electrical conduction through polyvinylidene fluoride
TRANSCRIPT
The Pennsylvania State University
The Graduate School
ELECTRICAL CONDUCTION THROUGH POLYVINYLIDENE
FLUORIDE: EXPLOITING THE INTERFACE AS A BARRIER TO
CHARGE TRANSPORT
A Dissertation in
Materials Science and Engineering
by
Michael Anthony Vecchio
© 2019 Michael Anthony Vecchio
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
August 2019
ii
The dissertation of Michael Anthony Vecchio was reviewed and approved* by the following:
Zoubeida Ounaies
Professor of Mechanical and Nuclear Engineering
ALP Fellow, Big Ten Academic Alliance
The International Society for Optics and Electronics Senior Member
Dissertation Co-Advisor
Co-Chair of Committee
Michael T. Lanagan
Professor of Engineering Science and Mechanics
Dissertation Co-Advisor
Co-Chair of Committee
Michael Hickner
Professor of Materials Science and Engineering, Chemical Engineering
Corning Faculty Fellow
Ramakrishnan Rajagopalan
Assistant Professor of Engineering in Applied Materials
John Mauro
Professor of Materials Science and Engineering
Associate Head for Graduate Education
Chair, Intercollege Graduate Degree Program
*Signatures are on file in the Graduate School
iii
ABSTRACT
Polymer capacitors exhibit a combination of unique properties including high dielectric
breakdown strength, light weight, flexibility, and low-cost production that make them appealing
candidates for film capacitor technologies. For high power capacitor applications, biaxially
oriented polypropylene (BOPP) is considered state of the art and exhibiting breakdown strengths
as high as 850 MV/m, however its low permittivity (εr = 2.2) prevents its use in high-energy density
demanding applications. In this dissertation, high permittivity (8 < εr < 12) polar polymers
poly(vinylidene fluoride) (PVDF) and its copolymer poly(vinylidene fluoride trifluoroethylene)
(P(VDF-TrFE)) are used as model materials to investigate the role of interfaces on low frequency,
high temperature and high electric field charge transport. This work demonstrates 1) the
importance of electrode/dielectric interface chemistry in controlling charge injection and
conduction, and 2) how interfaces in layered dielectrics block transport of impurity ions ultimately
influence dielectric performance.
The high field performance of hot-pressed layered dielectrics in pure PVDF laminates was
explored first. The solution processing and hot-press lamination procedure produced films
containing both alpha and beta crystal phases throughout the bulk. Impedance spectroscopy at
70oC combined with equivalent circuit (EC) modeling demonstrated a blocking effect in PVDF
films containing 4 layers relative to the 1-layer control. Finally, dielectric breakdown experiments,
analyzed via Weibull statistics, reveal a statistically significant 16% increase in the breakdown
strength of 3-layer films (490 MV/m) relative to 1-layer (415 MV/m) analyzed using a 90%
confidence interval. These initial results imply that multilayer lamination low frequency charge
migration which leads to higher dielectric breakdown strength, however hot-pressing proved to be
a disadvantageous processing procedure: layer counts beyond 4 were not possible due poor
repeatability and individual layer thickness is ~10μm.
A spin cast process was developed as an alternative to hot pressing for creating
reproducible P(VDF-TrFE) thin films. (~1μm thick). Electrode/dielectric interfacial chemistry on
high-field conduction was studied by controlling P(VDF-TrFE) surface chemistry using a CF4/O2
reactive ion plasma treatment. It was found that oxygen based chemical moieties detected using
X-ray photoelectron spectroscopy (XPS) grafted to the film surface cause Schottky barrier height
iv
lowering by approximately 0.05 eV. This reduction accounted for an order of magnitude increase
in leakage current at high fields.
The effect of a pure oxygen plasma treatment was then assessed in 10μm thick polyimide
(PI) films. PI exhibits a non-polar chemical structure and surface chemistry modification after
plasma treatment had a different effect on high-field conduction relative to P(VDF-TrFE). A
combination of hopping, Poole-Frenkle, and Schottky theoretical frameworks were used to analyze
charging current as a function of voltage and temperature in untreated and oxygen plasma treated
PI. It was found that oxygen moieties introduced via plasma treatment caused both decrease in the
leakage current of PI films at high temperature and delayed transition from bulk dominated
conduction (hopping) to interface dominated conduction (Schottky) by 50oC relative to untreated
PI. It is posited that the presence of electronic trapping centers introduced by chemical
modification at the electrode/dielectric interface are responsible for electronic charge scattering
and trapping at high temperatures.
The influence of interfaces on ionic transport in P(VDF-TrFE) multilayer films was
explored. Lithium perchlorate (LiClO4) is first doped into 1-layer P(VDF-TrFE), creating
dielectrics in which the impurity ion species is controlled and well known. Differential scanning
calorimetry (DSC) revealed a correlation between curie transition temperature of copolymer’s beta
phase and LiClO4 concentration added into the material. EC modeling of impedance spectra as a
function of temperature captured Li+ ion interaction with crystalline phases distributed throughout
the bulk as well as the electrode dielectric interface at low frequency and high temperature. It was
found that the impedance of crystalline interfaces and the electrode/dielectric interface is a major
contributor to the overall electrical response of the film, indicating that ions are blocked at
interfaces.
Finally, multilayered composites were created with alternating doped P(VDF-TrFE) and
thin (500nm) poly(vinyl alcohol) (PVA) layers used to develop an ion depleted interface. The EC
model used to describe 1-layer films of P(VDF-TrFE) was used to develop a model that predicts
the impedance behavior of doped layered composites. Impedance spectroscopy, EC modeling, and
TSDC are used to prove the occurrence of impurity ion blocking at the interface leading to
substantial space charge distribution through the bulk of the composites at low frequencies and
high fields. Finally, high voltage dielectric breakdown experiments were performed, and the defect
v
dominated breakdown mechanism showed most significant effect due to layering and is described
well using Weibull statistics.
vi
TABLE OF CONTENTS
LIST OF FIGURES ………………………………………………………………………………x
LIST OF TABLES ………………………………………………………………………………xv
ACKNOWLEDGEMENTS …………………………………………………………………….xvi
CHAPTER 1: INTRODUCTION ………………………………………………………………..1
1.1 HISTORY OF CAPACITORS ………….………………………………………………...1
1.2 CAPACITORS AND DIELECTRIC MATERIALS ……………………………………..2
1.2.1 Electrochemical Capacitors ……………………………………………….………...3
1.2.2 Ceramic Dielectrics …………………………………………………………………5
1.2.3 Polymer Dielectrics …………………………………………………………………7
1.3 PVDF AS A DIELECTRIC MATERIAL ……………………………………….………10
1.3.1 Material Structure ………………………………………………………….………10
1.3.2 Limitations of PVDF for Energy Storage ………………………………….………11
1.4 INTERFACES FOR IMPROVED DIELECTRIC PERFORMANCE ………………….15
1.4.1 The Electrode/Dielectric Interface ………………………………………………...15
1.4.2 The Bulk Distributed Interface ……………………………………………….……16
1.5 PROBLEM STATEMENT AND RESEARCH GOAL ………………………….……...19
1.6 ORGANIZATION OF DISSERTATION ……………………………………….………21
CHAPTER 2: MATERIAL PROCESSING, CHARACTERIZATION AND DATA
PROCESSING TECHNIQUES …………………………………………………………………23
2.1 INTRODUCTION ………………………………………………………………….……23
2.2 PVDF PROCESSING AND SAMPLE PREPERATION ………………………….……24
2.2.1 PVDF Casting and Hot-Press Lamination …………………………………….…...24
2.2.2 P(VDF-TrFE) Spin Casting …………………………………………………….….26
2.2.3 Electrode Deposition ………………………………………………………………27
2.2.4 Plasma Treatment ………………………………………………………………….29
2.3 MATERIAL CHARACTERIZATION EQUIPMENT AND METHODS ……………...32
2.3.1 Bulk Chemical Characterization …………………………………………………...32
2.3.1.1 Differential Scanning Calorimetry (DSC) …………………………….…32
2.3.1.2 Fourier Transform Infrared Spectroscopy (FTIR) …………………….....33
2.3.2 Surface Chemical Characterization …………………………………………….…..36
2.3.2.1 Optical Profilometry ………………………………………………….….36
2.3.2.2 H2O Contact Angle ……………………………………………………....36
2.3.2.3 X-ray Photoelectron Spectroscopy (XPS) ………………………………..38
2.3.2.4 Time of Flight Secondary Ion Mass Spectrometry (ToF-SIMS) ………....40
2.3.3 Electrical Characterization ………………………………………………………...41
2.3.3.2 Impedance Spectroscopy and Equivalent Circuit Modeling …………….41
2.3.3.2 Current Voltage (I(V)) Charging Experiments ……………………….….44
2.3.3.3 Thermally Stimulated Depolarization Current Measurement (TSDC) …..49
2.3.3.4 High Voltage Dielectric Breakdown Measurements and Weibull
Statistics………………………………………………………………………….52
2.4 SPECIAL ANALYTICAL TECHNIQUES ………………………………………….….54
2.4.1 Bootstrap Statistics Applied to I(V) Data …………………………………….……54
vii
2.4.2 Equivalent Circuit (EC) Modeling …………………………………………….…...57
2.4.2.1 Equivalent Circuit Theory ……………………………………………….58
2.4.2.2 Statistical Interpretation of Fit Parameters ………………………….…..61
2.4.3 TSDC Peak Deconvolution ………………………………………………………...62
CHAPTER 3: HOT-PRESSED PVDF LAMINATES ………………………………….………64
3.1 INTRODUCTION ………………………………………………………………….……64
3.2 MATERIALS AND METHODS ………………………………………………………..65
3.2.1 Materials Selection ………………………………………………………………...65
3.2.2 Multilayer Laminate Fabrication ……………………………………………….….65
3.2.3 Structural Characterization …………………………………………………….......66
3.2.4 Dielectric Measurement …………………………………………………………....66
3.3 RESULTS …………………………………………………………………………….….67
3.3.1 SEM Imaging the Interface …………………………………………………….…..67
3.3.2 DSC PVDF Thermal Analysis ………………………………………………….….67
3.3.3 FTIR PVDF Crystal Structure Analysis ……………………………………….…..68
3.3.4 Dielectric Breakdown ………………………………………………………….…..69
3.3.5 Impedance Spectroscopy …………………………………………………………..71
3.4 CONCLUSIONS …………………………………………………………………….…..73
CHAPTER 4: PLASMA SURFACE MODIFICATION OF P(VDF-TrFE): INFLUENCE OF
SURFACE CHEMISTRY AND STRUCTURE ON ELECTRONIC CHARGE INJECTION ...75
4.1 INTRODUCTION ...………………………………………………………………….….76
4.2 EXPERIMENTAL SECTION ……………………………………………………….…..78
4.2.1 Materials …………………………………………………………………………...78
4.2.2 Thin Film Fabrication ……………………………………………………………...78
4.2.3 Plasma Surface Modification ……………………………………………………....78
4.2.4 Investigated Processing Conditions ………………………………………………..79
4.2.5 Characterization Techniques ………………………………………………………79
4.2.5.1 Chemical and Structural ………………………………………………....79
4.2.5.2 Electrical ………………………………………………………………...80
4.3 RESULTS AND DISCUSSION ………………………………………………………....81
4.3.1 Differential Scanning Calorimetry ………………………………………………...81
4.3.2 X-ray Photoelectron Spectroscopy: Surface Chemistry and Structure Analysis …..82
4.3.3 ToF-SIMS Depth Profiling ………………………………………………………...84
4.3.4 Surface Roughness and Contact Angle Analysis …………………………………..85
4.3.5 Low Field Measurements ………………………………………………………......88
4.3.5.1 Dielectric Spectroscopy …………………………………………….……88
4.3.5.2 Ohmic Current – Voltage (I-V) Experiments ……………………….……90
4.3.6 High Field Current Density – Electric Field (J-E) Measurements …………………91
4.3.6.1 Conduction Mechanism Identification: Schottky and Poole-Frenkel
Modeling ………………………………………………………………………....91
4.3.6.1 Quantifying the Change in Barrier Height ………………………………94
4.4 CONCLUSIONS …………………………………………………………………….…..97
viii
CHAPTER 5: CONFUCTION THROUGH PLASMA TREATED POLYIMIDE: ANALYSIS
OF HIGH FIELD CONDUCTION BY HOPPING AND SCHOTTKY THEORY ………….....99
5.1 INTRODUCTION ……………………………………………………………………...100
5.2 ANALYTICAL METHODS …………………………………………………………...102
5.2.1 Bulk Dominated Conduction ………………………………………………….….103
5.2.2 Interface Dominated Conduction ………………………………………….……...104
5.3 RESULTS AND DISCUSSION ………………………………………………….…….104
5.3.1 Leakage Current Results …………………………………………………….……104
5.3.2 Data Analysis ……………………………………………………………….…….106
5.3.2.1 Modeling Bulk Conduction ………………………………….………….106
5.3.2.2 Interface Dominated Conduction ………………………………………112
5.4 DISCUSSION AND CONCLUSIONS ………………………………………….……..115
CHAPTER 6I: ANALYSIS OF LOW FIELD IMPURITY ION MIGRATION IN LiClO4
DOPED P(VDF-TrFE) THIN FILMS ………………………………….………………………117
6I.1 INTRODUCTION …………………………………………………………………….119
6I.2 MATERIALS AND METHODS …………………………………………….………..120
6I.2.1 Materials ………………………………………………………………….……...120
6I.2.2 Thin Film Fabrication …………………………………………………….……...121
6I.2.3 Structural Characterization ……………………………………………….……...121
6I.2.4 Electrical Measurements ………………………………………………………...121
6I.3 RESULTS AND DISCUSSION ………………………………………………………122
6I.3.1 Differential Scanning Calorimetry (DSC) ………………………………………122
6I.3.2 Dielectric Spectroscopy …………………………………………………………123
6I.3.3 AC Conductivity ………………………………………………………………...125
6I.3.4 Impedance Spectroscopy ………………………………………………….……..126
6I.4 EQUIVALENT CIRCUIT MODELING …………………………………….………..130
6I.4.1 Modeling the Capacitive Response …………………………………….………..132
6I.4.1.1 Electronic Polarization ………………………………………………..132
6I.4.1.2 Permanent Dipole Orientational Polarization ………….……………..132
6I.4.1.3 Ionic / Space Charge Conduction ………………………….…………..134
6I.4.1.4 Blocking Polarization ……………………………….…………………135
6I.4.2 Modeling the Resistive Response ……………………………………….……….137
6I.5 CONCLUSIONS …………………………………………………………….………...139
CHAPTER 6II: CONDUCTION IN MULTILAYERED LAMINATES: EXPLOITING THE
INTERFACE AS A BARRIER TO CHARGE TRANSPORT ………………….……………..141
6II.1 INTRODUCTION ……………………………………………………………………141
6II.2 MATERIALS AND METHODS ………………………………………….………….142
6II.2.1 Materials ………………………………………………………………………..142
6II.2.2 Multilayer Processing …………………………………………………………..143
6II.3 RESUTS AND DISCUSSION ………………………………………….…………….144
6II.3.1 Differential Scanning Calorimetry …………………………………….………..144
6II.3.2 Dielectric and Impedance Spectroscopy ……………………………….……….145
6II.3.3 Equivalent Circuit Modeling ………………………………………….………...148
6II.3.4 Thermally Stimulated Depolarization Current Measurements (TSDC) .………..151
ix
6II.3.4.1 TSDC on 1-layer P(VDF-TrFE) ……………………………………...151
6II.3.4.2 TSDC on 4-layer P(VDF-TrFE) ……………………………………...155
6II.3.5 High Voltage Dielectric Breakdown ………………………….………………...158
6II.4 CONCLUSIONS ……………………………………………………………………..160
CHAPTER 7: CONCLUSIONS AND FUTURE WORK …………………………….………162
7.1 CONCLUSIONS ……………………………………………………………….………162
7.2 SIGNIFICANT CONTRIBUTIONS …………………………………….……………..166
7.3 FUTURE WORK ………………………………………………………………………169
7.3.1 Tailoring the Electrode/Dielectric Interface for Controlled Current Injection …...169
7.3.2 Multilayer Dielectric Processing ………………………………………….………169
7.3.3 Composite Characterization Using High Voltage Techniques ………….………..171
APPENDIX A: Annotated Code for I(V) Nonlinear Regression ……………..……….………174
APPENDIX B: Annotated Code for TSDC Peak Fitting ………………………………………177
APPENDIX C: P(VDF-TrFE) Poole-Frenkel Analysis …………….…………………………180
APPENDIX D: Polyimide I(V) Nonlinear Regression Parameter Estimates ……………….…182
APPENDIX E: Equivalent Circuit Estimate Error Reports ……………………………………185
Bibliography…………………………………………………………………………………...187
x
LIST OF FIGURES
Figure 1-1: Ragone plot comparing the specific power and energy in common types of energy and
power storage devices. (figure credit [1]).
Figure 1-2: The charging and discharge cycle depicted for a typical electrochemical capacitor
(image credit [2]).
Figure 1-3: The dielectric constant of BaTiO3 as a function of temperature. Crystal phase
transitions are shown as a function of temperature above, corresponding to maxima in the
material’s dielectric constant (image credit [3]).
Figure 1-4: Breakdown strength of polymers, composites (nanodielectrics) and ceramics
organized by their dielectric constant (credit Tan et al. MatSc&App 2013).
Figure 1-5: a) impact of BaTiO3 nanoparticle inclusion and nanoparticle chain alignment on Ebd
in silicone composites [4] and b) impact of Al nanoparticle inclusion on Ebd in PP [5]. DC dielectric
breakdown data analyzed via 2 parameter Weibull statistics.
Figure 1-6: a) polymer chains of PVDF showing TTTT and TGTG¯ conformations associated
with the β- and α-phases respectively (image credit [6]), b) α-phase crystal structure and c) β-phase
crystal structure (image credit [7]).
Figure 1-7: Ragone plot for an ideal battery depicting the energy/power relation with and without
contributions from leakage RL (image credit [8]).
Figure 1-8: a) displacement current D vs electric field E loop depicting normal ferroelectric
behavior with purple shaded region representing recoverable energy density (image credit [9]) and
b) D-E behavior of biaxially oriented PVDF at 10Hz showing strong ferroelectricity (imag credit
[10]).
Figure 1-9: a) dielectric breakdown strength of P(VDF-TrFE-CTFE) as a function of electrode
material and b) high field leakage current as a function of electrode material. Major differences
are observed for Ag vs Al, reflecting significance of electrode/dielectric interface on controlling
high field conduction [11].
Figure 1-10: BOPVDF TSDC for a) Ag electrodes and b) Al electrodes. Both samples are
measured under the same conditions: Ep = 10 MV/m, Tp =50oC, heating rate = 5oC/min, and tp
ranges from 10 min – 50 min [12].
Figure 1-11: Leakage current as a function of electric field for untreated and plasma treated PI at
150oC. Both reduction of data scatter and magnitude of leakage current occurs after plasma
treatment [13].
Figure 1-12: a) SEM images of 32-layer and 256-layer P(VDF-HFP) / PC microlayer coextruded
multilayer laminate along with b) effect of composite structure on breakdown strength. The 32-
layer laminate from a) results in a composite dielectric breakdown strength surpassing P(VDF-
HFP) and PC alone [14].
Figure 1-13: a) schematic showing PC / P(VDF-HFP) coextruded composite (similar to Mackey
et al.) with PMMA tie layer incorporated. b) dielectric breakdown strength of multilayer composite
using multiple tie layer materials and c) dielectric loss tangent showing frequency reduction in
ionic relaxation for the PC / PMMA tie layer / P(VDF-HFP). Decrease in relaxation frequency is
related to decreased ionic mobility in the PMMA/P(VDF-HFP) interphase shown in a) [15].
Figure 1-14: Schematic of the proposed multilayer dielectric model system. The
electrode/dielectric interface is tailored by plasma treatment. Under electric field, Li+ cations
behave as a probe to target interfaces introduced via spin casting.
xi
Figure 2-1: PVDF processing procedure showing a) magnetic stirring (10-15% solid wt), b)
solution degassing, c) solution casting using Dr. Blade, d) film drying under vacuum 180oC for 1
hr stepped down to 60oC for 3 hrs with final 1-layer film, e) stacked multilayer sandwiched
between Kapton protective sheets and f) hot-pressing at 18-24 MPa at 150oC for 30 min.
Figure 2-2: Spin cast procedure depicting a) 3 – 15 wt% P(VDF-TrFE) solution deposition, wet
film drying at 100oC for 15 min and c) schematic of dried film with 10 scratched in profilometer
scan areas.
Figure 2-3: Spin cast parameter study varying spin speed (rpm) and solution solid wt%. a) contains
data for 15 and 7.5% wt films while b) shows thicknesses measured for 3% wt thin films.
Figure 2-4: Schematic showing stepwise process of determining plasma process etch rate. a) film
deposition onto square silicon substrate, b) scratched groove spanning substrate, c) profilometry
over step edge (5 scans total) d) application of 30s plasma treatment (see Table III for parameters),
and e) profilometry on treated film. Steps e – d are repeated 3 times.
Figure 2-5: Copolymer film thickness as a function of plasma treatment exposure time. A
calculated R2=9.97 indicates a linear relation between film thickness and exposure time where an
etch rate of 0.74 nm/s can is extracted.
Figure 2-6: Simplified model for a dielectric material under impedance spectroscopy test,
incorporating ideal capacitive and resistive components to describe electronic, orientational, and
ionic polarizations. The EC used in this dissertation is leveraged from this and discussed in section
2.4.2.
Figure 2-7: J(E) behavior of a) PI sample set displaying 45 data points taken at 25oC and b) the
bootstrapped sample set PI* containing 45 data points resampled with replacement from PI. The
outcome of resampling with replacement can be seen by gaps in data due to duplicate sampling.
Figure 2-8: EC model depicting polarization mechanisms associated with P(VDF-TrFE)
polycrystalline structure at frequencies spanning 10-1-105 Hz. Nested CPE3/R4 and CPE4
distributed circuit elements are incorporated to describe low frequency ion interaction with crystals
and the electrode/dielectric interface respectively.
Figure 2-9: Schematic of TSDC peak deconvolution procedure depicting a) fitting of strongest
signal in convoluted spectrum, b) subtraction of fit function from raw data over temperature range
and c) fitting deconvoluted low temperature peak.
Figure 3-1: a) SEM image of 2-layer cross section (displaying developed interface) and b) DSC
thermogram of un-pressed, 1-layer hot-pressed, and 4-layer laminate.
Figure 3-2: ATR-FTIR morphological characterization of hot-pressed laminates. Absorbance
spectra of PVDF films from top to bottom A) 1-layer, B) 1-layer hot-pressed, C) 2-layer D) 3-layer
and E) 4-layer laminates. Salient spectral features are discussed in this section. All PVDF
vibrational modes are presented in Chapter 2 section 2.3.1.2 Table 2-IV.
Figure 3-3: a) dielectric breakdown strength shown for 1-layer films, 1-layer hot pressed, 2- and
3-layer films and b) dielectric breakdown strength plotted as a function of t for 1-layer hot pressed,
2- and 3-layer films.
Figure 3-4: Dielectric permittivity and loss tangent calculated from impedance spectroscopy
measurements between 10-2 Hz – 105 Hz for 1-layer and 4-layer PVDF laminates. EC modeling
performed using the shown circuits for each sample set (bottom left) with parameter estimates
(bottom right).
xii
Figure 4-1: C1s and O1s spectra for as-spun control as well as post-anneal samples. Line shape
does not vary between 45s – 180s treatment times for all sample sets. Intensity axis heights share
the same scaling for easy comparison in each C1s and O1s group respectively.
Figure 4-2: ToF-SIMS depth profiles for CFO+ and CO+ ions in a plasma-treated film. This plot
was generated by taking the ratio of the corresponding profiles of the treated film to those of the
untreated control film (plasma treated signal/untreated signal).
Figure 4-3: a) Copolymer surface roughness (Sq) as a function of plasma exposure time for the
as-spun control and post-anneal sample sets. Data is averaged from 8 separate 87μm x 87 μm scans
per sample. b) Images of film surface topology for each treatment condition taken by optical
profilometry. Each image corresponds to the average Sq in a.
Figure 4-4: a) Contact angle measured as a function of treatment time as-spin and post-anneal
conditions. b) Images of H2O sessile drops at first impact on copolymer surface for each treatment
time.
Figure 4-5: Dielectric response under 1V rms AC electrical signal for the post-anneal processing
condition. Untreated, 45s, and 180s treated samples were tested.
Figure 4-6: Low field I(V) measurements for the post anneal sample set. The untreated control,
45s treated, and 180s treatment times are shown.
Figure 4-7: High field J(E) measurements for the post-anneal processing condition. Again, the
untreated control, 45s and 180s treatment times are shown.
Figure 4-8: Both Schottky and Poole-Frenkel plots for untreated and plasma treated P(VDF-TrFE)
J(E) data are displayed. The linear fit equation used to calculate permittivity from equations (4-5)
and (4-6) for each set of data are labeled and displayed next to the corresponding curve.
Figure 4-9: Parametric study of current generation during a Schottky emission process. Graph a,
b, and c demonstrate the impact of ranging A*, εr, and ϕS, respectively. Variables held constant and
their respective values are shown above each plot, while the upper and lower limits of the
parameter ranged is indicated within the plot.
Figure 5-1: Charging current measured for PI as a function of charge time at room temperature.
Figure 5-2: J(E) data for untreated and O2 plasma surface treated PI at a) 25oC, b) 100oC, and c)
150oC. Applied voltage is held for 20s (time required to obtain steady state conduction) before
current measurement at each field
Figure 5-3. Arrhenius plot for Pure P1 over the temperature range 25oC – 150oC. Activation
energy is extracted from linearization of equation (5-6) and the slope of the linear fit function
Figure 5-4 Hopping parameters Jo and d estimated using the bootstrapping statistical approach
Figure 5-5 J(E) data from Figure 5-2 transformed into Schottky plot format. Again, measurements
were taken at a) 25 oC, b) 100 oC and c) 150 oC. Here slope of linear fit corresponds to βs/kT in
equation (5-7)
Figure 5-6 Permittivity values calculated from linear fits from Schottky plots between 25oC –
175oC. A shaded region is marked indicating the range between high frequency permittivity (n2)
and permittivity measured between 100Hz – 100kHz.
Figure 6I-1: a) real part of the permittivity and b) loss tangent measured at 25oC as a function of
frequency for ionic content up to 1.0 wt%.
Figure 6I-2: a) real part of the permittivity and b) loss tangent measured at 100oC as a function of
frequency for salt wt %’s 0 – 1.0%.
xiii
Figure 6I-3: AC conductivity calculated using equation 1 calculated at 100 kHz – 0.1 Hz for each
of the measured LiClO4 wt %. Temperatures measured are a) 40oC, b) 80oC, and c) 100oC.
Figure 6I-4: Cole-Cole plot of impedance for 0 – 1.0% doped samples at 25oC. The material’s
bulk response cannot be fully resolved however a general trend between LiClO4 addition and
conductivity is observed.
Figure 6I-5: Complex impedance Cole-Cole plots at 100oC for a) all tested samples, b) 0.1% –
1.0% samples and c) 0.25 – 1.0% samples.
Figure 6I-6: Arrhenius plot of σb for LiClO4 range of 0.25% – 1.0%. Extracted activation energies
from linear fits are shown in the embedded table
Figure 6I-7: a) physical model of doped P(VDF-TrFE) with equivalent circuit (EC) model used
in fitting impedance spectra, b) raw impedance data (open symbols) with EC fit (solid lines) at
25oC and c) 100oC.
Figure 6I-8: Estimated parameter values for a) Q2 from CPE2 as a function of salt wt % and b)
approximated material permittivity. Each film tested is of equivalent thickness. Units for Q are
written in terms of conductivity as S sn given equation 61-5.
Figure 6I-9: The behavior of CPE3 as a function of LiClO4 wt % for the temperatures 25 oC – 100 oC. the standard deviation of parameter estimates in the sample set are reflected by error bars.
Fitting between 100oC – 110oC produces large standard deviations in n3 as well as erratic parameter
estimates for n3 and Q3.
Figure 6I-10: Fit results for CPE4 and n4 as a function of salt wt % at temperatures 25 oC – 110 oC. Q4 reflects ionic charge interaction associated with Li+ and the electrode where 0.5 < n4 < 1.0
estimates indicate electrode/polymer interfacial roughness and heterogeneous charge distribution
at the interfce.
Figure 6I-11: Fit results for resistive EC elements a) R3 associated with amorphous regions of the
bulk and b) R4 associate with the crystalline/amorphous interface plotted as a function of LiClO4
solid wt %. R4 is shown to dominate resistive response at low temperature and LiClO4 wt%.
Figure 6II-1: schematic of multilayer material system depicting material components and the
anticipated results.
Figure 6II-2: SEM image of 5-layer sample depicting PVA interfaces and P(VDF-TrFE)
copolymer layers.
Figure 6II-3: permittivity and loss tangent for pure 10μm copolymer cast from MEK. A strong
relaxation peak in the loss tangent is observed between 1-30 Hz in the vicinity of Tc.
Figure 6II-4: a) permittivity and loss tangent for a 0.25% doped 1-layer, b) permittivity and los
tangent for 0.25% 4-layer. Integration of interfaces reduces low frequency polarization as well as
lowers relaxation frequency and tan(δ) peak magnitude. c) M’’ relaxation of doped P(VDF-TrFE)
compared with pure PVA.
Figure 6II-5: AC conductivity at 100oC calculated using equation 6I-1 for a doped 1-layer, PVA
film, and doped 4-layer. Conductivity is reduced by two orders of magnitude at low frequency in
the layered film relative to the 1-layer control.
Figure 6II-6: a) EC used in fitting the composite impedance data. The model takes into account
bulk responses from doped P(VDF-TrFE), pure PVA and electrode/dielectric blocking
polarization. b) EC fits to 4-layer M’ data at 40oC, 80oC, and 110oC. Good qualitative fits enforce
model accuracy.
xiv
Figure 6II-7: a) CPE2 and CPE5 EC outputs corresponding to P(VDF-TrFE) and PVA bulk
capacitances as a function of temperature are shown (left) and converted to permittivity (right). b)
1-layer blocking CPE estimations (left) and 4-layer blocking CPE estimations (right).
Figure 6II-8: Depolarization current density for pure 1-layer 10μm films at a) Ep=10MV/m, b)
Ep=20MV/m, and c) Ep=30MV/m. All measurements were performed using tp=15min, Tp=50oC
and heating rate 5oC/min.
Figure 6II-9: fit TSDC spectrum using Bucci-Fieschi equation along with parameter estimates for
entire 1-layer sample set undergoing TSDC with the following parameters: Ep=20MV/m,
Tp=50oC, tp=15min, and scan rate 2.5oC/min. Individual fit components along with total synthetic
spectrum are shown.
Figure 6II-10: Depolarization current density for 0.25% doped 1-layer 10μm films at heating rates
a) 5oC/min, b) 2.5oC/min, and c) 1oC/min. All measurements were performed using tp=15min,
Tp=50oC and Ep=20MV/m.
Figure 6II-11: Depolarization currents for a) pure 1-layer sample set and b) pure 4-layer sample
set. Experimental TSDC conditions are Ep=20MV/m, Tp=50oC, tp=15min and a heating rate of
2.5oC/min for both sets.
Figure 6II-12: Depolarization currents for a) 1-layer doped sample set compared to 4-layered
doped sample set and b) 1-layer pure sample compared with 4-layer doped sample. Experimental
TSDC conditions are Ep=20MV/m, Tp=50oC, tp=15min and a heating rate of 2.5oC/min for both
sets.
Figure 6II-13: High voltage dielectric breakdown Weibull analysis for a) all breakdown fields
exhibiting bi-modal Weibull distributions separating defect and intrinsic type breakdown
mechanisms and b) 7 lowest Ebd events analyzed under IEEE standards for small sample sizes.
Figure 7-1: Space charge distribution between cathode and anode of INS3-SC1 cable insulation
under 60kV/mm stress as a function of measurement time. Packets of positive and negative charge
are clearly resolved [16]. Figure C-1: Linearized J(E) data into Poole-Frenkel plots for PI and PPIDS. Data is shown for
measurements at a) 25oC, b) 100oC, and c) 150oC.
Fig C-2: Permittivity values calculated from linear fits in PF plots. A shaded region is marked in
both plots that indicates the range between high frequency permittivity defined by polyimide’s
refractive index squared (n2) and permittivity measured at 1kHz.
Figure D-1: Histograms of parameter estimates from PI* after 10,000 iterations for Jo and d at
25oC, 75oC, and 100oC.
Figure D-2: Histograms of parameter estimates from PPIDS* after 10,000 iterations for Jo and d
at 25oC, 75oC, 100oC, 125oC, and 150oC.
Figure D-3. raw data from the PI data set (black points) displaying the converged fit result using
nonlinear regression (blue solid line) superimposed.
Figure D-4. Raw data from the PPIDS data set (red points) displaying the converged fit result
using nonlinear regression (blue solid line) superimposed
xv
LIST OF TABLES
Table 1-I: Dielectric breakdown strength of PVDF in relation to CF4/O2 plasma treatment time
and power [17].
Table 1-II: Dielectric breakdown strength of untreated, single side treated, and double side O2
plasma treated PI [13].
Table 1-III: A brief review of literature that investigates the impact of ionic conduction through a
variety of polymer dielectrics [18] [12] [19] [20] [21] [16] [22] [23] [24]. The fact that contributing
ionic species is frequently poorly defined or unknown is highlighted.
Table 2-I: calculated average film thickness, standard deviation and thickness variation in hot
pressed multilayers. 1-, 3- and 4- layer stacks consisted of ~8 micron thick single layer films, while
the 2-layer stack contained two ~13 micron films.
Table 2-II: Profilometer scan thicknesses t for two films processed at 3% solid wt., 1,000 rpm
measured as described in Figure 2-2.
Table 2-III: Plasma treatment parameters used on P(VDF-TrFE) and PI in this work.
Table 2-IV: Characteristic features of PVDF FTIR spectra by wave number, adopted from
Lanceros-Mendez et al. and other contributing authors [25, 26, 27, 28, 29, 30, 31].
Table 2-V: Surface polarity calculated for a number of common polymers using contact angle
experiments, depicting influence of plasma treatment on surface properties [32, 33, 34, 35, 36, 37,
38].
Table 4-I: Elements detected by XPS in atomic percentage
Table 4-II: Chemical species determined by XPS in atomic percentage
Table 5-I: parameter fit values and confidence intervals at each temperature for untreated PI
sample set. No values for Jo or d are provided because of the program’s inability to fit the data
using equation (5-2).
Table 5-II: parameter fit values and confidence intervals at each temperature for the plasma treated
PPIDS sample set
Table 6I-I: DSC results as a function of LiClO4 solid wt% for the first heating cycle. Endothermic
peak temperatures along with integration results are shown and the range in standard deviations of
the sample sets are given in italics.
Table 6I-II: Error % for CPE4 extracted from Q4 and n4 parameter estimates in 1 μm and 10 μm
samples.
Table 6II-I: DSC results for the first heating of 10 micron P(VDF-TrFE) without (pure) and with
0.25% LiCLO4 included. The solvent used was MEK, dried for 15 min at 100oC and annealed for
24 hrs at 142oC under vacuum.
Table E-I: Error percent associated with CPE2 parameter estimates out-put from the model
Table E-2: Error percent associated with CPE3 parameter estimates out-put from the model
Table E-3: Error percent associated with CPE3 parameter estimates out-put from the model
Table E-IV: Error percent associated with R3 and R4 parameter estimates out-put from the model
xvi
ACKNOWLEDGEMENTS
I would first like to acknowledge my two advisors, Dr. Zoubeida Ounaies and Dr. Michael
T. Lanagan, as well as colleague Dr. Amira Meddeb and former colleague Rudeger Wilke. These
individuals have provided unwavering support and guidance from the beginning of my career as a
graduate student, through a full repertoire of Materials Science coursework, departmental
candidacy and comprehensive exams and final defense of a 5 year long Ph.D. Simply put, it would
have been impossible to become the scientist I am today without their knowledge and willingness
to support my growth as a professional, teammate and individual. I hope they continue to impact
the lives of future students the same way they have done for me.
I am also grateful for support received from my committee members Dr. Ramakrishnan
Rajagopalan and Dr. Michael Hickner. Having well thought out critique early in the formulation
of this dissertation improved the rout my research took, teaching me the value of approaching
science from many different angles. My experience with them was positive, and I will continue to
seek input from experts in similar fields as mine to enhance my future work, whatever it may be.
My thanks go out to the individuals I have met along the way. EMCLab mates Hassene,
Nirmal, Saad, Masud, The Duke of Harrisburg Albert Forster III, Wei, John, Travis, Jess, Nick,
Lydia, Tahzib, new students Dash, Jai, and Xiaoue as well as High Energy Capacitor Group mates
Mengxue, Wutti, Maryam, Amir, Jiasheng, Rodger, Cesar, and Hossein have all made my graduate
school experience exceptional, as well as enabled me to add a particularly impressive run on
sentence in an otherwise well written document. In addition, Jeff Long, Steve Perini, Josh
Stapleton, Max Wetherington, and Jeff Shallenberger all played integral parts in my research and
were indispensable teammates during experimental planning and execution.
I am grateful for the National Science Foundation as part of the Center for Dielectrics and
Piezoelectrics under grant Nos. IIP-1361571 and IIP-1361503 for providing financial support.
Without this grant, research would have been impossible and this dissertation nonexistent.
Finally, my love and thanks go out to my mother Carmen Vecchio for her unyielding
strength, love, and encouragement for me to achieve the best in all aspects of my life and become
the best version of myself. She should know that distance will never lessen her presence in my life.
xvii
This chapter of my life is dedicated to my mother Carmen Vecchio and father Vincent Vecchio,
who I love very much
1
CHAPTER 1
INTRODCTION
1.1 HISTORY OF CAPACITORS
In the year 1745, German scientist Ewald Georg von Kleist developed the first known
capacitor technology which was then independently discovered at the University of Leyden,
Holland, by Dutch physicist Pieter van Musschenbroak. The technology was named the Leyden
jar, however both Ewald Georg von Kleist and Pieter van Musschenbroak are given credit for its
development [39] [40]. The capacitor consisted of a glass jar wrapped with metal foil inside and
out, where an electrostatic generator was used to deliver electrical charge to the inner foil while
holding the outer foil grounded causing a maintained state of charge which could be dissipated.
Although the current state of technology at the time did not allow for the capacitor as an electrical
component, its creation marked a turning point in the field of science concerning electricity and
energy storage.
As time and the field of electronics progressed, so did the complexity of capacitors and
their potential applications. Over one century after the Leyden jar’s development came the wax-
impregnated dielectric capacitor invented by Fitzgerald in 1876 [41]. Early capacitors used in radio
receivers typically employed the foil-wax paper capacitor as a power supply filter used for ripple
current reduction [39]. In 1909, William Dubliner invented the first capacitor employing mica as
the dielectric which was predominantly used in radio transmission applications, which was then
shortly followed by the aqueous electrolytic capacitor (first appearing in radio technology in the
late 1920s) [39].
As stated before, mica became a widely used ceramic for capacitors until the discovery of
barium titanate in 1941 [42]. The high permittivity of barium titanate in the range of 1,000 – 10,000
out matched the available repertoire of dielectrics of that time, catalyzing efforts to understand the
material’s crystal structure in relation to dielectric performance and optimize its family of materials
[43]. Around the same time frame (early 1950s) paper capacitors with vacuum deposited electrodes
entered the scene, making a debut in the telephone industry [39]. The process of vacuum metal
deposition termed “metallization” onto paper produced reliable electronic devices. Design
incorporating metalized paper dielectrics is believed by experts in the field of polymer dielectric
2
science including Janet Ho, T. Richard Jow and Steven Boggs to have paved the way for the
creation of initial polymer film capacitor concepts [39]. As mentioned in Ho et al. [39] and
reinforced by claims made from Tran Doan Huan et al. [44] research done through Bell Labs in
1954 on lacquer separation from paper dielectrics lead to the creation of the metalized lacquer film
capacitor [45], suggested to be the first metalized polymer film capacitor technology ever created.
Since work done by McLean and Wehe [45], metalized polymer film capacitors have become the
present-day cornerstone for high power static electrical storage and dissipation technology. A
significant amount of effort from the dielectric community is focused on improving polymer film
capacitor characteristics such as dielectric breakdown, leakage current, AC dielectric loss tangent
and permittivity, and degradation at high temperatures.
In the remainder of this chapter, a brief survey of current materials implemented in
capacitor design is presented along with their use in modern day applications. This is followed by
an in-depth overview of materials used in this dissertatio. The chapter concludes with outlining
the problem and motivation behind this research, as well as this work’s goal.
1.2 CAPACITORS AND DIELECTRIC MATERIALS
Since the invention of the Leyden jar, materials used in capacitor technologies have
expanded to include multiple types of materials depending on the desired applications. In this
section, three types of capacitors and their associated dielectric materials are reviewed:
electrochemical capacitors, ceramic capacitors and polymer capacitors. Each type exhibits unique
characteristics based on the underlying mechanisms for charge storage, which is portrayed in a
Ragone plot depicting specific power as a function of specific energy in Figure 1-1. It is seen that
polymer and ceramic based capacitors perform quite differently than electrochemical capacitors,
enabling higher power densities rather than energies. An even greater disparity is seen relative to
batteries and fuel cells which show the highest specific energy of all charge storage devices. In the
following sections, the mechanisms that drive capacitor performance will be reviewed from a
material’s perspective. Discussion will begin with a brief review of electrochemical capacitor
technology and end with polymer dielectrics.
3
Figure 1-1: Ragone plot comparing the specific power and energy in common types of energy and power storage
devices. (figure credit [1]).
1.2.1 Electrochemical Capacitors
Electrochemical Capacitor’s (EC) mechanism of energy storage involves charge separation
between the interface of an electrolyte and a solid electrode. A basic schematic of a typical EC is
presented in Figure 1-2 which depicts the components of the EC, the charging, and discharging
states of the device. The anode and cathode contain an electrolytic solution that provides the
available anionic and cationic charges distributed through the bulk of the device. In the charged
state, anion interaction with the anode and cation interaction with the cathode create charge
separation at the electrode/electrolyte interface termed double layer capacitance. EC’s exhibit
characteristically large capacitances which can be understood by considering the fundamental
equation for capacitance:
𝐶 =휀𝑟휀𝑜𝐴
𝑑 (1 − 1)
where εr is relative dielectric permittivity, εo is the dielectric permittivity of vacuum (=8.85.10-12
F/m), A is effective surface area and d is thickness. Small d associated with the electrical double
layer as well as large A values resulting from electrode texturing enable these devices to store large
amounts of charge.
Electrode materials can consist of carbon, metal-oxides, and conductive polymers. Each
electrode material influences capacitor performance. For example, the use of carbon-based
4
Figure 1-2: The charging and discharge cycle depicted for a typical electrochemical capacitor (image credit [2]).
electrode materials provide a high surface area which is controlled by the electrode porosity and
can be tuned to achieve optimal performance based on cation and anion size [46]. Metal-oxides
provide their own advantages including low electrical resistance and high specific capacitance.
Research done by the US Army Research Laboratory suggests both ruthenium-oxide and
manganese-oxide show promise in the development of high energy double layer capacitors based
on energy densities and power densities of 8.5 Wh/kg and 6 kW/kg (ruthenium-oxide) and
potential for low cost (manganese-oxide). A porous separator material is placed between anode
and cathode, allowing ionic passage through its thickness however preventing electrical interaction
between the anode and cathode.
Electrode material influences the choice of electrolyte used in the EC’s design. Electrolytes
used in EC’s fall under either organic or aqueous electrolytes. Aqueous electrolytes enable high
power operation of the EC due to a high concentration and conductivity of ionic carriers [47] [1].
Simple salts such as ACl, A2SO4, ANO, ALi, Na, and K eliminate the need for purification and
handling under controlled atmospheres, allowing for simple fabrication. Organic electrolytes
typically exhibit high breakdown voltages in comparison to aqueous, which increases the
attainable cell voltage of the EC (2-2.5V) and directly relates to the possible energy density
achievable by the capacitor [47]. Separator materials used in EC design are dependent on
characteristics of the electrolyte: organic electrolytes typically requiring polymer or paper
separators while aqueous electrolytes are usually coupled with either ceramic or glass fiber
5
separators. Regardless of material selection, high ionic conductance, high electrical resistance and
low thickness are required [47].
EC’s are typically used to “fill the gap” between capacitors and battery technology (see
Figure 1-1). Applications demanding energy within the time span of 10-2 s ≤102 s are ideal for
EC’s due to the ratio of stored energy to available power inherent of the device that prevents the
need to use oversized capacitor and battery components [48]. Another favorable attribute to EC’s
is a lack of toxic materials typically found in battery technologies and ability to withstand large
quantities of charge-discharge cycles without servicing [48, 49, 50]. Despite their benefits, EC’s
cannot support applications which require AC conditions or high ripple currents because high
internal resistance put the device at risk for thermal degradation [48]. By the same mechanism,
large internal resistances limit the device’s achievable peak power in comparison to conventional
capacitors, making them less desirable for pulse power applications.
1.2.2 Ceramic Dielectrics
Ceramic capacitors have become integral components of modern-day electronics such as
cell phones and computers, in which hundreds to thousands are present. In application,
multilayered ceramic capacitors (MLCC) are typically employed due to their compactness and
high capacitance [51]. MLCCs are predominantly used in resonant circuits and filters (requiring
low dielectric loss and high stability) and power supply bypass and decoupling (requiring high
capacitance but allowing for moderate dielectric loss) [51]. Ceramic dielectrics fall into three
classes: 1) low permittivity dielectrics offering superior temperature and voltage stability and low
losses for resonant circuit applications, 2) high permittivity dielectrics offering superior volumetric
efficiency, however nonlinear capacitance change over broad temperature and frequency ranges,
and 3) barrier layer dielectrics offering very high capacitance with limited operating voltage
(<25V). Since class 3 dielectrics operate based on different fundamental principles than the
materials used in this dissertation, the following will focus on dielectric classes 1 and 2.
Class 1 ceramic dielectric materials are characterized by their low relative dielectric
permittivity εr in the range of 5 to 100’s, and a low dissipation factor less than 0.01 [51]. Early
dielectric materials used to manufacture class 1 dielectrics have consisted of porcelain, steatite,
and mica. Modern day low permittivity ceramic dielectrics are based on simple oxides such as
TiO2 (rutile) as well as perovskite titanates such as CaTiO3 as well as modified (Ca,Sr) (Zr,Ti)O3.
6
Dopant elements may also be added during the dielectric fabrication process as needed to achieve
optimal dielectric performance. The influence of Nb, Ta, Al, Ca, Y, Ba, and Mn on TiO2’s optimal
processing conditions, microstructure, and electrical properties have all been extensively
investigated in past work [52, 53, 54, 55, 56, 57]. For example, work done by Chao and Dogan
[57] showed that 0.05 mol% Mn doped into TiO2 results in reduced conductivity and dielectric
loss, while increasing dielectric breakdown strength by 25% and energy storage efficiency from
93% - 98%. This was achieved in TiO2 dielectrics with low concentrations of dopant Mn that
maintain a linear permittivity – temperature relationship between 25oC – 200oC, which is a
characteristic of class 1 ceramic dielectrics [51].
Class 2 ceramic dielectrics exhibit characteristically high εr >1000 and thus higher
volumetric efficiency than class 1, however εr’s nonlinear temperature dependence as well as
voltage dependent capacitance make these materials suitable in applications where low loss and
high stability are not necessary. Unlike paraelectric class 1 dielectrics, class 2 dielectrics are
ferroelectric ceramics in which the material’s crystal symmetry allows for crystal domains
exhibiting spontaneous polarization to exist within the material’s bulk. Perhaps the most famous
ceramic, barium titanate (BaTiO3) exhibits an εr as high as 10,000 arising from ferroelectric
domain configurational entropy associated with the tetragonal – cubic phase transition portrayed
in Figure 1-3. Another characteristic of BaTiO3 is the material’s charge state dependence on strain
called piezoelectricity. This phenomenon arises from strain dependent displacement of Ti in the
tetragonal phase of the material. The piezoelectric nature of BaTiO3 as well as other class 2
ceramics such as lead zirconate titanate (Pb[ZrxTi1-x]O3 or “PZT”) extends their application
beyond power supply bypass and decoupling and into sensing and actuation. Previous work
demonstrates successful PZT based device development including cantilever actuators, atomic
force microscopy probes, micro pumps, and ultrasonic micromotors/transducers [58, 59, 60, 61,
62]. In recent years, concerns involving the toxicity of lead have catalyzed the development of
lead free piezoelectric ceramic materials such as sodium niobate [(K0.5Na0.5)NbO3 or “KNN”] [63],
and SrZrO3 modified sodium bismuth titanate [(Bi0.5Na0.5)TiO3-SrZrO3 or “BNT-SrZrO3”] [64].
Despite the development of lead-free high permittivity ceramics, their inherent mechanical
properties such as high stiffness as well as required high temperature processing conditions limit
their use in practical applications in which high strain and compatibility with polymers is required.
7
Figure 1-3: The dielectric constant of BaTiO3 as a function of temperature. Crystal phase transitions are shown as a
function of temperature above, corresponding to maxima in the material’s dielectric constant (image credit [3]).
1.2.3 Polymer Dielectrics
Polymer dielectrics are typically sought after in energy storage technologies for
applications requiring high breakdown strength. Unlike ceramic capacitors, polymers exhibit low
dielectric permittivities and high breakdown strengths, enabling them to achieve large energy
densities given equation 1-2 below:
𝐸𝑠𝑡𝑜𝑟𝑒𝑑 =1
2휀𝑟휀𝑜𝐸𝑏𝑑
2 (1 − 2)
where εr and εo are the material’s relative permittivity and permittivity of vacuum respectfully, and
Ebd is the dielectric breakdown field of the material. Another quality of polymer dielectrics is their
ability to undergo self-clearing defined as the electrical isolation of defect driven breakdown
events through excessive evaporation of the electrode. Self-clearing electrodes enable continued
device use after localized breakdowns and enhance the reliability of the overall device. Currently
the demand for improved polymer capacitor technologies comes from consumer-based markets
such as general consumer electronics and hybrid/electric vehicles, industrial markets requiring
improved power conversion electronics, and also military applications. The drive for improved
polymer dielectrics is centered on the following requirements: 1) improvement of high temperature
performance and thermal management in power electronics with emphasis on efficient thermal
regulation, and 2) increased achievable electrostatic energy density without increasing dielectric
losses or reducing operating field. This objective is focused around improvement of the dielectric
constant of the material as well as dielectric breakdown strength [44]. The following discussion in
8
this section is focused on requirement 2, improvement of electrostatic energy density in polymer
dielectric materials for power and energy density applications.
Initial dielectric development was centered around the consumer radio electronic market,
however a demand for improved high temperature performance and increased energy density while
maintaining low cost of production motivated development of polymer dielectrics. Since the Bell
Labs discovery of the metalized lacquer film capacitor [39] [44] [45], dielectrics manufactured for
polymer film capacitor applications expanded to include polyethylene (PE), polystyrene (PS used
as the main ingredient in Styrofoam), polytetrofluoroethylene (PTFE known as Teflon™),
polyimide (PI) and polypropylene (PP). The modern state of the art dielectric material for high
energy density applications is currently biaxially oriented polypropylene (BOPP). The maximum
achievable stored electrostatic energy density of a linear dielectric material is governed by the
following equation 1-2. Estored’s dependence on the square of Ebd make dielectric breakdown
strength the dominating parameter for determining a dielectric’s potential to succeed in energy
density applications. In this regard, the high dielectric breakdown field of BOPP measured as high
as 850 MV/m [65] makes BOPP a dominating material for energy density applications despite its
low permittivity of 2.2. High dielectric breakdown strength is an inherent quality of polymer film
dielectrics, surpassing that of their ceramic counterparts exemplified by Figure 1-4.
Figure 1-4: Breakdown strength of polymers, composites (nanodielectrics) and ceramics organized by their dielectric
constant (image credit [66]).
9
Although BOPP exhibits one of the highest breakdown strengths and thus achievable
energy densities of ceramics, nanodielectric composites and polymers, the material undergoes
considerable de rating at elevated temperatures due to its relatively low melting temperature of
80oC [44]. Other materials such as PI are being considered due to their high glass transition
temperatures in the range of 370 – 400oC [67] [68] (depending on chemical structure and
processing) as well as their considerably high dielectric breakdown strength reported as high as
615 MV/m [13].
Both BOPP and PI are non-polar polymers and exhibit low permittivities (2.2 and 3.2
respectively). In order to improve energy density in organic dielectric films, recent attention has
been placed on increasing the permittivity of materials used in high field applications. One
example of this is the introduction of high permittivity nanofillers embedded into a polymer matrix
to improve composite permittivity. Work done by Tomer and Randall [4] report increased
dielectric permittivity in silicone composites containing BaTiO3 nanoparticles in the range of 15 –
25% volume fraction, however reduction in Ebd of the composite (Figure 1-5a) detrimentally
impacts the maximum achievable electrostatic energy density by equation 1-2. Trends in reduced
dielectric breakdown strength in nano-composite polymer dielectrics is observed in other systems
as well including PP where increased Al nanofiller presence in the PP matrix cause reduction in
dielectric breakdown strength [5]. Another approach is to improve energy density capabilities of
polymer film by using high permittivity polar polymers as the main dielectric material to improve
Figure 1-5: a) impact of BaTiO3 nanoparticle inclusion and nanoparticle chain alignment on Ebd in silicone composites
[4] and b) impact of Al nanoparticle inclusion on Ebd in PP [5]. DC dielectric breakdown data analyzed via 2 parameter
Weibull statistics.
10
permittivity without detrimentally impacting Ebd by the inclusion of high εr particles. Examples of
recent work focusing on high field performance of organic dielectrics in the absence of nanofillers
implement poly(vinylidene fluoride) (PVDF) as the high εr material. PVDF along with recent
advances in dielectric material development incorporating PVDF is reviewed in the following
section.
1.3 PVDF AS A DIELECTRIC MATERIAL
1.3.1 Material Structure
PVDF is characterized by a chemical formula linking PE with PTFE in the following
sequence –(CH2–CF2)n–. The flexibility afforded by its structure along with stereochemical
constraint afforded by PTFE groups enable the formation of multiple molecular structures [69].
