FulltextThesis Interessant

Non-Destructive Characterisation Of Structural Ceramics
Using Impedance Spectroscopy
A Thesis Submitted For The Degree Of Doctor Of Philosophy
By
Xin Wang
Department Of Mechanical Engineering
Brunel University
September,2001
Abstract
The aim of this project was to explore the potential applications of impedance
spectroscopy(IS) in the non-destructive characterisation of structural ceramics.
A major advantage in the use of the IS technique is its capability in distinguishing the
properties of different microstructural origins in materials, e.g., grains and grain

boundaries, etc. In this thesis, a review of the theoretical aspectsof IS is presented. An
analytical approach and numerical analyses are conducted to illustrate how impedance
spectra become resolved and why the spectral resolution is dependent on the
representation formalism as well as on the difference between the electrical properties of
different microstructural features.
Three categories of structural ceramics, i. e. A1203/SiC nanocomposites, thermal barrier
coatings and clay-based ceramics, were used as the model materials for this IS study.
Both sintering and degradation phenomenon have been examined. Conventional
analytical techniques, such as SEM, XRD, EDS, TGA and dilatometry, were used to aid
in the understanding of IS and to prove the reliability of impedance measurements.The
researchresults indicate:
" For A1203/SiC nanocomposites, IS can be used to examine the conducting
mechanisms for the materials containing different quantities of SiC particles. The
content of SiC can be correlated to the impedance spectral features and dielectric
constant of the material. The oxidation scale formed at the surface gives a separate
relaxation process. Based on the capacitive effect of this relaxation process, the
thickness can be determined non-destructively using IS.
" For thermal barrier coatings, three relaxation processes can be found in the
impedance spectra, which correspond to the top coating, oxidation layer and
microstructural defects in the top coating, respectively. The thickness of the

oxidation layer can be quantitatively related to the diameter of the corresponding
semicircle in electrical modulus spectrum. In the meantime, the top coating
degradation could also be monitored using IS.
1
" For clay-based ceramics, the impedance spectra consist of a high frequency
semicircle and a low frequency tail, which correspond to bulk effect and electrode
effect, respectively. The variation of the bulk conductivity with sintering time can
be quantitatively correlated to the densification of the material during sintering.
The electrode effect tail is directly related to the capacitive effect of the
electrode/specimen interface, which could be an effective indicator of the

electrode temperature.
Therefore IS is a useful technique for non-destructive characterisation of structural
ceramics.
il
List Of Publications
This thesis describes research carried out by the author in the Department of Material
Engineering initially and then Department of Mechanical Engineering at Brunel
University, under the supervision of Dr. Ping Xiao. No part of this thesis has been
accepted or is being currently submitted for any degree or other qualification in this
university or elsewhere. The publications that have been generated during my PhD
studies are listed as follows:
1) X. Wang and P. Xiao, "Nondestructive characterisation of alumina/silicon
carbide nanocomposites using impedance spectroscopy", J. Euro. Ceram. Soc.,

20 (2000) 2591-2599.
2) X. Wang, J. Mei and P. Xiao, "Determining oxide growth in thermal barrier
coatings(TBCs) non-destructively using impedance spectroscopy", J. Mater. Sci.

Lett., 20 (2001) 47-49.
3) X. Wang, J. Mei and P. Xiao, "Non-destructive evaluation of thermal barrier
coatings using impedance spectroscopy", J. Euro. Ceram. Soc., 21 (2001) 855-

859
4) X. Wang, and P. Xiao, "Characterisation of the sintering process of clay-based
materials using impedance spectroscopy",J. Euro. Ceram. Soc., in press.
5) X. Wang, and P. Xiao, " In-situ monitoring of ceramic sintering using

impedancespectroscopy", British Ceramic Proceedings, accepted
6) J. Mei X. Wang, and P. Xiao, " Characterisation of oxide layers and thermal
barrier coatings using impedance spectroscopy", Materials Research Society
SymposiumProceedings, accepted.
7) S.-H. Song, X. Wang, and P. Xiao, "Effect of microstructural defects on the
electrical properties of Ti02, " Mater. Sci. Eng. B, accepted.
The majority of results presented in this thesis have been published as follows:
chapter 5 was written with the results that have been published in ref 1; chapter 6
was basedon the results published in the refs. 2,3 and 6; chapter 7 was written with
the results published in refs. 4 and 5.
ill
Acknowledgements
I would like to express my deepest gratitude to my supervisor, Dr. Ping Xiao, for his
encouragement, guidance throughout my PhD studies, and the painstaking effort of

seeking financial support for this project.
I am grateful to all the technical staff and secretaries in the Department of Materials
Engineering and Department of Mechanical Engineering, for their support and help in
various ways. Thanks are due to Dr. Sue Woodise for her friendly attitude and
professional instruction in setting up XRD, TGA and dilatometer etc.
I am deeply indebted to Dr Houzheng Wu, at Oxford University, for his kindness in
providing A1203/SiC samples,stimulating discussions,and most importantly, friendship.
I would like to thank my colleagues, Junqin Li, Shenhua Song, Ammer Jadoon, Md
Shawkat Ali and Ying Yuan, for the jokes, the funny anecdotesas well as the friendly
arguments. Thanks to them, my PhD has never become a `PermanentheaDache'
I want to thank my parents for their everlasting support and love. Special thanks to my
wife, Yanxiang, and my daughter, Yuanyuan, for their patience, support and
encouragementthroughout all these difficult times.
iv
Contents
Abstract i
List Of Publications iii
Acknowledgements iv
1 Introduction 1
1.1 The Need for Non-Destructive Characterisation (NDC) of Ceramics 1
1.2 The State-Of -The-Art Of NDC Techniques For Ceramic Materials 1
1.3 Impedance Spectroscopy As A NDC Technique 2

1.4 About This Study 3
2 Literature Review 5
2.1 Introduction 5
2.2 Impedance And Impedance Spectroscopy 7
2.2.1 ImpedanceMeasurement 7
2.2.2 Concept of Impedance 8
2.2.3 Various Formalisms Of IS 9
2.3 General Characterisation Procedure Using IS 10
2.4 Electrical Conduction In Ceramics Materials 12

2.4.1 Introduction 12
2.4.2 Electrical Conduction In Semiconductors 12
2.4.3 Ionic Conduction In Crystalline Materials 13
2.4.4 Ionic Conduction In Glass And Molten States 14

2.4.5 Ionic Conduction on Particle Surfaces 16
2.5 Physical Model And Equivalent Circuit 17
2.5.1 Brick Layer Model 17
2.5.2 Constriction Model 19
2.5.3 Multi-layer Model 21
2.5.4 Continuous Conductive Path Model 22
2.6 Representations Of Impedance Spectroscopy 24
2.7 Non-Ideality Of Realistic Materials 26
2.8 Electrode/Material Interfaces 29
V
2.8.2 Electrode/Semiconductor Interfaces 29
2.8.3 Electrode/Solid Electrolyte Interfaces 30
2.8.4 Effect Of HeterogeneitiesOf Interfaces On IS 31
2.9 Characterisation Of Structural Defects By IS 37
2.9.2 Zircoinia-Based Ceramics 38
2.9.3 Cementitious Materials 40
2.9.4 Debonding And Pullout Of Fibres In Composites 41
2.10 Summary 41
3 Experimental Procedures 43
3.1 Sample Preparation 43
3.2 Measuring Cell Preparation44
3.2.1 Electrode Preparation 44
3.2.2 Measuring Cell Containing A Dense Ceramic Sample 44
3.2.3 Measuring Cell For MeasurementsOf Green Form Ceramics 45
3.2.4 Measuring Cell For Impedance MeasurementsOf TBCs 46
3.3 Sample Holders 47
3.3.1 The Sample Holder For The Room TemperatureMeasurements 47
3.3.2 High Temperature Rig 47
3.4 Impedance Measurements 48
3.4.1 Impedance Spectroscopy 48
3.4.2 Correction Of Impedance Data At High Frequencies49
3.5 Other Analytical Techniques 51
3.5.1 Density Measurements 51
3.5.2 Thermogravimetric Analyses 51
3.5.3 Dilatometry 52
3.5.4 X-Ray Diffraction 53
3.5.5 Optical And Scanning Electron Microscopy 53
3.5.6 Energy Dispersive Spectroscopy 54
4 Impedance Data Analysis 55

4.1 Introduction 55
vi
4.2 Analytical Approach55
4.3 Numerical Analysis 59
4.4 Summary And Conclusions 64
5 Nondestructive Characterisation of Alumina/Silicon Carbide

Nanocomposites Using IS 65
5.1 Introduction 65
5.2 Experimental 66
5.2.1 Materials 66
5.2.2 Microstructure of Materials 67
5.2.3 Oxidation Treatments 68

5.2.4 Impedance Measurements 68
5.1 Results68
5.1.1 Impedance SpectraOf the Monolithic Alumina (MA) 68
5.1.2 Impedance SpectraOf A1203/SiC Nanocomposites 70

5.1.3 Impedance SpectraOf Oxidised Samples 74
5.1.4 Dielectric Properties At Room Temperature 78
5.2 Discussion 80
5.2.1 Equivalent Circuit Models And IS Parameters 80
5.2.2 Spectral Resolution 86

5.2.3 Dispersion Of Electrical Responses 87
5.2.4 DependenceOf Dielectric ResponseOn SiC Content 89
5.2.5 Nondestructive Detection Of Oxide Layer In A12O3/SiC Nanocomposites90
5.2.6 The Effect Of Particle Orientation On IS 95
6 Nondestructive Characterisation Of Thermal Barrier Coatings Using IS

99
6.1 Introduction 99
6.2 Experimental Procedure 100

6.2.1 Specifications Of Materials 100
6.2.2 Oxidation Treatment 101
6.3 Results 101
vii
6.3.1 Microstructure Of TBCs 101
6.3.2 Impedance SpectraAt 350° C 107
6.3.3 Impedance Spectra At Room Temperature 109

6.4 Discussion 111
6.4.1 Oxidation Kinetics 111
6.4.2 Equivalent Circuit Models 112
6.4.3 Fitting Of Impedance Spectra 116
6.4.4 Interpretation Of The Impedance SpectraAnd Fitting Results 121
6.4.5 Determining The Thickness Of The Oxide Layer 129
7 Characterisation Of Clay Sintering Process Using IS 132

7.1 Introduction 132
7.2 Experimental Procedure 133
7.2.1 Sample Preparation 133
7.3 Results 134
7.3.1 Impedance Spectra During Heating 134
7.3.2 Impedance Spectra During Cooling 140
7.3.3 Characterisation Of Clay Compacts Using Conventional Characterisation
Techniques 142
7.4 Discussion 146
7.4.1 Equivalent Circuit Models 146
7.4.2 Fitted Results 148
7.4.3 Sensitivity Of IS To The Formation Of Liquid Phases 150
7.4.4 Electrode Effects 152
7.4.5 Monitoring Densification Process 156
8 Conclusions And Suggestions For Future work 162

8.1 General Summary And Conclusions 162
8.2 Suggestions for future work 166
References 169
viii
Appendix I Representation Diagrams Of Impedance Spectroscopy 185
Appendix II The Impedance Spectra Of 10SA Samples 187
Appendix III The Impedance Spectra Of Clay/Graphite Composites 188
ix
CHAPTER 1 INTRODUCTION
1 Introduction
1.1 The Need for Non-Destructive Characterisation (NDC) of Ceramics
The need to characterise ceramic materials by non-destructive analytical methods is
growing rapidly, especially for in-process and in-service examination of ceramic

materials and components. This growth is mainly stimulated by the increasedautomation
of ceramic manufacturing processes, the demand for greater reliability in components
and more severedemands on the performance of ceramic materials. The exploration and
development in this field will undoubtedly lead to the improvement of materials quality
and competitiveness, as well as the reliability of componentsin service.
There are many conventional analytical methods available for the investigation of
ceramic materials, e.g. electron probe microanalysis, differential thermal analysis (DTA),
thermogravimetry (TGA), x-ray fluorescence and diffraction, and various types of
spectroscopy (infrared spectroscopy, Raman spectroscopy etc.). Recently some highly

sophisticated instrumentation such as nuclear magnetic resonance (NMR), scanning
tunnel microscopy (STM), atomic force microscopy (AFM), etc. have begun to be
applied to the characterisation of ceramic microstructures. Among these analytical

methods, some are costly and time-consuming, but yield detailed information, others are
relatively inexpensive and can be performed easily and rapidly, but yield limited
information. However, most of them are destructive and typically off-line in nature, and
it is difficult to adapt these measurementsinto the in-process and in-service examination
of materials.
1.2 The State-Of -The-Art Of NDC Techniques For Ceramic Materials
The art and science of non-destructive testing (NDT) is very old. In ancient China, a
pottery tradesmanwould perform a quality inspection on a ceramic article by patting and
listening for the characteristic responsefrom it. This is an effective technique that is still
in practice nowadays in China. This may be regarded as a primitive version of sonic
NDT.
P
CHAPTER I INTRODUCTION
Nowadays a wide range of NDT techniques have been developed for metallic materials
and components. By contrast, the modern non-destructive evaluation (NDE) of ceramic

materials is a relatively young field. So far the choices of techniques, for example, x-ray
and sonic techniques,which have appearedto be effective and efficient for characterising
ceramics in both green and sintered form, and for both in-process and in-service
circumstance, are very limited. In many cases,such NDE instrumentation is too delicate
and too expensive for routine use on the factory floor. Although advances in NDE
equipment have been made in the past decade,which have resulted in major reductions in
equipment costs, on line NDE sensorsare still rarely encountered in high volume ceramic
manufacturing applications at the present time (Schilling and Gray, 1998). Therefore
further investigation and development of robust, inexpensive NDE methods are highly
desirable.
1.3 Impedance Spectroscopy As A NDC Technique
Impedance spectroscopy (IS) has been used for several decades for investigating
electrochemical kinetics. This technique has demonstrated its capability in many

instances. It has been applied in fields ranging from metal or semiconductor
electrode/electrolyte interfaces, electrode/layer/electrolyte systems, to porous electrodes

(Gabrielli and Keddam, 1996). IS has been proven to be a powerful tool for the study of
the electrical properties and microstructure of ionic, electronic, or mixed conductor
ceramics (Lanfredi and Rodrigues, 1999). In the past decade,a number of works revealed
that IS could be used as a non-destructive tool for the detection of structural defects in
ceramics, especially in zirconia-based materials and cementitious materials (Kleitz et al.

1995, McCarter et al. 1988).
Besides being non-destructive, the technique of IS hasthe following advantages:
f IS canbe adaptedto on-line manufacturingor testingenvironments;
f IS does not require special sample preparation: theoretically the measurementscan be
performed on objects of any size and with any complex shape;
f IS equipment is relatively cheap, portable, and can provide quick and economical
measurementswithout slowing production or testing progress;
f Unlike x-ray and 7-ray techniques,IS is non-hazardous.
Page2
CHAPTER I INTRODUCTION
1.4 About This Study
Brittleness appears to be an intrinsic weakness of ceramic materials. The principal

limitations of applying ceramics for structural purpose are brittleness, low reliability and
difficulty in predicting fracture. As a consequence, the main issue concerning the
characterisation of structural ceramics is invariably the detection of various defects, such

as cracks, pores and other forms of inhomogeneities in structure, either in the
manufacturing process or in service.
Previous studies have revealed that IS is capable of detecting structural defects in some
ionically-conducting ceramic materials. However, these studies on zirconia-based-
materials are rather scattered, and far from being systematic and comprehensive. Very
limited success has been actually achieved when structural defects were correlated to
various IS parameters. Meanwhile, to our knowledge, no attempt seems to have been

addressedin applying IS to characterising structural ceramics other than zirconia-based
materials and cementitious materials.
The aim of this study is to investigate the feasibility of applying IS as an NDC method
for a wide range of structural ceramics. For this purpose, a number of different ceramic
materials have been covered, which fall into different electrically conductive categories,
e.g. insulator, semiconductor and insulator/semiconductor composites as well as
ionically-conducting materials. The circumstances of both in-service and in-process
examination of ceramic materials have been considered. Attempts have also been made
to establish quantitative as well as qualitative relations correlating microstructural
inhomogeneities in the materials to their electrical properties.
This thesis is concernedwith three representativematerial categories as following:
A1203/SiC nano-composites, a typical example of insulator/electronic conductor
composites and representative of a series of SiC/carbon fibre reinforced ceramics, with

the conductive phase being dispersed in an insulating ceramic matrix. The electrical
conduction behaviour ranges from being insulating to semi-conducting, depending on the
content of the conductive disperse phase.
P_agee3
CHAPTER I INTROD TION
Thermal barrier coatings (TBCs), consisting of yttria stabilised zirconia (YSZ) as the top
coating and metal alloys as the bond coating and substrate. After oxidation, a semi-
conducting oxide layer is formed. TBCs are a multi-layer solid-
electrolyte/semiconductor/metal system
A clay compact during sintering, a typical glassy phase/insulator composite is
representativeof the category of clay-based ceramics, such as white ware and porcelains
etc.. The glassy phase exhibits considerable alkali ion conductivity at high temperatures,
while the porosity and other solid particles, such as undissolved silica, are insulating in
nature.
The experimental results and theoretical analysis conducted in this work demonstrated
that:
" For the insulator/electronic conductor composites, the content and orientation of
the conductive phase are closely related to the IS parameters deduced from
impedance spectra. The oxidation scale formed at the surface can be determined
non-destructively by IS, a task that is very difficult for other techniques, such as
SEM, weight gain measurement,optical microscopical observation etc. to fulfil.
" In TBCs, the electrical responsefrom the oxide layer is well separatedfrom those
from other phases.The thickness of the oxide layer formed at the interface of the
top coat/ bondcoatis reflectedin the complexmodulusplots.Not only bondcoat

oxidation,but alsotop coatdegradationcanbe monitoredusingIS.
" During the sintering of clay-based ceramics, the formation of liquid phase and the
shrinkage rate of the ceramic body can be monitored by in-situ impedance

measurements.The electrode effect in the impedance spectra is representative of
the local temperature, which can be an effective indicator of the firing of clay-
based ceramics.
Therefore it is fair to say that IS is a powerful tool for the non-destructive
characterisation of structural ceramics.
Page 4
2 Literature Review
2.1 Introduction
Non-destructive testing (NDT) can be defined as "the application of physical principles

for detecting inhomogeneities in materials without impairing the usefulness of the
material" (McGonnagle, 1971). The primary purpose of NDT is to determine the existing
state or quality of a material, with a view to acceptanceor rejection (McGonnagle, 1971).
Similarly, non-destructive evaluation (NDE) or non-destructive characterisation (NDC)
refers to applying non-destructive techniques in the evaluation or characterisation of

materials or components.
The ceramic industries have already recognised the need for increased monitoring of
their manufacturing processes to measure microstructural evolution and detect the

presence of defects (Wadley and Mehrabian, 1984). The interest in materials' property-
determination by non-destructive techniques is increasing, especially for in-process and
in-service inspection of structural and electronic materials and components (Ruud, 1984).
Various non-destructive techniques have been employed in NDE for suspensions,
composites and films as well as green bodies and sintered ceramics. These techniques
include ultrasonic techniques (Komarenko et al. 1994; Challis and Tebbutt, 1998; Ahn,
1998; Kathrina and Rawlings, 1996; Boccaccini and Boccaccini, 1997), x-ray techniques
(Huss et al. 1998), optical methods (Goller and Pye, 1998) and sonic detection (Pape,
1998), etc.
No matter what type of NDT technique it is, a certain form of energy, e.g. ultrasonic
waves, optical light or x-radiation, etc. should invariably be utilised and some material
properties provide the base for detecting the inhomogeneities. For example, the basic
principle utilised in thermography is to apply heat to the test specimen and measure or
observe the resulting temperature distribution, by using a video-thermographic camera
(Curtis et al. 1993), or simply using surface-mounted thermal couples. Flaws alter the
temperature distribution on or in the specimen. In the radiography method, x-ray or
gamma radiation is used, and the radiated intensity from the sample is determined by the
Pages
CHAPTER 11 LITERATURE REVIEW
thickness, density and defects of the materials, therefore it can generate shadows on
photographic films (Nevadunsky and Lucas, 1975). In the ultrasonic technique,

ultrasonic impulses generated by a piezoelectric transducer are used. The echo signal
resulting from the reflection or scattering from the object is detected by the same
piezoelectric device or another receiver. The phase delay between the reflection from
different surfaces is correlated to the material properties (Dunyak et al. 1992).
In impedance spectroscopy, the energy applied is the electric field with varying
frequency, and the electrical responseduring measurementis determined by the electrical
properties of the materials under investigation. There are three variables, the real and
imaginary components of the electrical response and the frequency in impedance
measurement. A change in the ceramic microstructure and inhomogeneities leads to a

change of electrical properties of materials and thereby may cause a change of the
impedance spectrum. The electrical response from different phases, e.g., grain and grain
boundary, and different physical processes, for example, the electrode/materials
interfacial reaction, are normally reflected in different frequency domains. For instance,
in YSZ (yttria stabilised zirconia) and BaTiO3i the relaxation frequency for the grain
interior is always in a high frequency range, and the electrode effect is in a low frequency
range, with the response from the grain boundary phase being in an intermediate
frequency range. Therefore the electrical response from different phases can be
separated in the impedance spectra and the microstructure and properties of ceramic
materials can be correlated to the IS parameters. This constitutes the fundamental
attribute of using IS for characterising ceramics.
However, for our first impression, most ceramic materials, especially structural ceramics,
appear to be insulators, or at least have very poor electrical conductance. As a matter of

fact, most oxide ceramics are semiconductors, among which only a few examples are
listed as following:
n-type (metal-excess) semiconductors: BeO, WO3 CaO, ZnO, Ti02, MgAl2O4 etc.;
p-type (metal-deficit) semiconductors: Cr203, MgCr2O4, MnO, NiO, CoO, PbO,

etc.(Kubaschewski and Hopkins, 1962).
Page6
CHAPTERII LITERATUREREVIEW
Partially or fully stabilised zirconia ceramics are not only good ionic conductors, but also
typical phase-transformation-toughed structural materials that have many technical
applications. All the alkali glassesare alkali ion conductors. It should also be noted that
the conductivity enhancement of a semiconductor can be achieved by raising the
temperature and/or doping with a small amount of impurities in the materials. Many
ceramic materials that are insulators at room temperature may show ionic conduction at
high temperature (Kingery et al. 1976c). In this sense, most ceramics can be made
deliberately, at least partially, conductive to meet the requirement of impedance
measurements.Meanwhile, it is important to note that many ceramic composites contain

carbon and silicon carbide phases which exhibit significant electrical conductance at
room temperature (Zhang et al. 1992). This is why electrical resistance measurements
can be used to monitor the fracture (Wang et al. 1998) and detect the fatigue in ceramic
composites (Ishida and Miyayama, 1994). Therefore IS appearsto be applicable to the
characterisation of most ceramics, particularly those belonging to either ionic conductor,
semiconductor, or conductor/insulator composites.
2.2 Impedance And Impedance Spectroscopy
2.2.1 Impedance Measurement
IS measurementis usually made with cells having two identical electrodes applied to the
faces of a sample. There are several different electrical stimuli which are used in IS, but
the most common and standard one is to measure impedance directly in the frequency
domain by applying a single-frequency voltage to the interface and measuring the phase
shift and amplitude. The small amplitude of electrical stimuli ensuresthat there is a linear
relationship between the current and potential, and that the response is in a steady state;
that is, it does not change with time and many measurementscan be averaged. A wide
range of frequencies, usually over several orders of magnitude, is applied so that

processes having different time constants can be detected within one experiment
(Macdonald and Johnson, 1987).
P
CHAPTER 11 LITERA TURE REVIEW
2.2.2 Concept of Impedance
The concept of electrical impedance was first introduced by Oliver Heaviside in the
1880s and was soon after developed in terms of vector diagrams and complex
representation by A. E. Kennelly and especially C. P. Steinmetz. Impedance is a more

general concept than resistance because it takes phase differences into account
(Macdonald and Johnson, 1987).
In order to understand the basis of IS, it is helpful to consider some simple electronic
analogueswhich mimic closely the systems of interest (Christensen and Hamnett, 1994).
If we consider first a simple resistor and apply a steady, small sinusoidal potential,
Vsin(a ), across a resistor that has a resistance R, then the current through the resistor is
determined by:
V sin( wt)
(2- 1)
RR
When the same signal is applied across a capacitor that has a capacitance C, then the
current through the capacitor is given by:
dQ dV d (V sin( cot))
_ C_ C C coV sin( wt +/ 2) (2-2)
dt dt dt
ca
z
C
C
0Z Real Part Impedance
Figure 2-1 The impedance Z plotted as a planar vector using rectangular and polar
coordinates
Now we can see the current response from a pure resistor is in the same phase as the
applied potential signal, while the current response from a capacitor always leads the
Page8
CHAPTER II LITERATURE REVIEW
potential signal by 90°. It is not difficult to conceive that the current response from any
circuit that is composed of both a resistor and a capacitor would have a phase difference
with the applied sinusoidal potential signal. The phase difference is in between 0-90°,
and the impedance can be expressed in terms of in-phase component and out-of-phase
component. For this reason, impedance is usually expressed by a vector quantity,
Z(w)=Z'+jZ", where Z' is the in-phase component or real part of the complex, Z" is the
out-of-phase component or imaginary part of the complex, and j= -exp(jir/2).

Impedance can be plotted with either rectangular or polar coordinates (Macdonald and
Johnson, 1987), as shown in Fig.2-1. Here the two rectangular coordinate values are:
Z'=1 Z1cos(8) and Z"= (Z 1sin(6) (2-3)
With the phaseangle
e=tan'1(Z"/Z, ) (2- 4)
andthemodulus
IZI =[(Z,)2+(Z,,)2]"2 (2- 5)
2.2.3 Various Formalisms Of IS
From Eqs. 2-3-2-5, we can understand that the data from impedance measurementscan
be expressedin a number of different ways. Plotting impedance data in a complex plane
with real and imaginary coordinates gives a complex plane plot. Although complex
impedance plane plots are frequently called "Nyquist diagrams", this is thought to be a
misnomer by J. R. Macdonald (Macdonald and Johnson, 1987) and this term should be
IZI
avoided. We also can plot the total impedance and phase angle 0 as a function of
frequency w. The plots of log IZI vs. log w and 0 vs. log COare called Bode plots.
There are three other possible complex formalisms which can be derived from complex
impedance, the electric modulus M*, the admittance Y* (or A*), and the permittivity s*
and these quantities are interrelated (Hodge et al. 1976):
M* jwCoZ*=M'+iM" (2-6)
Page9
E*= (M*)"' (2- 7)
Y*=(Z*)-i (2- 8)
Y*= jwCos* (2- 9)
Where w is the angular frequency 21rf, Co is the vacuum capacitance of the empty
measuring cell, Co=eo/k, eo is the permittivity of free space (8.854x10-14 F/cm), and
k=1/A, the cell constant where 1is the thickness and A the area (Sinclair and West, 1989).
Among these four formalisms, the admittance Y*, the impedance Z* and the permittivity
s* have normal meanings as defined in many textbooks, while the concept of electrical
modulus M* is relatively new. It can be better understood by comparing it with the shear
modulus in mechanical relaxation. The vanishing of M" at low frequencies implies the
lack of a restoring force for flow of charge in a conducting dielectric under the influence
of a steady electric field (Provenzano et al. 1972). The electrical modulus (M) and the
complex permittivity (c*) bear the same relationship to one another as do the modulus
and the compliance in mechanical relaxation (Ambrus, et al. 1972).
If these different quantities are plotted in complex planes, then various complex plane
diagrams (Z" vs. Z', Y" vs. Y', M" vs. M' and s"vs. s') can be obtained. If they are
plotted as a function of frequency in a way similar to a Bode plot, then so called

`spectroscopic displays' (Y", Z" and M" etc. vs. logio(o) are obtained (Hodge et al.
1976). These different representation diagrams are illustrated in Appendix I. They are
sometimes called admittance spectrum (Y" vs. Y'), modulus spectrum (M" vs. M') and
dielectric permittivity spectrum (£"vs. E') etc., but impedance spectroscopy is a general
term that is usedto include and denote all these spectra(Macdonald and Johnson, 1987).
2.3 General Characterisation Procedure Using IS
IS involves a relatively simple electrical measurementthat can readily be automated and
whose results may often be correlated with many complex materials variables: from mass
transport in corrosion to defects and microstructure of solids (Macdonald and Johnson,
1987).
The essenceof the impedancetechniqueis that thecell is alwaysmimickedby a suitable

model system, in which the properties of the materials and the electrode/materials
by appropriateelectricalanalogues.The impedanceof the cell is
interfaceare represented
Page10
then measuredover a wide frequencyrange(ChristensenandHamnett,1994).Themodel

parameters are obtained by determining parametric values which give the best fit to the
impedance data. Finally the model must be correlated with the physical system to
establish the reliability of the model and to establish that the model values determined
from the fit are physically reasonable.If not, the model may have to be modified and the
entire analysis process repeated. Clearly the required analysis is not always
straightforward and is usually quite involved (Johnson and Worrell, 1987).
A flow diagram of a general characterisation procedure using IS is proposed by

Macdonald and Johnson (1987), as in Fig.2-2.
MATERIAL-
ELECTRODE
Is
EXPERIMENT
THEORY
EQUIVALENT
CIRCUIT
-V,
I PLAU SIBLE
MODEL
MATHEMATICAL CURVE FITTING

MODEL
SYSTEM
CHARACTERISATION
Figure 2-2 Flow diagram for the measurement and characterisation of a material-electrode
system.
P
It has been realised that the simulation of impedance experiments on its own generally
cannot provide a definite analysis of a system, since there is frequently more than one
equivalent circuit that fits, numerically, a given experimental data set (Sinclair and West,
1989). Therefore a set of impedance data could not be properly understood and
interpreted before we have a proper physical model for the particular system. An
appropriate model is always built basedon adequateknowledge of the particular system,

among which are the type of charge carriers, the migration path of charge carriers, the
charge transfer and masstransfer on the electrode/materials interface etc.
2.4 Electrical Conduction In Ceramics Materials
2.4.1 Introduction
The electrical conductivity associatedwith ceramics is, in most cases,semi-conductive or

ionic. To understand correctly the electrical conductance mechanism, it is necessaryto
establish an equivalent circuit model for the ceramic and thereby interpret the IS
properly.
2.4.2 Electrical Conduction In Semiconductors
The electrical conductance of a semiconductor can be understood in terms of the electron
energy band structure. As shown in Fig. 2-3, the electronic band structures of
semiconductor and insulators are fundamentally different from those of metals because
of the existence of the band gap which lies between a filled valence band and an empty
conduction band (Jiles, 1994). For metals there is always a finite concentration of
electrons in the conduction band, so metals always display high electronic conductivity.
For insulators, the band gap is sufficiently high so that normally there are no electrons
able to move through the crystal. For semiconductors, the concentration of electrons in
the conduction band depends on both temperature and composition (Kingery et al.
1976c).
Electronic conductivityin semiconductorsis due to the presenceof donor- or acceptor-

type point defects,which at room temperatureare fully ionised,giving rise to delocalised
electrons in the conduction band (n-type) and holes in the valence band (p-type)
Paw 12
CHAPTER-II LITERAT RE REVIEW
--
respectively. A typical order of magnitude for the concentration of donors or acceptors,

and hence of electronic free charge carriers, is between 1016and 1018cm-3, i.e., about six
orders of magnitude lower than in metals (Gomes and Vanmaekelbergh, 1996).
Conduction
Conduction
Energy Conduction Band
Band
Band
E, Band Gap Band Gap
Occupied Occupied
Metal Semiconductor Insulator
Figure 2-3 Simplified band structure diagrams of a metal, semiconductor and insulator.
Typical values of the band gap: 0eV in metals, 0.5-5eV in semiconductors and 5.0eV or
greater in insulators.
The conductivity of a semiconductor increases exponentially with temperature (Jiles,

1994):
E
a; ccexp(- (2- 10)
kT)
where Eg is the band gap and k is the Boltzmman constant.
For ceramics having low electrical conductivity (lower than 10 S"cm-1), even minor
variations in composition, impurity content, heat treatment, stoichiometry and other

variables can have a significant effect on electrical conductance. This is because the
impurity atoms introduce new localised energy levels for electrons intermediate between
the valence band and the conduction band. If the new energy levels are unoccupied and
lie close to the energy of the top of the valence band, it is easy to excite electrons out of
the filled band into these new acceptor levels (Kingery et al. 1976c).
2.4.3 Ionic Conduction In Crystalline Materials
One kind of charge carrier that is always present and can contribute to electrical
conductivity is the ions present in crystalline materials. Electrical conductivity resulting

from ion migration is important in many ceramic materials especially at high
Page13
temperatures (Kingery et al. 1976c). Among the oxide ionic conductors, the most well-
studied and useful materials to date are those with the fluorite-related structure,
especially ones based on Zr02, Th02, Ce02 and Bi203. The most common of these is the
Zr02 family, in which the mobile species is the oxide ion (Huggins, 1978). Pure Zr02 is
cubic only above -2400°C. 8 mol % of Y203 is needed to stabilise the cubic form to
room temperature, as in yttria stabilised zirconia (West, 1995).
For an ion to move through the lattice under the driving force of an electric field, it must
have sufficient thermal energy to pass over an energy barrier that corresponds to the
intermediate position between lattice sites. Therefore, similarly to the case of
semiconductors, the ionic conduction as a function of temperature can be expressedas
(Kingery et al. 1976c):
6; a exp(- (2- 11)

kT)
where OG is the potential barrier.
However it is important to note that some materials may exhibit mixed ionic-electronic
conductivity. For example, undoped nano-Ce02 electrolyte is an ionic conductor, but also
exhibits enhanced electronic conductivity (Chiang et al. 1997); bismuth oxide-based
materials are very good oxygen-ion conductors, but they develop electronic conductivity
in low oxygen partial pressures (Shuk and Wiemhofer, 1996; Badwal et al. 1998).
Meanwhile some semiconductors also exhibit ionic conductivity, for example, SrTiO3
exhibits both electronic and ionic conduction (Choi and Tuller, 1988). Titanium dioxide
exhibits either n or p type conductivity in addition to ionic conductivity, depending on
oxygen partial pressure (Tuller, 2000). Both single crystalline (Nowotny et al. 1998) and
nanocrystalline titania (Knauth and Tuller, 1999) have an ionic conductivity of about
2.5x10-5 S"cm' at 1000K, at high Poe etc.
,
2.4.4 Ionic Conduction In Glass And Molten States
The two most important constituents present in all glasses are network formers and
network modifiers. Network formers are compounds of a covalent nature such as Si02,
P203,B203,GeS2,P2S,Ge02 etc.,in which all the oxygenatomsare covalentlyboundto
Paze 14
the network former cations in tetrahedral or triangular elementary units. The
macromolecules are formed by an assembly of these units in which at least one atom of
oxygen, called a bridging atom, is shared (Ravaine and Souquet, 1978). Network
modifiers, such as Li20, Na20, K20 and Ag20 etc., can interact strongly with the
structure of network formers, leading to the breaking of the oxygen bridging linking two
network former cations. The addition of a modifier introduces ionic bonds, giving rise to
cationic conduction. For instance,in a silicate glass containing K20, the reaction between
silica and potassium oxide may be expressed schematically as in Fig. 2-4 (Souquet,
1995). In this way the charge carriers K+ are created.
-i-O-Si-+K20
K+
i-O- "O-Si-
K+
Figure 2-4 Schematic of reaction between silica and potassium oxide in silicate glass.
At the same time, some oxygen atoms that are not bridging can be negative charge
carriers and remain in the vicinity of alkali cations or alkaline earth cations of the
network modifier oxides (Ravaine and Souquet, 1978). According to the polarisation
study of glasses(Doi, 1987, Carlson et al. 1972 & 1974), it is clear that some oxygen-ion
conduction is possible by movement of non-bridging oxygen ions in the highly polarised
field near the anode. Normally alkali ions are the major charge carriers, while the
conductivity in the alkali-depleted-region is via non-bridging oxygen ions. The oxygen-

ion conduction in some glasses has been thought to be six or seven orders of magnitude
lower than that for alkali ion conduction (Badwal, 1995).
For a given silicate composition, little or no difference is seen between the structure of
the glassy and molten states as deduced from the diffraction and spectroscopy
measurements: i. e., the inter-ionic distances and oxygen co-ordination numbers for
cations are essentially the same for the glassy and molten states (Turdogan, 1983).
Therefore the electrical conduction mechanism the
of molten statescan be assumedto be
similar to that of glass. However, it is important to note that charge carriers in the liquid
are expected to be more mobile than in the solid state, for this reason the liquid is a much
Pa,ae 15
CHAPTERH LITERATUREREVIEW
better ionic conductor than the solid (Bockris et al. 1952). Above the glass transition
temperature (Tg), the ions move partly in responseto the viscous flow of the surrounding
medium; below Tg the mode of transport may involve the discrete hopping of ions past
effectively frozen barriers presented by the host glass matrix (Sidebottom et al. 1995).
An Arrhenius temperature dependence of electrical conductivity normally shows a
discontinuity at the transition range corresponding to the freezing of the glass structure at
this temperature, and the activation energy below Tg is normally substantially smaller
(Sidebottom et al. 1995).
The temperature dependenceof glass conductivity over a considerable temperature range
can be expressedas:
E
6=6,, exp(- (2- 12)
RT)
where E is an experimental activation energy for conductivity.
In terms of Eqs. 2-10,2-11,2-12, we can see that whatever the conduction mechanism,
electronic,ionic or mixed conduction,the electricalconductivityof ceramicmaterials(a)

is thermally activated and can be expressedby an Arrhenius law:
Ea
(2- 13)
- a-(,exp(_ RT
The activationenergyfor conduction(Ea)can be calculatedfrom the slopeof the straight

line given by logo vs. 1/T.
2.4.5 Ionic Conduction on Particle Surfaces
A typical example of the ionic conduction along a particle surface is the `surface proton
hopping' in ionic-type humidity sensors. Quite a few oxide ceramics demonstrate a
change in their electrical properties because of water-adsorption processes on their

