Haywang - Piezoelectricity Evolution and Future of A Technology
Haywang - Piezoelectricity Evolution and Future of A Technology
Haywang - Piezoelectricity Evolution and Future of A Technology
Piezoelectricity
Evolution and Future of a Technology
ABC
Prof. Dr. Walter Heywang Wolfram Wersing
Schwabener Weg 9a Wagnerfeldweg 10
85630 Grasbrunn 83346 Bergen
Germany Germany
Series Editors:
Professor Robert Hull Professor Jürgen Parisi
University of Virginia Universität Oldenburg, Fachbereich Physik
Dept. of Materials Science and Engineering Abt. Energie- und Halbleiterforschung
Thornton Hall Carl-von-Ossietzky-Strasse 9–11
Charlottesville, VA 22903-2442, USA 26129 Oldenburg, Germany
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
987654321
springer.com
Foreword
treatise of many exciting examples on the other. In all parts, the book has
been written by the internationally leading experts in the specific areas of
piezoelectricity and in material science. In addition to the aforementioned as-
pects this constitutes a high value on its own.
Jülich
August 2008 Rainer Waser
Foreword
Strasbourg
August 2008 Paul Siffert
Preface
The volume as presented has originated from discussions with the European
Materials Research Society. It is to fill a gap concerning the scientific and tech-
nical importance of piezoelectricity. We have to thank all who have supported
this project, especially the publisher and all our authors for their excellent
cooperation. We hope that we were able to meet the mentioned goal in spite
of the broad diversity of the whole field.
1 Introduction
L.E. Cross and W. Heywang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
5 Relaxor Ferroelectrics
L.E. Cross . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
8 Ultrasonic Imaging
W. Wersing and R. Lerch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
12 Piezoelectric Positioning
S. Arnold, P. Pertsch, and K. Spanner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
22 Thermodynamic Theory
G.A. Rossetti, Jr. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
Contents XIII
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571
List of Contributors
P. Muralt N. Setter
Ceramics Laboratory Ceramics Laboratory
EPFL – Swiss Federal EPFL – Swiss Federal
Institute of Technology Institute of Technology
(Ecole Polytechnique Fédérale (Ecole Polytechnique Fédérale
de Lausanne) de Lausanne)
Lausanne 1015 Lausanne 1015
Switzerland Switzerland
[email protected] [email protected]
P. Pertsch
K. Spanner
PI Ceramic GmbH
Physik Instrumente (PI) GmbH
Lindenstrasse
& Co. KG
D-07589 Lederhose
Auf der Roemerstrasse 1
Germany
D-76228 Karlsruhe
[email protected]
Germany
[email protected]
G.A. Rossetti, Jr.
Materials Science and
Engineering Program T. Steinkopff
Institute of Materials Science Siemens AG
University of Connecticut Corporate Technology
Storrs, CT 06269, USA Otto-Hahn-Ring 6
[email protected] D-81730 München
Germany
A. Safari [email protected]
The Glenn Howatt Electronic
Ceramics Research Laboratory
R. Theissmann
Department of Materials Science
University of Duisburg-Essen
& Engineering
Faculty of Engineering
Rutgers-The State University
Institute for Nano Structures
of New Jersey, Piscataway
and Technology (NST)
NJ 08854
Bismarckstr. 81
USA
D-47057 Duisburg
[email protected]
[email protected]
F. Seifert
University of Technology H. Thomann
Vienna Institute of Siemens AG
Sensor- and Actuatorsystems, Corporate Technology
retired Gusshausstrasse 27-29 München, Germany, retired
A-1040 Vienna Kössener Str. 15
Austria D-81373 München
[email protected] Germany
XVIII List of Contributors
K. Uchino S. Wagner
International Center for Actuators University of Karlsruhe
and Transducers Institute for Ceramics
Materials Research Institute in Mechanical Engineering
The Pennsylvania State University Haid-und-Neustrasse 7
University Park D-76131 Karlsruhe, Germany
PA 16802, USA [email protected]
Micromechatronics Inc.