PVDF can crystallize into four main phases labeled α, β, γ, and δ (referred to as forms II, I, III,
and IV respectively) depending on the conditions the material is exposed to during processing and
fabrication [70] [71]. The α-phase (form I) shown in Figure 1-6 is a non-polar crystal phase of the
material in which the molecular chain assumes a trans-gauche (TGTG¯) configuration. Unlike α,
β-phase crystals (Figure 1-6) exhibit an all trans configuration (TTTT), resulting in a polar phase
arising from aligned CF dipoles along the crystal’s c axis. The α-phase is typically converted to β-
phase through mechanical deformation of the material’s α-phase via stretching uniaxially or
biaxially [72] [73] [74]. Other methods have also been reported to result in the formation of β-
phase during film processing including quenching from the melt [75], controlling pressure and
temperature, and choice of solvent/solvent drying protocol required during solution processing
[76] [69]. The γ and δ phases are quite similar, both exhibiting a weakly polar crystal structure in
comparison to the β-phase, in which the TGTG¯ chain propagation is in the c direction (γ-phase).
The differentiation between γ-phase and δ-phase is a…TTTTTGTG¯TTTT… chain conformation
Figure 1-6: a) polymer chains of PVDF showing TTTT and TGTG¯ conformations associated with the β- and α-
phases respectively (image credit [6]), b) α-phase crystal structure and c) β-phase crystal structure (image credit [7]).
11
exhibited by δ, where the transgauche kinks in the polymer chain along the crystal’s c direction
disrupts long range TTTT configuration characteristic of β-phase. The γ-phase can be attained by
annealing near the melt and through solvent choice while δ-phase has been reported to arise from
poling α-crystals under strong electric fields [76, 77, 78, 79]. Regardless, the weak polarity of the
γ- and δ-phases in comparison to β-phase makes them unfit for practical applications requiring a
polar crystal structure.
PVDF’s electrical properties are controlled by the type and quantity of crystal phases
present in the material. For example, β-phase development within the bulk of the material results
in a ferroelectric PVDF due to permanent dipole orientation and upon poling using an external
electric field. Outside of the material’s ability to be processed into a ferroelectric film, the presence
of CF dipoles within the material also cause PVDF to have a high dielectric constant in the range
of 8-12 which makes it appealing as a material for energy storage applications. Additionally,
ferroelectric polymers including PVDF exhibit a nonconjugated polymer backbone structure,
making them highly insulating materials [69]. Good insulating properties combined with a melting
temperature Tm = 160oC enables PVDF to exhibit DC dielectric breakdown strength comparable
BOPP [44] (reported in the range of 720 – 770 MV/m [80]).
1.3.2 Limitations of PVDF for Energy Storage
Despite the material’s favorable properties such as high breakdown strength and dielectric
permittivity, large dielectric losses in the range of 0.5 – 1.0% as well as high leakage currents in
the range of 10-3 – 10-2 A/m2 at electric fields exceeding 20 MV/m limit its use in practical
applications [71] [44]. This point can be further understood by considering a Ragone plot for
power electronics where the boundaries of material performance are established by power and
energy density, which are controlled by material internal losses or leakage [8]. An example of this
is the Ragone plot for an ideal battery where the energy/power relation is theoretically derived
with and without leakage contributions (denoted by RL) in Figure 1-7. Other limitations damaging
unmodified PVDF as a viable material for high energy density are linked to its β-phase: there is a
strong ferroelectric hysteresis associated with polar domain switching leading to energy dissipation
and low recoverable energy densities especially in AC conditions (exemplified by Figure 1-8)
[81]. Recent work has targeted ferroelectric P(VDF-TrFE) and the relaxor ferroelectric
12
Figure 1-7: Ragone plot for an ideal battery depicting the energy/power relation with and without contributions from
leakage RL (image credit [8]).
Figure 1-8: a) displacement current D vs electric field E loop depicting normal ferroelectric behavior with purple
shaded region representing recoverable energy density (image credit [9]) and b) D-E behavior of biaxially oriented
PVDF at 10Hz showing strong ferroelectricity (imag credit [10]).
13
poly(vinylidenefluoride-trifluoroethylene-chlorofluoroethylene) [P(VDF-TrFE-CTFE)], with
promise of a reduced ferroelectric hysteresis above the coercive field [9]. Research on P(VDF-
TrFE) demonstrates electron beam irradiation and γ-irradiation result in narrower hysteresis loops
[82], however the high cost of (VDF-TrFE) based materials coupled with degraded mechanical
properties prevent large scale applications [12]. P(VDF-TrFE-CTFE), hereby referred to as
terpolymer, has uniquely high dielectric constant of 50 achieved by using defect modification via
introduction of Cl side groups to disrupt ferroelectric domains and reduce remnant polarization
[83] [84]. Despite chemical modifications, terpolymers exhibit limited energy density of ~10 J/cm3
linked to early saturation of polarization at elevated fields [83] [11].
Finally, work done by Chen et al. [11] and Yang et al. [12] highlight the significance of
electrode/dielectric interface in the high field performance of PVDF and its derivatives. The high
field performance of P(VDF-TrFE-CTFE) was analyzed by high voltage dielectric breakdown and
current vs. voltage measurements by Chen et al. [11]. Electrode material as well as electrode
deposition technique was varied to change contact properties between the electrode and the
polymer. It was found that dielectric breakdown strength ranged from 245 MV/m to 380 MV/m in
films with Ag and Al electrodes deposited via thermal evaporation respectively (Figure 1-9). The
leakage current during charging was also electrode material dependent with high dielectric strength
Al and Cr producing the lowest high field leakage currents relative to low dielectric strength Ag.
Yang et al. [12] reports on the high field performance of biaxially oriented PVDF. Thermally
stimulated depolarization current (TSDC) measurements performed with different electrode
materials indicate that both electronic injection (similar to reports by Chen et al. [11]) as well as
impurity ion concentration associated with ionic polarization in the samples depend on the nature
of electrode/dielectric contact (Figure 1-10). It was postulated based off research done by Eberle,
et al. [85] that HF gas emitted at high electric fields reacts with Al and Ag electrode metals to
produce Ag+ and Al3+ cations, as well as F- anions which contribute to the large ionic
depolarization peak in TSDC. The exact nature of these ionic species in PVDF are not well
understood, however their presence induces space charge accumulations that create high field
concentrations distributed heterogeneously throughout the material. Not only does this
demonstrate the electrode/dielectric’s dominating role on high field conduction in PVDF based
dielectrics, but also the prevalence of both electronic and ionic space charge that contribute to
conduction in the material. Although space charge migration and distribution at high fields has not
14
been extensively studied in PVDF, research on PE and crosslinked PE suggests space charges at
high fields significantly impact both conduction properties as well as field distribution through the
material [86, 87, 88]. Due to PVDF’s susceptibility to impurity ion and injected space charges at
fields required to achieve high energy densities, material development aimed at improving
dielectric breakdown as well as mitigating low frequency charge migration and space charge
concentrations under high fields is necessary prior to using PVDF in practical applications.
Figure 1-9: a) dielectric breakdown strength of P(VDF-TrFE-CTFE) as a function of electrode material and b) high
field leakage current as a function of electrode material. Major differences are observed for Ag vs Al, reflecting
significance of electrode/dielectric interface on controlling high field conduction [11].
Figure 1-10: BOPVDF TSDC for a) Ag electrodes and b) Al electrodes. Both samples are measured under the same
conditions: Ep = 10 MV/m, Tp =50oC, heating rate = 5oC/min, and tp ranges from 10 min – 50 min [12].
15
1.4 INTERFACES FOR IMPROVED DIELECTRIC PERFORMANCE
1.4.1 The Electrode/Dielectric Interface
Recent work done by Chen et al. [11] (Figure 1-9) and Yang et al. [12] (Figure 1-10)
highlights the dominant role of the electrode/dielectric interface in controlling electric conduction
through PVDF, however the concept of tailoring the material interface for improved high field
performance has existed for quite some time. Electron/proton irradiation and reactive plasma
treatments have proven effective in chemically tailoring polymer surfaces prior to electrode
deposition. A study done by Mammone et al. [17] in 1992 analyzes the effect of CF4/O2 gas plasma
on dielectric breakdown strength in PVDF films with results summarized in Table 1-I. Plasma
surface modification resulted in an 11% increase in the breakdown strength of treated films relative
to untreated control samples. Although the evolution of material structure as a result of plasma
treatment was not the focus, more recent reports by Adem et al. [89] indicate the formation of -
C=O and -COOH after electron and proton irradiation of PVDF. These results suggest that
improved dielectric breakdown strength reported by Mammone et al. [17] could be related to
surface chemical effects at the electrode/dielectric contact. In fact, evolved electrode/dielectric
interfacial chemistry as a result of plasma treatment is found to influence high field conduction
properties in other polymer dielectrics. Recent work by Meddeb et al. [13] correlates increased
oxygen at the electrode/dielectric interface after oxygen plasma treatment in PI with increased
dielectric breakdown (reported in Table 1-II) and reduced leakage current at high fields and
Table 1-I: Dielectric breakdown strength of PVDF in relation to CF4/O2 plasma treatment time and power [17].
16
temperatures, as shown in Figure 1-11. In this regard, turning surface chemistry of dielectric films
is considered an effective way to control electrical properties for high field applications,
warranting greater attention in polar materials considered for high energy density applications.
Table 1-II: Dielectric breakdown strength of untreated, single side treated, and double side O2 plasma treated PI
[13].
Figure 1-11: Leakage current as a function of electric field for untreated and plasma treated PI at 150oC. Both
reduction of data scatter and magnitude of leakage current occurs after plasma treatment [13].
1.4.2 The Bulk Distributed Interface
Recent work intended to advance organic capacitor technology for high energy density
applications not only use PVDF and its associated co- and ter-polymers for their unique properties,
but also take advantage of interfaces distributed throughout the dielectric’s bulk to mitigate the
effect of limitations discussed in section 1.3.2. Work done by Zhang et al. [83] shows that PVDF
blended with its relaxor ferroelectric counterpart P(VDF-TrFE-CTFE) produces a composite
dielectric exhibiting energy densities as high as 19.6 J/cm3 at 640MV/m which surpasses
performances reported for BOPP (1.2 J/cm3 at 640 MV/m). It was demonstrated by phase-field
17
simulation that interfaces between pure PVDF and P(VDF-TrFE-CTFE) domains play a role in
preventing low field polarization saturation and add interfacial polarization which both enhance
the achievable energy density of the composite [83]. Other efforts to improve high field dielectric
performance that involve bulk distributed interfaces are accomplished by combining low εr
polycarbonate (PC) with poly(vinylidenefluoride-co-hexafluoropropylene) (P(VDF-HFP)). In
work done by Mackey et al. [14], multilayer co-extruded PC/P(VDF-HFP) laminate dielectrics are
fabricated where the planar connectivity between PC and P(VDF-HFP) layers results in enhanced
dielectric breakdown strength as high as ~950 MV/m, surpassing either constituent material alone
(Figure 1-12). Due to the nature of breakdown initiation (electrical treeing) it was suggested that
the layered microstructure enables defect channel deflection and thus a greater time and electric
field required to achieve dielectric breakdown. This work was then followed up by Zhou et al. [15]
Figure 1-12: a) schematic showing PC / P(VDF-HFP) coextruded composite (similar to Mackey et al.) with PMMA
tie layer incorporated. b) dielectric breakdown strength of multilayer composite using multiple tie layer materials and
c) dielectric loss tangent showing frequency reduction in ionic relaxation for the PC / PMMA tie layer / P(VDF-HFP).
Decrease in relaxation frequency is related to decreased ionic mobility in the PMMA/P(VDF-HFP) interphase shown
in a) [15].
18
with work that assessed a PC/tie-material/P(VDF-HFP) laminate composite as a function of tie-
material. Use of PMMA as the tie material between PC/P(VDF-HFP) layers resulted in 25%
increased breakdown strength, 50% higher energy density and reduced hysteresis loop areas than
a 33-layer PC/P(VDF-HFP) control film. Enhanced dielectric breakdown strength was attributed
to a smoothing of the dielectric constant distribution along the thickness direction of the film due
to PMMA interdiffusion into adjacent PC and P(VDF-HFP) layers. These outcomes were also
coupled with slower ion migration through the film relative to the control sample, determined by
tracking the ionic relaxation peak in the loss tangent of dielectric spectroscopy measurements in
the range of 1 Hz – 50 Hz (Figure 1-13). It is postulated that PMMA/P(VDF-HFP) interphase
regions restrict ionic motion under the application of electric field. These advancements in
laminated polymer film technologies motivates continued research that targets the potential to
incorporate PVDF into all organic dielectric materials for energy storage. Similarly, the
dependence on composite structure highlights the impact interfaces have on high field performance
in layered dielectrics, however up to this point a controlled, systematic study that targets how
interfaces impact impurity ion and electronic charge transport in layered all organic dielectric
materials is absent in the current body of literature.
Figure 1-13: a) SEM images of 32-layer and 256-layer P(VDF-HFP) / PC microlayer coextruded multilayer laminate
along with b) effect of composite structure on breakdown strength. The 32-layer laminate from a) results in a
composite dielectric breakdown strength surpassing P(VDF-HFP) and PC alone [14].
19
1.5 PROBLEM STATEMENT AND RESEARCH GOAL
The proposed research is to study the influence of interfaces on degradation and dielectric
breakdown of multilayer dielectrics at high electric fields. We propose a model multilayer
dielectric system depicted in Figure 1-14 using P(VDF-TrFE) in which the effect of the interface
is enhanced via controllable ionic conductivity. The addition of ionic content into the multilayer
matrix is meant to create a system in which ionic transport at the interface can be separated from
transport through the bulk. Accomplishment of this objective will be done by accomplishing the
following tasks: 1) incorporation of a reactive plasma surface treatment to tailor film surface
chemistry and explore how electronic charge injection can be controlled by tailoring the
electrode/dielectric interface, 2) developing an additive spin-casting procedure which enables the
fabrication of thin (micron – submicron) layers, maximizing the ratio of interface to bulk present
within the structure. In this step, polyvinyl alcohol (PVA) is used as a sub-micron barrier layer to
construct P(VDF-TrFE)/PVA multilayered laminates. Finally, 3) doping individual layers in
multilayered dielectrics with ionic complex to increase conduction and accentuate the interface’s
influence on ionic transport through the material. Accomplishment of task 3 will address two
aspects of ionic conduction that are poorly understood in the current standing body of literature on
polymer dielectric performance. Although ionic charge migration in dielectrics is a well
documented phenomenon, the concentration of impurity species within dielectric films such as
PVDF are at or below the ppm level [90] [91]. Doping the model multilayer film in this dissertation
serves to increase the quantity of species contributing to ionic conduction, making their electrical
response highly recognizable during experimentation. The second aspect of ionic conduction
Figure 1-14: Schematic of the proposed multilayer dielectric model system. The electrode/dielectric interface is
tailored by plasma treatment. Under electric field, Li+ cations behave as a probe to target interfaces introduced via
spin casting.
20
through polymer dielectrics that doping will address is the wide-spread ambiguity surrounding the
actual chemistry of the impurity species contributing to ionic conduction. Impurity ions in the
dielectric will result from the process of suspension polymerization used by the polymer
manufacturing industry [92]. In the case of PVDF polymerization, chain transfer agents, H2O
soluble initiators such as persulfate salts, disuccinic acid peroxide, and β-hydroxyalkylperoxide or
alkylperoxybutyric acid are used [93] [94] [95] [96]. This is then compounded with additional
factors such as polymerization procedure temperature, pressure, ingredients, and post
polymerization processing steps that all influence final polymer chain characteristics and defect
species within the material. Because of the sub ppm levels of these impurities within the material,
their exact chemical nature is not possible to definitively measure which leads to either ambiguity
or complete disregard to the nature of impurities participating in ionic conduction in the literature.
This claim is exemplified by Table 1-III sampling literature that studies ionic conduction in a
variety of polymer materials where the contributing ionic species is not well defined or
unaddressed. This dissertation provides an avenue through doping by which the chemical species
of ionic impurity within the material is well defined and controlled to behave like space charge,
avoiding the ambiguity highlighted in Table 1-III.
Table 1-III: A brief review of literature that investigates the impact of ionic conduction through a variety of polymer
dielectrics [18] [12] [19] [20] [21] [16] [22] [23] [24]. The fact that contributing ionic species is frequently poorly
defined or unknown is highlighted.
21
1.6 ORGANIZATION OF DISSERTATION
Chapter 2 discusses polymer processing techniques implemented to create single layer,
plasma treated, and multilayered PVDF based dielectrics used in this work. An overview of
material characterization equipment used in this dissertation is then provided and broken into two
sections: structural/morphological characterization techniques (including microscopy,
profilometry, contact angle, differential scanning calorimetry, Fourier transform infrared
spectroscopy and X-ray photoelectron spectroscopy) and electrical characterization techniques
(including dielectric/impedance spectroscopy, thermally stimulated depolarization current
measurements, current voltage measurements, and high voltage dielectric breakdown). This
chapter is concluded with an overview of analytical techniques used to handle data.
Chapter 3 focuses on preliminary work done to determine the effect of the interface on
high voltage dielectric breakdown and ionic transport in pure PVDF hot pressed multilayer
laminates. The effect of multilayer lamination on PVDF crystal phase is addressed first. Dielectric
breakdown experiments at room temperature are then performed on 1- and 3- layer laminates.
Finally, an equivalent circuit model is developed to describe permittivity and loss tangent data at
70oC over a broad frequency range in 1- and 4-layer laminates.
Chapter 4 addresses the impact of surface chemical modification in P(VDF-TrFE). The
result of plasma treatment on P(VDF-TrFE) surface chemistry and morphology is analyzed as a
function of plasma treatment exposure time. Low and high field electrical properties are then
probed using dielectric spectroscopy, and current v voltage charging measurements. An analysis
by Poole-Frenkel and Schottky conduction theories inform on the mechanism of charge transport
through plasma modified P(VDF-TrFE). Finally, the Schottky equation is parametrically explored
where it is determined the Schottky barrier height dominates changes in conduction. Assumptions
involving measured plasma modified layer thickness as well as dielectric permittivity as a function
of plasma treatment conditions are used to simplify the Schottky equation and calculate the change
in barrier height caused by plasma treatment.
Chapter 5 focuses on the analysis of high field current voltage data in plasma treated PI.
Analysis is leveraged from results in Chapter 4 which demonstrate the interface’s importance on
limiting conduction. Current-voltage data analysis is performed by implementing hopping, Poole-
Frenkel, and Schottky conduction theories. Standard non-linear regression techniques combined
with bootstrap statistics are proposed as a method to extract the statistical significance of parameter
22
estimates during fitting, as well as identify breaks in the model’s ability to describe the data’s
behavior.
Chapter 6 is broken into two sections. Section 1 provides an in-depth analysis of how
material structure impacts impurity ion conduction of low quantities of Li+ through single layer
P(VDF-TrFE) films. Behavior of the materials curie transition temperature as a function of LiClO4
content in doped films is correlated to impedance, permittivity, and loss tangent over a broad
frequency range (105 Hz – 10-1 Hz) and temperature range (25oC – 110oC). Observations in data
are complemented by parameter estimates extracted from an equivalent circuit (EC) model that
describes polarization mechanisms associated with the material. Section 2 incorporates results
from chapter 5 to create a multilayered P(VDF-TrFE)/PVA laminate system exhibiting controlled
ionic conductivity. Concepts from impedance spectroscopy and EC modeling of doped 1-layer
films are applied to multilayered films to model the low field conduction behavior of multilayer
composites. Thermally stimulated depolarization current (TSDC) measurements focus analysis on
quasi DC ionic conduction in multilayered and single layered films. Depolarization mechanisms
are quantified using Bucci-Fieschi theory, providing a frame work for the mechanism responsible
for charge deflection in layered dielectrics. The section is concluded with high voltage dielectric
breakdown experiments to compare between 1-layer and 4-layer dielectrics of pure copolymer.
Chapter 7 outlines important conclusions and scientific contributions of this dissertation.
The section is concluded with suggestions for potential future work.
23
CHAPTER 2
MATERIAL PROCESSING, CHARACTERIZATION, AND DATA ANALYSIS
TECHNIQUES
2.1 INTRODUCTION
The contents of this chapter offer insight into processing procedures, characterization
techniques, and analytical techniques, however its contents are not imperative to understanding
subsequent chapters. The informed reader may skip this section and proceed with Chapter 3.
Contents of Chapter 2 address the following subject areas:
Processing Procedure Development – The processing procedures implemented in the
fabrication of hot pressed and spin casted samples in this dissertation are reviewed. An analysis to
determine quality of film produced as well as its repeatability is detailed in sections 2.2.1 and
2.2.2. Sections 2.2.3 explains the choice of electron beam evaporation as the electrode deposition
technique over sputtering and 2.3.4 details the specifics regarding plasma surface modification of
PI and P(VDF-TrFE).
Materials Characterization Equipment and Measurements – In this section, equipment and
the associated methods involved in data acquisition and analysis are reviewed. Section 2.3.1
presents characterization techniques used to analyze PVDF dielectric film bulk structure while
2.3.2 focuses on characterization techniques implemented to analyze film surfaces. The last section
2.3.3 gives a more detailed description of electrical characterization techniques and understanding
the electrical properties of PVDF and PI films at different frequencies and temperatures. Each
section includes a brief introduction to experimental and equipment set up involved in successful
implementation of the technique. From there, theory developed to describe data acquired using the
technique is discussed. Each technique has many different theories surrounding the production of
data and its interpretation, however only those theories implemented in this dissertation are
selected for discussion.
Special Data Analysis Techniques – Data interpretation required the use of advanced
analytical techniques to describe the behavior of electrical data. Section 2.4.1 reviews the
essentials of bootstrap statistics used to quantify parameter estimate significance during hopping
theory analysis of PI conduction in Chapter 5. In section 2.4.2, theoretical derivation of EC
24
impedance used to describe impedance spectroscopy data is reviewed, along with brief explanation
of the software used in complex nonlinear regression. Finally, section 2.4.3 reviews the concept
of peak deconvolution used to fit coalescing TSDC current peaks in P(VDF-TrFE) samples. Fitting
scripts were written in R-Studio and codes used in the analysis of data are annotated and presented
in their respective sections.
2.2 PVDF PROCESSING AND SAMPLE PREPARATION
The methods used to fabricate dielectric films are centered around solution casting
techniques. The following sections provide the specifics associated with solution casting and hot-
pressing PVDF films in greater detail than what is included in subsequent chapters.
2.2.1 PVDF Casting and Hot-Press Lamination
Single layer films of PVDF are fabricated by using a Doctor Blade casting tool depicted in
Figure 2-1. PVDF powder provided by Arkema is dissolved in anhydrous N,N
Dimethylformamide purchased from DrySolv and magnetically stirred for 2 hours. The solution is
then poured onto a glass plate previously cleaned using first IPA then Acetone. A Doctor Blade is
then used to spread the solution into a thin film which is then dried under vacuum for 1 hr, resulting
in a free-standing film. Thickness uniformity of the free-standing film is assessed by using a
Heidenhain Dial Gauge with 10-7m precision. Measurements were taken across the surface of an
approximately 5in x 7in casted film with approximately 1-2 cm between measured spots. Although
the average thickness of the film varied depending on solid wt% and blade height, each film
demonstrated reasonable uniformity in its thickness: standard deviations in film thickness fell
within the range of 3.9μm – 0.5μm for a film with average thickness of 79.3μm and 14.0μm
respectively. Thus the standard deviation in thickness of the entire film never surpasses 7% of the
average film thickness. Additionally, electrical measurements are taken using circular electrode
diameters of 1cm which is too small to capture local deviations in film thickness; therefore
thickness fluctuations were not though to be a significant contributor to electrical measurement
error in 1-layer casted films.
Hot pressed laminates were then constructed by stacking individual layers into multilayer
stacks ranging from 2 – 4 total layers. Two pieces of extruded polyimide Kapton films provided
25
Figure 2-1: PVDF processing procedure showing a) magnetic stirring (10-15% solid wt), b) solution degassing, c)
solution casting using Dr. Blade, d) film drying under vacuum 180oC for 1 hr stepped down to 60oC for 3 hrs with
final 1-layer film, e) stacked multilayer sandwiched between Kapton protective sheets and f) hot-pressing at 18-24
MPa at 150oC for 30 min.
Table 2-I: calculated average film thickness, standard deviation and thickness variation in hot pressed multilayers. 1-
, 3- and 4- layer stacks consisted of ~8 micron thick single layer films, while the 2-layer stack contained two ~13
micron films.
by Dupont sandwich the PVDF stacks as a protective non-stick barrier to the metal of the hot press
platens. The sandwich structure was then inserted into a uniaxial hot press with its platens held at
150oC. Pressing took place for 30 min under constant pressure (~20MPa). Hot-pressed films had
dimensions of approximately 5 cm x 7 cm, and thickness uniformity across the multilayered
laminates was measured in the same way as described for 1-layer films. Results of thickness
measurements for hot-pressed films 1- through 4-layers is presented in Table 2-I. Hot-pressing
increases the % variation of the thickness across the film which is thought to be due to a number
of different influencing factors: uneven pressure applied to the stack due to mis aligned platens,
minor variation in 1-layer thickness becoming more pronounced when stacking, fluctuating
26
pressure during the press (measured between 17 and 24 MPa), and wrinkling in the films. For these
reasons, each hot-pressed film was carefully examined prior to electroding to ensure only the most
uniform spots were used for measurements.
2.2.2 P(VDF-TrFE) Spin Casting
Challenges associated with repeatability in the hot-pressing procedure as well as limitations
on achievable thickness of cast PVDF films motivated the development of spin cast processing
protocols. Solutions of DMF and P(VDF-TrFE) were made using the processing procedures
outlined in Figure 2-1a and 1b. Figure 2-2 shows the spin cast processing and thickness
measurement procedure. Three solution batches were made with different copolymer solid wt% of
15%, 7.5% and 3%. The solution was then poured onto a 10.16 cm diameter silicon wafer and spun
for 50s at varying spin speeds ranging from 500 rpm – 2000 rpm. After spinning, wet wafers were
placed onto a hot plate exposed to atmosphere at a temperature of 100oC and dried for 15 min,
resulting in a thin P(VDF-TrFE) film adhered to the wafer.
The method used to assess film uniformity is shown in the schematic Figure 2-2. Thickness
measurements were performed by scanning a stylus profilometer over a step edge created by gently
scratching into the film with a razor blade. 10 step edges were created by scratching and each step
edge was scanned 3 times, resulting in a total of 30 scans per film. Two separate samples spun at
1,000 rpm using the 3% wt solution were fabricated to determine the repeatability in the processing
conditions. The results of thickness measurements across both samples are presented in Table 2-
II. It was found that the mean thickness of each sample was quite close (139 and 142 nm
respectively) with standard deviations (6 nm) of the mean thickness overlapping each other. An
unpaired Student’s T-test was performed using Microsoft Excell’s built in T-Test function to assess
the statistical significance between the two sample’s means. The analysis yielded a p value of
Figure 2-2: Spin cast procedure depicting a) 3 – 15 wt% P(VDF-TrFE) solution deposition, wet film drying at 100oC
for 15 min and c) schematic of dried film with 10 scratched in profilometer scan areas.
27
Table 2-II: Profilometer scan thicknesses t for two films processed at 3% solid wt., 1,000 rpm measured as described
in Figure 2-2.
0.035 which is less than 0.05 indicating the null hypothesis should be rejected and that there is a
statistically significant difference in sample thickness for the films processed at 1,000 rpm for 50s.
While blade casted films can be controlled by solution viscosity and blade height, spin
casting introduces a third parameter that can be used to control film thickness and uniformity:
wafer rotational speed. In order to understand the effect of solution viscosity (controlled by solid
wt %) and spin speed on resultant film thickness and uniformity, a spin cast parameter study was
performed within the range of copolymer solution wt% used in this dissertation. Results are shown
in Figure 2-3, suggesting a strong dependence on both solution wt% and wafer spin speed on
resultant film thickness. Due to statistical differences in individual films processed under the same
cast parameters (shown in Table II), the trends in copolymer thickness as a function of spin speed
and solution wt% are taken only as a general guide. A film thickness measurement protocol
(discussed in section 2.2.4) independent of the data shown in Figure 2-3 was developed and
implemented for each spin casted sample used in this dissertation prior to interpreting electrical
measurements.
2.2.3 Electrode Deposition
Methods used to deposit electrodes onto dielectric films for electrical measurements vary
depending on the material selection and desired measurement. As mentioned in the introduction
section of this dissertation, the polymer film capacitor industry typically uses metallization as the
electrode deposition process involving the “spraying” and adhesion of the electrode material
directly to the polymer film. One method typically employed in industry and research is RF
28
Figure 2-3: Spin cast parameter study varying spin speed (rpm) and solution solid wt%. a) contains data for 15 and
7.5% wt films while b) shows thicknesses measured for 3% wt thin films.
magnetron sputtering which utilizes a gaseous plasma to erode the target material via ionic
bombardment. This technique is inadequate for this work due to the presence of an active gas
plasma during the deposition process and PVDF’s surface sensitivity to plasma treatment [17].
The work in this dissertation implements electron beam evaporation using a Lab-18 tool provided
by Kurt J. Leskar instead of RF magnetron sputtering. In this technique, a beam of electrons is
focused on a crucible containing the electrode material of choice causing evaporation (or
sublimation) and consequentially deposition onto the substrate. Since the sample is held under
high vacuum (~1x10-6 Pa) in absence of reactive chemical species, metal/dielectric interactions
uncharacteristic of the two materials naturally coming in contact are prevented.
Pure PVDF films were used for hot-press lamination and ranged in thickness between
approximately 10 - 35μm depending on desired hot-press film thickness. After processing, these
films were cut into small squares and mounted onto an aluminum shadow mask using Kapton tape.
Electrode size was 1 cm in diameter. The electrode material used was Ag, deposited 100 nm thick
at a deposition rate of 2 A/s with the sample stage held at 0oC to prevent sample damage from
heating during deposition.
P(VDF-TrFE) films varied in thickness between thin (1 μm) and thick (10 μm) films. Thick
films of P(VDF-TrFE) were electroded using the same deposition parameters and shadow mask as
pure PVDF films. Typically films of 10 μm were intended for TSDC measurements, and thus Au
was used as the electrode material because of its good adhesion and chemical inertness under
poling conditions relative to other common materials such as Ag and Al, as demonstrated by
29
literature [12]. Thin films of 1 μm in thickness used electrode masks that were fabricated by laser
cutting circular holes into plastics such as mylar and taped down to the surface of the film.
Electrode diameters ranged from 1 mm – 3 mm. Due to the small electrode diameter, shadowing
during electrode deposition occurred. To account for shadowing, a Nikon Profile Projector V-12
optical comparator was used to approximate electrode diameter after evaporation and adjust
electrode area in calculations for permittivity and current density.
2.2.4 Plasma Treatment
A significant portion of this work is dedicated to understanding the influence of chemical
modification at the electrode/dielectric interface on high field conduction and charge injection in
P(VDF-TrFE) and PI. Reactive plasma treatment is used to chemically alter dielectric film surface
chemistry prior to electrode deposition. This process involves the generation of a plasma using RF
energy within a process reactor to ionize the reactive gas. A mixture of ions, radicals, neutral
species and charged species (electrons and protons) are accelerated in the direction of the sample
surface via application of an electric field. Decision of etching tool used is dependent on the desired
application. For example, tools such as the Alcatel Speeder 100 Si and SiO2 feature Bosch high
frequency fast etching and low frequency etching for silicon removal processes. Other tools such
as the Plasma-Therm 720 and Plasma-Therm Versalock exhibit a wider repertoire of etch material
capabilities that cover etching of silicon based dielectrics, metals (Au, Cr, Ti, Al), semiconductors
(GaAs, InP, InGaAs, AlGaAs, GaMnAs, poly and a-Si) and polymers (BCB, lift off resist, photo
resist, parylene, PDMS, etc.).
Equipment used for surface modification of organics must graft foreign moieties to the
dielectric surface without damaging the quality of the film surface and bulk. This requirement
makes systems such as the Plasma-Therm line of reactive ion etchers unsuitable for surface
modification in this study. An M4L RF gas plasma system provided by PVA TePla was used as
the reactive ion etching system in this study. The M4L is does not feature a strong plasma/sample
stage potential drop characteristic of the other previously mention systems, making the tool ideal
for the cleaning of organic surfaces and plasma surface modification where polymer film damage
is minimized. Plasma treatment parameters for the surface modification of P(VDF-TrFE) and PI
are presented in Table 2-III. Plasma treatment conditions were optimized for surface modification
for PI prior to data collection described in Meddeb et al. [13] and are not the focus of analysis of
30
Table 2-III: Plasma treatment parameters used on P(VDF-TrFE) and PI in this work
high field conduction in plasma treated PI films in this dissertation. Since analysis of PI films in
subsequent sections is leveraged off work previously done by Meddeb et al. [13], no further
discussion on plasma’s influence on PI is mentioned here.
Plasma treatment of P(VDF-TrFE) was performed as a function of treatment time ranging
from 45 – 180s shown in Table 2-III. A study measuring the etch rate of the treatment as a function
of treatment time (schematically shown in Figure 2-4) was performed to understand its effect on
thickness and surface heterogeneity of thin 1 μm copolymer films. A series of copolymer thin films
were deposited onto 1.25in x 1.25in square silicon wafers via spin casting 7.5% wt solution at 600
rpm for 50s. A step edge was gently scratched along the center and scanned with a stylus
profilometer confirming the film’s thickness to be 1μm. The film was then placed into the M4L
and administered plasma treatment using parameters specified in Table 2-III for P(VDF-TrFE) at
30s time intervals. After each 30s exposure its thickness was measured to develop film thickness
as a function of plasma exposure time presented in 2-5. Linear regression using Microsoft’s linear
regression function indicates film thickness obeys a linear relationship with plasma exposure time.
The slope represents the etch rate calculated to be 0.74 nm/s which must be accounted for during
film thickness dependent calculations. Although no significant change in the spread of
measurement as a function of treatment time was calculated (as indicated by nearly unchanging
standard deviation in measured thickness), the possibility that plasma interaction with copolymer
surface causes heterogeneity in surface roughness of the film remained a possibility. This was
studied using optical profilometry techniques and is discussed later in Chapter 4 section 4.3.4.
31
Figure 2-4: Schematic showing stepwise process of determining plasma process etch rate. a) film deposition onto
square silicon substrate, b) scratched groove spanning substrate, c) profilometry over step edge (5 scans total) d)
application of 30s plasma treatment (see Table III for parameters), and e) profilometry on treated film. Steps e – d are
repeated 3 times.
Figure 2-5: Copolymer film thickness as a function of plasma treatment exposure time. A calculated R2=9.97 indicates
a linear relation between film thickness and exposure time where an etch rate of 0.74 nm/s can is extracted.
32
2.3 MATERIAL CHARACTERIZATION EQUIPMENT AND METHODS
The following section presents material characterization methods used in material
characterization found in this dissertation. Each analysis technique is first presented from a
theoretical perspective where the sample set up and basic underlying physics is explained. Each
section ends with a brief review of the associated technique’s implementation for PVDF or PI
characterization when appropriate.
2.3.1 Bulk Chemical Characterization
2.3.1.1 Differential Scanning Calorimetry (DSC)
Processing’s effect on the crystal structure of pure PVDF and P(VDF-TrFE) films was
analyzed thermally using DSC. The experimental set up involves two identical Al pans, one
containing the test material and the other left empty as reference. The experiment measures heat
flows in and out of the sample in relation to the reference pan at a constant temperature ramp rate
(chosen to be 10oC/min). Heat flow can be approximated as enthalpy changes since the experiment
is performed at constant pressure:
𝑑𝑄
𝑑𝑡 𝑝=
𝑑𝐻
𝑑𝑡 (2 − 1)
Here, Q is heat, t is time, and H is the enthalpy. For a given sample in relation to the reference pan,
the change in enthalpy over time is defined as the following:
Δ𝑑𝐻
𝑑𝑡=
𝑑𝐻
𝑑𝑡 𝑡𝑒𝑠𝑡 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙−
𝑑𝐻
𝑑𝑡 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 (2 − 2)
As temperature of the experiment (dictated by temperature ramp rate) approaches a phase
transition associated with the test material’s structure, the sample will begin to consume or release
heat causing a change in enthalpy. This results in a peak in the DSC signal. Integration of the
change in sample enthalpy over the phase transition peak time will then yield enthalpy change of
the sample shown below.
Δ𝐻𝑠𝑎𝑚𝑝𝑙𝑒 = ∫𝑑𝐻
𝑑𝑡 𝑠𝑎𝑚𝑝𝑙𝑒
𝑡−𝑝𝑒𝑎𝑘 𝑒𝑛𝑑
𝑡−𝑝𝑒𝑎𝑘 𝑜𝑛𝑠𝑒𝑡
𝑑𝑡 (2 − 3)
In the context of performing DSC on polycrystalline organic materials, peaks recorded in the DSC
are due to crystalline domain phase transitions such as curie transitions in ferroelectric phases,
crystal melting, or re-crystallization (nucleation and growth). The % crystallinity of the sample
33
can be calculated by comparing integration of the melting peak to a theoretically derived melting
enthalpy ΔHo pertaining to the material’s crystal phase:
∆𝐻𝑠𝑎𝑚𝑝𝑙𝑒
∆𝐻0× 100 (2 − 4)
Since the initial amount of material loaded into the Al pan has a mass associated with it which will
affect the intensity of peaks in the DSC signal, calculations of crystallinity are normalized to
sample mass, ensuring accurate results (where ΔHsample and ΔHo have units J/g).
DSC’s sensitivity to polymer phase transformation has made the technique appealing for
work involving structural characterization of PVDF in relation to processing, specifically
mechanical deformation. One example is a study done by Lanceros-Méndez et al. [97] which
targeted structural changes occurring (crystallinity % and lamellae thickness) during mechanical
deformation in β-PVDF as a function of strain. Other research uses DSC to understand how
graphene oxide inclusion into the PVDF matrix as well as defect formation post exposure to
ionizing radiation (gamma) affect polymer crystalline structure. The technique is commonly used
in P(VF-TrFE) systems as well, informing on how nanofiller concentration in P(VDF-TrFE) films
effects crystalline phase % and order [98] and how poling and annealing conditions impact
ferroelectric phase formation [99]. Finally, DSC has also been implemented in the study of how
polymer crystallinity influences electrical properties of PVDF such as molecular mobility [100],
motivating its use to link polymer crystal structure to low and high field conduction properties in
this dissertation.
2.3.1.2 Fourier Transform Infrared Spectroscopy FTIR
Polymer phase development as a result of processing parameters in PVDF and P(VDF-
TrFE) is analyzed using FTIR. The technique involves the exposure of a test sample to a broad
frequency range of infrared radiation where either a) molecular bonds absorb incoming radiation
of a given energy dependent on its structure or b) transmission of IR radiation through the sample
occurs. A unique absorbance spectrum characteristic of the test material is formed depending on
crystal phase quantity, structure, and chemical bonding affiliated with the pristine material. The
characteristic bands pertaining to vibrational modes and crystalline phases of PVDF is presented
in Table 2-IV at the end of the section.
34
The process by which data is collected begins with a coherent black body radiation source
emitting in the IR wavelength spectrum (ranging from ~2.5x10-5 – 1.67x10-6 m corresponding to
400 cm-1 and 6000 cm-1 respectively). Radiation is passed into a Michaelson interferometer
containing one adjustable mirror panel of controlled position. Mirror panel movement causes the
periodic interference of radiation resulting from wave superimposition, thus modulating the
spectrum leaving the interferometer. This modulated beam is then exposed to the sample in which
its transmission or absorption is sensed by a detector as a function of the adjustable mirror panel’s
position. In order to convert light absorption/transmission and mirror panel position to light
absorption/transmission and radiation wavelength, a computer program performs a Fourier
transform. Mathematically this is represented by the following relation:
𝐺(𝑥) ↔𝐹
𝑔(𝜆) (2 − 5)
where the functions G(x) (a function of position x) and g(λ) (a function of wavelength λ) form a
Fourier pair and relate via the given relation.
𝐹[𝑔(𝜆)] = 𝐺(𝑥) = ∫ 𝑔(𝜆)𝑒−𝑖2𝜋𝑥𝜆 𝑑𝜆
∞
−∞
(2 − 6)
𝐹[𝐺(𝜆)]−1 = 𝑔(𝜆) = ∫ 𝐺(𝑥)𝑒−𝑖2𝜋𝑥𝜆 𝑑𝑥
∞
−∞
(2 − 7)
The results of the experiment are typically presented as a function of % transmission or %
absorption at the discretion of the author as a function of wavenumber with units cm-1.
Sensitivity to the abundance and conformation (during polarized measurements) of
crystalline phases in PVDF makes FTIR a powerful tool to determine bulk structure. Lanceros-
Méndez et al. [97] complemented the analysis of β-phase crystal development using DSC with
polarized FTIR in parallel and perpendicular modes. Measurements indicated a strong dependence
of polymer chain alignment direction within the crystalline phase on the axis of applied tension
while uniaxial drawing. Work done on the characterization of composite dielectrics formed by
compression molding P(VDF-TrFE) with natural rubber latex in the absence of solvent
implemented FTIR to measure the degree of interactions between P(VDF-TrFE) and additives
[101]. The analysis was able to conclude that there is no strong interactions between polymer and
additive materials due to an absence of absorption band intensity and peak wavenumber. Finally,
FTIR has also been used to understand how PVDF chain chemistry associated with CF2 chemical
35
groups impacts Li+ cation complexation in solid and wet polymer electrolytes [102]. In this regard,
the technique is seen as a powerful tool to link ionic interactions with polymer structure and
improve understanding of ionic conduction in polymer films.
Table IV: Characteristic features of PVDF FTIR spectra by wave number, adopted from Lanceros-Mendez et al. and
other contributing authors [25, 26, 27, 28, 29, 30, 31].
36
2.3.2 Surface Chemical Characterization
2.3.2.1 Optical Profilometry
Optical profilometry (also referred to as “optical profiling”) is an optical technique used to
measure variation in surface height of a sample over a specific scan area. In this technique, light
of a known wavelength is used as the source probing the sample surface, thus allowing for very
high resolution (~1 nm) in the vertical direction when measuring film properties such as surface
roughness. The optical profiler uses similar methods to detect variation in surface height as the
Michaelson interferometer (section 2.3.1.2). In this set up, the adjustable mirror is a light source
detector that is moved creates constructive and destructive interference with light reflected off the
test sample’s surface. When the distance from the sample surface to the interferometer’s beam
splitter equals the distance of the detector to the beam splitter, constructive/destructive interference
occurs. Optical path differences between sample and beam splitter and detector and beam splitter
are due to variations in height across the sample surface. Adjustment of detector height allows
focusing as a function of position along the sample surface, enabling reconstruction of sample
surface topology within the scan area.
Controlling dielectric film surface morphology is known to be important in order to realize
high breakdown strength and low loss films. Research done by Burlingame et al. [103] uses optical
profilometry to understand physical surface defects in Polythiourea films in relation to solvent
used during processing. Increased surface roughness and irregularity correlated directly to reduced
characteristic breakdown strength calculated using Weibull statistics. Similar results were obtained
in BaO – Al2O3 – B2O3 – SiO2 glass thinned by hydrofluoric acid etching reported by Lee et al.
[104] where roughness measured via optical profilometry was correlated to reduced breakdown
strength as well as Weibull modulus. Results reported by Burlingame and Lee motivate use of
optical profilmometry in this study to determine the effect of plasma surface treatment on P(VDF-
TrFE) film topology and is discussed in Chapter 4 section 4.3.4.
2.3.2.2 H2O Water Contact Angle
The contact angle between a liquid and sample surface quantifies the wettability of the
tested material. Test set up typically employs a syringe filled with the test liquid (in this case H2O)
connected to a hydraulic pump capable of depositing droplets in the range of 1 – 10μL. A water
37
droplet of known volume is deposited onto the surface of the test sample and imaged using a
camera and computer software that tracks the contact angle as a function of time.
The liquid’s contact with the solid surface is dictated by intermolecular forces arising from
sample surface chemistry as well as test liquid chemistry. A simplified model of the interaction
between test liquid and sample surface is the Young relation:
𝛾𝑆𝐺 = 𝛾𝑆𝐿 + 𝛾𝐿𝐺 cos 𝜃 (2 − 8)
where γSG is the interfacial energy between the solid phase (sample surface) and vapor phase
(liquid), γSL the solid-liquid interfacial energy, γLG the liquid-vapor interfacial energy, and θ is the
angle created between the liquid’s contact with the surface and the solid phase. Since the interfacial
energy is a quantification of molecular bond disruption occurring upon creation of a surface, the
contact angle θ is sensitive to surface state of the test liquid as well as sample. The contact angle
is also highly dependent on physical characteristics of solid surfaces including crystallinity, grain
size and shape, porosity, and surface roughness. Although the nature of how surface roughness
impacts wetting properties during experimentation is not well understood, effective surface area is
known to control surface energy, and thus roughness is determined to be a non-negligible
parameter contributing to the contact angle.
Literature using contact angle as a characterization tool covers wide range of applications
such as the surface characterization of rocks and minerals, and understanding wetting properties
in relation to sand blasted PMMA surface roughness to control cell adhesion and migration [105]
[106]. Some of the most pertinent literature surrounding the use of contact angle is the calculation
of surface polarity before and after plasma surface treatment in a variety of common polymers.
Table 2-V showing the polar component divided by the dispersive component of the free energy
extracted from literature is presented below and highlights contact angle’s sensitivity to surface
chemical properties in organic materials.
38
Table V: Surface polarity calculated for a number of common polymers using contact angle experiments, depicting
influence of plasma treatment on surface properties [32, 33, 34, 35, 36, 37, 38].
2.3.2.3 X-ray Photoelectron Spectroscopy
XPS is a universally used chemical characterization technique because of its applicability
to a broad range of test materials as well as its ability to provide quantitative information on the
chemical state of the surface of materials being studied. During experimentation, the sample
surface is excited using a monochromatic X-ray source (typically Al kα x-rays at 1486.6 eV) of
given energy hν which interacts with bound electrons at the sample’s surface. Inelastic scattering
events result in the liberation of photoelectrons from the material surface with kinetic energies KE
characteristic of the chemical species and surrounding molecular environment. The KE is related
to the binding energy of the detected electron Eb via the following relationship where φspec is the
XPS spectrometer’s work function:
𝐾𝐸 = ℎ𝜈 − (𝐸𝑏 + 𝜑𝑠𝑝𝑒𝑐). (2 − 9)
39
Spectral data is plotted as a function of Eb displaying peaks which correspond to the shell
(1s, 2s, 2p, etc.) by which the detected photoelectron originates. Similarly, high resolution scans
within the vicinity of a detected peak can be performed, revealing not only peak position on the Eb
axis, but also its salient features such as peak convolution or shifting. Peak shifting can be
indicative of the electronegativity of atoms such as fluorine in the local environment or electron-
nucleus attractions creating screening effects. Similarly, complexity in peak shapes are
characteristic of conformational and chemical characteristics of moieties containing elements that
emit photoelectrons within the specified energy range.
Data analysis of XPS spectra is performed by peak fitting, where synthetic peaks are
introduced to best fit XPS spectra. Information about chemical moieties present as well as
elemental %’s can be obtained through careful selection of peak position and integration. Accuracy
of data reduction by peak fitting depends on the following: 1) photoelectric spectral line
assignment validity (dependent both on equipment calibration and expertise of the scientist in
chair), 2) background signal treatment and 3) correctness of line shape implemented in the fitting
[107]. Much of the analysis involved in this dissertation involved peak fitting of the C1s spectra
allowing for a linear background to be used in fitting which is deemed adequate by past work done
by Beamson and Briggs [108]. Choice in line shape is more complex due to the line shape’s
dependence on a number of instrumental and physical factors: response function of electron
analyzer, X-ray line shape profiles, intrinsic lifetime broadening of core-level hole states, phonon
broadening, differential surface charging, and contributions from surface core-level shifts [109]
[107]. XPS fitting can be done using a number of different line shapes to account for these effects,
which are reviewed in Fairly [107], however only practical Gaussian and Lorentzian line shape
functions were used in this dissertation because of their simplicity and no demonstratable need for
more complex analytical interpretation. Casa XPS was used as the peak fitting software for XPS
spectra and employs the following Gaussian / Lorentzian functions for synthetic fitting:
𝐺𝐿𝑃(𝑥, 𝐹, 𝐸,𝑚) =𝑒𝑥𝑝 [−4𝑙𝑛2(1 − 𝑚)
(𝑥 − 𝐸)2
𝐹2 ]
1 + 4𝑚(𝑥 − 𝐸)2
𝐹2
(2 − 10)
𝐺𝐿𝑆(𝑥, 𝐹, 𝐸,𝑚) = (1 − 𝑚)𝑒𝑥𝑝 [−4𝑙𝑛2(𝑥 − 𝐸)2
𝐹2] +
𝑚
1 + 4(𝑥 − 𝐸)2
𝐹2
(2 − 11)
40
Both equations are characterized by a mixture of Gaussian and Lorentzian line shapes,
whose contributing quantity is controlled by the value of m. Other parameters involved are x
corresponding to the abscissa of a data point, E is the peak center, and F is the full width half max
parameter. Due to the symmetry of the GLP and GLS, asymmetries in the fit C1s signal relative to
control samples during analysis are assumed to arise due to introduction of foreign chemical
species and are fit accordingly in context of data interpretation in Chapter 4 section 4.3.2.
XPS has been used in past research to better understand the effect of various surface
treatments on polymer surface chemistry. One report by Vandencasteele et al. [110] implements
the technique to learn how fluorine containing polymers (PTFE, PVDF and PVF) react to N2 and
O2 rf plasma surface treatments. It was found that chemical species grafted to the surfaces was not
only gas plasma chemistry dependent, but also material dependent: O2 serving as an etchant to
PTFE while O uptake at the surface was detected in other tested films. Similarly, it was found that
fluorinated materials resulted in a de-fluorination at the sample surface after plasma treatment,
coinciding well with past work done by Duca et al. [111] that reports a similar phenomenon in
PVDF exposed to Ar plasma. Finally, XPS has been used to understand plasma treatment’s role in
surface modification of PVDF as a substrate for polyamide thin film composite membranes [112].
Despite the quantity of research done implementing XPS to better understand wetting and adhesion
properties of plasma treated organics, gaps exist in the literature concerning surface chemical
characterization in plasma treated PVDF dielectrics for high field applications which is addressed
in Chapter 4.
2.3.2.4 Time of Flight Secondary Ion Mass Spectrometry
ToF-SIMS is a destructive chemical characterization technique unlike XPS in which a
pulsed ion beam is used to liberate molecules from the sample surface. Sputtered material is
accelerated into an analyzer used to measure the mass of ions and clusters emitted from the sample.
Molecular and elemental identities can then be determined from the intensity of the detected signal
and exact mass of measured chemical constituents. ToF-SIMS has analysis depth capabilities on
the order of 2nm making it an ideal technique for surface chemical characterization, however its
ability to sputter test material enables chemical depth profiling. In this dissertation, ToF-SIMS is
used as a complimentary surface chemical analytical tool to XPS and to depth profile plasma
treated P(VDF-TrFE) thin films to estimate the length scale that M4L plasma modification affects
41
the material. This analysis is discussed when implementing Schottky theory to quantify Schottky
barrier height change after plasma treatment in Chapter 4 sections 4.3.3 and 4.3.6.2.
2.3.3 Electrical Characterization
2.3.3.1 Impedance Spectroscopy and Equivalent Circuit Modeling
Impedance spectroscopy is a low voltage analytical technique in which a sinusoidal voltage
is applied across a test material and charge measured as a function of signal frequency. The
impedance of the test material can be described using Ohm’s law via the following relationship
between impedance Z, voltage V, and current I:
𝑍 =𝑉
𝐼. (2 − 12)
The applied AC signal used is small (in the range of 10mV – 1V) so that the sample’s generated
current follows a linear relationship with the applied excitation signal. In this scenario, I will follow
V with a given phase shift φ:
𝑉𝑡 = 𝑉𝑜 sin(𝜔𝑡) (2 − 13)
𝐼𝑡 = 𝐼𝑜 sin(𝜔𝑡 + 𝜑) (2 − 14)
where ω is angular frequency calculated by 2πf, t is time and φ is a phase shift depending on the
properties of the device under test (φ = 90o for an ideal capacitor and 0o for an ideal resistor). The
impedance can then be expressed as a function of time:
𝑍𝑡 =𝑉𝑜 sin(𝜔𝑡)
𝐼𝑜 sin(𝜔𝑡 + 𝜑) (2 − 15)
and in complex notation:
𝑍(𝜔) = 𝑍𝑜(cos(𝜔𝑡) + 𝑗 sin(𝜔𝑡)) (2 − 16)
where j is (-1)1/2. Due to Z’s dependence on I which is strongly influenced by microscopic material
structure, impedance spectroscopy is a powerful tool used to characterize how morphology of the
test sample impacts conduction mechanisms dominating various frequency ranges.
The analysis of dielectric materials typically links current propagation through the material
to the dominating polarization mechanism occurring at different test frequencies. The impedance
for a capacitor is given by the following equation:
𝑍(𝜔) =1
𝑗𝜔𝐶 (2 − 17)
where C is the capacitance and follows the relationship
42
𝐶 =휀𝑟휀𝑜𝐴
𝑡. (2 − 18)
Since the impedance of the material is dependent on material permittivity εr, dielectric conduction
is usually analyzed by accounting for the polarization mechanisms associated with device charging
and discharging within the range of test frequencies. Spectroscopy used in analysis of PVDF and
P(VDF-TrFE) capacitors in this dissertation was performed in the broad frequency range of 0.1
kHz – 100 kHz, capturing the following polarization mechanisms: 1) electronic polarization
contributing at the highest frequencies of applied signal (f > 100kHz). In frequency regimes
dominated by electronic polarization (1014 – 1016 Hz [113]) , the permittivity of the material is
assumed to be equal to the refractive index squared (n2) because the active polarization mechanism
is electron cloud density displacement surrounding atoms. 2) permanent dipole polarization
dominates the room temperature impedance response in PVDF materials within ~100 Hz – 1 MHz.