surface, which permit their use as humidity sensors (Traversa, 1995). As dry oxides are
kept in contact with humid air, water molecules chemisorb to the available sites on the
oxide surface, to form hydroxyl ions. When the first layer of water molecules is formed,
subsequent layers of water molecules are physically adsorbed on to the first hydroxyl
Paffe16
CHAPTER II LITERATURE_R E
layer (McCafferty and Zettlemoyer, 1971). When only hydroxyl ions are present on the
oxide surface, the charge carriers are protons, which hop between adjacent hydroxyl
groups. When the first layer of water is continuous, charge transfer is governed by proton
hopping between neighbouring water molecules in the continuous film (Fleming, 1981).
Ionic conduction along particle surfaces is an important factor that needs to be
considered, especially in understanding the impedance spectra of very porous ceramic

materials. Frequently, because of surface conduction, the well-known phenomenon of
two-bulk-semicircles, corresponding to grain and grain boundary respectively, is not
present in the impedancespectrum.
2.5 Physical Model And Equivalent Circuit
2.5.1 Brick Layer Model
a) b
ciccuinics
Figure 2-5 (a) schematic of a real ceramic microstructure. (b) Simplified brick-layer model
and equivalent electrical network consisting of different RC elements. Indices:
el=electrode/ceramic interface, B=bulk, GB=grain boundary channels perpendicular to the
electrode interface (after Waser and Hagenbeck, 2000).
Pa e17
For a dense ceramic material, an ideally simplified microstructural model is the so-called
brick layer model (or brick-wall model), as shown in Fig. 2-5, where the ceramic is
assumed to consist of identical cubic grains separated by a continuous grain boundary
phase (Waser and Hagenbeck, 2000).
In a broad range of electroceramic materials which show ionic conduction, mixed ionic-
electronic conduction or pure electronic conduction, grain boundaries often act as

barriers to the transport of charge carriers (Waser and Hagenbeck, 2000), therefore
resistive layers at the grain boundary interfaces are formed. For example, at 500°C the
conductivity of grain boundaries in CaO-stabilised zirconia is at least two orders of

magnitude lower than that of the grains (Aoki et al. 1996). In such cases,the electrical
conduction along the grain boundary channels perpendicular to the electrode is very
small and assumed to be negligible, and therefore the equivalent circuit for the brick
layer model can be further simplified as shown in Fig. 2-6.
Rt RB RGB
Cel CB CGB
Figure 2-6 Simplified equivalent circuit of the brick layer model.
However the basic assumptions in the brick layer model, such as homogeneous grain
boundaries, cubic grains, identical grain boundary properties etc. are usually violated in
real ceramics. The influence of some deviation from the brick layer model on the
impedance of polycrystals have been investigated and calculated by a finite element
method (Fleig and Maier, 1998). It was found that the brick layer is a reasonable model,
as long as the sample has a spatially homogeneousdistribution of grain sizes, an isotropic
distribution of grain orientations, isotropic conductivities, a relatively narrow grain size
distribution and uniform grain boundary properties. The brick layer model fails when
there exist imperfect contacts (lateral inhomogeneities) between grains, or grain
Paze 18
boundary properties demonstrate a very broad or bimodal distribution (Fleig and Maier
1999a and Fleig, 2000).
2.5.2 Constriction Model
The constriction model was first proposed by Bauerle (Bauerle, 1969) for characterising
the grain and grain boundary effects in stabilised zirconia. According to this model
oxygen-ions are partially blocked at the grain boundaries by the segregatedphaseswhich

partially cover the grain boundaries. In their study of the conductivity of a Li superionic
conductor, Bruce and West (1983) also pointed out that the transport of charge carriers
takes place through a constriction mechanism as some sections of the diffusion paths are
blocked by the poorly conducting phases, such as glassy phase, voids, and any other
second insulating phases.
T
a)
Path a
Path b
RB
b) C)
RG
R(, Patha R1
ýý
ca CG CB
cc
Figure 2-7 (a) Schematic of the constriction model, with the dark areas representing
insulating phase. (b) Equivalent circuit of the constriction model. (c) Simplified equivalent
circuit of the constriction. Indices: G= grain, B= grain boundary, C= insulating phase.
(Note: the electrode effect is not included in this diagram).
Based on the constriction model, the grain boundary resistivity is actually the
constriction resistance at grain contacts, rather than that due to the presence of a
Pave 19
continuous grain boundary phase as in the brick layer model. As shown in Fig. 2-7a, b, in
such a case, the electrical conduction across the grain boundary actually consists of two
conduction paths, one crossing the insulating phase, another through the direct contact of
two grains. The equivalent circuit can be simplified as in Fig. 2-7c, if the conduction path
crossing the insulating phase is neglected.
Therefore the equivalent circuit for the constriction model is actually the same as that for
brick layer model. For this reason the constriction model sometimes is also called the
brick layer model. However, because the constriction model is established on a different
physical basis, the interpretation of the IS from it should be different from that for a brick
layer model with a continuous grain boundary phase.
According to the finite element calculation conducted by Fleig and Maier (1999b), the
grain boundary capacitanceobtained from a fit using two RC elements in series is close
to the geometric capacitanceof insulating inclusions, as long as the fraction of contacted
area is <25%, while the grain boundary resistance (Rgb) obtained from fitting is directly
proportional to the bulk resistance (Fleig and Maier, 1999b). This strongly supports the
validity of the constriction model. Based on the constriction model, the resistance of the
grain boundary is a consequenceof the constriction of bulk conduction. For this reason,
the temperature dependenceof the grain boundary resistance should be similar to that of
the bulk resistance, and the grain-size dependence of the grain-boundary resistance
should deviate from the inversely linear relationship suggestedby the brick layer model
(Fleig and Maier, 1999b). This is consistent with a number of experimental observations:
for example, the conductivity of grain boundaries in Li superionic conductor has a
comparable activation energy to that of the grain interiors (Bruce and West, 1983), and
the grain boundaries in un-doped BaTiO3 have a ferroelectric structure similar to the
grains (Hirose and West, 1996) etc.
Badwal (1995) and Dessemond et al. (1993) suggested that the constriction model is
more suitable for the study of the ZrO2-basedmaterials. It was found that the influence
on the IS of voids and secondphasesis similar to that of grain boundaries(Badwal,
" Please note that in order to avoid unnecessary repetitions, the electrode effect is not
included here, just for the sake of simplicity of discussion. For the detailed discussion on
electrode effect, see section 2.8.
Page20
CHAPTER II LITERA TURE REVIEW
1995). The blocking effect of grain boundaries is invariably found to disappear at high
temperature, which can only be explained by using the constriction model. According to
the brick layer model, a continuous grain boundary layer is assumedto be present in the
microstructure. Therefore the activation energy of the grain boundaries is expected to be

different from that of grains and the blocking effect of the grain boundaries should
always exist whatever the temperature. This has frequently been found not to be the case
according to many reports (Badwal, 1995; Dessemond et al. 1993; Bruce and West,
1983).
2.5.3 Multi-layer Model
b) RB Ri
a)
is
C) d)
Cg
RB
cl
II
Figure 2-8 (a) Series layer model. (b) Equivalent circuit of the series layer model. (c)
Parallel layer model. (d) Equivalent circuit for the parallel layer model. Indices: B= bulk
phase, I= interface between two layers.
This model is suitable for representing multi-layer materials and materials with well
aligned columnar grains. Typical cases are the series layer model in which layers are
arranged parallel to the electrodes, and the parallel layer model in which layers are
aligned perpendicular to the electrodes (Fig. 2-8 a-d). Two semicircles should appear in
Paize21
the complex impedance plot for the series layer model. By contrast there should be no
separatesemicircle for different phasesin the parallel layer model.
Nitrogen-stabilized-zirconia has a microstructure consisting of well aligned columnar
grains, for which the grain boundaries and bulk grains have been successfully
characterised using a multi-layer model by taking impedance measurementsalong and
perpendicular to the orientation of the columnar grains (Lee et al. 2000).
2.5.4 Continuous Conductive Path Model
Zicp:impedanceof the ìnsulating' conductive

paths
ZDCP:impedanceof the discontinuous
conductive paths
Zccp: impedanceof the continuous conductive
paths
Figure 2-9 Extracted electrical conduction model for concrete.
All the above-mentioned models are basedon the assumption that the grain boundary is a
poorly conducting phaseand that there are no continuous conductive paths present in the
microstructure. In the real world, however, there are many materials in which continuous
conductive paths may exist. One of the well-studied cases is cementitious materials, in
which there are basically three kinds of conduction paths which can be simplified as
shown in Fig. 2-9 (Song, 2000). Cement paste is a heterogeneousmixture of solid, liquid
and gaseous components (Macphee et al. 1997). The continuous conductive paths are
actually created by the continuous networks of pores filled with liquid that is not pure
water but a solution of moderate ionic strength and therefore capable of conducting
electricity (Christensen et al. 1994), while the un-reacted cement and entrapped air is
insulating, and the blocked liquid-filled pores are discontinuous conductive paths
(Macphee et al. 1997).
Page22
Some electroceramics, e.g., Ni-doped SrTiO3, may exhibit high grain boundary
conductivity under certain circumstances, and continuous conductive paths are thus
formed. In this case,the equivalent circuit can be expressedschematically as in Fig. 2-10.
Another example is ionic-type humidity sensors. In this type of device, the conductance
is created by proton hopping along particle surfaces (Fleming, 1981), and the porous
nature and high surface area of these materials provide an easy access of humidity and
the conductive paths are obviously continuous throughout the material.
RB RGB Rinv
1 ýj 1
CB CGB Cinv
11
1_1
L
----------- - -----------
Rlong
Figure 2-10 Schematic of the equivalent network describing the small-signal properties of a
ceramic taking conduction along the grain boundaries into account. The index ìnv' hints at
the RC-network elements of the inversion layer, the index `long' describes the longitudinal
GB resistance (after Waser and Hagenbeck, 2000).
Complications arise when there are short-circuit pathways present in materials. The
typical feature of the continuous path model is that normally there is only one semicircle
corresponding to the bulk effect, because in most casesthe continuous conductive paths
will dominate the electrical conduction and thus shield the effects from other elements in
the circuit. For this reason, care needs to be taken when interpreting the impedance
spectra, where electrical responsesfrom grain and grain boundary are not resolved so that
it is very difficult to distinguish the different effects from different phases.
Page23
CHAPTER II
2.6 Representations Of Impedance Spectroscopy
The complex impedance plane plot and the Bode plot appear to be the most commonly
used representations of IS. A semicircular arc in a complex plane plot normally

representsa particular physical /chemical process in the bulk or at the electrode/material
interface, while the Bode plot explicitly shows the dependence of the total impedance
and phase angle on the frequency. In the complex plane plot, detailed information, such
as relaxation frequency, semicircular diameters and depressedangles etc., sometimes can
be easily perceived.
However, as has been mentioned in section 2.2, there are four different immitance
functions related with IS and hence there are many different ways of representing
impedance spectra, with each one highlighting different features (Hodge et al. 1976).
Therefore, in order to have a deep understanding of IS and extract more useful, more
precise information from impedance measurement data and the proper representationsof
this data are important.
Bauerle (1969) demonstrated the usefulness of the complex admittance approach in solid
electrolyte studies, while Armstrong et al. (1974) have favoured the complex impedance
formalism for analysing data on ß-alumina and RbAg4I5. Macedo et al. (Macedo et al.
1972, and Ambrus et al. 1972) were among the first to exploit the modulus and used it
for analysing conductivity relaxation in glassesand in concentrated aqueoussolutions.
Owing to its simplicity, the electrical modulus formalism has found wide acceptance
among many researchers. This formalism suppresses the undesirable experimental

artifacts associatedwith electrode polarisation, leaving a clean data presentation (Macedo
et al. 1972). Hodge et al. (1975) extended the modulus treatment to the study of solid
electrolyte systems. It had been shown that a combined analysis using both the complex
impedance and the complex modulus formalisms have certain advantages(Hodge et al.
1976; Sinclair and West, 1989).
The Debye peak in spectroscopic plots of the imaginary components, Z" and M" versus
log w can be described by (Morrison et al. 2001):
Z"= R(o)RC ) (2- 14)

1+ (o)RC)2
Page24
CHAPTER II LITERATIURE REVIEW
E
ml_ ( wRC ) (2- 15)
C 1+(wRC)2
where co is the angular frequency and Eothe permittivity of free space. The frequency at
the Debye maxima for each RC element is given by
)max 27tfmax (RC)'= ti -1 (2- 16)
where ti is the time constant of the RC element, fn. is the relaxation frequency. Both are
independent of the geometry of the sample and are intrinsic properties of the RC element.
The electrical relaxation under consideration here is associated with the long-range
conductivity processes. To distinguish it from dielectric relaxation involving dipole

reorientation, an electrical relaxation of this sort is called conductivity relaxation
(Provenzano et al. 1972).
Themagnitudesof Debyepeaksat the peakmaximaaregiven by
Z"momR/2 (2- 17)
M"mom sc/2C (2-18)
Hence the magnitude of R and C can be estimated from either Z",,, ax or M",,, using
Eqs. 2-17,2-18.
From Eqs. 2-14,2-15, we can understand that the power of combined usage of
impedance and modulus spectroscopyis that the Z" plot highlights phenomena with the
largest resistance whereas M" picks out those with smallest capacitances (Sinclair and
West, 1994).
Badwal (1988) pointed out that the complex impedance plot is useful when the relaxation
times of various processesdiffer as a consequence of different capacitive components,
while the modulus or permittivity plane plots are more suitable for a system where the
relaxation times differ due to different resistive components.
As has been already mentioned in the previous section, the most frequently used physical
model is the ideal brick layer model (Fig. 2-5), for which, according to Abrantes et al.
(2000b),thereal and imaginarypartsof the impedance(Z*=Z' jZ") aredescribedby
Z'=RB/[l+(w/wB)2 ]+Rgb/[1+(c)/(Ogb)2]+ReI/[l+((O/c)el)2] (2- 19)
Z"=RB(w/wB /[1+(O)/o B)2 ]+Rgb(w/(Ogb) /[1+(w/(Ogb)2]+
Rel((o/O)ei)/[ 1+(O)/U)eI) 2]
(2- 20)
Page25
and the imaginary part (M") of modulus (M*=M'+jM")
M'' (CO/CB)(()/(Dg)/[I+((OI(OB)2]+(CO/Cgb)(O)/(t)gb) /[1+((D/Q)gb)2]+
(Co/Cei)(CO/(Oei)/[1+(c)/(wei)2] (2- 21)
where Co is the vacuum capacitanceof the cell.
Tangent loss is related to the real and imaginary parts of impedance:
tan(8) =2 (2- 22)
By plotting tg (S) vs. log f, we will have peaks at frequencies fp1,(fgbfb)1/2(Rgb/Rb)1rz,

indicating an approximate transition from the bulk to grain boundary contributions, and
indicating a transition from grain boundary to the electrode
fp2ý4fgbfei)1/2(Rgb/Rei)"1/2
contributions, respectively. Meanwhile we can also have minima at

fmin2=fel[(RB+Rgb+Rei)/(RB+Rgb)]1'2
fminI=fgb[(RB+Rgb)/RB]'1Z, respectively (Abrantes et al.
2000c). For this reason the alternative representation of log (tg(8)) vs. log(f) is believed
to be more useful for an overall analysis of impedance spectra and to obtain a wide range
of relevant parameters, even if the bulk and grain boundary arcs overlap in the complex
impedance plot. In the plot of log (tg(8)) vs. log(f), the peaks indicate electrode to grain
boundary transition, and/or the grain boundary to bulk transition. Local minima can be
used to extract the grain boundary or the electrode relaxation frequencies (Abrantes et al.
2000c).
Abrantes et al. (2000a) also suggestedthat the plots of Z' vs. (Z"/(O), or Z' vs. (Z"/f)
should be suitable to extract relaxation frequencies and resistance values relevant to the
bulk, grain boundary, and electrode effects, even for caseswhen the difference between
the bulk and grain boundary relaxation frequencies may not exceed one order of
magnitude.
2.7 Non-Ideality Of Realistic Materials
The impedance spectra of realistic materials usually cannot be fitted simply employing
an ideal RC circuit. For a wide range of materials, the complex plane plot of impedance
(Z*), electric modulus (M*) and complex dielectric constant (E*) for the bulk effect
normally does not show a perfect semicircle as predicted by an ideal RC element circuit,
instead it shows a depressedsemicircle with its axis depressedbelow the real axis by an
Page26
CHAPTER H LITERATURE REVIEW
angle a7t/2 (Cole and Cole, 1941, and Jonscher, 1978), as shown in Fig. 2-11. This kind
of non-ideal (non-Debye) behaviour can also be observed in the broadening of the Debye
peak in the modulus spectrum (M" vs. log (f)) (Hodge et al. 1976).
z,
Figure 2-11 A depressed semicircle in the impedance plane, the full circle completed by
dotted line can be expressed by Eq. 2-25.
The non-ideal response, i. e., the frequency dispersion from real materials, has not been
fully understood, but many studies point to chemical, physical, and even geometrical
inhomogeneities in material properties. According to Jonscher and Reau (1978), the
conducting species in the bulk moving by discontinuous jumps interact with one another
in a manner demanded by "screened hopping", which may lead to a frequency-
dependenceof electrical responseand therefore a non-Debye behaviour results. Hodge et
al. (1976) mentioned that the random orientation of anisotropically conducting crystals
and the presence of phasesof more than one composition or structure may contribute to
the broadening of the Debye peak. By carrying out finite element calculations, Fleig and
Maier (1998) demonstratedthat the varying thickness of the grain boundaries could lead
to the depressionof the grain boundary semicircle. Fleig and Maier (1998) also proposed
that an extreme grain size distribution and spatially inhomogeneous grain size
distributions could cause detours for the electrical current, and therefore could lead to a
change of the shape of the impedance spectra. Macdonald and Franceschetti (1987)
proposed that the non-ideal response could be described in terms of a distribution of
P
relaxation times which might be associatedwith a distribution of hopping-barrier-height

activation energies becauseof inhomogeneousmaterial properties. Such a distribution of
relaxation times will lead to frequency-dependenteffects.
A depressed semicircle in the impedance spectrum is fitted using a constant phase
element (CPE) connected in parallel with a resistor or a capacitor, or a CPE connected in

series with a resistor (Raistrick, 1987). The impedanceof the CPE is given by:
Y= A(jw)" =A w" [cos(w) +j sin(- )] (2- 23)
Where A is a constant that is independent of frequency, o is the angular frequency and
j=. When n=1, the CPE describesan ideal capacitor; when n=0, the CPE acts as a
pure resistor (Macdonald and Franceschetti, 1987). Thus a CPE can represent a wide
variety of non-ideal elements.
For a CPE connected in parallel with a resistor R, the total impedance can be expressed
by the Debye equation:
R
Z_ (2- 24)
1+ (ýwz)"
By combining Eqn. 2-23 and Eqn. 2-24, we can obtain an equation:
(Z'- + Z+tan a? 1+ tan 2(a)] (2-25)

224
which depicts a full circle which is displaced from the real axis as shown in Fig.2-11 (in
dotted line), with a diameter of RVI + tan Za and the centre at (2 (1+ tan 2a)),
,-2
and the depressedangle in the Fig. 2-11 is related to the exponent n:
a=(l -n)ic/2 (2- 26)
Therefore both the exponent, n, in the Debye equation and the depressedangle, a, of the
semicircle in the compleximpedanceplot describethe extent of the departureof real

materials from the ideal Debye behaviour.
Page28
2.8 Electrode/Material Interfaces
2.8.1 Introduction
In a typical complex impedance plot of ceramic materials, we often see either a
semicircle or a spike (tail) in the low frequency region that is due to the
electrode/material interface, which is also called the electrode effect. The electrode effect
is dependent not only on the properties of bulk materials, but also on what type of
electrode is used. The responsefrom the electrode/material interface often provides extra
information about the electrical properties of the bulk materials and even gives
indications of the surface state of the bulk materials. In terms of different conducting
charge carriers in bulk materials, we have electrode/semiconductor interfaces and

electrode/ionic conductor interfaces, while the electrode is normally always made from
metallic materials.
2.8.2 Electrode/SemiconductorInterfaces
The electrode /semiconductor interfaces can give either Ohmic or non-Ohmic contact. In
the case of an Ohmic contact, the electrode effect is hardly identifiable in the impedance
spectra, while a semicircle at low frequency can been seen for the case of non-Ohmic
contact. Strictly speaking, there is always a Schottky barrier present at the
electrode/semiconductor interface; that is even an Ohmic contact involves a thin
depletion layer. The capacitance of the Schottky barrier in some respects resembles a
parallel-plate capacitor, and the capactive effect of an Ohmic contact does not emerge in
impedance spectra because the depletion layer is so thin that the carriers can readily
tunnel through it (Rhoderick, 1978). Consequently, electrode /semiconductor interfaces
are comprised of both capacitive and resistive elements (Maiti and Basu, 1986) and can
be analysed using a parallel RC circuit, with the Ohmic contact being a special case with
an extremely low resistanceof the barrier.
The resistanceof the Schottky barrier is dependent not only on the oxidation potential of
the electrode metal, but also on the metal-metal interaction at the interface. For example,
the InGa/BaTiO3 interface is found to be an Ohmic contact, while the Pt/ BaTiO3
interface yields a large barrier resistance (Cann and Randall, 1996).
Paw 29
It is important to note that an Ohmic contact is a distinctively unique characteristic of
electrode/electronic conductor interfaces. The absence of a semicircular arc in the

impedance spectra due to the electrode/sample interface indicates that the conduction
process in the sample is electronic (Chiang et al. 1997). It is impossible for an ionically
conducting sample to have an Ohmic contact with a metal electrode.
2.8.3 Electrode/Solid Electrolyte Interfaces
From an electrochemical point-of-view, the electrode/solid electrolyte interface can be
either a blocking or a non-blocking one. An interface is blocking when the charge

carrying species cannot cross the interface or exchange charge with the electrode, and
there is no immediate equilibration of charged species. The interface across which the
charged species tend to equilibrate once contact has been made is called a non-blocking
interface (Armstrong and Todd, 1995). A non-blocking interface often behaves like a
resistance and a capacitance in parallel, which leads to a semicircle in the impedance

plane. This is quite like the case of non-Ohmic electrode/semiconductor contact. But an
idealised blocking interface, where there is no charge transfer, should behave like a
parallel plate capacitor, which should lead to a straight line perpendicular to the real axis
in the complex plane.
In a real electrode/solid electrolytic interface system, a non-blocking electrode may

behave like a blocking interface if an exchange current at the interface is very low. In
other cases, a blocking electrode may have charge transfer across the interface if an
extremely high potential is applied (Armstrong and Todd, 1995). But in impedance
measurementswith low applied voltages, the latter caserarely occurs.
Another important fact is that almost all the electrode/solid electrolyte interfaces exhibit
CPE behaviour to some extent. It is believed that even an ideally polarisable electrode
could exhibit CPE behaviour becausein the double layer region energy can be dissipated
in various ways other than by charge transfer (Lang and Heusler, 1998). Due to the
distributed nature of response, a non-blocking interface always shows a depressed
semicircle in the impedance spectra, which can be simulated by a CPE in parallel with a
resistance, while a blocking interface causesa straight line inclined to the real axis at an
angle smaller than 90°, which can often be simulated using a single CPE.
Page30
CHAPTER 11 LITERA TURE REVIEW
2.8.4 Effect Of Heterogeneities Of Interfaces On IS
The effect of heterogeneities of interfaces on IS is of particular interest and significance,

because the heterogeneities of the interfaces are closely associated with the bulk
properties. Hence the special features of an electrode effect in IS may provide

information not only about the interface itself, but also of bulk properties.
The CPEfor describingan electrode/electrolyte

interfaceshouldbe defined asEqn.2-23,
whereas an alternative definition:
YcPE=(Aj(,)" (2- 27)
which sometimesappearsin the literature shouldbe abandoned,since it doesnot scale

linearly with electrode area(Zoltowski, 1998).
Although the CPE is an empirical assumption, where constant À' and exponent `n' have
no well-defined physical meanings, many studies revealed that the value of the exponent
`n' is associated with heterogeneities, particularly the electrode roughness, the porosity
of a porous electrode, or a number of different charge accumulation modes such as deep

traps, inhomogenous doping etc. Many attempts have been made to seek quantitative
correlation between the exponent `n' obtained from the macroscopic response of the IS
and the microstructural features of the interface, e.g. the roughness of the electrode (Liu,
1985) or the pore size distribution of the porous electrode (Song et al. 1999). Therefore
the value of `n' is a very important parameter in the characterisation of materials.
As mentionedearlier,a blocking interfacewould showa straightline inclinedto real axis
at an anglea, the valueof `n' is relatedto the anglea:
a= (1-n) (2-28)
which is the sameasthe depressionangleasdefinedin Eq.2-26.

Thus the slope of an inclined line (frequently observed as a tail or spike), or the
depressionof a semicircle,which are presentin an impedancespectrum due to the

electrode effects, could be an indication of the extent of heterogeneitiesat the interface.
Paze 31
The impedance of a CPE can be transformed into an equivalent circuit containing

frequency-dependent resistor Rv in parallel with a frequency-dependent capacitor Cv
(McCannandBadwal,1982):
Y=R+J 'v (2-29)

v
By comparing Eq. 2-29 with the following:
Y= Ql (j, ))n =
QCOn {cos(-) +j sin( n;r )l
^
(2- 30)
22
we have:
Rv = Q-I w-" COs-'(2) (2- 31)
Cv = Qw"-' sin(-) (2- 32)
In other words, the CPE describes the extent to which both the resistive and capacitive
effects are frequency-dependent. The frequency dispersion of the interface is due to

various heterogeneities,such as roughness, pore size distribution, and even an anisotropic
surface structure (Motheo et al. 1997) and a surface disorder of the polycrystalline of
electrode (Kerner and Paijkossy, 1998). There is a large amount of literature covering
this field. However only the roughness effect and porous electrodes will be considered
here, because these are more relevant to the issue of non-destructive characterisation of
materials. The knowledge of these two aspects will certainly be helpful for a better
understandingand interpretation of the IS of ceramic materials.
2.8.4.1 Effect Of PorousElectrodes
The frequency dispersion of porous electrodes was first modelled by De Levie (1967).
De Levie's model shows that a cylindrical pore can be representedby a transmission line
as in Fig. 2-12.
With the transmission line, the following equations can be derived:
de + iRdx=0 (2-33)
Pay-e32
di +=0 (2-34)
where e stands for the potential, i stands for the electrical current, x is the length, R is the
electrolyte resistanceper unit length of the pore and Z is the Faradaic resistanceper unit
length of the pore. On combining Eqs. 2-33 and 2-34, we obtain
de-Re=0
(2-35)
X2
z"
-ZO (2-36)
Electrolyte resistance per unit

length of pore
ie i-di --ý
e+de
Rdx
di
Z/dx Faradaic resistance per unit length
of pore
Figure 2-12 Equivalent circuit of a pore.
The boundary conditions are:
x=0 e=e,
x=1 de/dx=0
The solution of Eqs. 2-35 and 2-36 can be expressed(De Levie, 1967):
cosh(px - pl) (2-37)

e-e
° cosh pl
i= pe° tanh (2-38)

pl
R
Paize33
CHAPTERII LIT RATURF.REVIEW
where
p=(R/Z)] ýz
and therefore
1°
Z. = = (RZ)'/2 cotanh pl (2-39)
eo
If the electrode/electrolyte interface is a blocking interface, then the impedance of the

double layer charging:
Z=- j/wC (2-40)
For a semi-infinite pore, the cotanh term in Eqn.2-39 can be dropped, thus, by combining
Eqs.2-39 and 2-40, a Warburg-like impedancecan be obtained:
[A
Z" =(1-j) = (2-41)
2RW
where R is the electrolyte resistance per unit pore length and C is the double layer
capacitance at the interface between electrolyte and electrode. This can explain why a
porous electrode in the absenceof Faradaic reactions follows a linear line with a phase
angle of 45°.
However many experimental data of impedance show that the phaseangle is not exactly
-
45° at high frequency or -90° at low frequency. It is believed that the impedance
frequency dispersion of the porous electrode can be affected by the shapeof the pore, and
more importantly, by the pore size distribution of a porous electrode (Song et al. 1999).
According to the model of Song et al. (1999), the impedance of a blocking porous
electrode is determined by two representativeparameters:penetrability and the standard

deviation of the pore size distribution. The penetrability au is the penetration depth X
divided by the length of pores and the penetration length X is a function of electrolyte
conductivity a, frequency coand double layer capacitanceCd:
6 (2-42)
oc
wcd
Page34
CHAPTER 11 ERATUREA EW
At high penetrability, i. e., when the penetration depth is larger than the length of the
pores, the impedance behaves like that of a planar electrode, the phase angle approaching
-90°. At low penetrability, the penetration depth is shallower than or similar to the depth
of the pore, the impedance behaves like a porous electrode. Meanwhile a wider pore size
distribution leads to a more porous characteristic of impedance, i. e., a smaller phase
angle at low frequency range (Song et al. 1999).
2.8.4.2 RoughnessEffect
The roughness effect can be understood better according to a theory of lateral charge
spreading in a diffusion double layer (Scheider, 1975). A schematic of a rough surface is

shown in Fig. 2-13 (after Scheider), here an ideally polarised electrode is assumed.
8e
\ý I' /
\IJ/
ab
Figure 2-13 Schematic of a "mound" type of surface irregularity; after Scheider (Scheider,
1975).
At the very first moment when an electrical field is applied, no concentration gradients
are present, and ionic flow follows the lines of electric force (Fig. 2-13 a) to the peak of
the mound. As charge accumulates in the double layer at the peak, the folding down of
the equipotential lines combines with lateral concentration gradients to cause tangential
spreading of ions to the remoter portions of the surface (Fig. 2-13b). As more surface
becomes accessible to charge with the elapse of time, the effective capacitance increases
(Scheider, 1975). This gives a clear picture of how capacitance is dependent on
frequency.
Paw 35
A rough electrode can be modelled by equivalent circuits composed of various RC
networks, among which the simplest one is the sameas that is used in the modelling of a
semi-infinite pore shown in Fig. 2-12. As the frequency decreases,more signal travels
through the resistive part of the network before being shunted across the interfacial
capacitance(Liu, 1985). This explains why the resistive effect of a rough interface also
increaseswith decreasingfrequency.
Thetransmissionline asshownin Fig.2-12 resultsin a Warburgimpedance:
Z oc (0) -o.s + jw -o.s) (2- 43)
which can be consideredto be a CPE with n=0.5.The transmission line also describes the
behaviour of a semi-infinite diffusion process (Raistrick, 1987). For this reason, the
electrical responseof a rough blocking interface sometimes is similar to that of a smooth,

non-blocking interface having the slow diffusion of a reactant species to the interface
(Armstrong and Todd, 1995)
However, for a real rough interface, a simple RC ladder network as shown in Fig. 2-12 is
still too simple to describe the complicated situation, and branched ladder networks
should be used to model the surface topology. The frequency dependence of these
branched ladder networks is completely determined by the branching type (Scheider,
1975), which in turn is dependenton the geometric topology of the interface.
A fractal model for CPE at rough interfaces was proposed by Liu (1985). A quantitative
relation between the exponent `n' and the geometry of the rough interface was
established as: n=1-d, where d is the fractal dimension of the rough surface that can be
determined by measurementof its areausing different scales.
Meanwhile, according to its physical model (in Fig.2-13), the lateral charge spreading
takes place within the bulk electrolyte adjacent to interface. Then it is not difficult to
understand that the resistive elements in the transmission line are actually bulk
electrolyte elements, and the conductivity varies directly with the bulk conductivity.
Thus an increase in the bulk conductivity is expected to have an effect on the CPE of the
interface by increasing the constant A. Therefore the impedance of a rough interface is
not only related to the roughness of the interface (via changing exponent `n'), but also
associated with bulk conductivity (via changing the constant A).
Paize36
Similarly to the concept of penetration depth, a propagation wavelength X

rc along the
transmission line was introduced by Scheider (1975):
2rß a (2- 44)

Ew
where a is the electrolyte conductivity, c is the dielectric constant and co is the angular
frequency. It is not difficult to seethat Eqs. 2-44,2-42 are exactly the same.
The concept of propagation wavelength (or penetration depth) is very important. It means
that a long propagation wavelength has a low `sensitivity' to the presence of the
roughness. According to Eqn. 2-44, the wavelength increasesas the bulk conductivity is
increased,therefore the impedance of an interface is expected to behave less `roughly' if
the conductivity of the bulk is increasedin some way.
It is important to note that all of the above mentioned results were based on the
electrode/liquid electrolyte systems, where the contact between the two phases are
always very intimate because of the fluidity of the liquid electrolyte. Very few studies
have been made on the electrode/solid electrolyte systems where the contact between the
two phases is not as intimate as in the electrode/liquid electrolyte systems. Obviously the
situation of the electrode/solid electrolyte system is much more complicated.

Nevertheless many basic principles of liquid electrolyte systems are applicable to solid
electrolyte systems.
2.9 Characterisation Of Structural Defects By IS
2.9.1 Introduction
Although IS has long been found to be a powerful tool for studying electroceramic and
solid electrolytic materials since Bauerle's pioneering work (Bauerle, 1969), it is only
recently that a number of studies have implied that IS could also be a promising
nondestructive tool for detecting structural defects in structural ceramics. So far it
appearsthat the majority of studies in this respect have been focused on zirconia-based
and cementitious materials, because it is of great commercial significance to apply the
simple and economic IS method for the nondestructive characterisation of these
materials.
Page3
2.9.2 Zircoinia-Based Ceramics
Zirconia-based ceramics have many important applications in industry. The
characterisation of the structural defects in zirconia-based materials using IS is desirable,

because zirconia-based ceramics, especially the partially stabilised zirconia materials
(PSZ), are often employed as structural materials
In zirconia-based materials, the ion conduction can be blocked by various microstructural

defects, such as cracks, pores and secondphase inclusions, as well as by grain boundaries
(Dessemond et al. 1993). Ion-blocking can be quantitatively characterised by the
blocking factor, which in turn can be calculated by comparing the diameters of grain and
grain boundary semicircles in impedance spectra(Kleitz et al. 1995).
The formation and evolution of cracks in structural ceramics are always a crucial issue.
Dessemond and Kleitz (1992) employed IS to characterise the cracks that were locally
induced by Vickers indentation in zirconia ceramics by applying a pin-shaped electrode
near an indentation fracture pattern. It was found that the intragrain resistance and
intragrain capacitance remain nearly unaffected by the indentation pattern, while the
intergrain resistance and intergrain capacitance vary with the indentation patterns and
with the load. Khalili and Kromp (1993), based on SEM observations and by referring to
the investigation of Dessemondand Kleitz (1992), further suggestedthat IS parameters,
such as intergrain resistance and capacitance, blocking factor etc. could be regarded
directly as parameters characterising the development of microcracks in zirconia
materials (Khalili and Kromp, 1993). A similar conclusion was reached in the report of
Tiefenbach and Hoffmann (2000): the crack has its effect on impedance in the low
frequency range, and the magnitude of the impedance increases with increasing crack
length. In an attempt to characterisemicrocracks in alumina/zirconia ceramic composites,
Smith et al. (1994) pointed out that the presence of microcracks decreasesthe effective
current carrying cross-section,and as a consequencethe resistanceis increased.
Based on the theoretical modelling, it is expected that, in addition to the relaxation
processesof grain and grain boundary, one or two separaterelaxation processesshould

appear in the impedance spectra after cracking in zirconia-based materials. But in real
cases, the responses from cracks and grain boundary tend to overlap and hence the
Pa-ae38
CHAPTER II LITERAT F.REVIEW
impedance spectra are characterised by only two relaxation processes before and after
cracking (Tiefenbach and Hoffmann, 2000).
Zirconia exists in three different phases: cubic, tetragonal and monoclinic. Phase
transformations may take place under various circumstances, for example, during high
temperature heat treatment, in a water vapour at low temperaturesor by thermal shock. In
the partially stabilised zirconia, high temperature annealing resulted in the formation of
monoclinic grains. In this case the impedance diagrams show the characteristic grain and
grain boundary semicircles, as well as an extra arc at intermediate frequencies
corresponding to monoclinic grains (Muccillo and Kleitz, 1996). Similarly, for a nano-
sized 2Y-TZP, isothermal ageing in a water atmosphere resulted in an increase of
monoclinic phase which in turn led to a change of the impedance spectrum shape
(Djurado et al. 2000). In addition, Muccillo et al. (1998) observed that an increase in
cubic/tetragonal phases as a consequenceof thermal shock produced a better resolution

of initially overlapping intergrain and intragrain semicircles All of these observations
imply that the phasetransformations in zirconia can be detected and characterisedby IS.
It is known that zirconia-based electrolytes acting as oxygen sensors are easily corroded
when in direct contact with glass melts. The inner corrosion of YSZ by glass melts was
investigated using IS by Rodrigues et al. (1997&1998). The electrical response of the
YSZ/glass composite materials was found to be strongly influenced by the presenceof a
YSZ-glass interface, which can be easily identified because of its typical relaxation
frequency. A large shift in the relaxation frequency of the intermediate frequency
impedance arc can be used to monitor the progress of glass through the zirconia-based
electrolytes (Rodrigues et al. 1998).
The porous effect of the YSZ samples with a fairly narrow pore size distribution can be
clearly identified in the impedance spectra. The diameter of the grain boundary
semicircle increases with the porosity present in the samples. It is documented that the
porosity of porous material can be characterisedby a blocking factor ((xR),a capacitance
ratio (ac) and a frequency ratio (af). The product of afa.R was found proportional to
porosity (Steil et al. 1997).
Page39
2.9.3 Cementitious Materials
Cement paste is a heterogeneousmixture of solid, liquid, and gaseouscomponents. The
solids are precipitated hydration products and un-reacted cement grains. The liquid is a
pore fluid which permeatesthe matrix through the pore structure and the gas is entrapped
air in empty or partially fluid-filled pores (Macphee et al. 1997). The hydration product
of cement paste is mainly an amorphous, hydrous gel, designated C-S-H. This phase is
the principal binder in Portland cement pastes and is characterisedby a high surface area,
a variable composition and a variable structure (Glasser et al. 1987). For this reason,
porosity and ionic conductivity are the main physicochemical parameters defining the
mechanical properties and lifetime of concrete structures (Andrade et al. 1999).
McCarter et al. (1988) were among the first who found that the hydration (hardening)
process of the cement paste could be monitored by IS. Since then there has been a large
amount of researchdevoted to the characterisation of cementitious materials using IS.
In the impedance spectra of cement paste there appears a depressed semicircular arc at
high frequencies that is due to a bulk effect and a low frequency tail due to the electrode
effect. As the hydration proceeds, the structure and composition of the hydration
products keep changing. It was found that in impedance spectra, the bulk resistance and
depression angle increases while the relaxation frequency bulk arc decreaseswith the
prolonged hydration time (McCarter et al. 1988).
Several IS parameters can be directly related to the microstructure of cementitious
materials, for example the bulk resistance can be related to the connectivity of the pore
structure and the arc depression factor appearsto be a measure of the inhomogeneity of
the underlying microstructure (Christensen et al. 1994). Use of the frequency at the top
of the bulk arc makes it possible to estimate the relative dielectric constant of the
material, which is in turn governed by the ratio of an average pore size to a blocking
layer thickness for the microstructure (Christensen et al. 1994). A linear dependencehas
been recently found between the high frequency dielectric constant and the cement paste
porosity (Andrade et al. 1999).
IS has also found application in the investigationof crack growth mechanismsin

cementitious materials. This is based on the fact that the high frequency arc (HFA)
Paize40
CHAýI LITERATUREREVIEW
diameter is dependent on the microstructural characteristics of the specimen. When
cracks are formed and saturated with liquid phase, small values of bulk resistance are
obtained. Consequently during compressive loading, the HFA diameter decreaseswith
increasing loading, due to the formation and propagation of cracks. The rate of decrease
of the HFA diameter reflects the rate of crack formation and propagation (Gu et al.
1993).
2.9.4 Debonding And Pullout Of Fibres In Composites
Two bulk arcs could appear in impedance spectra when chopped conductive fibres (steel
and carbon) are added to cementitious materials (Ford et al. 1998), which can be
explained by a `frequency switchable fibre coating' model (Torrents et al. 2000). The
high frequency arc representsthe bulk between wire tips and the electrodes, while the
low frequency bulk arc is the response from the ion spreading in the vicinity of the wire
tips (Torrents et al. 2001).
In a pullout test, a single steel fibre was being pulled out from a cement sample while
impedance measurements were made between the two electrodes. It was found the
resistance of the high frequency semicircle is sensitive to the debonding and pullout of
the embedded fibre. The high frequency bulk resistance increases progressively as the
fibre is pulled out and the extent of pullout can be related quantitatively to the increase in
the resistance. A significant increase in this resistance also occurs during the debonding
phase of the crack opening displacement (Torrents et al. 2001). Therefore IS can be
applied to detect and monitor both debonding and pullout of the embedded fibres in
fibre-reinforced cementitious materials.
2.10 Summary
According to this literature review, we know that IS is a cheap, quick and convenient
non-destructive technique for characterising electrical properties of solid electrolytes and

electroceramics. However, it has also been seen that the interpretation of IS needs a
sound understanding of electrical conduction mechanismsin materials, the establishment
of an appropriate physical model and proper representationsof IS.
Paize41
CHAPTER II LITE AT FRVIE-
The electrical conduction in the majority of polycrystalline ceramic materials is either
semi-conductive or ionic, while silicate glassesexhibit mainly alkali ion conduction. For
porous oxide ceramics the electrical conduction along free surfaces is an important factor
that hasto be considered.
There are dozens of ways of representing IS. Different IS representations highlight

different features; for example electrical modulus spectraemphasise the capactive effect,
while impedance spectra demonstrate more clearly the resistive effects of different
phases. Therefore to render a set of impedance data to be more informative, many
different representationsshould be tested before final representationsare decided.
All realistic materials show non-ideal behaviour, which can be fitted by using constant
phase elements (CPEs). The depression angle of a semicircle in complex impedance

spectra is an indication of the extent of the frequency dispersion of a material.
There are two kinds of electrode/material interface in terms of the charge carriers in the
material: electrode/semiconductor interface and electrode/electrolyte interface. The

heterogeneities at the interface, such as electrode porosity and roughness can be related
to a few IS parameters. The electrode effect is also reflective of the properties of bulk
materials. Therefore in analysing the impedancedata, attention should be paid not only to
the bulk semicircles, but also to the electrode tail in the impedance spectra.
The interest in using IS to characterise structural defects in materials is increasing.