State College, PA 16803 W. Wersing
USA Siemens AG
[email protected] Corporate Technology
München, Germany, retired
Wagnerfeldweg 10
R. Veith
D-83346 Bergen, Germany
EPCOS PTE LTD
[email protected]
166 Kallang Way
Singapore 349249
A. Wolff
[email protected]
Siemens AG
Corporate Technology
K. Wagner Otto-Hahn-Ring 6
EPCOS AG D-81730 München, Germany
R&D/Surface Acoustic Wave [email protected]
Components
Anzinger Strasse 13 Y. Yamashita
D-81671 Munich Toshiba Research Consulting Corp.
Germany Kawasaki 212-8582, Japan
[email protected] [email protected]
1
Introduction
time stability, reproducibility, reliability, cost and other market needs. Out of
the family of such requirement for a given device, and the measured material
properties, a figure of merit can be formulated. Now we have to compare these
figures of merit with the corresponding ones for already existing systems in
current applications.
From general experience, it is known that a new novel technology will not
succeed in application unless it offers at least a factor of three advantages.
Otherwise, in the inertia born tendency for existing solutions to survive, the
current approach will be improved by a factor of this order as soon as it is
perceived to be endangered.
To effectively enter the market piezoelectrics always had to compete with
well-known electromagnetic devices, as for example relays and electromagnetic
motors. Only in the case of substantial advantage would they be accepted
technically or commercially. Clearly this competition was the major reason
for the above-mentioned long delay of the technical break-through to large
scale applications.
First time a conceived major improvement in a practical device was
achieved through piezoelectricity in 1921 with the development of the quartz
crystal stabilized electrical oscillator. In this device, the weak piezoelectric-
ity in quartz is used to excite resonant elastic vibrations where frequency is
set by the ultra stable elastic properties of suitable crystal cuts. In 1924, the
first crystal stabilized radio transmitter was placed into service, and by late
1930s all high frequency radio transmitters were under crystal control. The
astounding success of this initially simple concept is attested by the fact that
even now more than 80 years later, the quartz crystal controlled oscillator is
still the secondary standard for timing and frequency control.
However, the avalanche of piezo technology starting in the market place
was not possible until it was realized that the mixed oxide compound barium
titanate (BaTiO3 ) was a ferroelectric, and it was demonstrated in 1946 that
BaTiO3 ceramic, which can be easily fabricated and shaped at low price,
can be made piezoelectric by an electrical poling process. First commercial
piezoelectric devices based on BaTiO3 ceramics were phonograph pickups and
appeared on the market about 1947.
The piezoelectric effect in ferroelectric titanates is much stronger than
in quartz because they exhibit different ferroelectric phases with only small
energy differences. This can lead to the unusual combination of high dielec-
tric permittivity in a strongly polar lattice in the broad vicinity of the phase
boundaries. These conditions are essential for high piezoelectric constants and
strong elasto-electric coupling. An advance of great practical importance was
the discovery in 1954 of very strong piezoelectric effects in lead zirconate ti-
tanate solid solutions. In this material system, a morphotropic phase bound-
ary exists between a tetragonal and rhombohedral ferroelectric phase, which
is nearly independent of temperature. Thus, the anomalously high piezoelec-
tric effect near to this boundary is almost constant over a wide temperature
1 Introduction 3
range, important for technical applications. However, one has to pay for the
anomalously large piezoelectric effect at this phase boundary with the risk of
higher instability of material properties. It was a long and arduous task to
control these instabilities by proper adjustment of dopants and grain struc-
ture. This problem was eased by the development of two basic material classes
adapted to quite different branches of applications: so called soft piezoelectric
with as high as possible piezoelectric properties for actuator and ultrasonic
applications and so-called hard piezoceramics with as high as possible stability
for high power and precise frequency filter applications.
In addition to the material property, considerations on optimum design of
device must consider the electrodes through which the field is to be applied.