In this mechanism, polarization is induced by the rotation of C-F dipoles associated with the
polymer backbone in PVDF. 3) Bulk ionic conduction arising from the migration of impurity ions
at low frequencies (0.1 Hz – 10 Hz). At these frequencies of measurement, the conduction through
the material is not intrinsically generated via the material structure and causes dielectric losses and
leakage current at low frequencies. Finally, 4) blocking polarization due to the build-up of ionic
species at the electrode/dielectric interface occurs at quasi DC frequencies and high temperatures.
This polarization process is typically not defined as a bulk response of the material.
An attempt to link material structure to impedance behavior as a function of
frequency is typically made via equivalent circuit (EC) modeling where an EC of ideal electrical
components is constructed in order to describe each polarization and conduction mechanism
associated with the material. EC’s used to describe the behavior of polymeric materials under test
typically account for polarization mechanisms 1 – 3, and in simplicity adhere to a general form
depicted in Figure 2.6. Contributions to the impedance arising from electronic polarization are
described by C1 which assumes an ideally capacitive response and takes permittivity as n2. Induced
dipole polarizations are described by C2 (8 < εr < 12 for PVDF) in series with a current limiting
resistor R2 which controls the frequency at which the polarization mechanism relaxes out by the
product τ=C2R2. Finally ionic conduction through the material at low frequencies is modeled by
R3 and takes on a value representing the bulk resistance of the material under test. Values for each
circuit element are usually determined via a nonlinear regression procedure in which the functional
43
Figure 2-6: Simplified model for a dielectric material under impedance spectroscopy test, incorporating ideal
capacitive and resistive components to describe electronic, orientational, and ionic polarizations. The EC used in this
dissertation is leveraged from this and discussed in section 2.4.2.
equivalent impedance of the chosen model is parametrically adjusted to best fit impedance data
over the entire tested frequency range. This process is described in greater detail in section 2.4.2.1.
Impedance spectroscopy techniques are a staple in literature studying electrical
conductivity in solid state polymer electrolytes. One such work by He et al. [114] prepares battery
separator membranes out of P(VDF-HFP)/HDPE for enhanced ionic conductivity using non-
solvent induced phase separation. Impedance spectroscopy was used to determine enhanced bulk
conductivity in P(VDF-HFP)/HDPE blends relative to pure P(VDF-HFP) suggesting improved
conductivity and battery performance was due to expanded amorphous area and high porosity.
Other literature uses impedance spectroscopy to describe conductivity contributions from material
structure. Work done by Marzantowicz et al. [115] performs impedance spectroscopy on PEO
dielectrics around the melting temperature of the crystalline phase. It was found that the ionic
conductivity was controlled by the presence of the crystal phase, impedance spectra exhibiting
unique characteristics above and below the melting temperature Tm = 324.4 K. EC modeling was
then implemented to describe this behavior. The EC used by Marzantowicz et al. [115] was of
similar form to that presented in Figure 2-6, however modification involving the introduction of
a constant phase element (CPE) parallel with a resistor in series with R3 was used to account for
ionic charge/polymer crystal interaction. This was verified by the necessity of the CPE/R element
to describe impedance below the melt temperature and its absence when the material is molten.
Similar analysis is presented in Chapter 6I section 6I.5.1 which also includes discussion of fit
parameter statistics for P(VDF-TrFE) impedance spectra fitting.
44
2.3.3.2 Current Voltage (IV) Charging Experiments
Current-voltage (IV) experiments are a powerful tool in interpreting the mechanism by
which conduction occurs in organic dielectrics. Typically, a metallized dielectric is inserted into a
test fixture where a voltage source is connected to apply a DC bias (Vapp) across the material. The
potential is held for a sufficiently long enough time t that allows for the achievement of steady
state current after the emptying of occupied trap states and ionic migration relaxation. Hold times
required to achieve true steady state currents are beyond time scales practical for measuring I vs
V, and thus experiments measuring I(t) at constant V are performed prior to I(V) measurements to
obtain a reasonable hold time in which a quasi-steady state in I is achieved. Once this time is
reached, a pA meter measures current. The applied voltage is then linearly increased, and the
process repeated until the relationship between current and voltage over a range of Vapp is obtained.
Analysis of I(V) characteristics of polymer dielectrics is not a straightforward task and
usually involves multiple conduction mechanisms within a given measurement range. One factor
which affects the I(V) characteristics of the material is the amorphous / polycrystalline structure of
organics. Unlike single crystals, disordered materials have trap energy levels which are distributed
according to specific distribution functions [116, 117]. Furthermore, the contact properties
between dielectric and metallization create their own unique trap states and distributions, typically
following different mathematical behavior in comparison to those distributed within the material’s
bulk. Since the analysis of high field conduction currents in this dissertation involves measurement
of two polycrystalline materials (PI and P(VDF-TrFE)), both bulk dominated and interface
dominated conduction mechanisms are considered. The following is a review of the three theories
used to describe I(V) data of plasma treated PI and P(VDF-TrFE) in this dissertation.
Interface Dominated Conduction – Schottky Theory: High field current injection effects resulting
from electric field assisted thermionic emission or tunneling through a potential barrier separating
electrode from dielectric is described by Schottky theory [117]. The potential barrier height
existing between electrode and dielectric under the application of electric field E is defined as the
following:
Ψ(𝑥) = ϕ𝑚 − 𝜒 −𝑞2
16𝜋휀𝑟휀𝑜𝑥− 𝑞𝐸𝑥 (2 − 19)
45
where φm is metal work function of the deposited electrode, χ is the electron affinity of the
dielectric, q is elementary charge, and x is distance into the dielectric from the surface of the
material. The term q2/16πεr εox is a potential energy term accounting for the image force generated
upon the emission of a charge from the metal into the dielectric, and qEx accounts for barrier height
lowering under field E. These two terms represent competing forces on the emitted charge:
attraction back to electrode from the image force and repulsion away from the electrode due to
field assisted barrier height lowering. The point at which net force acting on the charge is a
minimum and can be obtained by dΨ(x)/dx = 0, representing the point x = xmin such that the barrier
height is a minimum:
𝑥𝑚𝑖𝑛 = (𝑞
16𝜋휀𝑟휀𝑜𝐸)
12. (2 − 20)
At the position xmin, Ψ(xmin) equals the effective potential barrier height φB and the lowering of the
barrier height due to the application of electric field ΔφB can be derived by the following:
Δ𝜙𝐵 = (𝜙𝑚 − 𝜒) − 𝜙𝐵. (2 − 21)
By relationship Ψ(xmin) = φB:
Δ𝜙𝐵𝑆 = (𝜙𝑚 − 𝜒) − [(𝜙𝑚 − 𝜒) −𝑞2
16𝜋휀𝑟휀𝑜𝑥𝑚𝑖𝑛− 𝑞𝐸𝑥𝑚𝑖𝑛], (2 − 22)
substituting for xmin and canceling like terms:
Δ𝜙𝐵𝑆 =𝑞2
16𝜋휀𝑟휀𝑜(
𝑞
16𝜋휀𝑟휀𝑜𝐸)
12+ 𝑞𝐸 (
𝑞
16𝜋휀𝑟휀𝑜𝐸)
12, (2 − 23)
and by algebraic simplification, the following relation for field assisted barrier height lowering is
obtained:
Δ𝜙𝐵𝑆 = (𝑞3𝐸
4𝜋휀𝑟휀𝑜)
12
= 𝛽𝑠𝐸12. (2 − 24)
In this equation, the βS parameter is called the Schottky constant and depends on material
permittivity εr. Since this theory is descriptive of charge emission occurring on the time scale of
electronic processes, εr is usually thought to be representative of n2 for the material in question
(similar to electronically dominated polarization processes as discussed in section 2.3.2.1),
however fitting Schottky conduction behavior of polar polymers has suggested εr values
accounting for dipolar orientational polarization mechanisms to be more appropriate based on
goodness of fit.
46
Thermionic emission processes link current generation in the material to temperature via
an exponential relationship similar to the Arrhenius relationship:
𝐽 = 𝐴𝑅𝑇2𝑒𝑥𝑝 [−𝜙
𝑘𝑏𝑇]. (2 − 25)
In this formalism, J is the emitted current density at temperature T, φ is considered the work
function that must be energetically achieved for emission to occur, and AR is Richardson’s
constant. The AR term is a material dependent constant assuming the form AR = 4πqemekb2/h3
relating current to electrical particle mass (me) and charge (qe), however its derivation is not
discussed in this dissertation. In context of dielectrics during the I(V) experiment, the field assisted
thermionic emission assumes a similar form to that of pure thermionic emission that accounts for
field enhanced barrier height lowering:
𝐽 = 𝐴𝑅𝑇2𝑒𝑥𝑝 [𝛽𝑠𝐸
12
𝑘𝑏𝑇] 𝑒𝑥𝑝 [−
𝜙𝑆
𝑘𝑏𝑇]. (2 − 26)
In this equation, φS describes the electrode/dielectric contact’s barrier height at E=0. In systems
displaying Schottky type behavior, the natural logarithm of the current density ln(J) plotted as a
function of the square root of electric field E1/2 will yield a linear relationship from which the
slope can be related to βS by the following:
𝑙𝑛(𝐽) =𝛽𝑆𝐸
12
𝑘𝐵𝑇−
𝜙𝑆
𝑘𝐵𝑇+ 𝑙𝑛(𝐴𝑅𝑇2). (2 − 27)
Implementation of this relationship in the analysis of I(V) data will be discussed in context of
plasma treated P(VDF-TrFE) thin film and PI high field conduction analysis in Chapter 4 and 5
respectively.
Bulk Limited Conduction – Poole-Frenkel Theory: Poole-Frenkel (PF) charge emission is
descriptive of the filed enhanced thermionic emission of charges injected into the dielectric from
trap states distributed throughout the bulk of the dielectric under test [117]. The physics involving
charge emission in PF theory is very similar to that of Schottky since they both describe emission
of an electronic carrier from a bound trap state imposing coulombic interaction between escaping
electron and a positive charge. While Schottky describes coulombic interaction via creation of a
mobile image charge, the positive charge center involved in PF emission is fixed, resulting in an
altered field enhanced barrier height lowering term:
47
Δ𝜙𝐵𝑃𝐹 = (𝑞3𝐸
𝜋휀𝑟휀𝑜)
12
= 𝛽𝑃𝐹𝐸12 (2 − 28)
where βPF is called to Poole-Frenkel constant and is larger than βS by a factor of 2. The current
density described by PF conduction follows a similar double exponential relationship as Schottky:
𝐽 = 𝐸𝜎𝑜𝑒𝑥𝑝 [𝛽𝑃𝐹𝐸
12
𝑘𝑏𝑇] 𝑒𝑥𝑝 [−
𝜙𝑃𝐹
𝑘𝑏𝑇] (2 − 29)
with σo being a term proportional to the conductivity of the material and φPF representing trap site
barrier height at E=0. Linearization of the current density divided by the electric field produces a
similar result to that of Schottky emission:
𝑙𝑛 (𝐽
𝐸) =
𝛽𝑃𝐹𝐸12
𝑘𝐵𝑇−
𝜙𝑃𝐹
𝑘𝐵𝑇+ 𝑙𝑛(𝜎𝑜). (2 − 30)
In this format, linearization of the data and fitting to the above equation enables computation of
βPF and thus εr of the test material.
The analysis of Schottky and PF type conduction mechanisms relies heavily on
computation of permittivity from linearized I(V) data for both mathematical and physical reasons.
Both theories are described by the multiplication of two exponentials, making linearization through
taking the natural logarithm a mathematically simple task. Physically, both conduction
mechanisms give rise to behavior defined by a natural logarithm of the conductivity ln(σ) that is
proportional to E1/2. Therefor validation of the active conduction mechanism is reliant on
comparison of the permittivity extracted by linear fitting to known values of the test material’s
permittivity.
Bulk Dominated Conduction – Hopping Theory: Hopping conduction theory is based off the
structure of polycrystalline and amorphous organic dielectrics: intrinsic defects due to structural
disorders and extrinsic defects relating to impurities associated with material fabrication processes
acting as trap sites and in some instances charge carriers creating unique conduction
characteristics. These defects contribute to intermolecular charge transport processes which are
described using hopping theory. Derivation of the current density as a function of applied electric
field begins with an equation accounting for diffusional dependent conduction through the test
material written as:
48
𝐽 = 𝜎𝐸 + 𝑞𝐷𝑜
𝜕𝑛𝑐
𝜕𝑥+
𝜕𝐷
𝜕𝑡, (2 − 31)
where D is the diffusion coefficient for migrating charge species, q is the charge of the associated
carrier (typically taken as the elementary charge) and nc is a concentration of contributing charge
carriers. Considering measurement protocol which mandates measurement at steady state currents
as a function of applied field, the term ∂D/∂t = 0. Similarly, space charge and electronic charge
neutrality within the dielectric is assumed to be true, making ∂nc/∂x = 0 as well. This leaves the
following expression:
𝐽 = 𝜎𝐸 = 𝑛𝑞𝜇𝐸 (2 − 32)
where material conductivity σ is represented by the typical product of carrier concentration, charge,
and mobility (nqμ). Probabilities associated with charge trap state escape through random thermal
fluctuation are then considered and are written as:
𝑃𝑇 = 𝜐𝑜𝑒𝑥𝑝 [−𝜙𝐻
𝑘𝑏𝑇]. (2 − 33)
The term νo is the carrier vibrational frequency and φH typically referred to as an activation energy
or barrier height/trap depth for the hopping conduction process [113] [118].
Similar to Schottky and PF conduction theories, application of electric field E causes field
assisted barrier height lowering to occur. In context of carrier hopping, this reduces the energy
required to escape in the direction of applied field by - qeEd and increases it opposite of applied
field by +qeEd where d is hop distance. This makes the probability of escape (PE) to be the sum of
probabilities associated with escape in the direction of applied field and against applied field,
creating the following expression:
𝑃𝐸 = 𝜐𝑜𝑒𝑥𝑝 [−𝜙𝐻
𝑘𝑏𝑇] [𝑒𝑥𝑝 (
𝑞𝑒𝐸𝑑
2𝑘𝑏𝑇) − 𝑒𝑥𝑝 (−
𝑞𝑒𝐸𝑑
2𝑘𝑏𝑇)] (2 − 34)
and is simplified by using the definition of the hyperbolic function to the following:
𝑃𝐸 = 𝜐𝑜𝑒𝑥𝑝 [−𝜙𝐻
𝑘𝑏𝑇] [2 sinh (
𝑞𝑒𝐸𝑑
2𝑘𝑏𝑇)]. (2 − 35)
Considering the relationship between charge drift velocity, PE and d defined by vD = PEd as well
as μ= vD/E, and general formula for conductivity σ=nqμ, an expression for the conductivity can be
obtained:
𝜎𝐻 =2𝑛𝑣𝑑𝑞
𝐸𝑒𝑥𝑝 (−
𝜙𝐻
𝑘𝑏𝑇) [sinh (
𝑞𝑒𝐸𝑑
2𝑘𝑏𝑇)]. (2 − 36)
49
Multiplication of the conductivity by the electric field yields an expression for current density
𝐽 = 𝐽𝑜𝑒𝑥𝑝 (−𝜙𝐻
𝑘𝑏𝑇) [sinh (
𝑞𝑒𝐸𝑑
2𝑘𝑏𝑇)]. (2 − 37)
where 2nvdq is absorbed by the term Jo. Unlike previously discussed Schottky and PF conduction
theories, analysis of I(V) data implementing hopping theory is not straight forward due to its
complexity surrounding a multitude of unknown variables such as n, v, d, and φH. Similarly its
mathematical form does not permit linearization, leaving its use in analysis to be determined by
low and high field approximations, goodness of fit, and to the discretion of the scientist who passes
judgement on its validity based on the nature of the material being measured and past literature.
This dissertation provides an alternative approach to the fitting of I(V) data implementing hopping
theory not commonly found in literature, imparting a rigid statistical interpretation on parameter
estimates extracted via nonlinear regression to enhance the significance of the analysis. A detailed
discussion of the technique is found in section 2.4.1.
Each of the discussed conduction theories have been implemented to better understand I(V)
charging characteristics in polymers. One such study by Saxena and Gaur [119] implements I(V)
characteristics to explain high field conduction in PVDF-polysulfone (PDF) blends. PF and
Schottky modeling indicate that blended films exhibit a more PF bulk limited type behavior at high
temperatures relative to Schottky type. Bulk dominated conduction processes have been used to
describe high field conduction in other fluorocarbon-based materials such as PTFE [19] and also
polyimides where contributions made by Sawa et al. [120] and Sacher [121] suggest hopping is
the dominant charge conduction mechanism. Contributions by Meddeb et al. [13] and Vecchio et
al. [122] use I(V) in conjunction with PF and Shottky theory analysis to describe the effect of
surface chemical grafting on PI and P(VDF-TrFE) high field conduction properties. These works
are described in detail in chapters 4 and 5 of this dissertation.
2.3.3.3 Thermally Stimulated Depolarization Current Measurements (TSDC)
Equipment limitations on the lowest achievable frequency using impedance spectroscopy
(see section 2.3.2.1) motivate alternative approaches for measuring field induced space charge
conduction processes in polymer dielectrics. TSDC is a technique which enables the measurement
of currents generated by the buildup and release of charges induced by polarization of a capacitor
[123]. The general experimental procedure involves four main steps: 1) a DC voltage Vp is applied
50
across the dielectric material under test at a poling temperature Tp for set amount of time tp. 2) The
test material is then cooled to initial temperature To with Vp simultaneously applied. To is typically
below the glass transition of the material in question. 3) Applied bias Vp is changed to 0 and 4) the
material is short circuited, and the current is measured during heating from To to a final temperature
Tf. The heating rate chosen is constant (typically in the range of 1oC/min – 5oC/min) and current
is analyzed as a function of temperature.
Polarization occurring in dielectrics exposed to external DC fields can be achieved by a
variety of mechanisms including electronic polarization, atomic polarization, orientational
(dipolar) polarization, interfacial polarization, and space charge polarization [123]. In context of
this dissertation, orientational and various space charge polarizations are discussed. Dipole
polarization occurs in dielectrics containing permanent dipoles distributed through the bulk and
exhibit time scales as low as 10-12s. TSDC current generation during application of applied field
Ep for time t resulting from dipole depolarization is described by Bucci-Fieschi theory. The
following equation shows the build-up polarization in the material resulting from an applied field
Ep for time t at Tp:
𝑃(𝑡) = 𝑃𝑒 [1 − 𝑒𝑥𝑝 (−𝑡
𝜏)] (2 − 38)
where τ is the dipole relaxation time and Pe is the equilibrium polarization. Assuming polarization
times for polarizing and depolarizing are identical, polarization decay upon Ep removal is given
by the exponential:
𝑃(𝑡) = 𝑃𝑒𝑒𝑥𝑝 (−𝑡
𝜏). (2 − 39)
Since TSDC is performed with a constant heating rate over a set temperature range, t must be
transformed to T using the relationship T = To+qt where q is the heating rate dT/dt. Thus,
polarization decay can be rewritten as the following:
𝑃(𝑡) = 𝑃𝑒𝑒𝑥𝑝 [−∫𝑑𝑡
𝜏
𝑡
0
]. (2 − 40)
Assuming this relation holds over all temperatures, initial frozen in polarization equals polarization
at To and the temperature dependence of τ behaves the following Arrhenius relation:
𝜏(𝑇) = 𝜏𝑜𝑒𝑥𝑝 [𝜓
𝑘𝑏𝑇] (2 − 41)
51
with τo being relaxation time at infinite temperature, kb is Boltzmann’s constant and ψ is activation
energy of dipolar disorientation, the current density generated during a TSDC experiment can be
represented by the following:
𝐽(𝑇) =𝑃𝑒(𝑇𝑝)
𝜏𝑜𝑒𝑥𝑝 (−
𝜓
𝑘𝑏𝑇) 𝑒𝑥𝑝 [−
1
𝑞𝜏𝑜∫ 𝑒𝑥𝑝 (−
𝜓
𝑘𝑏𝑇′) 𝑑𝑇′
𝑇
𝑇𝑜
]. (2 − 42)
Due to the complexity of the integral contained within the second exponential term, an analytical
expression for the current density is typically employed in data analysis and fitting:
𝐽(𝑇) =̃𝑃𝑒(𝑇𝑝)
𝜏𝑜𝑒𝑥𝑝 (−
𝜓
𝑘𝑏𝑇) 𝑒𝑥𝑝 [−
𝑘𝑏𝑇2
𝑞𝜏𝑜𝜓𝑒𝑥𝑝 (−
𝜓
𝑘𝑏𝑇)]. (2 − 43)
Theories involving space charge polarization and depolarization processes are inherently
more complex than dipolar polarization processes. In this scenario, both injected charges occurring
during the poling process of the material as well as impurity ion polarization through the sample
resulting in heterocharging during poling contribute to the measured current. Space charge
relaxation processes are highly dependent on parameters that are not intrinsic to the material. In
the case of charge migration due to preexisting impurity ions within the material, counteracting
action of diffusion, charge recombination events causing migrating charge loss, blocking effects
of the dielectric/electrode interface and trapping properties of the material all influence the
measured current [123]. These aspects of space charge measurements by TSDC make analytical
analysis difficult of peaks arising from heterocharge processes (ionic relaxations). In this
dissertation, Bucci-Fieschi theory is used in the analysis of all relaxation peaks observed in TSDC
in order to make comparison between undoped, ionically doped, and multilayered films, however
initial assumptions surrounding the theory’s derivation must be respected during data
interpretation.
Literature typically uses TSDC as a supplementary technique used to further understand
polarization and depolarization mechanisms in relation to material structure. One example of this
is work done by Sauer and Kim [124] which uses TSDC to identify tacticity related molecular
relaxations occurring at the Tg of poly(methyl methacrylate). The data processing techniques
involved in this work are of interest to TSDC peak deconvolution and can be read about in detail
in work by Neagu et al [125]. More recent work on PVDF systems attempt to correlate structural
morphology, TSDC molecular relaxation, and the nature of d33 coefficient in PVDF/BaTiO3
composites [126]. This research reports TSDC peaks occurring in the temperature range of 135 –
52
170oC and is attributed to the merger of two relaxation processes: dipolar relaxations combined
with interfacial polarizations occurring between PVDF matrix and nanofillers to produce a single
relaxation. Other reports by Yang, et al. [12] also use TSDC to gain greater insight into molecular
relaxation processes involved in dipole polarization in BOPVDF. In this report, experiments are
done more systematically than in Gaur et al. [126] and perform TSDC with changing Ep, tp, and
electrode material [12]. It was found that current peaks occur within 50oC – 110oC and correspond
to electronic injection during poling as well as charge migration due to impurity ions controlled
by electrode metal chemistry.
2.3.3.4 High Field Dielectric Breakdown Measurements and Weibull Statistics
The characteristic dielectric strength of a material is measured by performing high field
dielectric breakdown experiments. In this technique, a dielectric film of known thickness is
exposed to a gradually increasing applied voltage until the material becomes conducting, marking
the point of dielectric breakdown. The method by which this is done is typically a ball and plate
experimental set up, where the sample is sandwiched between a grounded copper plate and
approximately 5mm diameter ball electrode whose electrical potential is ramped at a constant ramp
rate of 500 V/s. The small contact area characteristic of the ball electrode minimizes extrinsic
defects contribution to the breakdown measurements, enabling extraction of a breakdown strength
as close to the material’s intrinsic strength as possible given experimental limitations. During the
experiment, breakdown voltage (VBD) is measured many times on a given sample to form a
distribution of VBD values. Although no set requirement exists for the number of breakdowns
required for analysis, analysis in this dissertation involves >25 breakdown events.
The distribution of breakdown values is analyzed statistically by implementing Weibull
failure statistics to estimate parameters that quantify the material’s behavior. A two parameter
Weibull distribution is typically used to model the behavior of dielectric breakdown events and is
chosen in this dissertation based on of goodness-of-fit to data, use in other literature studying
similar systems and its IEEE standardization [127]. The probability distribution function describes
the distribution of breakdown events and exhibits the following form:
𝑓(𝑡, 𝛽, 𝛼) =𝛽
𝑡(𝑡
𝛼)𝛽
𝑒𝑥𝑝 [− (𝑡
𝛼)𝛽
] (2 − 44)
53
where α is the scale parameter (characteristic breakdown voltage), β is the shape parameter, and t
is the measured variable during tests (in this case measured VBD of the sample). The characteristic
breakdown strength is the dielectric breakdown strength (α) reported for the material and the shape
parameter is a quantity describing the spread in successive breakdown events. In order to calculate
these parameters from the data, the cumulative distribution function (CDF) is used which is
obtained by integration of the probability distribution function:
𝐹(𝑡, 𝛽, 𝛼) = 1 − 𝑒𝑥𝑝 [− (𝑡
𝛼)𝛽
]. (2 − 45)
By taking the double natural logarithm of F(t,β,α) the CDF can be linearized into the following
equation used for data analysis:
𝑙𝑛[−𝑙𝑛(1 − 𝐹(𝑡, 𝛽, 𝛼))] = 𝛽 ln(𝑡) − 𝛽ln(𝛼). (2 − 46)
When handling breakdown data, an approximation of F(t,β,α) is made based on the rank i of the
ith breakdown event from a total population of n breakdowns. An approximation formula for
F(t’β,α) takes on the following form:
𝐹(𝑡, 𝛽, 𝛼) =̃ 𝐹(𝑖, 𝑛) =𝑖 − 0.44
𝑛 + 0.25 (2 − 47)
which was empirically determined to be a good descriptor of dielectric breakdown data and listed
in the IEEE standard on breakdown analysis [127]. Successive breakdown events are ordered from
lowest VBD to highest. Plotting ln[-ln(1-F(i,n))] as a function of ln(t) enables the extraction of β
from linear fitting via the slope of the fit line. Characteristic breakdown voltage (α) is extracted
where ln(α) equals ln(t) at the point ln[-ln(1-F(i,n)] is zero. Since F(i,n) is an approximation of the
cumulative distribution function, its value represents probability of failure for a given voltage t, or
in this case rank i, and has a range 0 < F(i,n) < 1. The value α occurs at the point F(i,n) = 0, and
thus corresponds to the 63% probability of failure of the material.
Dielectric breakdown is an integral experiment in the understanding of electrical failure
and high field conduction properties of dielectric material. We have seen in the introduction of this
dissertation how dielectric breakdown experiments inform on organic dielectric material strength
when doped with nano-particles [4] [5], reactive plasma processing’s role on high field conduction
characteristics [13] [17], and related to IV experiments as a function of deposited electrode material
[11]. Other work done by Ahmed et al. [128] exemplifies breakdown’s ability to detect failure
mechanisms involved with breakdown events. Weibull analysis was used to study the phenomenon
54
of self-clearing on high field dielectric performance in metalized P(VDF-TrFE-CTFE) [128].
Clearing events were separated from intrinsic breakdown of the material by observation of bi-
modal behavior in the Weibull distribution of Ag electroded P(VDF-TrFE-CTFE). Extraction of
slope parameters for clearing events (β approximately 6) where twice as low as that for intrinsic
breakdown (β approximately 12) highlighting the techniques sensitivity to dielectric breakdown
mechanism. Work done by Wang et al. [129] also uses breakdown experiments with Weibull
analysis to extract high field characteristic strength of BaTiO3/PVDF composites, enabling the
calculation of field dependent energy density and BaTiO3 fraction dependent energy density of the
material. In this dissertation, dielectric breakdown is used to link blocking at low frequencies to
dielectric strength in multilayers structures, and is discussed in Chapter 6II and Chapter 7 in
greater detail.
2.4 SPECIAL ANALYTICAL TECHNIQUES
Theoretical interpretation of electrical data was used to draw a link between the behavior
of the material based on processing conditions to its structure. This was done by implementing a
variety of data fitting techniques with statistical interpretation of fit results to assess validity in the
model and in some cases, identify experimental conditions in which the model breaks. In this
section, three techniques are discussed in the context that they are presented in this dissertation: 1)
Bootstrap statistical methods to interpret the significance of hopping conduction theory parameters
estimated via non-linear regression, 2) Equivalent circuit analysis combining basic circuit theory
of ideal and distributed electrical components with complex non-linear regression and 3) TSDC
peak deconvolution used to extract Bucci-Fiechi theory fit parameter estimates for TSDC signals
characterized by overlapping relaxations.
2.4.1 Bootstrap Statistics Applied to I(V) Data
Non-linear regression using Hopping theory to fit current density vs. electric field (J(E))
data involves the estimation of unknown parameters Jo and d. Similarly, these fit parameters
exhibit unknown probability distributions. Nonlinear regression utilizes a sum of squares
minimization scheme through R-Studio software to estimate the two parameters in question,
however determining the accuracy of these estimates is not trivial because the variance of their
55
probability distributions is not clearly defined. This prevents the direct calculation of their standard
errors, necessitating the use of statistical tools implemented to score the accuracy of fit results.
In the bootstrap statistical procedure, a probability distribution for fit parameters is
empirically estimated via a Monte Carlo type procedure that effectively eliminates the need for a
formula to compute standard errors and confidence intervals. This method enables parameter
estimates to be scored in terms of confidence intervals which in turn reflect the behavior of the
total sample population, revealing information on the repeatability of measurement within the
sample set. The following sections focus on implementation of this procedure applied to the
analysis of J(E) data in this manuscript, however work done by Efron and Tibshirani [130] should
be referenced for further insight into bootstrap theory and application.
For simplicity, description of this method focuses on the PI sample set recorded at 25oC
only found in Chapter 5, section 5.3.2.1, Figure 5-2. In practice, it was applied to both PI and
PPIDS sample sets at all measurement temperatures during analysis. The procedure begins with
the raw J(E) data of each sample of the untreated polyimide. This set of data is represented blow
as PI:
PI:
𝑆𝑎𝑚𝑝𝑙𝑒 1 {(𝐸1𝐽1,1), (𝐸2𝐽1,2), … , (𝐸15𝐽1,15)}
𝑆𝑎𝑚𝑝𝑙𝑒 2 {(𝐸1𝐽2,1), (𝐸2𝐽2,2), … , (𝐸15𝐽2,15)}
𝑆𝑎𝑚𝑝𝑙𝑒 3 {(𝐸1𝐽3,1), (𝐸2𝐽3,2), … , (𝐸15𝐽3,15)}
Current is measured at a series of 15 progressively increasing electric fields labeled En where “n”
ranges from 1 (lowest applied field) to 15 (highest applied field). The fields used are the same
from sample to sample, however each sample responds uniquely to the applied field and produces
a unique current density Ji,n where “i” denotes the sample tested and “n” corresponds to the nth
electric field. PI is thus comprised of a total of 45 unique (En, Ji,n) pairs that describe the behavior
of the material at 25oC.
PI is then used to create a distribution of fit parameters Jo and d by employing the technique of
“resampling with replacement”. During a single iteration of this procedure, PI will be re-sampled
a total of 45 times to produce the following new data set:
𝑃𝐼∗: {(𝐸1𝐽2,1)∗, (𝐸13𝐽3,13)
∗, (𝐸2𝐽1,2)
∗, (𝐸2𝐽1,2)
∗, (𝐸7𝐽2,7)
∗… , (𝐸15𝐽1,15)
∗}
56
In this expression, the superscript “*” is used to denote the created bootstrap sample set. Due to
the concept of replacement, the PI remains unchanged during the sampling process. This produces
a new set PI* that can contain duplicate (En, Ji,n)* pairs since the state of PI is constant throughout
resampling. Visual representation of the PI set in comparison to a potential PI* resampled set is
shown below in Figure 2-7 for comparison.
Figure 2-7: J(E) behavior of a) PI sample set displaying 45 data points taken at 25oC and b) the bootstrapped sample
set PI* containing 45 data points resampled with replacement from PI. The outcome of resampling with replacement
can be seen by gaps in data due to duplicate sampling.
The set PI* is fit by standard nonlinear regression using R-studio’s built in non-linear regression
command “nls()”. The product yields a Jo* and d* estimated from PI*. The fit parameters are then
stored in a set of parameter estimates called set PE.
The generation of PI* from PI is then reinitiated, fit using nonlinear regression to produce
new estimates of Jo* and d*, which are then stored in set PE. This process is performed a total of
10,000 iterations, producing the following set of parameter estimates:
𝑃𝐸 = {(𝐽𝑜1∗ , 𝑑1
∗), (𝐽𝑜2∗ , 𝑑2
∗),… , (𝐽𝑜10,000∗ , 𝑑10,000
∗ )}
From the set PE, statistical information reported imperative to scoring the significance of
parameter estimates are calculated: a mean for Jo* and d* is computed using R-studio’s mean()
command, a histogram created for the parameter estimates using the hist() command, and 95%
confidence intervals extracted using the quantile() command1.
1 An annotated version of the R-Studio code used for bootstrap statistics can be found in Appendix A.
57
2.4.2 Equivalent Circuit Modeling
Impedance data analysis employed in this dissertation relies on EC fitting to link
conduction properties to material structure. This type of complex non-linear regression attempts
to fit the analytical expression for impedance derived from a user defined EC to dielectric data
recorded as a function of frequency. In doing so, polarization mechanisms occurring with the
material over a given frequency range can be quantified using estimated values of capacitive and
resistive components. Furthermore, the defined EC exhibits a form that links polarization events
to the material’s structure. The EC used in the analysis of P(VDF-TrFE) impedance as a function
of frequency and LiClO4 content (discussed in Chapter 6I section 6I.5) is shown below in Figure
2.7. Similar to Figure 2-6, the EC used accounts for typical polymer polarization behavior
associated with electronic, dipole orientation and ionic transport at low frequencies. To account
for P(VDF-TrFE)’s polycrystalline structure as well as included ionic content introduced into the
material by LiClO4 doping, two adjustments are made: a nested CPE3/R4 element is introduced to
account for Li+ cation interaction with crystalline structures at low frequency and a CPE4 in series
with the bulk circuit is added to account for blocking polarization associated with Li+ cation
interaction with the electrode at quasi DC frequencies. The following sections show the impedance
formula for each polarization mechanism associated with the EC in Figure 2-8, and derive the EC
equivalent impedance formula used in complex nonlinear regression of P(VDF-TrFE) data in
Chapter 6I section 6I.5.1.
Figure 2-8: EC model depicting polarization mechanisms associated with P(VDF-TrFE) polycrystalline structure at
frequencies spanning 10-1-105 Hz. Nested CPE3/R4 and CPE4 distributed circuit elements are incorporated to describe
low frequency ion interaction with crystals and the electrode/dielectric interface respectively.
58
2.4.2.1 Equivalent Circuit Theory
i) Electronic Polarization: This portion of the circuit is described by C1 and has the following
expression for impedance:
𝑍𝑒𝑙𝑒𝑐 =1
𝑗𝜔𝐶1 (2 − 48)
The electrical response is purely imaginary since the circuit element is a perfect capacitor. Thus
this element is not considered to contribute to the bulk resistance of the material involved in real
processes such as ionic transport.
ii) Permanent Dipole Orientation: The impedance caused by permanent dipole rotation under an
applied AC field is represented by the following expression:
𝑍𝑑𝑖𝑝 =1
(𝑗𝜔)𝑛𝑄𝑜2+ 𝑅2 (2 − 49)
The impedance for this leg of the bulk response has both real and imaginary components. The
imaginary component depends on distributed circuit element CPE2 imperfection factor n that
corresponds to distribution of relaxation times in dipole response of the material and is very close
to 1 (0.98 at the lowest from the fitting). For this reason, the impedance is approximated as that of
an ideal capacitor:
𝑍𝑑𝑖𝑝 =1
𝑗𝜔𝐶2+ 𝑅2 (2 − 50)
and extends purely into the imaginary plane. R2 is real, and controls the frequency at which the
polarization mechanism relaxes out by the product τ=C2R2.
iii) Ionic Polarization: The impedance for this leg of the bulk response is more complex,
depending on R-CPE3/R4 network. The ZCPE3,R4 is addressed first:
𝑍𝐼𝑜𝑛 = 𝑅3 + 𝑍𝑅4,𝐶𝑃𝐸3 (2 − 51)
First we focus on the impedance response of the nested CPE3/R4 circuit. The admittance of this
network is written as the following:
𝑌𝑅4,𝐶𝑃𝐸3 =1
𝑍𝑅4,𝐶𝑃𝐸3= [
1
𝑅4+
1
𝑍𝐶𝑃𝐸3] (2 − 52)
The impedance is then the reciprocal of the admittance
59
𝑍𝑅4,𝐶𝑃𝐸3 =1
[1𝑅4
+1
𝑍𝐶𝑃𝐸3] (2 − 50)
It is easy to see the admittance of CPE4 embedded within the denominator of the expression for
the impedance of the nested CPE3/R4 element. Invoking the definition of complex admittance for
a CPE circuit element Y*CPE, equation 6 can be transformed into the following expression:
𝑍𝑅4,𝐶𝑃𝐸3∗ =
1
1𝑅4
+ 𝑄3𝜔𝑛3[cos(𝑛3𝜋2
) + 𝑖 sin(𝑛3𝜋2
)] (2 − 53)
The complex impedance of the circuit is then broken into its real and imaginary components. This
equation is multiplied by the complex conjugate of the CPE admittance to remove imaginary
components from the denominator of the expression:
𝑍𝑅4,𝐶𝑃𝐸3∗ =
1
1𝑅4
+ 𝑄3𝜔𝑛3[cos(𝑛3𝜋2
) + 𝑖 sin(𝑛3𝜋2
)]×
1𝑅4
+ 𝑄3𝜔𝑛3[cos(𝑛3𝜋
2) − 𝑖 sin(𝑛3𝜋
2)]
1𝑅4
+ 𝑄3𝜔𝑛3[cos(𝑛3𝜋2
) − 𝑖 sin(𝑛3𝜋2
)]
Calculation of the numerator for Z*R4, CPE3 is trivial. We focus on simplification of the denominator
to a more useful form:
Define:
1
𝑅4= 𝐴
𝑄4𝜔𝑛3 cos (
𝑛3𝜋
2) = 𝐵
𝑄4𝜔𝑛3 sin (
𝑛3𝜋
2) = 𝐶
Simplified denominator multiplication using definitions:
(𝐴 + 𝐵 + 𝑖𝐶)(𝐴 + 𝐵 − 𝑖𝐶)
= 𝐴2 + 𝐴𝐵 − 𝑖𝐴𝐶 + 𝐵𝐴 + 𝐵2 − 𝑖𝐵𝐶 + 𝑖𝐴𝐶 + 𝑖𝐵𝐶 − 𝑖2𝐶2
= 𝐴2 + 2𝐴𝐵 + 𝐵2 + 𝐶2
Substituting back for physical quantities:
=1
𝑅42 + 2 [
1
𝑅4𝑄3𝜔
𝑛3 cos (𝑛3𝜋
2)] + [𝑄3
2𝜔2𝑛3cos2 (𝑛3𝜋
2) + 𝑄3
2𝜔2𝑛3sin2 (𝑛3𝜋
2)]
=1
𝑅42 + 2 [
1
𝑅4𝑄3𝜔
𝑛3 cos (𝑛3𝜋
2)] + 𝑄3
2𝜔2𝑛3 ; ∈ 𝐑𝐞 (2 − 54)
This simplified expression for Z*R4, CPE3’s denominator is now placed back into equation 7 and
further simplified:
60
𝑍𝑅4,𝐶𝑃𝐸3∗ =
1𝑅4
+ 𝑄3𝜔𝑛3[cos(𝑛3𝜋
2) − 𝑖 sin(𝑛3𝜋
2)]
1𝑅4
2 + 2 [1𝑅4
𝑄3𝜔𝑛3 cos (𝑛3𝜋2 )] + 𝑄3
2𝜔2𝑛3
(𝑅4
2
𝑅42)
=𝑅4 + 𝑅4
2𝑄3𝜔𝑛3[cos(𝑛3𝜋
2) − 𝑖 sin(𝑛3𝜋
2)]
1 + 2 [𝑅4𝑄3𝜔𝑛3 cos (𝑛3𝜋2 )] + 𝑅4
2𝑄32𝜔2𝑛3
(2 − 55)
It is now possible to include series resistance R3 to formulate an expression for the overall
frequency response of the ionic leg as expressed in equation 4:
𝑍𝑖𝑜𝑛∗ = 𝑅3 +
𝑅4 + 𝑅42𝑄3𝜔
𝑛3[cos(𝑛3𝜋2
) − 𝑖 sin(𝑛3𝜋2
)]
1 + 2 [𝑅4𝑄3𝜔𝑛3 cos (𝑛3𝜋2
)] + 𝑅42𝑄3
2𝜔2𝑛3
(2 − 55)
Equation (2-55) can then be broken into its real and imaginary components:
𝑍𝑖𝑜𝑛 ∈ 𝑹𝒆 = 𝑅3 +𝑅4 + 𝑅4
2𝑄3𝜔𝑛3[cos(𝑛3𝜋
2)]
1 + 2 [𝑅4𝑄3𝜔𝑛 cos (𝑛𝜋2 )] + 𝑅4
2𝑄32𝜔2𝑛
(2 − 56)
𝑍𝑖𝑜𝑛 ∈ 𝑰𝒎 =−𝑅4
2𝑄3𝜔𝑛3[𝑖 𝑠𝑖𝑛(𝑛3𝜋
2)]
1 + 2 [𝑅4𝑄3𝜔𝑛3 𝑐𝑜𝑠 (𝑛3𝜋2 )] + 𝑅4
2𝑄32𝜔2𝑛3
(2 − 57)
iv) Blocking Polarization: The impedance of CPE4 that represents polarization due to blocked
space charge at the electrode/dielectric interface is given by the following expression:
𝑍𝑏𝑙𝑜𝑐𝑘 =1
(𝑗𝜔)𝑛4𝑄𝑜4 (2 − 58)
This expression can be displayed in its complex form:
𝑍𝑏𝑙𝑜𝑐𝑘 =1
𝑄4𝜔𝑛4[cos(𝑛4𝜋2
) + 𝑖 sin(𝑛4𝜋2
)] (2 − 59)
Using a similar mathematical analysis as performed on circuit elements R4 and CPE3, the
blocking impedance can be broken into its real and imaginary impedance values shown below:
𝑍𝑏𝑙𝑜𝑐𝑘∗ =
cos(𝑛4𝜋2
) − 𝑖 sin(𝑛4𝜋2
)
𝑄4𝜔𝑛4 (2 − 60)
𝑍𝑏𝑙𝑜𝑐𝑘 ∈ 𝑹𝒆 =cos (
𝑛4𝜋2 )
𝑄4𝜔𝑛4 (2 − 61)
𝑍𝑏𝑙𝑜𝑐𝑘 ∈ 𝑰𝒎 =−i sin (
𝑛4𝜋2 )
𝑄4𝜔𝑛4 (2 − 62)
61
In the case n4 = 1, the impedance is purely capacitive and equation (14) goes to 0 while equation
(15) gives the impedance of an ideal capacitor (1/iωQ4). In the case n4 = 0, the response is purely
resistive and equation (15) goes to 0.
v) Complete EC Response: The total response as a function of frequency for the EC can be
derived by considering each individual component discussed. The general formula for the
impedance of the EC in Figure 2-8 is given by the following equation:
𝑍𝐸𝐶 = (1
𝑍𝑒𝑙𝑒𝑐+
1
𝑍𝑑𝑖𝑝+
1
𝑍𝑖𝑜𝑛)
−1
+ 𝑍𝑏𝑙𝑜𝑐𝑘 (2 − 63)
The total impedance for the model is formulated by substituting equations (2-48), (2-50), (2-55),
and (2-62):
𝑍𝐸𝐶 = (𝑗𝜔𝐶1 + [1
𝑗𝜔𝑄2
+ 𝑅2]−1
+ [𝑅3 +𝑅4 + 𝑅4
2𝑄3𝜔𝑛3[cos(𝑛3𝜋
2 ) − 𝑖 sin(𝑛3𝜋2 )]
1 + 2 [𝑅4𝑄3𝜔𝑛3 cos (
𝑛3𝜋2
)] + 𝑅42𝑄3
2𝜔2𝑛]
−1
)
−1
+ [cos(𝑛4𝜋
2 ) − 𝑖 sin(𝑛4𝜋2 )
𝑄4𝜔𝑛4
] (2−64)
Complex nonlinear regression using the electrochemical software Z-view is used to fit Q2, n2, R3,
Q3, n3, R4, Q4, and n4 from the above relation to impedance spectra for P(VDF-TrFE) in this
dissertation.
2.4.2.2 Statistical Interpretation of Fit Parameters
Complex non-linear regression fitting was performed on impedance data for the tested
P(VDF-TrFE) samples in the manuscript. During each fit, 4 impedance formalisms were assessed
to ensure a good fit to data: 1) real and imaginary capacitance, 2) complex impedance in cole-cole
plot format, 3) the magnitude of impedance and phase angle, and 4) real and imaginary modulus.
The impedance data is converted to each formalism using Z-View software. Each formalism is
easily displayed using the program as well. The goodness of fit was assessed in the following three
ways:
i) Regression fit result superimposed on the 4 mentioned impedance formalisms. This is a
qualitative assessment to ensure the behavior of the data is captured by using the proposed EC at
each temperature and LiClO4 concentration.
ii) Calculation of material and bulk properties from the appropriate circuit elements. This
was a useful step used to validate CPE2’s accuracy in estimating dipolar response by calculating
62
material permittivity from fit results and comparing to values calculated directly from capacitance
measurements within the frequency range 102 – 105 Hz.
iii) Reference to statistical reports generated by Z-view. The fitting of impedance data was
performed using built in complex non-linear regression software embedded in Z-view. The
statistical output used in data analysis of parameter value significance was % error of fit parameter,
which is explained in further detail in the Z-view Impedance/Gain Phase Graphing and Analysis
Software Operating Manual version 3.5. Calculation of statistical quantifiers such as % error is
sensitive to the weighting formalism used when fitting data. Using a weighting scheme such as
“unit weighting” will overemphasize data values of large magnitude, which is likely when fitting
a broad frequency spectrum. In this work, calc-proportional weighting is used since each data
point’s weight is normalized by its magnitude. In this weighting scheme, the real and imaginary
components are weighted separately.
2.4.3 TSDC Peak Deconvolution
As mentioned in section 2.3.3.3 Bucci-Feischi theory is used to describe relaxation
processes associated with pure copolymer in this dissertation. Similar to EC modeling and bulk
limited conduction analysis via hopping theory, the mathematical expression for TSDC is not
linearizable (equation 2-43) an requires the use of nonlinear regression for parameter estimation
and fitting. One challenge during fitting is the separation of coalescing TSDC signals. Equation
2-43 is intendend to describe only one depolarization event with a single relaxation time. Thus if
a borad temperature spectrum consisting of multiple peaks exists, peak deconvolution exposing a
single pronounced peak must be performed. A data subtraction technique was used, described
visually in a schematic presented in Figure 2-9a, 9b, and 9c. Initially the strongest peak in the
spectrum is selected in 9a. Nonlinear regression is performed producing parameter estimates that
describe the depolarization of peak 1. From these estimates, a function over the entire temperature
range is produced and subtracted from the raw data portrayed in 9b, exposing peak 2 without
contribution from peak 1. Fitting is then performed on peak 2 depicted in 9c and parameter
estimates recorded.
Non-linear regression performed using equation 2-43 necessitates the estimation of three
parameters: Po which is a polarization constant, activation energy ψ and relaxation time τo. Since
3 parameters must be estimated, a custom R-Studio script was written implementing nested “for
63
loops” which compute J for a user defined range of Po, ψ, and τo which follows the following
functional form:
𝜏𝑜 =𝑘𝑏𝑇𝑚
2
𝑏𝜓𝑒𝑥𝑝 (𝜓
𝑘𝑏𝑇𝑚) (2 − 65)
where b is the heating rate of the experiment.2
Figure 2-9: Schematic of TSDC peak deconvolution procedure depicting a) fitting of strongest signal in convoluted
spectrum, b) subtraction of fit function from raw data over temperature range and c) fitting deconvoluted low
temperature peak.
2 An annotated version of the R-Studio code used for TSDC peak deconvolution can be found in Appendix B.
64
CHAPTER 3
HOT-PRESSED PVDF LAMINATES – THE EFFECT OF INTERFACES IN SINGLE
MATERIAL LMINATED DIELECTRICS3
ABSTRACT
This chapter is focused on the development and characterization of multilayer PVDF
dielectric laminates for preliminary analysis of the effect of interfaces on charge transport and high
field dielectric breakdown in all organic dielectrics. Unlike work by Mackey et al. [14] and Zhou
et al. [15] on dielectric breakdown strength of P(VDF-HFP)/PC composites, this chapter only
incorporates PVDF as the laminated material, eliminating composite effects and focusing on how
the interface created by hot-press lamination impacts electrical properties. An outline of chapter
contents is provided below:
Introduction – motivates the work discussed along with the most impacted scientific field.
Materials and Methods – presents processing methods used to create multilayered
laminates (leveraged from process outlined in Chapter 2, section 2.2.1), as well as
structural and electrical characterization techniques.
Results – discusses outcomes of structural and electrical characterization of laminated
dielectrics in comparison to 1-layer control films.
Conclusions – summarizes results and provides avenues for further research.
Results obtained from this portion of this dissertation provided a proof-of-concept frame-work
demonstrating the impact interfaces can have on electrical conduction without need for composite
materials, leveraging research performed in subsequent chapters.
3.1 INTRODUCTION
Modern age high-powered electronic applications require the development of new
materials that exhibit high-energy storage capabilities. Although batteries currently have higher
electrical energy storage capabilities than capacitors, they lack the power output to fulfill most
3 A significant portion of this work was published at CEIDP Toronto 2016 conference [228].
65
high-powered applications. Currently, polymer capacitors are sought after for high-powered pulse
applications due to their high breakdown strengths, low losses, low cost, and ease of
manufacturability. Although these characteristics are seen as favorable, polymer capacitors suffer
from low dielectric permittivity and thus relatively low energy densities. Strategies used to blend
polymer composites and add inorganic inclusions to increase permittivity are met with limited
success, resulting in local electric field distortions that decrease the effective breakdown strength
[131].
In this study, we investigate the effect of added interfaces within the dielectric structure
created by lamination. Polyvinylidene fluoride (PVDF) is a widely used polymer for its excellent
mechanical, chemical and ferroelectric properties [132]. Due to its technological importance and
impact on many applications, we have selected it as the model polymer for our study. Past research
has shown that PVDF/Polycarbonate (PC) dielectric composite structures fabricated by layer
stacking improve dielectric breakdown strength in comparison to polymer blends [14]. Similarly,
it has been shown that self-assembled multilayered diblock4copolymers display significantly
higher breakdown strengths than stand-alone cast films [133]. Our main goal is to better understand
the effect of added interfaces on charge transport in all-organic laminates and provide insight into
conduction through layered organic dielectric media. We believe that with the addition of
interfacial elements created by a multilayer laminate structure, charge trapping at interfaces
distributed throughout the structure will enhance high field capacitor performance. This study
uses high voltage dielectric breakdown in conjunction with impedance spectroscopy to investigate
interfacial effects on capacitor performance.
3.2 MATERIALS AND METHODS
3.2.1 Materials Selection
Pure PVDF powder provided by Arkema, USA, and N,N-Dimethylformamide (DMF)
Anhydrous DrySolv were the raw materials used in this study. Protective Kapton PI sheets used
during hot pressing were provided by Dupont.
3.2.2 Multilayer Laminate Fabrication
This material is based upon work supported by NSF as part of the Center for Dielectris and Piezoelectrics under Grant N. IIP-1361503.
66
The procedure followed to fabricate laminates as well as dielectric film quality control is
explained in detail in Chapter 2, section 2.2.1, however the essentials are outlined here. Sample
fabrication is a two-step process: 1) solution casting to prepare PVDF monolayers, and 2) stacking
and hot pressing of monolayers to make multilayer laminates.
1: PVDF/DMF solutions are mixed in 100 ml beakers and magnetically stirred for
approximately 3 hours. The resulting solution is then de-gassed under vacuum for approximately
30 minutes prior to casting. The solution is then poured onto a glass plate (previously cleaned
with Acetone and Kimtech wipe) and spread uniformly over the plate using a doctor blade. Films
are dried at 180 oC under vacuum for 1 hour, leaving a freestanding film ready for hot pressing.
2: The cast films are cleaned using a Kimtech wipe damp with ethanol, then assembled into
a rectangular stacking of 1-4 layers. The stack is sandwiched between two Kapton sheets to
prevent the films from melting and sticking to the hot-press’s platens during pressure application.
The Kapton/PVDF assembly is placed into the vice and pressed at 150 oC for 30 min at
approximately 20 MPa. Samples are then removed and allowed to thermally equilibrate back to
room temperature.
3.2.3 Structural Characterization
Film thickness uniformity before and after hot pressing was assessed using a Heidenhain
model ND-280 dial gauge. SEM cross sectional images of a hot pressed 2-layer film were made
using by freeze fracturing in liquid nitrogen. The crystalline phase assemblage in PVDF for 1-
layer solution cast, 1-layer solution cast film subjected to hot-pressing, 2-, 3-, and 4-layer samples
is assessed by a combination of differential scanning calorimetry and Fourier Transform Infrared
Spectroscopy (FTIR) using a TA Instruments Q2000 DSC and Burker Vertex-70 FTIR
spectrometer.
3.2.4 Dielectric Measurements
Two forms of dielectric analysis were conducted for this study: 1) high voltage dielectric
breakdown at room temperature and 2) low voltage impedance spectroscopy at room and elevated
temperatures.
1: High voltage experimentation utilized a homemade configuration consisting of a TREK
amplifier model 30/20, National Instruments IO DAQ, and Labview software to record data. A
67
copper plate and ballpoint served as bottom and top electrode respectively. The sample is
submerged in insulating (Galden) oil throughout the duration of the experiment to prevent arcing
from top ball electrode to bottom copper plate.
2: Impedance spectroscopy was performed within the frequency range of 1 mHz – 100 kHz
using a Modulab impedance analyzer from Solartron Analytical. Temperature was swept using a
Delta Design 9010 oven. Values for the real and imaginary parts of impedance were collected with
Modulab and converted to capacitance and loss tangent using the general electrochemistry
software Z-view provided by Scribner Associates Inc.
Prior to dielectric measurement, electrodes are deposited onto each sample via electron
beam evaporation using Lab-18 (Kurt J. Lesker). Electron beam evaporation is used over
sputtering to avoid exposing PVDF samples to plasma and higher energy deposition characteristic
of RF magnetron sputtering, which may cause polymer surface degradation [111]. The electrode
material used is Ag, 100 nm in thickness, and is deposited in a circular pattern with a diameter of
1 cm. Lastly, capacitance and loss tangent for each sample are measured using a high-precision
LCR meter at 1 kHz and 1V. This step is done to provide a reference for acquired electrical data.
3.3 RESULTS
3.3.1 SEM Imaging the Interface
Imaging of the cross section of a 2-layer hot-pressed sample was performed to confirm the
presence of interfacial elements within the bulk of the laminate. The sample was prepared by freeze
fracturing in liquid nitrogen. The sample imaged was a 2-layer laminate of 30 μm in total thickness,
processed by hot pressing two individual 15 μm 1-layer films together. Imaging was able to capture
the interface generated by the hot-pressing procedure, confirming the presence of a layered
microstructure, as seen in Figure 3-1a.
3.3.2 DSC PVDF Thermal Analysis
Thermal characterization of PVDF was performed in order to understand the effect of
layering after the hot-pressing procedure. In these experiments, a monolayer exposed to the
outlined hot-pressing procedure as well as a 4-layer hot-pressed stack was measured between the
temperatures 25oC and 200oC with a heating ramp rate of 10oC/min and the results compared in
Figure 3-1b. It was found that the monolayer exhibited a single melting peak at Tm~160oC, which
68
is consistent with the expected value of Tm reported for PVDF films in the literature[17]. This is
also featured by the 1-layer film exposed to the hot-pressing protocol signifying not significant
structural changes occur due to hot pressing alone. The 4-layer stacked samples however showed
a much different thermogram than that of either monolayer. In these samples, the melting peak
split in two distinct signals, one at Tm1~160oC and the other at Tm2 ~170oC. Both crystal phase (α-
or β-phase) as well as crystal size can affect the melting temperature of the crystalline phase,
implying further structural characterization is required to fully understand how layering affects
PVDF crystal morphology.
Figure 3-1: a) SEM image of 2-layer cross section (displaying developed interface) and b) DSC thermogram of un-
pressed, 1-layer hot-pressed, and 4-layer laminate.
3.3.3 FTIR PVDF Crystal Structure Analysis
The dielectric behavior of PVDF’s polymorphic structure is sensitive to the phase (α, β, or
γ) composition of the material. Each phase exhibits a unique dielectric property, for example, β-
phase’s polar nature enabling PVDF’s characteristic piezo and ferroelectric properties in
comparison to the non-polar α-phase of higher dielectric constant [71] [134]. For this reason,
Attenuated Total Reflectance (ATR)-FTIR characterization was performed to compare phase
development across all samples (Figure 3-2).
The most common vibrational modes for the α-phase occur at wave numbers 532, 614,
764, and 796 cm-1. Vibrational modes corresponding to the β-phase are seen at wavenumber 840.
69
Figure 3-2: ATR-FTIR morphological characterization of hot-pressed laminates. Absorbance spectra of PVDF films
from top to bottom A) 1-layer, B) 1-layer hot-pressed, C) 2-layer D) 3-layer and E) 4-layer laminates. Salient
spectral features are discussed in this section. All PVDF vibrational modes are presented in Chapter 2 section
2.3.1.2 Table 2-IV.
Finally, 880 pertains to active vibrational modes for both α- and β-phases [135]. The combination
of these vibrational modes in the absorbance spectrum for the 1-layer PVDF sample exist,
suggesting a combination of α-phase and β-phases are present in solution cast films. Hot-pressing
did not cause a significant change in the 1-layer hot pressed film, however minor changes in peak
intensity increase of the α-phase occurs in 2- through 4-layer laminates. Ultimately the
combination of α- and β-phase in 1-layer and multilayer films is unchanged after hot pressing
suggesting DSC results for melting characteristics of 4-layer films are not due to a change in phase
distribution in the material.
3.3.4 Dielectric Breakdown
Samples undergo a 5-hour drying procedure at 100 oC under vacuum prior to breakdown
to ensure residual solvent does not alter material response. After releasing vacuum, samples are
immediately submerged in Galden oil to minimize exposure to atmosphere as much as possible.
Voltage is ramped 500 V/s until samples fail (determined by a current spike under test conditions).
All testing was performed at room temperature.
The electrical strength of PVDF laminates was analyzed using two parameter Weibull
statistics with the associated most likely probability of failure function:
𝐹(𝑖, 𝑛) =𝑖 − 0.3
𝑛 + 0.4 (3 − 1)
70
where i represents the ith failure in a stress test consisting of n total failures. Methods for estimating
statistical confidence intervals in Weibull modulus and characteristic breakdown strength are
based on the IEEE 930, 2004 standard [127]. Structures tested were a 1-layer solution cast, 1-layer
hot-pressed sample (HP), 2-layer, and 3-layer. The amount of data points recorded for each sample
ranged between 23-28 total breakdown events, which qualify the data as a “larger data set” as seen
in [127]. Computation of linear fits for Weibull data was done using Microsoft Excel’s built in
least squares regression function and correlation coefficients were compared in conjunction with
figure A.8 in [127]. The values of characteristic breakdown strength reported are extracted from
the best-fit line’s intersection with the Ln[Ln(1-F(i,n))] axis equal to 0, indicating the 63rd
percentile of failure. In order to derive statistical significance in this value, an associated 90%
confidence interval was computed for each characteristic breakdown value by methodology
offered in [10].
Figure 3-3a summarizes the characteristic breakdown values of tested samples extracted
from two parameter Weibull distributions. An attempt was made to maintain a similar total
thickness of each tested film to avoid the effect of thickness on dielectric breakdown results [136]
[137], however controlling sample thickness using stacking/hot pressing method was challenging.
The initial 1-layer sample achieved a characteristic breakdown field of 385 MV/m. After exposing
a 1-layer film to hot pressing, we note an increase in characteristic breakdown field to 415 MV/m.
This 8% increase in breakdown strength was expected due to the suspected healing of defects
formed through solution processing upon exposure to high temperatures around the melting
temperature (Tm) of the polymer. The 1-layer HP sample is most closely related to the 2-layer and
3-layer due to its exposure to thermal loading and pressure. With the addition of a single
interface (2-layer) characteristic breakdown field is seen to increase to a value of 480 MV/m
yielding a 16% increase that displays a 90% confidence range that falls outside of 1-layer HP. The
3-layer sample containing two interfaces achieved a characteristic breakdown field of 490 MV/m.
This is not considered a statistically significant increase in the durability of the dielectric structure
relative to the 2-layer. The breakdown strength of the films were then fit as a function of their
thickness t in order to test if their spread in film thickness (37μm for the 1-layer and 23μm for the
3-layer) contributed to the increase in breakdown strength in Figure 3-3b. Fitting the breakdown
strength as a function of t using Microsoft’s linear regression function revealed a poor fit
71
Figure 3-3: a) dielectric breakdown strength shown for 1-layer films, 1-layer hot pressed, 2- and 3-layer films and b)
dielectric breakdown strength plotted as a function of t for 1-layer hot pressed, 2- and 3-layer films.
dependent on t-0.38 which disobeys breakdown strength’s expected thickness dependence on t-1/2
[138]. Work measuring the breakdown strength of solution casted PVDF as a function of film t
has also shown no thickness dependence on the breakdown strength in films ranging 20 - 40 μm,
which spans the thickness range of films tested in this chapter [139]. The combination of fitting as
well as previous literature suggest that changes in dielectric breakdown strength are in fact due to
the introduction of planar interfaces by hot pressing.
3.3.5 Impedance Spectroscopy
Capacitance and loss data was captured for samples containing 1- through 4-layers and
converted to dielectric permittivity plotted as a function of frequency, as seen in Fig. 3. The
permittivity for a 1-layer sample is at 8.75 at 100 Hz and remains stable between 8 and 8.5 until
the onset of the dipolar relaxation regime at approximately 0.1 MHz. Samples ranging from 2- to
3-layers display increased permittivity as follows: 2-layer has permittivity of 9.10 at 100 Hz and
3-layer has 9.86 at 100 Hz. This trend breaks at 4-layers with saturation between 3- and 4-layers.
If we consider the 3-layer sample as having the highest achieved permittivity, we note a 13%
increase in dielectric permittivity through interfacial element addition. The loss tangent between
the range of 1 kHz – 100 Hz displays a decreasing trend with increasing layer count. This
observation motivates investigation of dielectric response at the quasi DC frequency range.
Analysis is expanded to include equivalent circuit (EC) modeling to characterize interfacial
effects at elevated (70 oC) temperatures. In Figure 3-4, room temperature permittivity and loss
72
Figure 3-4: Dielectric permittivity and loss tangent calculated from impedance spectroscopy measurements between
10-2 Hz – 105 Hz for 1-layer and 4-layer PVDF laminates. EC modeling performed using the shown circuits for each
sample set (bottom left) with parameter estimates (bottom right).
tangent are plotted against the equivalent circuit model corrensponding to 1- and 4-layer samples
as solid bold lines running thorugh the data. Fitting for each data set was done using Z-View
electrochemical circuit modeling software. The frequency range spans 100 kHz – 10 mHz.
The circuit model for the 1-layer consists of a parallel circuit element containing a
capacitor, resistor and constant phase element (CPE) in parallel. A similar circuit consisting of
CPE with a resistor in parallel is seen in another study modeling the electric response of bulk
poly(vinyl alcohol) (PVA) with yittrium modifyed barium zirconium titanate (BYZT) micro
particles [140]. Although similar, it does not completely incorporate circuit elements
corresponding to the three mechanisms typically seen in bulk polymer behavior: induced dipoles,
static dipoles, and ionic conductivity. This behavior can be rationalized through three respective
elements: capacitor, capacitor-resistor in series, and stand alone resistor [141]. For our proposed
model, the capacitor-resistor componant is replaced with a CPE; a capacitor like circuit element
with the following form of impedance:
𝑍′ =1
𝑄(𝑗𝜔)𝑛 (3 − 2)
where Q is the value of the CPE element, j the square root of negative one, ω is angular frequency
and n is an exponential imperfection factor ranging from zero to one. The precise origin of the
73
CPE is not wholly understood, however its occurrence in systems with electrode imperfections
such as roughness or porosity, as well as electrochemical diffusion mechanisms at play has been
studied [142] [143]. For systems involving polycrystalline solid state devices, CPE behavior has
been rationalized to arise due to alpha relaxation process. This EC describes the behavior of the 1-
layer by itself, yielding a good fit to data with accurate componant values diplaying low errors.
When performing the fitting for the 4-layer sample set, using the bulk model alone gives rise to
elevated errors in circuit elements, as well as deviation from the permittivity at 10 mHz. The
addition of a CPE in series (labeled CPE2) produced a more precice fit in the low frequency regime
as well as yielding reasonable values for circuit elements. This circuit element is believed to arise
from the development of added interfacial polarization afforded by the layered structure, and is
necessary in modeling the dielectric response of the 4-layer sample set.
3.4 CONCLUSIONS
We have successfully developed processing protocol to fabricate structurally robust
multilayered laminates comprised of pure PVDF. The interface was initially characterized using
IEEE standard dielectric breakdown protocol. Measurements yielded increased dielectric
breakdown strength with increasing layer count of the structure. 1-layer samples exhibit a 415
MV/m breakdown while 2- and 3-layers increase to 480 and 490 MV/m respectively. The origin
of this enhancement in breakdown strength is attributed to the blocking of mobile charged species
at high fields. Another potential cause for this increase in durability could be due to a reduction
in processing defects. Monolayers processed for lamination are thinner to match the equivalent
thickness of a single layer, and potentially contain less defects.
Room temperature dielectric spectroscopy data displays an increase in permittivity of the
samples, saturating at 4-layers. Likewise, the dielectric loss was seen to decrease within 1 kHz –
100 Hz with the 4-layer sample displaying lowest loss tangent at 100 Hz, similarly seen in [144].
Through impedance spectroscopy techniques, the DC limit was pushed yielding the effects of
lamination on space charge relaxation mechanisms. Below 10 Hz, layered samples exhibit an
increase in the permittivity in comparison to a 1-layer sample which may be linked back to slight
structural changes in hot pressed layered films seen in DSC and FTIR charaterization. Exposure
to elevated temperatures amplifies the effect of the interface in the 4-layer sample set. This
74
behavior was described by the addition of a CPE element in EC modeling, indicating layered
elements within the dielectric promote the blockage of mobile space charges at low frequencies.
In order to continue the study of interfaces in dielectric media on charge transport, total
thickness of the dielectric must be reduced while layer count increased. Increasing the ratio of
interfacial elements to total thickness will serve to amplify the effect of the interface on charge
transport. A switch to solution spin casting is made in subsequent chapters to reduce layer thickness
while increasing resultant dielectric film thickness uniformity and repeatability in processing.
Simialrly, DSC and FTIR results confirm difficulties surrounding PVDF crystal phase control
during processing. A switch the P(VDF-TrFE) as the model material is also made to enable
consistent phase distribution (predominantly β-phase) in the films to aid in accurate electrical data
interpretation.
75
CHAPTER 4
PLASMA SURFACE MODIFICATION OF P(VDFTrFE): INFLUENCE OF SURFACE
CHEMISTRY AND STRUCTURE ON ELECTRONIC CHARGE INJECTION5
ABSTRACT
In this chapter, the effect of the electrode/dielectric interface’s role on low and high field
conduction is studied by using reactive plasma treatment as a tool to change thin film surface
chemistry and electrical properties. P(VDF-TrFE) as a dielectric material as well as spin cast thin
film processing is introduced to 1) implement a material system in which crystallinity and crystal
phase are well controlled and 2) reduce film thickness relative to blade casted films to increase the
contribution of the interface to the electrical signal. An outline of the chapter contents is provided:
Introduction – A brief literature review emphasizing research targeting plasma
modification for dielectric applications is presented and concluded with the scope of the
work done in this chapter.
Experimental Section – materials and processing methods used to create plasma treated
P(VDF-TrFE) thin films are discussed. Specifics pertaining to chemical, structural and
electrical characterization techniques are listed.
Results and Discussion – outcomes of plasma surface modification on dielectric film
surface topology and chemical structure are first addressed, then linked to altered
conduction properties relative to untreated control samples.
Conclusions – the results are summarized in broader scientific context and used to inform
on avenues for further research as well as analysis of interfacial modified dielectric films.
The results obtained in this section of the dissertation highlight the impact of the
electrode/dielectric interface on electrical conduction in P(VDF-TrFE) and suggest its ability to
influence high field leakage currents in other systems.
5 A significant portion of this chapter has been published in Vecchio et al. [122] in the Journal of Applied Physics, a proceedings for the2017 18th
US-Japan Seminar on Dielectric and Piezoelectric Materials, Santa FE NM 2017 and a proceedings for CEIDP 2018 in Cancun Mexico [227].
76
4.1 INTRODUCTION
The modern age of high-powered electronic applications demands the development of new
dielectric materials that reliably support both high power densities and high energy densities.
Traditionally, polymer film capacitor technologies focus on the use of biaxially oriented poly
propylene (BOPP) because of its high breakdown strength (Eb) of 850 V/μm, self-healing
capabilities, low equivalent series resistance (ESR) and relatively low cost of production [145]
[65]. Although the aforementioned qualities of BOPP make it an excellent material for high-power
applications, its low dielectric constant (εr) of 2.2 hinders the functionality of the material for
devices in which high energy density is a necessity [145] [146] [147]. In this regard, the polar
dielectric polyvinylidene fluoride (PVDF) exhibits a breakdown field comparable to BOPP, as
well as a high permittivity of ~8-12, achieving εr as high as ~50 when polymerized with
trifluoroethylene and chlorotrifluoroethylene [71] [148] [149]. Although the polar nature of this
material allows for a desirably high dielectric constant, PVDF films exhibit large loss tangents that
approach ~0.5-1%, and consequentially high leakage currents [150] [149]. In order for PVDF to
realize its full potential as a material in high energy density and power applications, considerable
work needs to be done to understand conduction through PVDF at low and high fields. In this
work, the impact of surface chemistry on low and high field conduction through poly(vinylidene
fluoride trifluoroethylene) (P(VDF-TrFE)) is studied by using a reactive plasma surface
modification, with a focus on the role of surface chemistry on the electrical properties of polar
polymer dielectrics.
Surface modification via plasma treatment has proven to be both simple and effective in
altering dielectric properties of organic capacitor materials. The application of plasma to the
surface of organic films typically results in surface modification by chemical functionality grafting
that is dependent on both plasma composition and treated material [151]. The addition of these
moieties results in tailorability of the surface chemical composition [152] [153] [154], altered film
wettability [34] [155], adhesion to other surfaces [35] [156] [157] [158], and tuning of electrical
properties [152] [159] [17]. Meddeb et.al. [152] studied the effect of surface treatments using
oxygen based plasma on the wettability and electrical properties of polyimide Kapton® films at
room and elevated temperatures. It was found that increased presence of oxygen at the film’s
surface was coupled with enhanced hydrophilicity of the material post plasma exposure. High field
dielectric breakdown and current-voltage measurements also indicated increased Eb as well as
77
reduced leakage currents at 150 oC in treated films relative to the untreated control. The changes
measured in dielectric properties were associated with the evolved surface chemical structure,
leading to charge scattering effects of trapping regions at the electrode/material interface.
Similarly, Mammone et al. [159] studied the effect of plasma treating polypropylene resin with a
mixture of 96% CF4/4% O2 gas plasma prior to melt extrusion on dielectric breakdown strength.
The results demonstrate that 20% increase in the Eb occurred after the 4 min plasma treatment
procedure. These results and others motivate the incorporation of plasma treatment in the
processing of dielectrics for high field applications, however it should be noted that the majority
of current research focuses on nonpolar systems.
Past research performed on PVDF terpolymer focused on the electrode metal’s effect on
charge injection and dielectric breakdown of the material [11]. The high field dielectric strength
was observed to depend on both electrode material and electrode deposition technique. Similarly,
it was found that charge injection varied depending on electrode material selection, demonstrating
the importance of the metal/dielectric interface on the high field conduction properties of PVDF.
This study by Chen et al. [11] does not use plasma surface treatment to study high field properties
of PVDF, however the study by Mammone et al. [17] does. Mammone investigated the effect of
CF4/O2 plasma treatment on dielectric permittivity and breakdown strength of pure PVDF. The
researchers found that plasma treatment resulted in both an 11% increase in the Eb of the material
and reduction in the bulk dielectric constant of the system. IR measurements indicated a reduction
in absorbance in the alpha crystal absorbance bands (766 cm-1, 855 cm-1, and 978 cm-1), however
the study did not correlate the altered molecular structure to changes in polymer chemistry and
dielectric performance, nor did it identify the specific transport mechanism responsible for these
changes.
The goal of this study is to address the gap in the literature in terms of understanding how
reactive plasma treatments impact high-field performance of polar polymers, by providing greater
insight on the relationship between chemical structure, surface wetting properties and electrical
conduction in polar organic systems. We have devised a processing protocol for a CF4/O2 reactive
plasma surface treatment and successfully modified the surface chemistry of P(VDF-TrFE) thin
films. In doing so, the nature of film surface structure and electrode/dielectric contact for
copolymer dielectrics is changed. The effect of plasma treatment on the film’s surface chemistry
is first analyzed by X-ray Photoelectron Spectroscopy (XPS). The results are used to investigate
78
changes in surface chemical environment, and then correlate to changes in contact angle and
surface polarity of the polymer. Finally, current-voltage (I(V)) measurements were performed to
analyze how surface modification impacts conduction through the material. It is discovered that
the addition of carbonyl chemical moieties to the films via plasma treatment reduces net polarity
of the system and increases low and high field steady state conduction due to altered
metal/dielectric contact properties.
4.2 EXPERIMENTAL SECTION
4.2.1 Materials
The copolymer P(VDF-TrFE) 70/30 mol% was purchased from Arkema in powder form
with molecular weight Mn= 205 kg/mol, reported in the provided safety data sheet. The solvent
N,N-Dimethylformamide (DMF) purchased from DriSolv® was used to dissolve P(VDF-TrFE)
powder. P(VDF-TrFE)/DMF solution was spun onto platinum coated silicon wafers purchased
from Nova Electronic Materials.
4.2.2 Thin Film Fabrication
Solutions of P(VDF-TrFE)/DMF were prepared at 7.5% solid weight content. The solution
was then subjected to a degassing procedure to remove trapped air bubbles resulting from magnetic
stirring. The degassed solution was then transported into a clean room where it was spin casted
onto platinum coated silicon wafers forming a thin continuous film over the surface of the
substrate. A KLA-Tencor P16+ stylus profilometer was used to determine the thickness and
uniformity of the resultant films. Using a spin speed of 600 rpm for 50s consistently produced 1
μm thick films of good thickness uniformity (only a standard deviation of +/- 80 nm across the 4in
diameter substrate) and were used in all processing and experiments in this study. Samples
fabricated for electrical measurements had Ag electrodes deposited 50 nm thick, 1 mm in diameter
using a Lab-18 electron beam evaporator provided by Kurt J. Lesker. The sample stage temperature
was held at 0 oC during electrode deposition to prevent sample damage.
4.2.3 Plasma Surface Modification
All plasma treatments were performed in an M4LTM RF gas plasma system provided by
PVA TePla using a constant flow rate of 250sccm for CF4 and O2. An additional 50sccm of He to
79
control chamber temperature was flowed during the process. The chamber pressure was allowed
to naturally reach an equilibrium pressure between 720 – 740mTorr before plasma ignition. All
treatments were administered at 500W system power and treatment duration varied at 45s, 90s,
135s and 180s. After treatment, the samples remained under gas flow until chamber pressure re-
equilibrated and were then removed and immediately stored under vacuum until further
measurements.
4.2.4 Investigated Processing Conditions.
Samples were processed under two separate conditions for experiments involving plasma
treatments via methods outlined in sections 4.2.1 – 4.2.3 (as well as Chapter 2 section 2.2.2 –
2.2.4). The first is the “As-Spun” processing condition, where plasma treated films only underwent
the spin casting and 15min drying procedure outlined in section 4.2.2. The second is the “Post-
Anneal” processing condition, where films underwent the initial fabrication process in 4.2.2
followed by the plasma treatment protocol outlined in 4.2.3, and finally annealed after plasma
treatment for 24 hrs at 142 oC under vacuum. All plasma treated samples remained under vacuum
after processing until the time of electrical measurement and surface characterization.
4.2.5 Characterization Techniques
4.2.5.1 Chemical and Structural
The thermal properties of the spin casted films were measured using a Q2000 differential
scanning calorimeter provided by TA Instruments. Samples were prepared by scratching 5-7mg of
copolymer off of the wafer into an Al DSC pan using a carefully cleaned razor blade.
XPS experiments were performed using a Physical Electronics VersaProbe II equipped
with a monochromatic Al kα X-ray source (hν = 1,486.6eV) and a concentric hemispherical
analyzer. Charge neutralization was performed using both low energy electrons (less than 5eV)
and Argon ions. The binding energy axis was calibrated using sputter cleaned Cu foil (Cu 2p3/2 =
932.7eV, Cu 3p3/2 = 75.1eV). Peaks were charge referenced to CF2 band in the C1s spectra at
291.7eV. Measurements were made at a takeoff angle of 45° with respect to the sample surface
plane which resulted in a typical sampling depth of 3-6nm. Quantification was done using
instrumental relative sensitivity factors (RSFs) that account for the X-ray cross-section and
80
inelastic mean free path of the electrons. Carbon 1s and oxygen 1s spectra were analyzed using
CasaXPS data processing software.
Time of flight secondary ion mass spectrometry (ToF-SIMS) was used as a depth profiling
technique to estimate the depth at which plasma surface modification affects the material. Depth
profiling was performed using a PHI nanoTOF II SIMS. A pulsed 30 keV Bi32+ primary beam
was used for spectroscopy. Sample etching was achieved with a 5 keV Ar2500+ cluster ion beam.
Both electron and ion charge neutralization occurred during depth profiling, and the LMIG and
GCIB raster sizes were 100 x 100 microns, and 500 x 500 microns respectively.
Surface roughness of the copolymer before and after plasma treatment was measured on a
NextView 3D Optical Profilometer provided by Zygo. A series of measurements with scan areas
of 87μm x 87μm were taken on each sample. Sq values were then averaged over a series of 8 scans
per sample.
The H2O contact angle for each plasma treatment condition and exposure time was
measured using a Ramé-Hart Model 295 goniometer and deionized water. Sessile drop
experiments began by depositing a 5μL volume of H2O onto the sample using a small remote
controlled hydraulic pipette. The measured contact angle is taken as the mean between left and
right contact angle of the water droplet at first impact on the sample surface. A total of 5
measurements at separate locations on each sample were made to produce a standard deviation in
the measurement.
4.2.5.2 Electrical
A Cascade Probe Station equipped with DCM 210 series Precision Positioner 20μm probes
were used to capture low and high field dielectric behavior. Dielectric spectroscopy was performed
using an Agilent Precision LCR meter. Each sample was exposed to 1V rms ranging from 100Hz
– 1MHz.
I(V) experiments were performed using the same experimental set up as used for dielectric
spectroscopy. A 4140 pA meter provided by Hewlett-Packard that served as its own DC source
was used to measure current as a function of applied voltage. Initially, a series of low voltages (2V
– 10V) were applied to the samples and the current as a function of time was measured to determine
the required hold time to achieve a steady state current for current/voltage measurements. The hold
time of 20s was chosen by comparing current densities measured at 20s vs 200s. Showing
81
approximately a 2.4x10-6A/m2 difference in the current density at 2V, it was reasoned that a 20s
hold time is sufficient to record data with fluctuations in current being well outside the data’s
range. The samples were cycled from 5V – 100V a total of two times: the first cycle is used to poll
the ferroelectric copolymer and the second cycle is used in data analysis. This process eliminated
polarization domain wall movement as a contributing factor to displacement current in high field
measurements, allowing the focus to be the effect of plasma treatment on high field J-E behavior
and charge transport. All measurements were taken at room temperature (22oC) and averaged over
three separate electrode pads to produce a standard deviation in the data for each sample set.
4.3 RESULTS AND DISCUSSION
4.3.1 Differential Scanning Calorimetry
DSC was performed between a temperature range of 20oC – 200oC for 1μm thick
copolymer using a ramp and cooling rate of 10 oC/min. It is common for P(VDF-TrFE) to display
two endothermic phase transitions because of its ferroelectric nature [132]: one corresponding to
its Curie temperature (TCurie) and another marking the melting temperature (Tm) [160]. Although
pure PVDF exists in different crystalline phases, the copolymer P(VDF-TrFE) is chemically
modified such that it only crystallizes in the piezoelectric beta phase. The TCurie of the material
thus marks the temperature that differentiates between the ferroelectric (T<TCurie) and paraelectric
(T>TCurie) phase of the material. Our scans indicate two distinct peaks in the data pertaining to
TCurie=110oC and Tm= 148oC peaks that are congruent with similar values reported for P(VDF-
TrFE) fabricated via various processing techniques in literature [160] [161] [162]. Calculations on
percent crystallinity were made by integrating the melting endotherm of three samples originating
from the same batch of copolymer. A melting enthalpy of ΔHm = 91.45 J/g [98] estimated from a
70/30 ratio of melting enthalpies of fully crystalline PVDF and TrFE, respectively was used. For
the as-spun (control) sample, an average of 30% crystallinity is calculated using three samples
originating from the same film. Annealing temperature was then chosen based on previous work
that investigated annealing temperature’s influence on crystal percent and morphology [163].
Samples in this study were annealed at a temperature of 142 oC for 24 hours under vacuum.
Applying this procedure increased the percent crystallinity of our samples to approximately 33%,
which is comparable to crystallinity reported for P(VDF-TrFE) cited in literature [164].
82
4.3.2 X-Ray Photoelectron Spectroscopy: Surface Chemistry and Structure Analysis
As expected the untreated control film contained peaks (Figure 4-1) due to CH2 (from
PVDF), CHF and CF2 located at 287.2±0.1eV, 289.5±0.1eV and 291.7eV, respectively [165]. In
addition, small peaks at 294.1±0.1eV due to CF3 end groups and adsorbed hydrocarbons at ~285eV
were observed. The elemental composition for the material was determined from C1s, F1s, and
O1s peak integration and is presented in Table 4-I. Untreated samples yielded a 45.9% C and
53.1% F after peak integration. This is in good agreement with the theoretical values of 47.0% C
and 53.0% F computed assuming a 70/30 molar ratio of PVDF to TrFE.
Plasma treated samples exhibit a noticeably different line shape in the C1s spectra.
Examination of the line shape and synthetic peak distribution between the 45s – 180s treatment
times indicated no significant change with increasing plasma exposure beyond 45 seconds,
therefore comparisons in spectral shape are made exclusively between the control (untreated) and
180s treated samples for simplicity.
Table 4-II contains the atom% of carbon and oxygen in the different chemical
environments calculated from the curve fits for all processing conditions and treatment times.
Small shifts in the synthetic fit components are observed; CFx generally displaying a slight shift
downward in energy (0.5 – 0.9eV) and CHx components shifting upward in energy (0.3 – 0.4eV).
In general, a reduction of total C from 45.9% to 35.3% and increase in O (1.0% - 6.6%) and F
(53.1% – 57.9%) species are observed after treatment. Peak fitting for the untreated and treated
samples supports O and F increase, and also indicate a large decrease of PVDF CH2 from 19.0%
Figure 4-1: C1s and O1s spectra for as-spun control as well as post-anneal samples. Line shape does not vary between
45s – 180s treatment times for all sample sets. Intensity axis heights share the same scaling for easy comparison in
each C1s and O1s group respectively.
83
Table 4-I: Elements detected by XPS in atomic percentage
Table 4-II: Chemical species determined by XPS in atomic percentage
to 5.1% after treatment. Peak fitting at 287.4eV indicates the addition of C=O constituents at a
relative quantity of 3.5%, previously un-seen in the untreated copolymer film. Use of the CHx fit
component after plasma treatment was no longer necessary to improve fit quality in the as-spun
processing conditions. This indicates the eradication of hydrocarbon-based contaminants adsorbed
onto the film’s surface after plasma treatment. The CHx component’s return for the post-anneal
processing condition could indicate that the environment inside the oven used for annealing is
responsible for a small degree of surface contamination.
Samples tested from the post-anneal sample set yield a C1s line shape that exhibits qualities
of both the untreated control and plasma treated samples from the as-spun sample set. The peak
integration of the 292.9eV synthetic component corresponding to either CF2-O or CF3 is increased
from 0.7% to 2.4% relative to the untreated control, however it remains below that which is
detected in the as-spun condition. This trend is observed for the C=O component at 287.4 eV as
well, yielding 0.8% in comparison to 3.5% detected in the as-spun sample set. Finally, the decrease
84
in CH2-CFx is measured to be less than that detected in the previous samples. The general trends
in atom percent of the post-anneal films differ than that of the former for F. Detected F% shows a
slight drop from 52.9% (untreated control) to 50.6% after processing, which is consistent with the
less pronounced CF3 peak in Figure 4-1. Similarly, C% also shows a decrease (from 47.0% to
43.3%), which is approximately half the decrease in C% measured in as-spun films. Similar to C,
the measured amount of O shows the same trend as the former processing condition but in lower
quantity, reflected by decreased C=O synthetic component intensity.
High resolution scans of the O1s core level spectra also show considerable change with
plasma treatment conditions. A similar trend in the line shape as seen in the C1s spectrum is noted:
untreated and plasma treated samples differ in shape however there is no distinction between line
shapes corresponding to 45s – 180s exposure times. In general, ≤ 1.0% O is detected on the surface
of each of the three control samples. This low quantity is expected for untreated films and could
be an artifact of processing while exposed to atmosphere. Although low in quantity, the signal is
strong enough for the as-spun control sample to be fit with two separate synthetic components. A
lower binding energy component centered at 532.2eV is attributed to O-C species where a weak
adjacent peak at 535.0±0.2eV is attributed to O-CF. After plasma treatment the band in the O1s
spectra at 535.0±0.2 eV due to O-CF species is considerably increased [165]. Although there is
growth in O-C, its increase is small in comparison to O-CF, a phenomenon that may be driven by
the plasma composition. Again, the post-anneal sample set produces a plasma treated O1s line
shape that is between untreated and plasma treated samples in the as-spun set. In this set, the O-C
signal remains relatively unaltered, however a decrease in the O-CF relative to other 180s treated
condition is noted. Lastly, a small but significant percentage of N is detected seen in Table 4-I.
This element is only present in films that have been exposed to plasma treatment.
4.3.3 ToF-SIMS Depth Profiling
An untreated copolymer film was first analyzed to obtain a baseline profile representative
of unmodified material. Measurable CFO+ and CO+ ion signals were detected in the untreated
samples (data not shown). Although these ions would not be expected in the spectra of pristine
P(VDF-TrFE), the XPS data indicate the presence of small amounts of O within the first 10 nm of
the untreated films.
85
Figure 4-2: ToF-SIMS depth profiles for CFO+ and CO+ ions in a plasma-treated film. This plot was generated by
taking the ratio of the corresponding profiles of the treated film to those of the untreated control film (plasma treated
signal/untreated signal).
Figure 4-2 shows that after plasma treatment, the intensities of the CFO+ and CO+ peaks
increase by 1 to 2 orders of magnitude near the surface of the film. The SIMS and XPS data thus
both indicate the presence of increased oxygen concentrations in the near-surface region of the
plasma-treated samples. As sputtered material emanated from deeper within a sample, the signal
for both oxygen-containing ions decays rapidly. At a depth of roughly 3 nm, the profiles flatten
out around about a value of 1 indicating the plasma treated sample is giving off an identical signal
to the untreated control film. Although determining a plasma-treated layer thickness with high
accuracy is not possible due to limitations in the depth resolution of the technique, it can be
concluded that the plasma modified layer thickness is less than 5nm, demonstrating the plasma
treatment processing procedure only modifies dielectric/electrode contact properties. This
information will be referred to later in section 4.3.6.2 when modeling is used to quantify the effect
of plasma treatment on charge injection.
4.3.4 Surface Roughness and Contact Angle Analysis
The root mean squared height (Sq) is plotted as a function of plasma exposure time in
Figure 4-3a. The as-spun untreated control sample was measured to have a surface roughness of
86
Figure 4-3: a) Copolymer surface roughness (Sq) as a function of plasma exposure time for the as-spun control and
post-anneal sample sets. Data is averaged from 8 separate 87μm x 87 μm scans per sample. b) Images of film surface
topology for each treatment condition taken by optical profilometry. Each image corresponds to the average Sq in a.
8.8 nm which is within the standard error of the control sample for the post-annealed sample set.
Plasma treated films between 45s and 180s in the as-spun sample were not able to be accurately
measured using the optical profilometer, presumably due to significant surface damage during the
plasma treatment process. This could be due to the processing of the as-spun sample set: films
were not annealed for an extended period of time after processing leaving them less structurally
robust than films in the post-anneal set. For this reason, the untreated control is used as a baseline
to compare films from the post-anneal set that did undergo annealing prior to measurement. Post-
anneal samples exhibit minor variation in surface roughness as a function of plasma treatment
time. The maximum value of surface roughness is 11.9nm occurring at 135s, which is the only
measured roughness that displays a statistically significant increase in comparison to the untreated
control. It is believed that application of thermal annealing after plasma treatment serves to repair
a significant portion of surface damage caused by exposure to plasma that is reflected by the near
identical surface topology of the post-anneal samples to the untreated control. A potential
mechanism for this phenomenon could be due to the close proximity of Ta to Tm of the material
resulting in polymer chain ordering which increases crystallinity. This motion acts to restore
surface topology to a similar state as that of the undamaged control sample. In this scenario, the
annealing procedure can be viewed as a method to restore the surface topology of the film, similar
to what has been accomplished in other systems such as PLLA [166]. The similarity in surface
87
Figure 4-4: a) Contact angle measured as a function of treatment time as-spin and post-anneal conditions. b) Images
of H2O sessile drops at first impact on copolymer surface for each treatment time.
roughness can also be observed in 2D topographical micrographs of the untreated control in
comparison to the post-anneal samples in Figure 4-3b.
It is suggested by the nature the chemical species involved in the plasma treatment that a
grafting process dependent on plasma composition alters the surface chemistry of the films [110].
Figure 4-4a shows the contact angle for untreated and plasma treated samples. The as-spun sample
set shows a near identical value in contact angle after 45s exposure in comparison to the control
sample, followed by a sharp increase at 90s. The contact angle then gradually increases as a
function of plasma exposure time to a value of 103o after 180s exposure. The post-anneal sample
set shows different behavior, starting with a sharp increase from the untreated 84o to 109o for the
45s treatment and then gradually increasing to a value of 111o for the 180s treated sample. These
trends are prominent considering visible changes in the sessile drop images in Figure 4-4b,
88
showing the evolution of the contact angle as a function of plasma exposure time for both sample
sets.
The contact angle can be affected by many parameters including surface chemistry, texture,
and probe liquid used [105]. Given the chemical homogeneity for all treatment times measured in
XPS, as well as plasma’s damage to the surface of as-spun films, the data suggest that both sample
surface topology and chemistry influence contact angle. The nearly unchanged contact angle
measured for the 45s treatment time (Figure 4-4a) in the as-spun set is likely due to a combination
of surface damage and chemical modification caused by plasma exposure. This observation is
similar to reductions in contact angle observed in sessile drop experiments performed on
sandblasted surfaces of PMMA using diiodomethane and glycerol [106]. The steady increase in
the contact angle beyond 45s exposure times could indicate the chemical environment gradually
dominating the damaged surface, which implies surface damage could decrease for longer etch
times. The 180s treatment time in the post-anneal set exhibits the highest contact angle, as well as
a surface roughness similar to the untreated sample. With nearly identical film topology to the
untreated control sample, surface chemistry can dominate the contact angle. With surface
chemistry determining the outcome of the measurement and the highly polar nature of H2O,
increase in contact angle indicates decreased surface polarity afforded by the grafting of the
aforementioned chemical species via plasma treatment in section 4.3.2.
4.3.5 Low Field Electrical Measurements.
Electrical testing concentrated on the post-annealed P(VDF-TrFE) which demonstrated the
largest change in contact angle in comparison to the pre-anneal and as-spun sets. Similarly, the
uniform surface topology allows the discussion to focus mainly on how chemical modification at
the electrode/dielectric interface impacts electrical performance without contributions from
damaged film surface topology.
4.3.5.1 Dielectric Spectroscopy
Dielectric spectroscopy was performed over the frequency range of 100 Hz – 0.1 MHz in
order to analyze dipolar contributions to the permittivity. Typically, the permittivity of PVDF
based materials shows gradual reduction between 100Hz and 100kHz at room temperature[227].
This decrease in permittivity is due to the ferroelectric nature of the material arising from polar
89
Figure 4-5: Dielectric response under 1V rms AC electrical signal for the post-anneal processing condition. Untreated,
45s, and 180s treated samples were tested.
crystalline domains distributed throughout the material that exhibit a broad distribution of
relaxation times at room temperature [167]. As frequency increases, the permittivity decreases and
loss tangent increases (Figure 4-5) marking the onset of the dipolar relaxation regime where the
response from permanent dipoles distributed throughout the material no longer contribute to
measured displacement current.
Examination of the permittivity as a function of frequency indicates similar behavior in
dielectric data reported for plasma treated PVDF within the range of 100Hz – 10kHz: CF4/O2
plasma treatment resulting in slightly decreased permittivity [17]. Although the trend is similar,
the decrease in the permittivity is not statistically significant in the 45s treated sample determined
by considerable overlap of measurement standard error bars in the sample set. There is however a
slightly larger reduction in the permittivity observed for the 180s treated sample relative to the
control that is calculated to be lower by 7.3% at 1kHz. This reduction is analyzed by considering
the +/- 80nm standard deviation in 1 μm film thickness measured via stylus profilometry in section
2.2. The 80nm fluctuation indicates an 8% error in the thickness of the film which is larger than
the decrease in permittivity calculated for the 180s sample. Since relative permittivity εr scales
linearly with sample thickness t via the relation εr=Ct/Aεo, it is not appropriate to conclude that
the decrease in permittivity observed is significant. This indicates that the plasma treated layer is
not thick enough to impact bulk permittivity properties of the treated films. It should be noted that
90
this result is in good accordance considering the plasma processing technique, where plasma
treatments typically achieve very shallow depths on the order of 10’s of nm [151].
4.3.5.2 Ohmic Current – Voltage (I-V) Experiments
In Figure 4-6, linear fitting for low voltage I-V experiments was done using a built in
Microsoft Excel linear regression function. The fitting was forced to intercept the point that I=V=0
to better represent a 0 current situation given by Ohm’s Law when the applied voltage is 0. The R2
value for each fit is used to verify that the data accurately represents the sample’s ohmic behavior.
All linear fits yielded a R2 ≥ 0.97, indicating a good fit within the voltage range of 0-3V. A decrease
in material resistivity is calculated for plasma treated samples. The resistivity of P(VDF-TrFE)
70/30 has been cited at 1x1012 Ω-m for solvent cast films 200μm thick molded by a hot pressing
procedure [98]. Although we use a different processing, our result for annealed thin films falls
close in resistivity at 8x1011 Ω-m. The lowering of resistivity in plasma treated films is believed
to result from changed contact properties between the metal and dielectric surface. This result is
consistent with past research that focused on the interplay between conduction and electrode
contact in PVDF Terpolymer (PVDF-TrFE-CTFE) films, indicate that conduction depends on both
electrode metal reactivity with the surface of the material and also interfacial barrier height [11].
Although Chen, et al. performed exclusively high field measurements, there is still a contact
Figure 4-6: Low field I(V) measurements for the post anneal sample set. The untreated control, 45s treated, and 180s
treatment times are shown.
91
limiting effect in low field I-V present in our systems. In treated films, a lowering of
metal/insulator contact resistance could result from a change in polymer surface chemistry prior
to electrode deposition increasing conduction even at low fields.
4.3.6 High Field Current Density – Electric Field (J-E) Measurements
4.3.6.1 Conduction Mechanism Identification: Schottky and Poole-Frenkel Modeling
I(V) data at high voltages is converted to current density as a function of electric field (J(E))
and plotted in Figure 4-7. Measurements display a similar trend to low field measurements that
indicate increased conduction in plasma-modified samples. Typically, the addition of polar
chemical moieties after plasma treatment to non-polar polymers result in lower conduction. This
is observed in polyimide (PI) films measured at high temperature after surface modification via O2
plasma treatment [152]. In this system, XPS and water contact angle experiments confirm the
surface of PI is chemically modified after plasma treatment. It is believed that this evolved
chemical environment at the surface of the film controls contact resistance between electrode and
material and enhances charge trapping and scattering effects at high fields and temperatures.
Another similar study examined the effect of non-reactive Ar plasma treatment on poly-p-xylylene
(PPX) films [168]. In this study, argon plasma was generated in a plasma reactor during the
polymerization process of PPX films and was expected to introduce chemical and physical defects
to the film. Current density was then measured as both a function of field ranging from 1 to 7
MV/m, and temperature ascending from -170 oC to room temperature at a rate of 6 oC/min. Results
indicated a two orders of magnitude reduction in the current density for all measured temperatures
in the plasma treated PPX relative to untreated, as well as increase in breakdown strength form 7
MV/m to 10 MV/m. Similar to Meddeb, et. al [152], suppression in the current density and increase
in the breakdown strength for plasma treated PPX is attributed to introduced defects acting as
scattering centers or localized charge trap sites. This conclusion coincides with literature that
suggests high
field performance will improve with increased quantity of dipolar groups within the material to
serve as electron scattering centers, as well as by adding polar moieties such as C-F groups in
plasma polymerized PPE, PPEF and PPET [169] [170].
92
Figure 4-7: High field J(E) measurements for the post-anneal processing condition. Again, the untreated control, 45s
and 180s treatment times are shown.
Clearly, P(VDF-TrFE) supports the trends reported in the literature, exhibiting a direct
correlation between high field conduction and surface environment altered by plasma treatment.
Due to the proximity of the treated layer to the metal electrode in contact with the film during
measurement, changes in high field conduction are likely related to Schottky emission, i.e., field
assisted thermionic emission from electrode contact into the dielectric under high electric fields
[113]. The functional form of the current density as a function of electric field is expressed in
equation (4-1):
𝐽 = 𝐴∗𝑇2𝑒𝛽𝑆𝐸
12
𝑘𝑏𝑇 𝑒−𝜙𝐵𝑘𝑇 (4 − 1)
where J is the current density, E is the electric field, T is absolute temperature in kelvin, ϕB is the
Schottky barrier height of the electrode/material contact at E=0, and A* is the Richardson constant
for the material under test. The βS parameter represents field-assisted barrier height lowering and
is expanded as such:
𝛽𝑆 = (𝑞3
4𝜋𝜖𝑜𝜖𝑟)
12
(4 − 2)
q being the value of elementary charge, and εoεr are vacuum and relative permittivity respectively.
Equations (4-1) and (4-2) indicate that current is dependent on the polarity of the material linked
by the relative permittivity εr.
93
Both Schottky and Poole-Frenkle (PF) conduction mechanisms must be considered in
plasma treated P(VDF-TrFE). Past research has applied PF theory to describe high field
conduction through bulk polytetrafluoroethylene (PTFE) under the assumption that its
polycrystalline and amorphous structure and high density of impurity centers lends itself to bulk-
limited PF conduction [171]. In this case, current emission as function of electric field is expressed
in the following manner:
𝐽 ∝ 𝐸𝑒𝛽𝑃𝐹𝐸
12
𝑘𝑏𝑇 𝑒−𝜙𝑃𝐹𝑘𝑇 (4 − 3).
The primary difference between Schottky and PF emission functional forms are ϕPF that now
describes the trap-state barrier height at E=0 and βPF that pertains to field-assisted trap-state barrier
height lowering. The functional form for βPF is the following:
𝛽𝑃𝐹 = (𝑞3
𝜋𝜖𝑜𝜖𝑟)
12
(4 − 4)
and differs by βs by a factor of 2. Typically high field data for Schottky and PF emission is handled
by plotting Ln(J) vs. E1/2 and Ln(J/E) vs. E1/2 respectively, and is plotted using the J(E) data
presented in Figure 4-7 for E > 60MV/m in Figure 4-8. A linear fit using Microsoft Excels linear
regression function is applied for each curve in Figure 4-8 with its associated equation shown. In
this format, the slope m of the data is equal to the quantity βs,PF/kbT. Given the expressions for βs
and βPF in equations 4-2 and 4-4, a value of relative permittivity for the films can be back
calculated from linear fit slopes found in the Schottky and PF plots via the two mathematical
relationships:
𝜖𝑟,𝑃𝐹 = [(𝑚𝑘𝑏𝑇)2𝜋𝜖𝑜
𝑞3]
−1
(4 − 5)
𝜖𝑟,𝑆 = [(𝑚𝑘𝑏𝑇)24𝜋𝜖𝑜
𝑞3]
−1
(4 − 6)
where m is the slope of the data obtained from linear fits to the data in Figure 4-8. Using equation
(4-5), permittivity from the PF plot for untreated films yield εr=124 and for plasma treated films,
94
Figure 4-8: Both Schottky and Poole-Frenkel plots for untreated and plasma treated P(VDF-TrFE) J(E) data are
displayed. The linear fit equation used to calculate permittivity from equations (4-5) and (4-6) for each set of data are
labeled and displayed next to the corresponding curve.
εr=171, both of which are an order of magnitude beyond the measured material permittivity for
P(VDF-TrFE) (Figure 4-5). Calculation of the material permittivity from the Schottky plot using
equation (4-6) yields εr=9.4 and εr=11.2 for untreated and plasma treated films respectively. This
is in very good agreement with the values of permittivity measured in Figure 4-5, supporting that
high field conduction is due to a Schottky type emission mechanism and not bulk limited under
high electric fields.
4.3.6.2 Quantifying the Change in Barrier Height
It is thought that the increase in the high field current density after plasma treatment is the
result of field emission across the metal/insulator boundary due to surface states created at the
plasma modified interface. The influence of surface states on high field conduction have been
documented for a number of systems including polymers (PI by Meddeb et al. [152]) and also
inorganic semiconducting materials such as Indium Phosphide (InP) and Zinc Oxide (ZnO) [172]
[173]. In the case of InP, the presence of surface states and defects at the interface between metal
and semiconductor limits the achievable barrier height due surface Fermi level pinning [174].
Similarly in the case of ZnO, a 20% O/80% He based plasma surface treatment showed both
evolved surface chemistry and a change in the contact properties between ZnO and Au electrodes
[172].
95
In the case of P(VDF-TrFE), a similar process is considered where increased high field
conduction is posited to be induced by decreased Schottky barrier height relating to grafted
chemical species at the electrode interface. Since the charge conduction mechanism in P(VDF-
TrFE) thin films fits well to Schottky injection theory, the Schottky equation is parametrically
explored to determine the most influential parameter in current density generation. The results of
a parametric study are shown in Figure 4-9 where the effect of each variable on the conduction
current was analytically determined. In this study, values for A* ranged two orders of magnitude:
from an upper value of 1.2x106 A(m-1K-2) calculated from its theoretical form of A*=4πqemekb2/h3
down to a lower value of 1.2x104 A(m-1K-2) [173]. The relatively small change in current density
is expected considering A* is a pre-exponential factor and is shown in Figure 4-9a.
The εr ranged from 2-20, where a permittivity of 2 reflects a non-polar material such as
polytetrafluoroethylene and 20 approaches values reported for terpolymer P(VDF-TrFE-CTFE)
[175]. In this analysis, current density has a larger dependence with the permittivity than the
Richardson constant; however, a large change in material permittivity would be required to
significantly alter high field emission. In the case of low values of εr and high fields, permittivity
must double from 2 to 4 to cause a change in current density by an order of magnitude. Similarly,
as the permittivity becomes large, as is the case for PVDF, the value of the exponential term,
(βSE1/2/kbT), from equation (4-1) becomes less sensitive to small changes in permittivity because
εr is in the denominator of the exponent. Here, doubling the εr from 10 to 20 increases current
density by less than an order of magnitude from 5x10-3 – 1.3x10-3 A/m2 which are values taken
from Fig 4-9b. The Schottky barrier height at E=0, ϕs. ranges from 0.55eV – 1.45eV and results in
16 orders of magnitude change in the current density with all other parameters from equation (4-
1) held constant. This result is reflective of the direct relationship between barrier height and
current density, where J α EXP(-ϕs), and indicates that the Schottky barrier height has the greatest
impact on the injected current under high electric fields.
The change in permittivity calculated from Schottky fitting using equation (4-6) in section
4.3.6.1 is now considered. Calculated permittivities from equation (4-6) for untreated and plasma
treated films yield εr=9.4 and εr=11.2 respectively, resulting in less than a factor of 2 decrease in
calculated current density. It is also noted that the permittivity calculated directly from capacitance
and loss measurements of the material shows no statistically significant change due to plasma
treatment by Figure 4-5 in section 4.3.5.2, indicating that permittivity is not an dominating factor
96
Figure 4-9: Parametric study of current generation during a Schottky emission process. Graph a, b, and c demonstrate
the impact of ranging A*, εr, and ϕS, respectively. Variables held constant and their respective values are shown above
each plot, while the upper and lower limits of the parameter ranged is indicated within the plot.
in controlling conduction. This suggests that a change in barrier height controls the increase in
conduction in plasma treated films.
The geometry of the applied surface coating is used to enable the calculation of
barrier height change in treated films relative to the untreated control. ToF-SIMS experiments
(section 4.3.3) indicate the plasma treatment process has little to no impact on chemical structure
of the material beyond the first 5 nm of the sample surface, thus behaving as a surface coating
only. This allows A* to be treated as a constant term which is similar to work done by Reddy [173]
where A* for n-type Indium Phosphide (n-InP) is used to calculate change in barrier height in films
both with and without a 30nm thick surface coating of PVDF. The change in material properties
for plasma treated samples relative to untreated films are isolated by way of a curve subtraction
procedure, where the linear fits in Figure 4-8 for the Schottky plot are subtracted as follows:
𝑙𝑛(𝐽1) − 𝑙𝑛(𝐽2) = [𝛽𝑆1
𝑘𝑏𝑇𝐸
12 −
𝜑𝑆1
𝑘𝑏𝑇+ 𝑙𝑛(𝐴∗𝑇2)] − [
𝛽𝑆2
𝑘𝑏𝑇𝐸
12 −
𝜑𝑆2
𝑘𝑏𝑇+ 𝑙𝑛(𝐴∗𝑇2)] (4 − 7)
where the subscripts 1 and 2 represent plasma treated and untreated curves in Figure 4-8
respectively. In this format, both Ln(A*T2) terms cancel in equation (4-5) due to constant
measurement temperature and A* between the samples, leaving the following expression:
𝑙𝑛(𝐽1) − 𝑙𝑛(𝐽2) =1
𝑘𝑏𝑇[(𝛽𝑆1 − 𝛽𝑆2)𝐸
12 + (𝜑𝑆2 − 𝜑𝑆1)] (4 − 8)
and at E=0, the change Schottky barrier height φs is solved in terms of change in ln(J):
∆𝜑𝑠 = 𝑘𝑏𝑇[∆𝑙𝑛(𝐽)]𝐸=0 (4 − 9)
Change in the Schottky barrier height can be simply computed by multiplying the curve’s change
in Ln(J)E=0 intercept by kbT. Using equation (4-9), a barrier height lowering of 0.05eV is
calculated. This reduction in barrier height is in good agreement with an order of magnitude
97
increase in current density computed using equation (4-1) from the parametric study when A* and
εr are held constant.
Due to similarity in the surface topology of the post-annealed plasma treated film tested
and the untreated control sample, the calculated change in barrier height is attributed
predominantly to surface chemical modification by plasma treatment. In this scenario, grafted
foreign chemical groups to the surface of the film behave like defects, decreasing the barrier height
from its natural level. Another cause for lower barrier height and increased emission could relate
to the increased amount of electronegative O and F based chemical moieties detected at the surface
of the film which may behave as acceptor type states. An increase in acceptor states at the surface
would provide a higher density of hole carriers at the plasma modified interface in comparison to
the bulk. This increase in available carrier density could relate to enhanced hole injection under
high fields, correlating well to past research that demonstrates conduction through PVDF based
polymers is dominated by a predominantly hole transport process [11] [176].
4.4 CONCLUSIONS
Plasma treatment with a 50/50 CF4/O2 gas composition was applied to P(VDF-TrFE)
polymer film between 45s and 180s at a constant power of 500W. Synthetic peak fitting of the
XPS C1s spectrum indicated the addition of carbonyl C=O groups on treated films. Fitting of the
O1s spectrum also indicated the uptake of CF-O and C-O moieties. It was determined that the
treatment time duration had no effect on the quantity of chemical species detected by XPS,
however annealing after plasma treatment does affect the ratio of F/C detected.
The water contact angle showed a statistically significant change as a function of treatment
time for as-spun and post-anneal samples. The near constant contact angle for various plasma
treatment durations in the post-anneal set reflects both constant surface chemistry detected in XPS
after treatment as well as uniform surface roughness measured via profilometry. Constant surface
topology allows the impact of introduced chemical species to dominate, revealing that of the
grafted chemical species after plasma treatment cause surface polarity to decrease.
Low field dielectric spectroscopy was used to further investigate the effect of grafted
chemical species on the polarity of the films. Data indicates that there is no statistically significant
relationship between exposure time to CF4/O2 plasma and change in the material’s dielectric
constant. This indicates that plasma treatment only impacts P(VDF-TrFE) material structure local
98
to the surface of the film, leaving the bulk of the films unaltered. Plasma treatment duration was
linked to low and high field I-V measurements where leakage current increased as a function of
plasma treatment time. Lowered contact resistance of the electrode/dielectric interface is thought
to influence conduction through the films at low fields. Enhanced high field current generated in
plasma treated films is found to be describable by Schottky theory. Mathematical analysis
implementing approximations acquired by a parametric study of the Schottky equation as well as
work done by Reddy [173] linked increased conduction to a lowered Schottky barrier height by
0.05 eV after plasma treatment. Finally, two potential scenarios linking enhanced conduction to
processing and surface chemistry are proposed: 1) grafted chemical constituents behave as defects
that lower Schottky barrier height and 2) increased F and O moieties behaving as acceptor type
states that increase hole carrier density at the electrode contact, linking enhanced conduction to
hole transport across the metal dielectric interface.
The effect of plasma surface treatment on high field conduction emphasizes the importance
of interface chemistry between metal electrode and dielectric in thin P(VDF-TrFE) films. Although
application of CF4/O2 plasma results in degraded high field performance of the material, the plasma
processing technique shows the potential to significantly impact leakage current under large
electrical loading. This study demonstrates the viability of surface chemical modification
techniques to change conduction properties in P(VDF-TrFE) and should be explored further to
develop new ways of limiting current leakage in materials for high field applications.
In the next chapter, the effect of plasma treatment is studied on a non-polar polymer
currently being considered for high voltage and high temperature applications: polyimide.
Techniques used for high field data analysis are expanded upon to incorporate curve fitting
procedures that can account for hopping theoretical analysis in which linearization is not possible.
4.5 ACKNOWLEDGEMENTS
The Authors of this publication would like to acknowledge the support of the National
Science Foundation as part of the Center for Dielectrics and Piezoelectrics under grant Nos. IIP-
1361571 and IIP-1361503.
99
CHAPTER 5
CONDUCTION THROUGH PLASMA-TREATED POLYIMIDE: ANALYSIS OF HIGH-
FIELD CONDUCTION BY HOPPING AND SCHOTTKY THEORY6
ABSTRACT
In this chapter, analysis on high field current density vs electric field (J(E)) on plasma
treated P(VDF-TrFE) is expanded upon to describe conduction through plasma treated Polyimide
(PI) as a function of electric field and temperature. Unlike copolymer, PI has a non-polar material
structure as well as excellent thermal stability to temperatures upwards of 400oC. These
characteristics cause the material to exhibit a different high field conduction dependence on plasma
treatment than that of P(VDF-TrFE), requiring the use of bulk dominated conduction theories to
accurately describe its behavior. Hopping theory is now considered by using nonlinear regression
techniques because of its non-linearizable mathematical form. Chapter 5 is broken into the
following sections:
Introduction – presents the scope in which PI is considered as a dielectric for high power
applications as well as introducing work by Meddeb et al. [13] which is the parent study in
which data in this chapter originates.
Analytical Methods – past work concerning PI conduction is reviewed, as well as the basics
of theory implemented in high field conduction analysis of PI in this chapter. A more
detailed description of theory can be found in Chapter 2 section 2.3.3.2.
Results and Analysis – high field J(E) data from Meddeb et al. [13]
https://doi.org/10.1016/j.cplett.2016.02.037) are analyzed using PF, Hopping and Shottky
theory and the results discussed.
Discussion and Conclusions – results from theoretical analysis are discussed in context of
plasma treatment as a tool to tailor dielectric surface properties.
6 This chapter is based off of published work by Vecchio et al. in the Journal of Materials Science. Data analyzed via Hopping, PF, and Shottky
theories is acquired from Meddeb et al. [13]
100
The results of analysis not only inform on plasma treatment’s impact on electrical properties of PI,
but also provide more rigorous analytical techniques to quantify material properties associated
with conduction using hopping theory.
5.1 INTRODUCTION
There exists a need to develop next generation materials for power electronic applications
with lighter weights and smaller volumes, yet capable of operating with high efficiency at high
temperatures. Organic dielectric materials are one option, due to their ease of manufacturability,
light weight and favorable electrical properties such as high breakdown strength. Currently,
biaxially oriented polypropylene (BOPP) is considered the state-of-the art in capacitor films,
exhibiting a breakdown strength on the order of 850 MV/m [65] and theoretical power densities
calculated to be in the range of 108 W/cm3 [177]. Other materials relevant to high voltage dielectric
applications are polar polymers such as polyvinylidene fluoride (PVDF) along with its associated
copolymer poly(vinylidene fluoride-trifluoroethylene) (P(VDF-TrFE)) and terpolymer
poly(vinylidene fluoride-trifluoroethylene-chlorotrifluoroethylene) (P(VDF-TrFE-CTFE)).
Recent research done on blended PVDF/ P(VDF-TrFE-CTFE) composites show impressive
breakdown strengths (640MV/m) and energy densities measured as high as ~20 J/cm3 [9] [83].
Routes to improve the high field performance of pure organic dielectrics that depend on more
involved processing procedures have also been studied. Past research done on organic dielectrics
fabricated by the coextrusion of P(VDF-HFP) and polycarbonate (PC) demonstrate the ability for
the composite material to achieve a breakdown strength that exceeds that of either constituent
material [14].
Although these materials show promise to develop the class of next generation power and
energy dense organic electronics, their low melting temperatures near 160oC limits their ability to
withstand energy dissipation attributed to high dielectric loss and leakage currents. This can be
conceptualized by a device’s performance represented by a Ragone plot: in the case of power
electronics the boundaries of capacitor performance are represented by power vs. energy density
and are determined by the material’s internal losses and/or leakage [8]. In the case of PVDF, its
polar nature causes the material to be inherently susceptible to high dielectric loss [149] exhibiting
loss tangents that reach as high as ~0.5-1.0% [71].7 As a result, PVDF experiences high leakage
7 See Chapter 4, sections 4.3.5.2 and 4.3.6 for conduction characteristics at low and high field of P(VDF-TrFE).
101
currents at DC frequencies even at low temperatures, within the range of 10-3 – 10-2 A/m2 at fields
exceeding 20 MV m-1 depending on material geometry and processing technique [71] [122]. In the
case of non-polar BOPP, dielectric absorption is very low, exhibiting dielectric loss tangents as
low as 10-4 at low fields [178]. However, the application of high electric fields required for high
power and energy applications causes energy dissipation in the form of leakage current. This leads
to joule heating, a precursor to dielectric breakdown [179] [180], causing temperature elevation of
the material regardless of ambient operating conditions, and bringing the material closer to its
relatively low derating temperature of 80oC [149].
Another influence which limits the applicability of these materials is ambient operating
temperatures for applications that require thermal stability within the range of PVDF’s and BOPP’s
melting temperatures. For example, passive electronics fabricated for automotive applications
must retain continuous structural and electrical stability within 150oC – 200oC, measured on-
engine and in-transmission during vehicle operation [181] [182]. This creates a demand for
volumetrically efficient capacitors that can maintain high voltage/temperature stability
manufactured at low costs. Unlike BOPP, PC, and PVDF blends and composites, polyimide (PI)
exhibits a combination of large breakdown strength measured at 615 MV m-1 [13] and high glass
transition temperatures reported in the range of 370oC – 470oC depending on chemical structure
and processing [67] [183]. Low dielectric dissipation factor in the range of 10-3 (provided by
DupontTM Kapton® HN material data sheet as well as literature [184]) and a potential of high
energy and power density add to its appeal as a next generation high field dielectric material.
Despite these favorable properties of PI, high temperature environments combined with charge
injection at the electrode/dielectric interface put the material at risk for joule heating and high
leakage currents during operation. Implementation of processing procedures that can limit the
leakage current through PI at high temperatures are required to achieve its full potential as a high
temperature dielectric polymer.
Reactive plasma treatments of polymers have been shown to have a significant impact on
material chemistry [13] [153] [154], and properties such as wettability [34] [155] and adhesion
[35] [156] [157] [158]. In addition, plasma treatment has been demonstrated to influence electrical
properties. In previous work by Mammone et al. [159], plasma treatment processing using 96%
CF4/ 4% O2 on polypropylene resin prior to melt extrusion was shown to increase the dielectric
breakdown strength of the material by 20%. Similar work done that uses 96% CF4/ 4% O2 plasma
102
treatment on the surface of 12-micron PVDF films led to 11% increase in dielectric breakdown
strength relative to untreated control samples [17]. Recently, research done in our group by
Meddeb, et al. [13] showed that plasma surface treatment using pure O2 on PI resulted in reduced
spread in dielectric data, improved Weibull modulus at low temperatures and improved breakdown
strength at elevated temperatures. Among these observations, an order of magnitude reduction in
leakage current at 150oC was also noted. This study provided a thorough chemical analysis using
XPS and H2O sessile drop experiments to characterize the surface of the films after plasma
treatment and surmised that the inclusion of O containing species at the surface of the film resulted
in charge trapping and scattering effects at the electrode / dielectric interface at high temperatures.
Meddeb et al. [13] also posited that the mechanism dominating conduction through the material
could potentially be changed and controlled using surface chemical modification. To this date and
to the best of our knowledge, this hypothesis remains untested. In the present work, the current
density vs electric field [J(E)] behavior of PI films measured in Meddeb et al. [13] (DOI:
10.1016/j.cplett.2016.02.037) is analyzed using a series of theoretical frameworks that describe
conduction through insulators and semiconductors to investigate the effect that O2 plasma surface
treatment has on high field conduction in polyimides over a broad temperature range. This study
will leverage further research that focuses on the role of dielectric/electrode interface in controlling
conduction in PI and other organic dielectrics.
5.2 ANALYTICAL METHODS8
Past research on PI has demonstrated the material’s propensity to conduct via various
charge transport mechanisms. Work by Diaham et al. [24] investigated conduction characteristics
of thin (1.4 μm) PI films under high electric fields (in the range of 100MV/m) and high
temperatures (320 – 400oC). In Diaham’s work [24], analysis of pre-breakdown currents as a
function of field and temperature suggests conduction at very high temperatures and fields
preceding dielectric breakdown is a space charge dominated process. This result coincides well
with results obtained by Kishi, et al. [185] where pulsed electroacoustic measurements
demonstrate hetero space charge formation increase as a function of field intensity and temperature
prior to breakdown. Both of these studies involve either application of very high electric fields for
8 This section provides a general framework for Hopping, PF, and Schottky theory used in this section, however a detailed explanation high field
conduction theory can be found in Chapter 2 section 2.3.3.2.
103
long times (Kishi ~260 MV/m at t>10min [185]) or sample exposure to extremely high temperature
(Diaham T=400oC [24]) to stress the material in the range of pre-dielectric breakdown. Space
charge measurements under less extreme conditions by laser intensity modulation method in 10
μm PI films for long time scales (t~30 min) and moderate fields ranging 0.5 – 2 MV/m show
charge accumulation within the samples dependent on electronic charge injection at the anode and
cathode [186]. Given these results, the analysis in this manuscript will be carried out using a
combination of theoretical frameworks that describe both bulk dominated and interface dominated
processes which are reviewed in the following section.
5.2.1 Bulk Dominated Conduction
Conduction through polymer dielectrics are typically analyzed using one of two theoretical
frameworks: Poole-Frenkel (PF) or Hopping [187]. The PF model for conduction describes field-
enhanced thermal de-trapping of electrons or holes from impurity centers distributed through the
bulk of the material [113]. In the case of PI and other organics such as polytetrafluoroethylene
(PTFE), this framework is typically considered because of the high density of impurity centers
located within the bulk of the film arising from its partly amorphous structure [19]. The PF
transport current is defined by equation (5-1):
𝐽𝑃𝐹 = σ𝑜𝐸𝑒𝑥𝑝 (𝛽𝑃𝐹𝐸
12
𝑘𝑏𝑇)exp (−
𝜙𝑃𝐹
𝑘𝑏𝑇) (5 − 1)
where JPF is the current density, σo is a constant related to charge, carrier density, and carrier
mobility, E is the electric field, kb is Boltzmann’s constant, T is temperature, ϕPF is the trap barrier
height at zero field, and βPF is a constant. In equation (5-1), βPF = (q3/πεoεr)1/2 where q is the
elementary charge and is related to field assisted barrier height lowering in the material and εr is
the material’s dielectric constant.
Hopping conduction theory is similar to PF theory in that it describes a conduction process
occurring through the material’s bulk. In hopping, the charge carrier undergoes a diffusion based
carrier hopping process [187] [113] [188]. The mathematical framework that governs this process
is presented in equation (5-2):
𝐽𝐻 = 𝐽𝑜𝑒𝑥𝑝 (−𝜙𝐻
𝑘𝑏𝑇) sinh [
𝑞𝑑𝐸
2𝑘𝑏𝑇] (5 − 2).
104
In equation (5-2), Jo is a constant of the form nfdq with n defined as a carrier density, f carrier
vibration frequency, d the distance between trap sites termed the “hop distance”, and q the
elementary charge. The term ϕH is typically referred to as an activation energy or barrier height for
the hopping conduction process [113] [118].
5.2.2 Interface Dominated Conduction
In this work we also consider the possibility of Schottky type emission. Schottky theory
assumes that conduction is dominated by charge emission through the dielectric / electrode
interface. This framework is considered in the analysis of plasma treated PI for two reasons: 1)
the fact that plasma treatment is a surface modification process, and 2) the treatment’s influence
on conduction that has been observed at high temperatures [13]. The mathematical framework
for Schottky emission is very similar to PF type conduction and is presented in equation (5-3)
below.
𝐽 = 𝐴∗𝑇2𝑒𝑥𝑝(𝛽𝑆𝐸
12
𝑘𝑏𝑇)exp (−
𝜙𝑆
𝑘𝑏𝑇) (5 − 3)
Equation (5-3) differs from (5-1) by the three main parameters A*, ϕS and βS. The constant A* is
the Richardson constant and has the form 4πqmek2/h3, with me being the mass of an electron and h
is Plank’s constant. ϕS is the Schottky barrier height defined by the difference between metal
contact work function and electron affinity of the test material, and βS is a constant related to field
assisted barrier height lowering. In Schottky, βS = (q3/4πεoεr)1/2 and is slightly different than that
of PF emission because it takes into account the image force potential associated with charge
leaving the electrode during injection.
5.3 RESULTS AND ANALYSIS
5.3.1 Leakage Current Results
Kapton® PI samples, 13 μm-thick, were obtained from Dupont. The PI and PPIDS sample
sets consists of three untreated and plasma treated samples (respectfully) electroded using 100nm
evaporated silver. During electrode evaporation, each sample set was exposed to high vacuum (10-
6 Pa) for approximately 2 hours. Electrodes were circular and 1 cm in diameter for each sample.
The three samples labeled PPIDS are plasma treated with O2 on both sides of the film prior to
electrodes deposition. The flow rate was held at 200 sccm and the treatment power was 200W that
105
lasted 5 min/side. During plasma treatment, low vacuum was applied, stabilizing between 0.42 –
0.45 Pa. Details concerning specifics about PI surface chemical characterization as well as J(E)
data acquisition are present in the parent study and made available to the dedicated reader in
Meddeb et al. [13] (DOI: 10.1016/j.cplett.2016.02.037).
Current voltage measurements were performed using a Hewlett Packard 4140B pA meter
that served as its own DC source connected to a Trek voltage amplifier model 10/10B-HS. Initial
measurement of the charging current transients in PI was done at the lowest fields of measurement
used in this study (8 MV/m – 23 MV/m) and at room temperature. Figure 5-1 shows charging
current as a function of time. The lowest current measured was 40pA (after a 100s voltage hold
time) with 8 MV/m applied, which is well above the minimum 1pA that can be measured by the
pA meter. Steady state in the measurement begins at 20s indicated by a plateau in charging current.
At higher temperature and fields, plateauing in the current occurs at shorter times, making 20s a
reasonable hold time where a) steady state in the current is achieved and b) the chance for sample
degradation at high fields and high temperatures is minimized.
A significant reduction in leakage current through PI was measured following O2 plasma
surface treatment in our previous study by Meddeb et al. [13] and is shown in Figure 5-2. At low
temperatures the average current density of both the untreated PI and O2 plasma treated PPIDS
data sets are within the same range. As temperature increases to 150oC, the average current density
for PI at 115 MV m-1 is 4.04x10-4 A m-2 which is an order of magnitude larger than the PPIDS
Figure 5-1: Charging current measured for PI as a function of charge time at room temperature.
106
measured to be 4.07x10-5 A m-2. At all temperatures there is less scatter in the data for plasma
treated samples relative to the untreated control set, however this is especially true at high
temperatures and electric fields. Due to the similarities in sample storage and processing
between the PI and PPIDS sample sets, changes in current magnitude and scatter are linked to a
change in surface chemistry after the plasma treatment procedure: increases in hydroxyl as well as
carbonyl and carboxyl groups due to treatment are detailed in Meddeb et al. [13], and have also
been reported by Inagaki et al. [189] to be present in Kapton® films post oxygen plasma treatment.
The impact that plasma treatment has on the magnitude of leakage current through
the film as a function of temperature and field can be seen from the data presented in Figure 5-2,
however the exact mechanisms responsible for this reduction are not yet determined. It is important
to seek out the exact mechanisms by which plasma surface treatment limits conduction so that
similar results can be exploited to decrease leakage currents and improve high field and high
temperature performance of dielectric films. This requires quantitative analysis under a set of
theoretical frameworks that describe high field conduction through dielectrics. The following
sections use the theoretical frameworks that describe bulk dominated and interface dominated
conduction to quantify how plasma surface treatment achieves reduced leakage current and discuss
its impact for future high temperature capacitor development.
5.3.2 Data Analysis
5.3.2.1 Modeling Bulk Conduction
Initially, PF theory is considered. The linearization of equation (5-1) leaves the following
expression for conduction under PF framework:
𝐿𝑛 (𝐽𝑃𝐹
𝐸) =
𝛽𝑃𝐹𝐸1/2
𝑘𝑏𝑇−
𝜙𝑃𝐹
𝑘𝑏𝑇+ 𝐿𝑛(𝜎) (5 − 4)
The slope of linearized data can be directly related to βPF at a single temperature T. Since J(E) data
was recorded isothermally at temperatures ranging between 25 oC – 175 oC, the permittivity of PI
can be extracted directly from the slope of the PF plots using the expression for βPF stated in
section 5.2.1. This procedure is done at each temperature by applying a linear regression fit to the
data plotted in PF plot format (Ln(J/E) vs. E1/2). The slope m of the regression function is then
converted to permittivity using the following equation:
107
Figure 5-2: J(E) data for untreated and O2 plasma surface treated PI at a) 25oC, b) 100oC, and c) 150oC. Applied
voltage is held for 20s (time required to obtain steady state conduction) before current measurement at each field.
108
𝜖𝑟,𝑃𝐹 = [(𝑚𝑘𝑏𝑇)2𝜋𝜖𝑜
𝑞3]
−1
(5 − 5)
Treatment of the data in this manner produces εr,PF values that range between 500 and 20
depending on temperature. Considering that PI’s accepted range in permittivity is between ~2.25
at high frequencies and ~3.15 at 1kHz, it is concluded that PF theory inadequately describes
behavior of the material.9
Past work done on PI by Sawa, et al. [120] and Sacher [121] suggest that room temperature
conduction through the material is a result of proton hopping originating from carboxyl groups in
unreacted polyamic acid (PAA) within the bulk of the material. Hopping conduction is described
using equation (5-2) and is more complex than equations (5-1) and (5-3). This equation cannot
be linearized to calculate fit parameters Jo, ϕH, and d. Instead, fitting is performed using the
statistical software R-Studio where a non-linear regression technique optimizes the parameters Jo
and d to minimize the sum of squares between equation (2) and raw data.
In hopping conduction theory, the value of ϕH is taken as a low field activation energy. In
this study, low field activation energy is estimated using the following Arrhenius type equation for
conduction
𝐼 = 𝐼𝑜𝑒𝑥𝑝 (−𝜙𝐻
𝑘𝑏𝑇) (5-6)
where I is measured current, and I0 is a constant associated with conduction as ϕH approaches 0.
Figure 5-2 demonstrates how surface chemical modification reduces leakage current, potentially
impacting the conduction mechanism at high temperatures and fields. For this reason, the PI set is
used to extract a low field activation energy to exclude effects introduced by plasma treatment.
I(V) data for the PI set are plotted as a function of Ln(I) vs. T-1 and presented in Figure 5-3.
Measurements taken at 100V at each temperature were used for low field currents. The
data is described well using the Arrhenius relation in equation (5-6) and returns an activation
energy of 0.25 eV, which is within range reported in studies where PI was processed by solution
9 More information on PF analysis including treatment of data and calculated values of εr,PF are provided in Appendix C
109
Figure 5-3. Arrhenius plot for Pure P1 over the temperature range 25oC – 150oC. Activation energy is extracted
from linearization of equation (5-6) and the slope of the linear fit function.
casting and vapor deposition under AC and DC field conditions [190] [118]. Similarly, ϕH < 1.0
eV is indicative of electronic based conduction in the film, implying electrons may play a role as
charge carriers during DC conduction in the film [190]. This result can be related back to work
done by Ito, et al. [118], Sawa et al. [120], and Sacher [121] where conduction through PI films
was found to be dependent on proton hopping stemming from carboxyl group presence due to
unreacted PAA. In the current study, PI films are obtained from Dupont and are of high quality.
Although the complete absence of proton conduction from carboxyl entities is not guaranteed, the
quality of the films suggests limited proton conduction resulting from unreacted PAA compared
to PI films studied by Sacher [121]. This analysis coincides well with work done by Tu and Kao
[191], where C-V experiments performed on 100% imidized PI suggest conduction is dominated
by an electron transport process, potentially occurring via hopping along acceptor states both along
the polymer chain’s backbone and between adjacent chains.
A combination of physical arguments as well as statistical analysis was used to assess the
validity of the hopping model to explain the J(E) behavior. Due to the form of the function used to
fit J(E) data, calculation of P values should not be implemented to determine parameter estimate
significance. Non-linear models do not have a well-defined relationship between parameter (in this
case d and Jo) and predictor variables (applied field E) which inhibits the creation of a single
hypothesis test that can represent all nonlinear models. A more appropriate method to determine
parameter significance for our purposes would be computation of a confidence interval for each
110
Figure 5-4 Hopping parameters Jo and d estimated using the bootstrapping statistical approach.
parameter estimate. Unfortunately, the distributions of fit parameters describing the sample sets
are unknown and difficult to extract from measurements given the sample size N = 3 in each sample
set. To estimate a distribution for the parameters, bootstrap statistics are employed where variation
of fit parameters that describe the data set can be estimated by sampling with replacement. This
method yields a confidence interval for Jo and d that reflects variation among the samples
comprising PI and PPIDS sample sets.10
Physical arguments concerning the model’s validity begin by examining the hop distance.
Values of d are first considered to assess how well the hopping model describes sample behavior
and are plotted in Figure 5-4 for both sample sets as a function of temperature. Hop distance
estimated for both sample sets ranges between ~10 Å – 30 Å, which is within the range of hop
distances reported in the literature for PI and also BOPP [118] [179] [192] [193]. At room
temperature, d is estimated to be approximately 12 Å, which is quite close to the length of the
chemical repeat unit of PI (~15 Å measured by x-ray diffraction in Ito et al. [118]) which suggests
hopping conduction propagates in the polymer chain direction through the material. At higher
temperatures the hop distance increases in both sample sets: 25 Å in PI at 75oC and 24 Å at 125oC
in PPIDS. This exceeds the chemical repeat unit for the material and may reflect a transition from
intrachain to interchain charge carrier hopping at higher temperatures.
10 A more detailed outline of the bootstrap statistical approach can be found in Chapter 2 section 2.4.1 along with annotated R-Studio script used
for fitting and bootstrap analysis
111
Jo is also analyzed across the measurement’s temperature range. Using the relation Jo =
nfqd, a nf (product of carrier density and vibration frequency) can be calculated from fit estimates.
The product nf ranges from 1.6x1025 to 4.4x1025 m-3s-1 in PI between 25oC – 75oC, and from
5.6x1024 and 8.0x1025 m-3s-1 in the PPIDS set between 25oC – 125oC. These ranges are compared
to previous work done by Ito, et al. [118] that focus on the effect of curing vapor deposited PI
films. Films undergoing 4 hours of curing returned nf values after fitting on the order of 1x1021 m-
3s-1 while uncured films returned 1x1035 m-3s-1. The films used in this study produce a value of nf
within the range reported by Ito, et al. Small differences can be expected considering the difference
in sample preparation, as well as fitting procedures used: Ito, et al. [118] obtains higher activation
energy of 0.34 eV and uses a series of low and high field approximations on equation (5-2) in
fitting.
Statistical interpretation of the fit results was used to comment on the behavior of the
sample set as a function of temperature. At 25oC there is very little change in the estimated values
of either fit parameters between sample sets. This is an expected outcome considering similarity
in behavior between PI and PPIDS sample sets at low temperature reported in Figure 5-2. It is
also noted that the estimates of d and Jo are centered within their respective confidence intervals
computed by bootstrapping which are presented in Table 5-I for PI. This implies that the estimated
parameters follow a symmetric distribution, which is most likely caused by the samples within the
data sets behaving similarly to one another as a function of electric field. As temperature is
increased to 100oC, a significant separation in the values of Jo and d estimated from PI and PPIDS
sets occurs. Both parameters for the PI set exceed values computed for PPIDS. Greater change can
be noted in the 95% confidence intervals for both fit parameters in PI. At 100oC, the confidence
interval pertaining to d ranges a full order of magnitude. Similarly, the estimated value of d is no
longer centered within the confidence interval, indicating an asymmetric distribution for hop
distance. A similar outcome exists for Jo displaying an asymmetric distribution with confidence
interval ranging 7 orders of magnitude at 100oC, indicating a major variation between the samples
comprising the PI data set. These outcomes of fitting d and Jo at high temperature diminish the
significance of their estimated values, indicating a poor fit to the data. Unlike PI, the PPIDS sample
set does not show a pronounced change in fit parameters within 25oC – 100oC. The confidence
intervals for both fit parameters are also significantly tighter and do not display asymmetry, as
shown in Table 5-II, suggesting that samples within the set behaving quite similarly as a function
112
of field within this temperature range, coinciding nicely with small standard deviations calculated
for PPIDS J(E) data in Figure 5-2.11
Table 5-I: parameter fit values and confidence intervals at each temperature for untreated PI sample set. No values
for Jo or d are provided because of the program’s inability to fit the data using equation (5-2).
Temperature
(oC):
25 75 100 125 150 175
Joestimated
: 6.98x104 7.36 x10-4 2.56 x10-3 --- --- ---
2.5% CI 3.66 x10-4 1.03 x10-5 1.67 x10-9 --- --- ---
97.5% CI 1.16 x10-3 2.23 x10-3 8.25 x10-3 --- --- ---
destimated
: 1.26 x10-9 2.50 x10-9 3.18 x10-9 --- --- ---
2.5% CI 9.52 x10-10 1.38 x10-9 1.03 x10-9 --- --- ---
97.5% CI 1.60 x10-9 4.66 x10-9 1.04 x10-8 --- --- ---
Table 5-II: parameter fit values and confidence intervals at each temperature for the plasma treated PPIDS sample
set
Temperature
(oC):
25 75 100 125 150 175
Joestimated
: 7.36 x10-4 1.53 x10-3 7.12 x10-4 3.94 x10-4 2.34 x10-4 2.70 x10-4
2.5% CI 6.17 x10-4 9.09 x10-4 4.24 x10-4 2.45 x10-4 1.61 x10-4 1.84 x10-4
97.5% CI 8.70 x10-4 2.45 x10-3 1.13 x10-3 5.74 x10-4 3.03 x10-4 3.58 x10-4
destimated
: 1.19 x10-9 1.04 x10-9 1.44 x10-9 1.86 x10-9 2.55 x10-9 2.87 x10-9
2.5% CI 1.10 x10-9 7.59 x10-10 1.11 x10-9 1.57 x10-9 2.34 x10-9 2.64 x10-9
97.5% CI 1.27 x10-9 1.32 x10-9 1.75 x10-9 2.18 x10-9 2.84 x10-9 3.18 x10-9
5.3.2.2 Interface Dominated Conduction
Schottky analysis was performed in a similar manner to PF by the linearization of
equation (5-3):
𝐿𝑛(𝐽𝑆) =𝛽𝑆𝐸1/2
𝑘𝑇−
𝜙𝑆
𝑘𝑇+ 𝐿𝑛(𝐴∗𝑇2) (5-7)
In equation (5-7), the J(E) data are plotted in a Shottky plot (Ln(Js) vs. E1/2) where linear fit slope
can be used to calculate material permittivity using the form of βS discussed in section 5.2.2. The
11 Histograms of parameter estimates from which confidence intervals are derived are presented in Appendix D along with raw J(E) data plotted
with superimposed nonlinear fits.
113
data from Figure 5-2 are transformed into Schottky plot format and shown in Figure 5-5 along
with their respective linear fit functions.
Figure 5-5 J(E) data from Figure 5-2 transformed into Schottky plot format. Again, measurements were taken at a)
25 oC, b) 100 oC and c) 150 oC. Here slope of linear fit corresponds to βs/kT in equation (5-7).
114
Both PI and PPIDS data sets behave similarly at 25oC, which is expected given previous
observations, as seen in Figure 5-2. At 100oC a change in the sample sets’ behaviors occurs where
the slope of linear fit corresponding to PI becomes steeper compared to that of PPIDS. As
temperature is increased to 150oC, the difference between these slopes still exists however it is less
dramatic. This implies that the value of permittivity calculated using Schottky theory will change
depending on both temperature and sample processing condition. Calculation of permittivity under
Schottky formalism is done via the following equation:
𝜖𝑟,𝑆 = [(𝑚𝑘𝑇)24𝜋𝜖𝑜
𝑞3]
−1
(5 − 8)
Permittivity values calculated for each sample group as a function of measurement
temperature are shown below in Figure 5-6. At low temperatures, Schottky theory produces
permittivity values within the range of 18 to 4 for PI, indicating that this model inadequately
describes conduction through the material. This outcome coincides well with modeling performed
using hopping theory in section 5.2.1 that confirms that low temperature conduction is better
described by a bulk-dominated process. At elevated temperatures, starting at 100oC, the PI sample
set undergoes a transition in behavior and returns a calculated permittivity that lies within the range
considered acceptable for PI. The value of permittivity calculated using Schottky theory remains
within the range of expected permittivities for the material (determined by the high frequency
permittivity given by refractive index squared (n2) and εr measured at 1kHz) up until 150oC
indicating that high temperature conduction is no longer dominated by hopping but rather injection
through the electrode/dielectric interface.
Unlike the PI, PPIDS samples do not display Schottky type behavior at low and moderate
temperatures. In fact, Schottky theory does not return a permittivity within the expected range of
PI until 150oC and is located at the upper limit of acceptable permittivity for polyimide. At 175oC,
the calculated permittivity for the PPIDS films are well within the expected range for PI, displaying
a value similar to untreated PI at 100oC. This demonstrates that high field conduction in plasma
treated polyimide can be described by a Schottky dominated process only after the film is exposed
to 75oC higher temperature than untreated PI. This result is in good agreement with the results
obtained from section 5.3.2.1 that yielded good fits using hopping theory spanning the entire
temperature range for the PPIDS set.
115
Figure 5-6 Permittivity values calculated from linear fits from Schottky plots between 25oC – 175oC. A shaded region
is marked indicating the range between high frequency permittivity (n2) and permittivity measured between 100Hz –
100kHz.
5.4 DISCUSSION AND CONCLUSIONS
Plasma surface treatment using a mixture of oxygen and helium gas demonstrates the
ability to chemically alter the surface environment of PI, changing both wetting and electrical
properties of the film. Results obtained from XPS measurements on plasma treated PI from the
previous study by Meddeb et al. [13] indicate the increase of hydroxyl groups and suggest the
formation of added carbonyl and carboxyl moieties to the surface of the films.
In the study by Meddeb et al. [13], the addition of these chemical moieties is surmised to
result in high field conduction suppression, suggesting that their presence serves to trap or scatter
charges at the interface. In this present work, it is shown that low temperature conduction through
untreated PI and plasma treated PI is dominated by a hopping type mechanism, governed by the
diffusion of charge carriers through the bulk of the film. PI is shown to diverge from this model at
100oC, exhibiting conduction that is well described by Schottky theory, indicating charge injection
at the electrode/dielectric interface. Calculation of the activation energy in the low field regime
indicates the activation energy for DC conduction through an untreated film is 0.25 eV implying
the presence of electronic charge carriers. This result strengthens the outcome of high temperature
conduction modeling, suggesting electronic injection dominates leakage current through the
material under the combination of sufficiently high electric field and temperatures.
116
Plasma treatment increases the hopping-Schottky conduction transition temperature from
100oC to 175oC. This result indicates that the combination of hydroxyl, carbonyl and carboxyl
groups grafted by plasma treatment could be preventing charge injection at the interface. Previous
mention of the conduction at high temperature dominated by electronic injection at the
electrode/dielectric interface coincides well with this theory: electronegative oxygen containing
moieties create electron trapping centers at the interface and limit current injection at high
temperatures. This finding correlates well with past research that demonstrates conduction is
electronically dominated [194] [191], indicating that the high temperature performance of PI can
be improved by simple surface chemical modification using O2 to suppress charge injection of
electronic carriers. In context of Chapter 4, it seems the carrier type dominant in the material
(holes for P(VDF-TrFE and electrons for PI) plays a significant role in determining the outcome
of plasma surface treatments of a given chemistry. Plasma treatment can potentially limit the
formation of electronic space charges in dielectrics in the context of work done by Locatelli et al.
[186], thereby reducing field enhancements at the electrode/dielectric interface however the
dominant charge carrier must be taken into account prior to surface modification.
Future research concerning PI’s applicability for high temperature dielectrics should
emphasize the role of the electrode/dielectric interface and its potential to be used in tailoring
electrical properties of the material. The outcomes from theoretical analysis of data are used to
provide further insight on how surface chemistry may be used to tailor high field leakage current
in PI, and other materials. Suppressing leakage current by plasma surface treatment will potentially
lead to improved capacitor performance by limiting joule heating, suppressing electronic space
charge development under high fields, and extending the operable temperature range of dielectric
materials for power applications.
ACKNOWLEDGEMENTS
The authors of this publication would like to acknowledge the support of the National
Science Foundation as part of the Center for Dielectrics and Piezoelectrics under Grant Nos. IIP-
1361571 and IIP-1361503. We also would like to acknowledge Adam Walder for his expertise in
R-Studio script writing and input on statistical analysis of parameter estimates from non-linear
regression. Adam is a Ph.D. student in the department of statistics as part of the Eberly College of
Science at Penn State.
117
CHAPTER 6
IMPURITY ION AND SPACE CHARGE CONDUCITON AT LOW AND HIGH FIELDS:
ANALYSIS OF IMPURITY ION MIGRATION AND INTERACTION WITH INTERFACES
IN P(VDF-TrFE)
ABSTRACT
Chapter 6 takes concepts from Chapters 3-5 and applies them to the development of a
system in which interfaces in P(VDF-TrFE) can be probed through controlled quantities of
impurity ions. This chapter is broken into two sections: Chapter 6I and 6II. In Chapter 6I an
equivalent circuit model was introduced to describe low frequency polarization mechanisms
dominating ionic conduction through P(VDF-TrFE). An outline for this section is provided below:
Chapter 6-I Outline
Introduction – a short literature review highlighting work which emphasizes the impact of
both injected and ionic space charges on electrical properties, field distribution and material
degradation is presented. The section is completed with statement of Chapter 6-I’s content:
quantification of low frequency ionic charge migration and interaction with P(VDF-TrFE)
polycrystalline morphology.
Materials and Methods – the processes by which LiClO4 doped P(VDF-TrFE) films are
fabricated is explained as well as film structural and electrical characterization methods.
Results and Discussion – raw structural and electrical experimental data is analyzed and
the results discussed prior to the implementation of equivalent circuit modeling techniques.
Equivalent Circuit Modeling – impedance data is fit with an EC descriptive of polarization
mechanisms contributing to the electrical signal. Fit parameter estimation combined with
statistical interpretation using Z-view electrochemical software is used to strengthen the
link between ionic conduction and material structure described by the model.
Conclusions – analysis of data in Chapter 6-I is discussed in context of its impact in
understanding conduction through P(VDF-TrFE) as well as how it will be used in Chapter
6-II.
118
Chapter 6-II takes the information on ionic conduction through P(VDF-TrFE) gained in 6-
I and applies it to further knowledge gained in Chapter 3 by controlling space charge interaction
with planar interfaces created by multilayer spin casting. Thermally stimulated depolarization
current (TSDC) is introduced as an experimental technique to compliment low frequency
impedance spectroscopy by capturing electrical depolarization responses at equivalent frequencies
in the quasi DC range. Although 6-I emphasizes low frequency interaction with interfaces in
layered films, the high electric fields and long poling times characteristic of TSDC necessitate
information obtained from Chapter 4 and 5 to accurately interpret results. An outline of Chapter
6-II is provided below:
Chapter 6-II Outline
Introduction – Concepts pertinent to the understanding of conduction obtained from
chapters 3 to 6-I are briefly reviewed in place of the typical literature review. The section
is concluded with the proposed layered dielectric system outlining the mechanism by which
dopant impurities contribute to the understanding of how interfaces impact electrical
properties.
Materials and Methods – processing implemented to create undoped and doped layered
dielectrics along with relevant experimental parameters and techniques are explained.
Results and Discussion – the results of impedance spectroscopy of layered structures as
well as TSDC are presented and discussed in context of relevant literature.
Conclusions – the results of data analysis are discussed to further analysis of the interface
on charge transport through organic dielectrics.
Chapter 6 signifies the end of work done in this dissertation. This chapter is followed by potential
avenues for future work suggested by the author to continue studying charge transport in organic
dielectrics.
119
CHAPTER 6I
ANALYSIS OF LOW FIELD IMPURITY ION MIGRATION IN LiClO4 DOPED P(VDF-
TrFE) THIN FILMS
6I.1 INTRODUCTION
In recent work published by Yang, et al. [12] the performance of biaxially oriented PVDF
(BOPVDF) under high electric fields in relation to electrode chemistry is analyzed. Thermally
stimulated depolarization current (TSDC) measurements performed at high fields and long poling
times indicate high ionic polarization in BOPVDF. The amount of ionic polarization observed was
higher than expected considering only impurity ion concentrations associated with sample
fabrication. It was postulated based off research done by Eberle, et al. [85] that HF gas emitted at
high electric fields reacts with Al and Ag electrode metals to produce Ag+ and Al3+ cations, as well
as F- anions which contribute to the large ionic depolarization peak in TSDC. The exact nature of
these ionic species in PVDF are not well understood, however their presence induces space charge
accumulations that interact with the materials polycrystalline structure and can create
heterogeneous field concentrations distributed throughout the material.
Although the effect of space charge migration has not been extensively investigated in
PVDF, the high voltage cable insulation community has placed emphasis on this area of research
in polyethylene (PE) and crosslinked PE (XLPE). Past work by Hozumi et al. [86] demonstrates
via pulsed electroacoustic (PEA) measurements that under electric fields exceeding 0.2 MV/cm
heterocharges related to impurities from crosslinking byproducts (acetophenone) and antioxidant
form within the material. At fields beyond 0.7 MV/cm, the formation of packet charges is observed
and thought to be due to local ionization of impurities assisted by acetophenone through solvation.
Similar observations have been made by Ren et al. [195] where crosslinking byproducts create
non-uniformity in the conductivity across the thickness of XLPE cable insulation. Ultimately,
work by Cao et al. [87] demonstrates that these mobile impurity species contributing to charge
heterogeneity across the film lead to dielectric breakdown in XLPE films at temperatures between
50 – 90oC. Impurity charge migration is also shown to adversely affect the performance of other
material systems such as mineral insulating oil. Gabler et al. [88] demonstrates the impact of
heterogeneous charge build-up due to injected charges as well as intrinsic charge carriers at DC
120
frequencies on electric field distribution in mineral oil/paper arrangements. Calculation of field
distributions in the oil/paper arrangements indicate charge separation at the paper/oil interface,
resulting in field drops 7 times larger than the mean field strength through the material. This
ultimately leads to breakdown initiation in the vicinity of charge accumulation.
In this chapter, the effect of ionic charge migration through P(VDF-TrFE) on AC
conductivity at low frequencies and high temperatures is investigated. Films are doped with low
content of LiClO4 to create a model material in which the quantity and species of ionic charge
carrier is well understood and dominates the electric response of the film. This specific ionic
complex is chosen because of past research implementing a combination of transference number
measurements [196] and electrolysis experiments [197] that demonstrate Li+ is the dominant ionic
charge carrier in LiClO4 doped PVDF films. Thus, lightly doping the film enables characterization
of ion migration where the impurity ion is both controlled well defined unlike studies mentioned
in Chapter 1 Table 1-III. and distribution through the bulk of the film without significantly
altering material morphology. Differential scanning calorimetry (DSC) is used to structurally
characterize doped films as a function of LiClO4 wt %. Impedance spectroscopy over a broad
frequency (100 kHz – 0.1 mHz) and temperature (25oC – 110oC) captures the electrical response
of P(VDF-TrFE) due to molecular and space charge polarization mechanisms. Finally, an
equivalent circuit (EC) to model impedance behavior is developed based off the structure of the
material and fit to impedance data using complex nonlinear regression. A combination of results
from DSC, the known structure of P(VDF-TrFE), and statistical outputs from the EC model are
used to enhance the understanding of ionic charge migration and accumulation in P(VDF-TrFE)
thin films.
6I.2 MATERIALS AND METHODS
6I.2.1. Materials
The copolymer P(VDF-TrFE) 70/30 mol% was purchased from Poly K in powder form
with a molecular weight Mn = 205 kg/mol. Battery grade 99.99% metals basis LiClO4 of 106.39
g/mol was purchased from Sigma Aldrich. Both materials were dissolved in N,N-
Dimethylformamids (DMF) purchased from DriSolv® into a solution that was spun onto platinum
coated silicon wafers purchased from Nova Electronic Materials.
121
6I.2.2 Thin film Fabrication
Doped copolymer films were fabricated by solution deposition. The weight of LiClO4 was
measured in comparison to solid weight of P(VDF-TrFE) powder: grams(LiCLO4)/gramsP(VDF-
TrFE). This parameter was tuned to create 6 separate batches containing 0% wt (pure copolymer),
0.1%, 0.25%, 0.50%, 0.75%, and 1.0% LiClO4. The total solid wt % of the solution ((wt P(VDF-TrFE)
+ wt LiClO4) / wt solution) was ~7.5%. A degassing procedure to removed trapped air bubbles from
magnetic stirring was performed prior to spin casting onto platinum coated silicon wafers. A KLA-
Tencor P16+ stylus profilometer was used to determine film thickness and uniformity. It was found
600 rpm for 50s yielded a 1μm film of good thickness uniformity. Silver electrodes (50nm thick
by 3 mm diameter) were deposited using a Lab-18 electron beam evaporator provided by Kurt J.
Lesker. The stage temperature was held at 0oC during electrode deposition to prevent sample
damage.
6I.2.3 Structural Characterization
The thermal properties of the spin casted films were measured using a Q2000 differential
scanning calorimeter by TA Instruments. The samples were prepared by scratching 5–7 mg of
copolymer off the wafer into a T-Zero aluminum DSC pan using a carefully cleaned razor blade.
The temperature ramp rate was 10oC/min and spanned a range of -60 – 180oC. Analysis in this
paper used the first heating cycle in order to directly capture the effect of processing and LiClO4
content on copolymer crystal structure. Each film produced three samples and the data for each
film condition is taken to be the average between the three samples from the originating single
film.
6I.2.4 Electrical Measurements
A Cascade Probe Station equipped with DCM 210 series Precision Positioner 20μm probes
were used to capture low field behavior of the samples. Impedance spectroscopy was performed
within the frequency range of 100kHz – 0.1Hz using a Modulab impedance analyzer from
Solartron Analytical. Temperature was swept from 25 – 110oC using a Temptronic TP03000
thermal chuck vacuum wafer probe chiller interfaced with the probe station during measurement.
Values for the real and imaginary parts of impedance were collected with Modulab and converted
to capacitance and loss tangent using the general electrochemistry software Z-view provided by
122
Scribner Associates Inc. Impedance fitting and equivalent circuit (EC) modeling was also
performed using Z-View where complex nonlinear regression implemented a calc proportional
data weighting scheme.
6I.3 RESULTS AND DISCUSSION
6I.3.1 Differential Scanning Calorimetry (DSC)
Crystallinity is calculated using a 91.45 J/g melting enthalpy for a theoretically 100%
crystalline sample of P(VDF-TrFE) 70/30 mol% [98] via melting peak integration. The results of
DSC measurement and endotherm integration are presented in Table 6I-I below.
Table 6I-I: DSC results as a function of LiClO4 solid wt% for the first heating cycle. Endothermic peak temperatures
along with integration results are shown and the range in standard deviations of the sample sets are given in italics.
Sample Average Tc
(oC) Int Tc (J/g)
Average
Tm (oC)
Int Tm
(Crystal %)
Pure 110.1 ± 0.6 15.9 ± 0.7 151.3 ± 0.2 32.0 ± 0.6
0.10% 109.3 ± 0.7 16.2 ± 0.9 151.3 ± 0.2 32.5 ± 0.8
0.25% 109.9 ± 0.9 20.1 ± 0.7 152.2 ± 0.3 32.6 ± 1.1
0.50% 107.0 ± 0.5 21.6 ± 1.1 152.7 ± 0.4 33.2 ± 0.7
0.75% 107.0 ± 0.9 21.1 ± 2.0 153.2 ± 0.8 33.6 ± 1.6
1.00% 104.6 ± 0.4 22.7 ± 0.9 153.1 ± 0.3 32.4 ± 1.0
The pure copolymer film shows two peaks a temperatures T1 and T2 in the first heating ramp of
DSC: T1 = 110oC = Tc signifying the -phase crystals transitioning from a ferroelectric phase to
the paraelectric phase [198] and T2 = 151oC = Tm corresponding to polymer crystal melting [160].
As LiClO4 is introduced to the film, changes in Tm and Tc are measured. An increase in Tm occurs
between films containing 0.1% - 1.0% LiClO4, however it is only by 1.85oC. Integration of Tm
reveals a slight increase in % crystallinity of the film as salt content is increased. Similar to Tm,
change in the Tc endotherm is observed which is more pronounced: as LiClO4 % is increased, Tc
peak position decreases from 110.1 – 104.6oC and integrated Tc signal increases from 16.2 – 22.7
J/g. Work by Oliveira et al. [199] that studied P(VDF-TrFE) structure via in-situ X-ray diffraction
(XRD) as a function of temperature shows contributions to both α (2θ = 19.9) and β (2θ = 17.8 )
crystal phases at temperatures below Tc. Above Tc only α remains, indicating the Tm transition
pertains exclusively to α crystal melting in the material. Because of LiClO4 addition’s minimal
123
effect on Tm and prominent effect on Tc, it is suggested that salt-ion pair interaction is occurring
predominantly between polar crystal domains associated with the β-phase of the films.
6I.3.2 Dielectric Spectroscopy
Permittivity and loss tangent for each ionic content at room temperature is shown below in
Figures 6I-1a and 1b respectively. The pure copolymer shows a relatively stable permittivity and
loss tangent at room temperature that agrees well with values reported in literature [200]. At high
frequency, a dip in the permittivity can be seen corresponding to the relaxation of the orientational
polarization mechanism associated with permanent C-F dipole rotation [201]. This dip is coupled
by a slight increase in loss tangent and is characteristic of the dipole relaxation mechanism. When
the frequency is between the range of 10Hz – 10kHz, the permittivity and the loss tangent is
Figure 6I-1: a) real part of the permittivity and b) loss tangent measured at 25oC as a function of frequency for ionic
content up to 1.0 wt%.
124
considered stable. At frequencies below 10Hz, an increase in the permittivity and loss tangent is
measured for all samples indicating the onset of space charge migration through the dielectric. An
increase in permittivity and loss tangent within this frequency range occurs as greater quantity of
LiClO4 is introduced into the films suggesting that cation Li+ species serve as the dominant
contributor to space charge in doped films.
Figure 6I-2a and 2b show the permittivity and loss tangent for the same films as seen in
Figure 6I-1, at 100oC. At higher temperatures, dipole mobility increases resulting in larger
permittivity at frequencies related to C-F bond rotation. This increase in dipole mobility also
creates a shift in dipole relaxation to higher temperatures. Temperature’s influence on chain
mobility and free volume also impacts low frequency ionic charge migration by increasing ionic
mobility through the bulk [202, 203]. The pure copolymer film shows an order of magnitude
increase in the permittivity at 0.1Hz at 100oC relative to the 25oC measurement, however change
Figure 6I-2: a) real part of the permittivity and b) loss tangent measured at 100oC as a function of frequency for salt
wt %’s 0 – 1.0%.
125
in LiClO4 containing samples are more prominent. The permittivity in the 0.1% doped film
increases from εr = 14 at 25oC to εr = 8.1x103 at 100oC which is coupled by a fully resolved loss
tangent relaxation centered at 10Hz (Figure 6I-2b). Further addition of LiClO4 increases the low
frequency permittivity, reaching a maximum at 1.2x105 for the 0.75% sample, and shifts the
position of the loss tangent peak within the frequency range of 40Hz – 100Hz. The presence of a
fully resolved peak in the loss tangent as well as plateauing in the permittivity give evidence that
Li+ migration through the film interacts with the electrode dielectric interface at the lowest
frequencies of measurement and highest temperatures. The absence of this behavior in the pure
copolymer film suggests that the saturated permittivity in ion containing films is indicative of a
pseudo double layer capacitance between Li+ space charge and the anode.
6I.3.3 AC Conductivity
The real part of the permittivity and loss tangent are used to calculate the AC conductivity
of the films as a function of temperature and ionic content. Permittivity and loss tangent are related
to the AC conductivity of the material via the following relationship:
𝜎𝐴𝐶 = 휀𝑟휀𝑜𝜔𝑇𝑎𝑛(𝛿). (6𝐼 − 1)
The conductivity values for T = 40oC, 80oC, and 100oC using equation (6I-1) are shown in Figure
6I-3a, 3b, and 3c.
Figure 6I-3a represents low temperature behavior of the samples. At 40oC, LiClO4
addition to the polymer has a negligible effect on the conductivity at the highest frequencies. In
this temperature/frequency domain, the dominating conduction mechanism does not depend on
ionic impurities but rather losses due to dipole polarization dominates. LiClO4 begins to have an
effect on the calculated conductivity at frequencies below 10Hz. There is considerable change in
conductivity from low wt% samples, increasing from 2.0x10-11 S/m in the pure film to 5.3x10-11
S/m in the 0.1% doped film at 0.1Hz. An order of magnitude increase is calculated when increasing
from 0.1% wt (5.3x10-11 S/m) to 0.25% wt (2.4x10-10 S/m) in the film. At higher concentrations,
the conductivity saturates at 0.1Hz, where 0.5%, 0.75%, and 1.0% doped films display
conductivities of 1.1x10-9 S/m, 1.1x10-9 S/m, and 1.0x10-9 S/m respectively.
Figure 6I-3b represents sample behavior at moderate temperatures. A plateau in the
conductivity of doped samples occurs at 80oC in ion rich films as a function of frequency,
indicating conduction through the sample is primarily dominated by ionic conductivity due to Li+
126
Figure 6I-3: AC conductivity calculated using equation 1 calculated at 100 kHz – 0.1 Hz for each of the measured
LiClO4 wt %. Temperatures measured are a) 40oC, b) 80oC, and c) 100oC.
cation migration. Saturation in the calculated conductivity also takes place at all frequencies for
samples containing 0.50% - 1.0% wt LiClO4, suggesting increased carrier concentration due to
added LiClO4 does not produce large changes in the conductivity of the samples within the range
at the measurement temperature.
Saturation in ion containing samples is less prominent at 100oC than it was at lower
temperatures. A decrease in conductivity is initiated between 1Hz and 0.1Hz in samples containing
LiClO4. This relaxation in the AC conductivity is related to the development of another capacitive
effect in the material caused by the build-up of ionic charges at the electrode/dielectric interface
only observed at high temperatures and low frequencies. Drop in conductivity could arise from the
electrodes blocking ionic charge transport at low frequencies. This interpretation of the response
is supported by no observable change in the conductivity for the pure sample which exhibits
gradual plateauing conductivity as a function of frequency in the absence of Li+ charges.
6I.3.4 Impedance Spectroscopy
The impedance of films containing 0 -1.0% LiClO4 was characterized as a function of
temperature using ac impedance spectroscopy. In this analysis, the imaginary impedance Z” and
real impedance Z’ are measured as a function of frequency. Figure 6I-4 shows the -Z” impedance
as a function of Z’ in Cole-Cole plot format. The ideal impedance response of a capacitor can be
approximated by an RC equivalent circuit which traces a semicircular arc in Z’, Z” space. The
extension of the semicircle in the Z’ direction is representation of the bulk resistivity Rb of the
material while extension into -Z” represents capacitive behavior. In Figure 6I-4, the pure and low
ion containing 0.1% sample extend predominantly into the -Z” direction, indicating a capacitive
response with low leakage and high Rb. This is supported by AC conductivity calculations at low
127
Figure 6I-4: Cole-Cole plot of impedance for 0 – 1.0% doped samples at 25oC. The material’s bulk response cannot
be fully resolved however a general trend between LiClO4 addition and conductivity is observed.
temperature in Figure 6I-3a where the pure and 0.1% films display conductivities 1 – 2 orders of
magnitude below their highly doped counterparts. As the doping concentration increases, a greater
portion of the impedance arc is resolved in the Z’ direction indicating the motion of Li+
contributing to the bulk response of the material and lowering the value of Rb.
Higher temperature facilitates the movement of ionic carriers (similar to that seen in the
permittivity and loss tangent results in Figure 6I-2) and yields more informative impedance
diagrams. Figure 6-I-5a, 5b, and 5c show Cole-Cole plots for doped copolymer films at 100oC.
Figure 6I-5a shows the impedance of all samples super imposed on one another. Only the pure
copolymer can be observed at the scale shown in Figure 6I-5a because of its large impedance in
comparison to doped films. Similarly, the pure film only shows a single feature in its impedance
response: a depressed semicircle representing the bulk response of the material under test common
to many polymers and solid polymer electrolyte (SPE) materials [202, 204, 205]. Figure 6I-5b
and 5c show the impedance of the doped copolymer films by progressively decreasing the scales
on the Z’ and -Z” axes. In Figure 6I-5b, the impedance of the 0.1% film shows two features: a
depressed semicircle representing the bulk response of the material and low frequency line that
extends upward in the -Z” direction. This behavior is strong evidence of Li+ blocking at the
electrode/dielectric interface which was interpreted as pseudo double layer capacitance in section
6I.3.2 [205, 206]. The lowest measured impedances are shown in Figure 6I-5c which reveals the
same two features representing bulk and blocking behavior for the 0.25 – 1.0%. This set of films
128
display the strongest separation between bulk behavior and blocking and have bulk impedance
responses with a well-defined Z’ magnitude.
Figure 6I-5: Complex impedance Cole-Cole plots at 100oC for a) all tested samples, b) 0.1% – 1.0% samples and c)
0.25 – 1.0% samples.
129
Bulk resistance Rb of the material is taken as the point in which the bulk semicircle touches
down on the Z’ axis, magnitudes marked by open brackets for 0.25 – 1.0% films in Figure 6I-5c.
The bulk conductivity of the material can be calculated by the following equation:
𝜎𝐵 =𝑡
𝑅𝑏𝐴 (6𝐼 − 2)
where t and A are the thickness and area of the film under test. This calculation is used to compute
the bulk conductivity of ion containing film within the temperature range 60oC – 100oC. Only ion
containing films between 0.25 – 1.0% are used because they were the only to consistently exhibit
full resolution of the bulk response.
Arrhenius type behavior of the conductivity is reported in many ion containing systems
with a host polymer matrix such as poly(methyl-methacrylate) (PMMA), poly(vinyl pyrollidone)
(PVP), poly(vinyl acetate) (PVA) and chitosan (CS) [207, 208, 209, 210]. The Arrhenius law
portrays conduction as a thermally activated process and relates the conductivity to the activation
energy via the following relation:
𝜎𝐵 = 𝜎𝑜𝑒−𝐸𝑎𝑘𝑏𝑇 (6𝐼 − 3)
where σo is a constant related to the conductivity at 0K, kb is Boltzmann’s constant, T is the
temperature in K, and Ea is the activation energy. Figure 6I-6 shows the variation of ln(σB) as a
function of T-1 for LiCLO4 0.25% – 1.0%. Linear regression reveals a good fit over the specified
temperature range indicating Li+ migration through the bulk is governed by hopping type
conduction from adjacent trap sites distributed thought the material [207]. As LiClO4 % is
increased, the conductivity is observed to abruptly increase in value at lower temperatures for
concentrations ≥ 0.5%. Conduction through polymer electrolytes is dependent on polymer
chemical structure, crystallinity, free volume, and qualities of the contributing carrier [202]. In
order to address changes in conductivity the activation energy extracted from Figure 6I-6
(embedded in the Arrhenius plot in table format) must be considered. Calculated activation energy
for 0.25% film is 2.32 eV. Following the abrupt low temperature increase in conductivity starting
with the 0.5% samples is a decrease in the calculated Ea to 1.29, 1.59, and 1.52 eV for the 0.50%,
0.75%, and 1.00% samples respectively. In general, high values of Ea is a characteristic feature of
ionic conducting solid polymer electrolytes [202, 211] that are not exhibited by their aqueous
electrolyte counterparts such as tetrabutylammonium triflate (TbaTf) dispersed in propanol. The
aqueous TbaTf system displays
130
Figure 6I-6: Arrhenius plot of σb for LiClO4 range of 0.25% – 1.0%. Extracted activation energies from linear fits are
shown in the embedded table.
activation energies in the range of 0.34 – 0.42 eV, which depends on TbaTf concentration
[212]. Values calculated in the P(VDF-TrFE) – LiClO4 system are slightly higher than those
reported for other non-aqueous SPE systems such as in CS: Lithium Triflate (LiTf) [211], however
the level of doping studied in this manuscript is an order of magnitude lower than much of the
literature surrounding SPE ionic conduction. Similarly, P(VDF-TrFE) is unique in the fact it has
polar crystal domains which give the material its piezoelectric nature. The 0.5% LiClO4
concentration serves as a critical point in the conductivity in which the following occurs: 1) a
decrease in the activation energy associated with conduction and 2) a leap upward in low
temperature conductivity and will be discussed in context of EC modeling in the next section.
6I.4 EQUIVALENT CIRCUIT MODELING
The impedance response of P(VDF-TrFE) 1 μm films pure through 1.0% LiClO4 doped
into the polymer matrix was analyzed using a modified Debye like equivalent circuit (EC) shown
in Figure 6I-7a. The EC is broken into 4 main parts: 1) a capacitor element labeled C1 which
accounts for electronic polarization at high frequencies. The value of this element is proportional
to the refractive index squared (n2) of the material and remains constant for each fit at all
temperatures. 2) A constant phase element CPE2 in series with resistor R2. This leg of the circuit
represents the dielectric response attributed to permanent dipole polarization induced by C-F bond
rotation. The element R2 dictates the frequency at which dipole contributions to the impedance
131
Figure 6I-7: a) physical model of doped P(VDF-TrFE) with equivalent circuit (EC) model used in fitting impedance
spectra, b) raw impedance data (open symbols) with EC fit (solid lines) at 25oC and c) 100oC.
begin to relax out. 3) A resistor R3 in series with a nested R4/CPE3 element accounts for
predominantly low frequency ionic motion through the bulk of the dielectric. Considering P(VDF-
TrFE) is a semi crystalline material, it is proposed that R3 directly relates to bulk charge carrier
transport through the amorphous region of the material while the nested R4/CPE3 corresponds to
charge carrier transport through the amorphous/crystalline interphase region. A similar model is
proposed by Marzantowicz et al. where a nested CPE/R element is implemented represent the
influence of the crystalline phase on ionic conductivity in poly(ethylene oxide) (PEO) [115]. 4)
The
final element is a stand-alone CPE4 element in series with legs 1 through 3 that represents
low frequency polarization caused by impurity ion build up at the electrode/dielectric interface in
the quasi DC frequency regime. It should be noted that components 1 through 3 describe the bulk
response of the dielectric while component 4 arises in the special case films are doped with LiClO4
and measured at sufficiently high temperatures.12 The quality of fit for each sample at each
temperature was determined by a combination of three ways: 1) how well the fit result matches
raw (Z’, Z”) (M’, M”) (C’, C”) and (|Z|, θ) formalisms, 2) how well parameter estimates reflect
physical properties of the material system and 3) statistically computed parameter estimate
errors.13 Impedance fits using the EC in Figure 6I-7a are shown at 25oC and 100oC in (Z’, Z”)
format as solid lines over raw data.
12 Mathematical interpretation of the EC model used in nonlinear regression is found in Chapter 2 section 2.4.2 13 Parameter estimate error% are compiled for each sample and measurement condition presented in Appendix E.
132
The impedance response of pure and on containing P(VDF-TrFE) films were modeled
using Z-view electrochemical impedance fitting software. In this regard, each circuit element is
regarded as a parameter that is optimized via complex nonlinear regression using a calc-
proportional weighting scheme. All plots concerning individual circuit element contain data points
that are direct outputs of complex nonlinear regression parameter estimation. Statistical
interpretation of the fit results incorporates percent error that is generated by Z-views fitting
statistical report and indicate parameter estimate significance. Each data point shown in fit
parameter estimate figures was taken as the average between fit results extracted from the
responses of three individual samples measured at each temperature and LiClO4 content. The
standard deviations of the fit parameter estimates are also plotted, however in most instances they
are relatively small and covered by the data points.
6I.4.1 Modeling the Capacitive Response
Capacitive elements of the EC are first analyzed to inform how Li cations interact with the
material’s bulk structure. In this section, each capacitive circuit element is discussed in relation to
the polarization or conduction mechanism it describes: electronic polarization, permanent dipole
orientational polarization, ionic / space charge conduction, and blocking polarization.
6I.4.1.1 Electronic Polarization
The ideal capacitor C1 is chosen to represent the dielectric response of the material due to
electron cloud displacement under the applied electric field and has an impedance described by
equation (61-4):
𝑍𝐶1=
1
𝑗𝜔𝐶1 (61 − 4)
Electronic polarization dominates at frequencies on the order of THz, well beyond the
measurement range used to study impedance of P(VDF-TrFE). Thus, the value of C1 is assumed
to remain constant as a function of frequency and temperature for each sample fit. Using the high
frequency approximation εr = n2 and sample electrode area and thickness, C1 was approximated
using the equation C = εrεoA/t and held constant at 1.25x10-11 F.
6I.4.1.2 Permanent Dipole Orientational Polarization
133
Figure 6I-8: Estimated parameter values for a) Q2 from CPE2 as a function of salt wt % and b) approximated material
permittivity. Each film tested is of equivalent thickness. Units for Q are written in terms of conductivity as S sn given
equation 61-5.
The dipolar contributions to the overall bulk impedance response are described by CPE2 in
series with R2. Unlike the ideal capacitor, CPE2 has a more complex definition for impedance
described by equation (6I-5):
𝑍𝐶𝑃𝐸2=
1
(𝑗𝜔)𝑛2𝑄2 (61 − 5)
where the value Q2 is the estimated value of CPE2 acquired by fitting and n2 is the imperfection
factor. In this regard, Q2 must be taken into consideration along with n2 during analysis. Since the
impedance response due to dipole rotation dominates the measured signal within the frequency
range kHz – MHz, CPE2 is defined as an optimizable parameter during fitting.
The values of CPE2 are plotted as a function of LiClO4 wt % in Figure 6I-8a from 25 oC
– 110 oC. The parameter n2 represents a distribution of relaxation times due to heterogeneity of the
molecular environment surrounding permanent dipoles in the material. For all fits, n2 is held
between 0.98 – 0.99, chosen by statistical interpretation of the goodness of fit for the sample being
measured. The value of the circuit element increases as a function of temperature which can be
explained by an increase in dipole mobility due to temperature increase. As LiClO4 content is
increased, there is no significant change in the value of Q2 for any given temperature. Considering
each sample exhibits the same geometry for all salt concentrations, it can be concluded that ionic
inclusions do not have an impact on CPE2 parameter estimation. Because CPE2 describes
polarization associated with dipole orientation, the value of Q2 should correlate to the material
permittivity from Figure 6I-1 and 6I-2 for frequencies that space charge effects are not observed.
134
Figure 6I-9: The behavior of CPE3 as a function of LiClO4 wt % for the temperatures 25 oC – 100 oC. the standard
deviation of parameter estimates in the sample set are reflected by error bars. Fitting between 100oC – 110oC produces
large standard deviations in n3 as well as erratic parameter estimates for n3 and Q3.
Due to the value of n2 being 0.98 – 0.99, the response of CPE2 can considered as that of an ideal
capacitor using equation (6-I-2). Thus, Q2 ≈ C2 and εr ≈ Q2t/εrεo. This approximation uses
parameter estimates of Q2 from Figure 6I-8a to approximate the material permittivity in Figure
6I-8b. The results of permittivity approximation agree with the permittivity calculated from raw
capacitance data in Figures 6I-2 and 6I-3. The orientational polarization is dominant at ~10kHz,
where the permittivity is independent of temperature for all salt ionic contents.
6I.4.1.3 Ionic / Space Charge Conduction
Ionic space charge migration dominates at low frequencies and is described by R3 in series
with the nested CPE3/R4 circuit element. In this portion of the EC, R4 represents conduction of
ionic charge through the amorphous regions of the polymer and CPE3/R4 represents charge
migration through the crystalline/amorphous interphase. Like CPE2, the impedance of CPE3
depends on Q3 and n3. In this scenario, both parameters were estimated by fitting.
The response of CPE3 fit parameters to temperature and LiClO4 solid wt% is plotted in
Figure 6I-9. Parameter estimation of Q3 indicates that salt concentration has an impact on the
behavior of CPE3. As the amount of LiClO4 is increased, Q3 increases in value. At 25 oC, the model
yields a value of 2x10-10 S sn for the pure sample. This increases by two orders of magnitude to
1.21x10-8 S sn in the film containing 1.00% salt. This general trend is seen at each temperature and
can be related to increased interaction between charges and the crystalline regions of the material
as a greater quantity of ionic charge carriers are introduced. P(VDF-TrFE) crystallizes primarily
135
in the polar β-phase, thus interaction of LiClO4 molecules with the crystalline/amorphous
interphase region is anticipated to occur due to internal charge compensation.
The value of n3 is also plotted as a function of LiClO4 % in Figure 6I-9b. In general, the
value of n3 is centered around 0.5. With an n3 of 0.5, the CPE phase angle is constant at 45o and
its magnitude of impedance is proportional to the inverse of the square root of frequency (ω-1/2).
In this scenario, the CPE behaves similar to a Warburg impedance element, which typically models
diffusion-based processes that are characteristic of charge transfer resistance and double layer
capacitance. This outcome of the fit supports that the CPE3 element is descriptive of interaction
between crystalline regions and Li+ cations associated with doping, also described in PEO by
Marzantowicz et al. [115].
It should be noted that the value of n3 becomes erratic as measurement temperature
approaches 110 oC. Table 6I-I from section 6I-3.1 indicates that the Tc for the material ranges
between 104oC – 110oC depending on LiClO4 wt%. At temperatures surrounding the phase
transition temperature of the β-phase, parameter estimation begins returning uncharacteristically
high values of Q3 as well as erratic values of n3 without definable physical meaning. Erratic
parameter estimation of Q3 and n3 is also coupled by large parameter estimate error % (ranging
from 15.6% – 54.8% for Q3 and 8% - 62% for n3 depending on salt concentration)14. Large error
% is indicative of parameter estimate insignificance and inaccuracy when fitting, indicating that
the nested CPE3/R4 circuit element’s contribution to the model breaks around temperatures
associated with the ferroelectric paraelectric phase transition of the crystal. Large standard
deviations within the sample sets of estimated n3 values also occur at high temperature. This
outcome suggests that the nested CPE3/R4 circuit element is unique to polar crystalline sites
through the film and that Li+ transport is significantly impactd by interaction of Li+ with
crystalline/amorphous interphase regions associated with ferroelectric crystals in the material.
6I.4.1.4 Blocking Polarization
The final capacitive circuit element is CPE4 and describes the blocking of ionic space
charges at the lowest, quasi DC frequencies and highest temperatures of measurement. Fit results
for CPE4 are plotted in similar fashion to the others with results of Q4 and n4 parameter estimates
in Figure 6I-10. From T = 25 – 60 oC, the value and n4 for CPE4 are held at n4 = 1. This results in
14 Error % mentioned is presented in Appendix E.
136
Figure 6I-10: Fit results for CPE4 and n4 as a function of salt wt % at temperatures 25 oC – 110 oC. Q4 reflects ionic
charge interaction associated with Li+ and the electrode where 0.5 < n4 < 1.0 estimates indicate electrode/polymer
interfacial roughness and heterogeneous charge distribution at the interface.
impedance of CPE4 to equal that of an ideal capacitor. Similarly, Q4=1F within 25 – 60oC except
for the 0.5% sample at 60oC. This high capacitance indicates the contributing impedance of CPE4
is very low (by equation 6I-4) and does not interfere with the impedance response of the circuit
during fitting within this temperature range. For temperatures beyond 60oC, a value for Q4 is
estimated via fitting. Figure 6I-10a shows these parameter estimates as a function of LiClO4 wt%
and temperature. The pure copolymer samples do not deliver fit results for Q4 until 110oC, however
all LiClO4 containing films return Q4 parameter estimates reflecting charge blocking. The
values of Q4 are within the range of 2x10-7 S sn estimated for 0.1% containing film and gradually
rise to 8x10--6 S sn for the 1.0% containing film. Each sample displays a gradual rise in Q4 as a
function of LiClO4 wt% present in the film which demonstrates the blocking layer capacitance is
proportional to the amount of separated charges at the interface. This follows the traditional
definition of capacitance C = qV where C is capacitance, q is stored charge and V applied voltage.
Values of n4 were also subjected to fitting at temperatures above 60o for all LiClO4
containing samples and are plotted in Figure 6I-10b. The estimated values of n4 range between
0.95 for the 0.25% samples measured at 80oC to 0.72 for the 1.0% samples measured at 100oC.
This n4 range is neither indicative of ideal capacitor behavior, nor charge diffusion. It has been
demonstrated that the impedance response of 2 dimensional systems that show deviation away
from 2D geometry can be modeled using constant phase elements with n = 1/(D-1) where D is the
fractal dimension. In the case that charge accumulation is homogeneous and surface roughness
negligible, D = 2 making n = 1. In the case heterogeneity in charge accumulation occurs and
137
surface roughness is non-negligible, 2 < D < 3, causing a range in n between 0.5 and 1 [213].
Fitting of n4 suggests fractal dimensions indicative of heterogeneity and roughness, which reflects
microscopic irregularities suspected to be present at the electrode/dielectric interface.
In order to verify CPE4’s validity in capturing the effect of space charge blocking at low
frequencies, 0.25% samples fabricated 10 μm thick were measured and compared. Increasing the
thickness of the sample had little effect on the behavior of bulk impedance data and related
capacitive circuit elements. At low frequencies however, blocking polarization in the 10 μm film
is comparatively less pronounced than its 1 μm counterpart. This subtle difference is captured by
calculations of error % in parameter estimates for Q4 and n4 which are presented in Table 6I-II.
Table 6I-II: Error % for CPE4 extracted from Q4 and n4 parameter estimates in 1 μm and 10 μm samples.
Q4 Error % n4 Error %
Temperature
(oC) 1 μm 10 μm 1 μm 10 μm
25 --- --- --- ---
40 --- --- --- ---
60 --- --- --- ---
80 4.1 17.8 4.8 19.1
90 3.1 11.7 3.1 13.1
100 5.1 13.7 2.9 9.6
110 2.3 2.3 1.1 1.9
In the case of CPE4 fitting for the 1 μm sample, strong low frequency polarization response seen
in the impedance data is coupled by Q4 and n4 parameter estimates with low error %’s. On the
contrary, error % is an order of magnitude higher in the 10 μm between 80 – 100 oC, indicating
parameter estimates of Q4 and n4 in films with a comparatively weaker low frequency polarization
response are less significant, verifying that CPE4 captures space charge blocking in doped
specimens.
6I.4.2 Modeling the Resistive Response
In this section, parameter estimates of the bulk EC’s purely resistive components R3 and
R4 are analyzed as a function of temperature for salt wt % ranging from the pure material to 1.00%
138
Figure 6I-11: Fit results for resistive EC elements a) R3 associated with amorphous regions of the bulk and b) R4
associate with the crystalline/amorphous interface plotted as a function of LiClO4 solid wt %. R4 is shown to dominate
resistive response at low temperature and LiClO4 wt%.
doped film. R3 and R4 are examined individually as a function of temperature and compared.
During curve fitting of impedance data, both R3 and R4 were set as optimizable parameters at each
temperature and salt %. The values of R3 and R4 obtained from parameter estimation are plotted as
a function of LiClO4 content in Figure 6I-11a and 11b. The R3 circuit element is a standalone
resistor in series with nested CPE3/R4 and is plotted in Figure 6I-11a. Without an associated
parallel capacitor element, R3 is taken to represent energy dissipation due to Li+ cation migration
through the amorphous regions of the material. Unlike R3, R4 exists in parallel with CPE3 and thus
represents resistance associated with the crystalline/amorphous interphase. Both circuit elements
exhibit a reduction in resistance with increasing temperature. Fitting estimates correlate well with
increases in conduction calculated in Figure 6I-4 at mid – low frequency.
Some differences are observed when comparing the values of parameter estimates obtained
by fitting for each resistive component. At 40oC, R4 exceeds R3 by an order of magnitude at all
LiClO4 contents, indicating that the crystalline/amorphous interphase dominates the bulk
resistance of the material especially at low temperature and low doping %.15 At high doping %’s
and low temperatures (40oC), a plateau in both resistive parameter estimates is initiated at 0.5%.
Starting at 60oC, R3 estimates begin to approach estimates of R4, finally showing virtually no
distinction in estimated values of the two parameters at T > 90oC and LiClO4 concentrations
between 0.50% - 1.00%. Within this range, R4 is not considered the dominating parameter,
15 The bulk resistance for the EC model used is equal to the summation R3 + R4 (can be reasoned mathematically by taking the bulk impedance
derived in Chapter 2 section 2.4.2.1 in the limit frequency approaches 0). At low temperatures R4 dominates thus it is considered the
crystalline/amorphous interphase limits conduction.
139
implying that both R3 and R4 contribute equally to bulk conduction and the crystalline/amorphous
interphase no longer limits ionic carrier migration through the bulk after eradication of the β-phase.
Further evidence of this can be concluded by relating fit parameter estimation to the material’s
LiClO4 dependent thermal response discussed in section 6I.3.1. As temperature is increased in the
range of Tc (100 – 110oC) the crystal structure of the material undergoes the ferroelectric to
paraelectric phase transition which disrupts the nature of the crystalline/amorphous interphase
regions within the material. Evidence of this is reflected by fit parameter statistical analysis:
parameter estimates for R4 exhibiting erroneously large error %’s spanning multiple orders of
magnitude in all samples16. The behavior of R3 and R4 coupled with large error % in the model
specifically for CPE3/R4 elements suggests ionic migration is no longer limited by the crystal
phases of the material at high temperature.
6I.5 CONCLUSIONS
Lithium doped P(VDF-TrFE) were successfully fabricated. The impurity Li+ ion migration
through amorphous and amorphous/crystalline interphases played a key role in the electrical
properties. DSC indicated that addition of ions did not have a significant impact on the %
crystallinity but significantly increased -phase integrated signal at Tc. This is surmised to be due
to ions interaction with the polar β-phase by ionic complexation within the crystalline/amorphous
interphase to maintain charge neutrality within the sample.
Dielectric spectroscopy at room temperature indicates an increase in the low frequency
permittivity and loss tangent suggesting added dissipation through space charge polarization in
doped samples. High temperature dielectric spectroscopy measurements of the permittivity
showed a plateau in the low frequency response as well as the formation of a fully resolved peak
in tan(δ). This behavior is evidence of the formation of a pseudo double layer capacitance at the
electrode/dielectric interface in doped films at low frequency, and later verified by high
temperature impedance spectroscopy where linear extension of data along the -Z” axis at low
frequencies clearly indicates blocking.
Increasing ionic content resulted in a reduction in bulk resistance of the films within the 25
– 100oC temperature range. Bulk resistance extracted from (Z’,-Z”) plots showed Arrhenius type
behavior that was governed by Li+ hopping type conduction through the film. The activation
16 Statistical error % reports for all fit parameters can be found in Appendix E.
140
energy abruptly decreased for a LiClO4 content of 0.5%, indicating a transition point in conduction
behavior where that is attributed to combination of polymer morphological change and salt
percolation in the amorphous region of the film.
The EC used to fit impedance spectra captures conduction processes related to the four
polarization mechanisms in the system: electronic, orientational, ionic space charge migration, and
blocking polarization mechanisms. Fit outputs associated with the bulk of the EC highlight
permittivity’s independence from salt concentration at mid – high frequency, indicating that added
Li+ behave as impurity ions and do not interrupt natural polarization processes associated with the
virgin material. The low frequency nested circuit element describing ionic charge migration
reflects space charge diffusion through the crystalline amorphous interfaces of the bulk. Similarly,
at high temperatures the nested CPE3/R4 parameter estimate accuracy and physical meaning begin
to degrade. When compared with DSC results, this degradation between 100 – 110oC of parameters
is tied to the ferroelectric – paraelectric phase transition, reflecting space charge’s dependence on
polar domain interactions associated with the β-phase. At low frequencies, space charge
polarization at the electrode/dielectric interface observed in impedance Cole-Cole plots were fit
using a stand-alone CPE4 element that suggest the formation of an ionic space charge layer. Values
of n4 indicate the non-ideal behavior of the circuit element is attributed to surface roughness effects
at the interface. Ultimately space charge accumulation at DC frequencies at the electrode dielectric
interface creates large field drops at the electrode/dielectric boundary, potentially serving as a point
for breakdown initiation.
In the following section, an outline demonstrating how impedance spectroscopy will be
used in conjunction with TSDC measurements to inform on impurity ion migration and interaction
in layered P(VDF-TrFE) dielectrics will be presented along with preliminary data on copolymer /
polyvinyl alcohol (PVA) multilayered films.
141
CHAPTER 6II
CONDUCTION IN MULTILAYERED LAMINATES: EXPLOITING THE INTERFACE AS
A BARRIER TO CHARGE TRANSPORT
6II.1 INTRODUCTION
In Chapter 6I an equivalent circuit model was introduced to describe low frequency
polarization mechanisms dominating ionic conduction through P(VDF-TrFE). It was found that
the electrode dielectric interface impacts ionic charge transport at low frequencies and high
temperatures (in addition to influencing electronic injection discussed in Chapters 4 and 5). The
crystalline/amorphous interphase regions associated with the material’s β-phase also impacts ionic
charge carrier propagation through the amorphous regions of the material in single layer films. In
this regard, interfaces associated with electrode contact (Chapters 4 and 5) as well as those within
the dielectric (Chapter’s 3 and 6I) impact dielectric performance under DC conditions and must
be understood to improve polymer capacitor performance.
Chapter 6II incorporates results from all subsequent chapters to create a multilayered
composite system in which low frequency charge interaction with bulk distributed interfaces can
be controlled. In this section, a layered composite is fabricated by depositing P(VDF-TrFE) doped
with 0.25% LiClO4 via spin casting, and then adding a thin layer of polyvinyl alcohol (PVA),
which serves as a barrier layer during P(VDF-TrFE) solution re-deposition. This creates a series
of planar interfaces within the material that are in series to the electrode/dielectric boundary. A
schematic of the multilayered test material is displayed in Figure 6II-1 demonstrating PVA
capping layers (interfaces) intended to limit ionic migration at quasi DC frequencies.
Figure 6II-1: schematic of multilayer material system depicting material components and the anticipated results.
142
It should be noted, that the effects of multilayer lamination have been demonstrated in past
work by Mackey et al. [14], where microlayer coextrusion was implemented to create P(VDF-
HFP) / PC layered structures. In this work, the benefits in terms of dielectric performance were
exemplified by large enhancements of breakdown strength (Chapter 1, Figure 1-10) than either
constituent material alone. Similar results were then obtained by related work performed by Zhou
et al. [15] which suggested layered coextruded dielectrics can limit ionic propagation by reducing
the mobility of impurity carriers within interphase regions between layers. Regardless of both
studies, the quantity and species of impurity ions are not well understood or controlled.
The following sections in this chapter present the following: 1) processing techniques for
fabrication of composite multilayer dielectrics thin films. 2) equivalent circuit (EC) modeling of
impedance for layered films. This portion focuses on the development of EC that accounts for the
added effect of PVA charge interacting interfaces distributed throughout the bulk of the film and
will offer greater insight into understanding the impact of layers on low frequency conduction
characteristics. 3) Thermally stimulated depolarization current measurements (TSDC) will be
implemented for both single and multi-layered films. Due to the technique’s high poling fields and
long poling time at DC frequencies (details in Chapter 2 section 2.3.3.3), TSDC offers
information on charge migration due to a) electronic charges associated with charge injection from
electrode into P(VDF-TrFE) and b) ionic charge migration characteristics observed at lower
frequencies than traditional impedance techniques. In this sense, TSDC can be seen as a technique
to characterize space charge interaction with interfaces from multiple contributing charge carrier
types, serving as a framework to apply the technique to similar material systems in which
dominating charge carrier species contributing to conduction is not well known. Finally, the
chapter is concluded with 4) high voltage dielectric breakdown experiments on 1-layer and 4-layer
films to show the impact of interfaces on dielectric breakdown strength and consequentially energy
density.
6II.2 MATERIALS AND METHODS
6II.2.1 Materials
The copolymer P(VDF-TrFE) 70/30 mol% was purchased from Poly K in powder form
with a molecular weight Mn = 205 kg/mol. 87 – 89% hydrolyzed high molecular weight Polyvinyl
alcohol (PVA) was purchased from Alpha Aesar. Battery grade 99.99% metals basis LiClO4 of
143
106.39 g/mol was purchased from sigma Aldrich. P(VDF-TrFE) and LiClO4 were dissolved in
electronic grade 2-Butanone (MEK) purchased from Alfa Aesar while PVA was dissolved into
deionized H2O. Both solutions were spun onto silicon wafers purchased from Nova Electronic
Materials.
6II.2.2 Multilayer Processing
The processing procedure for multilayered dielectrics was an additive spin casting protocol
involving alternating depositions of P(VDF-TrFE) and PVA. Ultimately the ionic content within
the P(VDF-TrFE) is meant to simulate the presence of impurity ions within the material. Due to
their low quantities present in the pure dielectric, it is important to choose a concentration of
LiClO4 that enhances qualities of ionic conduction through the material and does not disturb the
microstructure of the bulk. The 0.25% LiClO4 concentration was chosen over the 0.1% because it
was the lowest amount of added salt that clearly displays space charge polarization during
dielectric spectroscopy (Figure 6I-1 and 6I-2) as well as blocking polarization at high
temperatures in impedance and EC modeling (Figure 6I-5 and 6I-10) relative to the 0.1% sample
set. Also, DSC results from Table 6I-1 indicates that LiClO4 concentrations greater than 0.25%
begin to reduce Tc of the polymer, implying disturbance in its crystal phase. For these reasons,
0.25% was seen as the optimal LiClO4 concentration to use in the study.
Fabrication of pure and 0.25% LiClO4 doped P(VDF-TrFE) was the same as that described
in Chapter 6I section 6I.2.2 with DMF replaced by MEK. PVA was dissolved into deionized H2O
by magnetic stirring in a bath of silicon oil on a hot plate. The oil bath was heated to a temperature
of approximately 85-90oC (measured using a thermocouple) to fully dissolve PVA. Total mixing
time for a 5% solid wt. PVA/H2O solution was 4 hrs. The completely mixed solution was then
Figure 6II-2: SEM image of 5-layer sample depicting PVA interfaces and P(VDF-TrFE) copolymer layers.
144
rested for time spans up to 1 week prior to use. This procedure was determined to be most effective
in the elimination of bubbles from the PVA/H2O solution and yielded best results when spin
casting. Multilayered films of P(VDF-TrFE)/PVA were fabricated by an additive spin cast
procedure. First an 8% solid wt P(VDF-TrFE) solution was deposited onto a silicon wafer at
600rpm for 15s and dried at 100oC plat temperature for 15 min. Next, PVA solution was deposited
using a wafer spin speed of 2000 rpm for 20s and then dried under the same conditions. The process
was repeated a total of 7 times to produce a sample set consisting of 3 layers of PVA 0.56μm thick
and 4-layers of copolymer 2.40μm thick. Samples were freeze fractured in liquid nitrogen so their
cross section may be imaged by SEM. Figure 6II-2 shows the fractured cross section depicting
well defined P(FDF-TrFE) and PVA regions of uniform thickness where interfaces created via
additive spin casting are better defined than those through hot-pressing PVDF in Chapter3,
Figure 3-3.
6II.3 RESULTS AND DISCUSSION
The main purpose of this section is to elucidate the effect of the dielectric interfaces in low
frequency conduction at low and high fields. The experimental results of a layered film are
compared to a 1-Layer control film of equivalent thickness All films are maintained at ~10 micron
thick regardless of layer count and changes in data are compared with maintaining constant device
volume.
6II.3.1 Differential Scanning Calorimetry
Pure P(VDF-TrFE) and 0.25% LiClO4 doped P(VDF-TrFE) samples were measured using
DSC to compare thermal properties to MEK cast films with DMF cast films in Chapter 6I section
6I.3.1. The same measurement parameters as section 6I.3.1 were used. The results of Tc / Tm
location and integration for cast with MEK are presented below in Table 6II-I.
Table 6II-I: DSC results for the first heating of 10 micron P(VDF-TrFE) without (pure) and with 0.25% LiCLO4
included. The solvent used was MEK, dried for 15 min at 100oC and annealed for 24 hrs at 142oC under vacuum.
Sample Average Tc
(oC) Int Tc (J/g)
Average Tm
(oC)
Int Tm
(Crystal %)
Pure 101.6 ± 0.3 23.4 ± 1.4 151.5 ± 0.3 32.8 ± 1.7
0.25% 105.5 ± 0.2 24.6 ± 0.4 152.1 ± 0.2 34.5 ± 0.6
145
Both the pure and 0.25% doped films display similar behavior to those cast from DMF; however,
a slightly higher % crystallinity is calculated for the 0.25% film in comparison to those discussed
in Chapter 6I section 6I.3.1. A more noticeable difference is seen in behavior of Tc for the pure
sample. Table 6II-I indicates the Tc of the pure film decreased by 9oC when cast using MEK in
comparison to films cast with DMF. Similarly the integrated signal associated with the beta phase
is stronger (23.4 J/g in comparison to 15.9 J/g). Addition of LiClO4 causes an increase in Tc to
105.5oC however this is below that measured for films cast with DMF (109oC). Considering the
effect of Tc on material impedance behavior and EC model parameter estimation (Chapter 6I)
thermal characteristics of P(VDF-TrFE) cast using MEK will be considered for the remainder of
this section.
6II.3.2 Dielectric and Impedance Spectroscopy
The frequency dependent capacitance and loss of pure 10 μm films were measured in the
range of 10-1 – 105 Hz between 25oC – 110oC in similar fashion to films discussed in Chapter 6I
section 6I.3.2. The permittivity and loss tangent are calculated from real and imaginary
capacitance data and shown in Figure 6II-3. Permittivity of samples cast from MEK are within
the range of those reported for copolymer cast from DMF in section 6I.3.2. In films cast from
MEK, a fully resolved loss tangent peak is observed with 1 – 10 Hz at T = 100oC ≈ Tc of the film
(correlating to Tc in Table 6II-I). This is a similar feature observed in loss tangent data for DMF
cast films within the same frequency range at T = 110oC ≈ Tc. In both materials, features of the
dielectric response depend on transition temperature of the crystalline phase from its ferroelectric
to paraelectric form, signifying the β-phases importance on conduction regardless of solvent
choice.
Impedance spectroscopy is used to analyze the effect of PVA barrier interfaces on ionic
conduction in layered films containing 0.25% LiClO4 in P(VDF-TrFE) layers. Measurements as a
function of temperature are performed on a 0.25% LiClO4 doped 1-layer film and compared to a
4-layer sample set of equivalent total thickness presented in Figure 6II-4a, 4b, and 4c. Addition
of Li+ ionic carriers into the material results in 2 orders of magnitude increase in the low frequency
polarization response as seen in the permittvity between 0.1-1Hz, as well as increase in tan(δ)
relaxation peak magnitude and frequency. This outcome is expected given results reported in
Chapter 6I.3.2 and implies conduction at low frequency is dominated by impurity ion transport
146
Figure 6II-3: permittivity and loss tangent for pure 10μm copolymer cast from MEK. A strong relaxation peak in the
loss tangent is observed between 1-30 Hz in the vicinity of Tc.
Figure 6II-4: a) permittivity and loss tangent for a 0.25% doped 1-layer, b) permittivity and los tangent for 0.25% 4-
layer. Integration of interfaces reduces low frequency polarization as well as lowers relaxation frequency and tan(δ)
peak magnitude. c) M’’ relaxation of doped P(VDF-TrFE) compared with pure PVA.
147
through the material’s bulk. The addition of PVA interfaces into the dielectric impacts the
dielectric response in the following ways:
1) The Composite Effect: PVA contributes to composite polarization at high temperatures and high
frequencies. This effect can be clearly portrayed by observing the imaginary portion of the
complex modulus (M’, M’’) response of a 1-layer 0.25% doped film superimposed on the 1-layer
pure PVA film shown in Figure 6II-4c.
2) The Blocking Effect: at low frequency between 0.1-1Hz, the dominating contributing conduction
mechanism is due to ionic migration through the film. In the 1-layer sample (Figure 6II-4a),
εr=2.5x105 at 110oC and 0.1Hz, indicating polarization due to Li+ accumulation at the electrode
dielectric interface. Layered structures (Figure 6II-4b) exhibit a low frequency εr=4.7x103, which
is two orders of magnitude reduced relative to the 1-layer control sample. A similar phenomenon
is observed in the tan(δ) of layered films where both relaxation peak frequency and magnitude is
lowered. The lower relaxation frequency corresponds to longer ionic relaxation time that suggests
Li+ impurity ion mobility is reduced by the addition of PVA barriers. Similar phenomena are
reported in Zhou et al. [15] that demonstrates lowered tan(δ) relaxation frequency in layered
P(VDF-TrFE)/PMMA/PC fabricated using a multilayer extrusion technique.
The AC conductivity is related to εr and tan(δ) by equation 6I-1 and is used to calculate
the conductivity of 1-layer and 4-layer doped films at 100oC in Figure 6I-5. The conductivity for
the 1-layer sample is in the same range as the 0.25% doped 1-layer film measured in Figure 6I-3.
Figure 6II-5: AC conductivity at 100oC calculated using equation 6I-1 for a doped 1-layer, PVA film, and doped 4-
layer. Conductivity is reduced by two orders of magnitude at low frequency in the layered film relative to the 1-layer
control.
148
The conductivity for the 1-layer pure PVA sample lies beneath that of doped P(VDF-TrFE)
following the same general trend in reduction as a function of frequency. The composite material
behaves very differently than either constituent material alone. There is no difference between 1-
and 4-layered films at high frequency between 104-105Hz because the 4-layer sample is
predominantly P(VDF-TrFE) based (~82% by volume) and loss through the film is dominated by
dipole rotation polarization. At low frequencies where Li+ migration dominates conduction, the
conductivity in layered films is significantly reduced. It is concluded that the interface plays a
significant role in limiting conduction due to impurity ion migration by reducing ionic mobility
and blocking at internal P(VDF-TrFE)/PVA interfaces.
6II.3.3 Equivalent Circuit Modeling
Data analysis from the previous section suggests the presence of PVA interfaces limits Li+
ion conduction through blocking at the interface. In this section, an EC is developed (Figure 6II-
6a) which accounts for the series combination of impedance contributions from bulk P(VDF-
TrFE), bulk PVA, and electrode polarization caused by ionic charge build up. The model is fit to
impedance spectra over the temperature range 40oC – 110oC capturing the evolution from intrinsic
bulk polarization to extrinsic impurity ion transport at low frequency and high temperature. A
similar analysis was done in Chapter 6I. The complex modulus (M’M’’) is considered because of
its sensitivity to both P(VDF-TrFE) and PVA bulk relaxations at high frequency. Figure 6II-6b
shows M’ and M’’ at T=40, 80, and 110oC with EC model fits super imposed on the raw data to
demonstrate the model’s ability to capture the behavior of the composite at all frequencies and
temperatures. Due to the two bulk EC components in series, both P(VDF-TrFE) and PVA
relaxations are represented by the model. To determine the model’s ability to accurately predict
composite behavior, the permittivity of the model is calculated from the series combination of
CPE2 and CPE5 fit estimations. Considering total series capacitance behaves as the reciprocal of
added capacitive responses, and the layer thickness of P(VDF-TrFE) and PVA layers are
equivalent for each material, the total capacitance of the EC can be derived as the following
equation:
𝐶𝐶𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑒 =휀𝑃𝑉𝐷𝐹휀𝑃𝑉𝐴휀𝑜𝐴
3𝑡𝑃𝑉𝐴휀𝑃𝑉𝐷𝐹 + 4𝑡𝑃𝑉𝐷𝐹휀𝑃𝑉𝐴 (6𝐼𝐼 − 1)
149
Figure 6II-6: a) EC used in fitting the composite impedance data. The model takes into account bulk responses from
doped P(VDF-TrFE), pure PVA and electrode/dielectric blocking polarization. b) EC fits to 4-layer M’ data at 40oC,
80oC, and 110oC. Good qualitative fits enforce model accuracy.
where εPVDF and εPVA are computed by taking the estimated values of CPE2 and CPE5 along with
total P(VDF-TrFE) and PVA layer thicknesses (tPVDF and tPVA) respectively and inserting into the
equation εr=Ct/εoA. The composite’s permittivity is then calculated by taking into consideration
CComp and total thickness of the multilayer assembly tComp by the following relationship:
휀𝑟,𝐶𝑜𝑚𝑝 =𝐶𝐶𝑜𝑚𝑝𝑡𝐶𝑜𝑚𝑝
휀𝑜𝐴 (6𝐼𝐼 − 2)
CPE2 and CPE5 estimates from the model are plotted in Figure 6II-7a and permittivity estimates
using Equation 6II-2 are compared to those calculated directly from C’ measurements. It is noted
that estimates for CPE5 corresponding to PVA capacitance contributions are an order of magnitude
150
Figure 6II-7: a) CPE2 and CPE5 EC outputs corresponding to P(VDF-TrFE) and PVA bulk capacitances as a function
of temperature are shown (left) and converted to permittivity (right). b) 1-layer blocking CPE estimations (left) and
4-layer blocking CPE estimations (right).
higher than those of CPE2 corresponding to P(VDF-TrFE). This is due to the geometry of PVA
relative to P(VDF-TrFE): total PVA layer thickness is ~1.7μm while P(VDF-TrFE) is ~9.6μm
causing a near order of magnitude disparity of capacitance, considering the relation for capacitance
C= εrεoA/t. The frequency of 10kHz is chosen to compare measured and modeled permittivities
because this frequency showed the most stable measurements of capacitance for P(VDF-TrFE),
PVA, and the composite samples. It is found that the estimated permittivity for the composite is
comparable to that of the measured value, reinforcing the ability of the EC model in Figure 6II-
6a to describe the behavior of the 4-layer system.
The EC element CPE1 represents blocking/interfacial polarization. Figure 6II-7b shows
the behavior of the blocking CPE’s Q value and its imperfection constant n (see equation 6I-5) as
a function of temperature for a 1-layer 0.25% doped control sample and its 0.25% 4-layer
equivalent. In both cases, the value of n is between 0.5 and 1.0 correlates to electrode/polymer
interfacial roughness and heterogeneous charge distribution at the interface. This is similar to
151
results reported in Chapter 6I Figure 6I-10. At low temperatures, the blocking CPE is not
required to accurately fit the impedance spectra. At high temperatures, the blocking CPE fit
estimations are required to obtain a good fit to the impedance data and progressively increase as a
function of temperature for the 1-layer film reflecting increased electrode polarization at the
electrode/dielectric interface. The 4-layer films show a similar behavior however Q values are
estimated two orders of magnitude lower in comparison to the 1-layer control. This is consistent
with low frequency permittivity measurements shown in Figure 6II-3 that exhibit two orders of
magnitude reduced low frequency polarization at high temperature. At high temperatures, the
response of the blocking CPE in layered films plateaus. This can be related to a reduced presence
of charge at the dielectric/electrode interface given the fundamental relationship for capacitance
that is charge divided by voltage (C=Q/V). Since the ratio of Li+ carriers to P(VDF-TrFE) is
maintained between the 1- and 4-layer films, reduced charge at the electrode/dielectric interface is
reasoned to be due to charge blocking and deflection at PVA barrier interfaces.
6II.3.4 Thermally Stimulated Depolarization Current Measurements (TSDC)
6II.3.4.1 TSDC on 1-layer P(VDF-TrFE)
Initial measurements focused on understanding 1-layer control sample sets of pure and
0.25% doped films. A parametric study was performed where Ep and the heating rate of the TSDC
experiment were systematically changed between 10 – 30MV/m and 5 – 1oC/min respectively. All
samples tested were spin casted from MEK and were 10μm in total thickness.
The depolarization current as a function of temperature for pure 1-layer P(VDF-TrFE)
films for Ep ranging from 10 – 30MV/m are shown in Figure 6II-8a, 8b, and 8c. Each graph is
plotted on both a linear and logarithmic axis to accentuate low current peaks otherwise not visible
due to scaling. A prominent shoulder in the current density that increases in magnitude with
increasing poling field strength between 30oC < T < 55oC is visible on graphs scaled
logarithmically. The origin of this feature is associated with interfacial polarization at the electrode
during poling due to its dependence on Ep as well as its vicinity to T ≈ Tp = 50oC. For temperatures
between 60oC and 125oC, there are two overlapping current peaks corresponding to separate
depolarization processes. Research by Faria et al. [214] found similar TSDC signals in P(VDF-
TrFE) to those in Figure 6II-8. In this work, the low temperature peak for P(VDF70/TrFE30) is
attributed to a ferroelectric-ferroelectric phase transition and the high temperature peak
152
Figure 6II-8: Depolarization current density for pure 1-layer 10μm films at a) Ep=10MV/m, b) Ep=20MV/m, and c)
Ep=30MV/m. All measurements were performed using tp=15min, Tp=50oC and heating rate 5oC/min.
corresponds to a ferroelectric-paraelectric phase transition related to Tc of the material. The origin
of primary and secondary β-phases of copolymer has been investigated in past work by Gregorio
Jr. and Botta [161] where the ratio of each phase developed depends on the vicinity of the
crystallization temperature to the crystallization onset temperature and melt temperature of the
alpha phase To. Although this special temperature To is not well defined for this material system,
Gregorio Jr.’s and Botta’s work suggests 1-layer films have been crystallized above To causing the
development of two β-phases with unique TSDC contributions.
The peaks in Figure 6II-8 do not exhibit repeatable maximum current magnitudes,
however their maximm temperatures (Tm) are unchanging within each sample set. For Ep between
10 and 20 MV/m, a low temperature peak is situated within the range of 83 – 86oC and high
temperature peak around 103 – 105oC (in the vicinity of Tc measured via DSC). Similarities
between the measured data and results from Faria et al. [214] suggest the high temperature TSDC
features in pure copolymer are predominantly due to molecular relaxations associated with
conformational changes within the crystal phase of the material. This is supported when the electric
field is increased to 30 MV/m (approaching coercive field Ec = 50MV/m) causing peak amplitude
increase. [215]. It should be noted that results obtained by Faria et al. [214] involve pellets pressed
at high temperature while samples in this dissertation involve the use of solvent. In this case,
153
depolarization from impurity ions cannot be neglected and could account for dissimilarity between
samples.
The heating rate was then reduced by half to 2.5 oC/min for an Ep = 20MV/m. It was found
that the magnitude of depolarization current density was unchanged, however a slight shift in the
high temperature peak occurred from the range of 103-104oC to a range of 107-109oC. Shifting in
the peak allows for greater distinction between low and high temperature features, enabling peak
fitting using peak deconvolution techniques. Peak fitting implemented Bucci-Fieschi theory which
describes depolarization currents in TSDC experiments arising from orientational dipole rotation
events as a function of temperature.17 The fitting procedure to perform peak deconvolution was
performed using a custom R-Studio script by a process described in Chapter 2 section 2.4.3. The
procedure begins by fitting the strongest signal. Parameter estimates of τo (relaxation time), Po
(polarization) and Ψ (activation energy) are then used to generate a function over the entire range
of the convoluted peak from which raw data can be subtracted, leaving the low temperature peak
without contribution from the high temperature signal. This deconvoluted spectrum is then fit and
parameter estimates for both signals are reported.
Figure 6II-9 shows the result of peak fitting for 3 samples along with fit parameter
estimations using the 2.5 oC/min, Ep=20MV/m, Tp=50oC, tp=15min measurement condition. Initial
fitting was performed on the most prominent signal which was the high temperature peak
corresponding to Tc transition (with the exception of sample 2). Fitting on this current peak shows
a good sum of squares residual of 2.1x10-6 A/m2 to 7.0x10-6 A/m2 which is an order of magnitude
below the smallest values present in the TSDC spectrum within the temperature range 60oC –
140oC. The activation energy estimated for high temperature peak associated with the ferroelectric-
paraelectric phase transition ΨF-P was estimated in the range of 1.8eV to 2.1eV with characteristic
relaxation time τo ranging from 7.74x10-28s to 3.51x10-24s, which are in the same range reported
for corona polled P(VDF-TrFE) [216]. Large values of activation energy are typically attributed
to space charge contributions to the TSDC signal and observed in work done by Fedosov et al.
[216], Gaur [126], and Indolia [217], however good fits using Bucci-Feischi theory suggest high
ΨF-F estimates reflect large scale motion associated with crystal phase transition. Similar results
have been recorded for the α relaxation of epoxy in both TSDC and dielectric spectroscopy [218].
17 Specifics regarding Bucci-Fieschi theory including equation and general derivation with annotated assumptions can be found in Chapter 2
section 2.3.3.3 while R-Studio Script used in peak fitting is found in Chapter 2 section 2.4.3.
154
Figure 6II-9: Fit TSDC spectrum using Bucci-Fieschi equation along with parameter estimates for entire 1-layer
sample set undergoing TSDC with the following parameters: Ep=20MV/m, Tp=50oC, tp=15min, and scan rate
2.5oC/min. Individual fit components along with total synthetic spectrum are shown.
Peak deconvolution is then performed using parameter estimates generated by fitting the
strongest signal to expose characteristics of the weak signal. The low temperature peak returned
lower parameter estimates for ferro-ferroelectric phase transition ΨF-F and larger estimates on τo
that are closer to the range of those reported for lightly doped PVDF/BaTiO3 nanocomposite
systems reported by Gaur [126]. The sum of squares calculation of the low temperature peak fit
however is calculated to an average of 2.30x10-5 A/m2 +/- 3.36x10-5A/m2 across the sample set
which is larger than the lowest current signals measured within the fit temperature range. This is
indictive that Bucci-Fieschi theory inadequately describes the ferro-ferroelectric phase transition
depolarization process. Characteristics of the ferro-ferroelectric depolarization peak that may
contribute to poor fit results could be a distribution in relaxation times associated with transition
of the crystal phase, or more likely the existence of a third peak not visible in the measured
spectrum that resides within the temperature range associated with space charge depolarization
events.
Addition of 0.25% LiClO4 in 1-layered films has a significant impact on the measured
TSDC and is displayed in Figure 6II-10a-10c for ramp rates 5, 2.5, and 1oC/min at Ep=20 MV/m.
Unlike pure P(VDF-TrFE) the doped films are characterized by a broad depolarization current
peak that dominates the signal spanning 60oC – 115oC. Generally, the peak maximums are
155
Figure 6II-10: Depolarization current density for 0.25% doped 1-layer 10μm films at heating rates a) 5oC/min, b)
2.5oC/min, and c) 1oC/min. All measurements were performed using tp=15min, Tp=50oC and Ep=20MV/m.
unchanging in temperature given individual measurement conditions indicating the sample sets are
characteristic of a single relaxation mechanism. The addition of ions produces an order of
magnitude increase in depolarization current density for a scan rate of 5oC/min relative to pure
control films. Reducing the scan rate causes a decrease in the magnitude of the current measured
which is a distinct characteristic of doped films relative to pure. Unlike pure films, there was no
definitive change in measured current density by increasing Ep from 20MV/m to 30MV/m,
indicating the TSDC of doped 1-layered films is dominated by Li+ impurity carrier migration
instead of depolarization events associated with P(VDF-TrFE)’s crystal phase. It should be noted
that irregularity and asymmetry in the TSDC of doped films prevents accurate fitting due to
phenomena pertaining to space charge depolarization in electrets [215], and thus the analysis of
doped 1-layer and 4-layer films in the next section will be done by comparison.
6II.3.4.2 TSDC on 4-layer P(VDF-TrFE)
Incorporation of thin PVA into the P(VDF-TrFE) structure is first analyzed in pure films.
Figure 6II-11a and 11b shows the TSDC of a pure 1-layer sample set compared to a pure 4-layer
sample set of equivalent thickness measured with Ep=20MV/m and heating rate of 2.5C/min.
156
Figure 6II-11: Depolarization currents for a) pure 1-layer sample set and b) pure 4-layer sample set. Experimental
TSDC conditions are Ep=20MV/m, Tp=50oC, tp=15min and a heating rate of 2.5oC/min for both sets.
Decrease in the high temperature depolarization current peaks occurs by an order of magnitude in
layered films relative to the pure control sample. The highest temperature peak associated with
ferro-paraelectric phase transition in the material decreases from 1.8x10-4 A/m2 in the pure 1-layer
to 4.0x10-5 A/m2 in the pure 4-layer composite. This can be explained by considering 3 features of
the composite sample. First, P(VDF-TrFE) individual layer thickness is ~2.40 μm and PVA layer
thickness is 0.56 μm, indicating the total volume of the sample is comprised of only 82% P(VDF-
TrFE). This indicates reduced amount of polarizable material present in composites relative to pure
1-layer control films. The second feature takes into consideration field distribution through PVA
relative to P(VDF-TrFE). This can be done conceptually considering the simplified case of a
multilayered dielectric consisting of two materials where the only free charge density contributions
+/-σ occur at the electrodes. Then the displacement can be related to total charge enclosed in the
system via Gauss’s law:
∫ �⃑⃑� ∙ 𝑑𝑎 = 𝑄𝑒𝑛𝑐 (6𝐼𝐼 − 3)
𝐷 = 𝜎 (6𝐼𝐼 − 4)
Equation 6II-4 can then be related to the electric field E in either material via the following
relationship for a linear dielectric:
𝐸 =𝜎
휀𝑟휀𝑜 (6𝐼𝐼 − 5)
157
By equation 6II-5 the ratio of field distributed through the P(VDF-TrFE) relative to PVA becomes
the ratio of εPVA/ εP(VDF-TrFE). This was approximated by using εr measured at 10kHz at 25oC for
each constituent material and is calculated to be 0.58 indicating the electric field drop in the PVA
layer is approximately 2 times larger than P(VDF-TrFE). Given depolarization current’s
dependence on Ep observed in pure 1-layer films discussed in section 6II.3.4, field reduction
within P(VDF-TrFE) layers as a result of integrated PVA must be considered as a potential
contributor to reduced currents measured in the TSDC Figure 6II-11. The third feature of the
composite is considered by combining results from impedance spectroscopy to TSDC data in
Figure 6II-11. Current suppression is most extreme for the peak occurring between 86-88oC,
where integration of PVA barrier layers cause the extinction of the ferro-ferroelectric
depolarization peak. TSDC fitting performed in section 6II.3.4.1 indicates that the depolarization
current peak in this temperature range is poorly described using the Bucci-Fieschi equation and is
ultimately linked to the likely hood of a conflicting relaxation process occurring in the same
temperature range. Considering the breadth of Li+ peaks observed in ionically doped samples, as
well as reduced ionic polarization in layered films initiating at T=90oC measured by EC fitting in
section 6II.3.3 Figure 6II-7b, the suppression of TSDC peaks in pure 4-layer composites are
believed to be influenced by the blocking of migrating impurity ions. This is supported by a
prominent peak occurring at low temperatures around Tp=50oC suggesting the presence of
enhanced interfacial polarization in layered samples relative to the 1-layer control.
The effect of interfaces on ionic depolarization in layered films is more thoroughly
investigated by analysis of the doped 4-layer composite sample set. Figure 6II-12a shows 4-layer
doped samples in comparison to the 1-layer doped control sample set for the measurement
conditions Ep=20MV/m and heating rate 2.5C/min. The current in doped layered films is reduced
by nearly an order of magnitude relative to their 1-layer control group, displaying current densities
in the same range of the low temperature peak measured for pure 1-layer films. It is also noted that
signals spanning a broad temperature range characteristic of doped 1-layer films is not featured by
layered films suggesting Li+ contributions to the depolarization current are suppressed in layered
films. Figure 6II-12b shows a single sample characteristic of the 1-layer pure film set in
comparison to a characteristic sample from the 4-layer doped sample set. Unlike the pure 1-layer
sample set that exhibits two depolarization features, doped layered films exhibit a single well-
defined peak centered at 95oC caused by Li+ ion migration. This comparison reveals impurity ion
158
Figure 6II-12: Depolarization currents for a) 1-layer doped sample set compared to 4-layered doped sample set and
b) 1-layer pure sample compared with 4-layer doped sample. Experimental TSDC conditions are Ep=20MV/m,
Tp=50oC, tp=15min and a heating rate of 2.5oC/min for both sets.
contributions to the TSDC is centered between the ferro-ferroelectric and ferro-paraelectric
depolarization peaks in the pure material. This observation explains Bucci-Fieschi theory’s
inadequate representation of low current peaks in the TSDC of pure copolymer: interference from
impurity ion depolarization.
6II.3.5 High Voltage Dielectric Breakdown
Breakdown experiments of pure 1-layer P(VDF-TrFE) film and a pure 4-layer samples
done using the procedure described in Chapter 2 section 2.3.3.4. F Dielectric breakdown results
are dependent on the total thickness of the sample tested [136] [139], thus the thickness of 1-layer
and 4-layer samples used were spin cast to near equivalent thicknesses of 9.6μm and 11.0μm
respectively. The total number of breakdown events was n=30 for both films and the results are
plotted in the Weibull plot in Figure 6II-13a. Both the 1-layer and 4-layer films display a bimodal
Weibull distribution, suggesting the presence of defect-controlled breakdown events at low electric
field and breakdown events associated with the material’s intrinsic behavior at higher fields. This
is a phenomenon observed in the similar material P(VDF-TrFE-CTFE) in both electroded and
unelectroded samples undergoing ball and plate dielectric breakdown under the same IEEE
standards [128]. In the case of layered films,
159
Figure 6II-13: High voltage dielectric breakdown Weibull analysis for a) all breakdown fields exhibiting bi-modal
Weibull distributions separating defect and intrinsic type breakdown mechanisms and b) 7 lowest Ebd events analyzed
under IEEE standards for small sample sizes.
the dielectric breakdown at the lower tail of the distribution is most significantly impacted where
a substantial gap in ln(Ebd) occurs. Linear regression was performed on the first 7 data points in
each distribution and plotted in Figure 6II-13b. The R2 values for pure and layered films were
0.91 and 0.95 respectively, which satisfies the minimum R2 criteria of 0.91 for a sample population
of 7 as defined by IEEE standards. This indicates that the low field values can be described
sufficiently using Weibull statistics for analysis. Analysis was performed using a procedure
outlined for small data sets of singly censored data found in IEEE standards for statistical analysis
of dielectric breakdown [127]. The estimated modulus (β) and characteristic breakdown strength
(α) for the 1-layer and 4-layer defect driven distribution are listed adjacent to Figure 6II-13b in
table format. With the addition of interfaces, estimated β is calculated to increase by 59% from 23
(1-layer) to 37 (4-layer). Similarly, a 27% increase in α is also calculated from 394MV/m (1-layer)
to 501MV/m (4-layer). The increase in modulus and characteristic breakdown strength indicate
interfaces have a significant effect on defect driven breakdown events. In the context of impurity
ion blocking observed in impedance spectroscopy and TSDC in sections 6II.3.2-6II.3.4, layers
can be viewed as barriers to defect propagation through the dielectric, limiting the effect of defects
160
on high field performance by their deflection at barrier interfaces. It should be noted that field
intensity reduction in P(VDF-TrFE) afforded by the presence of PVA should also be considered
in the analysis of breakdown results (briefly discussed in section 6II.3.4.2) and should be
addressed in future work.
6II.4 CONCLUSIONS
An additive spin casting procedure was developed to fabricate P(VDF-TrFE)/PVA
multilayered laminates with reproducible layer thickness. SEM cross sections of P(VDF-
TrFE)/PVA layered dielectrics exhibit a more definitive interface than hot-pressed laminates
(Chapter 3). Thermal analysis of P(VDF-TrFE) films cast from MEK measure a decrease in the
Tc of 9oC in the pure material and 4oC in the doped films indicating the ferroelectric phase is
affected by solvent used to cast films.
Impedance spectroscopy captures the effect of the interface on low frequency conduction
in two ways: 1) blocking polarization at low frequency is decreased by 2 orders of magnitude and
2) tan(δ) relaxation frequency is lowered reflecting reduced ionic mobility in layered films. AC
conductivity is also impacted in layered films, exhibiting a two order of magnitude reduction in a
doped layered film relative to the 1-layer control. An EC model is developed that captures the
impedance response of both doped P(VDF-TrFE) as well as PVA interface layers and can be used
to predict material properties with high accuracy. Analysis of blocking circuit element estimates
reveal that 4-layered films reduce electrode polarization by two orders of magnitude, reinforcing
observations made on dielectric permittivity and loss data at high temperature and low frequencies.
TSDC on pure P(VDF-TrFE) revealed depolarization events associated with two β-phase
transitions: ferro-ferroelectric transition at 84oC<T<86oC and ferro-paraelectric transition in the
vicinity of the material’s Tc measured by DSC. High temperature peaks at Tc were described well
using Bucci-Fieschi theory, coinciding well with the assumption this signal occurs due to dipole
depolarization associated with the ferroelectric crystal. The low temperature peak was separated
from the raw signal implementing a peak deconvolution technique however ultimately returned
poor fits using the Bucci-Fieschi equation. TSDC on pure 4-layer films show significantly
suppressed depolarization current densities which is through to be due to a combination of 1) the
blocking of low levels of impurity species distributed thought the material and 2) composite effects
including reduced volume% of polarizable P(VDF-TrFE) and increased electric field drop in the
161
PVA layers during poling. The addition of Li+ impurity carriers enable more effective comparison
between 1-layer and 4-layer composites. Depolarization currents in doped 4-layer films were
measured to be an order of magnitude lower than the doped 1-layer control indicating Li+ blocking
at PVA interfaces.
Comparison to pure 1-layer films with the 4-layer doped group show a depolarization peak
situated between the ferro-ferroelectric and ferro-paraelectric peak caused by Li+ migration. This
demonstrates that high temperature depolarization in P(VDF-TrFE) TSDC is caused by a mixture
of β-phase crystal transitions combined with impurity ion depolarization contributions. This is in
good agreement with fitting results that show poor results when Bucci-Fieschi theory is applied to
low current level broad peaks at 84oC-86oC.
Finally, dielectric breakdown on pure 1-layer and pure 4-layer films of equivalent thickness
was performed. Results show bimodal Weibull distributions for both films indicating defect and
intrinsic type behavior mechanisms at play in both films. The defect dominated breakdown
mechanism showed most significant effect due to layering and is described well using Weibull
statistics. Statistical Analysis intended for small sample sizes indicate increased Weibull modulus
and characteristic breakdown strength increase by 27%, suggesting that internal barrier layers
within the dielectric mainly block and deflect defect propagation. This conclusion agrees well with
results from impedance spectroscopy, EC modeling and TSDC which demonstrate strong impurity
ion species blocking at internal PVA interfaces.
162
CHAPTER 7
CONCLUSIONS AND FUTURE WORK
7.1 CONCLUSIONS
In Chapter 1 of this dissertation, a brief history of capacitor technology is presented and
the performance criteria for polymer-based capacitors are reviewed. Various attempts at achieving
high energy density in polymer-based dielectrics by targeting their low permittivity are discussed.
Alternative approaches for improving dielectric breakdown strength in polar organics are also
reviewed. The chapter concludes with a description of the research goals and a general outline of
the thesis document.
Chapter 2 provides an in-depth description of material processing, characterization
equipment, and analytical techniques used in this research. Multilayered laminates were fabricated
via hot-pressing and spin casting techniques. Spin casting was found to produce films with greater
repeatability, control, and uniformity. The material processing section is concluded with an
overview of plasma surface modification and electrode deposition techniques. Each piece of
equipment used in chemical, structural, and electrical analysis is listed with a cursory explanation
of underlying fundamental physics required to understand its purpose. Greater detail is provided
for equipment that this dissertation relies on most heavily. The last section of this chapter provides
a description of analytical techniques used in impedance spectroscopy and thermally stimulated
current discharge data analysis.
In Chapter 3, hot-pressing pure PVDF multilayered laminates was explored. SEM
imaging of the cross section of layered films verified the existence of interfaces between hot
pressed films. High voltage dielectric breakdown using IEEE experimental and statistical analysis
standards were usedn. 1-layer samples exhibited a 415 MV/m breakdown while 2- and 3-layers
increase to 480 and 490 MV/m, respectively. The higher dielectric breakdown strength for
laminates were attributed to barrier effects at internal interfaces and processing defect reduction in
multilayer films with thinner individual layers. It was concluded that in order to continue the study
of role of interfaces on charge transport, total thickness of the dielectric must be reduced while
layer count must be increased. This motivated the start of spin cast film fabrication discussed in
Chapter 4.
163
Chapter 4 presents a plasma surface modification procedure (describes in Chapter 2) for
altering the electrode/dielectric contact chemistry. The intent was to study the impact of electrode
contact properties on high field charge injection of thin films. A considerable amount of surface
chemical and structural characterization is performed as a function of plasma treatment and
annealing. First, polymer surface chemistry is analyzed using XPS for untreated and
50%CF4/50%O2 gas plasma treated thin films. Synthetic peak fitting of the XPS C1s spectrum
indicated the addition of carbonyl C=O groups on treated films. Fitting of the O1s spectrum also
indicated the uptake of CF-O and C-O moieties. It was determined that the treatment time duration
had no effect on the quantity of chemical species detected by XPS; however, annealing after
plasma treatment did affect the ratio of F/C detected. Annealing also impacted surface topology:
films annealed after plasma treatment resulted in a surface roughness equivalent to an untreated
film. This is seen as a procedure that can repair damage done to the surface of the film during
plasma treatment, allowing the surface chemistry to dominate during contact angle and electrical
measurements.
Water contact angle experiments showed a near constant contact angle for various plasma
treatment durations in the post-anneal set, corresponding with constant surface chemistry after
treatment as well as uniform surface roughness measured via profilometry. With copolymer
surface chemistry and structure well understood, electrical analysis involved low field
spectroscopy and I(V) measurements. Plasma treatment had no significant impact on dielectric
constant; however, the resistivity decreased from 8x1011 Ω-m to 0.8x1011 Ω-m after plasma
treatment. The reduced contact resistance was attributed to grafted chemical species. The results
for high field I(V) were similar: leakage current increased after plasma treatment. Electronic
conduction models (Poole-Frenkel theory and Schottky theory) were considered and the dominant
high field conduction mechanism is Schottky emission. Finally, similar work performed by Reddy
[173] were applied to develop a comparative mathematical technique used to extract approximate
barrier height change due to grafted chemical species. The barrier height was lowered by 0.05eV,
causing enhanced emission in plasma treated films. It is posited that the inclusion of acceptor type
states (F and O) to the surface of P(VDF-TrFE) increase hole density at the electrode dielectric
contact, likely causing enhanced hole transport at the electrode dielectric contact.
164
Chapter 518 investigates similar phenomenon as Chapter 4, however using PI as the
material system. Unlike PVDF copolymer, PI is a non-polar high glass transition temperature
material typically considered for high temperature dielectrics. Thus I(V) measurements were
performed as a function of temperature beginning at 25oC and ending at 150oC. I(V) data was
analyzed using a combination of PF, Hopping, and Schottky theories. PF-theory returned
unrealistic estimations of the materials permittivity based on spectroscopy data recorded in
Meddeb et al.[1], suggesting Hopping is more appropriate mechanism for bulk limited conduction
analysis. Hopping analysis was performed by using a bootstrap statistical procedure by which the
behavior of the sample set itself imparts the framework for statistical analysis. It was found that
Hopping theory describes the conduction characteristics well to 100oC, at which point the model
breaks down. TUnlike P(VDF-TrFE), PI has been observed to undergo predominantly electronic
dominated [194] [191] conduction processes. It is determined that the conduction properties of the
material under test play an intimate role in determining the effect of surface chemical modification
on high field conduction.
Chapter 6 is broken into two separate studies: 1) Chapter 6I which focuses on the role of
Li+ impurity ion transport in P(VDF-TrFE) and 2) Chapter 6II which exploits concepts developed
in 6I to study impurity ion transport in Li+ impregnated layered dielectrics.
In Chapter 6I, copolymer films are spin cast with varying wt% of LiClO4 ranging from
0% – 1.0%. DSC measurement indicate that the β-phase is significantly impacted by LiClO4
inclusion into the material, resulting in reduction of Tc from 110oC in the pure film to 104oC in the
1.0% doped sample. Impedance spectroscopy was then used to study dissolved Li+ ion transport
through the film as a function of frequency between 25oC – 110oC. The high frequency behavior
of the samples remained constant as a function of salt% in the film while low frequency
polarization and dissipation increased indicating that ionic contributions to the permittivity are
only active at low frequency. High temperature impedance analysis showed saturation in the
permittivity and fully resolved tan(δ) relaxations, implying the presence of Li+ impurity ion
blocking at the electrode dielectric interface. Observations in the behavior of the impedance as a
function of temperature and frequency along with the known P(VDF-TrFE) material structure were
used to develop an EC model to describe the doped system. Analysis of capacitive circuit elements
18 Data acquisition for LI I(V) analysis provided in Chapter 5 was performed in a separate study by Meddeb et al. [13] independent of the
mathematical analysis provided in this dissertation.
165
as well as resistive circuit elements determines that Li+ interactions with the crystalline/amorphous
interphase region limit bulk transport at low frequency. In the quasi DC regime at high
temperatures, the impedance is dominated by Li+ polarization at the electrode dielectric interface.
In this regard, low frequency and high temperature conduction in doped P(VDF-TrFE) films are
found to: 1) show space charge conduction mechanisms similar to those reported in literature for
a wide variety of polymer dielectric materials (Table 1-III), 2) be controlled by quantity of LiClO4
introduced into the films and 3) be dominated by interfaces in the bulk of the film
(amorphous/crystalline interphase) as well as the electrode/dielectric contact interface.
Chapter 6II begins with the establishment of processing protocol used to fabricate
multilayer copolymer films with PVA barriers. SEM imaging of the film’s demonstrates superior
uniformity as well as repeatability in layer geometry relative to the hot-pressing technique
discussed in Chapter 3. Impedance spectroscopy captured the effect of PVA interfaces on low
frequency conduction in two ways: 1) decreasing electrode/dielectric blocking polarization by two
orders of magnitude and 2) lowering tan(δ) relaxation frequency. Both outcomes suggest the
deflection of Li+ carriers at low frequencies prevent substantial charge build up at the
electrode/dielectric interface. An EC model is developed that captures the impedance response of
both doped P(VDF-TrFE) as well as PVA interface layers. Analysis of blocking circuit element
estimates reveal that 4-layered films reduce electrode polarization by two orders of magnitude,
reinforcing observations made on dielectric permittivity and loss data at high temperature and low
frequencies. TSDC measurements were introduced as a method to investigate charge transport at
elevated electric fields and effective DC frequencies. The TSDC of pure 4-layer films result in
reduced depolarization currents at high temperatures and show an additional low temperature peak
centered around Tp=50oC indicating interfacial polarization in layered films. Addition of Li+
increases the depolarization current by an order of magnitude in 1-layered doped films, however
similarly doped 4-layer films display current densities on the order of pure 1-layer films. Finally,
dielectric breakdown reveals both intrinsic and defect breakdown behavior in 1 and 4-layer
samples. Weibull analysis for small sample sizes 5<n<15 is performed on the defect related
distribution, reporting an increase in Weibull modulus by 60% and characteristic breakdown field
by 27% in layered films. The analysis suggests that internal barrier layers within the dielectric
mainly block and deflect extrinsic related conduction mechanisms and reduces their impact on the
dielectric’s electrical performance.
166
7.2 SIGNIFICANT CONTRIBUTIONS
Throughout the completion of this work, the following contributions were made to the field
of dielectrics for high energy storage high powered electrical insulation:
1) Effect of Plasma Treatment on High Field Performance: A significant contribution of this
work was to systematically investigate the effect of chemical surface modification on
polymer surface structure and on high field properties. In this study, it was found that both
polymer surface chemistry and surface texture are altered by the application of plasma
treatment. The application of a thermal anneal at 142oC for 24 hr post plasma treatment
restored the surface roughness to that of the untreated control while maintaining the
modified surface chemistry. To the best of my knowledge, this processing procedure has
not been implemented in other investigations studying interfacial dominated charge
emission. Also, this study compares the behavior of PI to P(VDF-TrFE). Unlike
copolymer, PI is a non-polar low permittivity polymer thought to exhibit conduction of
predominantly electronic carriers instead of holes. Grafting acceptor type O species to the
surface of PI results in suppressed leakage current which contradicts the behavior of plasma
modified copolymer. These findings agree with arguments linking the combination of the
material’s primary charge carrier and surface chemistry to high field interface-dominated
injection.
2) Data Analysis for Surface Modified Thin Films: A considerable effort was put forth in the
development of data analytical techniques to quantify high field leakage current behavior
in polymer dielectrics. One contribution was the development of a mathematical technique
to comparatively quantify material property change in plasma treated thin films using
Schottky theory. In this method, results from ToF-SIMS on plasma treated P(VDF-TrFE)
that approximate the chemically modified surface of the film to be ~1-3nm thick was used
to justify an unchanged Richardson constant for the material. This enabled the derivation
of Schottky barrier height change from two linearized Schottky plots of an untreated film
and plasma treated equivalent. Although this method is not used in any other literature, I
have encountered, similar assumptions were made by Reddy [173] who investigated the
effect of 30nm coatings of PVDF on charge injection properties of InP. Other contributions
have been made on the analysis of high field conduction using non-linearizable conduction
167
theories such as hopping. The hopping equation implemented to describe bulk limited
transport in PI depends on the hyperbolic function, preventing linearization that enables
simple linear regression for fitting. Non-linear regression is required, complicating the fit
parameter estimations. Non-linear models do not have a well-defined relationship between
parameter (in this case d and Jo for hopping theory) and predictor variables (applied field
E) which inhibits the creation of a single hypothesis test that can represent all nonlinear
models. This eliminates the ability to use the “P-values” for statistical interpretation of
goodness of fit. In this work, bootstrap statistics were employed where the behavior of the
sample set itself was used to effectively create the statistics involved in parameter
estimation analysis. Although a widely accepted technique, boot strapping has not been
rigorously implemented to describe nonlinear leakage current behavior in the current
standing body of literature. Introducing this technique can improve how non-linear
processes to small sample sets are treated and improve methodology used to determine
transitions between conduction mechanisms in dielectric materials over broad temperature
ranges in future research.
3) Impurity ion transport in P(VDF-TrFE): As discussed in Chapter 6I, the migration of
impurity ion species in PVDF and in other materials such as XLPE contribute to
degradation at high electric fields at long time scales. This work provides an in-depth
investigation of impurity ion transport through P(VDF-TrFE) which is achieved by doping
with low concentrations of LiClO4.Two unique contributions that stand out relative to other
publications investigating similar phenomena: 1) Ionic concentrations are kept low enough
that the structure of the material remains unaltered, enabling quantification of their
interaction with amorphous and crystalline regions of the bulk. Most literature focuses on
improving the ability to ionically conductors with high ionic concentrations, which causes
structural irregularities unassociated with the pristine material. These changes in
microstructure, including increased porosity or integration of aqueous electrolytes, mask
the response of impurity species on the material’s natural structure. Eliminating this effect
by keeping impurity concentration low allows for an accurate portrayal of how ionic charge
carriers impact electrical performance in dielectrics used for capacitor applications. 2) The
dominant ionic charge carrier is well known both in quantity and chemical species. The
nature of impurity ion species in polymers contributing to extrinsic conduction is
168
ambiguous and dependent on the material and its processing. In the case of solution cast
films, residual solvent is typically assumed to contribute to the response; however,
literature studying conduction through PVDF also cites the possibility that electrochemical
interactions between the electrode and dielectric surface contribute to the available ionic
carriers at high fields [12]. By controlling the quantity of LiClO4 added to the material, the
available quantity of Li+ carriers contributing to conduction can be tailored such that they
dominate low frequency conduction at high temperature. In this sense, I have created a
model material in which impurity ionic conduction can be fundamentally studied to
understand the effect of interfaces on impurity ion transport in layered films.
4) Creation of a Model Material to Understand the Role of Interfaces: the study of interfaces
on high and low field conduction PC/P(VDF-HFP) composites wasinvestigated by
Mackey and Zhou et al. [14] [15]. Mackey et al. hypothesized that increases in dielectric
breakdown strength in layered structures are caused by blocked ionic charge interfaces.
The local electric field is influence by planar discharge and breakdown channel deflection.
Zhou et al. continued this work and reports similar enhancements of dielectric breakdown
strength in layered films, and also notes slower ion migration behavior in films containing
a ‘tie layer” of PMMA between the PC/P(VDF-HFP) interface. My work creates a PVDF
matrix with controlled Li+ dopant concentration of, enabling direct investigation of charge
interaction with PVA interfaces. Unlike either work produced by Mackey et al. or Zhou et
al., this dissertation implements TSDC to bridge results obtained by low field impedance
spectroscopy measurements to results of high voltage dielectric breakdown experiments.
Poling at fields in the range of 20-30MV/m introduces depolarization current from dipole
orientation, electronic charge injection (Chapter 4 and 5), and Li+ impurity ion
migration/interfacial polarization (Chapter 6I and 6II). Simultaneous contribution of these
mechanisms provides a complete picture of how interfaces impact both intrinsic and
extrinsic aspects of low frequency conduction in layered composites. This is the first time
a model system has been developed which controls extrinsic conduction for understanding
charge blocking within a multilayered dielectric. A definitive charge blocking mechanism
was discovered for multilayering with increased resilience to defect driven breakdown
processes.
169
7.3 FUTURE WORK
7.3.1 Tailoring the Electrode/Dielectric Interface for Limited Current Injection
In this dissertation, I showed how electrode/dielectric contact chemistry impacts both low
and high field leakage current in Chapter 4. Grafting of O and F acceptor type chemical moieties
to the surface of the films via reactive plasma treatment causes a reduction in the Schottky barrier
height. It is posited that the introduction of acceptor states promotes the accumulation of carriers
in the valence band near the electrode/dielectric interface and enhances current emission due to
P(VDF-TrFE)’s tendency to conduct holes. This hypothesis is supported by observations found in
PI which is dominated by electronic conduction and had leakage current suppressed by the
introduction of acceptor type chemical moieties at the surface of the film. In order to reduce charge
emission in P(VDF-TrFE) via chemical surface modification, the following should be taken into
consideration:
1) Plasma gas Chemistry: The gas chemistry was limited to a 50/50 mixture of CF4/O2. and
other plasma chemistries were not considered due to the availability of gasses offered in
our facilities. Both O and F have relatively high electron negativities of 3.44 and 3.98
respectively. To move away from electron accepting states, gas plasmas with low electron
negativity elements should be considered, such as hydrogen-based plasma. Similarly, inert
gasses including He and Ar should also be considered to investigate their effect on material
structure in comparison to what was reported in Chapter 4.
2) Additional Surface Coating Techniques: This dissertation limits surface chemical
modification to the use of reactive ion plasma treatments, however other avenues to
achieving thin surface coatings should be considered. Both molecular layer deposition
(MLD) and atomic layer deposition (ALD) processes yield high quality thin film growth
with excellent thickness control. Ultimately this technique can be optimized to produce
high quality thin films of acceptor or donor type materials (such as B or Al doped Si vs. Bi
or P doped Si) at the surface where the effect of surface chemical state on P(VDF-TrFE)
injection can be more rigorously investigated.
7.3.2 Multilayer Dielectric Processing
A major hurdle that needed to be overcome in the processing of multilayer dielectrics was
solvent from consecutive solution depositions destroying the surface of preexisting films. This was
addressed by spin casting a thin film of PVA/H2O on top of P(VDF-TrFE) as a capping layer.
170
PVA’s chemical resistance to the solvent MEK enabled the capping layer to serve as protection to
the P(VDF-TrFE) residing underneath. Despite the practicality afforded by PVA’s chemical
resistance to most solvents, certain aspects of the material should be addressed in future work to
bolster the significance of claims made referring to the “composite effect” in Chapter 6II:
1) Processing P(VDF-TrFE) Thin Films Using MEK: Multilayer laminates cast using MEK
as the solvent for copolymer show porosity in the bulk microstructure. It is believed that
the volatility of MEK causes a rapid dry time that impedes the’s ability to form a dense
copolymer film. Future work should address this issue by re-visiting processing conditions
used to fabricate thin films of copolymer using MEK specifically 1) hot plate temperature,
2) solution weight%, and 3) solvent drying time.
2) Integration of New Barrier Layers: multilayer laminates were realized via a solution
casting technique, however similar to suggestions made in Section 7.2.1, other methods to
creating internal interfaces should be explored. ALD and MLD provide an avenue to create
ultra-thin interfaces on the order of 10’s-1nm thick. Other techniques such as vapor
deposition can also be used to create multilayered dielectrics. For example, Perylene
coating is a common and viable option for interface layer deposition, however it was not
tried in this dissertation. It should be noted that a barrier layer of different chemistry could
support P(VDF-TrFE)/DMF solution deposition, which would reduce copolymer porosity
through solvent selection.
3) Field Distribution in Composites: a conceptual derivation using Gauss’s law and the
equation for electric field in a linear dielectric is used to approximate field drop across the
PVA relative to P(VDF-TrFE) layers. This derivation makes the assumption that 1) the
permittivity of the material is taken at room temperature, 2) free charges do not contribute
to free charge density at the PVA/copolymer internal interface and 3) both PVA and
P(VDF-TrFE) behave linearly with a poling field of 20MV/m and 220V applied across the
composite sample. In order to accurately portray the field distribution, modeling
implementing either COMSOL or ABAQUS should be implemented. In this study, the
geometry and permittivity of P(VDF-TrFE) and PVA layers can be controlled.
Conductivity of the PVDF layer in relation to PVA can also be considered. Similarly, free
charge density at the PVA/copolymer interface can be added in accordance with the amount
of charge introduced by controlled LiClO4 doping. The free charge state on the electrodes
171
can be controlled by a user defined function empirically derived from TSDC results of
layered films as a function of temperature and ramp rate. From this the model can predict
field distribution and as a function of temperature during TSDC, giving an internal “in-
situ” view of laminate depolarization.
7.3.3 Composite Characterization Using High Voltage Techniques
TSDC is able to capture a gamut of charge injection, conduction, and polarization events
simultaneously including injected electronic polarization, interfacial polarization, dipole
orientational polarization, and defect ion polarization. These conduction processes are all captured
using quasi DC frequencies and high poling electric fields, making TSDC a powerful link between
low and high field experiments (in this case linking Spectroscopy to I(V) and dielectric
breakdown). Future work should focus on implementation of TSDC to develop greater
understanding in layered dielectrics, as well as incorporate other high voltage techniques via the
following suggestions:
1) Thorough Investigation of TSDC Experimental Conditions: Depolarization processes in
polled electrets are strongly influenced by the experimental conditions during TSDC. In
this dissertation, heating rate and electric field intensity were varied to understand the
origins of TSDC features for 1-layer and 4-layer films. More work can be done
investigating the effect of Tp and tp. Measuring impedance data as a function of temperature
could capture the conditions that ionic space charge contribution dominates. Then, Tp and
tp can be adjusted to explore the TSDC of the material on time scales above and below the
ionic relaxation as a function of polling temperature.
2) Comparison to Pulsed Electric Acoustic Measurements: In the case of studying the effect
of barrier layers on charge blocking, pulsed electroacoustic measurements (PEA) would
enhance analysis by providing an internal depiction charge distribution of a poled, unpoled,
and de-poled layered composite. Typically, PEA measurements are performed on bulk
films ~100μm in thickness. From a processing point of view, this is easily achievable with
careful selection of solution deposition parameters during spin casting. PEA will capture
the distribution of charges through the films as a function of poling condition. Given
features found in TSDC that give good indication of carrier type (peak location and shape)
and relative quantity (through TSDC peak integration), features measured in PEA can be
easily identified based on location and relative charge quantity. Another attribute of PEA
172
Figure 7-1: Space charge distribution between cathode and anode of INS3-SC1 cable insulation under
60kV/mm stress as a function of measurement time. Packets of positive and negative charge are clearly
resolved [16].
measurement is for the technique’s ability to capture both charge location as well as charge
sign simultaneously. This is exemplified in Figure 7-1 which shows a map of distributed
charge recorded by PEA measurement through the thickness of high voltage cable
insulation exposed to constant 60kV/mm as a function of measurement time. Regions of
positive and negative charge are easily resolved by the technique, and can be applied to
LiClO4 doped PVDF specimens as a method to determine if Li+ ions are the dominating
ionic carrier (as suggested by Tsuchida et al. [197]) or if ClO4- anions are also mobile. It is
expected that significant charge proportional to the quantity of Li+ introduced by doping
will be located at the interface of P(VDF-TrFE) and PVA, however PEA is necessary to
verify the accuracy of the model presented in Chapter 6II Figure 6II-1.
3) Transference Number Measurements: In this dissertation, it is assumed that the main
contributor to low frequency conduction at high temperatures is dominated by the passage
of impurity ions through the bulk of the dielectric. This assumption is made by interpreting
features present in dielectric data and comparing to literature that makes similar claims (see
Table 1-III). The possibility that conduction through the material could be a mixture of
both cations, anions and electronic carriers has not been scientifically tested. In order to
prove conduction under low frequency and high temperature is predominantly ionic,
transference number measurements should be performed in both doped and undoped
dielectric films. Similar analysis has been performed in P(VDF-TrFE) gel electrolytes by
173
Saikia and Kumar [219] that calculates transference numbers within 0.90-0.98 which imply
the conductivity is due to a single ionic carrier. A number of acceptable methods used to
calculate transference numbers in aqueous electrolyte solutions already exist [220] [221]
[222], however measurement in lithium salt doped non-aqueous systems are not well
documented [223]. Currently used methods such as potentiostatic polarization and nuclear
magnetic resonance are typically implemented [224] [225] [226] however reported results
need be interpreted cautiously due to the measurements inherent difficulty. This
measurement will be a powerful tool when combined with PEA measurements, enabling
visual representation of the space charge distribution throughout the dielectric film as a
function of time, but also conclusive evidence the measured charges are purely due to a
single ionic species: Li+ cations or CLO- anions.
4) High Field P-E Loops: Although the dielectric breakdown strength can be related to stored
energy density of the material via equation (1-2), a major assumption that the material is
a linear dielectric in the derivation of this formula renders its use inappropriate for our co-
polymer. Polarization/electric field (P-E) loops are a better way to characterize energy
density of nonlinear dielectrics and P-E can be controlled. In context of this work, these
experiments should be done as a function of both temperature and frequency (related to
Figure 6II-4a and 4b). The ferroelectric hysteresis of doped and undoped 1-layer and 4-
layer films should be measured initially at temperatures and frequencies space charge
polarization is not featured in the dielectric spectrum. This can then be compared to
equivalent measurements under experimental conditions where it is known space charge
dominates the impedance response. Integration of the polarization with respect to field will
return the effect blocking layers have on recoverable energy density and inform on how
charge blocking can be implemented to improve dielectric performance.
174
APPENDIX A
Annotated Code for I(V) Nonlinear Regression
The following script was used in the fitting and bootstrap procedure for all hopping
nonlinear analysis. Annotations in the code explaining non-trivial steps are written in green font
and initiated by the # symbol in the code. Comments are not observed by the program when fitting.
R function definitions imperative to the operation of the code are also presented in the comments.
Definitions are taken from a combination of R Documentation, a “Table of Useful R Commands”
found on www.calvin.edu, and user-based forums that were referenced when creating the script.
Procedures are explained in context of PI and PI* explained in the Bootstrap Procedure section
of this document.
Annotated Code
#Loaded Library:
library(nlstools)
#Defined physical constants associated with hopping theory:
q = 1
kb = 8.617*10^-5
t = 298.15
Phi = 0.25
#Compiling raw data:
#vect_E1 – vect_E3 are vectors containing electric fields associated with current measurement in the J(E)
experiments using the function c(a1, a2,…, an). This stores data a1,…,an as a vector of length n. These are considered
“independent variables”
vect_E1 = c(7692307.692, 15384615.38, 23076923.08, 30769230.77, 38461538.46, 46153846.15, 53846153.85,
61538461.54, 69230769.23, 76923076.92, 84615384.62,
92307692.31, 100000000, 107692307.7, 115384615.4)
vect_E2 = c(7692307.692, 15384615.38, 23076923.08, 30769230.77, 38461538.46, 46153846.15, 53846153.85,
61538461.54, 69230769.23, 76923076.92, 84615384.62,
92307692.31, 100000000, 107692307.7, 115384615.4)
vect_E3 = c(7692307.692, 15384615.38, 23076923.08, 30769230.77, 38461538.46, 46153846.15, 53846153.85,
61538461.54, 69230769.23, 76923076.92, 84615384.62,
92307692.31, 100000000, 107692307.7, 115384615.4)
vect_E = c(vect_E1, vect_E2, vect_E3) #creates a vector representing PI set’s independent variables
#y_PI1 – y_PI3 are vectors containing current densities at each field value for samples 1, 2, and 3 respectively. These
#are considered “dependent variables”.
y_PI1 = c(1.30E-08, 1.45E-07, 2.71E-07, 4.05E-07, 5.72E-07, 7.06E-07,
8.84E-07, 1.05E-06, 1.27E-06, 1.54E-06, 1.86E-06, 2.22E-06,
2.74E-06, 3.27E-06, 4.03E-06)
y_PI2 = c(6.74E-08, 1.46E-07, 2.27E-07, 3.25E-07, 4.08E-07, 5.38E-07,
6.13E-07, 7.73E-07, 9.68E-07, 1.16E-06, 1.41E-06, 1.75E-06,
2.10E-06, 2.62E-06, 3.22E-06)
175
y_PI3 = c(1.35E-07, 3.04E-07, 4.75E-07, 6.46E-07, 8.43E-07, 9.92E-07,
1.24E-06, 1.48E-06, 1.77E-06, 2.06E-06, 2.48E-06, 3.07E-06,
3.78E-06, 4.65E-06, 5.71E-06)
y_PI = c(y_PI1, y_PI2, y_PI3) # creates a vector representing PI set’s dependent variables
df = data.frame(y=y_PI,x=vect_E) #creates a data frame that is set PI. The data.frame() function is used to store the
#data as a table with their associated properties intact.
#Hopping theory function used to fit raw data:
JPI_nonlin = nls(y~(Jo*exp(-Phi/(kb*t))*sinh((q*d*vect_E)/(2*kb*t))),
start = list(Jo = 0.007,
d = 1*10^-9), data = df
) #The nls(f(x, Pn)) function determines the nonlinear least-squares estimates of a parameter or set
of #parameters Pn, using a nonlinear model f(x, Pn). Here, Jo and d are parameters being fit to equation (2) in this
#documents associated manuscript.
#Plotting raw data and fit function (used in the creation of nonlinear regression fits in the sextion “Raw Data with
#Non-linear Fits”)
plot(vect_E,y_PI,pch=19,col="black",main="Polyimide 25C", xlab="E (V/m)", ylab="J (A/m^2)",ylim=c(0,6*10^-
6))
lines(vect_E,predict(JPI_nonlin),col = "green") #The predict(f(x, Pn)) function yields the predicted values as a
#function of x, of the model function f(x, Pn).
#Bootstrap procedure to acquire confidence interval on parameter estimates:
iters = 10000
bs.samples = matrix(0,ncol=2,nrow=iters) #a defined unfilled matrix of 2 columns and 10,000 rows
#initiation of loop used fit 10,000 distributions formed by sampling with replacement.
#for(i in 1:γ) {procedure} – a flow control statement initiating a loop. It indicates to perform the ith iteration of the
#outlined procedure for a total of γ iterations spanning from 1 to γ.
for(i in 1:iters){
bs.inds = sample(45,45,replace=T) #sample(α, β, replace = T/F) – takes a sample from a vector. α represents total
#number of possible indices to select from a matrix, β represents number of elements to choose (in this situation,
#number of indices), replace = T/F indicates sampling with replacement yes or no, respectively.
bs.y = y_PI[bs.inds] #samples with replacement from vector y_PI to create PI* set’s dependent variables
bs.x = vect_E[bs.inds] #samples with replacement from vector vect_E to create PI* set’s independent variables
bs.df = data.frame(y=bs.y, x=bs.x ) #creates PI* set
#performs non-linear regression using hopping theory to PI*
bs_fit = nls(y~(Jo*exp(-Phi/(kb*t))*sinh((q*d*x)/(2*kb*t))),
start = list(Jo = 0.007,
d = 1*10^-9), data = bs.df
)
bs.samples[i,] = as.vector( summary(bs_fit)$coefficients[,1] ) #defines bs.samples matrix as vector containing
#10,000 non-linear regression outputs of fit parameters Jo and d. The The summary(object) – produces a result
#summary of various model fitting functions represented by “object”.
}
#Extraction of statistical parameter estimates
Jo_est = mean(bs.samples[,1])
d_est = mean(bs.samples[,2]) #obtains the mean for the indicated fit parameters
Jo_CI = quantile(bs.samples[,1], probs=c(0.025,0.975))
d_CI = quantile(bs.samples[,2], probs=c(0.025, 0.975)) #obtains the 95% confidence interval for the indicated fit
#parameters where quantile(α, prob = c(β,γ)) yields sample quantiles of an object or vector α. The quantile produced
#corresponds to probabilities defined by β (lower range) and γ (upper range).
176
Jo_hist = hist(bs.samples[,1], freq = TRUE, breaks = 500, main="Jo Histogram 25C",xlab="Range of
Jo",ylab="Counts")
d_hist = hist(bs.samples[,2], freq = TRUE, breaks = 500, main="d Histogram 25C",xlab="Range of
d",ylab="Counts")#displays the indicated fit parameter’s empirically derived probability distribution for visual
#representation.
177
APPENDIX B
Annotated Code for TSDC Peak Fitting
The following script was used in the fitting peak deconvolution procedure for pure 1-layer
P(VDF-TrFE) TSDC analysis. Annotations in the code explaining non-trivial steps are written in
green font and initiated by the # symbol in the code. Comments are not observed by the program
when fitting. Definitions are taken from a combination of R Documentation, a “Table of Useful R
Commands” found on www.calvin.edu, and user-based forums that were referenced when creating
the script.
Annotated Code
## Performs peak fitting (Bucci Theory) on initial TSDC peak. estimated fit parameters are then used to generate a
function over the entire temperature range data is collected (via a second input file) which is then subtracted from the
data to deconvolute peaks within the T range selected. ##
install.packages("cubature")
install.packages("nnet")
install.packages("stringi")
install.packages("fit.models")
install.packages("seewave")
install.packages("gdata")
library(cubature) #package cubature
library(nnet) #package nnet
library(seewave) #package seewave
library(stringi)
library(fit.models)
library(gdata)
data <- read.csv(file.choose()) # Input csv file location, Temperature should be in Kelvin, strongest peak
chosen
T <- na.omit(data$T) # Need to tailor data range to peak of interest
J <- na.omit(data$J)
Ea <- seq(1, 3, by=0.005) # Activation energies vector [eV]
P <- exp(seq(log(1e-3), log(2e-0), length.out=length(Ea))) # Polarization vector
kb <- 8.165656e-05 # Boltzmann constant [eV]
Tm <- T[which.max(J)] # Current peak temperature
b <- 2.5/60 # Heating rate [deg/s]
Tau <- c(1:length(Ea)) # Relaxation time
result <- c(1:length(T)) # Fitting vector
err <- matrix(nrow = length(P), ncol = length(Ea)) # Error matrix
#Peak fitting for loops using equation 2-43
for (j in seq(from=1, to=length(Ea), by=1)) {
Tau[j] <- kb*(Tm^2)/(b*Ea[j]*exp(Ea[j]/(kb*Tm)))
for (k in seq(from=1, to=length(P), by=1)) {
for(i in seq(from=1, to=length(T), by=1)){
result[i] <- P[k]/Tau[j]*exp(-Ea[j]/(kb*T[i]))*exp(-kb*T[i]^2/(b*Tau[j]*Ea[j])*exp(-Ea[j]/(kb*T[i])))
}
178
err[j,k] <- rms(result-J)
}
}
I = which(err == min(err), arr.ind = TRUE)
E = Ea[I[1]] # Fitted Best value for Ea
P0 = P[I[2]] # Fitted Best value for P
Tau0 = kb*(Tm^2)/(b*E*exp(E/(kb*Tm))) # Relaxation time using E
##### CHECK POINT: fit check (make sure it fits well visually and call out min(err)) #####
for(i in seq(from=1, to=length(T), by=1)){
result[i] <- P0/Tau0*exp(-E/(kb*T[i]))*exp(-kb*T[i]^2/(b*Tau0*E)*exp(-E/(kb*T[i])))
}
plot(T,J, type="b", ylim=c(0, 2e-4),xlab="",ylab="")
par(new = TRUE)
plot(T,result, ylim=c(0,2e-4), type="l", col="red",xlab="Temperature (K)", ylab="J (A / m2)")
####### fit check (make sure it fits well visually) ########
###### Import New Data set (Spans full TSDC T range ) ######
data <- read.csv(file.choose())
T <- na.omit(data$T)
J <- na.omit(data$J)
#generates fit function for data subtraction over TSDC temperature range
fit1 <- function(t){
fit1_1 <- P0/Tau0*exp(-E/(kb*t))*exp(-kb*t^2/(b*Tau0*E)*exp(-E/(kb*t)))
return(fit1_1)
}
df_fit1 <- data.frame(T, fit1(T))
plot(df_fit1, type="l", col="blue")
#exports fit function data points into its own file
x_name <- "fit T (K)"
y_name <- "fit J (A/m2)"
df_fit <- data.frame(T,fit1(T))
colnames(df_fit) <- c(x_name, y_name)
setwd(choose.dir())
write.csv(df_fit,file="S10 High T Peak Fit Function")
#subtracts data from fit function
df_raw <- data.frame(T, J)
plot(df_raw, type="b", col="black")
df_merged <- data.frame(within(merge(df_raw, df_fit1, by = "T"), J_sub <- J - fit1.T.))
df_subtracted <- data.frame(T, df_merged$J_sub)
plot(df_subtracted, type="b", ylim=c(0,2e-4)) # VISUAL CHECK: Make sure filtered data is clean
#exports subtracted data
x_name <- "T"
y_name <- "J"
df <- data.frame(T, df_merged$J_sub)
colnames(df) <- c(x_name, y_name)
179
setwd(choose.dir())
write.csv(df,file="0.25% Subtracted dat_function generated")
############ END creation of subtracted data set #############
### USER REQUIRED STEP: Select T range for Subtracted Peak ###
############ START fitting of subtracted data set ############
#import subtracted data set and undergo same fitting procedure
data <- read.csv(file.choose())
T <- na.omit(data$T)
J <- na.omit(data$J)
Ea2 <- seq(0.001, 2, by=0.002)
P2 <- exp(seq(log(1e-5), log(1e1), length.out=length(Ea2)))
kb <- 8.165656e-05
Tm2 <- T[which.max(J)]
b <- 5/60
Tau2 <- c(1:length(Ea2))
result2 <- c(1:length(T))
err2 <- matrix(nrow = length(P2), ncol = length(Ea2))
for (l in seq(from=1, to=length(Ea2), by=1)) {
Tau2[l] <- kb*(Tm2^2)/(b*Ea2[l]*exp(Ea2[l]/(kb*Tm2)))
for(m in seq(from=1, to=length(P2), by=1)) {
for(i in seq(from=1, to=length(T), by=1)){
result2[i] <- P2[m]/Tau2[l]*exp(-Ea2[l]/(kb*T[i]))*exp(-kb*T[i]^2/(b*Tau2[l]*Ea2[l])*exp(-
Ea2[l]/(kb*T[i])))
}
err2[l,m] <- rms(result2-J)
}
}
I = which(err2 == min(err2), arr.ind = TRUE)
E2 = Ea2[I[1]]
P02 = P2[I[2]]
Tau02 = kb*(Tm2^2)/(b*E2*exp(E2/(kb*Tm2)))
for(i in seq(from=1, to=length(T), by=1)){
result2[i] <- P02/Tau02*exp(-E2/(kb*T[i]))*exp(-kb*T[i]^2/(b*Tau02*E2)*exp(-E2/(kb*T[i])))
}
#plots deconvoluted peak with fitted function
plot(T,J, type="b", ylim=c(0, 1.5e-4), xlab="", ylab="")
par(new = TRUE)
plot(T,result2, ylim=c(0, 1.5e-4), type="l", col="red",xlab="Temperature (K)", ylab="J (A / m2)")
legend("topleft",
legend = c("Exp", "Fit"),
col = c(rgb(0.1,0,0),
rgb(0.9,0,0.1)),
pch = c(19,NA_integer_),
bty = "n",
pt.cex = 2,
cex = 1.2,
text.col = "black",
horiz = F ,
inset = c(0.1, 0.1))
180
APPENDIX C
P(VDF-TrFE) Poole-Frenkel Analysis
C.1 Poole-Frenkel Analysis
Both PI, and PPIDS sample sets are considered using Poole-Frenkel theory. The linear regression
is performed on the mean of the three measurements in each sample set at the temperatures of
25oC, 75oC, 100oC, 125oC, 150oC, and 175oC. The PF plots for data taken at 25oC, 100oC, and
Figure C-1: Linearized J(E) data into Poole-Frenkel plots for PI and PPIDS. Data is shown for measurements at a)
25oC, b) 100oC, and c) 150oC
181
150oC are shown in Figure C-1 to illustrate the plasma treatment’s effect on raw data through at
the indicated temperatures using PF theory. At low temperatures the effect of the plasma treatment
is not observable which is similar to what is seen during Schottky analysis. Unlike linearization
performed using Schottky theory, linearization using PF theory yields non-linear data at 25oC
indicating poor description of sample behavior in both sets at room temperature. At higher
temperatures, linearization produces a linear relationship between Ln(J/E) vs E1/2. To assess
accuracy of PF theory in describing conduction in the samples, material permittivity is calculated
using the slope of linear fit functions pertaining to the data. The slope of the fit can be converted
to permittivity under PF theory by equation (A-1) in the manuscript and reproduced below:
𝜖𝑟,𝑃𝐹 = [(𝑚𝑘𝑇)2𝜋𝜖𝑜
𝑞3]
−1
(A − 1)
Permittivity values calculated from the linear regression as a function of measurement
temperature are shown below in Figure C-2. Calculation of the permittivity from PF plots returns
values that are greater than the acceptable range defined by the limits n2 and εr at 1 kHz by at least
an order of magnitude at all temperatures. This indicates that PF theory is insufficient to describe
conduction through the material for either processing condition at all measurement temperatures.
Fig C-2: Permittivity values calculated from linear fits in PF plots. A shaded region is marked in both plots that
indicates the range between high frequency permittivity defined by polyimide’s refractive index squared (n2) and
permittivity measured at 1kHz.
182
APPENDIX D
Polyimide I(V) Nonlinear Regression Parameter Estimates
The following sections show parameter estimates as well as estimated model fits using the
procedure and code outlined in Chapter 2 section 2.4.1 for data discussed in Chapter 5 section
5.3.2.1 Figure 5-2.
D.1 Hopping Conduction Parameter Estimate Histograms
Figure D-1: Histograms of parameter estimates from PI* after 10,000 iterations for Jo and d at 25oC, 75oC, and 100oC.
183
Figure D-2: Histograms of parameter estimates from PPIDS* after 10,000 iterations for Jo and d at 25oC, 75oC, 100oC,
125oC, and 150oC.
184
Figure D-3: raw data from the PI data set (black points) displaying the converged fit result using nonlinear
regression (blue solid line) superimposed.
Figure D-4: Raw data from the PPIDS data set (red points) displaying the converged fit result using nonlinear
regression (blue solid line) superimposed.
185
APPENDIX E
Equivalent Circuit Estimate Error Reports
E.1 Capacitive EC Element Error Percent
The error percent of all parameter estimates for capacitive circuit elements at each
temperature and LiClO4 quantity is listed below in Tables E-I through E-III.
Table C-I: Error percent associated with CPE2 parameter estimates out-put from the model
Table C-2: Error percent associated with CPE3 parameter estimates out-put from the model
Table C-3: Error percent associated with CPE3 parameter estimates out-put from the model
186
E.2 Resistive EC Element Error Percent
The error percent of all parameter estimates at each temperature and LiClO4 quantity is
listed below in Tables E-IV.
Table E-IV: Error percent associated with R3 and R4 parameter estimates out-put from the model
187
Bibliography
[1] W. J. Long, D. Belanger, T. Brousse, W. Sugimoto, M. B. Sassin and O. Crosnier,
"Asymmetric Electrochemical Capacitors-Stretching the Limits of Aqueous Electrolytes,"
MRS Bulletin, vol. 36, no. 7, pp. 513-522, 2011.
[2] N. S. A. Manaf, M. S. A. Bistamam and M. A. Azam, "Development of High Performance
Electrochemical Capacitor: A Systematic Review of Electrode Fabrication Technique
Based on Different Carbon Materials," ECS Journal of Solid State Science and Technology,
vol. 2, no. 10, pp. M3101-M3119, 2013.
[3] J. Su and J. Zhang, "Recent development on modification of synthesized barium titanate
(BaTiO3) and polymer/BaTiO3 dielectric composites," Journal of Materials Science:
Materials in Electronics, vol. 30, pp. 1957-1975, 2019.
[4] V. Tomer and C. A. Randall, "High field dielectric properties of anisotropic polymer-
ceramic composites," Journal of Applied Physics, vol. 104, p. 074106, 2008.
[5] Z. Guoqiang, Y. Li, S. Tang, R. D. Thompson and L. Zhu, "The Role of Field Electron
Emission in Polypropylene/Aluminum Nanodielectrics Under High Electric Fields," ACS
Applid Materials & Interfaces, vol. 9, pp. 10106-10119, 2017.
[6] J. L. Wang, X. J. Meng and J. H. Chu, "New Properties and Applications of Polyvinylidene-
Based Ferroelectric Polymer," Ferroelectric Materials-Synthesis and Characterization,
vol. IntechOpen, 2015.
[7] A. J. Lovinger, "Ferroelectric Polymers," Science, vol. 220, no. 4602, pp. 1115-1121, 1983.
[8] T. Christen and M. W. Carlen, "Theory of Ragone Plots," Journal of Power Sources, vol.
91, pp. 210-216, 2000.
[9] L. Zhu and Q. Wang, "Novel Ferroelectric Polymers for High Energy Density and Low
Loss Dielectrics," Macromolecules, vol. 45, pp. 2937-2954, 2012.
[10] F. Guan, J. Wang, J. Pan, Q. Wang and L. Zhu, "Effects of Polymorphism and Crystallite
Size on Dipole Reorientation in Poly(vinylidene fluoride) and Its Random Copolymers,"
Macromolecules, vol. 43, pp. 6739-6748, 2010.
[11] Q. Chen, B. Chu, X. Zhou and Q. M. Zhang, "Effect of metal-polymer interface on the
breakdown electric field of poly(vinylidene fluoride-trifluoroethylene-
chlorofluoroethylene) terpolymer," Applied Physics Letters, vol. 91, p. 062907, 2007.
188
[12] L. Yang, J. Ho, E. Allahyarov, R. Mu and L. Zhu, "Semicrystalline Structure-Dielectric
Property Relationship and Electrical Conduction in a Biaxially Oriented poly(vinylidene
fluoride) Film under High Electric Fields and High Temperatures," Applied Materials &
Interfaces, vol. 7, no. 36, pp. 19894-19905, 2015.
[13] A. B. Meddeb, Z. Ounaies and M. T. Lanagan, "Enhancement of electrical properties of
polyimide films by plasma treatment," Chemical Physics Letters, vol. 649, pp. 111-114,
2016.
[14] M. Mackey, A. Hiltner, E. Baer, L. Flandin, M. A. Wolak and J. S. Shirk, "Enhanced
breakdown strength of multilayered films fabricated by forced assembly microlayer
coextrusion," Journal of Physics D: Applied Physics, vol. 42, p. 175304, 2009.
[15] Z. Zhou, J. Carr, M. Mackey, K. Yin, D. Schuele, L. Zhu and E. Baer, "Interphase/Interface
Modification on the Dielectric Properties of Polycarbonate/Poly(vinylidene fluoride-co-
hexafluoropropylene) Multilayer Films for High-Energy Density Capacitors," Journal of
Polymer Science Part B: Polymer Physics, vol. 51, pp. 978-991, 2013.
[16] D. Fabiani, G. C. Montanari, C. Laurent, G. Teyssedre, P. H. Morshuis, R. Bodega, L. A.
Dissado, U. H. Nilsson and A. Campus, "Polymeric HVDC Cable Design and Space Charge
Accumulation. Part 1: Insulation/Semicon Interface," IEEE Electrical Insulation Magazine,
vol. 23, no. 6, pp. 11-19, 2007.
[17] R. J. Mammone and M. Binder, "Effects of CF4/O2 gas plasma power/exposure time on
dielectric properties and breakdown voltages of PVDF films," Journal of Applied Polymer
Science, vol. 46, no. 9, pp. 1535-1538, 1992.
[18] M. Mackey, D. E. Schuele, L. Flandin, M. A. Wolak, J. S. Shirk, A. Hiltner and E. Baer,
"Reduction of Dielectric Hysterisis in Multilayered Film via Nanoconfinement,"
Macromolecules, vol. 45, pp. 1954-1962, 2012.
[19] W. Vollmann and H.-U. Poll, "Electrical Conduction in Thin Polymer Fluorocarbon Films,"
Thin Solid Films, vol. 26, no. 2, pp. 201-211, 1975.
[20] R. Bartnikas, "Performance Characteristics of dielectrics in the Presence of Space Charge,"
IEEE Transactions on Dielectrics and Electrical Insulation, vol. 4, no. 5, pp. 544-557,
1997.
[21] N. Hussin and G. Chen, "Analysis of Space Charge Formation in LDPE in the Presence of
Crosslinking Byproducts," IEEE Transactions on Dielectrics and Electrical Insulation ,
vol. 19, no. 1, pp. 126-133, 2012.
[22] Y. Lin, W. Du, D. Tu, W. Zhong and Q. Du, "Space charge distribution and crystalline
structure in low density polyethylene (LDPE) blended with high density polyethylene
(HDPE)," Polymer International, vol. 54, pp. 465-470, 2005.
189
[23] H. Smaoui, M. Arous, H. Guermazi, S. Agnel and A. Toureille, "Study of relaxations in
epoxy polymer by thermally stimulated depolarization current (TSDC) and dielectric
relaxation spectroscopy (DRS)," Journal of Alloys and Compounds, vol. 489, pp. 429-436,
2010.
[24] S. Diaham and M.-L. Locatelli, "Space-charge-limited currents in polyimide films,"
Applied Physics Letters, vol. 101, p. 242905, 2012.
[25] G. C. Chen, J. Su and L. J. Fina, "FTIR-ATR Studies of Drawing and Poling in Poly-mer
Bilaminate Films," Journal of Polymer Science Part B: Polymer Physics, vol. 32, no. 12,
pp. 2065-2075, 1994.
[26] V. Bharti, T. Kaura and R. Nath, "Ferroelectric Hysterisis in Simultaneously Stretched and
Corona-Poled PVDF Films," IEEE Transactions on Dielectrics and Electrical Insulation,
vol. 4, no. 6, pp. 738-741, 1997.
[27] N. Betz, A. Le Moel, E. Balanzat, J. M. Ramillon, J. Lamotte, J. P. Gallas and G.
Jaskierowicz, "A FTIR Study of PVDF Irradiated by Means of Swift Heavy Ions," Journal
of Polymer Science Part B: Polymer Physics, vol. 32, pp. 1493-1502, 1994.
[28] B. Hilczer and J. Kulek, "The Effect of Dielectric Heterogeneity on Pyroelectric Response
of PVDF," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 5, no. 1, pp.
45-50, 1998.
[29] B. Mattsson, H. Ericson, L. M. Torell and F. Sundholm, "Micro-Raman Investigations of
PVDF-Based Proton-Conducting Membranes," journal of Polymer Science Part A:
Polymer Chemistry, vol. 37, pp. 3317-3327, 1999.
[30] R. M. Silverstein and F. X. Webster, Spectrometric Identification of Organic Compounds,
New York: John Wiley and Sons, 1998.
[31] A. Garton, Infrared Spectroscopy of Polymer Blends, Composites and Surfaces, Munich:
Hanser Editorial, 1992.
[32] S. Wu, Polymer Interface and Adhesion, New York: Marcel Dekker, Inc., 1982.
[33] M. Xanthos, M. W. Young and J. A. Biesenberger, "Polypropylene/polyethylene
terephthalate blends compatibilized through functionalization," Polymer Engineering and
Science, vol. 30, no. 6, pp. 355-365, 1990.
[34] A. Kaminska, H. Kaczmarek and J. Kowalonek, "The influence of side groups and polarity
of polymers on the kind and effectiveness of their surface modification by air plasma
action," European Polymer Journal , vol. 38, no. 9, pp. 1915-1919, 2002.
190
[35] S. J. Park and H. Y. Lee, "Effect of atmospheric-pressure plasma on adhesion characteristics
of polyimide film," Journal of Colloid and Interface Science, vol. 28, no. 5, pp. 267-272,
2005.
[36] H. Kaczmarek, J. Kowalonek, A. Szalla and A. Sionkowska, "Surface modification of thin
polymeric films by air-plasma or UV-irradiation," Surface Science, vol. 507, pp. 883-888,
2002.
[37] Y. Li, J. Q. Pham, K. P. Johnston and P. F. Green, "Contact angle of water on polystyrene
thin films: effects of CO2 environment and film thickness," Langmuir, vol. 23, pp. 9785-
9793, 2007.
[38] S. M. Pawde and K. Deshmukh, "Surface characterization of air plasma treated poly
vinylidene fluoride and poly methyl methacrylate films," Polymer Engineering & Science,
vol. 49, no. 4, pp. 808-818, 2009.
[39] J. Ho, T. R. Jow and S. Boggs, "Historical Introduction to Capacitor Technology," IEEE
Electrical Insulation Magazine, vol. 26, no. 1, pp. 20-25, 2010.
[40] "https://ethw.org/Capacitors".
[41] D. G. Fitzgerald, "Improvements in electrical condensers or accumulators," British Patent
No. 3466/1876, September 2, 1876..
[42] G. F. Mansbridge, British Patent No. 19,451, 1900.
[43] G. F. Mansbridge, "The manufacture of electrical condensers," IEEE, vol. 41, p. 535, 1908.
[44] T. D. Huan, S. Boggs, G. Teyssedre, C. Laurent, M. Cakmak, S. Kumar and R. Ramprasad,
"Advanced polymeric devices for high energy density applications," Progress in Material
Science, vol. 83, pp. 236-269, 2016.
[45] D. A. McLean and H. G. Wehe, "Miniature lacquer film capacitors," Proceedings IRE, vol.
42, no. 12, pp. 1199-1805, 1954.
[46] E. Frackowiak and F. Beguin, "Carbon materials for the electrochemical storage of energy
in capacitors," Carbon, vol. 39, pp. 937-950, 2001.
[47] P. Sharma and T. S. Bhatti, "A Review on Electrochemical Double-Layer Capacitors,"
Energy Conversion and Management, vol. 51, pp. 2901-29012, 2010.
[48] R. Kotz and M. Carlen, "Principles and applications of electrochemical capacitors,"
Electrochemica Acta, vol. 45, pp. 2483-2498, 2000.
[49] T. C. Murphy and W. E. Kramer, "Proceedings of The 4th International Seminar on Double
Layer Capacitors and Similar Energy Storage Devices," Florida Educational Seminar, 1994.
191
[50] M. Bartsch, A. Braun, B. Schnyder, R. Kotz and O. Haas, "Bipolar glassy carbon
electrochemical double-layer capacitor: 100,000 cycles demonstrated," Journal of New
Materials for Electrochemical Systems 2, vol. 2, no. 4, pp. 273-277, 1999.
[51] M.-J. Pan and C. A. Randall, "A Breif Introduction to Ceramic Capacitors," IEEE Electrical
Insulation Magazine, vol. 26, no. 3, pp. 44-50, 2010.
[52] W. Y. Wang, D. F. Zhang, T. Xu, L. F. Li, T. Zhou and X. L. Chen, "Non-linear Electrical
Behavior and Dielectric Properties of (Ca, Ta)-Doped TiO2 Ceramics," Journal of Alloys
and Compounds, vol. 335, pp. 210-215, 2002.
[53] R. Guo, Y. Jiang and A. S. Bhalla, "Processing and annealing conditions on the dielectric
properties," of (Ta2O5)0.92(TiO2)0.08 ceramics, vol. 57, pp. 270-274, 2002.
[54] M. A. McCormick and E. B. Slamovich, "Microstructure Development and Dielectric
Properties of Hydrothermal BaTiO3 Thin Films," Journal of the European Ceramic
Society, vol. 23, no. 12, pp. 2143-2152, 2003.
[55] J. A. S. Ikeda and Y. M. Chiang, "Space Charge Segregation at Grain Boundaries in
Titanium Dioxide: I, Relationship Between Lattice Defect Chemistry and Space Charge
Potential," Journal of the American Ceramic Society, vol. 76, no. 10, pp. 2437-2446, 1993.
[56] Q. L. Wang, O. Varghese, C. A. Grimes and E. C. Dickey, "Grain Boundary Blocking and
Segregation Effects in Yttrium-Doped Polycrystalline Titanium Di-oxide," Solid State
Ionics, vol. 178, pp. 187-194, 2007.
[57] S. Chao and F. Dogan, "Effects of Manganese Doping on the Dielectric Properties of
Titanium Dioxide Ceramics," Journal of American Ceramic Society, vol. 94, pp. 179-186,
2011.
[58] T. Fujii, S. Watanabe, M. Suzuki and T. Jujiu, "Application of lead zirconate titanate thin
film displacement sensors for the atomic force microscope," Journal of Vacuum Science &
Technology B: Microelectronics and Nanometer Structures Processing, Measurement, and
Phenomena, vol. 13, pp. 1119-1122, 1995.
[59] P. Luginbuhl, G.-A. Racine, P. Lerch, B. Romanowicz, K. G. Brooks, N. F. de Rooij, P.
Renaud and N. Setter, "Piezoelectric cantilever beams actuated by PZT sol-gel thin film,"
Sensors and Actuator A, vol. 54, pp. 530-535, 1996.
[60] P. Murlat, M. Kohli, T. Maeder, A. Kholkin, K. Brooks, N. Setter and R. Luthier,
"Fabrication and Characterization of PZT Thin-film Vibrators for Micrometers," Sensors
and Actuators A, vol. 48, pp. 157-165, 1995.
192
[61] M.-A. Dubois and P. Murlat, "PZT Thin Film Actuated Elastic Fin Micromotor," IEEE
Transactions on ultrasonics, Ferroelectrics, and Frequency Control, vol. 45, no. 5, pp.
1169-1177, 1998.
[62] J. J. Bernstein, S. L. Finberg, K. Houston, L. C. Niles, D. Chen, E. Cross, K. K. Li and K.
Udayakumar, "Micromachined High Frequency Ferroelectric Sonar Transducers," IEEE
Transactions on Ultrasonics, Ferroelectrics, and Frenquency Control, vol. 44, no. 5, pp.
960-969, 1887.
[63] J. F. Li, K. Wang, F. Y. Zhu, L. Q. Cheng and F. Z. Yao, "(K,Na)NbO3‐Based Lead‐Free
Piezoceramics: Fundamental Aspects, Processing Technologies, and Remaining
Challenges," Journal of American Ceramic Society, vol. 96, no. 12, pp. 3677-3696, 2013.
[64] A. Maqbool, J. u. Rahman, A. Hussain, H. K. Park, T. G. Park, T. S. Song and M. H. Kim,
"Structure and temperature dependent electrical properties of lead-free Bi0.5Na0.5TiO3-
SrZrO3 ceramics," IOP Conference Series: Materials Science and Engineering, vol. 60, no.
1, 2014.
[65] S. J. Laihonen, U. Gafvert, T. Schutte and U. W. Gedde, "DC Breakdown Strength of
Polypropylene Films: Area Dependence and Statistical Behavior," IEEE Transactions on
Dielectrics and Electrical Insulation , vol. 14, no. 2, pp. 275-286, 2007.
[66] D. Tan and P. Irwin, "Polymer Based Nanodielectric Composites," in Advances in Ceramics
- Electric and Magnetic Ceramics, Bioceramics, Ceramics and Environment, IntechOpen,
2011, p. 116.
[67] P. M. Hergenrother, K. A. Watson, J. G. Smith Jr, J. W. Connel and R. Yokota,
""Polyimides from 2,3,3',4'Biphenyltetracarboxylic Dianhydride and Aromatic Diamines,"
Polymer, vol. 43, no. 19, pp. 5077-5093, 2002.
[68] P. M. Hergenrother, "The Use, Design, Synthesis, and Properties of High
Performance/High Temperature Polymers: an Overview," High Perfomance Polymers, vol.
15, pp. 3-45, 2003.
[69] F. Polymers, Ferroelectric Polymers Chemistry, Physics, and Applications, Boca Ranton.
FL: CRC Press, taylor & Francis Group, 1995.
[70] R. Gregorio Jr and M. Cestari, "Effect of Crystallization Temperature on the Crystalline
Phase Content and Morphology of Poly (vinylidene Fluoride)," Journal of Polymer
Science: Part B: Polymer Physics, vol. 32, pp. 859-870, 1994.
[71] R. Gregorio Jr and E. M. Ueno, "Effect of Crystallinie Phase, Orientation and Temperature
on Dielectric Properties of Poly (vinylidene fluoride) (PVDF)," Journal of Materials
Science, vol. 34, pp. 4489-4500, 1999.
193
[72] T.-C. Hsu and P. H. Geil, "Deformation and transformation mechanisms of poly(vinylidene
fluoride) (PVF2)," Journal of Materials Science, vol. 24, pp. 1219-1232, 1989.
[73] J. C. McGrath and I. M. Ward, "High effective draw as a route to increased stiffness and
electrical response in poly(vinylidene fluoride)," Poymer Communications, vol. 21, pp.
855-857, 1980.
[74] B. Mohammadi, A. A. Yousefi and S. M. Bellah, "Effect of tensile strain rate and elongation
on crystalline structure and piezoelectric properties of PVDF thin films," Polymer Testing,
vol. 26, pp. 42-50, 2007.
[75] D. Song, D. Yang and Z. Feng, "Formation of B-phase microcrystals from the melt of
PVF2-PMMA blends induced by quenching," Journal of Materials Science, vol. 25, pp. 57-
64, 1990.
[76] M. Schwartz, "Encyclopedia of Smart Materials, vols 1-2," New York, Wiley, 2002, p. 811.
[77] M. Benz and W. B. Euler, "Determination of the Crystalline Phases of Poly(vinylidene
fluoride) Under Different Preparation Conditions Using Differential Scanning Calorimetry
and Infrared Spectroscopy," Journal of Applied Polymer Science, vol. 89, pp. 1093-1100,
2003.
[78] S. L. Hsu, F. J. Lu, D. A. Waldman and M. Muthukumar, "Analysis of the Crystalline Phase
Transformation of Poly(vinylidene fluoride)," Macromolecules, vol. 18, pp. 2583-2587,
1985.
[79] P. D. Southgate, "Room-temperature poling and morphology changes in pyroelectric
polyvinylidene fluoride," Applied Physics Letters, vol. 28, no. 5, pp. 250-252, 1976.
[80] T. R. Jow and P. J. Cygan, "Dielectric breakdown of polyvinylidene fluoride and its
comparisons with other polymers," Journal of Applied Physics, vol. 73, no. 10, pp. 5147-
5151, 1993.
[81] T. Furukawa, K. Nakajima, T. Koizumi and M. Date, "Measurements of Nonlinear
Dielectric in Ferroelectric Polymers," Japanese Journal of Applied Physics, vol. 26, pp.
1039-1045, 1987.
[82] L. Zhu, "Exploring Strategies for High Dielectric Constant and Low Loss Dielectrics,"
Macromolecules, vol. 45, pp. 2937-2954, 2012.
[83] X. Zhang, Y. Shen, Z. Shen, J. Jiang, L. Chen and C.-W. Nan, "Acheiving High Energy
Density in PVDF-Based Polymer Blends: Suppression of Early Polarization Saturation and
Enhancement of Breakdown Strength," Applied Materials & Interfaces, vol. 8, no. 40, pp.
27236-27242, 2016.
194
[84] S. Wu, M. Shao, Q. Burlingame, X. Chen, M. Lin, K. Xiao and Q. M. Zhang, "A high-K
ferroelectric relaxor terpolymer as a gate dielectric for organic thin film transistors,"
Applied Physics Letters, vol. 102, p. 013301, 2013.
[85] G. Eberle, H. Schmidt and W. Eisenmenger, "Influence of Poling Conditions on the Gas
Emission of PVDF," in Proceedings of IEEE Conference on Electrical Insulation and
Dielectric Phenomena, Pocono Manor, PA, USA, 1993.
[86] N. Hozumi, H. Takeda, H. Suzuki and T. Okamoto, "Space Charge Behavior in XLPE
Cable Insulation under 0.2 - 1.2 MV/cm dc Fields," IEEE Transactions on Dielectrics and
Electrical Insulation, vol. 5, no. 1, pp. 82-90, 1998.
[87] L. Cao, L. Zhong, Y. Li, H. Ren, J. Gao and G. Chen, "Influence on Dicumyl Peroxide
Content on DC Performance of Polyethylene for DC Cables," in IEEE Conference on
Electrical Insulation and Dielectric Phenomenoa, Cancun, Mexico, 2018.
[88] T. Gabler, K. Backhaus, T. Gotz and S. Grobmann, "Effect of the non-linear electric
conductivity of mineral insulating oil on the dielectic strength at high DC voltage stress,"
in IEEE Conference on Electrical Insulation and Dielectric Phenomena, Cancun, Mexico,
2018.
[89] E. Adem, G. Burillo, E. Munoz, J. Rickards, L. Cota and M. Avalos-Borja, "Electron and
proton irradiation of poly(vinylidene fluoride): characterization by electron paramagnetic
resonance," Polymer Degredation and Stability, vol. 81, pp. 75-79, 2003.
[90] M. Mackey, D. E. Schuele, L. Zhu and E. Baer, "Layer confinement effect on charge
migration in polycarbonate/poly(vinylidene fluorid-co-hexafluoropropylene) multilayered
films," Journal of Applied Physics, vol. 111, p. 113702, 2012.
[91] Y. Lianyun, E. Allahyarov, F. Guan and L. Zhu, "Crystal Orientation and Temperature
Effects on Double Hysteresis Loop Behavior in a Poly(vinylidene fluoride-co-
trifluoroethylene-co-chlorotrifluoroethylene)-graft-Polystyrene Graft Copolymer,"
Macromolecules, vol. 46, pp. 9698-9711, 2013.
[92] B. Ameduri, "From Vinylidene Fluoride (VDF) to the Applications of VDF-Containing
Polymers and Copolymers: Recent Developments and Future Trends," Chemical Reviews,
vol. 109, pp. 6632-6686, 2009.
[93] J. Blaise and E. Grimaud, "(Elf Atochem S.A.". U.S Patent 4025709, 1977.
[94] X. Bacque and P. Lasson, "Solvay S.A.". Eur Patent 387938, 1990.
[95] H. Iserson, "Pennwalt Corp.". U.S Patent 3245971, 1966.
195
[96] J. A. Baxter, C. O. Eddy and H. C. Stevens, "PPG Industries, Inc". U.S Patent 3642755,
1972.
[97] S. Lanceros-Menendez, J. F. Mano, A. M. Costa and V. H. Schmidt, "FTIR and DSC
Studies of Mechanically Deformed B-PVDF Films," Journal of Macromolecular Science,
Part B, vol. 40, no. 3-4, pp. 517-527, 2001.
[98] A. Lonjon , L. Laffont, P. Demont, E. Dantras and C. Lacabanne, "Structural and electrical
properties of gold nanowires/P(VDF-TrFE) nanocomposites," Journal of Physics D:
Applied Physics, vol. 43, p. 345401, 2010.
[99] H. Ohigashi, N. Kagami and G. R. Li, "Formation of ferroelectric domains in a copolymer
P(VDF-TrFE)," Journal of Applied Physics, vol. 71, no. 1, pp. 506-508, 1992.
[100] G. Teyssedre, A. Bernes and C. Lacabanne, "Influence of the Crystalline Phase on the
Molecular Mobility of PVDF," Journal of Polymer Science Part B: Polymer Physics, vol.
31, pp. 2027-2034, 1993.
[101] R. D. Simones, M. A. Rodriques-Perez, J. A. De Saja and C. J. L. Constantino, "Tailoring
the Structural Properties of PVDF and P(VDF-TrFE) by Using Natural Polymers as
Additives," Polymer Engineering and Science, vol. 49, no. 11, pp. 2150-2157, 2009.
[102] Y. J. Shen, M. J. Reddy and P. P. Chu, "Porous PVDF with LiClO4 complex as 'solid' and
'wet' polymer electrolyte," Solid State Ionics, vol. 175, pp. 747-750, 2004.
[103] Q. Burlingame, S. Wu, M. Lin and Q. M. Zhang, "Conduction Mechanisms and Structure-
Property Relationships in High energy Density Aromatic Polythiourea Dielectric Films,"
Advanced Energy Materials, vol. 3, pp. 1051-1055, 2013.
[104] H. Lee, N. J. Smith, C. G. Pantano, E. Furman and M. T. Lanagan, "Dielectric Breakdown
of Thinned BaO-Al2O3-B2O3-SiO2 Glass," Journal of the American Ceramic Society, vol.
93, no. 8, pp. 2346-2351, 2010.
[105] C. M. Miller, "Adhesion and the surface energy components of natural minerals and
aggregates," Diss. Texas A&M University, 2010.
[106] M. Lampin, R. Warocquier-Clerout, C. Legris, M. Degrange and M. F. Sigot-Luizard,
"Correlation between substratum roughness and wettability, cell adhesion, and cell
migration," Journal of Biomedical Materials Research: An Official Journal of The Society
for Biomaterials and The Japanese Society for Biomaterials, vol. 36, no. 1, pp. 99-108,
1997.
[107] N. Fairley, "XPS Line Shapes and Curve Fitting," in Surface Analysis by Auger and X-ray
Photoelectron Spectroscopy, Trowbridge, UK, IM Publications, 2003, pp. 397-420.
196
[108] G. Beamson and D. Briggs, The XPS of Polymers Database, Manchester, UK:
SurfaceSpectra, Ltd, 2000.
[109] C. J. Powell, "Energy Calibration of X-ray Photoelectron Spectrometers. II. Issues in Peak
Location and Comparison of Methods," Surface Science and Interface Analysis, vol. 25,
pp. 777-787, 1997.
[110] N. Vandencasteele, M. Delphine and F. Reiners, "XPS and contact angle study of N2 and
O2 plasma-modified PTFE, PVDF and PVF surfaces," Surface and Interface Analysis, vol.
38, no. 4, pp. 526-530, 2006.
[111] M. D. Duca, C. L. Plosceanu and T. Pop, "Surface modifications of polyvinylidene fluoride
(PVDF) under rf Ar plasma," Polymer Degredation and Stability, vol. 61, pp. 65-72, 1998.
[112] E.-S. Kim, Y. J. Kim, Q. Yu and B. Deng, "Preparation and characterization of polyamide
thin-film composite (TFC) membranes on plasma-modified polyvinylidene fluoride
(PVDF)," Journal of Membrane Science, vol. 334, pp. 71-81, 2009.
[113] K. C. Kao, Dielectric Phenomena in Solids, Elsevier, 2004.
[114] J. He, J. Liu, J. Li, Y. Lai and X. Wu, "Enhanced ionic conductivity and electrochemical
capacity of lithium on battery based on PVDF-HFP/HDPE membrane," Materials Letters,
vol. 170, pp. 126-129, 2016.
[115] M. Marzantowicz, J. R. Dygas, W. Jenninger and I. Alig, "Equivalent circuit analysis of
impedance spectra of semicrystalline polymer," Solid State Ionics, vol. 176, pp. 2115-2121,
2004.
[116] A. Sussman, "Space-Charge-Limited Currents in Copper Phthalocyanine Thin Films,"
Journal of Applied Physics, vol. 38, p. 2738, 1967.
[117] K. C. Kao, "Electrical Conduction and Photoconduction," in Dielectric Phenomena in
Solids: With Emphasis on Physical Concepts of Electronic Processes , London, Elsevier
Academic Press, 2004, pp. 381-509.
[118] Y. Ito, M. Hikita and T. Kimura, "Effect of Degree of Imidization in Polyimide Thin Films
Prepaired by Vapor Deposition Polymerization on the Electrical Conduction," Japanese
Journal of Applied Physics, vol. 29, no. 6, pp. 1128-1131, 1990.
[119] P. Saxena and M. S. Gaur, "Electrical conduction mechanism of polyvinylidenefluoride
(PVDF) – polysulfone (PSF) blend film," Journal of Electrostatics, vol. 67, pp. 844-849,
2009.
197
[120] G. Sawa, S. Nakamura, K. Iida and M. Ieda, "Electrical Conduction of Polypyromellitimide
Films at Temperatures of 120-180 C," Japanese Journal of Applied Physics, vol. 19, no.
453, 1980.
[121] E. Sacher, "Dielectric Properties of Polyimide Film II," IEEE Transactions on Electrical
Insulation 2, Vols. EI-14, no. 2, pp. 85-93, 1979.
[122] M. A. Vecchio, A. B. Meddeb, M. T. Lanagan, Z. Ounaies and J. R. Shallenberger, "Plasma
Surface Modification of P(VDF-TrFE): Influence of Surface Chemistry and Structure on
Electronic Charge Injection," Journal of Applied Physics, vol. 124, no. 11, pp. 114102(1-
11), 2018.
[123] P. Braunlich, "Field-Induced Thermally Stimulated Currents," in Thermally Stimulated
Relaxation in Solids, New York, Springer-Verlag, 1979, pp. 135-219.
[124] B. B. Sauer and Y. H. Kim, "Structural heterogeneity in poly (methyl methacrylate) glasses
of different tacticity studied by thermally stimulated current thermal sampling techniques,"
Macromolecules, vol. 30, no. 11, pp. 3323-3328, 1997.
[125] R. M. Neagu, E. R. Neagu, I. M. Kalogeras and A. Vassilikou-Dova, "Evaluation of the
dielectric parameters from TSDC spectra: application to polymeric systems," Materials
Research Innovations, vol. 4, no. 2-3, pp. 115-125, 2001.
[126] M. S. Gaur, P. K. Singh, A. Ali and R. Singh, "Thermally stimulated discharge current
(TSDC) characteristics in b-phase PVDF–BaTiO3 nanocomposites," Journal of Thermal
Analysis and Calorimetry, vol. 117, pp. 1407-1417, 2014.
[127] N. Hampton, R. Ross, G. Stone and et al, "IEEE Std 930™-2004: IEEE Guide for the
Statistical Analysis of Electrical Insulation Breakdown Data," in IEEE American National
Standard (ANSI), New York, 2004.
[128] S. Ahmed and Z. Ounaies, "SPIE Smart Structures and Materials + Nondestructive
Evaluation and Health Monitoring," in A study of metalized electrodes self-clearing in
electroactive polymer (EAP) based actuators, Las Vegas, Nevada, United States, 2016.
[129] Y. Wang, J. Cui, Q. Yuan, Y. Niu, Y. Bai and H. Wang , "Significantly Enhanced
Breakdown Strength and Energy Density in Sandwich‐Structured Barium
Titanate/Poly(vinylidene fluoride) Nanocomposites," Advanced Materials, vol. 27, no. 42,
p. 6658–6663, 2015.
[130] B. Efron and R. Tibshirani, " Bootstrap Methods for Standard Errors, Confidence Intervals,
and Other Measures of Statistical Accuracy," Statistical Science, vol. 1, no. 1, pp. 54-77,
1986.
198
[131] P. Khodaparast and Z. Ounaies, "On the impact of functionalization and thermal treatment
on dielectric behavior of low content TiO2/PVDF nanocomposite," IEEE Transactions on
Dielectrics and Electrical Insulation, vol. 20, 2013.
[132] H. S. Nalwa, Ferroelectric Polymers: Chemistry, Physics, and Applications, New York:
Marcel Dekker, 1995.
[133] S. P. Samant, C. A. Grabowski, K. Kisslinger, K. G. Yager, G. Yuan, S. K. Satija, M. F.
Durstock, D. Raghavan and A. Karim, "Directed Self-Assembly of Block Copolymers for
High Breakdown Strength Polymer Film Capacitors," ACS Applied Materials and
Interfaces, vol. 8, no. 12, pp. 7966-7976, 2016.
[134] W. Li, Q. Meng, Y. Zheng, Z. Zhang, W. Xia and Z. Xu, "Electric energy storage properties
of poly(vinylidene fluoride)," Applied Physics Letters, vol. 96, p. 192905, 2010.
[135] M. Kobayashi, K. Tashiro and H. Tadokoro, "Molecular Vibrations of Three Crystal Forms
of Poly(vinylidene fluoride)," Macromolecules, vol. 8, no. 2, pp. 158-170, 1975.
[136] D. H. Park, K. Yoshino, K. Okuyama and et al, "Dependence of dielectric breakdown of
thin poly (vinylidene fluoride) film, on temperature and thickness," Electrical Engineering
in Japan, vol. 109, no. 4, pp. 1-6, 1989.
[137] H. Bluhm, "Static and dynamic breakdown strength of dielectric materials," in Pulsed
Power Systems: Principles and Applications, 2006, pp. 7-54.
[138] J. J. O'Dwyer, London, England: Oxford University Press, 1964.
[139] A. B. Meddeb, Optimization of Polymer-Based Nanocomposites for High Energy Density
Applications, Texas A&M University: Ph.D Dissertation, 2012.
[140] V. Senthil, T. Badapanda, S. N. Kumar, P. Kumar and S. Panigrahi, "Relaxation and
conduction mechanism of PVA: BYZT polymer composites by impedance spectroscopy,"
Journal of Polymer Research , vol. 19, p. 9838, 2012.
[141] A. K. Jonscher, "Dielectric relaxation in solids," Journal of Physics D: Applied Physics,
vol. 32, no. 14, p. R57, 1999.
[142] J.-B. Jorcin, M. E. Orazem, N. Pebere and B. Tribollet, "CPE analysis by local
electrochemical impedance spectroscopy," Electrochimica Acta , vol. 51, pp. 1473-1479,
2006.
[143] J. R. Macdonald, "Impedance Spectroscopy," Annals of Biomedical Engineering, vol. 20,
pp. 289-305, 1992.
199
[144] L. L. Sun, B. Li, Y. Zhao, G. Mitchell and W. H. Zhong, "Structure-induced high dielectric
constant and low loss of CNF/PVDF composites with heterogeneous CNF distribution,"
Nanotechnology, vol. 21, no. 30, p. 305702, 2010.
[145] M. a. G. P. Rabuffi, "Status quo and future prospects for metalized polypropylene energy
storage capacitors," IEEE transactions on plasma science 30.5, pp. 1939-1942, 2002.
[146] J.-K. e. a. Yuan, "Fabrication and dielectric properties of advanced high permittivity
polyaniline/poly (vinylidene fluoride) nanohybrid films with high energy storage density,"
Journal of Materials Chemistry, vol. 20.12, pp. 2441-2447, 2010.
[147] Z.-M. e. a. Dang, "Fundamentals processes and applications of high-permittivity polymer-
matrix composites," Progress in Materials Science, vol. 57.4, pp. 660-723, 2012.
[148] C. H. a. Q. Zhang, "Enhanced Dielectric and Electromechanical Responses in High
Dielectric Constant All-Polymer Percolative Composites," Advanced Functional Materials,
vol. 14, no. 5, pp. 501-506, 2004.
[149] T. D. Huan, S. Boggs, G. Teyssedre, C. Laurent, M. Cakmak, S. Kumar and R. Ramprasad,
"Advanced Polymeric Dielectrics for High Energy Density Applications," Progress in
Materials Science, vol. 83, pp. 236-269, 2016.
[150] J. a. E. M. U. R. Gregorio, "Effect of Crystallinie Phase, Orientation and Temperature on
Dielectric Properties of Poly (vinylidene fluoride) (PVDF)," Journal of Materials Science,
vol. 34, pp. 4489-4500, 1999.
[151] C.-M. T.-M. K. a. H. H. Chan, "Polymer surface modification by plasmas and photons,"
Surface Science Reports, Vols. 24.1-2, pp. 1-54, 1996.
[152] A. B. Z. O. a. M. L. Meddeb, "Enhancement of electrical properties of polyimide films by
plasma treatment," Chemical Physics Letters, vol. 649, pp. 111-114, 2016.
[153] F. D. Egitto, F. Emmi and R. S. Horwath, "Plasma etching of organic materials. I. Polyimide
in O2-CF4," Journal of Vacuum Science & Technology B: Microelectronics Processing
and Phenomena, vol. 3.3, pp. 893-904, 1985.
[154] H. Yaghoubi and N. Taghavinia, "Surface chemistry of atmospheric plasma modified
polycarbonate substrates," Applied Surface Science, vol. 257.23, pp. 9836-9839, 2011.
[155] C. Yang, X.-M. Li, J. ilron, D.-f. Kong, Y. Yin, Y. Oren, C. Linder and T. He, "CF 4 plasma-
modified superhydrophobic PVDF membranes for direct contact membrane distillation,"
Journal of Membrane Science, vol. 456, pp. 155-161, 2014.
[156] B. T. Ginn and O. Steinbock, "Polymer surface modification using microwave-oven-
generated plasma," Langmuir, vol. 19.19, pp. 8117-8118, 2003.
200
[157] M. J. Shenton, M. C. Lovell-Hoare and G. C. Stevens, "Adhesion enhancement of polymer
surfaces by atmospheric plasma treatment," Journal of Physics D: Applied Physics, vol.
34.18, p. 2754, 2001.
[158] C.-Y. Li and Y.-C. Liao, "Adhesive stretchable printed conductive thin film patterns on
PDMS surface with atmospheric plasma treatment," ACS applied materials & interfaces,
vol. 8.18, pp. 11868-11874, 2016.
[159] R. J. Mammone and M. Binder, "Increased breakdown strengths of polypropylene films
melt-extruded from plasma-treated resin," Journal of applied polymer science, vol. 46.9,
pp. 1531-1534, 1992.
[160] T. Yagi, M. Tatemoto and J.-i. Sako, "Transition behavior and dielectric properties in
trifluoroethylene and vinylidene fluoride copolymers," Polymer Journal, vol. 12, no. 4, pp.
209-223, 1980.
[161] R. a. M. M. B. Gregorio, "Effect of crystallization temperature on the phase transitions of
P(VDF-TrFE) copolymers," Journal of Polymer Science-B-Polymer Physics Edition, vol.
33.6, pp. 403-414, 1998.
[162] G. A. B. C. L. Teyssedre, "Coopertative movements associated with the curie transition in
P(VDF-TrFE) copolymers," Journal of Polymer Science Part B: Polymer Physics, vol.
33.6, pp. 879-890, 1995.
[163] K. e. a. Lau, "Effect of annealing temperature on the morphology and piezoresponse
characterization of Poly (vinylidene fluoride-trifluoroethylene) films via scanning probe
microscopy," Advances in Condensed Matter Physics, 2013.
[164] Y. L. K. L. a. Q. W. Jason Claude, "Electrical Storace in Poly(vinylidene fluoride) based
Ferroelectroc Polymers: Correlating Polymer Structure to Electrical Breakdown Strength,"
Chemistry of Materials, vol. 20, no. 6, pp. 2078-2080, 2008.
[165] G. a. D. B. Beamson, High resolution XPS of organic polymers, Wiley, 1992.
[166] C. e. a. Ribeiro, "Surface Roughness dependent osteoblast and fibroblast response on poly
(l-lactide) films and electrospun membranes," Journal of Biomedical Materials Research
Part A, vol. 103.7, pp. 2260-2268, 2015.
[167] V. B. a. Q. M. Zhang, "Dielectric Study of the Relaxor Ferroelectric Poly(vinylidene
Fluoride-trifluoroethylene) Copolymer System," Physical Review B, vol. 63, pp. 184103-1,
184103-6, 2001.
[168] T. T. M. a. M. I. Mori, "High-field phenomena in poly-p-xylylene thin films: effects of
plasma treatment," Journal of Physics D: Applied Physics, vol. 22.10, p. 1518, 1989.
201
[169] Y. S. B. a. R. R. Sun, "The effect of dipole scattering on intrinsic breakdown strength of
polymers," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 22, no. 1, pp.
495-502, 2015.
[170] T. e. a. Nakano, "Effect of polar groups on the electrical breakdown strength of plasma-
polymerized films," IEEE Transactions on Electrical Insulation, vol. 25, no. 6, pp. 1085-
1091, 1991.
[171] W. V. a. H.-U. Poll, "Electrical Conduction in Thin Polymer Fluorocarbon Films," Thin
Solid Films, vol. 26, no. 2, pp. 201-211, 1975.
[172] e. a. H. L. Mosbacker, "Role of Near-Surface States in Ohmic-Schottky Conversion of Au
Contacts ZnO," Applied Physics Letters, vol. 87, 2005.
[173] V. R. Reddy, "Electrical Properties of Au/Polyvinylidene Fluoride/n-InP Schottky Diode
with Polymer Interlayer," Thin Solid Films, vol. 556, pp. 300-306, 2014.
[174] H. C. a. E. Ayyildiz, "Temperature Dependence of Electrical Parameters of the Au/n-InP
Schottky Barrier Diodes," Semiconductor Science and Technology, vol. 20, pp. 625-631,
2005.
[175] O. M. Z. a. J. S. Cheng Huang, "High-Dielectric-Constant all-Polymer Percolative
Composites," Applied Physics Letters, vol. 82, no. 20, pp. 3502-3504, 2003.
[176] T. N. a. M. I. T. Mizutani, "TSC due to Space Charge in Polyvinylidene Fluoride," Journal
of Physics D: Applied Physics, vol. 17, pp. 1883-1887, 1984.
[177] J. Zhu, "Theory of Ragone Plots for Electrostatic Energy Storage," B.S. Thesis in
Engineering Science and Mechanics, 2018.
[178] T. Umemura, K. Akiyama and D. Couderc, "Morphology and Electrical Properties of
Biaxially-Oriented Polypropylene Films," IEEE Transactions on Electrical Insulation,
Vols. EI-21, no. 2, pp. 137-144, 1986.
[179] J. Ho and R. T. Jow, "HIgh Field Conduction in Piaxially Oriented Polypropylene at
Elevated Temperature," IEEE Transactions on Dielectrics and Electrical Insulation, vol.
19, no. 3, pp. 990-995, 2012.
[180] J. J. O'Dwyer, "Breakdown in Solid Dielectrics," in Conference on Electrical Insulation &
Dielectric Phenomena - Annual Report, Amherst, MA, USA, 1982.
[181] R. W. Johnson, J. L. Evans, P. Jacobson and R. Thompson, "HIgh-Temperature Automotive
Electronics," in Proceedings of International Conference for Advanced Packaging and
Systems, Reno, Nevada, 2002.
202
[182] R. W. Johnson, "The Changing Automotive Environment: High-Temperature Electronics,"
IEEE Transactions on Electronics Packaging Manufacturing, vol. 27, no. 3, pp. 164-176,
2004.
[183] P. M. Hergenrother, "The Use, Design, Synthesis, and Properties of High
Performance/High Temperature Polymers: an Overview," HIgh Performance Polymers,
vol. 15, pp. 3-45, 2003.
[184] R. Khazaka, M. L. Locatelli, S. Diaham, P. Bidan, L. Dupuy and G. Grosset, "Broadband
Dielectric Spectroscopy of BPDA/ODA Polyimide Films," Journal of Physics D: Applied
Physics, vol. 46, pp. 065501(1-7), 2013.
[185] Y. Kishi, T. Hashimoto, H. Miyake, Y. Tanaka and T. Takada, "Breakdown and space
charge formation in polyimide film Breakdown and space charge formation in polyimide
film," Journal of Physics: Conference Series, vol. 183, p. 012005, 2009.
[186] M.-L. Locatelli, C. D. Pham, S. Diaham, L. Berquez, D. Marty-Dessus and G. Teyssedre,
"Space Charge Formation in Polyimide Films and Polyimide/SiO2 Double-Layer Measured
by LIMM," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 24, no. 2, pp.
1220-1228, 2017.
[187] D. Das-Gupta, "Conduction Mechanisms and High-Field Effects in Synthetic Insulating
Polymers," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 4, no. 2, pp.
149-156, 1997.
[188] D. K. Das Gupta and T. Noon, "Phototransient Spectral Shifts in Polyethylene with High
Fields," Journal of Physics D: Applied Physics, vol. 8, pp. 1333-1340, 1975.
[189] N. Inagaki, S. Tasaka and K. Hibi, "Surface Modification of Kapton Film by Plasma
Treatments," Journal of Polymer Science Part A: Polymer Chemistry, vol. 30, pp. 1425-
1431, 1992.
[190] S. Muruganand, S. Narayandass, D. Mangalaraj and T. M. Vijayan, "Dielectric and
Conduction Properties of Pure Polyimide Films," Polymer International, vol. 50, pp. 1089-
1094, 2001.
[191] N. R. Tu and K. C. Kao, "High-Field Electrical Conduction in Polyimide Films," Journal
of Applied Physics, vol. 85, no. 10, pp. 7267-7275, 1999.
[192] K. Ikezaki, T. Kaneko and T. Sakakibara, "Effect of Crystallinity on Electrical Conduction
in Polypropylene," Japanese Journal of Applied Physics, vol. 20, no. 3, pp. 609-615, 1981.
[193] G. G. Raju, R. Shaikh and S. U. Haq, "Electrical Conduction Processes in Polyimide Films-
I," IEEE Transactions on Dielectrics and Electrical Insulation, vol. 15, no. 3, pp. 663-670,
2008.
203
[194] G. M. Sessler, B. Hahn and D. Y. Yoon, "Electrical Conduction in Polyimide Films,"
Journal of Applied Physics, vol. 60, no. 1, pp. 318-326, 1986.
[195] R. Hiyang, Z. Lisheng, W. Zhao, M. Liu, X. Yang, Y. Li, Q. Yu and L. Cao, "Influence of
Crosslinking Byproducts on DC Conductivity of HVDC XLPE Cable Insulation," in IEEE
Conference on Electrical Insulation and Dielectric Phenomena, Cancun, Mexico, 2018.
[196] "Ionic conduction in P(VDF-HFP)/PVDF–(PC+DEC)–LiClO4 polymer gel electrolytes,"
Electrochemica Acta, vol. 49, p. 2581–2589, 2004.
[197] E. Tsuchida, H. Ohno and K. Tsunemi, "Conduction of Lithium Ions in Polyvinylidene
Fluoride and its Derivatives-I," Electrochemica Acta, vol. 28, no. 5, pp. 591-595, 1983.
[198] H. S. Nalwa, Ferroelectric Polymers: Chemistry: Physics, and, CRC Press, 1995.
[199] F. Oliveira, Y. Leterrier, J.-A. Manson, O. Sereda, A. Neels, A. Commann and D.
Damjanovic, "Process Influences on the Structure, Piezoelectric, and Gas-Barrier
Properties of PVDF-TrFE Copolymer," Journal of Polymer Science, vol. 52, pp. 496-506,
2014.
[200] C. V. Chanmal and J. P. Jog, "Dielectric relaxations in PVDF/BaTiO3 nanocomposites,"
Express Polymer Letters, vol. 2, no. 4, pp. 294-301, 2008.
[201] V. Bharti and Q. M. Zhang, "Dielectric study of the relaxor ferroelectric poly(vinylidene
fluoride-trifluoroethylene) copolymer system," Physics Review B, vol. 63, p. 184103, 2001.
[202] S. B. Aziz, T. J. Woo, M. Kadir and H. M. Ahmed, "A conceptual review on polymer
electrolytes and ion transport models," Journal of Science: Advanced Materials and
Devices, vol. 3, pp. 1-17, 2018.
[203] T. Uma, U. Mahalingam and U. Stimming, "Solid Polymer Electrolytes Based on
Poly(vinylchloride)-Lithium Sulfate," Material Chemistry Physics, Vols. 245-249, p. 2005,
90.
[204] A. Eid and S. Mahmoud, "The A.C. Impedance of Polycrystalline CuInS2 Thin Films,"
Journal of Materials Science: Materials in Electronics, vol. 8, pp. 259-264, 1997.
[205] S. B. Aziz, R. M. Abdullah, M. A. Rasheed and H. M. Ahmed, "Role of ion dissociation on
DC conductivity and silver nanoparticle formation in PVA: AgNt based polymer
electrolytes: deep insights to ion transport mechanism," Polymers, vol. 9, no. 8, p. 338,
2017.
[206] M. Jacob, S. Prabaharan and S. Radhakrishna, "Effect of PEO Addition on the Electrolytic
and Thermal Properties of PVDF-LiClO4 Polymer Electrolytes," Solid State Ionics, vol.
104, pp. 267-276, 1997.
204
[207] L. Othman, K. W. Chew and Z. Osman, "Impedance spectroscopy studies of poly (methyl
methacrylate)-lithium salts polymer electrolyte systems," Ionics, vol. 13, no. 5, pp. 337-
342, 2007.
[208] M. Ravi, Y. Pavani, S. Bhavani, A. K. Sharma and V. V. R. N. Rao, "Investigations on
structural and electrical properties of KClO4 complexed PVP polymer electrolyte films,"
International Journal of Polymer Materials, vol. 61, pp. 309-322, 2012.
[209] R. Baskaran, S. Selvasekarapandian, N. Kuwata, J. Kawamura and T. Hattori, "Structure,
thermal and transport properties of PVAceLiClO4 solid polymer electrolytes," Journal of
Physics and Chemistry of Solids, vol. 68, pp. 407-412, 2007.
[210] M. F. Z. Kadir, S. R. Majid and A. K. Arof, "Plasticized chitosane-PVA blend polymer
electrolyte based proton battery," Electrochimica Acta, vol. 55, pp. 1475-1482, 2010.
[211] S. B. Aziz, M. F. Z. Kadir and Z. H. Z. Abidin, "Structural, Morphelogical and Electro-
chemical Impedance Study of CS:LiTf Based Solid Polymer Electrolyte: Reformulated
Arrhenius Equation for Ion Transport Study," International Journal of Electrochemical
Science, vol. 11, pp. 9228-9244, 2016.
[212] M. Petrowsky and R. Frech, "Salt concentration dependence of the compensated Arrhenius
equation for alcohol-based electrolytes," Electrochemica Acta, vol. 55, p. 1285–1288, 2010.
[213] U. Rammelt and G. Reinhard, "On the Applicability of a Constant Phase Element (CPE) to
the Estimation of Roughness of Solid Metal Electrodes," Electrochimica Acta, vol. 35, no.
6, pp. 1045-1049, 1990.
[214] R. M. Faria, J. M. Guimaraes Neto and O. N. Oliveira Jr, "Thermal Studies on VDF/TrFE
Copolymers," Journal of Physics D: Applied Physics, vol. 27, pp. 611-615, 1994.
[215] P. Braunlich, Thermally Stimulated Relaxations in Solids, Berlin Heidelberg: Springer-
Verlag, 1979.
[216] S. N. Fedosov, A. E. Sergeeva and A. F. Butenko, "Depolarization currents in Fresh and
Aged Corona Poled P(VDF-TrFE) Films," arXiv preprint arXiv:0704.3993, 2007.
[217] A. P. Indolia and M. S. Gaur, "Investigation of structural and thermal characteristics of
PVDF/ZnO nanocomposites," Journal of Thermal Analysis and Calorimetry, vol. 113, no.
2, pp. 821-830, 2013.
[218] H. Smaoui, M. Arous, H. Guermazi, S. Agnel and A. Toureille, "Study of relaxations in
epoxy polymer by thermally stimulated depolarization current (TSDC) and dielectric
relaxation spectroscopy (DRS)," Journal of Alloys and Compounds, vol. 483, pp. 429-436,
2010.
205
[219] D. Saikia and A. Kumar, "Ionic conduction in P(VDF-TrFE)/PVDF-(PC+DEC)-LiClO4
Polymer Gel Electrolytes," Electrochemica Acta, vol. 49, no. 16, pp. 2581-2589, 2004.
[220] M. A. Duncan and M. Dole, "MacInnes, Duncan A., and Malcolm Dole. "The transference
numbers of potassium chloride. New determinations by the Hittorf method and a
comparison with results obtained by the moving boundary method," Journal of the
American Chemical Society, vol. 53, no. 4, pp. 1357-1364, 1931.
[221] H. S. Harned and E. C. Dreby, "Properties of Electrolytes in Mixtures of Water and Organic
Solvents. IV. Trans-ference Numbers of Hydrochloric Acid in Water and Dioxane-Water
Mixtures from 0 to 50°," Journal of the American Chemical Society, vol. 61, no. 11, pp.
3113-3120, 1939.
[222] M. Bester-Rogac, R. Neueder and J. Barthel, "Conductivity of Sodium Chloride in Water
+ 1,4-Dioxane Mixtures at Temperatures from 5 to 35°C I. Dilute Solutions," Journal of
Solution Chemistry, vol. 28, no. 9, pp. 1071-1086, 1999.
[223] S. Zugmann, M. Fleischmann, M. Amereller, R. M. Gschwind, H. D. Wiemhofer and H. J.
Gores, "Measurement of transference numbers for lithium ion electrolytes via four different
methods, a comparative study," Electrochemica Acta, vol. 56, pp. 3926-3933, 2011.
[224] L. F. Li, H. S. Lee, H. Li, X. Q. Yang, K. W. Nam, W. S. Yoon, J. McBreen and X. J.
Huang, "New electrolytes for lithium ion batteries using LiF salt and boron based anion
receptors," Journal of Power Sources, vol. 184, pp. 517-521, 2008.
[225] Y. Aihara, T. Bando, H. Nakagawa, H. Yoshida, K. Hayamizu, E. Akiba and W. S. Price,
"Ion Transport Properties of Six Lithium Salts Dissolved in g-Butyrolactone Studied by
Self-Diffusion and Ionic Conductivity Measurements," Journal of The Electrochemical
Society, vol. 151, no. 1, pp. A119-A122, 2004.
[226] K. M. Abraham, Z. Jiang and B. Carroll, "Highly Conductive PEO-like Polymer
Electrolytes," Cheemistry of Materials, vol. 9, no. 9, pp. 1978-1988, 1997.
[227] M. A. Vecchio, A. B. Meddeb, Z. Ounaies, M. T. Lanagan and J. Shallenberger, "Schottky
Barrier Height Quantification of Plasma Treated P(VDF-TrFE) Thin Films," in Conference
on Electrical Insulation and Dielectric Phenomena, Cancun Mexico, 2018.
[228] M. A. Vecchio, Z. Ounaies, M. T. Lanagan and A. B. Meddeb, "Conference on Electrical
Insulation and Dielectric Phenomena," in Polymer Laminates for High Energy Density and
Low Loss, Toronto, Canada, 2016.
206
VITA
Michael Anthony Vecchio
Michael Anthony Vecchio was born in New York City, ESA in 1991, and was raised in
Long Valley New Jersey. Inspired by his father’s passion for science, Michael pursued a degree in
Physics and Mathematics at Dickinson College in Carlisle Pennsylvania, where he obtained his
Bachelor’s of Science in 2014. Unsatisfied, he enrolled at the Pennsylvania State University to
pursue a Ph.D. in Materials Science and Engineering in Fall of 2014. He served as a Graduate
Assistant in both the Electroactive Materials Characterization Laboratory lead by Dr. Zoubeida
Ounaies and the High Energy Capacitor Group by Dr. Michael T. Lanagan, which he is
particularly proud of.
Listed below are his publications during his tenure at Penn State University:
Journal Manuscripts
1. *“Conduction through Plasma Treated Polyimide: Analysis of High-Field Conduction by
Hopping and Schottky Theory” Michael A. Vecchio, Amira Barhoumi Meddeb, Michael
T. Lanagan, and Zoubeida Ounaies. Journal of Materials Science 54. 14 (2019): 10548-
10559.
2. *“Plasma Surface Modification of P(VDF-TrFE): Influence of Surface Chemistry and
Structure on Electronic Charge Injection.” Michael A. Vecchio, Amira Barhoumi Meddeb,
Michael T. Lanagan, Zoubeida Ounaies, and Jeff Shallenberger. Journal of Applied Physics
124, 114102 (2018); doi: 10.1063/1.5042751.
3. “Fabrication of Solid-State Multilayer Glass Capacitors.” Rudeger H. T. Wilke, Harlan
Brown-Shaklee, Adrian Casias, Billy Cunningham, Jr., Amanda Dean, Michael A.
Vecchio, and Rohith Vudatha. IEEE Transactions on Components, Packaging and
Manufacturing Technology 7.11 (2017): 1906-1910.
Conference Proceeding Papers
4. *“Schottky Barrier Height Quantification of P(VDF-TrFE) Thin Films” Michael A.
Vecchio, Amira Barhoumi Meddeb, Michael T. Lanagan, Zoubeida Ounaies, and Jeff
Shallenberger. (2018) IEEE Conference on Electrical Insulation and Dielectric
Phenomena, Cancun Mexico.
5. *“Plasma Surface Modification of P(VDF-TrFE) Dielectrics.” Michael A. Vecchio, Amira
Barhoumi Meddeb, Michael T. Lanagan, Zoubeida Ounaies, and Jeff Shallenberger. (2017)
18th US-Japan Seminar on Dielectric and Piezoelectric Materials, Santa Fe NM.
6. *“Polymer Laminates for High Energy Density and Low Loss.” Michael A. Vecchio,
Zoubeida Ounaies, Michael T. Lanagan, and Amira Barhoumi Meddeb. (2016) IEEE
Conference on Electrical Insulation and Dielectric Phenomena, Toronto Canada.