Previous studies have demonstrated that structural defects such as cracks/microcracks,
porosity, phase changesand other second phase inclusions etc, can be characterised by IS
parameters. However, so far only zirconia-based materials and cementitious materials
have been investigated in this respect. The present study is an attempt to extend IS as a
NDC technique to a wider range of structural ceramics both in-process and in-service
applications.
Paze42
CHAPTER III EXPERIMENTAL PROCEDURES
3 Experimental Procedures
3.1 Sample Preparation
A number of different ceramic materials have been used in this project, some of which
were provided by other research groups or industrial companies that will be separately
specified later on at appropriate occasions and others, unless otherwise denoted, were
prepared by the following processing procedure:
Raw Materials
Granulating
& Sieving
Weighing
Forming by
Pressing
Ball Milling
Sintering
Drying
Sintered Pellets
Figure 3-1 Flow chart for preparing sintered ceramic pellets.
The green form pellets were 13 mm in diameter and 3-4 mm in thickness, formed by uni-
axial pressing using a pressureof 120 MPa. The pellets were sintered in a tubular furnace
(Lenton, with SiC heating elements), using a heating rate of 3 K"minute 1 and a cooling
Page 43
CHAPTER LII EXPERIMENTAL PROCEDURES
rate of 5 K"minute-1. The dwell time depended on different circumstances that will be
given in particular situations in the following chapters.
3.2 Measuring Cell Preparation
3.2.1 Electrode Preparation
An electrode can be applied either by pressing a metal foil directly to a surface of the
solid material under investigation, or by applying a thin conductive paint to the surface.
Since it is extremely difficult to make an absolutely flat surface for solid materials, the
contact between the electrode and the material using the former method is usually very
poor. To ensure a good contact, painted electrodeswere normally used.
To apply a painted platinum electrode, the sample was painted firstly with a thin layer of
platinum conductiveink (EngelharhClal, UK) over a definedareaon the samplesurface,

followed by baking at 1100°C for 45 min.
Likewise, a painted silver electrode was applied by painting a thin layer of silver
conductive ink (E8100, Johnson Matthew Electronics, UK) on the sample surface, and
this was followed by baking at 450°C for 15 min.
3.2.2 Measuring Cell Containing A Dense Ceramic Sample
Platinum foil
Paint Electrode
pellet
_Sample
Platinum lead
Figure 3-2 A typical measuring cell with a sintered ceramic pellet.
A typical measuring cell for impedance measurementsis shown in Fig. 3-2. Here the cell
is composed of a pellet specimen sandwiched by two platinum foils. The cell is
Paze 44
CHAPTER III LXPFRIA/ENN7_ILPRO('E/)('RES
symmetric, with two identical electrodes. The electrodes are either painted platinum or
painted silver electrodes.
Two platinum foils were applied to the painted electrodes to ensure a good contact
between the painted electrode and the lead.
The procedure for preparation of the measuring cell is as follows:
(a) a sample pellet was ground using SiC papers of P180 and P800 grit;
(b) the sample pellet was polished using a polishing disc with 6µm and 1µm diamond
pastes;
(c) the sample pellet was cleaned using acetone and then dried in air;
(d) the sample pellet was painted with platinum or silver pastes.
3.2.3 Measuring Cell For MeasurementsOf Green Form Ceramics
A measuring cell for green form ceramics is shown in Fig. 3-3. Becausethe green form is
porous and fragile, it is difficult to apply the painted electrodes. Therefore, two platinum
foils were used here as electrodes.
Insulating block
Platinum foil
Platinum lead
Sample pellet :xt,
Figure 3-3 A measuring cell with a green powder compact.
In order to make as a good contact between the sample and the electrodes as possible, the
contacting surfaces of both the green compact pellet and electrodes need to be as flat as
possible. For this reason the green compact was carefully polished with fine SiC paper
Paze 45
CHAPTER 111 EXPERIMENTAL PROCEDURES
(P1200 grit) to ensure both surfaces were flat and smooth enough. Meanwhile, the
platinum foils were bonded to insulating blocks by the following procedure:
(a) two A1203blocks were polished using 6µm and 1µm diamond pastes;
(b) the A1203blocks were painted with a thin layer of platinum ink;
(c) the platinum foils (200µm in thickness) were attached to the painted area on the
A1203blocks before the drying of the platinum ink;
(d) the A1203/foil joint were formed after sintering at 1250°C under a pressureof 5N for
60 minutes.
3.2.4 Measuring Cell For Impedance MeasurementsOf TBCs
The cell for impedance measurementsof TBCs was asymmetric (Fig. 3-4). In this cell a
silver electrode was painted at the top surface of the YSZ coating, while another one was
the bond-coat beneaththe YSZ top layer in TBCs (see Fig. 3-4).
Lead /-Painted electrode
op YSZ coat
xide product
3ond coat
>ubstrate
Figure 3-4 A measuring cell with TBCs.
In the preparation stage, care must be taken to prevent mechanical damage to the
coatings, especially during the cutting of the samples. To avoid damaging the top coat
(YSZ), the top surface was not polished before applying electrodes, but the sampleswere
cleaned using acetonebefore the application of the painted silver electrode.
Pa,ae 46
3.3 Sample Holders
3.3.1 The Sample Holder For The Room TemperatureMeasurements
For the room temperature measurements,a simple sample holder as shown in Fig. 3-5
was used, where an adjusting bolt was used to ensure an intimate contact between
specimen and the platinum electrodes.
432
1. STEEL FRAME
2. ADJUSTING BOLT
3. MEASUREMENT CELL
4. INSULATING BLOCK
Figure 3-5 Sample holder for room temperature measurements.
3.3.2 High TemperatureRig
When the room temperature sample holder (Fig. 3-5) was used for impedance
measurementsat high temperatures in a tube furnace, there was a strong electromagnetic

interference from the electric heating furnace. For this reason, a special measurement rig
was designed, as shown in Fig. 3-6. Here a Fecralloy tube that was connected to ground
was used as an electromagnetic shield to prevent the interference from the furnace
winding. A wire spring was used to apply a constant pressure of about 5N to the
specimen to ensure an intimate contact between electrodes and the specimen. The
electrodes were two platinum foils bonded to A1203 plates and connected to the
impedance analyser. A thermocouple (Type R, Carbolite, with a digital 2132 indicator)
was placed next to the specimen in order to measure accurately the temperature of the
specimen.
Paze 47
18
1. ELECTRODES 2. THERMOCOUPLE 3. LEADS

4. SPRING 5. CERAMIC HOLDER 6. SILICA TUBE
7. SHIELDING 8. TUB FURNACE
Figure 3-6 A Rig for impedance measurement at high temperatures.
Since both the Fecralloy tube and ceramic holder used in this rig are capable of
withstanding a high temperature environment (higher than 1250°C), this rig can be used
for impedancemeasurementsup to 1200°C.
3.4 Impedance Measurements
3.4.1 Impedance Spectroscopy
The set up for impedance spectroscopy is shown in Fig. 3-7, in which a Solartron 1296
dielectric interface was used in conjunction with a Solartron Frequency Response
Analyser (FRA) 1255. The 1296 interface/1255 FRA unit was controlled by a personal
computer.
The basic function of the 1296 hardware is to extend the current measurement range of
the FRA, thus enabling the measurementof very high impedances (Solartron, 1999). This
is essential when we conduct a measurement on a dielectric material with a high
resistivity.
Paw 48
The main features of the 1296 interface/1255 FRA combination are:
Frequency range: I OµHz to 10MHz
Signal amplitude: 0 to 3V rms
Current measurement: 1fA to 1OOmA
Tan 8 range: <10"4to 103
Impedancerange: 100 S2to 100TS2(1014Q)
Capacitancerange: 1pF to >0.1F
1296 INTERFACE /
PC SI 1255 ANALYSER
Connection via
MEASUREMENT
the GPIB r-
CELL.
iký .ý
ý.. ".. .p
Figure 3-7 Schematic of the set up of impedance spectroscopy
3.4.2 Correction Of Impedance Data At High Frequencies
At high frequencies, especially when the frequency is greater than 1MHz, an abnormal
feature was always present in the impedance spectra. This is because at high frequencies
impedance data can be seriously affected by stray or residual capacitance and inductance
effects due to the cabling and sample holder. For a real measurement system, the
measured impedance Z,, can be considered to be a mixed contribution of the impedance
of the sample ZS,the impedanceof the short circuit Z, and the impedance of open circuit
Zo,as schemed as in Fig. 3-8 (Christensen et al. 1994).
Pa2e 49
Zc
2
Hrn
Zo ZS
Figure 3-8 Schematic representation of the elements contributing to a typical impedance

measurement.
In an ideal case, Z. should be infinite and Z,,should be zero. However, Z, increases and
Z. decreaseswith increasing frequencies. So, at high frequency, errors resulting from the
effects of stray capacitanceand inductance can be very large, which have to be corrected
by a proper procedure. One of the most efficient ways is by the following procedure
(Christensen et al. 1994).
f Perform IS of Z,., with a high conductivity material shorting the sample holder.
f Perform IS of Zo,with only air in the sample holder.
f Collect IS with a sample in the sample holder.
f Insert real (R) and imaginary (X) components for the shorted (s), open (o), and
measured(m) responsesinto the following equations at each frequency:
R, R (R' +X')-R(R' +X') (3- 1)
R2 =RJR2 +X(2)+R,, (R2 +X? ) (3-2)
RR, + 2R0XcXm
R3= 2Rc, (3-3)
n
+X2) X0 (Rm
X, =X. (Ro - +X2)
(3-4)
X2 = XJR2 +Xz)+X, (R2 +X? ) (3-5)
X3 = 2RcX,,Rm+2XcX,, X,,, (3- 6)
D (R, -R, -Ro)2 +(X. -X,. -X)2 (3-7)
Z real= (R, - R2 + R3)/D (3-8)
n___ en
CHAPTER III EXPERIMENTAL PROCESS
Z`mag = (XI -X2 +X3)/D (3- 9)
where RI, R2, R3, X1, X2, X3 and D are intermediate functions, Z, `°' is the real part and
Z;"' the imaginary part of the impedance of the sample under measurement. This
procedure has been used for eliminating the high frequency abnormality in some casesin
this project. Meanwhile there is also a correction procedure provided in the Solartron
Òperating Manual', which is called `normalising a measurement' (Solartron, 1999). This
Solartron approach is much less time-consuming, but according to my personal
experience, it seemedless effective in eliminating the errors.
Nevertheless it is important to note that the impedance data at the frequencies under
I MHz were reliable even without undergoing a correction procedure. This had been
proven by, in several cases,our comparing the raw experimental data with the corrected
data obtained following the correction procedures.
3.5 Other Analytical Techniques
3.5.1 Density Measurements
The density measurementswere conducted by measuring the weight and the geometrical
dimensions of the samples. The samples were in a circular disk shape. The faces of the
disk were carefully polished using P800 grit SiC paper, then the diameter and the
thickness of the disk were measured using a micrometer. The disks were dried at 200°C
for 24h before being weighed using a digital balance. The density was calculated by:
D=D (3-10)
2,5
where W is the weight (g), D (cm) the diameter and S (cm) the thickness of the sample
disk.
3.5.2 Thermogravimetric Analyses
Thermogravimetric analyses (TGA) were carried out in a thermogravimetric analyser

(FGS-2, Perkin-Elmer) to examine the weight loss of a ceramic compact during heating.
Pame51
CHAPTER 111 EXPERIMENTAL PROCEDURES
-
The thermogravimetry instrument consists mainly of a precision balance, a furnace
controlled by a temperature programmer and a data acquisition system as shown

schematically in Fig. 3-9. TGA allows monitoring and precisely measuring the thermal
weight changeof the sample at varying temperaturesup to 1100°C.
Furnace
Sample container
Figure 3-9 Schematic of a thermogravimetry instrument.
3.5.3 Dilatometry
The shrinkage behaviour of a green powder compact as a function of temperature was
measuredusing a dilatometer (402E, Netzsch-GeratebauGmbH).
Temperature
Sample programmer
Figure 3-10 Schematic of a dilatometer instrument
Pa,ae 52
The dilatometer consists of an electronic transducer, a furnace controlled by a
temperature programmer and a data acquisition system as shown schematically in Fig. 3-

10. This instrument allows the continuous measurement and observation of the thermal
change in length of a specimen at temperatures ranging from room temperature to

1700°C. This test needsa sample with dimensions in the range of 25 to 50mm in length
and 3 to 14mm in diameter.
3.5.4 X-Ray Diffraction
X-Ray Diffraction (XRD) is the most effective technique for the identification of
crystalline phases present in materials. In this project, XRD was used to examine the
phases present in the sintered clay compacts and heat-treated TBCs samples. Both bulk
and powdered sampleswere used. The powdered samples were obtained by crushing the
bulk samples using a pestle and mortar. XRD measurementswere performed using a
Philips diffractometer (PW1050), with Ni-filtered CuKci radiation. The step size was
0.01' and angular scan speedwas 0.8°/minute. The angular scan range was 15 - 90°.
3.5.5 Optical And Scanning Electron Microscopy
Optical microscopy and Scanning Electron Microscopy (SEM) were used to study the
microstructures of the materials. The samples for the optical microscopy and SEM
observationswerepreparedby the following procedure:
(a) samples were mounted in resin and cured at room temperature for 24h;
(b) the mounted samples were ground with SiC papers progressively from P180 grit to
P800 grit on a rotating plate;
(c) the samples were polished with first 6µm, and then I µm diamond paste.
Specimensfor SEM were preparedby gluing the samplesonto a sampleholder with a

carbon tape and sputter-coating them with a gold or carbon conductive layer to prevent
electron charging effects.
Pa,ae 53
SEM observations were carried out with a JEOL, JXA-840A instrument.
3.5.6 Energy Dispersive Spectroscopy
Energy dispersive spectroscopy (EDS) is an electron probe microanalysis technique. In

this technique, an electron beam is focused on the surface of a specimen. X-rays are
emitted as a consequenceof interactions between the electron beam and materials. The
X-ray spectrum consists of characteristic peaks at different energy levels. Qualitative and
quantitative analyses of the component elements can be made by measuring the energies
and intensities of thesecharacteristic lines in the spectrum.
In this project, the microanalysis was made with an energy-dispersive X-ray analyser
(AN 10000, Oxford Instrument) coupled to a scanning electronic microscopy (JEOL
JXA-840A). All the microanalysis was conducted qualitatively.
P
4 ImpedanceData Analysis
4.1 Introduction
Once we have obtained a set of experimental impedance data, we can conduct equivalent
circuit modelling by using computer software to get the values of elements in the
equivalent circuit. In this project we have used `Zview for Windows' (Scribner
Associates, Inc) in simulation. Obviously equivalent circuit modelling is only useful
when an appropriate model is selected. Software cannot select a model for you, although
it can determine the values which best fit your data based on a model you design
(Scribner, 1998).
To select an appropriate equivalent circuit, we need to rely on our own knowledge of the
physical processesin the measuredsystem. Also, and almost of equal importance, is that
we need to make many trials to see what features of the spectrum will change if we
adjust certain value(s) of elements in the equivalent circuit. In this way we can examine
whether or not the equivalent circuit model on trial is `reasonable'. If not, elements need
to be either added, or cancelled, or rearranged basedon the changes in spectrum features
when values of elements are changed.
The analytical approach to IS analysis to be presentedin the following section is not only
necessaryfor a better understandingof IS, but also very useful for the fitting of
impedance data and the proper representationsof IS.
4.2 Analytical Approach
In reference to a brick layer model, we assume that grain, grain boundary and
electrode/material interface are ideally represented by RIC,, R2C2, R3C3 elements

respectively, which are connectedin seriesas in Fig. 4-1.
Page55
CHAPTER IV IMPEDANCE DATA ANAL SIS
Rl R2 R3
cl CZ C3
Figure 4-1 Equivalent circuit for an idealised brick layer model.
The impedanceof this electrical circuit can be expressedas:
Z= Z'- jZ" (4-1)

11+11+11=
+j wC, -+ jCOC2 -+ jwC3
R, RZ R3
where w is the angular frequency, w=2irf. From this equation, we have:
R' R2 R3
Z'= + + (4-2)
1+(wR, C, )Z 1+(wRZC2)2 1+(COR33)Z
and
toR2iC, O)R22C2 tOR23C3

Z_ + + (4-3)
1+(wR, C, )Z 1+(wR2C2)2 1+(wR3C3)Z
The relaxation frequency of an RC element is determined by its resistance and
capacitance:
I
w, (4- 4)
R,C,
then we have
Paize56
CHAPTER IV IMPEDANCE DATA ANALY 1
R' R2 R3
Z'= + + (4-5)
1+(0)2 1+(w )2 1+(Ü 2
Q)1 C02 033
Co ww
R, - R2- R3-
Z" = ý' ý2 (03
+ + (4-6)
J+( 03)Z 1+(u' )Z 1+()zCo
0)1 wZ C03
These expressionsare in the sameform as those given by Abrantes et al. (2000b).
Based on Eq.2-6 (M* j(oCoZ*), it is not difficult for us to get similar equations for
modulus components:
Co*(0)2 C', Co Co*(?

)2 )2
C' C2
M'
O)'
+ C0Z C3 (03
= + (4-7)
1+(0)2 1+(CO )2 1+(w)2
CVO Cv2 0)3
C', w
C1
C0 w c', w
O)l C2 w2 C3 (03
M"= + + (4-8)
1+(_)2 1+(w)2 1+(w)z

0)1 0)z W3
Now a demonstration will be given, mathematically, that there can be three separate
semicircles in complex impedance and modulus plots, if the relaxation frequencies of

different RC elements are orders of magnitude different to one another.
Assuming w>»wz»w3, in high frequency range, we have: (o/w2»1 and 0w3» l. R2
and R3 usually have finite values. If R2 is assumedto be much smaller than o)/o)2and R3
is assumedto be much smaller than w/w3, then we can have two approximations:
R2 R3
0 (4-9a) and =:> 0 (4- 9b)
1+-
w2
1+- 0
w3
Based on these two approximation, then according to Eqs. 4-5 and 4-6, we have
Paize57
CHAPTER IV IMPEDANCE DATA ANAL.Y, I
R'
z, = (4- 10)
1+(, )Z
Oil
R Co
i
Z"= 0)' (4- 11)
1+(``')Z
Co]
Upon combining Eqs. 4-10 and 4-11, we have
1)2
(Z")2 +(Z, _ (RI )2 (4- 12)
2 =
This means in the high frequency range, the contributions from the R2C2and R3C3 is
negligible and the impedance of the R1C, element follows a circular equation, with a
diameter R1 and the center on the x-axis (R1/2,0).
In the medium frequency range, w/w1«1, w/w3»1 and by assuming R3 «w/w3 then
Rl R3
=* R, and
1+03 1+-
wl C03
Eqs. 4-5 and 4-6 can be approximated by
RZ
Z'=R, + (4-13)
1+(0-))2
(0z
R2 0)
Z" _ 0)2
(4- 14)
1+(0) )2
(t) 2
By combining Eqs.4-13 and 4-14, we can obtain another circular equation:
(Z")2 +[Z'-(R, +Z )]2 = (22 )2 (4- 15)
Paze 58
CHAPTER IV IMPEDANCE DATA ANALYSIS
In this frequency range, the impedance of the R2C2 element also follows a circular
equation, with a diameter R2 and the center on the x-axis [R1+(R2/2), 0].
Likewise at low frequency, co/coI«1 and w/cw2«1,
The impedance of R3C3element follows a circular equation, with a diameter of R3 and
the center at [R1+R2+(R3/2),0]:
(Z")'+ [Z'-(R, + R2 +3 )]Z = (23 )2 (4- 16)

2
By following similar deduction procedures,the circular equations for modulus spectra in

different frequencyrangescanbe obtained:
Ca Co
(MI? ) 2+ (M \z lz
/ lY[ _ 1l _ (4- 17) at low frequencies
2C, 2C,
co
)] 2= (Co )2
2+IM'_(- 1,
+
lý
1z (4- 18) at mediumfrequencies
l
z\
C', Cc) CO
(M"ý2 )Z
+[M'-(C + + (4- 19) at high frequencies
C, C2 2C3 2C3
4.3 Numerical Analysis
According to the above mathematical analyses, the electrical responses from different
phasesin a specimen can be distinguished in impedance and modulus spectra as resolved

semicircles, if the relaxation frequencies of different phases are `sufficiently' different.
But how different should the relaxation frequencies be to get a well-resolved spectrum?
Meanwhile in the light of Eqn. 4-4, we know the relaxation frequency is associatedwith
both the resistance and capacitance. The relaxation frequency can differ as a
consequenceof the difference in either resistanceor capacitance,or both. These different

factors may have different effects on the resolution and other features of the spectra. To
answer these questions,numerical analysis is necessary.
Page 59
Impedance spectra Z' Spectra
6.E+08
6. E+08
4.E+08 i4 4.1.
2.E+08
0.E+00 E-04 1.E-02 1.E+00 1.E+02 1.E+04 1.E+06

0.E+00 4. E+08 8.E+08 1.E+09
Frequency
n,
r Spectra
2E402
2.E02
Rr tEa¢ wE6.
E-C3
O.&00
o.E+ao 1.E02 2Ea2 s.e-a¢ 4E-C2 5.E-M 1.E04 1.E-01 1.E+02 1.E+06
M
Bode plot
Tan delta vs. Frequency plot
100.00 1.E+01
80.00 r'ý S 1.E+00

e
av 7sý-ý
60.00 1 E-01
trI
IL 40.00
1.E-02
1.E400 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E-01 1.E+01 1.E+03 1.E+05
Frequency
Frequency
Fig. 4-2 Various representations of IS corresponding to equivalent circuit model as shown

in Fig. 4-1, with """"i.....""representing R3/R2=R2/R,=5, - representing R, /R2=R,/R, =25,
" representing R3/R2=R2/R1=100. The values of elements employed in this simulation
are listed in Table 4-1.
Table 4-1 Values of the elements in the equivalent circuit model (Fig. 4-1) used for
numerical analysis.
R, R2 R3 C, C2 C3 Co R3/R2=R2/R,
3x10 1x10 1x10 1x10 IXIO-10 IXIO-10 IXIO-11 100
3x108 4x107 1x109 1x10-10 1x10-10 1x10-10 1x10-" 25

3x108 2x108 1x109 1x10-10 1x10-10 1x10-10 1x10-"
* Empty capacitance of the measuring cell.
Page60
Impedance Z' Spectra

spectra
3. E+08
4.E+08
3.E+08 2. E+08
_
jl 2.E+08 f" N
1.E+08
1.E+08
rY
0.E+00 0. E+00
0. E+00 2. E+08 4. E+08 6. E+08 8. E+08 1. E+09 1.E-03 1.E-01 1. E+01 1.E+03
T Frequency
M" Spectra
Modulus Spectra
6.E-02 5.E-02
4.E-02
4.E-02
3.E-02
N 2.E-02
2E-02 tIII1.
E-02
0.E+00 0.E+00
0.E+00 2.E-02 4.E-02 6. E-02 8. E-02 1.E-01 1.E-04 1.E-01 1.E+02 1.E+05
M' Frequency
Bode plot Tan delta vs. Frequency plot
100.00 1.0E+04
80.00
1.0E+02
C 60.00
0 F
40.00 1.0E+00
20.00
1.0E-02
0.00
1.0E-04 1.0E-02 1.0E+00 1.0E+02
1. E-03 1. E-01 1. E+01 1. E+03
Frequency
Frequency
Fig. 4-3 Various representations of IS corresponding to equivalent circuit model as shown

in Fig. 4-1, with ."."a......" representing C, /C2=C2/C3=5, - representing Cl/C2=C2/C3=25,
" representing Cl/C2=C2/C3=100. The values of elements employed in this simulation
are listed in Table 4-2.
Table 4-2 Values of the elements in the equivalent circuit model (Fig. 4-1) used for
numerical analysis
RI R2 R3 Cl C2 C3 Co Cl/C2=C2/C3
3x10 3x10 3x10 2x10 2x10 2x10 1XIO-11 100

RIO' 3x108 3x108 2x10-" 5x10-10 7.5x10-9 1x10-" 25
3x108 3x108 3x108 2x10-" 1x10"10 5x10-'° 1x10-" 5
* Empty capacitance of the measuring cell.
Page 61
CHAPTER IV IMPEDANCEDATA ANALYSIS
Two caseshave been considered here. In the first case (hereafter denoted as case one),
we assumed that the relaxation frequency differs as a consequenceof the differences in

resistances. In order to highlight the influence of the resistance, all three capactances
were assumedto be the same.The differences in resistanceswere described by the ratios
(R3/ R2, R2 /R1, R3 >R2 >R1) which vary from 5 to 100. The parameters corresponding to
the equivalent circuit model (Fig. 4-1) that have been used in the numerical analysis are
given in Table 4-1. The simulation results are displayed in Fig. 4-2.
In another case (hereafter denoted as case two), we assumed that the relaxation
frequencies are different due to the differences in capacitances. The differences of
capacitances were described by the ratios C1 /C2, C2 /C3, C>> C2> C3, which also vary
from 5 to 100, while all of the three resistances were the same. The values of the
elements in the equivalent circuit are given in Table 4-2. The simulation results are
displayed in Fig. 4-3.
A number of interesting points are evident from Fig. 4-2 and 4-3, which are summarised
as follows:
a. When the difference in relaxation frequency is due to the differences in

resistances among RC elements, as in case one, the modulus spectra (both
complex modulus plot and M" spectroscopic diagram) show well-resolved
semicircles corresponding to different elements in the equivalent circuit, while
semicircles in the impedance spectra (both complex impedance plot and Z"
spectroscopic diagram) entirely overlap. By contrast, in case two where the
capacitancesof the RC elements are different, the semicircles in the impedance
spectra are resolved but modulus spectra are not resolved. This means that the
difference in resistance leads to resolution in the modulus spectra, while the
difference in capacitanceleads to the resolution in the impedance spectra.
b. In both cases,we can seepeaks and valleys in the phase angle (Bode) plot and tan
(S) vs. lg(f) diagram. This means that the phase angle (Bode) plot and tan (S) vs.
1g(f)diagramaresensitiveto the differencesboth in resistancesandcapacitances.
Pare 62
CHAPTER IV IMPEDANC ED ATA ANALYSIS
c. In both cases, spectral resolution depends on the degree of difference in

resistancesor capacitances.When the ratio between resistancesor capacitancesis
larger than 25, semicircles can be distinguished. However, when the ratio
between resistancesor capacitances is smaller than 5, the semicircles overlap to
such an extent that only one depressed semicircular arc can be seen in the
complex impedanceor modulus spectra.
d. When semicircles are well resolved (i. e., if the relaxation frequency difference is
larger than 25-times), the resistances can be worked out straightaway from the
complex impedance plot, as the resistances are equal to the diameters of the
semicircle. However in practice the semicircles are not perfect due to the overlap,
the resistancesneed be worked out with reference to the cusp, i. e. the lowest point
where two semicircular arcs meet. The real component at the first cusp from the
left is the resistanceof the RC element with the highest relaxation frequency, the
length in the real componentsbetween the first cusp and second cusp corresponds
to the second RC element and so forth.
e. The resistances can also be easily obtained using the Z" spectroscopic diagram.
According to Eqn. 4-6, the peaks in the Z" spectroscopic diagram are R1/2, R2/2
and R3/2 respectively.
f. In a similar way, the capacitancecan be obtained by using the complex modulus
plots. According to Eqs.4-17 - 4-19, the diameters of the semicircles are Co/C1,
Co/C29Co/C3,respectively, where Co is the capacitanceof the empty space. Here
again the diameters of arcs should be determined by reference to the cusps of the
spectrum.
Also the capacitances can be obtained by examining the M" spectroscopic

diagram. According to Eqn. 2-8, the peaks in the M" spectroscopic diagram are
Co/(2CI), C0/(2C2),Co/(2C3),respectively.
Page 63
CHAPTER IV IMPEDANCE DA TA ANALYSIS
4.4 Summary And Conclusions
The representations of IS need to be made in as many formalisms as possible. To
represent IS only in one formalism, for example only in the impedance or modulus
formalism, may miss valuable information that could have been available in impedance
data.
When the relaxation frequencies of different RC elements are close (the difference is
below 5 times), semicircles will overlap to form one single semicircular arc similar to a
depressedsemicircle. This will seriously complicate the impedance analysis. Care must
be taken when we attempt to draw any conclusions in this situation.
The Bode (phase angle) plot and tangent loss plot are sensitive to the differences both in
resistances and capactances.They are useful in determining the presence of relaxation

processes.However a data analysis which combines the modulus and impedance spectra
is more effective. The advantage of this method is that not only qualitative information,
such as how many relaxation processesare present and how well resolved is the spectra,
but also quantitative information, such as the values of resistances, capacitances and
relaxation frequencies can be obtained from both complex plane and spectroscopic plots
of impedance and modulus.
Pam
CHAPTER V NDC OF AL UMINA/SILICONCARBIDENANOCOMPOSITES
5 Nondestructive Characterisation of Alumina/Silicon

Carbide Nanocomposites Using IS
5.1 Introduction
Non-destructive characterisation (NDC) of ceramic matrix composites is essential for

developing reliable ceramics for industrial applications. The brittleness, catastrophic
failure behaviour and therefore, the low reliability of ceramics have been the main
obstacles for the application of ceramics. Without NDC, a slight improvement in the
reliability of ceramics would require a much higher cost of manufacture, because
expensive processing techniques have to be used to ensure the high performance of
products, which quite often is unnecessary. A high performance does not necessarily
provide a high reliability, whereas an effective NDC technique will certainly help
improve the applicability of ceramic materials. Extensive studieshave been carried out to
investigate the processing and properties of advanced structural ceramics, which have
resulted in significant improvements in the performance of ceramics. However, very few

studies have been done on NDC of structural ceramics, which is critical in controlling the
quality of ceramics (Van de Voorde, 1998). This largely is due to the lack of availability
of an efficient and comprehensive nondestructive technique that could be applied to
evaluating ceramics.
In this study, in order to investigate the feasibility of using impedance spectroscopy as a

NDC technique for characterising structural ceramic materials, A1203/SiC
wereselectedasmodelmaterialsfor the following reasons:

nanocomposites
(a) SiC particles and whiskers are important reinforcing components to strengthen and
toughen ceramics and glasses, but reports about the electrical properties of the
materials of this category are limited (Wang et al. 2000; Zhang et al. 1992). Under
certain circumstances, the electrical properties may be of concern in the application
of these composites, therefore an investigation into the electrical properties of these
materials is necessary.
Page 65
CHAPTER V NDC OF AL UMINA/SILICON CARBIDE NANOCOMPOSITES
(b) A1203/SiC nano-composites represent a series of insulator/electronic conductor
composites, for which particular examples are the SiC or Carbon-fibre-reinforced

ceramics and glasses. A successin applying IS as a NDC technique for A1203/SiC
nanocomposites means that IS could be extended to the whole series of
insulator/electronic conductor composites.
(c) A1203/SiC nanocomposites have been attracting growing interest because significant
improvements in strength and toughness have been achieved by the addition of SiC
nanoparticles (Niihara, 1991; Ohji et al. 1996). They are amongst the most promising
structural materials that could be applied at high temperaturesup to 1000°C.
In addition, it was revealed that the oxidation rates of A1203/SiCnanocompositeswere
several orders of magnitude higher than that of monolithic SiC (Luthra and Park,
1990; Sciti and Bellosi 1998). Oxidation at elevated temperature could be a crucial
factor in controlling the degradation of A1203/SiC nanocomposites at high
temperatures. Therefore it is highly desirable to develop a nondestructive evaluation

technique for studying the oxidation of the Al203/SiC nanocomposites at high
temperature.
The purpose of this work is to study the electrical properties of AI2O3/SiC
nanocomposites and their dependence on the microstructure, and thereby to

characterise the microstructural features of composites and microstructural changes
when subjected to high temperature heat treatment. It has been found that IS can be
used to (i) examine the conducting mechanisms of ceramic materials; (ii) determine
the content and orientation of conductive phase in insulating materials; (iii) determine
the thickness of oxide scalesformed at the surface of ceramic composites.
5.2 Experimental
5.2.1 Materials
A1203/SiC nanocomposites containing 5 vol %, 10 vol %, 20 vol % SiC (denoted

hereafter as 5SA, 1OSA and 20SA, respectively), and a monolithic alumina (hereafter
denoted as MA) were obtained by hot-pressing at 1600-1700°C, of which the detailed
Page66
fabrication procedures had been documented in the literature by Sternitzke et al. (1998)
and Wu et al. (1998). All the A12O3/SiCnanocomposites used in this investigation were
provided by the solid mechanics group of the Department of Materials, Oxford
University. Samples in plate form of about 10mmx6mmx2mm in size were cut from the
as-hot-presseddiscs.
5.2.2 Microstructure of Materials
The microstructure of the AI203/SiC nanocomposites consists of A1203 grains and
superfine SiC particles (Fig-5-1). The A1203grain size is 2- 4 µm, while the SiC particle
size is 100- 300 nm. The SiC particles are uniformly dispersed in alumina grains or at
the alumina grain boundaries.
Figure 5-1 TEM picture of 5SA, where the dark grey speckles are SiC particles (courtesy of
Houzheng. Wu, Department of Materials Science at Oxford University)
P
5.2.3 Oxidation Treatments
Before oxidation treatments all the A1203/SiC nanocomposite samples were polished
using 6µm diamond paste and then cleaned with acetone. Oxidation treatments were
conducted in a chamber furnace (Carbolite) at different temperaturesfrom 800°C to 1400