Electrodes must be robust and in very intimate contact with the high permit-
tivity ceramic. Barely any other functional material exists with such a plethora
of application-specific shaping requirements: plates, rods, disks, foils, tubes,
all designed for either static, dynamic, or resonant applications. There are
also composites made from combinations of piezoceramics and plastics, now
very widely used in medical ultrasound and in hydrophone systems. Further
more, to achieve enhanced elongation in actuators at smaller driving voltages,
an inexpensive multilayer technology has been developed. In this way, it has
become possible to achieve elongations on 0.1 mm in compact monolithic de-
vices, a technology now used for modern fuel valves. These valves react ten
times faster than conventional valves and are just revolutionizing motor man-
agement in the automotive industry. Thin film technologies are of developing
interest for micro electronics and micro electro mechanical (MEMs), for high
permittivity dielectric layers, in nonvolatile computer memories, pyroelectric
thermal images, and biosensors.
It is very important to recall that beside polycrystalline piezoelectrics,
ferroelectric piezo-crystals have also been developed. Today we have the sta-
ble ferroelectric phase single crystal lithium niobate (LiNbO3 ) and lithium
tantalate (LiTaO3 ). Both are commercially available and widely used for sur-
face acoustic wave devices, which play a major role in signal detection and
processing in modern telecommunication systems.
Finally it should be mentioned that there are plastics that can be made
ferro- respectively piezoelectric by stress and/or by electrical polarization be-
cause of an alignment of asymmetric molecules. They find use as inexpensive
foils wherever cheapness or plasticity are relevant., e.g. in cheap microphones
or shock gauges. Recent studies following analogous paths to electroceramics
have demonstrated effective control of microstructure and properties, stepping
stones in bridging to the growing field of biological technologies.
An over-all compilation of existing application of piezoelectricity is given
in Table 1.1. The heavy-typed fields indicate where the number of installed
piezoelectric samples have exceeded already several billions. Other fields are on
the way to follow like the field of automotive applications, where piezoelectric
sensors, valves, and injectors are key elements for clean and fuel saving motor
4 L.E. Cross and W. Heywang
A. Ballato
2.1 General
2.1.1 Commercial and Technological Usages of Quartz
Although material quartz is of scientific interest in its own right, its vol-
ume of usage and variety of applications dictate its technological importance.
The technological prominence of α-quartz stems largely from the presence
of piezoelectricity, combined with extremely low acoustic loss. It was one of
the minerals with which the Brothers Curie first established the piezoelectric
effect in 1880. In the early 1920s, the quartz resonator was first used for fre-
quency stabilization. Temperature-compensated orientations (the AT and BT
shear cuts) were introduced in the 1930s, and assured the technology’s suc-
cess. By the late 1950s, growth of cultured bars became commercially viable,
and in the early 1970s, cultured quartz use for electronic applications first ex-
ceeded that of the natural variety. The discovery of cuts that addressed com-
pensation of stress and temperature transient effects occurred in the 1970s,
and led to the introduction of compound cuts such as the SC, which has
both a zero temperature coefficient of frequency, and is simultaneously stress-
compensated [1–5]. Between 109 and 1010 quartz units per year were produced
by 2000 at frequencies from below 1 kHz to above 10 GHz. Categories of ap-
plication include resonators, filters, delay lines, transducers, sensors, signal
processors, and actuators. Particularly noteworthy are the bulk- and surface-
wave resonators; their uses span the gamut from disposable timepieces to
highest precision oscillators for position-location, and picosecond timing ap-
plications. Stringent high-shock and high-pressure sensor operations are also
enabled. Table 2.1 shows the major applications of quartz crystals. These ap-
plications are discussed subsequently in greater detail. For general background
and historical developments, see [1, 6–11].
Crystal quartz for timekeeping currently has a global market of over 109
USD per year, and that for cellular communications exceeds 50 × 109 USD
10
A. Ballato
per year. Useful rules-of-thumb for the temporal evolution of this technology
are the following:
• Upper frequency limit: f0 = 10+r , where f0 is nominal frequency, r = 4 +
0.075 (Y – 1920), and Y is year.
• Frequency accuracy (total fractional absolute frequency variations over all
environmental ranges such as temperature, mechanical shock, and aging):
∆f /f0 = 10−a , where ∆f = (f − f0 ) is frequency shift, and a = 3.5 + 0.05
(Y – 1940).
• Aging: <10−8 /day for production units, and <10−11 /day for high preci-
sion units.
• Frequency stability: ∆f /f0 = 10−s , where s = 6 + 0.1(Y − X), with
X = 1920 for laboratory versions, X = 1940 for commercial versions, and
X = 1960 for large-scale production models. Observation times are in the
range 0.1–10 s. As with Moore’s Law, a saturation of these exponential
rates is predicted, but not yet observed.
Polymorphs of Silica
10
Stishovite
Tetragonal (4/m mm)
Coesite
Monoclinic (2/m)
6
Pressure (GPa)
β– quartz
hexagonal (622)
2 α– quartz
trigonal (32) 1470 1710
573
870 lite
e Cristoba Liquid
Tridymit
0
0 400 800 1200 1600 2000
Temperature ( C)
move only 0.03 nm, changing the trigonal axis to hexagonal, and the point
group symmetry to 622/D6 . Beta quartz is also enantiomorphic, composed of
space groups P62 22/D6 4 and P64 22/D6 5 ; handedness is preserved in the tran-
sition. Enantiomorphism implies the presence of piezoelectricity. Beta quartz
is not stable below 573◦ C. At 600◦ C, unit cell dimensions are a0 = 0.501
and c0 = 0.547 nm, and density ρ = 2.533. α-tridymite (point group un-
certain; mmm or 2/m) is metastable below 870◦ C, to ∼117◦ C; ρ = 2.28.
β-tridymite (6/m mm) is stable between 870 and 1, 470◦ C, and metastable
down to ∼163◦ C; ρ = 2.24. α-cristobalite (422) is metastable to below
200◦ C; ρ = 2.32. β-cristobalite (m3m) is stable between 1,470 and 1, 710◦ C,
and metastable down to 268◦ C; ρ = 2.24. Density units are Mg/m . Coesite
3
is of point symmetry 2/m; that of stishovite is 4/m mm; both are metastable
under ordinary conditions. With the exception of α-cristobalite, the other
phases are not piezoelectric. The symmetry diagram for β-quartz is given in
Fig. 2.4.
Transitions between different phases of a single polymorph (α-β transfor-
mations) involve only bond warping, and not bond breaking; these displacive
transformations proceed rapidly. Cooling through the α-β quartz transition in-
variably results in α-quartz regions having oppositely-directed x1 axes: “elec-
trical twinning.” Transitions between polymorphs require breaking of bonds;
these reconstructive transformations are slow, and subject to kinematic con-
straints.
Hydrothermal growth of cultured quartz takes place in aqueous NaOH or
Na2 CO3 solutions at temperatures of about 350◦ C and pressures between 80
and 200 MPa, driven by a temperature gradient from 4 to 10 K. The process
14 A. Ballato
takes from one to nine months. Processing of quartz crystals includes the steps
of cutting, X-ray orienting, lapping, polishing and etching (usually NH4 F ·
HF or HF(aq) ). Growth and etch rates are highly anisotropic, being most
rapid along z, with the z rate between 100 and 1,000 times that along x or
y axis.
Table 2.2. Van Dyke matrix for the mixed coefficients cE , e, and εS of α-quartz
Table 2.3. Van Dyke matrix for the homogeneous coefficients sE , d, and εT of
α-quartz
Mixed sets:
(3) T = cE S − e E and D = eS + εS E
(4) S = sD T + g D and E = −gT + β T D
Constants d and g are “strain” coefficients; e and h are “stress” coefficients.
The [sE , d , d, εT ] and [cD , −h , −h, β S ] matrices are inverses, as are the
[cE , −e , e, εS ] and [sD , g , −g, β T ] matrices.
Point-group symmetry dictates the number of components appearing in
the constitutive equations coupling the field tensors, and any relations among
the components. Tables 2.2–2.5 display the Van Dyke matrices (or “term
schemes”) representing point group symmetry 32 and 622. Factors of 2 in
the s, d, and g matrices arise from a convention regarding the definition of
strain, and, in these tables, c12 = c11 − 2c66 and s12 = s11 − s66 /2. The Van
Dyke matrices of Tables 2.2 and 2.3, and of Tables 2.4 and 2.5 are inverses of
each other.
Table 2.4. Van Dyke matrix for the mixed coefficients cE , e, and εS of β-quartz
Table 2.5. Van Dyke matrix for the homogeneous coefficients sE , d, and εT of
β-quartz
unit: (TPa)−1
a
Slightly different values are given in [36] and [37]
Table 2.12 contains values for the piezoelectric coefficients. Table 2.13 lists val-
ues for the dielectric permittivities and the impermeabilities, while Table 2.14
gives values for the thermoelastic (thermal expansion) coefficients.
unit: 10−6 /K
a
Averages gleaned from the literature
b
Cultured quartz at 23◦ C
c
−2, 637 for consistency
unit: 10−6 /K
a
−1, 290 for consistency
units: α(1) in 10−6 /K; α(2) in 10−9 /K2 ; α(3) in 10−12 /K3
a
−50◦ C to +150◦ C
λµ → 11 13 14 33 44 66 (12)
η 1.37 0.72 0.01 0.97 0.36 0.32 0.73
The comprehensive and readable works of Brice [24] and James [32] contain
careful and insightful reviews of measurement methods, and of various data
sets appearing in the literature. For the linear elastic, piezoelectric, and dielec-
tric constants of quartz, the set obtained by Bechmann [30] using resonator
methods is a good choice. The averaged values of Brice [24], and the set derived
by James [32] from time-domain method measurements are good alternatives.
For the temperature coefficients, the set given in [39], obtained from resonator
measurements is a good choice, with the averaged values of Brice [24], and the
set derived by James [32] from time-domain method measurements as good
alternatives.
2.2.2.7 Beta-Quartz
cE
λµ [50] sE
λµ [50]
c11 116 s11 9.41
c13 33 s13 −2.6
c33 110 s33 10.6
c44 36 s44 27.7
c66 50 s66 20.02
(c12 ) 16 s12 −0.60
and insertion loss; in resonators and resonant MEMS devices, they establish
pole-zero spacing and adjustment range; in actuators and sensors, they control
electromechanical conversion efficiency. In applications, they usually appear
as k 2 . For a given type of resonant motion, effective values of elastic, piezo-
electric, and dielectric coefficients arise. There are four basic forms expressing
coupling, depending upon the constitutive equation set chosen. The generic
homogeneous forms are the following:
√ √
• k = |d|/√ (εT sE ) and |h|/√ (β S cD ). The generic mixed forms are:
• k = |e|/ (εS cE ) and |g|/ (β T sD ) [28].
Explicit formulas for some α-quartz resonator plates are the following:
√
• X-cut (plate normal to x1 axis): k11 = |e11 |/ [ε11 S c11 E + (e11 )2 ] = 9.23%;
pure extensional mode, driven by electric field√ in plate thickness direction.
• Y-cut (plate normal to x2 axis): k66 = |e11 |/ [ε11 S c66 E +(e11 )2 ] = 13.55%;
pure shear mode, driven by electric field in plate
√ thickness direction.
• Z-cut (plate normal to x3 axis): k44 = |e14 |/ [ε11 S c66 E ] = 2.69%; pure
shear mode, motion along direction of driving electric field in x2 −x3 plane.
• Rotated-Y-cuts (plate normal in x2 − x3 plane), pure shear mode: k26 =
√
|e26 |/ [ε22 c ]; primes denote rotated values. AT cut: k26 = 8.80% and
Quartz
Log(Q) 8
Refractories
Water-
4
Solubles
Ceramics
2
Piezo Semiconductors
& Piezopolymers
0
0 20 40 60 80 100
K(%)
Fig. 2.5. Acoustic quality factor (Q) vs. piezoelectric coupling factor (k) for various
types of piezoelectric materials
delay lines, and filters) for the mobile telecommunications, multimedia, and
industrial-scientific-medical (ISM) bands; wireless passive identification tags,
sensors and transponders, and microbalances. BAW applications include res-
onators in precision clock oscillators; front-end GPS filters for cell phones;
thin-film, solidly mounted resonators (SMRs), and stacked crystal filters
(SCFs) formed as SMRs, for integration with microwave heterojunction bipo-
lar transistor (HBT) VCOs [5, 8–11].
Why quartz? Quartz is the only material known that possesses the following
combination of desirable attributes:
• Piezoelectric
• Zero temperature coefficient cuts
• Stress/thermal transient compensated cuts
• Low acoustic loss/high acoustic Q
• Capable of withstanding 35,000–50,000 g shocks
• Very stable
• Capable of integration with micro- and nano-electronic components
• Easy to process; hard but not brittle; under normal conditions, low solu-
bility in everything except fluoride etchants
• Abundant in nature; easy to grow in large quantities, at low cost, and with
relatively high purity and perfection. Of the man-made single crystals,
quartz, at over 3,000 tonnes per year (2000), is second only to silicon in
quantity grown [7].
• 1880
– Piezoelectric effect discovered: Brothers Pierre & Jacques Curie
• 1900s
– First hydrothermal growth of quartz in laboratory
• 1910s
– First application of piezoelectricity, in sonar (Rochelle salt)
– First quartz crystal oscillator, filter
• 1920s
– Introduction of quartz resonator; requires “telescope making” technol-
ogy: rough X-raying; grinding, lapping, polishing
– Quartz resonator used for frequency stabilization and control
– First crystal control of radio station
– BVD equivalent circuit representation
– Electrical measurements of resonators using oscillators
– Compound resonator for measuring constants of nonpiezoelectrics
– First quartz crystal-controlled clock
28 A. Ballato
• 1930s
– Temperature compensated bulk acoustic wave (BAW) AT/BT cuts de-
veloped; requires X-raying to minutes of arc
– Modal surface motions investigated using lycopodium powder patterns
– Equivalent circuit incorporating mechanical ports for transducers
• 1940s
– Hydrothermal growth of quartz [74]; (Czochralski growth of Si and Ge)
– Electrical measurements using passive transmission networks
– Electrode system: A
evaporation directly on resonator surfaces
– Surface finish found to affect quality of resonator
– First contoured, high-Q, resonator designs
– Discussion of coupled plate modes [49]
• 1950s
– Hydrothermal growth of cultured bars commercially viable
– Solder-sealed metal enclosures
– Processing ambient: soft vacuum
– Quartz crystal microbalance (QCM) developed
– Equivalent circuit in form of acoustic transmission line
– Temperature-compensated crystal oscillator (TCXO) described
– Theory of vibrations of anisotropic plates [110]
• 1960s
– “Energy trapping” theory used to reduce unwanted modes [75–80]
– Monolithic crystal filter (MCF) developed
– Solidly mounted resonator (SMR) described
– Glass enclosures introduced for low aging
– Electrolytic sweeping reduces effects of ionizing radiation and neutron
flux
– Application of microelectronics technology enabling batch-processing
– High coupling refractories, such as lithium niobate and tantalate grown
– Anisotropic plate theory extended [111–113]
• 1970s
– Cultured quartz use exceeded natural quartz; (Czochralski growth of
III-V binaries)
– Electrical measurements of crystal parameters using bridges
– Stacked-crystal filters (SCFs) described
– UV-ozone cleaning to produce atomically clean surfaces and very low
aging [81]
– Chemical polishing reduces subsurface damage; greatly increased shock
resistance [82]
– Temperature and nonlinear effects compensation; requires X-raying to
seconds of arc
– Interdigital transducer (IDT) enables efficient transduction of surface
waves [83–87]
– Surface acoustic wave (SAW) cuts; temperature compensated ST cut
– Stress-compensated compound cut developed (SC) [2–4]
2 Basic Material Quartz and Related Innovations 29
The search for new materials is fed by practical demands for devices with
improved characteristics, e.g., lower loss (higher resonator quality factor (Q),
lower filter insertion loss (IL)), higher piezocoupling (k 2 ) for increased fil-
ter bandwidth, better temperature stability, greater miniaturization (higher
acoustic velocity), etc. Materials receiving recent attention include the quartz
analog langasite and its isomorphs, generically called LGX.
The langasite (LGS, La3 Ga5 SiO14 ) family consists of Czochralski grown,
congruently melting, class 32 materials such as langanite (LGN, La3 Ga11/2
Nb1/2 O14 ) and langatate (LGT, La3 Ga11/2 Ta1/2 O14 ) [70, 71]. Table 2.20 pro-
vides pertinent phenomenological data. These materials have acoustic Qs
higher than quartz, but have disordered structures, due to the difficulty of
satisfying both ionic size and charge compensation constraints. In the fu-
ture, totally ordered (Ca, Sr)3 (Nb, Ta)Ga3 Si2 O14 crystals, and similar mem-
bers, are expected to offer higher elastic stiffnesses, as well as lower dielectric
permittivity and higher piezocoupling. Candidate materials include CNGS
(Ca3 NbGa3 Si2 O14 ), CTGS (Ca3 TaGa3 Si2 O14 ), SNGS (Sr3 NbGa3 Si2 O14 ), and
STGS (Sr3 TaGa3 Si2 O14 ). Main market possibilities are greater bandwidth
(versus quartz), high stability IF filters; high temperature sensors (no twin-
ning); and high Q BAW and SAW resonators [72].
References
1. W.G. Cady, Piezoelectricity (McGraw-Hill, New York, 1946)
2. E.P. EerNisse, in Proc. 29th Ann. Freq. Control Symp., pp. 1–4, US Army
Electronics Command, Ft. Monmouth, NJ, May 1975
3. J.A. Kusters, J. Leach, in Proc. IEEE. vol. 65, no. 2, pp. 282–284, Febru-
ary (1977)
4. A. Ballato, in Physical Acoustics: Principles and Methods, vol. 13, ed. by W.P.
Mason, R.N. Thurston. (Academic Press, New York, 1977), Chap. 5, pp. 115–
181, ISBN: 0-12-477913-1
5. S. R. Stein, J. R. Vig, in The Froelich/Kent Encyclopedia of Telecommunica-
tions, Vol. 3, ed. by F.E. Froehlich, A. Kent (Marcel Dekker, New York, 1992)
pp. 445–500
6. R.B. Sosman, The Properties of Silica (Chemical Catalog Co., New York, 1927)
7. “Quartz crystal,” Bulletin 667, Mineral Facts and Problems, US Department
of the Interior, Bureau of Mines, 1975
32 A. Ballato
32. B.J. James, “Determination of the Elastic and Dielectric Properties of Quartz,”
PhD Dissertation, Royal Holloway and Bedford New College, University of
London, Spring 1987, 231 pp. See also B. J. James, “A new measurement
of the basic elastic and dielectric constants of quartz,” IEEE Intl. Frequency
Control Symp. Proc. (42nd Ann.), pp. 146–154, Baltimore, MD, June 1988
33. J. Kushibiki, I. Takanaga, S. Nishiyama, IEEE Trans. Ultrason., Ferroelect.,
Freq. Contr. 49(1), 125–135 (January 2002)
34. J.V. Atanasoff, P. J. Hart, Phys. Rev. 59(1), 85–96 (1941)
35. W.P. Mason, Phys. Rev. 55, 775–789 (April 1939)
36. R. Bechmann, Proc. Phys. Soc. (London), B64, 323–337 (April 1951)
37. R. Bechmann, Archiv der Elektrischen Übertragung 5, 89–90 (1951). (present
name: Archiv für Elektronik und Übertragungstechnik)
38. I. Koga, M. Aruga, Y. Yoshinaka, Phys. Rev. 109, 1467–1473 (March 1958)
39. R. Bechmann, A. Ballato, T. J. Lukaszek, Proc. IRE 50(8), 1812–1822 (August
1962); Proc. IRE 50(12), 2451 (December 1962)
40. P.C.Y. Lee, Y.-K. Yong, J. Appl. Phys. 60(7), 2327–2342 (1986)
41. J. Zelenka, P.C.Y. Lee, IEEE Trans. Sonics Ultrason. SU-18(2), 79–80 (1971)
42. R.K. Cook, P.G. Weissler, Phys. Rev. 80(4), 712–716 (15 November 1950)
43. A. Ballato, M. Mizan, IEEE Trans. Sonics Ultrason. SU-31(1), 11–17
(January 1984)
44. J.A. Kosinski, J.G. Gualtieri, A. Ballato, IEEE Trans. Ultrason., Ferroelec.,
Freq. Control 39(4), 502–507 (July 1992)
45. R. Bechmann, in Landolt-Börnstein, Numerical Data and Functional Relation-
ships in Science and Technology, New Series, Group III: Crystal and Solid
State Physics, ed. by K.–H. Hellwege, A.M. Hellwege (Springer Verlag, Berlin,
New York, vol. III/1, pp. 40–123, 1966; and vol. III/2, pp. 40–101, 1969)
46. J. Lamb, J. Richter, Proc. Roy. Soc. (London) A293, 479–492 (1966)
47. W.R. Cook, Jr., H. Jaffe, in Landolt-Börnstein, Numerical Data and Functional
Relationships in Science and Technology, New Series, Group III: Crystal and
Solid State Physics, vol. III/11, ed. by K.–H. Hellwege, A.M. Hellwege (Springer
Verlag, Berlin, New York, 1979) pp. 287–470
48. B. Parzen, Design of Crystal and Other Harmonic Oscillators (Wiley, New
York, 1983). ISBN: 0-471-08819-6
49. R.A. Sykes, in Quartz Crystals for Electrical Circuits: Their Design and Man-
ufacture, ed. by R.A. Heising (D. Van Nostrand, New York, 1946) Chap. 6,
pp. 205–248
50. E.W. Kammer, T.E. Pardue, H.F. Frissel, J. Appl. Phys. 19(3), 265–270
(March 1948)
51. B.K. Sinha, Ferroelectrics 41(1), 61–73 (1982)
52. B.K. Sinha, Proc. 35th Annu. Freq. Control Symp. 213–221 (May 1981)
53. M. Valdois, B.K. Sinha, J.-J. Boy, IEEE Trans. Ultrason., Ferroelec., Frequency
Contr. 36(6), 643–651 (November 1989)
54. B.K. Sinha, IEEE Ultrason. Symp. Proc. 557–563 (December 1990)
55. B.K. Sinha, IEEE Trans. Sonics Ultrason. SU-32(4), 583–591 (July 1985)
56. B.K. Sinha, IEEE Trans. Ultrason., Ferroelec., Frequency Contr. 34(1), 64–74
(January 1987)
57. S. Locke, B. K. Sinha, IEEE Trans. Ultrason., Ferroelec., Frequency Contr.
UFFC-34(4), 478–484 (July 1987)
58. R.N. Thurston, H.J. McSkimin, P. Andreatch, Jr., J. Appl. Phys. 37(1), 267–
275 (January 1966)
34 A. Ballato
59. R. Stern, R.T. Smith, J. Acoust. Soc. Am. 44, 640–641 (1968)
60. R.F.S. Hearmon, in Landolt-Börnstein, Numerical Data and Functional Rela-
tionships in Science and Technology, New Series, Group III: Crystal and Solid
State Physics, vol. III/11, ed. by K.–H. Hellwege, A.M. Hellwege (Springer
Verlag, