hz )

Figure E.1c TTS expansion of 1 axis piezo compliance, 50( C.

M a s te r Cur ve P ie zo Los s vs Fre q 0 6 2 2 -2

7%

6%
% L os s Co m plian ce

5%

4%

3%

2%

1%

0%
- 10 -5 0 5 10

L o g Fr e q u e n cy ( Hz )

Figure E.1d TTS expansion of 1 axis piezo loss, 50( C.


164

Master Curve Compliance vs Freq 0622-4

1.0
Com plia nce (Gpa ^ -1) 0.8

0.6

0.4

0.2

0.0
-10 -5 0 5 10
Log Frequency (Hz)

Figure E.2a TTS expansion of 1 axis elastic compliance, 50( C.

M a ste r Cu rve Ela stic L o ss vs F re q 0622-4

7%

6%
% Ela stic Loss

5%

4%

3%

2%

1%

0%
-10 -5 0 5 10

L o g F re q u e n cy (Hz )

Figure E.2b TTS expansion of 1axis elastic loss, 50( C.


165

M a ste r Curve d31 vs Fre q 0622-4

25

20
d31 pC/N

15

10

0
-10 -5 0 5 10

Log Fre que ncy (Hz )

Figure E.2c TTS expansion of 1 axis piezo compliance, 50( C.

M a ste r Cu rve P ie z o Lo ss vs Fre q 0622-4


7%
6%
% Pie z o Loss

5%
4%
3%
2%
1%
0%
-10 -5 0 5 10

L og Fre q ue ncy (Hz )

Figure E.2d TTS expansion of 1 axis piezo loss, 50( C.


166

M a ste r Curve Com plia nce vs Fre q 1205-3B

Co m p lian ce (Gp a^-1) 0.8

0.6

0.4

0.2

0
-10 -5 0 5 10

Lo g F re q u e n cy (Hz)

Figure E.3a TTS expansion of 1 axis elastic compliance, 50( C.

M a ste r C u rve Ela stic L o ss vs F re q 1205-3B

7%

6%

5%
% Elas tic Lo s s

4%

3%

2%

1%

0%
-10 -5 0 5 10

L o g F re q u e n cy (H z )

Figure E.3b TTS expansion of 1 axis elastic loss, 50( C.


167

M a s te r C u r ve d 3 1 vs Fr e q 1 2 0 5 -3 B

25

20

15
d 31 p C/N

10

0
-10 -5 0 5 10

L o g F r e q u e n c y (H z )

Figure E.3c TTS expansion of 1 axis piezo compliance, 50( C.

M a ste r C u rve P ie z o L o ss vs F re q 1205-3B

7%

6%

5%
% Pie z o L os s

4%

3%

2%

1%

0%
- 10 -5 0 5 10

L o g F re q u e n cy (H z )

Figure E.3d TTS expansion of 1 axis piezo loss , 50( C.


168

M a ste r Cu rve Co m plia n ce vs Fre q 1205-4

1.0

Com p lian ce (Gp a^-1) 0.8

0.6

0.4

0.2

0.0
-10 -5 0 5 10

L og F re q ue ncy (Hz )

Figure E.4a TTS expansion of 1 axis elastic compliance, 50( C.

M a ste r C u rve Ela stic L o ss vs F re q 1205-4

7%

6%

5%
% Elas tic Los s

4%

3%

2%

1%

0%
- 10 -5 0 5 10

L o g F re q u e n cy (H z )

Figure E.4b TTS expansion of 1 axis elastic loss , 50( C.


169

M a s te r C u r ve d 3 1 vs Fr e q 1 2 0 5 -4

25

20

15
d 31 p C/N

10

0
-10 -5 0 5 10

L o g F re q u e n c y (H z )

Figure E.4c TTS expansion of 1 axis piezo compliance, 50( C.

M a ste r C u rve P ie z o L o ss vs F re q 1205-4

7%

6%

5%
% p ie z o Los s

4%

3%

2%

1%

0%
-10 -5 0 5 10

L o g F re q u e n cy (H z )

Figure E.4d TTS expansion of 1 axis piezo loss, 50( C.


170

M a ste r Cu rve Com p lia nce vs F re q 1205-5

1.0

Com p lian ce (Gp a^-1) 0.8

0.6

0.4

0.2

0.0
-10 -5 0 5 10

L og F re q ue ncy (Hz )

Figure E.5a TTS expansion of 1 axis elastic compliance, 50( C.

M a ste r Cu rve Ela stic L o ss vs F re q 1205-5


7%

6%

5%
% Elas tic L os s

4%

3%

2%

1%

0%
-6 -4 -2 0 2 4 6

L o g F re q u e n cy (Hz )

Figure E.5b TTS expansion of 1 axis elastic loss, 50( C.


171

M a s te r C u r ve d 3 1 vs Fr e q 1 2 0 5 -5

25

20

15
d31 pC/N

10

0
-10 -5 0 5 10

L o g F re q u e n c y (H z )

Figure E.5c TTS expansion of 1 axis piezo compliance, 50( C.

M a ste r C u rve P ie z o L o ss vs F re q 1205-5


7%

6%

5%
% Pie z o Los s

4%

3%

2%

1%

0%
-6 -4 -2 0 2 4 6

L o g F re q u e n cy (H z )

Figure E.5d TTS expansion of 1 axis piezo loss, 50( C.


172

M a ste r Cu rve Com plia n ce vs F re q 1206-4


1.0

Co m plian ce (Gpa^-1) 0.8

0.6

0.4

0.2

0.0
-10 -5 0 5 10

L og F re q ue ncy (Hz )

Figure E.6a TTS expansion of 2 axis elastic compliance, 50( C.

M a ste r C u rve Ela stic L o ss vs F re q 1206-4

7%

6%

5%
% Elas tic L os s

4%

3%

2%

1%

0%
- 10 -5 0 5 10

L o g Fr e q u e n cy (Hz )

Figure E.6b TTS expansion of 2 axis elastic loss, 50( C.


173

M a s te r C u r ve d 3 2 vs Fr e q 1 2 0 6 -4
2 .4
2 .2
2 .0
1 .8
1 .6
d32 pC/N

1 .4
1 .2
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-10 -5 0 5 10

L o g Fr e q u e n c y ( Hz )

Figure E.6c TTS expansion of 2 axis piezo compliance, 50( C.

M a ste r C u rve P ie z o L o ss vs F re q 1206-4

5%

0%
% Pie z o Lo s s

- 5%

-10%

-15%

-20%
- 10 -5 0 5 10

L o g Fr e q u e n cy ( Hz )

Figure E.6d TTS expansion of 2 axis piezo loss, 50( C.


174

M a ste r Cu rve Com plia n ce vs F re q 1206-6


1.2

1.0
Com p lian ce (Gp a^-1)

0.8

0.6

0.4

0.2

0.0
-10 -5 0 5 10

L og F re q ue ncy (Hz )

Figure E.7a TTS expansion of 2 axis elastic compliance, 50( C.

M a ste r C u rve Ela stic L o ss vs F re q 1206-6


7%

6%

5%
% Elas tic Los s

4%

3%

2%

1%

0%
- 10 -5 0 5 10

L o g F re q u e n cy (H z )

Figure E.7b TTS expansion of 2 axis elastic loss, 50( C.


175

M a s te r C u r ve d 3 2 vs Fr e q 1 2 0 6 -6
2 .4
2 .2
2
1 .8
1 .6
d31 pC/N

1 .4
1 .2
1
0 .8
0 .6
0 .4
0 .2
0
-10 -5 0 5 10

L o g F r e q u e n c y (H z )

Figure E.7c TTS expansion of 2 axis piezo compliance, 50( C.

M a ste r C u rve P ie z o L o ss vs F re q 1206-6

5%

0%

- 5%
% Pie z o L os s

-10%

-15%

-20%

-25%

-30%
- 10 -5 0 5 10

L o g F re q u e n cy (Hz )

Figure E.7d TTS expansion of 2 axis piezo loss, 50( C.


176

Vita

Mitchell L. Thompson
Born: 11/13/54, Pensacola, Florida
Citizenship: United States of America

EDUCATION
Ph.D. 2002 Drexel University, Philadelphia, Pennsylvania
Mechanical Engineering

M.S. 1983 The Pennsylvania State University, State College,


Pennsylvania
Mechanical Engineering

B.S. 1977 The Pennsylvania State University, State College,


Pennsylvania
Meteorology

PROFESSIONAL EXPERIENCE
1998 - 2002 Measurement Specialties, Inc., Norristown, PA
Director, Development Engineering

1993 - 1998 AMP Incorporated, Norristown, PA


Manager, Product Engineering
Manager, Materials Engineering

1989 - 1993 Elf Atochem, NA, Norristown, PA


Program Manager, Product Development
Manager, Ferroelectric Materials Development

1986 - 1989 Pennwalt Corporation, King of Prussia, PA


Manager, Manufacturing Engineering and Production

1983 - 1986 IBM Corporation, East Fishkill , NY


Senior Engineer, Advanced Optical Photolithography
Equipment Engineer, Resist Systems Engineering.

1977 - 1981 Officer, United States Navy, Pacific Fleet.

PATENTS: Four issued

Das könnte Ihnen auch gefallen