°C, for a period ranging from 1h to 100h. The heating and cooling rates were 3 K"minute"'
and 15 K. minute"1respectively.
5.2.4 Impedance Measurements
The measuring cells were about 10mm by 6mm by 2mm in size, with silver or platinum
paste on the two parallel surfaces of the plate specimen as electrodes. For impedance
measurements at temperatures above 400°C, platinum was used as electrodes. For

impedance measurementsof oxidised samples,platinum foils were applied directly to the
surfaces of the specimens as the electrodes.
Impedance measurements were made over a range from room temperature (R. T) to
800°C. Data were recorded over frequency ranges of 106 down to 10"4-10-1 Hz with 4
data readings per decade, using an 0.1 Vrms signal. A special rig as shown in Fig. 3-6
was usedfor impedancemeasurements

at high temperatures.
5.3 Results
5.3.1 ImpedanceSpectraOf the Monolithic Alumina (MA)
Pure alumina is a very good insulator. At low temperatures, say below 400°C, the
resistivity of alumina was so high that the electrical current through a thin plate sample
was too low to be measured using IS. As a consequence, at low temperatures MA
behaved like a pure capacitor: as shown in Fig. 5-2a, a straight line is almost vertical to
the x-axis, indicating the electrical response is almost purely capacitive.
Pa,ae 68
CHAPTERV NDC OF AL UMINA/SILICONCARBIDENANOCOMPOSITES
1.0x1011
-- -150
8.0x1010 2.5x108
" 4009C
A 6009C
2.0x109
"" 800`k
-"-
6.0x1010
"
1.5xI09
"
4.0x1010 1.0x100
"
5.0x108 "
0.
1.0x109 2.0x109 3.0x100 4axI0a 5.0x10a
Za
z
2.0x10104.0x101°
z
H
s.oxio'-
400'C
s.o1o'
. 600'C
4.0x10'- BOOS
N 3.0x10'-
1.0X107
8.0x107 8.0x107 0iiri

2.Ox1e 4.0x107 -r-ter
0 1x10'2 2x10'2 3x10'2 4x10'2 Sx10'2 8x10-2
z M
Fel
Ff] 25x1Qz
Mq 400°C
................ ................. ..........
2.Ok1(P
20c1Qe ""'A 1.@c10P

2.0x10'2
a.., ß... 600`C
1.5x10 12c10'
1.5X10"2
Ce
GP
10x10.2 awcf
l
Pb v ý!
5.0X10,3 P bo 5c, 0'3 a 4. k10°
ýýýýýio000QQe ;"
od f I"
O.D f "
0.0 0.0
10'3 10'2 10.1 10° 10ý 102 10,10,105 106 10ý 1v 10' ip 10' 0= 10p , 0' 103 10°
Frequency (Hz) Frequency (H
Figure 5-2 Impedance spectra of monolithic alumina: (a) complex impedance plot at 150°C;
(b) and (c) complex impedance plots at various measurement temperatures, note (c) is a
zoomed view of (b); (d) complex modulus plane plots; (e) modulus spectroscopic plots; (f)
comparison of modulus and impedance spectroscopic plots, where the hollow square
symbols represent the impedance data while the solid square symbols represent the
modulus data.
Pay-e69
At temperaturesabove 400°C, the resistanceswere still very large, but were measurable.
The complex impedance plots at different temperatures are given in Figs.5-2b-c.
Obviously there is only one semicircle at all measuring temperatures. The resistance
decreaseddrastically with increasing temperature. For this reason,the impedance spectra
at different temperaturesneed be displayed in different scales.By contrast, the diameters

in complex modulus plots, or the height of the M" peaks in the modulus spectroscopic
plots (Fig. 5-2d-e) at different temperatures, are similar. By reference to Fig. 5-2e, the
relaxation frequencies corresponding to the M" peaks at 400°C, 600°C, 800°C were
2Hz, 70Hz, and 1000Hz respectively. The capacitive effect remained almost unchanged
as the temperature was increased.Because o)=1/RC, the variation in relaxation frequency

(w) is entirely attributed to the changes of resistance. The product of frequency and
resistance ((°R) was calculated to be about 6x1010 for MA at all the measurement
temperatures.
The full-width at the half-maximum (FWHM) in the Z" and M" spectroscopic plots is
about 1.2-1.25 decades,which is very close to the ideal Debye peak that is 1.14 decades
(Morrison et al. 2001). In addition, the semicircles present in the complex plane are
nearly perfect, with only a slight depression. The M" peak and Z" peak coincide at the
same frequency as shown in Fig5-2f. All of these features indicate that the response of
MA is nearly ideal.
5.3.2 Impedance Spectra Of A1203/SiC Nanocomposites
5.3.2.1 ImpedanceSpectraOf 5SA
A1203/SiC nanocomposite with 5vol% SiC (5SA) exhibited very poor conductivity at
temperatures below 400°C. At low temperatures, the impedance spectra were similar to
that of MA. At the temperaturesabove 400°C, the impedancespectra of 5SA are given in
Fig. 5-3. The spectra of 5SA are similar to those of MA in a few aspects: only one
relaxation process can be seen in all forms of representations;the resistance decreased

drastically with increasing temperature; modulus spectra at different temperatures were
similar.
Paze 70
F-a] 1.0x109
ý1.
OXAI
" 400CC " 400
6
8.0x100 " 6009C
8 0x108
A . 800CC
"
6.0x10° 6.0x106
4.0x10° ' " "

'" 4.0x!08
2.Oxl0° " "

2.0x10° " ".
"
" "
0.0 ti
0.0 4.0ý 0° 8.0x10° 1.2x109 1.8x10° 0.0
0.0 4.010° 8 0x108 1.2x10 1.6x107
Z,
0.05 0.025
C
i d
0.04 0.020 46 """"
4009C ..
_ " ".
o 600CC --"-- 600°C ", .
0.03 v 800CC 0.015-
0.010 M
0.02- "
t
"
0 0.005 ý
° "
a °tr
0.01 o: "a ."
" """
ý
G4P
0
0.00 Oa 102 10.1 10° 101 102 103 104 105 100
0.00 0.02 0.04 0.06
Frequency (Hz)
M
a 40C 4.0x10"
5.0x106
4.OtI0y
ý
ý CCPC
°
ýc
° .0 we
o
s.a1o°
4.010"
o
O.ý.
0.025
0020
3.Ox10e 4.Oe105
3 o
% d 3 10° 0.015
o. .
N 3.0x105 1 ot
O
°1
2.0x106 ßl1
2on05 . ..... ....
P...
ö ° 2010° OO" 0.010
20x108 p
° ä o "
°..
t 0x10° 1.0x106 n
00
0° o 1.0[10° " 0.005
° 0C0
° 0
o o OO
Quo' COD
10"3 10.2 10"1 100 101 102 103 104 10s 106 102 10,10° 101 10' 102 10' 1 108
Frequency (Hz) Fr&UenO'(Hz
Figure 5-3 Impedance spectra of 5SA: (a) and (b) complex impedance plane plot, please
note the different scales used; (c) complex modulus plane plots; (d) modulus spectroscopic
plots; (e) impedance spectroscopic plots, note the different scales of y-axis; (f) comparison
of modulus and impedance spectroscopic plots, where small squares represent modulus
data, while the hollow circles represent the impedance data.
Page 71
However, with regard to the spectral features, 5SA is different from MA in the following
aspects:the semicircles in the spectraare more depressed;Z" and M" peaks are broader;
and Z" and M" peaks no longer coincide at the same frequencies. For example, FWHM
in the Z" and M" spectroscopic plots were measured as 1.9-2.0 decades, which are
much broader than those of MA. The frequencies corresponding to the Z" peak in
impedance spectroscopic plots were 7Hz, 400Hz, and 5000Hz, while the frequencies at
the M" peak were 20Hz, 900Hz, 10000 Hz, at 400°C, 600°C and 800°C respectively.
The average mismatch between the Z" peak and M" peak was about 0.2 decade. All
these features indicate that the response from 5SA was more distributed, or in other
words, more frequency-dependent,than MA.
The impedance spectra of IOSA were similar to those of 5SA, except that the resistance
was lower and the depression angles of impedance semicircles were noticeably larger
than that of 5MA. To save space, the results of the measurementson IOSA (please refer
to Appendix 11for the impedancespectra of 1OSA) are not to be presentedin detail here.
5.3.2.2 Impedance Spectra Of 20SA
A1203/SiC nanocomposites with 20 vol% SiC (20SA) had a distinctive behaviour
compared to those with lower SiC content. Firstly this material was `measurable' even at
R.T, because the resistance of the measuring cells was lower than 109 Q. However, at
low temperatures (e.g., below 300°C), the reproducibility of the impedance data was
poor, this could be a consequence of either a contact problem between electrode and
material, or the surface conductivity of moisture, or both.
Secondly, the electrical responsewas strongly dependent on the electrodes used. When
painted silver or painted platinum electrodes were used, there were, consistently, two
fairly well-resolved semicircles in the complex impedanceplots. But when platinum foils
were used as electrodes,there was only one depressedsemicircle, as illustrated in Fig. 5-

4a.
Thirdly the responsefrom 20MA was the most frequency-dispersedamong all the
composites investigated. At 400°C, FWHMs in the impedance and modulus
spectroscopicplots were about 2.0 decades,and the frequency at the Z" peak was
Paize 72
CHAPTERV NDC OF ALUMINA/SILICONCARBIDENANOCOMPOSITES
900Hz, while the frequency at the M" peak was 4500 Hz. The mismatch between them
was 0.6 decade,compared to 0.2 decade for the caseof 5SA.
Lii Fb
--"--room temperaWrowith siI celea'odes

20x108 room temperature vAih sil%erNeutodes
--o- room temperaturewith platinum foil electrodes 3.O,AOs
" 40CPC
1.5x10°
jV 2.Ox105
7.0x1OB .. """ "
1.00105 "" "

5.0x107 .
ý"
0.0
7.0x10° 2.0x108 3.0x10° 4.0x108 0.0 1.0xlo5 2.0x10 3.Ox10 4OxlCP 6.aa0° 6.0xlo5
z z
F-cl 001
a 1QQý ,2oo`C
oo9C
" 40OC
001 " -4001 ° Göf
°
°f
!E ý
hf
o"
1.01 " ° ° g 0.00 hf
00
0.00
1.00
0.00 0.01 012 0.03 0.04 i°
Frequency (Hz)
M
tard
loo^c r ".,40o°c
'ý
2ox1d
a 241O^
aaw'
t. sztd
°
I
b}py
N e.oXia'
0' 7.ox7d tann,
00.
90n0ý
o° 30k1O°
4.m1o' s.mxla °
°
4 ola
o
0.0 0.0 °
' 1
1O. 1O. 1o. 1o' IC? 102 1C iW i ao o0
lo- 'O a2 w' i& lö' io id id w' iCf
Frequency(Hz) Frequency(Hz)
Figure 5-4 Impedance spectra of 20SA: (a) complex impedance plane plot; (b) a partly
zoomed view of (a); (c) complex modulus plane plots; (d) modulus spectroscopic plots; (e)
impedance spectroscopic plots, note the different scales of y-axis; (f) comparison of modulus
and impedance spectroscopic plots
Page 73
5.3.3 Impedance SpectraOf Oxidised Samples
5.3.3.1 Impedance MeasurementsAt Room Temperature
A1203/SiC nanocomposites can be oxidised at elevated temperature in air. During
oxidation, SiC is oxidised into silica and carbon dioxide.
SiC +202=SiO2+C021' (5- 1)
Silica stays where it is produced by oxidation, or further reacts with alumina to form
mullite (Luthra and Park, 1990), whereas carbon dioxide evaporates. Since SiC is a
conductor, while Si02 and mullite are insulators, the electrical properties of the oxide
layer are expected to be different from those of the bulk materials.
For the case of 20SA, there was a drastic increase of impedance after oxidation. As can
be seen from Fig. 5-5a, at R.T, the Z" peak of the sample that had been oxidised at 1100
°C for 1h was more than four orders of magnitude higher than that of the same sample
without oxidation. In the meantime it has been noticed there are two fairly resolved arcs
in the modulus spectra of the oxidised samples,as shown in Fig 5-5b&c.
It is also very interesting to note that not only the resolving of the arcs, but also the
diameterof the low frequencyarc dependedon oxidation temperatureandduration,asis
shown in Fig. 5-5c.
Basedon the aboveobservation,we can confidently attribute the LF circular arc to the
oxide layer formed after oxidation, while a HF semicircle is attributed to the bulk effect.
The diametersweredeterminedaccordingto thecriteria describedin section4.3.
Paize74
CHAPTER V NDC OF ALUMINA/SILICON CARBIDE NANOCOMPOSITES
a 4.Ox1 Oxidised Q
Withoutoxidation 1.5x 107
Q
ö
i
3.0x1
ä
ö 1.0x107 N
N
2.Oxl
o°
Q0 5.0x108
1.0x1
Po
13
-D--l nn
0-* 10-3 101 10-1 100 101 102 103 104 105 10
Frequency (Hz)
51
M"
M'
F-cl
1400t for60 h and thinnedbygrinding
0.0015 14004Cfor 1h
1100`Cforlh
0.0010
A0
0.0005 ",
o U.
"l, o am
M-0: M OEM
Aý
0.0000
0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030
AA'
Figure 5-5 (a) Impedance spectroscopic plots for the samples with and without an oxidation
treatment, please note the different scales used in y-axes; (b) modulus complex plane plots
for the same samples; (c) the low frequency arcs for the samples oxidised under different
conditions.
Paze 75
Table 5-1 gives a summary of semicircle diameters of both LF and HF for specimens
oxidised at various temperatures.The diameter of the LF arc increases with an increase

of oxidation temperature, whereas the diameter of the high frequency semicircle is nearly
constant. This indicates that little change occurred in the bulk specimens, in contrast
there was clearly a monotonic change of the oxide layer formed at the surface as a
function of oxidation temperature.
Table 5-1 Semicircle diameters for 20SA oxidised for Ih at various temperatures.
Oxidation LF semicircle diameter (DLF) HF semicircle diameter (DHF)

DLF/D
condition .....................
F
._.......
NOT ö. 1h ...
_ 2.5x10' 3.7x10' 6.8x10
950°C Ih 1.0x104 3.7x10'2 2.7 x10'3
1100°C lh 3.0x10 3.7x10'2 8.2x10"3
1250°C lh 6.Ox10-4 3.6x10'2 1.7x10'2
1325°C Ih 1.2x10'3 3.6x10"2 3.4x10"2
1400°C Ih 2.2x10'3 3.5x10'2 6.3x10'2
5.3.3.2 Impedance MeasurementsAt High Temperatures
The impedance spectraof the oxidised 20SA at high temperaturesare displayed in Fig. 5-
6. The sample had been oxidised at 1100°C for Ih. It is important to note that the spectral
resolution, i. e., the separation of semicircles in the complex impedance plots increased,
while the spectral resolution in the complex modulus plots decreasedas the measuring
temperature was raised. At 200°C, the spectral resolution in the complex impedance plot
(Fig. 5-6a) was very poor, but that in the complex modulus plots was better as shown in
Figs. 5-6e-f. When the temperature increased to 400°C, there appeared two clearly
resolved semicircles in the complex impedanceplots in Fig. 5-6b, while the separation of
semicircles in the complex modulus plots disappeared in Fig. 5-6 e-f. At 600°C the
spectral resolution was obviously better than that at 400°C, by comparing Fig5-6b with
Fig. 5-6c. Meanwhile, the FWHM in modulus spectroscopic plots decreased with
temperature.
Page 76
CHAPTER V NDC OF AL UMI NA/SILICON CARBIDE NANOCOMPOSITES
a a "
6.Ox10ý
5.Oxie
4.Ox1e
"4
° 600'C
2DOC
1.5x106
1.0x108
" 400f
60M
"
" °
" N"
3.Ou1e
o "
"
2.0XIe " 5.0x105
"".. " """
1.OX1e ý öopO°o
0.0
0 0-ý 5.0x105 tool 8 1.5x106 2.0x106 2.5ý 06 3.0x106
2.0x107 4CO e 6.0ý e 6.O;Ae 1.0X10°
Z
Fc]
6.0x10°
o
400
600ì
o
[d] 006
-- 200t
4009C ffQ
°Q
P
::: -
°°°° , (
0 O
ý
O
f 93
0+0 a "
0.0 2.0x104 4.0x10° 6.0x104 6.0x10° 1.0x105 Dam
001 01 1 10 10D Y00 10A0 1( 1®0
Z Frequency (Hz)
Fee]0.040
171
"
0.035
0.00 °
" 4001
0.030 "
-°-600"c 0.006 .
0.025 -f- 200°C 0
0.020 M" 0.004

0.015- A 406-'
06 60(t
0.010- f---f " -A--- 0.00 0 '
@_" 20(r,
0.
0" A f 4"_ A 'fyffýf
. "
fý . "'
0.00
0 010 0.008 0.010
0.04 0.05 0.000 0.002 0.004 0.006
0.01 0.02 0.03
M M.
Figure 5-6 Impedance spectra of oxidised 20SA: (a), (b) and (c) complex impedance plots,
note that (b) is a zoomed view of (a) and (c) is a zoomed view of (b); d) modulus
spectroscopic plots; e) and f) complex modulus plane plots, (f) is a zoomed view of (e). The
measuring temperatures are indicated in each figure.
Paue 77
CHAPTERV NDC OF ALUMINA/SILICONCARBIDENANOCOMPOSITES
5.3.4 Dielectric Properties At Room Temperature
As mentioned previously, it is difficult to measure the impedance of 5SA and 1OSA
materials at R.T, becauseof the extremely high resistance of these materials. In order to
obtain a stable current from materials with an extremely high resistance to determine the
dielectric properties of the 5SA and 1OSA materials, impedance measurements were
taken by connecting a1 GS) resistor in parallel with the specimens as shown in Fig. 5-7a.
The capacitanceof the resistor (3x10"12F) is less than 10% of that of the 5SA and IOSA
specimens (5x10.11F) to be measured,while the resistanceof the resistor (1GS) ) is much

lower than that of the 5SA and IOSA specimens. The capacitive effect of this circuit is
mainly contributed by the specimen, while conduction mainly occurs across the resistor
in parallel with specimens. The modulus spectra of this circuit (Fig. 5-7b) show that all
semicircles have a depressedpart at high frequency. The conduction of the circuit (the
imaginary part of the electrical modulus) arises from the conduction of the resistor at
lower frequency and conduction from the specimen at high frequency. Measurements of
IOSA specimens without being connected to the resistor-generated spectra at high
frequency as indicated in the amplified diagram in Fig. 5-7b (denoted a). No spectra could
be obtained at low frequency since the impedance at low frequency is too high to be
detected. The increasing conductivity of specimens results in more depressedsemicircles
in the modulus spectra.
Fig. 5-7c shows the dielectric constants obtained from measurementsshown in Fig. 5-7b
as a function of the content of SiC in the A1203matrix. The small error bars indicate that
the data are quite consistent with repeat measurements.The dielectric constant increases
with increasing content of SiC in A1203.According to Fig. 5-7c, the dielectric constant is
11.8 for pure A1203,which is a little higher than the value of 6-10.2 that can be found in
the literature (Buchanan, 1991), which might be due to the edge effect.
Page 78
L1
RESISTOR
SAMPLE CELL
F-b]
6
MA//R --c-- 53A//R
10SA1/R 20SA//R
5 --ix-
o 1OSA
-+-
4 X
0.4
°Oo
0 3
Xx- 0.2
5.0 5.5 6.0

1 a
0
24 `ö 8 10 12
M'x 102
FC-1 30
21
$
2(
1(
SiC content vol%
Figure 5-7 (a) A circuit with a resistor connected in parallel to the measured sample; the
resistance of the resistor (R) is 1x109 S2, while its capacitance is about 2--4x10'" F; (b)
modulus spectra obtained at room temperature from measurements of the circuit shown in
(a) with specimens containing different amounts of SiC; (c) dielectric constant as a function
of SiC content, where the standard deviation is based on six sets of impedance
measurements for each sample.
Page 79
5.4 Discussion
5.4.1 Equivalent Circuit Models And IS Parameters
Rey RB
Cel CB
Figure 5-8 Equivalent circuit for monolithic alumina. Indices: el= electrode effect, B=bulk.
Our experimental results demonstratedthat monolithic alumina (MA) gave a nearly ideal
Debye response and there was only one semicircle attributed to the bulk effect (Fig. 5-2).
At high temperatures (say 800°C), a tail appears at low frequency in the complex
impedanceplane (Fig. 5-2c). This is most likely due to the Schottky barrier in the material
at the electrode / material interface. From this observation, the equivalent circuit of MA
is supposedto be like the one given in Fig. 5-8. No grain boundary effect was seen in the
impedance spectra, this might be either due to very thin grain boundaries in hot pressed
MA materials, or because of the similar electrical properties of the grain boundary and
grain. Therefore from the viewpoint of electrical properties, the MA is monolithic with
little inhomogeneity. This also explains why MA exhibits a nearly ideal behaviour during
impedance measurements.
Based on the microstructure of A1203/SiC nanocomposites(seeFig. 5-1), the equivalent
circuit should be like that in Fig. 5-9. Based on this model we may expect two
semicircles, but the experimental results showed that all the A1203/SiC nanocomposites
without an oxidation treatment had only one relaxation process in impedance spectra.
The relaxation frequencies of 5SA and MA at various temperaturesare listed in Table 5-
2. The relaxation frequency does not change greatly when a small amount of SiC is
incorporated. Therefore it can be assumed that the impedance spectra of 5SA and 1OSA
mainly reflect a response from alumina. No separatespectrum corresponds to the SiC,
Paize80
becauseSiC is conductiveand it is supposedto behavelike an electricalconnector.This

situation can also be ascribed to the relaxation frequency of SiC which is much higher
than the upper frequency limit of the impedanceanalyser (107 in this case).
Rs Rn i
CA
Figure 5-9 Equivalent circuit for alumina/silicon carbide nanocomposites. Indices: el=
electrode effect, S=silicon carbide, A=alumina between electrodes, Al= alumina between
SiC particles.
Table 5-2 Comparison of relaxation frequencies for MA and 5SA
Temperature Relaxation frequency* of Relaxation frequency of

(°C) MA (Hz) 5SA (Hz)
400 27
600 60 400
800 500 5000
* All the relaxation frequencies here are the frequencies corresponding to the Z" peaks in
impedance spectroscopic plots.
The conductivity relaxation frequency (f) of a material can be calculated from:
(5- 2)
27cRC 2;r.---,,
where6 is the conductivity,c is the dielectric constantof thematerialand Eo=8.85x10"14

F/cm. The conductivity of SiC is about 10-1-10"3(Qcm)-' (Kingery et al. 1976c), while
the dielectric constant of ceramics are normally about 5-20. Thus the relaxation
Pa e81
frequency of SiC can be calculated to be 1x 108- 2x 109Hz. For this reason the relaxation
of SiC cannot be detected in the frequency range for IS measurements,which is normally

106_10"4Hz.
However for the caseof 20SA, the relaxation frequencieswere much higher than those of
MA, 5SA and IOSA. For example at 400°C, the relaxation frequency of 20SA was 900
Hz, more than two orders of magnitude higher than those of MA and 5SA at the same
temperature. This implies the conduction mechanism in this material might be different
from other materials.
5SA
22
MA
20 ide layer
18
/
16
" 20SA
/
14 i
8 12 16 20 24 28
104/T
Figure 5-10 Arrhenius plot of resistances for different materials
Fig. 5-10 shows the Arrhenius plots of the resistances of 20SA, 5SA and MA as well as
the oxide layer formed at the surface of 20SA, where we can seethat the linear slopes are
similar for all the cases except for bulk 20SA. According to the linear slopes in the
Arrhenius plots, the activation energies of electrical conduction can be calculated as
shown in Table 5-3, where we can see the activation energy for the 20SA specimen is
only abouthalf of thosefor othermaterials.
Paze82
Table 5-3 Activation energies for different materials
Materials MA 5SA Oxide layer Bulk of 20 SA
Activation energy (eV) 0.93 1.05 1.12 0.49
The low activation energy and high relaxation frequency may be due to the percolating
network of SiC formed in the alumina matrix when the SiC content is 20 vol%.
When a composite is composed of insulator as a matrix and conductor as a filler, the
overall electrical behaviour of the composite is governed by percolation behaviour

through the formation of a continuous network of conductive particles in the insulating
matrix. It has been observed that there exists a threshold filler volume fraction, f*, above
which the conductive filler forms a continuous network. When the fraction of the filler is
close to f*, the resistivity changes drastically. The threshold volume fraction strongly
dependson the particle sizesand particle distribution (Kim et al. 1998).
In the A1203/SiC nanocomposites,the grain size of alumina (RA) is about 3µm, while the
grain size of SiC particles (Rs) is about 200 nm (refer to Fig. 5-1). So the ratio of RA/Rs is
about 15. If all of the SiC particles are located at grain boundaries, the threshold volume
percentage can be estimated to be about 10 vol% according to Kim et al. (1998). In the
present A1203/SiC nanocomposites,not all of the SiC particles are segregatedat the grain
boundaries. It may be reasonable to assume that the threshold volume percentage is
between 10 vol% to 20 vol%.
Page83
1010
109
.5
108
N
107
0
106
05 10 15 20 25 30
SiC content vol%
Fig. 5-11 Resistivity at 400 °C of the alumina/silicon carbide nanocomposites as a function of

SiC content.
Fig. 5-11 shows the resistivity of the A12O3/SiCnanocomposites at 400 °C as a function
of SiC content. The composite shows a more significant change in resistivity when the
SiC content is increasedfrom 10 vol% to 20 vol%. At the same time, the mapping of the
silicon element in A1203/ SiC nanocomposites illustrated that SiC particles were
clustered in the alumina matrix and these clusters are more densely populated in 20SA
(Fig. 5-12b) as compared to 5SA (Fig. 5-12a). This may imply that the threshold volume
percentage for the A1203/SiC nanocomposites with a microstructure as in Fig. 5-1 is

between 10vol% to 20 vol%. The formation of the percolation of SiC particles can
explain why 20SA is distinctive in conductivity and relaxation frequency as compared

with the composites with lower SiC content.
Pake 84
1ý1
Figure 5-12 Silicon element mapping of SiC phase in a) 5SA and b) 20SA, where white
pixels represents the silicon content in the alumina matrix.
Paize 85
5.4.2 Spectral Resolution
Spectral resolution is an important factor when different physical processesneed to be

de-convoluted. As discussed in chapter 4, the spectral resolution in the modulus spectra
depends on the difference in resistances. Based on Figs. 5-6 a-c, the ratio of the
resistanceof the oxide layer (RoX)to that of the bulk (RBU)was calculated to be 50 and 12
respectively at 400°C and 600°C. According to these data, the ratio of R,,,, /RBu was
calculated to be 900 at 200 °C, by assumingthe activation energies of the oxide layer and
the bulk are constant in the range of 200-400'C. Therefore it can be understood that the
better resolution in the modulus spectra at lower temperatures was a consequence of
larger difference in resistances. For an ideal case, as has been discussed in chapter 4,
when the resistance ratio is larger than 25, a distinct resolution should appear in the
modulus spectra. However, current experimental results indicate that for a non-ideal case
such as 20SA, the resistance difference needs be greater than 50 to get a resolved
spectrum.
On the other hand, the spectral resolution in the impedance spectra not only depends on
the difference between capacitances as has been pointed out in chapter 4, but also is
associated with the conductivities of different microstructural components in the

material. The higher the conductivities, the better resolved are the spectra. According to
the analysis presented in chapter 4, whether or not the responses from different
microstructural factors can be resolved in impedance spectrum depends on how well the
conditions specified in Eqs 4-9 a-b, i. e., R;/(I+ca/(o;)=:>0, are fulfilled. Eqs. 4-9 a-b are
obviously better approximated when R; is smaller, given w; to be constant. If R; is very

large and in the same order of magnitude with, or even larger than the co/co;in value, we
can see that Egs.4-9 a&b will never be fulfilled, and therefore the spectrum will never be
resolved. As we can see from Figs. 5-6 a-c, the impedance spectra were shown to be
better resolved at higher temperatures, because the conductivities of both the bulk and
oxide layer increase with temperature. This is a reason why IS is more powerful in
characterising a material that has a reasonable conductivity. It is impossible for IS to
distinguish the electrical responsesfrom different microstructural factors if the material
is totally insulating.
Page 86
5.4.3 Dispersion Of Electrical Responses
5.5
22
20SA
5.0
C 20i C
rn 4.5
rn
cu
c -------------------- co
0 18 O 0
y 4.0
9 Cl)
N
1 a5
16
O' 3.0
14
300 400 500 600 700 800
Temperature(°C)
Figure 5-13 Dependence of depression angle of MA and 20SA on the temperature.
As discussed in the literature review (Chapter 2), the non-ideality of a response from a
real material, i. e., the distribution of relaxation times, or a dispersion of electrical

response, depends on either chemical, or physical or geometrical inhomogeneities in the
material. The extent of departure from an ideal response can be expressed by a
depression angle (a) of the semicircle in the complex impedance plot or the exponential
index (n) in a CPE. The larger the depressed angle (i. e., the smaller the n value), the
greater the dispersion. The departure from ideality also results in the broadening of the
Z" and M" peaks in spectroscopic plots and the displacement of M" peaks from
corresponding Z" peaks.
Page 87
1E
1E
° 14
0) 1i
C
C 1(
C
O_
E
I-
E
0.
VJ IV IJ LV
SiCcontent(wI%)
Figure 5-14 Depression angle at 400°C of A1203/SiC nanocomposites as a function
of the SiC content.
Based on the understanding detailed in section 2.7, the depression angle a is related to
exponential index n by the equation: a=(1-n) Tc/2,whereas the exponential index, n, can
be obtained by simulation using `Zview' software. Here the depression angle of the
A12O3/SiCnanocompositesdepends not only on the temperature, but the SiC loading as
well. Fig. 5-13 demonstrates a strong dependence of the depression angle on the
measuring temperature. The dispersion increases with increasing temperature, this is

probably because at high temperatures more heterogeneities are activated which make
the conductivity relaxation times more distributed. The variation of depressionangle as a
function of the SiC content is shown in Fig. 5-14. The depression angle increases as the
SiC content increases. The increase in the dispersion of composites probably mainly
results from the incorporation of impurities, i. e., the increase in chemical inhomogeneity.
The impurity substance that is most likely introduced is Si02. It is well established that
there is always a thin layer of Si02 formed at the surface of SiC particles, thus Si02 can
Pa-ze88
CHAPTER V NDC OF AL UMINA/SILICON CARBIDE NANQCQMPOSI TES
be introduced as a kind of impurity as a consequenceof an addition of SiC particles into
the alumina matrix. The higher the SiC loading in composites, the more the Si02 is
introduced, then the more dispersedthe materials will become. As mentioned earlier, the
SiC phase behaves like a conductive connector, so it is not expected to change the
distribution of the relaxation times of the materials.
5.4.4 DependenceOf Dielectric ResponseOn SiC Content
Experimental results showed there was a strong correlation between modulus spectra and
the content of SiC in the A1203 matrix as shown in Fig. 5-7, indicating that modulus
spectroscopy can be used to determine the content of the conducting phase in an

insulating matrix nondestructively.
C2
F-a
] Cl SiC C2 A1203
HHH.HHH
-1 Hr.......
--I H
-11
Figure 5-15 a) The idealized microstructure of the SiC/A1203 composites and b) its
equivalent circuit model.
To simulate the dielectric property of the composites containing both conductive and
insulating materials, the microstructure of the AIZO3/SiC nanocomposites is simplified as
cubic units packed as shown in Fig. 5-15a where SiC cubes are packed inside A1203
uniformly. The conductive SiC cubes behave as connection lines between capacitors with
A1203 as the dielectric material. Where there is no SiC present along the electrical field
direction, A1203 behaves as a capacitor with small area but large thickness. There is no
edge effect being considered in this case. SiC cubes are assumed to be of uniform size
inside the A1203.Assume the particle size of the SiC (i. e. each edge of the SiC cubes) is
Pake 89
1, the volume percentageof the SiC particles in the alumina matrix is V1, Based on the
.
model in Fig. 5-15, the dielectric constant of the composite, c, i. e. the capacitanceof a
cubic configuration of Icm by 1cm by Icm is derived as:
F3 Z
Vpe`
C=o(1+
E- VP",) 5-3
1-3 Vpcc
where we can see the dielectric constantof the compositesis independentof the SiC
particle size, becausethe edge effect is neglected here.
As the dielectric constant of the matrix (A1203),so was determined to be 11.8 in Fig. 5-7c,
the dielectric constants (s) of the composites containing different volume percentageof
the SiC particles can be calculated according to Eqn. 5-3 and these values are shown in
Table 5-4.
Table 5-4 Dielectric constants of nanocomposites
Materials Calculated F Measured e The difference between measured c and

calculated c
MA 11.8 11.8 -
5SA 12.7 14.7 2.0
1OSA 14.0 17.9 3.9
20SA 17.4 27.0 9.6
Apparently the predicted data are lower than the experimental data becausethis model is
too simplified to give an accurateprediction of the capacitanceof the composites. Factors
such as the irregular shape, non-uniform distribution of SiC particles and edge effects
have not been taken account in this model, so further work is needed to predict the
capacitanceof the A1203/SiC composites accurately.
5.4.5 Nondestructive Detection Of Oxide Layer In A1203/SiC Nanocomposites
Previous research on the oxidation of SiC/A1203 composites was carried out at

temperatures above 1300°C because it is difficult to use weight gain measurementsor
Pa e 90
CHAPTER V NDC OF AL UMINA/SILICON CARBIDE NANOCOMPOSITE-S
electron microscopy for characterising the oxide scales formed at lower temperatures
(Luthra and Park, 1990; Sciti and Bellosi 1998). In the present study modulus
spectroscopy was used to examine oxide scales formed at temperatures from 800 to
1400°C. The thickness of oxide scales can be predicted from the measurements of
modulus spectra. For a sample of thickness t with an oxide layer of thickness S at its
surface (Fig5-16a), the equivalent circuit is like a series layer model (Fig. 5-16b).
ROx Rbk
F-b
1
-C3- --cl]-7
---iC Cbk
Fig. 5-16 SiC-AI203 composite of thickness t with an oxidation layer of thickness S, À'
representing the electrode area. Indices: ox=oxidation layer, bk=bulk of 20SA.
Here, we can have:
C"x°2SA (5-4)
Cbk 6n A\
Cbk (5-5)
_
t- 2ý
C _ ý`tA (5-6)
where C.X: capacitanceof the oxide scale; Cbk: capacitanceof the bulk specimen without
the oxide scale; Co: capacitanceof empty spaceoccupied by the specimen.
The low frequency (LF) semicircle diameter is given by
28
DI. = Ce/ Cox= (5- 7)
F 6nst
Pane 91
The high frequency (HF) semicircle diameter is given by
D,,,; =Ce/Cbk=
t- 28
(5-8)
ehk t
The ratio of the LF semicircle diameter to HF semicircle diameter is given by:
2 CSEhk
D,,,: / DHIý_ (5- 9)
EX(t- 28)
Since the thickness of oxide layer S is much smaller than that of the sample t, the above
formula can be approximated by:
2öSnk
D,.,,/DHF_ (5- 10)
Boxt
Then the thickness of oxide layer is calculated as
t(DLr,
28= /DHI:) (5- 11)
Ebk
where t co, and sbk can be assumed to be unchanged during oxidation. The values of
DLF/DHFare listed in Table 5-1.
The dielectric constant of the 20SA sample was determined as 27.0 (Table5-4).
According to Eqn. 5-11, Eoxcan be obtained if the thickness of the oxide scale is known.
Fig. 5-17 shows an optical micrograph of the cross-section of oxide scales formed at the
surface of the 20SA sample, which had been oxidised at 1400°C for 64h. The bright
phase layer is the mullite/alumina composite and the dark phase layer is mainly alumina
(Luthra and Park, 1990; Sciti and Bellosi, 1998)). Alumina and mullite have similar
dielectric constants (6.2-6.8 for mullite and 6-10.2 for alumina) (Buchanan, 1991). SEM
observations showed a large amount of porosity in the surface layer.
Paize92
I iulk
Scale
Fig. 5-17 Optical micrograph of cross-section of sample 20SA after oxidation at 1400°C for
64h.
Ea
]
'1ti ''I
1.5
0 1.0
x
0.5 _; ".
ý U.
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
M' X102
Fig. 5-18 (a) optical micrograph (b) modulus spectrum of the 20SA sample after oxidation
at 1400°C for 64h. The oxide scale thickness was reduced to -20µm by polishing.
Paize93
The oxide layer on one of the faces of this sample was totally removed, meanwhile the
oxide scale on another face was thinned by polishing. According to the optical
micrograph (Fig. 5-18a), the thickness of the thinned oxide layer was determined to be
20µm. The overall thickness of this plate sample is 1.1mm. Based on the modulus
spectrum of this sample (Fig. 5-18b), DLF/DHFwas obtained approximately as 0.15. By

using Egn. 5-11, c, was calculated as 3.5, which is lower than the dielectric constant of
alumina or mullite. This could be due to the high porosity in the oxide scale.
I\
cO
C -0,,
ý-J-1
-2
6789
104/ToK-1
Fig. 5-19 Plot of the oxidation thickness S against 1/T, with the linear slope in the range of
1250-1400°C determined to be -2.3 by linear regression.
The thicknessof the oxide scalesformed at different oxidation conditionswas calculated

basedon Eox=3.5 and Ebk=27(Table 5-4). Fig. 5-22 shows a plot of the calculated
thickness of oxide scales (8) vs. inverse temperature. To compare the activation energy
with the reported value in the literature, the activation energy in the temperature range
1250-1400°C is calculated as 378kJ/mol, which is close to the reported value of 450
kJ/mol (Luthra and Park, 1990). The thickness of the oxide scales is very difficult to
measureusing SEM due to the low contrast in the image between the oxide scale and the
substrate. Meanwhile it is also very difficult to measurethe thin oxide scale using optical
microscopy, whereas IS is sensitive to a very thin oxide scale that is formed at as low
Page 94
temperature as 800°C. Therefore, IS is a powerful tool for determining oxide growth due
to oxidation of ceramics.
5.4.6 TheEffect Of ParticleOrientationOn IS
In composite materials, the preferential orientation of the dispersed particles is
unavoidable during the manufacture, if they are non-equiaxed, e.g., in plate or fibre form.
The orientation of fibres or plates in composites is an important issue, becauseit leads to
anisotropy in both mechanical and electrical properties of the materials. IS has been
found to be sensitive to the content of conductive SiC particles. Likewise IS may be also
sensitive to the orientation of particles.
Since SiC particle in A1203/SiC nanocomposites are equiaxed, to investigate the

dependence of IS on the orientation of the dispersed conductive phase in an insulator
matrix, the clay/graphite composites containing different amount of flaky graphite were
fabricated for impedance measurements.As shown in Appendix III, the experimental
results clearly show that in the direction parallel to the orientation of the particles, the
resistancewas lower and dielectric constant was higher.
(11)
to particle
tatrix
Fig. 5-20 An idealised model for clay/graphite composites.
To explain this phenomenon, an idealised model is given in Fig. 5-20 for the
clay/graphite composites. The graphite particles are assumed to be in the form of a

Paw 95
square plate. The average distance between conductive particles is shorter in the lateral
direction than in normal direction (81>82), then the probability of forming a continuous
path along the lateral direction is higher than in the normal direction, thus the
conductivity is higher in the lateral direction.
The explanation for the higher dielectric constant in the lateral direction is not so
straightforward. Based on the idealised model, the dielectric constant should be the same
along both directions. However, as mentioned in the previous section, factors such as the
irregular shape,non-uniform particle distribution and edge effects will also have effects
on the dielectric constant. The difference between measured e and calculated s can be
considered to be an extra contribution from these factors which cannot be explained by
the idealised model. As can be seen in Table 5-4, the extra contribution increaseswith
SiC loading, i. e., it increasesas the distance between particles is decreased.This can be
usedto explain the observation of higher dielectric constant in the lateral direction.
Form the viewpoint of NDC, the significance of this finding is that we can use the IS
technique to detect non-destructively the orientation of conductive particles in composite
materials. Recently a study (Wang et al. 2000) has been published on the effect of
whisker orientation on the electrical properties of Si3N4/SiC whiskers composites.
Similarly to the results presented here, it was found that the impedance spectra exhibited
a significant dependenceon whisker orientation.
Monolithic alumina and A1203/SiC nanocompositescontaining small amounts of SiC are
very insulative at room temperature, thus it is impossible to get measurable signals in the
whole frequency measuring range. Impedance measurements have to be made at high
temperaturesto obtain a full spectrum in the whole frequency range.
A1203/SiC nanocomposites exhibit only one relaxation process. The conductive SiC
phase dispersed in the alumina matrix does not have a corresponding semicircle in the
Page96
impedance spectrabecauseit hasa very high relaxation frequency (outside the measuring
range of the instrumentation used here).
The response of monolithic alumina is very close to an ideal Debye response, while
A1203/SiC nanocomposites exhibit considerable frequency-distribution. The depression
angle in the complex impedanceplane shows a clear increasing trend as the SiC content
is increased.As the electrical responsebecomes more distributed, Z" and M" peaks in
the spectroscopic plot become broader, and the M" peak is separatedfurther from the Z"
peak.
20SA has distinctive electrical properties as compared to those with lower SiC loading,
which is attributed to a percolating network of SiC particles formed in the material. The
activation energy of 20SA is much lower than that of MA, 5SA, and the oxide layer,
indicating that 20SA has a distinctively different conduction mechanism from the other
materials.
There were two relaxation processesin the impedance response after 20SA is oxidised,
one of which corresponds to the bulk effect, the other is from the oxide layer. The
resolution of the modulus spectra at low temperature is better, due to a larger difference
in resistance between the bulk and oxide layer at low temperatures. By contrast, the
impedance spectra are better resolved at high temperatures because of the higher
conductivity.
The effect of the SiC in composites on the dielectric properties of the composite was
examined by room temperature measurements.A microstructural model was developed

to calculate the dielectric properties of composites with insulating matrices and
conductive inclusions. Impedance spectra strongly depend on the orientation of
conductive particles, a similar microstructural model was used to explain the effect of the
orientation of conductive particles.
Oxidation kinetics of 20SA was examined by room temperature measurements. The
activation energy was determined to be 378kJ/mol, which is close to the reported value
using other methods. The oxide growth was predicted based on models developed for
calculating capacitanceof multi-layer structures.
P
CHAPTER V NDC OF AL UMINA/SILICON CARBIDE NANOCOMPOSITE 4
This study demonstrates that IS is a useful technique for non-destructive evaluation of
ceramiccompositeswith insulatingmatricesandconductiveinclusions.
Page98
CHAPTER VI NDC OF THERMAL BARRIER COATINGS
6 Nondestructive Characterisation Of Thermal Barrier

Coatings Using IS
6.1 Introduction
Thermal barrier coatings (TBCs) have been extensively applied to engine components to
allow engines to operate at higher temperatures; hence achieving higher thrusts and
greater thermodynamic efficiency of the engine (Lee et al. 1996). Due to its low thermal
conductivity and comparatively high thermal expansion coefficient, together with its
chemical and thermal stability, yttria stabilised zirconia (YSZ) has been widely used as a
thermal insulating material (Bartlett and Maschio, 1995; Sergo and Clarke, 1998; Wolfe
and Singh, 1998).
A TBC normally consists of a top coat of YSZ and a bond coat typically of the formula
MCrAIY where M represents either Co, Ni or Fe. As an oxygen ion conductor at high
temperature, the YSZ top coat does not block the oxidation of the underlying bond coat.
A protective alumina layer forms at the interface between the bond-coat and the YSZ due
to oxidation of the bond coat at high temperature (Tawancy et al. 1998). Therefore, the
bond coat between YSZ and the substrate superalloy is designed not only to promote
adhesion of the YSZ to the substrates,but also to provide oxidation resistance (Meier and
Gupta, 1994).
The oxidation of the bond coat is a dominant factor leading to the failure of TBCs
(Demasi-Marcin et al. 1990; Wu et al. 1989). As the oxidation proceeds,more and more
aluminium element diffuses outward to the bond coat/YSZ interface to form an alumina
layer. When the aluminium in bond coat is being depleted, other elements, such as
chromium and nickel, can be oxidised at the same time, and then form mixed spinel
phases. The spallation of YSZ was found to be associated with the phase conversion of
the initially-formed a-alumina to a mixed spinel phase (Wu et al. 1989; Schillington and
Clarke, 1999), one of the reasons is that the interfacial fracture resistance of the
TBC/spinel interface is lower than that of the TBC/a-A1203 interface (Schillington and
Clarke, 1999).
Page99
CHAPTER VI NDC OF THERMAL BARRIER CQ TINE S
As the alumina layer grows to a certain thickness, usually about 10µm, the TBC fails,
either at the bond coat /YSZ interface or in the YSZ close to the interface due to thermal
mismatch between the alumina layer and the bond coat. Failure of TBCs will expose the
underlying metallic components to hostile environments, which might result in a
catastrophic failure of the component. For application of TBCs on engine components, it
is vital to predict the coating lifetime which is controlled by this oxide growth at the bond
coat/YSZ interface. Finding a method of measuring and monitoring the thickness of the
oxide layer non-destructively will provide a way of predicting the lifetime of TBCs;
henceavoiding unexpectedfailure of TBCs and prevent any disastrous consequence.
By making impedancemeasurementsat 469K, Ogawa et al. (1999) found that impedance
spectra were sensitive to the thickness of different layers in TBCs. In the present work,
impedance measurementswere made at room temperature and 350°C in order to monitor
the oxide growth in TBCs due to oxidation at 1100°C for a period up to 1500hr. At a
measuring temperature of 350°C, YSZ is a moderate ionic conductor whereas the

alumina layer is a good insulator. The large difference in electrical properties of the
alumina and YSZ lead to a distinction of the electrical responseof the alumina layer from
that of YSZ in the impedance spectra of TBCs. In order to examine the detailed
microstructure of TBCs after oxidation, different equivalent circuit models representing

different microstructures of TBCs were established to simulate impedance spectra of
TBCs after oxidation for different periods.
6.2 Experimental Procedure
6.2.1 Specifications Of Materials
Table 1 The Composition of Haynes 230 Alloy (wt%)

Ni Cr W Mo Fe Co Mn Si Al C La B
57** 22 14 2 3* 5* 0.5 0.4 0.3 0.10 0.02 0.015*
*maximum, ** asbalance.
The Haynes 230 alloy (for composition seetable 1) with a thickness of 2mm was sprayed
with a HVOF MCrAlY bond coat (38.5 wt%Co-32 wt%Ni-21 wt%Cr-8 wt%A1-0.5
Page100
wt%Y) and a Metco 204 NS TBC top coat (8 wt% yttria stabilised zirconia) at
Sermatech,Lincoln, UK.
6.2.2 Oxidation Treatment
The plates with thermal barrier coatings were cut into 10 mm x 10 mm squares with a
precision diamond cutting saw at low speed to avoid any mechanical damage to the
coating layers. The samples were placed in a chamber furnace at 1100°C for periods
ranging from 2h to 1500 h. The heating and cooling rates were 3 K"minute" and 15
K"minute-1 respectively. Four samples for each oxidation time were prepared, of which
one sample was examined by SEM coupled with energy dispersive x-ray spectroscopy
(EDS) and the remaining three sampleswere used for impedance measurements.
6.2.3 Impedance Measurements
The measurementcells were about 10mm x 10mm x 2mm in size, which were obtained
by painting silver electrodes on the top coat (YSZ). The painted area was 9mm x 9mm
and was I mm away from the edge to avoid possible current leakage. The substrate
(Haynes 230 alloy) acted as the other electrode. Impedance measurementswere made at
room temperature and 350°C in air. Data were recorded over frequency ranges of 106
down to 10"3Hz with 5 readings per decade,using a1 Vrms signal. To ensure the steady
state of the specimen, impedance measurementswere made after the samples had been
placed in a furnace at 350°C for 30 minutes. ZView Impedance Analysis software

(Scribner Associates, Inc.) was used to simulate the impedance spectra based on
equivalent circuit models.
6.3 Results
6.3.1 Microstructure Of TBCs
There were two coating layers in the as-received TBCs samples: the top coat and bond
coat, as shown in Fig. 6-la. The top coat was about 300 pm in thickness, while the bond
coat was about 100 µm thick. The typical as-sprayed top coat exhibited about 10-15
Page 101
vol% porosity (Haynes et al. 1996) and the t'-phase zirconia in the top coat has a micro-
cracked structure (Miller et al. 1981 & 1983). For this reason we observed there were a
lot of pores and small cracks in the top coat. By contrast, the bond coat was fully dense.
After oxidation at 1100°C, an oxide layer was formed at the interface between the top
coat and bond coat, as shown in Fig. 6-lb.
Fa]
bQ
Figure 6-1 SEM images for (a) cross section of an as-received sample; (b) cross section of a
sample oxidised for 1500h at 1100°C.
Page 102
20h
200h
400h
9ooh
1500h
Figure 6-2 SEM observation of the oxide layer growth between the top coat and bond coat.
Please note that all the images were taken at the same magnification.
Pau 103
ý`ý TopciaL
14
! ti 'I rxle5
. a.: ol
ýr MNx
+Y
d'
Ct ,
f '_
" fir`
.
_______
Lbj ýM
EDS microanalysis (b)
"
Ry ...
'
ýi. i7 -.
fir. - ý. ^ â: ` Tiýý
.ice' `_, ,"`"º1+

[+1.
j,,, -...
IMIRNPw
lmua
Rstdi.: 1
Figure 6-3 Back-scattered images of the cross section of TBCs after oxidation (a) for 100h
and (b) 800h at 1100°C, showing the presence of the alumina layer (dark layer) and mixed
oxide (grey), the magnification being different between a and b.
Page 104
Fü:
L 3o sp esf t: JQ Rena ini n9: 03
Rea L 57s 3_'r. ['tad
< -. 2 4. -440 Kelp 10.1 r

FS= 32K eh 257= 652 cts
r'IEM1:
X-RAY: 0- 20 IceV
bI Live:
L ive: 100s
1003 Preset:
Pr'e_ct: 100s R nairiir O
u Real
Peal: : 220
220SS 55% Oeed
4.940 keV
,ch 257=
Figure 6-4 Energy dispersive mirco-ananlysis of (a) a selected point in the dark layer and
(b) a selected point in the grey region as shown on Fig. 6-3b.
The oxide layer grew along the rough interface between the top coat and the bond coat.
As can be seen in Fig. 6-2, the oxide layer appeared to be continuous and became
Page 105
thickened with prolonged oxidation time. No cracks were observed within the oxide
layer.
50h
r.
' 100h
_y
ti ý,
200h
, ,_-
ýý
1300h
tý v
.r-
. will
Figure 6-5 Back-scattered images of the oxide layer, the grey regions being mixed oxides
and dark layers representing the alumina. All the images were taken using the same
magnification.
Figs.6-3 a&b show the back-scatteredimages of TBCs after oxidation at 1100°C for 100
and 800 hours respectively. EDS microanalysis (in Fig. 6-4) indicated the dark layer
between the top coat and bond coat is of nearly pure alumina (Fig, 6-4a), while the dark
grey region consist of mixed oxides, i. e. alumina, chromium oxide, cobalt oxide and
nickel oxide (Fig. 6-4b). This observation is in agreement with previous studies on
oxidation of TBCs (Haynes et al. 1996; Wolfe and Singh, 1998). In most areas, the
alumina layer is continuous whereas the mixed oxides are present between the alumina
layer and top coat, with a discrete distribution and non-uniform growth. Images with
enhanced contrast are displayed in Fig. 6-5. It appeared that both the alumina layer and
the mixed oxides grew with increase in oxidation time. In the samples that have been
oxidised for less than 200 h, the alumina layer is not uniform or continuous; in some
parts the mixed oxides were formed between the top coat and the bond coat. For the
samplesthat have been oxidised for more than 400h, the alumina layer is continuous and
relatively uniform.
Page 106
CHAPTERVI NDC OF THERMAL BARRIERCOATINGS
6.3.2 Impedance SpectraAt 350°C
The impedance spectra obtained from impedance measurements of various samples at

350°C are shown in Fig. 6-6.
The differencein the impedancespectrabetweenTBCs oxidised for lessthan 200 hours

and more than 400 hours (hereafter denoted as L200 and M400 respectively) is apparent.
For M400 samples, three relaxation processes were shown in the phase angle plots
(Fig.6-6a) where their relaxation frequencies were about 2x105Hz, 4x102Hz and 10 Hz.
However, for L200 samples, there were only two relaxation processes,at about 2x105Hz
and 4x 102Hz.
In the complex impedanceplots (Figs. 6-6c-d), two semicircles are present and a tail-like
curve for both M400 and L200 samples. It can be easily found that the high frequency
arcs for all samples were similar in diameter, the resistances were consistently in the
range of 5-9x103 Q. However the low frequency semicircles increased gradually with
oxidation time from 0 to 200h, and then jumped abruptly from 5x104 ) at 200h oxidation
time (Fig6-6c) to greater than 1x107S2at 400h oxidation time (Fig. 6-6 d).
For L200 samples,two relaxation processescan be identified in both complex impedance
and phase angle plots. However, for M400 samples, the complex impedance plot shows
only two semicircles, while the phase angle plot shows three relaxation processes.The
relaxation frequencies for M400 samples are estimated as 4xlO5Hz and 6Hz for the high
frequency semicircle and low frequency semicircle in the complex impedance plot,
which match the 2x105 Hz and 10Hz relaxation processesshown in the phase angle plot.
The relaxation process at 4x102 Hz shown in the phase angle plot does not appear in the
complex impedanceplot.
In the complex modulus plots (Fig. 6-6e-f), again we can see the distinction between
L200 and M400 samples: there are two well-resolved semicircles for M400 samples,but
only one semicircle for L200 samples.The peak frequencies are about 6x105 Hz and 5Hz
(see Figs. 6-6 b, e) respectively for high frequency semicircle and low frequency
semicircle.
Page 107
aa 80
z.ox1o'ý
1.5x10
-"-Oh
/11
6- 60
"-200h po
9 4o
t. Ox1Q
-°-4h
-. -too0h
--15ooh
5.0x10 1
20 yh /
"
i_
0 P 10-2 101
1 10° 10' 102 103 10° 105 106
Frequency (I-h) Frequency (Hz)
d
Ic 3.0x107
zaýC ".
- --800h
e-1000, do
-. -a, 1.5x10 ..
-v-5m -°--150011
---2OCh 2Ox107 _, gOqt tatd ý.
. 400h
ro,
s.atd
°" 0%o
z.o,no' 1.0xi tonte adla
p,
0.o 0
0.0 2.ono' 4.0x10' e.oxlo' 8.0x10' 1.0x107 2.0x10? 3.0x107 4.0x107 5.0x107
z
z
HI
6x103
5x104 ý"
5x10'
4x10'3
-0-400h 0
-ý-Oh 3x10 -ý -icy ".
!2 -"- 200h
2x10'4 "
-o- 400h
-A- 800h o-
1000h 1x10' bo _ mow
-- oo
-+-1500h wo a
0 0
2.0x10'3 4.0x10'3 6.0x10 8.010"3 1.Ou1O 0.0 2.0x10-' 4.0x104 6.0x10' 8.0x104
M M
Figure 6-6 impedance spectra at 350°C: (a) phase angle plots; (b) modulus spectroscopic
plots; (c) complex impedance plots for L200 samples; (d) complex impedance plots for
M400 samples, with an enlarged presentation for high frequency arcs; (e) complex modulus
plane plots and (f) enlarged presentation for low frequency semicircles
It is important to note that the diameter of the low frequency semicircle in the complex
modulus plot increases monotonically with oxidation time, while the high frequency
semicircle does not appear to vary with oxidation time. Similarly the low frequency peak
Page 108
CHAPTER VI NDC OF THE MA/ BARRIER COATINGý
height in the modulus spectroscopic plot also increases with oxidation time, indicating
that the capacitive effect of the low frequency process decreaseswith the oxidation time.
With the evidence presented in Fig. 6-6, and in combination of our SEM observations
presented in 6.3.1, we can quite confidently ascribe the low frequency process to the
oxide layer and the high frequency process to the top coat (YSZ). What is responsible for
the intermediate frequency process (i. e. at4x102 Hz) will be dealt with in detail later in
this chapter.
6.3.3 Impedance SpectraAt Room Temperature
At room temperature (R. T), the reproducibility of impedance data was found to be much
poorer than that at 350°C. This is due to the low conductivities of the samples at R.T.
Materials with a conductivity of lower than 10"5(S2.cm)"l are thought to be susceptible to
the influences of other unpredictable factors (Kingery et al. 1976c), such as moisture in
the atmosphere and impurities that were probably brought in by the painted electrodes.
Typical impedance spectraobtained by R.T measurementsare shown in Fig. 6-7.
At R. T, quite different from the impedance spectra at 350°C, the as-received (i. e.,
denoted as Oh) sample showed only one relaxation process in the phase angle plot (Fig.
6-7a), one semicircle and one low frequency tail in the complex impedance plot (Fig. 6-
7b). All the oxidised samples exhibited two relaxation processes,one at a frequency in
the range of 5-100Hz and the other at a frequency in the range of 0.004-1Hz. The
scatteredrelaxation frequencies were probably a consequenceof the poor reproducibility
of the impedance data. Nevertheless, similarly to the caseof the measurementsat 350°C,
it was consistently found that the high frequency semicircular arc in the complex
impedance plot was similar in diameter, while the low frequency semicircular arc was
very different. For instance, the low frequency semicircular arc for L200 samples were at
least two orders of magnitude smaller than those for M400 samples.
Similarly to the case of 350°C spectra,the complex modulus plots at R.T had two
semicircles for M400 samples, but only one semicircle for L200 samples (Figs. 6-7c-d).
The high frequency semicircle in the complex modulus plot became more depressed
(Fig. 6-7c), and the high frequency peak became broader (Fig. 6-7e) when the oxidation
Paze 109
CHAPTER VI NDC OF THERLIAL. BARRIER COATINGS
time was extended above 800h, indicating the top coat microstructure might change as
well during heat treatment.
[7] Fb-
° "
1.0.10'
GJ 3"V-
00
0000 a0-
. Oh
200h
60- 2ovf0' " IOQ
6 ýý l
---
-o-200h
40 Jýý"ýi ý"-400h
800n
N
Co "-
L 1300h öam° +ö, o
O -ý-
CL 20 - ,
oýýY_
o00oooo. ' 1500h) °
"" -"-
'"iýa ý"N ýýaaa a"a a.
0
ink in, In2 1,3 ; n4 ns 1r 5

2.0x10" 4.0,10' 6.0,10' 6.0x10" 1.0.10" 12x10"
Frequency (Hz)
Fd-]
Hi 6x10'3
5x10
4z10'3
- -Oh
-O- 200h
"- 400h
°- 800h A
-a- 1300h
3x10 - 1500h Q It -.
". ___ ___. _.. !E Q40__... o
-. "__. 400h
2x10 '- os "-+_"_ý_-_o_. ý.
. "___ý-ý. ý,.. -: _.8001
ý-QýAÖ-QQ
1900h
_ý _.. 1500h
2.0x10'' 4.0x10"3 B.Ox10"3 1.0x10'= 2.0x104 4.0X104 6.0x10 8.0x10 1.0x103
hA'
le
-"- Oh
-0- 20th
1-71 -. -- Oh
-0--200h
--°- 800h
-0-m
ý_ 130th M" -'-- 1300h
1.5x10" t
1.0x10"3
5.Ox1o
0.0
iä' 102 1ö' Id Id Id
10'2 10-' 10010' 102 102 14 103 10° Frequency(Hz)
Frequency (Hz)
Figure 6-7 Impedance spectra at R. T: (a) phase angle plots; (b) complex impedance plots,
with an enlarged presentation for high frequency arcs; (c) complex modulus plots and (d)
enlarged presentation for low frequency semicircles of modulus spectra; (e) modulus
spectroscopic plots for the whole frequency range; (f) modulus spectroscopic plots focused
on high frequency processes.
Page 110
6.4 Discussion
6.4.1 Oxidation Kinetics
Based on the analysis of SEM images as shown in Fig. 6-2, the oxide layer thickness was
found to be a function of oxidation time (Fig.6-8a). By plotting oxidation layer thickness
against the square root of time, we can obtain two linear regions in the plot (Fig. 6-8b):
one is below an oxidation time of 25h, the other above 50h. Thus the oxidation kinetics
of the bond coat can be assumedto follow a parabolic law:
8= kt (6- 1)
Fa-] 12
ýc
:L
iE
V_
'
s "
d
ag
"
-200 0 200 400 600 800 1000 1200 1400 1600
Oxidation Time (h)
12
10
U
E8 a
y6
a
U
f. 4 2
a)
n
5
Squareroot of time(h12)
Figure 6-8 Oxide thicknessdetermined by image analysisplotted vs. (a) time and (b) square
root of time.
Page111
CHAPTERVI NDC OF THERMALBARRIERCOATINGS
According to Fig. 6-8b, the parabolic rate constant is calculated to be 1.8x10-12cm2s"' for
the initial 25h and 0.3x10.12cm2S'' for the later oxidation. The faster initial oxidation rate
is commonly known as a transparent effect in oxidation where Ni or Cr-rich oxide may
form before a steady-state alumina film is established. This is in agreement with our
SEM observation. Our SEM observation revealed that mixed oxides (grey areas in Fig. 6-
5) were formed in the initial stage. In the initial oxidation, the alumina layer was not
continuous, other elements such as Ni, Cr, etc., in the bond coat were oxidised to form a
mixed oxide along the interface. In the later stages the alumina layer was literally
continuous, which is supposedto prohibit the diffusion of other elements, except around
defected areas, such as cracks, pores or other forms of defects, which could allow other
elements to pass across the alumina layer. For this reason mixed oxides grew like
mushrooms in the caseof long oxidation period (in Fig. 6-5).
6.4.2 Equivalent Circuit Models
Based on microstructural observations, the electrical conduction paths in TBCs are

illustrated schematically in Fig. 6-9, where the alumina layer is continuous for M400
samples, but discontinuous for L200 samples. Electrical conduction first passes the
electrode/top coat interface and then the top coat itself. The top coat is composed of
polycrystalline YSZ, where grains and grain boundaries would respond differently to
electrical signals in different frequency domains (Steil, et al. 1997; Badwal, 1995),
therefore two R-CPE elements should be used for grain and grain boundary, respectively
in the top coat. Then the electrical conduction is divided into two paths: path one and
path two. For M400 samples, path one and path two meet again in the alumina layer
becauseit is continuous, whereas, for L200 samples,path one and path two will not come
together until they reach the bond coat. In the light of the analysis of conduction paths,
the comprehensive equivalent circuits corresponding to the two cases are illustrated in
Fig. 6-10.
Obviously we could not use these equivalent circuits directly to fit the impedance data in
order to get meaningful IS parameters. Normally it is hoped that there is a separate

relaxation process corresponding to each R-C (or R-CPE) element in an equivalent
circuit. Since only two or three relaxation processesappeared in the impedance spectra
for TBCs, this means that electrical responsesfrom some microstructural factors may
Page 112
CHAPTER VI NDC OF THERMAL BARRIEKCQA_TING'
overlap, or do not give significant contributions to the impedance in the measuring

frequency range. To understandbetter the impedance spectra and generate meaningful IS
parameters by fitting impedance spectra, the equivalent circuits shown in Fig. 6-10 need
be simplified.
Firstly, equivalent circuits can be simplified by putting aside the electrical responses
from all the interfaces, e.g., the YSZ/electrode, alumina/YSZ and alumina/bond coat.
This is because, in most cases, the responsesfrom the interfaces either occur at low
frequencies that will be clearly separated from the bulk effect, or make only minor
contributions to the impedance. Secondly, when there are parallel conduction paths, the
electrical conduction always finds the easiest path. Thus the equivalent circuit can be
further simplified to the easiestpath.
Electricalconduction
path
coat
Mixed Oxide
iina
COAL
Electrical conduction
path
coat
Mixed Oxide
nina
coat
Figure 6-9 Schematic of electrical conduction paths for (a) M400 samples and (b) L200
samples.
Page 113
Electrode/YSZ YSZ grain

aI YSZ grain
interface boundary
Path
PE
interface
Electrode/YSZ YSZ grain

YSZ grain
interface boundary
Path
YSZJalumina Alumina layer Alumina / bond

Pathtwo
interface coat interface
Figure 6-10 Equivalent circuits for (a) M400 samplesand (b) L200 samples
Page114
By neglecting all the interfaces in the equivalent circuits, there are only four components
left for consideration: YSZ grains and grain boundaries in YSZ, the mixed oxide and the
alumina layer. For the M400 samples,the alumina layer is in series connection with the
mixed oxide, while for L200 samples, the alumina layer is connected in parallel to the
mixed oxide. As mentioned earlier on, three relaxation processes appeared in the
impedance spectra of M400 samples at 350°C, of which the high frequency and low
frequency relaxation processes corresponded to the top coat and the alumina layer
respectively. As a consequenceof this, the intermediate frequency process can only be

attributed to a mixed response from the grain boundaries in YSZ and the mixed oxide,
which will be discussedin detail later on in this chapter. Therefore the equivalent circuit
shown in Fig. 6-1Oacan be simplified to that shown in Fig. 6-1 la.
For the L200 samples, as shown in Fig. 6-9b, the mixed oxide and alumina layer are
parallel connected. The mixed oxide can be assumed to be semi-conductive, and would
show much higher conductivity than alumina at high temperatures. Therefore the
conduction through the mixed oxide in L200 samples at 350°C (path one in Fig. 6-10a) is
an easy path and the responsefrom alumina is shielded. The low frequency process in the
impedance spectra of L200 samples measured at 350°C has the same relaxation
frequency as the intermediate frequency process of M400 samples, indicating they
represent a similar physicochemical process. For these reasons, the high frequency
relaxation can be attributed to the top coat and low frequency relaxation process to the
mixed response from the grain boundaries in the YSZ and the mixed oxide layer. The
equivalent circuit shown in Fig. 6-1Ob can be simplified that shown in Fig. 6-11b
At R.T, both the mixed oxide and alumina are assumed to be very insulating and the
electrical responses from the mixed oxide and alumina can overlap. In this case the
equivalent circuit can be simplified to that shown in Fig. 6-11c.
In practice, the impedancespectra usually cannot be fitted simply by employing ideal RC
elements, because of the distributed nature of the electrical response from realistic
materials. In terms of the various representationsof impedance data as shown in Fig. 6-6
and Fig. 6-7, it is not difficult to seethat the impedance and modulus spectra are rather
depressed, so it is necessary to use constant phase elements (CPEs), rather than ideal
capacitances,in the equivalent circuits.

Page 115
Rc Rm R.
CPE, CPEý, CPEa
R, Rm R,
Fcý
lýOX
CPE, CPEm CPE, CPEOX
Figure 6-11. Equivalent circuits of (a) M400 samples at 350°C; (b) L200 samples at 350°C
and (c) all samples at R.T, t--top coat, m=mixed oxides, a--alumina and ox=oxide layer.
6.4.3 Fitting Of Impedance Spectra
The equivalent circuit models shown in Fig. 6-11 were used to fit the impedance data.
These models fit very well the experimental data as shown in Figs. 6-12a-c. The IS
parameters that can be obtained by fitting are R, A and n. R represents the resistance,
while A and n are the parameters in CPE (Y=Aae ). Since the fitted results obtained
from three sets of experimental data are different, error bars are used to indicate the
standarddeviation of the three sets of fitted data.
Page 116
CHAPTER VI NDC OF THERAMAI BARRIER OATING4
8.Ox10-3
-Fitted results
° Experimental data
6.0x10'3
4.0x10"3
/.
2.0x10''
0.0 i
0.0 e n, n n-3 a no 01 an n-2
3.0x10°
-Fitted results
o Experimentaldata
2.0x104
N1
0.0 ,r; -- ' i---i----I

0.0 1.0x104 2.0x104 3.0x104 4.0x104 5.0x104 6.0x104
ZI
a 5.0x1010
4.0x10
3.0x1010
N
- Fitted results
2.0x1010 " Experimentaldata
1.0x1010
1.0x1010 101D
Figure 6-12 Fitted experimental data of (a) 1000h oxidised sample measured at 350°C, (b)
200h oxidised sample measured at 350°C (c) 1000h oxidised sample at R.T.
Page 117
The IS parameters obtained from the fitting of impedance data measured at 350°C are
shown in Fig. 6-13. Although the standard deviation for the resistance data is very large,
the resistance of the top coat appearsto have a fixed pattern: it increases drastically for
the first 10 hours, and then appears to gradually increases with oxidation time. In
contrast, the resistancesfor the oxide layer (alumina) and mixed responsesdo not appear
to show a clear trend.
Top coat
V.W
0.65-
080-
0.75-
s
0.70-
0.65-
0.60
0 200 400 600 800 1000 1200 1400 lean
-200
Ndabon Time(h) O,ddationTime(h) O)ddation
Time (h)
Alumina layer
,i
rl? e -e
Gmdatlon
Time(h) CmdationTime(h) OoädaAion
Time(h)
Mixedresponse
,2x,0f 0.80
1.ozio' 0.75 --arreceh. d

ddd
8.mno' ' 0.70
6.oxno' 0.65
}
0.60
I
0 200 400 800 400 10D01200 1400 1800

O,ddatcm Time (h) (Mdation Time(h) OOddatlon
Time (h)
Figure 6-13 IS parameters obtained from the fitting of impedance spectra measured at
350°C by using equivalent circuits as shown in Figs. 6-11a & b.
Page118
The n values for the top coat and oxidation layer do not appear to vary with oxidation
time, while the n value of the mixed response show a steep increase in the initial
oxidation, then it decreasesgradually with oxidation time
The parameter A of the top coat shows a drastic decreasein the first 10h of oxidation,
followed by a less significant and unsteady change. The parameter A of the mixed
response shows a sharp drop within the first 200h oxidation, then stays constant during
heat treatment, while an apparent decreasing trend is shown for the parameter A of the
alumina layer.
Top coat
ý:ý
`
i--
1.85
Ozdaton Time (h) O)ddatoh Time (h) OddMon Time (h)
Oxidation layer
Ono
2
tI
1ono"
QO ..
.1
OAdationTime (h) O»»daIonTime (h)

O daaon
Time
(h)
Figure 6-14 IS parameters obtained by fitting the impedance spectra measured at R.T by
using the equivalent circuit as shown in Fig. 6-l1c.
The IS parameters resulting from fitting the impedance data measured at R.T are shown
in Fig. 6-14. Here, of particular interest is that the n value of the oxide layer demonstrates
a rapid increase within the first 400h of oxidation and then stays unchanged afterwards,
while the parameter A of the oxide layer decreasesabruptly in the first 200h of oxidation
time and showslittle changeafterwards. Other IS parametersdo not seemto show a
clear trend.
Page 119
Ch1APTER VI NDC OF THERMAL BARRIER COATINGS
It is important to note that the parameter A does not have a defined physical meaning. By
contrast, capacitance representsthe capability of accumulating electric charge, which is

related to the geometric aspectsof a given microstructural factor. The capacitance of an
ideal capacitor is frequency-independentand can be defined by:
A
C=c ° (6-2)
"E
where c,, is the vacuum permitivity: 8.85x10"'4, c is the dielectric constant, Ao is the area
and 1 is the thickness of the microstructural factor of concern. Therefore capacitance is
more meaningful than the parameter A. However for a real material with a non-ideal
behaviour, the capacitance is frequency-dependent, i. e., the lower the frequency, the
larger the capacitance. In order to make the capacitance of a non-ideal system
comparable with that of an ideal system, the capacitancecorresponding to the frequency

at the top of the arc is regarded as the equivalent capacitance. According to Jacobsenet
al. (1996), the equivalent capacitancefor a R-CPE circuit can be calculated by:
C= (RI-nA)'ln (6- 3)
where R is the resistance,n and A are parameters for the CPE.
Based on the fitting results shown in Fig. 6-13 and Fig. 6-14, the capacitances of the top
coat and oxide layer at R.T and 350°C were calculated employing Eqn. 6-3. The results
are presented in Fig. 6-15. The capacitance of the top coat measured at both RT and
350°C (Fig. 6-15a) shows a sudden drop in the first 10h of oxidation. Afterwards the
capacitance does not seem to change much with oxidation time. The mixed response
(Fig. 6-15b) shows a steep rise in capacitancefor the first 50h oxidation, then decreases
gradually with oxidation time. The capacitance of the alumina layer measured at 350°C
decreaseswith oxidation time (Fig. 6-15c). The capacitance of the oxide layer at R.T
(Fig. 6-15d) drops quickly in the initial stageand then decreasesgradually.
Page 120
F7a] F-b]
"_.
o-
room temperaturemeasurement
--- 350°measuemeM
2.0x10-10
I ... " .C
10 i°"
Co
K
() 1wo-10 0° °
CL
C)
5.0x10""
1400
O,ddation Time (h) Oddation rime (h)
Fcý L1
LL_ U-
N
d
Odation Time (h) OOddat

on Time (h)
Figure 6-15 Equivalent capacitances of (a) top coat derived from both R. T and 350°C
impedance data; (b) mixed response at 350°C (c) alumina layer at 350°C; (d) oxide layer at
RT.
6.4.4 Interpretation Of The Impedance Spectra And Fitting Results
6.4.4.1 HF Relaxation Process- ResponseOf Top Coat
There are three pieces of evidence supporting the attribution of the high frequency
relaxation process to the top coat (YSZ bulk). Firstly, the high frequency relaxation
process of TBCs is similar to that of a free-standing top coat, for which the impedance
spectrum will be presented later on (in Fig 6-17). Secondly, the relaxation frequency
reflected in the impedance spectra is in agreement with the calculation based on the data
of conductivity and dielectric constant. The relaxation frequency can be calculated as:
f" =1_a 27CEQE L6-4)

2nRC
Page 121
CHAPTER VI AMC OF THERMAL BARRIER COATINGS
where a is the conductivity and other symbols have the normal meanings. The
conductivity of a free-standing top coat was measured to be 6x10-6 S"cm, c0 8.85x10-

14F/cm,the dielectric constant (e) for YSZ can be assumed as 30 (Bonanos et al. 1987),
then the relaxation frequency is calculated as 3x105 Hz, which is good agreement with
the value of 2x 105Hz that is observed in the impedance spectra.
Thirdly, the capacitancededuced from IS parameters is in accordancewith the calculated
value using Equ.6-2. The electrode area of the measuring cell is 0.9 cm2, the thickness of
top coat is about 300 µm, c=30. Thus the capacitance of top coat is calculated to be
8.85x10'', which is lower, but fairly close to the value of 1-1.2x10-10 derived from IS
parameters for R.T impedance spectra, as shown in Fig. 6-15a. This further justifies the
ascription of the high frequency relaxation process to the top coat.
In the as-received TBCs, the top coat (YSZ) is mainly composed of the t' phase (Miller
et al. 1981&1983). According to the Zr02-Y203 phase diagram (Scott, 1975), for the
YSZ containing 8 wt % (4.5 mol %) of Y203, the equilibrium phases should be in
tetragonal (t) phase plus cubic (c) phase region. However, on the rapid cooling of the
molten droplets of YSZ during plasma spraying, the diffusional process, which leads to
the formation of the equilibrium phases: a low-Y203 -content t-phase and a high-Y203-
content c-phase, is prohibited. Therefore the c-phase in the cubic single-phase field at
high temperature undergoes a displacive transition to a non-equilibrium phase that is
crystallographically identical to the tetragonal phase, but with higher Y203 content. This
phase is commonly referred as t'-phase (Scott, 1975).
The t' phase has a higher electrical conductivity than the cubic and tetragonal phase
(Drennan and Auchterlonie, 2000). The t' phase is a metastable, distorted fluorite-type
phase (Badwal, 1998). When thermally aged, it rapidly dissociates into c-phase and t-
phase (Miller et al. 1981 &1983). Thus the dissociation of the t' phase is expected to be
accompanied by an increase in electrical resistance. The oxidation temperature used in
this research is 1100°C. According to the phase diagram (Scott, 1975), the equilibrium
phase at this temperature for 8 wt% PSZ is c-phase plus t-phase. So it is possible that the
t' phase transformed into the equilibrium phases when heat-treated at this high
temperature. Therefore it can be speculated that the drastic increase of top coat resistance
in the initial heat treatment (refer to Fig. 6-13) might be due to the dissociation of the t'
Page122
phase. This also can explain the drop of capacitance of the top coat (Fig. 6-15a) during
initial 10h oxidation. According to the discussion in chapter 5, the dielectric constant
increaseswith the content of conductive phase. When the t' phase is transformed into c-
phase and t-phase, the conductive phase is reduced and then the dielectric constant is
expected to become smaller.
The X-ray diffraction analysis of the top coat after heat treatment revealed that prolonged
oxidation led to an increase in monoclinic (m) phase as shown in Fig. 6-16. Since m-
phase is less conductive than the tetragonal and cubic phases (Muccillo and Kleitz,
1996), the increase in the m-phase may explain the gradual increase in resistance of the
c- andt-phase
o m-phase
C oo
N
CD
b
20 30 40 50 60 70
2e (deg)
top coatastheoxidation is prolonged.
Figure 6-16 X-ray diffraction analysis of top coat after heat treatment using various heat
treatment schemes (a) at 1 100°C for 2h; (b) at 1100°C for 100h and (c) 1150°C for 2000h.
The responseof the top coat correspondsto a small semicircle in the complex impedance
plot and a large semicircle in the complex modulus plot, becausethe top coat has a high
Page123
conductivity and low capacitanceas compared with the oxide layer. At R.T, the modulus
spectra (Figs. 6-7 c-e) demonstrate that a prolonged oxidation time results in a more
frequency-dispersedresponseof the top coat. When the oxidation time exceeds 800h, the
peak corresponding to the top coat becomes abnormally broadened. It is quite likely that
it consists of two overlapping peaks (Fig. 6-7 e), becausewe can seethe semicircle of the
top coat in complex modulus plot appearsto separateinto two semicircles. This might be
due to the presence of a considerable amount of monoclinic phase when the oxidation
time is prolonged. Becausethe electrical response from monoclinic zirconia is expected

to be different from those from tetragonal and cubic zirconia, even an additional
semicircle is possible (Muccillo and Kleitz, 1996).
The degradation of the top coat (YSZ) is strongly related to the polymorphisms of
zirconia. For example it is well established that the t' phase is the major reason for the
superior TBC performance of YSZ (Miller, 1987). The phase transitions: t' phase -+ t-
phase + c-phase, and t-phase -* m-phase, have promotion effects on the degradation of
the top coat. The results presented here indicate that impedance measurements can be
used to monitor the phase changes in the top coat. Therefore IS should be informative
regarding the degradation of the top coat. Further detailed work, however, is needed to
establish an explicit correlation between IS parameters and the degradation status of the
top coat.
6.4.4.2 Low FrequencyProcess- ResponseOf Oxide Layer
The electrical response from the oxide layer is at low frequencies. At 350°C, the
relaxation frequency of the oxide layer is about 10 Hz. However the relaxation frequency
at room temperature is much lower and varies from one measurementto another. In order
to verify the assignment of the low frequency relaxation process to the oxide layer, the
capacitanceof the oxide layer is estimated basedon microstructural observations. For the
case of the 1500h oxidised samples, the observed average thickness of oxide layer is
11.4µm, the dielectric constant of oxides can be assumed to be about 10. Using Eqn. 6-3,
the capacitance can be calculated to be 7x10"10 F, which is close to the equivalent
capacitance of 1.06xI0"9 derived from the IS parameters for the R.T impedance spectra
(see Fig. 6-15c).
Page 124
CHAPTER VI NDC OF THERMAL BARRIER COAT]N GS
At 350 °C, the response from the oxide layer is split into two responses,one from the
mixed oxide and the other one from the alumina layer, because the differences of
relaxation frequency between the mixed oxide and the alumina layer are expected to be
very large. At 350°C, the relaxation frequency of the mixed oxide, which will be
discussed in detail later on, should be at a very high frequency, therefore the mixed
oxides may not contribute to the low frequency relaxation process.
Table 6-2 shows the equivalent capacitance of the low frequency relaxation process of
TBCs at 350 °C and R. T respectively. On average, the capacitance of the alumina layer at
350°C is about twice that of the oxide layer at R.T. The dielectric constant of the oxide
layer could not be doubled from RT to 350°C, so the increase in the capacitance of the
low frequency relaxation process must be due to the decreasein thickness. This justifies
the speculation that the low frequency relaxation process in the R.T measurement
corresponds to the overall oxide layer, including both alumina and mixed oxide, while
the low frequency relaxation process at 350°C `senses' only the alumina layer. The
capacitance of low frequency process at 350°C is mainly from the continuous alumina
layer. The effective thickness will be smaller than the overall oxide thickness and the
capacitance is expected to be higher than that at R.T. This is in accordance with the
capacitance values derived from the IS parameters as shown in Table 6-2. This also
suggests that the thickness of both alumina and mixed oxides can be detected by IS
measurementsat different temperatures.
Table 6-2 Equivalent capacitance of low frequency process at R. T. and 350°C
Oxidation time Equivalent capacitance (F) of low Equivalent capacitance (F) of low
(h) frequency process at R.T. frequency process at 350°C
400 1.8x 10'9 3.7x 10'9
800 1.5xl0'9 2.5x 10'9
1300 1.1x109 1.9x10"9

1500 1.0X10-9 1.8X10-9
Page125
6.4.4.3 Intermediate Frequency Process -A Mixed Response
In the impedance spectra measured at 350°C, an intermediate frequency relaxation
process can be observed at about 400Hz. As mentioned earlier on, when the alumina
layer is not continuous, the response from the alumina layer is suppressedby the short-
circuiting of the mixed oxide. There are only two relaxation processes for the L200
samples at 350°C, of which the low frequency process is at about 400Hz which
corresponds to the IF process in the impedance spectra for the M400 samples. The
intermediate frequency process was ascribed to the mixed oxide by Ogawa et al. (1999),
as it was observed there was a mixed oxide layer present between the top coat and bond
coat in the microstructure of TBCs. However this ascription is questionable.
In order to make clear whether or not the mixed oxide is mainly responsible for the
intermediate frequency relaxation process, a free-standing top coat (YSZ), in which the
presence of mixed oxide can be ruled out, was measured using two identical silver
electrodes.The free-standing top coat was as-sprayedand provided by the same company
as our TBCs. The impedance spectra of the free-standing top coat (Fig. 6-17) was very
much the same as that of Oh TBCs sample, except the electrode effect which was absent
with silver electrodes.This is understandablebecause a silver electrode has been found to
show less polarisation resistance. The high solubility of oxygen in silver induces the
spreading of the electrode reaction zone over the whole electrolyte interface, thus
reducing the blocking of oxygen ions at the interface (Drevet et al. 2000). This also
indicates that the low frequency tail in the impedancespectra for L200 samples should be
due to the response of the top coat/mixed oxides interface. The electrode effect curve
becomes larger as oxidation time is prolonged. This implies the YSZ/mixed oxides
interface is changing with oxidation time. It is possible that more meaningful information
can be extracted by analysing systematically the change of the electrode effect. This,
however, is beyond the scope of the presentstudy.
The free-standing YSZ has a relaxation process at about 800 Hz, which is very close to
that of the YSZ as top coat in TBCs after oxidation. In the meantime, the capacitance
associated with this relaxation process of the free-standing YSZ is also similar to that of
the intermediate frequency relaxation of TBCs after oxidation. This indicates that the
intermediate frequency relaxation processmay not be associatedwith mixed oxide.
Page126
6.0c1
04
-o- Freestandingtop at
-"- oh
-v- 50h
4.0x104
N
V-V
2.0x104
-v V-v
ýA.,, rº, v,, Vi

0.0
0.0 2.0x104 4.0x104 6.0x104 8.0x104
Figure 6-17 Comparison of the IS of a free-standing top coat with those of oxidised samples.
The capacitance of the intermediate frequency relaxation process (i. e., the mixed
response)does not seem to decreasemonotonically with the oxidation time (Fig. 6-15b),
although our SEM observations clearly revealed that the amount of mixed oxide
increasedwith the oxidation time. This suggeststhe mixed oxides should not be the main
contributor to the intermediate frequency process. In addition, our impedance

measurements on Ni203 and Cr203 revealed that Ni203 and Cr203 exhibit quite high
conductivity and the relaxation for these materials are well above 106 Hz at 350°C.
Assuming the mixed oxide have similar electrical properties to nickel oxide and
chromium oxide, since both of which are the main constituents of the mixed oxide
according to EDS micranalysis, the relaxation frequency of the mixed oxide should be
much higher than the intermediate relaxation frequency. Therefore it seems more
reasonable that the intermediate frequency relaxation process can be ascribed to the
blocking effect of microstructural defects in the top coat. It is well established that
microstructural defects: including grain boundaries, pores, microcracks and impurity

inclusions etc., normally have a combined contribution to the same relaxation process
Page 127
that is called intergrain (or grain boundary) semicircle in the complex impedance plot
(Steil et al. 1997;TiefenbachandHoffmann,2000).
A typical YSZ microstructure is composedof grains a few microns in size and with grain
boundaries a few nanos thick (Badwal, 1995). Thus the capacitive effect of grain
boundaries should be at least 2-3 orders of magnitude higher than that of the bulk. The
capacitive effect of the intermediate frequency process for the L200 sample is about one
order of magnitude higher than that of top coat. This further supports the assumption that
the intermediate frequency process is not only from grain boundaries, but from other
microstructural defects as well.
According to the fitted results in Fig. 6-13, the n value of the mixed response shows a
sharp increase during the initial 200h oxidation and the parameter A shows a sudden drop
for the same oxidation period. After 400h, the n value decreases gradually with
oxidation, while parameter A does not appear to change with oxidation. Meanwhile the
resistancealso exhibits an increase in the first 200h oxidation. It is difficult in this stage
to find out what is exactly responsible for these variations of IS parameters. However,
there are a few speculationsthat are worth mentioning. The first is that high temperature
annealing leads to not only the disappearing of the t' phase, but also grain boundary
degradation (Badwal et al. 1998). For example, segregation of yttrium along grain
boundaries has frequently been observed (Whalen et al. 1987; Hughes and Badwal,
1991). This would increase the resistanceof grain boundaries. The second is that Cr203 is
gaseous at high temperature, due to a reaction of Cr203 with oxygen (Sterns et al. 1974).
GaseousCr203 could reach, through the porous top coat, to the inside surfaces of cracks
and pores, as well as at the grain boundaries. This would aggravate the degradation of
grain boundaries and increase the inhomogeneities, and therefore result in the decreaseof
n value with prolonged periods of time. The third is that the tetragonal to monoclinic
phase transformation is rapid and accompanied by 5-7 % of volume expansion (Lange,
1982a & b), which could result in an increase of microcracks in the top coat becauseof
the volume mismatch between monoclinic zirconia and tetragonal/cubic zirconia. The
increase of microcracks in the top coat is expected to increase the resistance and decrease
the capacitive effect of the intermediate frequency process, which seems to agree with
the variations of IS parameters of the top coat with oxidation time as shown in Figs. 6-13
and Fig. 6-15b. However, the steep increase of the n value and abrupt drop of the
Page 128
parameter A for the initial I00h oxidation still cannot be explained based on current
knowledge. Obviously, further detailed work is neededto elucidate the relation between
the impedanceand the degradation of the top coat.
6.4.5 DeterminingTheThicknessOf TheOxide Layer
1.tä109 7x104
1 Capacitance
at RT ý
9(10d
8.Ox10g -7
5x10
6.Cä108
o 4x104
Q
VAn,,, 3XjCj4
ns
at 35ß°C/'
Capacitance 1
2x104
2.OxlCP
1x1()'4
0.0 1311" 0
TT TýT rTTý1
0123456789 10 11 12
Thickness (µV
Figure 6-18 Linear relationships of the reciprocal capacitance vs thickness and diameter in
modulus spectra (D. ) vs thickness, where Dm C0/Cox,a dimensionless parameter.
One of the main objectives of this work was to detect nondestructively the thickness of the
oxide layer. As discussedin section 6.4.3, the thickness of the oxide layer is quantitatively
related to the capacitance by Eqn.6-3. To examine the linear relationship between the
thickness and the reciprocal capacitanceas indicated by Eqn.6-3, the capacitancederived
from the IS parameters is plotted in Fig. 6-18 as a function of the thickness that was
determined based on the SEM observations. A linear relationship is well established for
both casesat R.T and 350°C, indicating that it is feasible to determine the thickness using
capacitance.However, to work out the equivalent capacitanceof the oxide layer is not that
straightforward. To obtain the IS parameters, tedious fitting procedures are needed. By
contrast, the method using the complex modulus plot to determine the thickness of the
Page129
oxide layer is simple and effective, and has been successfully applied in the detection of
the oxide layer in A1203/SiC nanocompositesas describedin chapter5.
Figs. 6-6e,f give the complex modulus plane plots at 350°C of the as-receivedand oxidised
samples, where the low frequency process is at the left hand and the high frequency
process is at the right hand; correspondingto the oxide layer and the top coat respectively.
The diameter of the right semicircle is similar for all the samples,while the diameter of the
left semicircle increaseswith the oxidation time, indicating the thickness of the oxide layer
increasedwith the oxidation time.
By referring to the discussion in section 4.3, the diameter (Dm) for the left semicircle is
equal to C0/C0,, where Co is the capacitanceof the empty cell, Cox is the capacitanceof
oxide layer. In the light of Eqn. 6-3, the diameter Dm is related to the thickness of the oxide
layer by the following equation:
C', "8
where the electrode area (Ae) was the same for all the impedance measurements,Co was
assumed to be constant when the electrical modulus data were calculated by using the
equation: M*=-2TtfCoZ*. Therefore, the diameter in the modulus spectrum (D,, ) should be
linearly related to the thickness (S) of the oxide layer in TBCs:
Dm= Cr8
with the constant:
C"
C, _
E.-, Ae
As shown in Fig. 6-18, Dm is in a good linear relationship with the thickness; by linear
r
regression fitting, the slope, i. e., C,=5.5x10-5 µm
Page 130
CAPTER VI
The bond coat in TBCs exhibited a fast oxidation in the initial stageswhen heat-treated at
1100°C in air. The oxide grew steadily after 20h oxidation. The kinetics of oxidation
followed a parabolic law. SEM observations and EDS analyses revealed the oxide layer
formed in the initial stages was composed of both mixed oxides and alumina and a
continuous alumina layer was not formed until after 400h oxidation.
Impedance measurementsat R.T demonstrated that there were two relaxation processes
for oxidised samples, one corresponding to the top coat and the other to the oxide layer.
The impedance measurements at 350°C showed that there were three relaxation
processes in the samples oxidised for more than 400h, while only two relaxation
processes for the samples oxidised for less than 200h. The high frequency process
corresponds to the top coat and the low frequency process to the oxide layer. It is
postulated that the third relaxation process at about 400Hz was a response mainly from
grain boundaries, pores and cracks in the top coat, rather than from the mixed oxide.
Comprehensive equivalent circuits were proposed and these can be significantly
simplified basedon the analysis of the microstructure of TBCs. The simplified equivalent
circuits fit experimental data very well, indicating the IS parameters obtained by fitting
are reliable. Based on the analysis of impedance spectra and IS parameters, it has been
found that not only the characteristics of the oxide layer, but also the degradation of the
top coat, can be monitored non-destructively by impedance measurements. However,

further work is neededto clarify the relation between the IS parameters and the top coat
degradation.
The capacitance derived from impedance measurementsat both RT and 350°C shows a
monotonic dependence on the oxidation time. The reciprocal capacitance and the
semicircle diameter in the complex modulus plane have a linear relation with the
thickness of the oxide layer. Modulus spectra have been proven to be a very useful non-
destructive approach to examine the growth of the oxide layer in TBCs. Therefore it is
highly feasible to determine the degradation of TBCs using IS.
Page 131
7 Characterisation Of Clay Sintering Process Using IS
7.1 Introduction
In-situ monitoring of sintering processesis extremely important for understanding the
sintering mechanismsand improving the efficiency of ceramic manufacturing processes.

Clay is the principal raw material for producing conventional ceramics, such as
whiteware and porcelain, so the majority of conventional ceramics are clay-based

ceramics. Clay-based ceramics have a wide range of applications and they are the most
complex ceramic materials. Whiteware, i. e., floor and wall tile, artware and pottery,
dinnerware and fine china, sanitaryware and food-serviceware etc., embraces an
important industry, which accounts for about 7% of the entire ceramic market worldwide
(Carty and Senapati, 1998). Because the complicated reactions occurring in clay based
ceramics are kinetically governed, rather than thermodynamically equilibrated processes,

the measurement of temperature alone often does not provide sufficient information for
monitoring and thereby controlling the sintering processes (Carty and Senapati, 1998).
Traditionally pyrometric cones were developed to follow the firing progress. Pyrometric
cones are three-sided pyramids made from mixtures of kaolin, feldspar, quartz and other
raw materials resembling the body composition that function as true indicators of
reaction kinetics in porcelain bodies (Fairchild and Peters, 1926; McCaughey and Neff,
1944). Modern commercial firing processesare still often checked with pyrometric cones
for verification of the heat work accomplished. For this reason, it is desirable to develop
an effective approach for the in-situ characterisation of the sintering of clay-based

ceramics.
The purpose of this work is to use IS as a non-destructive approach for characterising the
real-time sintering process of clay-based ceramics. Liquid phase sintering is the dominant
mechanism in the sintering of clay-based ceramics (Kingery et al. 1976b). The formation
of the liquid phaseis a crucial step in the process in which the fluidity of liquid initiates
compact shrinkage. The quantity and the viscosity of the liquid phase are the two
important factors that determine the kinetics of sintering (Anseau et al. 1981). The liquid
phase formed in the ceramic compact at high temperaturesis an ionic conductor and thus
Page132
CHAPTERVII CHARATERISATION
OF CLAYSINTERINGPRO FSS
acts as an electrolyte (Morgan and Koutsoutis, 1986). Microstructural parameters, such

as the quantity of liquid phase, the porosity content and the distribution of different
phases as well as the chemical composition of the liquid phase would affect sintering
process and these parameters would also keep changing as sintering proceeds.
Monitoring the sintering of clay-based ceramics using IS will provide important
information on sintering phenomenain clay-basedceramics.
In this study, impedance measurementswere made in-situ while a clay compact was
being fired at different temperatures. The impedance spectra of clay compacts under
firing consist of a high frequency semicircle and a low frequency tail, which have been
proven to correspond to bulk specimen and electrode effect respectively. The high
frequency semicircle is informative regarding the electrical properties of the bulk
specimen. The variation of bulk conductivity with sintering time, which can be deduced
from isothermal impedance spectra, was found to be in agreement with the shrinkage
curve. An equation relating the electrical conductivity to the density of the specimen has
been established.The electrode effect curve has a strong temperature dependence, which
is directly related to the capacitive effect of the electrode /specimen interface. This
capacitive effect has been proven to correspond to the blocking anode where a double
layer charging is built up by the accumulation of the non-bridging oxygen ions nearby
anode. The low frequency tail can be an indicator of local temperature. In addition,
variations of the IS parameters with sintering temperature were found to be sensitive to
the formation of liquid phase.
7.2 Experimental Procedure
7.2.1 Sample Preparation
Ball clay (Dorset, HYMOD PRE) was ball-milled in the dry mode for 5h to produce a
powder of uniform particle size. The chemical composition and particle size distribution
specified by the powder supplier are given in Table 7-1 & 7-2 respectively. Pellets 13
mm in diameter and 3-4 mm in thickness were prepared by uni-axial pressing using a
pressure of 120 MPa. Before sintering the green clay compacts were carefully polished
with SiC paper to ensure every sample was in a regular geometric shape.
Page133
Table 7-1 Chemical composition of ball clay (wt %)
Si02 A1203 Fe203 Ti02 CaO MgO K20 Na20 L.O.I* C
54 301.4 1.3 0.3 0.4 3.1 0.5 8.8 0.3
" Loss of ignition
Table 7-2 Particle size distribution of ball clay (wt %)
< 5µm < 24m < lµm < 0.5µm
96 88 79 67
7.2.2 ImpedanceMeasurements
The rig shown in Fig. 3-6 was used for impedance measurementsat high temperature,
Since the resistanceof the leads connecting the electrodesto the impedance analyser was
much smaller than that of the tested samples,the influence of the leads on the impedance
measurementswas negligible. Data acquisition was undertaken using a PC computer in
real time when the clay compact was being fired at different temperatures. Measurements
were taken over a frequency range of 106 to 0.1 Hz using aI Vrms applied voltage with
3-6 readings per decade of frequency. Our preliminary experiments revealed that
measurementswith voltages lower than 0.2 Vrms resulted in an unstable signal, while
measurements with voltages 0.2-3 Vrms gave stable signals and reproducible results.
Therefore, the voltage of I Vrms was chosen for all the impedance measurements.
7.3 Results
7.3.1 Impedance SpectraDuring Heating
Impedancemeasurements needto be madeusing a set of frequencies,usually scanning

from high frequencydown to low frequencywith a few data readingsfor eachdecade.
For this reason, a single cycle of measurementwill take a few minutes, sometimes even a
few days,dependingon the chosenfrequencyrangeand the numberof datareadingsfor
Page134
CHAPTER VII CHARATERISATION OF CLAY SINTERING PROCESS
each decade. Therefore the sample under investigation needs to be kept in a steady state
during the impedance measurementsso that it does not change with time and many
measurements can be averaged (Macdonald and Johnson, 1987). For this reason, the
measurement needsto be made isothermally, and the measuring temperature needs to be
controlled to be as steady as possible.
Fa-]
6.0x105
----c - 700°C
A 800°C
4. Ox105 -0- 900°C
N
2. Ox105 4.64Hz
46.4Hz
464 oo 0
§Mie
0.0
0.0 2.0x105 4.0x1056.0x1058.0x10' 1.0x106 1.2x1061.4i10ß
T
rbl
5.0x104-
4.0x 104
--o--l OOOOC
3.0x 104 -¢--900°C
N 2.0x104
464Hz
1. Ox104 Hz
O_ 1
Pf, _00
0.0
0.0 2.0x104 4.0x104 6.0x104 8.0x10" 1.Ox10'
Z'
Figure 7-1 Complex impedance plots of the clay compact at (a) 700°C, 800°C, 900°C; (b)
900°C11000°C.
In this work, isothermal impedance measurementswere made at various temperatures in
the range of R.T-1200°C during heating. It was found that the impedance spectra
Page 135
measured at temperatures above 500°C are reproducible, but at temperatures below
500°C, the reproducibility was poor. Several factors could affect the reproducibility of
impedance measurements,e.g. incomplete contact between sample and electrode, short
circuit through a less resistive path in the clay (for example, because of the presence of
moisture in clay), as well as possible physicochemical reactions occurring in the
compact, which will be dealt with later on.
Typical complex impedance plots are shown in Fig. 7-1 where the spectra at 700°C,
800°C, 900°C and 1000° are given. There is a high frequency semicircular arc and a low
frequency tail in each of the complex impedance plots. Both the high frequency arc and
the low frequency tail become smaller with increasing temperature and the cut-off
frequency between the high frequency arc and low frequency tail increases from 4.6 Hz
at 700°C to 4600 Hz at 1000°C. The cut-off frequency represents the transition from one
relaxation process to another. Measurements made by using samples with different

thicknesses showed that the thickness of the sample only affected the high frequency arc
radii, but had little effect on the low frequency tail. Therefore the high frequency arc may
be attributed to bulk effect and the low frequency tail to the effect of electrode.
0.12:i
-a- 5000:
0.10C -o- 7OO
-A- 800°c
-v- 900ct
0.07;i
1000`
0.05(
0.02°
0.00(
0.00 0.05 0.10 0.15 0.20
tvT
Figure 7-2 Complex modulus plots at different temperatures of green clay form during
heating.
Fig.7-2 shows the complex modulus plots at different temperatures.The modulus
spectrumat 500°C seemsto consistof two overlappedarcs,and the high frequencyare
Page136
CHAPTER VII CHARATERISATION OF CLAY JNTERING PROMS
(right hand) gradually disappeared when the temperature increased. When the
temperature reached 1000°C, even the low frequency arc became smaller. The
disappearanceof the bulk arc in the modulus spectra indicates that the clay compact
becomes more conductive with increasing temperatures.
In order to examine the time dependence of the impedance spectra, the impedance
measurements were made repeatedly every 5 minutes at a given temperature. As

mentioned earlier on, a single cycle of impedance measurements needs at least a few
minutes to finish. Here the frequency range of 106 -103 Hz was selected with 6 data
acquisitions per decade, which would take 2.62 minutes for a single cycle to finish.
Therefore the total time elapsed between two cycles was 7.62 minutes.
Fig. 7-3a shows the multi-cycle impedancemeasurementsat 800°C, where the impedance
spectra for different cycles entirely overlap. This suggests that there could not be
significant physicochemical change occurring in the clay compact at this temperature.
On the other hand, the impedance spectra measuredat temperatures above 950°C (Figs.
7-3b-c) show a gradual shift towards the left. As shown in Figs.7-3b-c, the spectra shift is
faster at the initial stagesof sintering than at the later stages of sintering. In addition, the
spectra shift is faster at the higher sintering temperature. The trend of spectrum shifting is
consistent with the sintering behaviour of ceramics where the densification rate is higher
at higher temperature and it slows down when the shrinkage curve approaches an
asymptote.
As will be discussed in detail later in this chapter, the shrinkage curve measured by
dilatometry shows the clay compact starts to shrink at a temperature of about 950°C, the
shifting of the spectrum can be a consequenceof the densification of the compact under
firing. During sintering, the density increases with the sintering time. The increase in
density means the effective conduction area in the bulk is increased,which gives rise to a
decreasein resistance.
Page137
2.0x105
16x105 -0-1st cycle

-o- 2nd cycle
1.2x105 --- 3rd cycle
-v- 4th cycle
N 8.0x10" 5th cycle
-x-
4.0x10°
0.00.0
1.0x105 2.0x105 3.0x105 4.0x105 5.0x105
z
b]
1.(
8.(
6.(
N
4.(
2.(
[C] 8.0x102
6.0x102
ýýmg 1MHz
4.0x102 ý; ýý
2.0x102 00 000OO,
ýýý '000
ý_ 1
0.0
1.500' 1.800' 2.1x10' 2.4x10' 2.7x10"
z
Figure 7-3 Multiple impedance measurements at time intervals of 5 min, each measurement
taking 2.62 minutes at (a) 800°C; (b) 1050°C; (c) 1100°C.
Page 138
It is worthwhile noting that during isothermal sintering, some spectrum features, e.g., the
dielectric loss, the cut-off frequency and the height of the low frequency tail, etc., hardly
changed as the sintering time was increased.This might indicate that there was not much
change in the physical and chemical properties during isothermal sintering, except the
change in the size of the specimen.
It is also important to point out that the height of the low frequency tail appeared to be an
indicator of temperature. As in Fig. 7-4, the height of the low frequency tail is referred to
as the imaginary component of the impedance at 0.1Hz, the lowest frequency used in
these measurements.For all the cases, the tail height changed little during isothermal
sintering, but changed significantly with the temperature. Table 7-3 gives the tail height
at different temperatures.
4.0x103
- -10min
-o- 40min
3.0x103- -A- 70min
-v-100min
0.1
2.0X103 / ý.
1. Ox103 1(ýI
ýý-
ýo/o Imaginary component
ýý
ý of LFtail
oo
0.0-,
1.0x103 2.0x103 3.0x103 4.0x103 5.0x103 6.0x103
Figure 7-4 Complex impedance plots for clay compact under firing at 1100°C, at different
sintering times, frequency range used being 106_10"1Hz.
Strictly speaking,it is impossibleto obtain a set of impedancedata correspondingto a

steady state of the material under sintering, becausethe material is undergoing a
Page139
continuing change at a given temperature, so the electrical properties under this
circumstance should change continuously. However we found that the main features of
the impedance spectrum obtained during sintering remained more or less the same. In
impedance measurements,the high frequency data takes less time than low frequency
data to acquire. For example in this study, data acquiring for the bulk semicircle (in the
high frequency range) took less than 0.5 minute. The measurements for the bulk effect
were very quick, so the impedance spectra thus obtained can be regarded as representing
an average state of the specimen over a single cycle of measurement, which was 2.62
minutes in the caseshown in Fig. 7-3.
Table 7-3 Tail height at different temperatures
Temperature ( °C) 700 800 900 1000 1100
Tail height(Q)* 1x105 4x104 1.5x104 4.2x103 1.8x103
* The electrode area was the same for all the measurements: equal to 0.64cm2.
7.3.2 Impedance SpectraDuring Cooling
After having been sintered at 1200°C for 4h, the sintered body was cooled to R. T. To
observe the variation of the impedance spectra with the change in temperature, similarly
to the case of the green compact during heating, isothermal impedance measurements
were made at every interval of 100°C from 1200 °C to R.T, during the cooling of the
sintered body.
Figs.7-5a-c show the impedance spectra at different temperatures. The spectra of the
sintered specimen consisted of a high frequency semicircle and a low frequency tail, as in
the spectra of the green compact. The semicircle diameter increased with the decreasein
temperature. As expected, there was no change in the spectra during the isothermal
measurements.
As a whole, the featuresof the impedancespectraare very similar to thoseof the green
compact,except that the bulk semicircleof the sinteredbody is much smaller, and the
bulk andelectrodeeffectsare betterresolved. The complexmodulusplots of the sintered
Page 140
OF CLAYSINT IN PRO ESS
body shown in Fig. 7-6 are also similar to those of the green compact: as measurement
temperature decreased,the bulk arc becamemore complete.
Fa-]
-ý- Wt
5.0x105
4.0x105
3.0x105
2.0x105
1.0x10 .
0.0 2.0x105 4.0x105 6.0x105 8.0x105
a 1.0x105
8.0x104 -0--
700`C
8000
900°C "
"
--o--
6.0x104
IV
4.0x104 O "
O
O
0,0 ! '; M=3W r-r- i

0.0-" 4,0x104 8.0X10 1.DA05
fc]
4.Dxl
N
0.0 iY-'T'' 7
0.0 2.0x103 4.0x103 6.0x1U3 8.0x103 1.0x1U°
Z
Figure 7-5 Complex impedance plots for sintered bodies measured isothermally at different
temperatures.
Page 141
0.20
- -500°C
-o-700°C
0.15 -0-800°C
900°C
ý_-_1000°C
0.10
0.05
0.00
0.00 0.05 0.10 0.15 0.20 0.25
M
Figure 7-6 Complex modulus plots for sintered bodies at various temperatures during
cooling.
7.3.3 Characterisation Of Clay Compacts Using Conventional Characterisation

Techniques
In order to examine what physicochemical processes occurred in clay compacts during

heating and sintering, and relate these to the impedance spectra, some conventional
characterisation techniques, i. e., TGA, XRD, dilatometer measurements and SEM were
usedto examinethe clay compact.
Fig. 7-7 gives the TGA result from the clay powder, where we can see two steep drops,
one in the temperature range of 50-150°C, the other one between 420 and 550°C.
Kaolinite is the main constituent of clay mineral. The crystal structure of kaolinite
contains hydroxyl groups and the dehydroxylation of these groups to form metakaolin
occurs at about 550°C (Brindley and Nakahira, 1959a&b):
A1203"2S102"2HZO -* A1203-2S102+ 2H201' (7- 1)
Thereforethe weight loss occurring in the range 420-550°C can be attributed to the
dehydroxylationof kaolinite. The weight loss in the range50-150°Cis probably due to
Page14
the loss of moisture that is physisorbed at the particle surfaces. Since both physisorbed
and chemisorbed moisture can significantly affect the electrical properties of the powder
compact (Traversa, 1995), the poor reproducibility of impedance data we observed at
temperatures below 500°C was largely due to the presence of moisture in the clay
compact.
10(
9E
0 100 200 300 400 500 600 700 800 900 1000
Temperature (°C)
Figure 7-7 Thermogravimetric analysis of clay powder from R. T to 950°C
0
o.
c
-0.
o. a
9
(3 -0
-o 500 600 700 800 900 1000 1100 1200

Temperature (°C)
Figure 7-8 Shrinkage curve of clay compact obtained from dilatometer measurements.
The shrinkage curve of the clay compact measured by dilatometer is shown in Fig. 7-8,
where we can see noticeableshrinkagestarts at about 950°C and the shrinkage rate
increasesas the temperature is then raised.
Page143
Q Mca
Kadire
" Q
D
4
"
PC
1C2 Q Q
" p
N
ý3 9ecPC
. "
Q A
" o
16 18 23 22
20
Figure 7-9 X-ray diffraction patterns of the clay compacts after 15minutes sintering at
different temperatures.
The XRD analyses (Fig. 7-9) after heat treatment at 850°C and 900°C shows that no
apparent phase transformationsoccurred in the clay compacts, whereas after heat

treatment at 950°C the amount of kaolinite and mica appeared to decrease. The heat
treatment at 1020°C led to almost complete disappearanceof kaolinite and mica.

Meanwhile the phasetransformationin the clay also involved the disappearanceand
formation of glassyphases.It is believedthat the diminishingmica and kaolinite contents
in the clay compactsare related to the formation of a liquid phase (Shuller, 1964;
Sonuparlaket al. 1987;Okada et al. 1986). Therefore the XRD analysessuggestthe
presenceof the liquid phasein clay at temperaturesbetween900 and950°C.
The dominant mechanism in the sintering of clay-based ceramics is vitrification where
significant shrinkage cannot occur unless a certain amount of liquid phase is present
(Anseau et al. 1981). In this sense,the XRD analyses in Fig. 7-9 are in agreement with
the shrinkage measurement shown in Fig-7-8, both suggesting that the liquid phase
occurs at about 950°C. This is further confirmed by the SEM observations (Fig. 7-10).
Page 144
CHAPTER VII CHARATERISATION OF CLAY SINTERIN(; PRO('ESS
phase region
Figure 7-10 SEM micrographs of polished surfaces of the sintered or semi-sintered clay
compacts after being fired for 15minutes at (a) 850°C; (b) 900°C; (c) 960°C; (d) 1020°C.
Clay mineral has a platy habit, because the main constituent of clay-kaolinite is
composed of [Si205]2- layers and A12[(OH)4]2 layers (Kingery et al. 1976a). At

temperatures below 950°C (Fig. 7-10a, b) when there is no liquid phase in the clay
compact, the polished surface shows a rough morphology and the platy nature of the clay
mineral can be identified. At 960°C (Fig. 7-10c), most of the surface area is covered by
smooth morphology, which is a consequenceof liquid drenching, but some regions still
have not been touched by the liquid phase. When the temperature goes up to 1020°C
(Fig. 7-10d), liquid phase seems to have reached everywhere except pores, and therefore
the clay compact is composed of a continuous network of glass drenched-regions and
many small pores.
Page 145
CHAPTER VI! CHARATERISATION OF CLAY SINTF,RING PROCESS
_______. ___
7.4 Discussion
7.4.1 Equivalent Circuit Models
According to a brick layer model as discussedin Chapter 2, the complex impedance plot
of a polycrystalline ceramic material is expected to have two semicircular arcs, one

corresponding to grains and the other to grain boundaries. However, there was only one
semicircle corresponding to the bulk effect of the clay, according to the impedance
measurement results shown in Figs.7-1 a-b. This might be due to the continuous
conduction path in the clay compact.
At relatively low temperature before the liquid phase was formed, the conduction is
through the surfaces of the clay particles. Clay minerals consist of hydrated aluminum
silicates that are fine-grained and usually have a platy habit where positive ions, such as
alkali ions and alkali earth ions, can fit on the surface of the particles or between
different layers (Kingery et al. 1976a). The fine microstructure leads to a large free
surface area while the mobile ions at the surface may make a major contribution to the
electrical conduction in the clay. Materials with high porosity, if exhibiting considerable
particle surface conduction, normally do not show resolved grain and grain boundary
effects in their impedance spectra, becausethe conduction path along particle surfaces is
an èasypath'. This is similar to the case of ionic-type humidity sensors. In these sensors,
the proton hopping between neighbouring water molecules on particle surfaces is the
dominant conduction mechanism (Traversa, 1995). Because of the èasy path' at the
surface, there is usually only one semicircle corresponding to these sensors, which have
been witnessed in the works of Hwang et al. (1995) and Gusmano et al. (1996).
At temperatures above 950°C when a continuous network of glass phase is present in the
clay, the charge carriers, e.g., the alkali ions, can transport along the network without
experiencing a significant blocking effect. Therefore there is usually only one semicircle
in the complex impedance plot for glass samples. Such behaviour has already been
observed by other researcherswith the glass containing crystallised particles and pores
(Murugan and Varma, 2001; Abrahams and Hadzifejzovic, 2000). Therefore the
continuous conductive path model as mentioned in Section 2.5 is applicable to the clay
compact during sintering.
Page 146
A tail (sometimes called a `spike') is always present in the complex impedance plot at
low frequencies, which is an electrode effect. The electrode effect is frequently seen in
the impedance measurementsof cement paste (McCarter et al. 1988; Christensen et al.
1992), humidity sensors(Hwang et al. 1995; Gusmano et al. 1996), and other ionically
conducting materials (Steil et al. 1997). According to the equivalent circuit fitting, a high
capacitive effect is associated with the low frequency tail. The electrode effect is a
response from the electrode/material interface which is very thin in thickness and the
electrode effect normally has a very high capacitive effect.
Rb
CPE.
CPEb
Figure 7-11 Equivalent circuit model for clay compact.
The equivalent circuit for the clay compact can be given as in Fig. 7-11, where a parallel
R-CPE element represents the bulk clay compact, and a CPE represents the electrode
effect. Here a single CPE, rather than a parallel R-CPE, is used to represent the electrode
effect. This is because the platinum/clay interface is believed to be a blocking interface,
i. e., the resistance can be regarded as infinitely large. The electrode effect should have
been purely capacitive and representedby a straight line vertical to the x-axis. Here a
tail-like line, rather than a straight line normal to the x-axis, is obtained in the complex
impedance plane. This is because the electrode/material interface is very rough and the
electrode/material contact is far from ideal.
Page 147
OF CLAY SINTER NG PROCES
7.4.2 Fitted Results
Table 7-4 Fitted results for clay compact during heating
Temperature Rb Ab nb Ae n,
(0C) (ohm) (ohm'"""F") (ohm '"""F'")
700 1.0x106 2.0x10'1° 0.73 2.6x10"6 0.17
800 3.2x105 2.7x10'1° 0.73 7.0xl0"6 0.16
900 7.6x104 3.6x101° 0.74 2.6x10"5 0.16
1000 7.8x103 4.5x10'10 0.75 7.8x10"5 0.25
1100 2.2x103 -------- ------- 1.Ox10-4 0.27
The fitted results for the green clay compact during heating are given in table 7-4. The
resistance of the specimens decreaseswith increasing temperature while the parameters

for CPEs increase with increasing temperature. The parameter Ae (for the electrode
effect) is about five orders of magnitude larger than Ab (for the specimen). The
exponential index for the bulk, nb, is much closer to 1 than that for the electrode effect,
ne. This indicates that the CPEe in Fig. 7-11 assignedfor the electrode effect is far from
being an ideal capacitor. For a perfect flat blocking electrode, the electrode effect is
expected to be purely capacitive (Armstrong and Todd, 1995; Raistrick, 1987). But under
the experimental conditions, the contact between the clay compact and platinum
electrode was poor and the electrode/clay interface was rough. It is well established that
the porosity of specimen or roughness of the electrode surface leads to a frequency
dispersion of the interfacial impedance and produces a constant phase angle (Armstrong
and Todd, 1995; Raistrick, 1987). The ne obtained here is only 0.157-0.274, which
implies that the electrode/clay interface is very rough.
Page 148
Table 7-5 IS parammeters of the sintered specimen
Temperature Rb Ab nb A. n,
(0C) (ohm) (ohm'-""F') (ohm'"W")
500 3.6x105 1.9x10"° 0.76 2.2x10"6 0.62
700 4.5x104 2.OxlO"° 0.80 1.1x10'5 0.52
800 1.9x104 8.5x 10" 0.86 2.2x 10"5 0.47
900 5.1x103 1.9x10'10 0.80 6.3x10"5 0.38
1000 3.5x103 1.2x10"1° 0.85 1.5x10"5 0.36
1100 2.1x103 ------- --------- 1.0x104 0.27
Table 7-5 shows the IS parameters obtained from fitting the impedance spectra of
sintered specimens at different temperatures. The bulk resistance increased with a

decrease in temperature. Temperature had little effect on Ab, however Ae decreased
considerably with a decrease in temperature. It is interesting to note that ne shows an

increasing trend as the temperature is decreased, which cannot be interpreted at the
present stage. However, by comparing Table 7-4 and Table 7-5, it is obvious that the ne
value for the sintered body is higher than that for the green clay compact. This is
reasonable, because the sintered specimen has a smoother contact with the electrodes,
while the green clay compact had a poor contact with electrode due to the rough compact
surface. Therefore the steepnessof the low frequency tail seems to be able to represent
the condition of the electrode/compactinterface.
The dielectric constant of the bulk specimen can be obtained by employing an equation
developed by Christensen et al. (1994):
c=(27tfcop)(zein->)OIýSEo (7- 2)
where ft,p is the frequency at the top of the impedancearc corresponding to the specimen,
8 is the depressionangle, R is the resistance, 1-28/it is equal to the n value in Eqn. 7-2
Page149
yý CHARATERISATION
OF CLAYSINTERrN GPRO c
____ ____
__
and 1/ASrepresentsthe geometrical factor of sample. According to Eqn. 7-2, the dielectric
constants of the clay compact at 700-900°C were calculated as 400-500. At
temperatures above 900°C, the frequency at the top point of the bulk arc is out of the
measurementrange. Therefore the dielectric constant cannot be calculated. The dielectric
constant of 400-500 is unusually high, which can only be explained by an underlying
"dielectric amplification factor" (DAF) mechanism (Christensen et al. 1994). DAF
usually results from a microstructural arrangement where isolated conductive domains

are imbedded in a continuous insulating matrix. This factor is responsible for the large
effective dielectric constants of a barrier layer capacitor, where conductive grains are
surrounded by the insulating grain boundary. DAF was also found to exist in cements
where a conductive pore solution is separated by an insulating product C-S-H layer
(McCarter et al. 1988; Christensen et al. 1992) and as mentioned in chapter 6, in
A1203/SiC nano-composite where conductive SiC particles imbedded in A1203 matrix
result in larger capacitance. However, further work is needed to clarify what is

responsible for DAF in clay compact at high temperatures.
7.4.3 Sensitivity Of IS To The Formation Of Liquid Phases
Fig. 7-lb shows an incomplete bulk semicircle where the high frequency part is
incomplete when the measurementswere made at 900°C and 1000°C. The relaxation
frequency of the clay at 1000°C was well above IMHz. This suggests that the clay is
very conductive at 1000°C, which is due to the formation of a continuous conductive

liquid phasein the clay compact.
Based on the data in Table 7-4, the bulk conductivity can be calculated by: Go=1/Rb. The
logarithmic of the conductivity against the reciprocal temperature is plotted in Fig. 7-12.
The data in the range of 600-900°C shows a linear relationship, indicating a consistent
conductivity mechanism in this temperature range. The data at temperatures above 900°C
apparently do not fit the same linear relation, which implies a new physicochemical
process may occur between 900-1000°C. Meanwhile, based on impedance spectra, the
dielectric loss factor (at 1MHz) in the clay compact (Fig. 7-13) also shows a steep
increase at temperatures above 900°C. All of this suggests that IS is sensitive to the
formation of liquid phase in clay compacts.
Page 150
-6
1100°C
N -8
c 1000°C
-10
900°C
-12
CD
? 700C
-14
600°C
-1647 89 10 11 12
1/TX104 (K"1)
Figure 7-12 Plot of logarithmic conductivity vs. reciprocal temperature.
CIO
c
sU_
U
N
'ra
60 700 800 900 1000 1100 121

00
Temperature()C)
Figure 7-13 Loss tangent (at1MHz) obtained from impedance measurements as a function
of temperature.
havealreadybeenusedto detectthe presence

D.C. electricalconductivitymeasurements
of liquid phase in the A1203-1mol% Ti02-0.5 mol% NaOin system (Morgan and
Koutsoutis, 1986). D.C. electrical measurementswere found to be sensitive to the
presenceof a small amountof liquid phasethat may influenceor control the sinteringof
Page15!
CHAPTERVIL__ CHARATERISATION
OF CLAY SINTFRI G ROC S
=,._:, ___
many ceramic compositions. IS is a much more sophisticated technique than the D. C.

conductivity measurementtechnique. More information, e.g. capacitive effects, dielectric
loss and electrode polarisation, etc., can be obtained from impedance spectra. Also the
electrode effect can be excluded to examine the electrical properties of the specimen
only. In this work, the formation of liquid phase during firing was not only detected by
the change in the activation energy for electrical conduction, but also determined by a
steep increase in the dielectric loss (Fig. 7-13) and spectrum shifting (Fig. 7-3).
7.4.4 Electrode Effects
The electrode effect corresponds to a low frequency tail in the complex impedance plane
plot. It has been observed that the height of the low frequency tail remains unchanged
during sintering at a constant temperature as shown in Fig. 7-4. The height appeared to
have a strong temperature-dependence.This is of interest from the viewpoint of in-situ
monitoring of sintering, because this seems to provide the extra information regarding
local temperature, i. e., the temperature at the contact between electrode and specimen.
This means the low frequency tail could become an effective indicator of temperature,
which used to be checked by pyrometric cones. For this reason, it is necessary to

elucidate what is responsible for the presenceof the low frequency tail and what kind of
information can be sought basedon the featuresand parametersof the low frequency tail.
The so-called electrode effect here is a response from the platinum electrode/clay
compact interface. At high temperatures when clay is vitrified, the glass phase is the
predominant phase in the clay compact. In alkali oxides-silica glasses, some Si-O-Si
bonds are broken and non-bridging oxygen ions are generated.The ionic species, such as
alkali ions and non-bridging oxygen are distributed among the strongly covalent network
or chain structure of Si-O-Si. Those ionic species are the charge carriers for conduction
(Doi, 1987). Since alkali ions have much greater mobility than the non-bridging oxygen
ions, alkali ions will first conduct to the cathode and generate the alkali-depleted region
(ADR) near the anode as shown in Fig. 7-14 (Carlson et al. 1974; Doi, 1987). In the
ADR, the non-bridging oxygen ions are not charge-compensated,and tend to move to the
anode due to electrostatic attraction. Thus oxygen ions accumulate underneath the
blocking anode, and a double layer is formed at the anode. In the meantime, alkali ions
gather near the cathode, so a double layer can be built up at the cathode. The double layer
Page 152
OF CLAY SINTERINGPROCESS
at either anode or cathode could be responsible for the low frequency tail in impedance
plot. The platinum electrode is supposedto be blocking to both oxygen and alkali ions at
a small voltage signal and a pure capacitive effect from electrode is expected. However,
as mentioned previously, the non-ideal situations arising from effects, such as roughness
and porosity of the electrode, the poor contact between electrode and the material etc.,
could lead to a constant phase angle (CPA) behaviour and produce a `tail' at low
ADR
R+
Anode Cathode
O2' R
,R'
02- Electriccurrent
R+
02-
R' R'
02-
R+
R, R'
02.
frequencies.
Figure 7-14 An alkaline depleted region (ADR) underneath anode, R+ representing alkali
ions.
The constant angle behaviour means the capacitance is dependent on frequency. The
capacitanceof a blocking electrode can be calculated by employing Egn. 2-31:
C=Aw"-'sin(nn/2).
Based on the fitted results as shown in Table 7-5, the capacitance at a given frequency
can be obtained. In order to investigate the temperature-dependenceof capacitance of the

electrode effect in detail, we take the capacitance at 0.1Hz as representative of the
behaviour throughout the frequency range. This is reasonable because: (i) the
capacitance at 0.1E1z is reliable in representing the electrode effect, as at such a low

frequency the influence of the bulk effect on the low frequency tail can be ruled out; (ii)
the capacitances of' electrode effect at different temperatures thus obtained are
Page 153
C j4p_TER vi! CHARATERISATION OF CLAYSINTFRING PROCESS
comparable, since the influence of frequency on capacitanceis excluded. The logarithmic

capacitanceof the electrode effect is plotted against the reciprocal temperature in Fig. 7-
15. Obviously, the capacitanceis thermally activated, i. e., the capacitance increases with
-7
-8 M
LL -9
4'
-10 ",
J
-11
-12
-13
7 a 0 1n I1 17 1I' ._
1/Tx104(K )
increasing temperature.
Figure 7-15 Arrhenius plot of capacitive effect of electrode / sintered body interface.
A thermally activated capacitance is most likely due to some transport processes.

Jonscher and Reau (1978) proposed an interpretation for this phenomenon: capacitance
arises from a near-equilibrium situation near an incompletely replenishing electrode,

where the interfacial field is screened within a Debye length X* and the resulting
capacitanceper unit areais givenby theexpression(Jonscherand Reau,1978):
Ch = E,./A= [e2E,,N/ kT ]"' (7-3)
0 There is balancebetween two factors that affect the distribution of the mobile ions near
electrode:the electrostaticattractionthat causesthe mobile ions to be as nearto the interfaceas

possible,and thermalmotion that causesa spreadinto the bulk. The Debyelength is an important
parameterdescribingthe spreadingof the mobile ions into the bulk: X=(ckT/e2N)'/2(Armstrong
and Todd, 1995).The largerthe Debyelength,the more spreadarethe mobile ions.
Page154
CUfZ & JL CKARATERISATION OF CLAY SINTERING PROCESS
where e is the magnitude of the electronic charge and N is the density of the mobile ions,
c,, is dielectric constant and other symbols have normal meanings. The density N is
clearly thermally activated and the activation energy for density N should be similar to
the conductivity activation energy (Jonscher and Reau, 1978). Therefore, according to
Eqn. 7-3, the capacitance is expected to have an activation energy that is half of the
conductivity activation energyt. Based on Fig. 7-15, the slope of the linear relation: Ln(C)
vs. I/Tx 10° is -0.846. The activation energy of capacitance (E, =R*slope) is calculated as
-8 I_#
-10
v -12
-14
J
-16
-18 68 10 12 14 16 18
1/Tx104
16.8 kcal/mat, from which the activation energyof ion density N is calculatedas 33.6
kcal/mol (1.46eV).
Figure 7-16 Arrhenius plot of bulk conductivity for sintered clay body.
Based on the fitted results as shown in Table 7-5, the conductivity of the sintered
specimen is plotted against the reciprocal temperature in Fig. 7-16. The slope of the
linear relationship: Ln(G) vs. I/Txl04 is 0.89, so the activation energy for the bulk
conductivity is calculated as 17.7kcal/mol (0.77eV). This is in good agreement with the

results reported in the literature. The activation energy of bulk conductivity for
commercial soda lime glass was reported as 18.2kcal/mol (i. e., 0.79eV) (Doi, 1987) and
N is thermally activated:N=Ndexp(E/kT),whereE. is the activationenergy.On combiningthis
equationwith Eqn. 7-3, we haveCb=(e2sNa/kT)'nexp(Ea/2kT).

Thereforethe activation energy
for capacitanceis E./2.
Page 155
lTEBVJL CHARATERISATION
OF CLA Y SINTERING p ROOF
__ --_.
that of a normal glass was reported as about 17.Okcal/mol (i. e., 0.74eV)(Carlson et al.
1974).
The activation energy of the movement of non-bridging oxygen ions in the ADR
was
reported to be 41.4 kcal/mol (i. e., 1.8eV) (Carlson et al. 1974). The activation energy of
the moving ions obtained based on Fig. 7-15 is very close to the reported activation
energy of the movement of non-bridging oxygen ions, but much higher than that of the
bulk conductivity. This indicates that the electrode effect is mainly from the blocking
anode' where non-compensatedoxygen ions in the ADR are mobile, but have a much
higherenergybarrierthan thealkali ions in thebulk.
The assignmentof the low frequencytail to the blocking anode is also reasonableon
physicalgrounds.Thecathodeis expectedto havea much higher capacitancethan anode,
becausealkali ions are much more mobile than the non-compensated oxygen ions in the
ADR (Carlsonet a!. 1974).Sufficient alkali ions can approachthat cathodeas near as
they canoncean electricalfield is establishednearthe electrode.By contrast,the oxygen
ions in the ADR are less mobile and tend to spreadin the bulk near the interface,
thereforethe capacitiveeffect is smaller.The anodeand cathodeare connectedin series,
hence the impedanceresponsefrom the cathodedouble layer that has a much greater
capacitiveeffect is negligible.
If the electrodeareaand the contactbetweenelectrodeandmaterialremain constant,the

capacitive effect of the anode double layer is only dependenton the temperature
accordingto Eqn.7-3. This is the reasonwhy the heightof the tail can be an indicator of
local temperature of the electrode/specimencontact area.
7.4.5 Monitoring DensificationProcess
Our experimentalresultsshowedthat whentemperatureis increasedto above950°C,the
clay compact startedto shrink. Meanwhile the impedancespectrumbegan to show a

gradual shift towards the left, as shown in Figs.7-3b-d. This indicated that the
densificationoccurringat temperatures
above950°Ccanbe monitoredusing IS.
= During impedance measurements, the anode and cathode keep switching sides, due to the
alternating current used. However at any particular moment, there is always an electrode acting
as the blocking anode, while the other one acting as the cathode.
Page 156
OF CLAY SINTERING PRCJCE
As is well established, the diameter of a semicircle in the complex impedance plot is
representative of the resistance. The semicircle corresponding to the bulk specimen is

well resolved from the spectra corresponding to the electrode effect. The diameter of the
semicircle can be taken as the real part at the cut-off frequency. By neglecting the
resistanceof leads that is less than 0.1% of the resistance of bulk clay, the diameter of the
semicircle can be taken as the bulk resistance. The electrical resistance can be described
as:
R= S/ß"Ae (7-4)
where ß is the conductivity, S is the distancebetweenelectrodesand A, is the area of

electrodes.During impedancemeasurements,the parameterAe can be consideredas a
constant,but both 6 and 8 changeddueto densificationof the clay compact. Therefore
both increasing electrical conductivity and decreasingdistance between electrodes
contributeto the decreasingresistanceof the specimen.
According to SEM observation and XRD analyses, the sintered clay compact mainly
contains a continuous glass phase and isolated pores where the quartz and mullite
particles are imbedded in the glass phase. To analyse the electrical conductivity of the
clay, we can treat this material as a two-phase composite. The continuous glass phase
containing crystal particles is treated as one single conducting phase while the isolated
pores can be regarded as an insulating phase. The structure of the clay is very similar to
the 3-0 connectivity pattern shown in a piezoelectric-pyroelectric composites studied
previously (Newnham et al. 1978). The conductivity of the insulating phase, e.g. pores,
can be assumedto be zero, the electrical conductivity of the composite can be estimated
using the equation (Mclachlan et al. 1990):
Gm 6h(1-f)m (7- 5)
where a,,, is the electrical conductivity of the composite, ah is the electrical conductivity
of the continuous conductive network in the composite, f is the fraction of the insulating
phase and m here is an exponent which is associated with the particle shape and
distribution (Mclachlan et al. 1990).
Another approachfor calculatingthe transportpropertiesof porous materials has been
developedas(Fanet al. 1992;Fan,1996):
cc /Dranfac (7- 6)
where V refers to the transport properties of a porous material, b,,, refers to the transport
properties of the matrix without porosity and fa,c is a topological parameter. The
topological parameter facdepends not only on the volume fraction of the matrix phase,
but also on the microstructure parameters, such as pore size, pore shape and pore
distribution. In general, it can be assumedto follow a power law (Fan et al. 1992; Fan,
1996):
fa, fa' (7- 7)
wherefa is the volume fraction of matrix, the m here is a parameterwhich is dependent

on severalmicrostructuralparametersas mentionedabove.Combining Eqs.7-6 and 7-7,
we have:
(De=(Dm fam (7- 8)
It is obviousthat Eqn.7-8 is the sameasEqn.7-5.
In Eqn.7-8, fa is equal to the ratio (Db/Dt) of the actual bulk density (Db) to the
theoretical density (Di) of the sample. Therefore, by combining Eqs. 7-4 &7-8, we
expressthe electricalconductance(G) in termsof the density,Dband the thicknessof the
specimen,S:
G=1/R=C"Dbm/8 (7- 9)
where C=Ae"a,,,/Dt', which can be assumed to be constant during sintering at a given

temperature.
Based on the impedance spectra measured at 1100°C (Fig. 7-3c), the conductance G of
the bulk sampleunder firing was obtainedas a function of sintering time (Fig. 7-17a.)
For comparison,the densityand dimensionalchangeof the clay compactsthat have been
sintered at 1100°C for various periods were measured.The isothermal densification

curve and isothermal shrinkage curve are shown in Figs. 7-17b and c, respectively.
Obviously, the three plots show the same trend.
&cIQ4
a n....
... a ...............
7x10' ..........
0c1CF4
Ll
basedon inpedace spectra(Fig.7-3c)
5o -4 ° predictionbasedon Egn7-9
4 0-4
25 b
11
18 C
-_---
17
ö 16-
15-
SinteringTirre (h)
Figure 7-17 (a) the dependence of bulk conductivity on sintering time; (b) and (c) the
density and dimensional variations as a function of sintering time, each value being
obtained by averaging five samples.
By using the data of Db and S shown in Figs 7-17b,c and assuming m=2, the
conductance of the sample was calculated according to Eqn. 7-9. The change of
conductance is displayed in Fig. 7-17a, showing the calculated data of the conductance
fits the measured curve from the impedance measurementsvery well. It is interesting to
note that the assumption of m as a constant of 2 gives a very good fit to the experimental
results, while m is dependenton the microstructural parameters,e.g. pore shapeand pore
distribution.
The value of m often has been found to be in the range of 1.65 to 2.0 (Mclachlan et at.
1990), so m=2 in this work is reasonable. When m=3-5, this implies extreme geometries
or complex structures for the dispersing phase (Balberg, 1987). In the clay during
sintering where the continuous phase is a glass phase and the material is quite pliable,
pores are more likely to have regular shapes and be homogeneously distributed, which
can be observed in Fig. 7-10 d. This might be the reason why m remains constant during
the densification process.
Impedancemeasurementsfor both green clay compacts during heating and sintered

bodies during cooling were conductedin-situ isothermally at high temperatures.In both
cases,the effect of the bulk specimenis well distinguished from the electrode cffect.
Instead of two semicircles, there is only one semicircle present in the complex impedance
plot corresponding to the bulk effect. This is because continuous conductive paths are
present in both cases.In the green clay compact, the continuous conductive path is due to
the surface conduction of mobile ions adsorbed on the surface of clay particles. In the
sinteredbodythe continuousglassyphaseforms a continuousconductivepath.
By employing the continuous conductive path model to simulate the impedance spectra,
IS parameters for both the bulk and electrode effects were obtained. Based on these IS
parameters, a lot of useful information can be generated, for example, the conductivity,
dielectric constant and dielectric loss of a bulk specimen, the capacitive effect and
inhomogeneity of electrode / specimen interface etc. A discontinuity in the Arrhenius
plot of the bulk conductivity and a drastic increase in the dielectric curve coincide with
the formation of liquid phase in clay compacts which occurs at --950°C. This was
confirmed by all the conventional analytical techniques used, i. e., XRD, dilatometer
measurement and SEM. Therefore IS is sensitive to the formation of a phase transition
during heating. An unusually large dielectric constant was found to be associated with
the bulk effect which needsfurther detailed study.
The electrodeeffect was found to have a strong temperaturedependence.

The activation
energiesfor the bulk conductivity and mobile ions associatedwith electrodeeffect were
calculated. Based on IS parameters, it has been found that the activation energy of the
latter is very close to the reported activation energy for non-bridging oxygen ions in an
alkali-depleted region (ADR). Therefore it can be assumed that the electrode effect is
reflective of the double layer charging in the ADR. The mobility of the non-bridging
oxygen ions in the ADR is thermally stimulated. Therefore the height of the low
frequency tail in the complex impedanceplot is indicative of the local temperature.
A shifting of impedance spectrum was observed for clay compacts under firing at
>950°C, which is a direct consequenceof a densification occurring in the clay compact.

The variation of the bulk conductivity with the sintering time, which can be deduced
from isothermal impedance spectra, was found to be in agreement with the shrinkage
curve. An equation relating the electrical conductivity to the density of the specimen has
been established.
In conclusion, IS can be used to detect the formation of a liquid phase, sense the local
temperaturearound the electrode, and monitor the densification of the specimen.

Therefore impedance spectroscopy is a promising tool for investigating the sintering
process of clay-basedceramics.
Page161
8 Conclusions And Suggestions For Future work
8.1 General Summary And Conclusions
The development of a simple non-destructive technique for characterisation of structural
ceramics is desirable. Although it has already been recognised by some scientists that IS
has an important area of application in the study of structural ceramics, most of the
previous studieshave been restricted to zirconia-based and cementitious materials. So far

no attempt seems to have appeared in the literature to characterise a wide range of
structural ceramics using IS. This project was designed to contribute to this new research
field, and to explore the potentiality of the IS technique in an attempt to extend its
application to the majority of structural ceramics. For this purpose, materials studied
represent a wide range of structural ceramics. The microstructure, phase distribution and
composition of thesestructural materials have been characterised.
To set up an equivalent circuit model for a specific system is essential for the correct
interpretation of its impedance spectra, which in turn strongly relies on a clear
understanding of electrical conduction mechanisms. Knowledge of the microstructure-

electrical conduction relationship is essential. For structural ceramics, the electrical
properties are not the major concern for their application, but provide the basis which can
be used to extract microstructural information from the corresponding impedance
spectra. Various conduction mechanisms and physical models have been reviewed in
chapter 2. The fundamentals of IS, such as the non-ideality of response and electrode
effects have been discussedin detail.
The real power of the IS technique lies in its capability of distinguishing the electrical
responses from different microstructural origins. An analytical approach and numerical

analyses have been conducted to illustrate why impedance spectra become resolved and
how different parameters corresponding to an equivalent circuit affect the spectral
resolution. It has been revealed that different representationshighlight different features,

therefore representations in various formalisms are useful in order to obtain more
information of the measuredsystem.
Page162
CHAPTER VIII CONCLUSIONS
AND SUGGESTIONSFOR FUTURE WORK
With regard to the electrical properties of a material, whether or not the electrical
responsesfrom the different microstructural origins can be distinguished depends on the

extent of the difference in their electrical properties, as different electrical properties
result in different relaxation frequencies. For example, IS does not discriminate the
grains and grain boundaries in A1203/SiC nanocomposites, presumably due to the
similarity of conductivities of grains and grain boundaries. Whereas for thermal barrier
coatings, there are three separated relaxation processes which correspond to the YSZ
bulk, the blocking effect from various defects in the top coat and oxide layer
respectively, due to their distinct relaxation frequencies (time constants).
The existence of continuous conduction paths masks the responses from grains and grain
boundaries in ceramics. For example, in the case of clay powder compacts, surface ionic
conduction is assumed to be a dominant conduction mechanism before the formation of

liquid phase, while after the appearanceof liquid phase,the liquid phase tends to form a
continuous network which can transport the charge carriers without experiencing the
blocking effect of other phase(s). For this reason there are no grain and grain boundary
effects shown in the impedancespectraof clay compacts.
The existence of a conductive phase could not lead to a separatecurve in the impedance
spectrum, as its relaxation frequency is too high to be measured. For example, the
dispersed SiC particles in A1203/SiC nanocomposites and mixed oxide in TBCs at 350°C
are quite conductive; in both casesit is believed that their relaxation processesshould be
outside of the measurable frequency range. However, the presence of the conductive
phase leads to a change of the capacitive effect. In this sense IS still is sensitive to the
existence of the conductive suspendedphases. This has been shown in the experimental
observations that the dielectric constant of A1203/SiC nanocomposites depended on the
SiC content and, more interestingly, the orientation of suspendedconductive particles in
composites had an influence on bulk capacitanceas well. This observation offered a base
on which a qualitative or even quantitative characterisation of a given microstructural
factor can be conducted. A quantitative establishment was attempted to explain the
dependenceof dielectric constant on SiC content as presentedin chapter 5.
Page 163
Microstructuralchangesare supposedto be relatedto the variations of correspondingIS
parameters, so the core issue in characterising structural ceramics using IS is the
establishment of the relationship between microstructure and IS parameters.
The exponential index n for CPE is often related to heterogeneities in materials. It has
been found that pure, monolithic A1203 material gives a nearly ideal Debye response,
while the incorporation of SiC particles apparently introduces frequency dispersion to the
electrical responseof A12O3/SiCnanocomposites.The depression angle increased as the
SiC content is increased.Meanwhile it was also found that the n values corresponding to
both the top coat and oxidation layer in TBCs varied with oxidation time. This gives
extra information concerning the degradation of materials, although the exact underlying
mechanism hasyet been established.
The capacitive effect has been found to be closely related to the geometric dimensions of
a given microstructure. This is useful especially for the characterisation of the multi-layer
structure. The inverse capacitanceof the oxide layer in TBCs has been illustrated to have
a linear relationship with the layer thickness. Modulus spectra can be used to highlight
the capacitive effects of different layers having different conductivities. The diameter of
a semicircle in a complex modulus plot is proven theoretically to be proportional to the

thickness. Meanwhile in practice examination of modulus spectra have been found to be
a very useful approach for non-destructive examination of the growth of the oxide layer
in both TBCs and A1203/SiC nanocomposites.
The change in resistivity (or conductivity) may reflect the change in phase composition
of a material. In TBCs, the top coat was found to show a decreasein its conductivity with
the heat treatment time. This was speculatedto be associatedwith the phase changes: t'-
phase -). t-phase + c-phase, and t-phase-*m-phase, which in turn could be an effective
indicator of the degradation of the top coat. In an impedance study of clay sintering, the
conductivity increase was also related to the formation of liquid phase in the clay. During
isothermal sintering, clay compacts show an increase in conductivity with sintering time,
which representsthe densification progress of clay compacts. A quantitative relationship

has been established to relate the variation of bulk conductivity to the bulk density.
Page 164
The electrode effect is an electrical response from the electrode/material interface.
Although complications may be involved in the analysis of the electrode effect, the
electrode effect does provide extra information regarding the microstructure and
electrical properties of the bulk material itself. For the case of clay compacts under firing,
the equivalent capacitanceof the electrode effect (platinum/clay interface) was found to
be closely related to the temperature at the electrode/material interface, which means it
may be usedto give indication of the local temperature.
A lot of valuableinformationconcerningmicrostructuralchangesin ceramicsthat occur

in service or in manufacturing processes can be generated non-destructively using IS.
Based on the researchresults presented in this thesis, the following conclusions can be
drawn:
1) The content and orientation of the conductive phase in insulator/electronic
conductor composites,are closely related to the IS parameters deduced from impedance

spectra. This proves that IS is capable of identifying the concentration, orientation and
distribution of the suspendedconductive phase(s).
2) A dramatic change in the electrical properties of insulator/electronic conductor
composites occurs when a percolating network of the conductive phase(s) is formed.

The oxidation scale formed at the surface of the composites with a percolating network
of the suspendedconductive phase(s) can be determined non-destructively using IS,

which is a difficult task for other test techniques, such as, SEM, weight gain
measurements,optical microscopic observations etc. to accomplish.
3) In TBCs, the electrical response from the oxide layer of the bond coat is well
separated from those from other phases. Well-defined semicircles in a complex

modulus plot were obtained. The oxide layer consists of two main components: mixed
oxide and alumina. The impedance spectrum obtained by room temperature
measurement detects the whole oxide layer, while at 350°C the impedance spectrum
measuresthe alumina layer.
Page165
4) There are two relaxation processes corresponding to the top coat in TBCs
measuredat 350°C; one corresponds to the bulk effect, the other to the blocking factors,
such as grain boundaries, cracks, pores and impurities etc. Both of them show a clear
trend of variation with heat treatment time, which can be used to indicate the extent of
degradation of the top coat.
5) During the sintering of clay-based ceramics, the formation of liquid phase and
the shrinkage rate of the ceramic body can be monitored by in-situ impedance
measurements.The electrode effect in impedance spectra is representative of the local
temperature, which can be an effective indicator of heat work for the firing of clay-
basedceramics.
8.2 Suggestions For Future Work
It is a new areaof researchto employ IS to characterisestructural ceramics. The research
presented in this thesis has indicated that this is a promising research field of scientific
interest, which may also bring great benefit to the manufacturing and applications of
structural ceramics. However this work representsonly a first step into this area, there is
a lot of space left for future research. With regards to the issues concerned in this thesis,
the following aspectsare suggestedfor future work.
1) In the fibre or particulate reinforced ceramic composite materials, it has been
recognised that the distribution of the dispersed phase, such as SiC whiskers, SiC
particles and carbon fibres, etc., is an important concern in the fabrication of these
materials. As mentioned earlier on, IS is sensitive to the distribution and orientation of
the dispersed particles. Further work should focus on the relationship between IS and
the anisotropic mechanical properties resulting from different distribution modes (such
as homogenous and clustered, oriented and isotropic etc.) of the dispersed conductive
particles.
Page166
2) For the top coat in TBCs, the change in phase composition is an important factor
leading to the degradation of the top coat, which then may contribute to the failure of
TBCs. In the research presented in chapter 6, it has been demonstrated that both the
conductance and capacitive effect of the top coat varied with the heat treatment
duration. This was postulated to be due to the phase transformation of zirconia: t'-
phase -* t-phase + c-phase and t-phase-m-phase. However, more detailed work on

phase identification and quantitative determination of phase composition is required to
substantiate this proposition.
3) In the impedance spectra of TBCs, the mixed response was believed to be a
combined contribution from all of the blocking factors contained in the top coat (YSZ)
in TBCs, such as grain boundaries, cracks, pores and impurities etc. Our research
results showed clearly that all the three IS parameters:R, n and A corresponding to this
response had clear trends of variation as a function of heat treatment time. However, an
unambiguous explanation for these variation patterns is not yet available. Further work
is necessary to differentiate the influences from different blocking factors in order to
find a satisfactory interpretation for this phenomenon, which may also lead to more in-
depth knowledge about the degradation of the top coat in TBCs.
4) As has been addressedin chapter 6, there was always a tail-like curve in the low
frequency range in the impedance spectrum of TBCs. The author attributed this tail-like
curve to a response of the top coat/mixed oxide interface. This curve appeared to
change with the heat treatment duration, which indicates the interface might keep
changing during heat treatment. It has already been established that most of the TBCs
failures occur at, or near, the top coat/bond coat interface. Further detailed study on the
impedance of this electrode effect and its dependenceon interface microstructure might
be able to provide insight into the failure mechanism and possibly even a new approach
to monitor the failure of TBCs.
5) A non-destructive technique of characterising the cracks and microcracks in
ceramics is always desirable. Quartz is the most commonly used filler in clay-based
ceramics. One of the issues in the production of whiteware and porcelain is the
deterioration in mechanical properties as a result of the ß-)-ocquartz inversion at 573°C
during the cooling process, which results in a volume decrease of 2% and hence
Page 167
produces sufficient strain to cause cracking of the glassy matrix and quartz particles
(Carty and Senapati, 1998). This cracking should lead to a substantial change in
electrical conductivity of ceramics. In the researchpresented in chapter 7, we have not

found a change in conductivity of sintered clay bodies during the cooling process. This
might be due to the lack of large quartz particles in the sintered body. Further work is
needed to confirm whether or not IS is capable of detecting the presence of cracks and
microcracks in traditional, clay-based ceramics.
Page 168
References
Abrahams I. and Hadzifejzovic, E., (2000) Lithium ion conductivity and thermal
behaviour of glasses and crystallised glasses in the system Li20-Al203-TiO2-P205, Solid
State Ionics, 134,249-257.
Abrantes, J. C. C., Labrincha J. A. and Frade, J. R., (2000a), An alternative
representation of impedance spectraof ceramics, Mater. Res. Bull., 35,727-740.
Abrantes, J. C. C., Labrincha J. A. and Frade, J. R., (2000b), Representations of

impedance spectra of ceramics part I. simulated study cases,Mater. Res. Bull., 35,955-
964.
Abrantes, J. C. C., Labrincha J. A. and Frade, J. R., (2000c), Representationsof

impedancespectraof ceramicspart II. Spectraof polycrystalline SrTiO3, Mater. Res.
Bull., 35,965-976.
Ahn, H. - S., (1998), Detection of contact damage in ceramics by an ultrasonic method, J.

Mater. Res., 13,1899-1903.
Ambrus, J. H., Moynihan, C. T. and Macedo, P. B., (1972), Conductivity relaxation in a
concentrated aqueouselectrolyte solution, J Phys. Chem., 76,3287-3295.
Andrade, C., Blanco V. M., Collazo, A., Keddam, M., Novoa, X. R. and Takenouti, H.,
(1999), Cement paste hardening process studied by impedance spectroscopy,
Electrochimica. Acta., 44,4313-4318.
Anseau, M. R., Deletter, M. and Cambier, F., (1981), The separation of sintering
mechanismsfor clay-basedceramics,Trans.J Br. Ceram.Soc.,80,142-146.
Aoki, M., Chiang, Y. -M., Kosacki, I., Lee, I. J. R., Tuller, H. and Liu, Y. P., (1996),
Solute segregationand grain boundary impedancein high-purity stabilisedzirconia, J
Am. Ceram.Soc.,79,1169-1180.
Armstrong, R. D., Dickinson, T. and Willis, P. M., (1974), The A. C. impedance of
powdered and sintered solid ionic conductors, J. Electroanal. Chem., 53,389-405.
Page 169
Armstrong, R. D. and Todd, M., (1995), Interfacial electrochemistry, p. 265, in Solid
P. G. Bruce,(ed),CambridgeUniversity Press,Cambridge.
State Electrochemistry,
Badwal, S. P. S., (1988), Proceedings of the international seminar on solid state ionic
devices, World of Scientific Publishing, Singapore, p165.
Badwal, S. P. S., (1995), The grain boundary resistvity in zirconia-based materials: effect
andimpurities,SolidStateIonics, 76,67-80.
of sinteringtemperatures
Badwal, S. P. S., Ciacchi, F. T., Rajendran, R. and Drennan, J., (1998) An investigation
of conductivity, microstructure and stability of electrolyte compositions in the system 9

mol% (Sc203-Y2O3)-ZrO2(A1203),
SolidStateIonics, 109,167-186.
Balberg, I., (1987), Tunneling and nonuniversalconductivity in compositesmaterials,

Phys.Rev.Lett., 59,1305-1308.
Bartlett, A. H. and Maschio, R. D., (1995), Failure mechanisms of a zirconia-8 wt%
yttria thermal barrier coating, J. Am. Ceram. Soc., 78,1018-24.
Bauerle, J. E., (1969), Study of solid electrolytepolarisation by a complex admittance

method,J. Phys.Chem.Solids,30,2657-2670.
Boccaccini D. Nand Boccaccini, A. R. (1997), Dependence of ultrasonic velocity on
porosity and pore shapein sintered materials, J. Nondes. Eva., 16,187.
Bockris, J. ON, Kitchener, J. A., Igatowicz, S. and Tomlinson, J. W., (1952) Electric
conductance in liquid silicates, Trans. Faraday Soc., 48,75-91.
Bonanos, N., Steele, B. C. H. and Butler, E. P., (1987), Application of impedance
spectroscopy,p191, in ImpedanceSpectroscopy,J. R. Macdonald,(ed.), John Wiley &

Sons, New York.
Brindley, G. W. and Nakahira, M., (1959a), The kaolinite-mullite reaction series: 1, a
survey of outstanding problems, J. Am. Ceram. Soc., 42,311-314.
Brindley, G. W. and Nakahira, M., (1959b), The kaolinite-mullite reaction series: II, a
survey of outstanding problems, J. Am. Ceram. Soc., 42,314-318.
Page170
Bruce P. G. and West,A. R. (1983),The A-C conductivity of polycrystalline LISICON,
Lie+2,Zni.,,GeO4 and a model for intergranular constriction resistance, J. Electrochem.
Soc., 130,662-669.
Buchanan, R. C., (1991), Properties of ceramic insulators, p1, in Ceramic Materials for
Electronics: Processing,Properties and Applications, R. C. Buchanan, (ed), Marcel

Dekker, Inc., New York.
Cann D. P. and Randall, C.A., (1996), Electrode effects in positive temperature
coefficient and negative temperaturecoefficient devices measuredby complex-plane

impedanceanalysis,J. Appl. Phys.,80,1628-1632
Carlson, D.E., Hang K. W. and Stockdale,G. F., (1972), Electrode `polarisation' in

alkali-containing glasses,J. Am. Ceram. Soc., 55,337-34.
Carlson, D.E., Hang K. W. and Stockdale,G. F., (1974), Ion depletion of glass at a
blocking anode II properties of ion depleted glasses,J.. Am. Ceram. Soc., 57,295-300.
Carty, W. M. and Senapati, U. (1998), Porcelain-raw materials, processing, phase
evaluations, and mechanical behaviour, J. Am. Ceram. Soc, 81,3-20.
Challis, R., andTebbutt,J., (1998),Ultrasonicwave interactionswith colloidal mixtures,

p23, in Nondestructive Evaluation of Ceramics, C. H. Schilling and J. N. Gray, (eds)
(Vol 89, Ceramic Transaction), American Ceramic Society, Westerville, Ohio.
Chiang, Y.-M., Lavik E. B. and Blom, D. A., (1997), Defect thermodynamicsand

electrical properties of nanocrystallineoxides: pure and doped CeO2,NanoStructure
Materials, 9,633-642.
Choi G. M., and Tuller, H. L., (1988), Defect structure and electrical-properties of single-
J. Am. Ceram.Soc.,71,201.
crystalBa0.03Sr0.97TiO3
,
Christensen, B. J., Coverdale, R. T., Olson, R. A., Ford, S. J., Garboczi, E. J., Jennings,
H. M. and Mason, T. 0., (1994), Impedance spectroscopy of hydrating cement-based
materials: measurement, interpretation and application, I Am. Ceram. Soc., 77,2789-

2804.
Page 171
Christensen, P. A. and Hamnett, A., (1994), Techniques giving mechanistic information,
p154, in Techniques and Mechanisms in Electrochemistry, Blackie Academic &

Professional,London.
Christensen, B. J., Mason, T. 0. and Jennings, H. M., (1992), Influence of silica fume on
the early hydration of portland cements using impedance spectroscopy, J Am. Ceram.
Soc,75,939-945.
Cole, K. S. and Cole, R. H., (1941), Dispersion and adsorption in dielectrics I, alternating
current characteristics,J. Chem. Phys., 9,341-351.
Curtis, P. T., Gates, J. and Margerison, G. G., (1993), The selection of cyclic load
frequencyfor the fatiguetestingof fibre reinforcedpolymeric composites,Rae Technical
Report 93017(Farnborough,
Guildford, UK).
De Levie, R., (1967), Electrochemical responseof porous and rough electrodes, p 329, in
Advancesin Electrochemistryand ElectrochemicalEngineering, vol. VI, Interscience,
New York.
Demasi-Marcin, J. T., Sheffler, K. D. and Bose, S., (1990), Mechanisms of degradation
and failure in a plasma-deposited

thermal barrier coating,J. Eng. Gas TurbinesPower,
Vol. 112,521-526
Dessemont L. and Kleitz, M., (1992), Effects of mechanical damage on the electrical
propertiesof zirconiaceramics,J. Euro. Ceram.Soc.,9,35-39.
Dessemond,L., Muccillo, R., Henault M. and Kleitz, M., (1993), Electric conduction-
blocking effects of void and secondphasesin stabilisedzirconia, Appl. Phys., A57,57-
60.
Djurado, E., Dessemond,L. and Roux, C., (2000), Phasestability of nanostructured

tetragonal zirconia polycrystals versus temperature and water vapour, Solid State Ionics,
136-137,1249-1254.
Doi, A., (1987), Ionic-conductionand conductionpolarisation in oxide glass,J Mater.

Sci., 22,761,
Page 172
Drennan, J. and Auchterlonie, G., (2000) Microstructural aspects of oxygen ion
conduction in solids, Solid State Ionics, 134,75-87.
Drevet, C., Henault, M. and Fouletier, J., (2000), Oxygen electrode reaction on stabilised
zirconia under high oxygen pressure (up to 100bar), Solid State Ionics, 136-137,807-
812.
Dunyak, T. J., Stiuchcomb, W. W. and Reifsnider, K. L. (1992), Examination of selected

NDE techniques for ceramics composites damage detection, p3, in Composite materials,
ASTM, STP 1128, J. E. Masters, (ed), American Society for Testing And Materials,
Philadelphia.
Fairchild, C. 0. and Peters,M. F. (1926), Characteristicsof pyrometric cones, J. Am.

Ceram.Soc.,9,701-742
Fan, Z., (1996), A microstructural, approach to the effective transport properties of
multiphase composites,Philosophical Magazine A, 73,1663-1684.
Fan, Z., TsakiropoulosP. and Miodownik, A. P., (1992), Prediction of young's modulus
of particulate two phasecomposites, Mater. Sci. Technol., 8,922-929.
Fleig, J., (2000), The influence of non-ideal microstructures on the analysis of grain
boundary impedance,Solid State Ionics, 131,117-127.
Fleig, J. and Maier, J., (1998), A finite element study on the grain boundary impedance
of different microstructures, J Electrochem. Soc, 145,2081-2089.
Fleig, J. and Maier, J., (1999a), The impedance of ceramics with highly resistive grain
boundaries:validity and limits of the brick layer model,J Euro. Ceram. Soc, 19,693-
696.
Fleig, J. and Maier, J., (1999b), Finite element calculations on the impedance of
electroceramics with highly resistive grain boundaries: I, laterally inhomogeneous grain

boundaries, J. Am. Ceram. Soc, 82,3485-3493.
Fleming, W. J., (1981), A physical understanding of solid-state humidity sensors, Proc.
Int. AutomotiveMeet.,SAE, Detroit,USA, paperno. 810432,51-62.
Page 173
Ford, S. J., ShaneJ. D. and Mason, T. 0., (1998), Assignment of features in impedance
spectra of the cement paste/steelsystem,Cem. Concr. Res., 28,1737-1751.
Gabrielli. C and Keddam. M., (1996) Contribution of electrochemical impedance
spectroscopy to the investigation of the electrochemical kinetics, Electrochemica Acta,

41,957-965.
Glasser, F. P., Lachowski, E. E. and Macphee, D. E., (1987), Compositional model for
calcium silicate hydrate (C-S-H) gels, their solubilities and free energies of formation, J.
Am. Ceram. Soc., 70,481-485.
Goller M. H. and Pye, L. D., (1998), Light scatteringof particles in glasses,p327, in

NondestructiveEvaluation of Ceramics,C. H. Schilling and J. N. Gray, (ed) (Vol 89,
Ceramic Transaction), American Ceramic Society, Westerville, Ohio.
Gomes W. P. and Vanmaekelbergh, D., (1996), Impedance spectroscopy at

semiconductorelectrodes:review and recent development,ElectrochemicaActa, 41,
967-973.
Gu, P., Xu, Z., Xie P. and Beaudoin, J. J., (1993), An A. C. impedance spectroscopy
study of micro-cracking in cement-basedcomposites during compressive loading, Cem.

Concre. Res.,23,675-682.
Gusmano, G., Bianco, A., Montesperelli, G. and Traversa, E., (1996), An EIS study of
the humidity-sensitive electrical conduction of alkali-doped Ti02 films, Electrochemica.
Acta, 41,1359-1368.
Haynes, J. A., Gigney, E. D., Ferber, M. K. and Porter, W. D., (1996), Oxidation and
thermalbarrier coatingsystem,Surf. Coat. Tech., 86-
degradationof a plasma-sprayed
87,102-108.
Hirose, N. and West, A. R., (1996), Impedance spectroscopy of undoped BaTiO3

ceramics, J. Am. Ceram. Soc., 79,1633-1641.
Hodge,I. M., IngramM. D. and West,A. R., (1975),New methodfor analysingthe A.C.
behaviour of polycrystalline solid electrolytes, J. Electroanal. Chem., 58,429-432.
Page 174
Hodge, I. M., Ingram M. D. and West, A. R., (1976), Impedance and modulus
spectroscopy of polycrystalline solid electrolytes, J. Electroanal. Chem., 74,125-143.
Huggins, R. A., (1978), Crystal structure and fast ionic conduction, p27, in Solid
Electrolytes, General Principal, Characterisation, Materials, Applications, P.
Hagenmuller & W. Van Gool, (eds), Academic Press,London.
Hughes,A. E. and Badwal,S. P. S., (1991),Impurity and yttrium segregationin yttria-

tetragonal zirconia, Solid State Ionics, 46,265-274.
Huss, S. P., Schilling, C. H., Gray, J. N., and Maranville, C. W., (1998), Recent advances
in x-ray characterisation of composites and tape cast ceramics, p45, in Nondestructive
Evaluation of Ceramics, C. H. Schilling and J. N. Gray, (eds), (Vol 89, Ceramic
Transaction), American Ceramic Society, Westerville, Ohio.
Hwang, J.-H. Mason,T. 0., Buehler,M. F., Darab,J. G., Matson,D. W. and Linehan, J.
C., (1995), Characterisation of humidity-sensing NiO-Ni(OH)2 nanocomposites by
impedance spectroscopy,Mater. Sci. Eng., A204,252-257.
Ishida, A. and Miyayama, M., (1994), Prediction of fracture and detection of fatigue in
ceramic composites from electrical resistivity measurements,J. Am. Ceram. Soc., 77,
1057-1061.
Jacobsen, T., Zachau-Christiansen, B., Bay, L., and Skaarup, S., (1996) SOFC cathode
mechanisms, p29, in Proceedings of the 17l" Riso International Symposium on Materials

Science, F. W. Poulsen, N. Bonanos, S. Linderoth, M. Mogensen, and B. Zachau-
Christiansen, (eds), Riso National Laboratory, Roskilde, Denmark.
Jiles, D., (1994), Electronicpropertiesof semiconductors,p130, in Introduction to the

Electronic Propertiesof Materials,by D. Jiles, Chapman& Hall, London.
Johnson, W. B. and Worrell, W. L., (1987), Solid state devices, p238, in Impedance
Spectroscopy, J. R. Macdonald, (ed), John Willey & Sons,New York.
Jonscher, A. K., (1978), Analysis of the alternating current properties of ionic
conductors, J. Mater. Sci, 13,553-562.
Page 175
Jonscher A. K. and Reau, J.-M., (1978), Analysis of the complex impedance data for p-
PdF29J. Mater. Sci, 13,563-570.
Kathrina, T. and Rawlings, R. D., (1996), Ultrasonic non-destructive evaluation of
alumina powder during compaction, Bri. Ceram. Trans., 95,233.
Kerner Z. and Paijkossy, T., (1998), Impedance of rough capacitive electrodes: the role
of surfacedisorder,J. Electroanal.Chem.,448,139.
Khalili A. and Kromp, K., (1993), On the correlationbetweenmechanicaldamageand

the electrical properties of zirconia ceramics,J. Euro. Ceram. Soc., 11,455-461.
Kim, W. J., Taya, M., Yamada, K. and Kamiya, N., (1998), Percolation study on
electrical resistivity of SiCISi3N4composites with segregateddistribution, J. App!. Phys.,

83,2593-2598.
Kingery, W. D., Bowen, H. K. and Uhlmann, D. R., (1976a), Structure of Crystals, p77,
in Introductionto Ceramics,SecondEdition, Burke,E., Chalmers,B., and Krumhansl,J.
A., (eds), John Wiley & Sons, New York.
Kingery, W. D., Bowen and H. K. Uhlmann D. R., (1976b), Grain Growth, Sintering and
Vitrification, p448, in Introduction to Ceramics, Second Edition, Burke, E., Chalmers,
B., and Krumhansl, J. A., (eds), John Wiley &Sons, New York.
Kingery, W. D., Bowen, H. K. and Uhlamann, D. R., (1976c), Electrical conductivity,
p851, in Introduction to Ceramics, Second Edition, Burke, E., Chalmers, B., and
Krumhansl, J. A., (eds) john Wiley & Sons, New York.
Kleitz, M., Dessemond, L. and Steil, M. C., (1995), Model for ion-blocking at internal
interfaces in zirconias, Solid State Ionics, 75,107-115.
Knauth, P. and Tuller, H. L., (1999),Electrical and defectthermodynamicpropertiesof

nanocrystalline titanium dioxide, Appl. Phys. 85,897-902.
Komarenko, P., Messier, R. W. and Scarton, H. A., (1994) Ultrasonic in-situ monitoring
of sintering in alumina, Intl. J. Powd. Meta., 30,67-76.
Page 176
Kubaschewski 0. and Hopkins, B. E., (1962), Electrical properties of metal compounds-
basic information, p25, in Oxidation of Metals and Alloys, Butterworth & Co. Limited,
London.
Lanferdi, S. and Rodrigues, A. C. M., (1999), Impedance spectroscopy study of the
electrical conductivity and dielectric constant of polycrystalline LiNbO3, J Appl. Phys.,

86,2215-2219.
Lang, G. and Heusler, K. E (1998), Remark on the energetics of interfaces exhibiting
constant phaseelement behaviour, J. Electroanal. Chem., 457,257-260.
Lange, F. F., (1982a), Transformation toughening 1: size effects associated with the
thermodynamics of constrained transformation, J. Mater. Sci., 17,225-234.
Lange,F. F., (1982b),Transformationtoughening2: contributionto fracturetoughness,

J. Mater. Sei., 17,235-239.
Lee, J., Chung, T-J. and Kim, D-Y., (2000) Electrical and microstructural
characterisation on nitrogen-stabilised zirconia, Solid State Ionics, 136-137,39-44.
Lee, W. Y., Stinton, D. P. and Berndt, C. C., Erdogon, F., Lee, Y. D. and Mutasim, Z.,
(1996), Concept of functionally graded materials for advanced thermal barrier coating
applications, J. Am. Ceram. Soc., 79,3003-12.
Liu, S. H., (1985), Fractal model for the AC response of a rough interface, Phys. Rev.
Lett., 55,529-532..
Luthra, K. L., Park, H. D., (1990), Oxidation of silicon carbide-reinforced oxide-matrix
compositesat 1375°to 1575°C,J. Am. Ceram.Soc.,74,1014-1022.
Macdonald, J. R. and Franceschetti, D. R., (1987), Physical and electrochemical model,
p84, in Impedance Spectroscopy,J. R. Macdonald, (ed), John Willey & Sons, New York.
Macdonald, J. R. and Johnson W. B., (1987), Fundamentals of impedance spectroscopy,
pl, in Impedance Spectroscopy,J. R. Macdonald, (ed), John Willey & Sons,New York.
Macedo, P. B., Moynihan, C. T. and Bose, R., (1972), Role of ionic diffusion in
polarisation in vitreous ionic conductors, Phys Chem. Glass, 13,171-176.
Page 177
Macphee, D. E., Sinclair, D. C. and Cormack, S. L., (1997), Development of an
equivalent circuit model for cement pastes from microstructural considerations, J. Am.
Ceram.Soc.,80,2876-2884.
Maiti H. S. and Basu, R. N., (1986), Complex-plane impedance analysis for semi-
conducting barium titanate, Mater. Res. Bull., 21,1107-1114.
McCafferty E. and Zettlemoyer, A. C.,(1971), Adsorption of water vapour on a-Fe203,

Discuss. Faraday Soc,. 52,239-263.
McCann J. F. and Badwal, S. P. S., (1982), Equivalent circuit analysis of the impedance
response of semiconductor/electrolyte/counterelectrode cells, J. Electrochem. Soc.,129,

551-559.
McCarter, W. J., Garvin S. and Bouzid, N., (1988), Impedance measurementson cement
paste,J Mater. Sci. Lett., 7,1056-1057.
McCaughey, W. J. and Neff, J. M. (1944), Notes on constitution of pyrometric cones,
Am. Ceram.Soc.Bull., 23,373-375.
McGonnagle, W J., (1971), Preface, page vii, in Nondestructive Testing, Gordon and
Breach Science Publishers, London.
Mclachlan, D. S., Blaszkiewicz, M. and Newnham, R. E., (1990), Electrical resistivity of
composites,J. Am. Ceram. Soc., 73,2187-2203.
Meier, S. M. and Gupta, D. K., (1994), The evolution of thermal barrier coating in gas
turbine engine applications, Transaction of the ASME, 116,250-257.
Miller, R. A., (1987),Currentstatusof thermalbarriercoatings-an overview,Surf Coat.

Technol.,30,1-11.
Miller, R. A., Garlick, R. G., and Smialek, J. L., (1983), Phase distributions in Plasma-
sprayedzirconia-yttria,Am. Ceram.Soc.Bull., 62,1355-1358.
Miller, R. A., Smialek, J. L., and Garlick, R. G., (1981), Phase stability in plasma-
sprayed zirconia-yttria, partially stabilised zirconia-yttria, p241, in Advances in
Page178
Ceramics, Vol 3, Science and technology of zirconia, Heuer A. H. and Hobbs. L. W.,
(eds), American Ceramic Society, Columbus, Ohio.
Morgan, P. E. D., and Koutsoutis, M. S., (1986), Electrical conductivity measurementsto

detect suspectedliquid phasein the A1203-lmol% Ti02-0.5 mol% NaO1r2and other
systems,J. Am. Ceram. Soc., 67, c254-c255.
Morrison, F. D. Sinclair D. C. and West, A. R., (2001), Characterisation of lanthanum-

doped barium titanate ceramics using impedance spectroscopy, J. Am. Ceram. Soc., 84,
531-538.
Motheo, A. J., Sadkowski, A. and Neves, R. S., (1997), Electrochemical immittance

spectroscopyapplied to the study of the single crystal gold/aqueousperchloric acid
interfaceJ. Electroanal.Chem.,430,253.
Muccillo, E. N. S. and Kleitz, M., (1996), Impedance spectroscopy of Mg-stabilised
zirconia and cubic phasedecomposition, J. Euro. Ceram. Soc., 16,453-465.
Muccillo, R., Muccillo, E. N. S. and Saito,N. H., (1998), Thermal shockbehaviourof

Zr02: MgO solid electrolytes,Mater. Lett., 34,128-132.
Murugan, G. S. and Varma, K. B. R., (2001), Characterisation of lithium borate-bismuth
tungstate glassesand glass-ceramicsby impedance spectroscopy, Solid State Ionics, 139,

105-112.
Nevadunsky, J. J. and Lucas, J. J. (1975), Early fatigue damage detection in composite
materials, J. Composite Mater., 9,394-408.
Newnham, R. E. Skinner, D. P. and Cross, (1978), L. E., Connectivity and piezoelectric-
pyroelectric composites,Mater. Res. Bull., 13,525-36.
Niihara, K., (1991), New design concept of structure ceramic-ceramic nanocomposites,J.

Ceram.Soc.Jpn., 99,974-982.
Nowotny, J., Redecka, M., Rekas, M., Sugihara, S., Vance, E. R. and Weppner, W.,
(1998), Electronic and ionic conductivity of Ti02 single crystal within the n-p transition
range, Ceramics Int., 24,571-577.
Page 179
Ogawa, K., Minkov, D., Shoji, T., Sato, M. and Hashimoto, H., (1999), NDE of
degradation of thermal barrier coating by means of impedance spectroscopy, NDT&E
International, 32,177-185.
Ohji, T., Hirano, T., Nakahira,N., and Niihara, K., (1996), Particle/matrixinterfaceand
its role in creep inhibition in alumina/silicon carbide nanocomposites., J. Am. Ceram.
Soc.,79,33-39.
Okada,K., Otsuka,N. and Ossaka,J., (1986),Characterisationof spinel phaseformedin

the kaolin-mullite thermal sequence,.1 Am. Ceram. Soc, 69, c251-c253.
Pape, D. A., (1998), Sonic detection of manufacturing flaws in ceramic components,
p291, in. Nondestructive Evaluation of Ceramics, C. H. Schilling and J. N. Gray, (eds),

(Vol 89, Ceramic Transaction), American Ceramic Society, Westerville, Ohio.
Provenzano,V. Boesch, L. P., Volterra, V., Moynihan, C. T. and Macedo, P. B., (1972),
Electricalrelaxationin Na2O.3SiO2glass,J Am. Ceram.Soc.,55,492-496.
Raistrick, I. D., (1987), Electrical analogsof physical and chemical processes,p27, in

Impedance Spectroscopy,J. R. Macdonald, (ed), John Willey & Sons, New York.
Ravaine D. and Souquet, J. L., (1978), Ionic conductive glasses, p277, in Solid
Electrolytes, General Principal, Characterisation, Materials, Applications, P.
Hagenmuller& W. Van Gool, (eds), Academic Press,London.
Rhoderick, G. H., (1978), Introduction and synopsis, page 1-17, in Metal-Semiconductor

Contacts,ClarencePress,Oxford.
Rodrigues,C. M. S., LabrinchaJ. A. and Marques,F. M. B., (1997), Study of yttria-

stabilised zirconia-glass composites by impedance spectroscopy, J. Electrochem. Soc.,
144,4303-4309.
Rodrigues, C. M. S., Labrincha J. A. and Marques, F. M. B., (1998), Monitoring of the
corrosion of YSZ by impedance spectroscopy,J. Euro. Ceram. Soc., 18,95-104
Ruud, C. 0., (1984), Preface, page v, in Nondestructive Methods for Material Property
Determination, C. 0. Ruud and R. E. Green, Jr., (eds), Plenum Press,New York.
Page 180
Scheider, W., (1975), Theory of the frequency dispersion of electrode polarisation:
topology of networks with fractional power frequency dependence,J. Phys. Chem., 79,
127-136.
Schilling C.H. and Gray, J. N. (1998), Needs and opportunities for NDE in ceramic
processing", pl, in Nondestructive Evaluation of Ceramics, C. H. Schilling and J. N.

Gray, (eds) (Vol 89, Ceramic Transaction), American Ceramic Society, Westerville,
Ohio.
Schillington, E. A. G., and Clarke, D. R., (1999), Spalling failure of a thermal barrier
coating associated with aluminium depletion in the bond coat, Acta. Meta., Vol 47 No. 4,
1297-1305.
Sciti, D., and Bellosi, A., (1998), Oxidation behaviour of alumina-silicon carbide
J. Mat. Sci., 33,3823-3830.

nanocomposites.
Scott,H. G., (1975), Phaserelationshipsin the zirconia-yttria system,J. Mater. Sci., 10,
1527-1535.
Scribner Associates, (1998), Impedance /gain phase graphing and analysis software-
operating manual, Version 2.1, p5.1.
Sergo,V. and Clarke, D. R., (1998), Observationof subcritical spall propagationof a

thermal barrier coating," J Am. Ceram. Soc., 81,3237-42.
Shuk, P., Wiemhofer, H. D., Guth, U., Gopel, W. and Greenblatt, M., (1996) Oxide ion
conducting solid electrolytes basedon Bi203, Solid State Ionics, 89,179-196.
Shuller,K. H., (1964),Reactionbetweenmullite and glassyphasein porcelains,Trans.

Br. Ceram. Soc., 63,102-107.
Sidebottom, D. L., Green, P. F. and Brow, R. K., (1995), Comparison of KWW and
power law analysis of an ion-conducting glass,J. Non-Cryst. Solids, 183,151-160.
Sinclair, D. C. and West, A. R., (1989), Impedance and modulus spectroscopy

of
semiconducting BaTiO3 showing positive temperature coefficient of resistance, J. Appl.
Phys., 66,3850-3856.
Page181
Smith, D. S., Lavielle D. and Orange, G., (1994), Electrical response of microcracks in
alumin/zirconia ceramic composites, Silica. Indus., 5-6,165-167.
Solartron Group, (1999), 1296 Dielectric Interface Operating Manual, Solartron,

Farnborough, UK.
Song, G. -L., (2000), Equivalent circuit model for AC electrochemical impedance
spectroscopyof concrete, Cem. Concr. Res., 30,1723-1730.
Song, H. Dao, L. -H., (1999), Electrochemical impedance

-K., Jung, Y. -H., Lee K. -H. and
spectroscopy of porous electrodes: the effect of pore size distribution, Electrochimica.
Acta., 44,3513-1519.
Sonuparlak, B., Sarikaya M. and Aksay, I. A., (1987), Spinel phase formation during
980°C exothermic reaction in the kaolin-mullite reaction series, J. Am. Ceram. Soc., 70,
837-842.
Souquet, J. L., (1995), Ionic transport in glassy electrolytes, p74, in Solid State
Electrochemistry, P. G. Bruce, (ed), Cambridge University Press, Cambridge.
Steil, M. C., Thevenot F. and Kleitz, M., (1997), Densification of yttria-stabilised
zirconia: impedance spectroscopyanalysis, J. Electrochem. Soc., 144,390-398.
Sternitzke, M., Derby, B. and Brook, R. J., (1998), Alumina/silicon carbide

nanocomposites by hybrid polymer/powder processing: microstructures and mechanical
properties, J Am. Ceram. Soc.,81,41-48.
Sterns, C.A., Kohl, F.J. and Fryburg, G. C., (1974), Oxidative vapour kinetics of Cr203
in oxygenfrom 1000C to 13000,J. Electrochem.Soc.,121,953-959.
Tawancy, H. M., Sridhar, N., Abbas N. M. and Rickerby D., (1998), Failure mechanism
of a thermal barrier coating system on a nickel-base superalloy, J. Mater. Sci., 33,681-

686.
Tiefenbach, A. and Hoffmann, B., (2000), Influence of a crack on the electrical

impedance of polycrystalline ceramics, J. Euro. Ceram. Soc., 20,2079-2094.
Page182
Torrents, J. M., Easley, T. C., Faber, K. T., Mason, T. 0., and Shah, S. P., (2001),
Evolution of impedance spectra during debonding and pullout of single steel fibres from
cement,J. Am. Ceram.Soc.,84,740-746.
Torrents, J. M., Mason T. 0. and Garboczi, E. J., (2000), Impedance spectra of fibre
reinforced cement based composites: a modelling approach, Cem. Concr. Res., 30,585-
592.
Traversa,E., (1995), Ceramic sensorsfor humidity detection: the state-of-the-artand

future developments, Sensorsand Actuators, B23,135-156.
Tuller, H. L., (2000), Ionic conduction in nanocrystalline materials, Solid State Ionics,
131,143-157.
Turdogan, E. T., (1983), Chapter 2 Structural aspects, p57, in Physicochemical

Properties of Molten Slags and Glasses,The Metals Society, London.
Van de Voorde, M. H. (1998), Non-destructive evaluation (NDE) of carbon-carbon (C-C)
andceramiccompositematerial,Bri. Ceram.Trans.,97,287-292.
Wadley, H. N. G., and Mehrabian, R., (1984), Acoustic emission for in-process
monitoring and microstructure control, p207, in Nondestructive Methods for Material

Property Determination, C. 0. Ruud and R. E. Green, Jr., (eds), Plenum Press,London.
Wang, S. -K., Fu. X. -L. and Chung, D. D. L., (1998) Nondestructive evaluation of
brittle-matrix composites during loading by electrical resistance measurement, p145, in
Nondestructive Evaluation of Ceramics, C. H. Schilling and J. N. Gray, (eds) (Vol 89,
Ceramic Transaction), American Ceramic Society, Westerville, Ohio.
Wang. C. -A., Huang, Y., Xie, Z. -P., Li, Y. and Zhang, Z. -T., (2000), Complex
impedanceanalysis on the orientation effect of whiskers in oriented silicon carbide
whisker/silicon nitride composites,J. Am. Ceram. Soc., 83,2689-2692.
Waser R. and Hagenbeck,R., (2000), Grain Boundary in dielectric and mixed conducting
ceramics, Acta. Mater., 48,797-825.
Page183
West, A. R., (1995), Crystalline solid electrolytes I: general considerations and the major
materials, p7, in Solid state Electrochemistry, P. G. Bruce, (ed), Cambridge University

Press.
Whalen P. J., Reidinger, F., Correale, S. T. and Marti, J., (1987), Yttria migration in Y-
TZP during high-temperature sintering, J. Mat. Sci., 22,4465-4469.
Wolfe, D., and Singh, J., (1998), Functionally gradient ceramic/metallic coatings for gas
turbine components by high-energy beams for high temperature applications, J. Mater.

Sci., 33 3677-92.
Wu, B. C., Chang, E., and Chang, S., (1989), Degradation mechanisms of Zr02-8wt%
Y203/Ni-22Cr-1OA1-1Y thermal barrier coatings, J. Am. Ceram. Soc.,72,212-218.
Wu, H. Z., Lawrence, C. W., Roberts, S. G., and Derby, B., (1998), The strength of
A1203/SiC nanocompositesafter grinding and annealing, Acta Mater., 11,3839-3848.
Zhang, J. -S., Huang, H., Cao, L. -H. and Xia, F., (1992), Semiconductive property and
impedancespectraof alumina-siliconcarbidewhisker composites,J. Am. Ceram.Soc.,
75,2286-2288.
Zoltowski P, (1998),On the electricalcapacitanceof interfacesexhibiting constantphase

elementbehaviour,J. Electroanal.Chem.,443,149-154.
Page 184
Appendix I
Representation Diagrams Of Impedance Spectroscopy:
For an equivalent circuit:
Rt RZ
-f-
CI C2
FigA1-1. An ideal equivalent circuit
R' Rz
Z'_ + (A1- 1)
l+(wR, C, )Z 1+(wR2C2)2
Z_ WR21C1 wR22C2 (A1- 2)

+
1+(wR, C, )z 1+(wwR2C2)2
M* jwCoZ*=M'+jM" (A1- 3)
E*= (M*)4 (Al- 4)
Y*=(Z*)"l (Al- 5)
Y*= jwCoE* (Al- 6)
If assuming R1=1x107 SZ,C, =1x10"11F, R2=5x1070, C2=1x10'10F, and based on above
equations, various IS representation diagrams for this circuit can be illustrated as in

Fig. A1-2a-i.
Page 185
abc
Y'
de", f
e Lo9io<o)
Lopw(ý,)
N`m
Loglohn)
Login(. )
L09m(o,)
Fig. A1-2(a) complex impedance plot; (b) complex admittance plot; (c) complex modulus
plots; (d) complex permittivity plot; (e) impedance spectroscopic plot; (1) admittance
spectroscopic plot; (g) modulus spectroscopic plot; (h)Bode plot (IZI vs log1o(c,
»)); (i) Bode
plot (0 vs logio(w)).
Page 186
Appendix II
The Impedance Spectra Of 10SA Samples

The Impedance Spectra Of IOSA SamplesseeFig. A2- 1.
Faý
b 5.0x10°T-
eMý
6.ox,o°
N
N 4omo°
z.mno"
o.o.
Co 2.010' 4.0x,0° 6Axlo° 80do° 1.0x10° 1210'
2 z
I-cl
o.oe
,oo*c
F-d]
004- ýE
h
0.02-
oA0.00
0.00 0.02 0,04 0.06 0.08 0110 0.12 to' td m' td td td td td
M Frequency(F¢)
LI1 3.50CF
a 0.035
so td
aoxtd 0.000
4.o tdý 25.1d i oars
-'- M p°r" & d "~ 0.030
tl4 zo(td
30.td 1 0
-o-r°sr`reat
1.5ttd N /r o 0.015
zoxtd
tO td I b 0010
toktd ý
so td iý y (1005
"
00 00 3035
ta' tv° ta' td to' .\ o' td ta' td td Io' 1Q" id 16'0d 10 100 ld 1M
Reese y0t) ' Frequency (Hz),
FigA2-1 (a) complex impedance plot at 100°C; (b) complex impedance plot at 100°C; (c)
complex modulus plots; (d) modulus spectroscopic plots; (e) impedance spectroscopic plots,
note the different scales of y-axis; (I) comparison of Z" peak and M" peak in modulus and
impedance spectroscopic plots.
Page187
Appendix III
The Impedance Spectra Of Clay/Graphite Composites
In order to investigate the dependence of IS on the orientation of the dispersed
conductive phase in an insulator matrix, clay/graphite composites containing 3 vol %, 6

vol% and 12 vol% flaky graphite were fabricated for impedance measurements.All the
clay/graphite composites were sintered in argon to protect the graphite inclusions from
oxidation. The sintering temperatures and dwell time were 1100°C-1250°C and 1-4h
respectively.
b
a Pressingdirection Normal Late al
us wo "" ". * am ý
Figure A3-1 a) A schematic of the microstructure of the clay/graphite composites in cross

section, with the flaky graphite particles oriented parallel to the top surface; b) definition of
cutting direction.
Optical microscopy observations showed that the flaky graphite particles were oriented
preferentially parallel to the top surface, i. e., perpendicular to the pressing direction as a
consequence of uni-axial pressing, as shown in Fig. A3-la. Samples were cut from the
sintered thick pellets along two directions: parallel and normal to the top surface, as
shown in Fig. 5-lb. All the sampleswere ground to the samedimension: 10mm (long) by
6mm (wide) by 2mm (thick).
The general IS features of clay/graphite composites were similar to those of A1203/SiC
nanocomposites: only one relaxation process can be identified in the IS; the conductivity
and dielectric constant increased with an increase in graphite particle loading. It was
found the composite containing 5vol% graphite was insulative, but the composite became
conductive when graphite content was 12 vol%, this indicates that the threshold
Page 188
conductive volume fraction should be between 5 vol% and 12 vol%. Unlike SiC particles
in A1203/SiC nanocomposites,the graphite particles are not equiaxed. The non-equiaxed
nature of conductive particles could result in a lower threshold volume fraction.

Therefore the threshold volume of clay/graphite composites should be lower than that of
o s.ono° o- Normaa
0.20
0.15
Labnl
Namur
4.0X10°
!E D.10 °
. ". _ _".
2.0x10° "" 0.05

ý" R
"ö
Sö 0.00
0.0 0.00 0.05 0.10 0.15 020 0.25 0.30 0.35
0.0 2.ad0° 4.0,10° 8.o,n0° 8.0x1 1.0x10° M
Z
5.WO' \
-"-Lateral erN
Normal
ö -o-Normal
I. Ox10° ý
0,12-
3.OX105
0.08-
2.0x10° ", ö
ý\ö 0.04 d/ "ý
1.0X101
°° e eo-E
,.. s=
0.00
0.o iv' io° id 10° 1o' IV io' lo"
lo 1o° 1o' 1o= io'
Frequency (Hz) Frequency (Fiz)
A1203/SiC nanocomposites.
Fig. A3-2 impedance spectra in different directions of a clay/graphite composite containing

3 wt% flaky graphite particles (a) impedance complex plane plot; (b) modulus complex
plane plot; (c) impedance spectroscopic plot and (d) modulus spectroscopic plot. Please note
that all of these quantities have been normalised in terms of the specimen's geometrical
dimensions.
The impedance spectra of clay/graphite composite with 3 vol % graphite particles in two
different directions are given in Fig. A3-2. We can seethat in the direction parallel to the
orientation of the particles, the resistance was lower and dielectric constant was higher.
However there was not much difference in relaxation frequency in different directions.
Page 189

FulltextThesis Interessant

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

FulltextThesis Interessant

Uploaded by

Copyright:

Available Formats

Non-Destructive Characterisation Of Structural Ceramics

Using Impedance Spectroscopy

A Thesis Submitted For The Degree Of Doctor Of Philosophy

Department Of Mechanical Engineering

spectroscopy(IS) in the non-destructive characterisation of structural ceramics.

properties of different microstructural origins in materials, e.g., grains and grain

Three categories of structural ceramics, i. e. A1203/SiC nanocomposites, thermal barrier

" For A1203/SiC nanocomposites, IS can be used to examine the conducting

microstructural defects in the top coating, respectively. The thickness of the

electrode/specimen interface, which could be an effective indicator of the

Therefore IS is a useful technique for non-destructive characterisation of structural

1) X. Wang and P. Xiao, "Nondestructive characterisation of alumina/silicon

carbide nanocomposites using impedance spectroscopy", J. Euro. Ceram. Soc.,

2) X. Wang, J. Mei and P. Xiao, "Determining oxide growth in thermal barrier

coatings(TBCs) non-destructively using impedance spectroscopy", J. Mater. Sci.

3) X. Wang, J. Mei and P. Xiao, "Non-destructive evaluation of thermal barrier

coatings using impedance spectroscopy", J. Euro. Ceram. Soc., 21 (2001) 855-

4) X. Wang, and P. Xiao, "Characterisation of the sintering process of clay-based

materials using impedance spectroscopy",J. Euro. Ceram. Soc., in press.

5) X. Wang, and P. Xiao, " In-situ monitoring of ceramic sintering using

7) S.-H. Song, X. Wang, and P. Xiao, "Effect of microstructural defects on the

electrical properties of Ti02, " Mater. Sci. Eng. B, accepted.

encouragement, guidance throughout my PhD studies, and the painstaking effort of

I am deeply indebted to Dr Houzheng Wu, at Oxford University, for his kindness in

providing A1203/SiC samples,stimulating discussions,and most importantly, friendship.

arguments. Thanks to them, my PhD has never become a `PermanentheaDache'

encouragementthroughout all these difficult times.

List Of Publications iii

1.2 The State-Of -The-Art Of NDC Techniques For Ceramic Materials 1

1.3 Impedance Spectroscopy As A NDC Technique 2

2.4 Electrical Conduction In Ceramics Materials 12

2.4.4 Ionic Conduction In Glass And Molten States 14

4 Impedance Data Analysis 55

5 Nondestructive Characterisation of Alumina/Silicon Carbide

5.2.3 Oxidation Treatments 68

5.1.2 Impedance SpectraOf A1203/SiC Nanocomposites 70

5.2.2 Spectral Resolution 86

5.3 Summary And Conclusions 96

6 Nondestructive Characterisation Of Thermal Barrier Coatings Using IS

6.2 Experimental Procedure 100

6.3.3 Impedance Spectra At Room Temperature 109

7 Characterisation Of Clay Sintering Process Using IS 132

8 Conclusions And Suggestions For Future work 162

1.1 The Need for Non-Destructive Characterisation (NDC) of Ceramics

The need to characterise ceramic materials by non-destructive analytical methods is

growing rapidly, especially for in-process and in-service examination of ceramic

and competitiveness, as well as the reliability of componentsin service.

spectroscopy (infrared spectroscopy, Raman spectroscopy etc.). Recently some highly

applied to the characterisation of ceramic microstructures. Among these analytical

1.2 The State-Of -The-Art Of NDC Techniques For Ceramic Materials

and components. By contrast, the modern non-destructive evaluation (NDE) of ceramic

1.3 Impedance Spectroscopy As A NDC Technique

electrochemical kinetics. This technique has demonstrated its capability in many

electrode/electrolyte interfaces, electrode/layer/electrolyte systems, to porous electrodes

the electrical properties and microstructure of ionic, electronic, or mixed conductor

ceramics, especially in zirconia-based materials and cementitious materials (Kleitz et al.

Besides being non-destructive, the technique of IS hasthe following advantages:

f IS canbe adaptedto on-line manufacturingor testingenvironments;

f IS does not require special sample preparation: theoretically the measurementscan be

performed on objects of any size and with any complex shape;

measurementswithout slowing production or testing progress;

f Unlike x-ray and 7-ray techniques,IS is non-hazardous.

1.4 About This Study

E= (M)"' (2- 7)

Y= jwCos (2- 9)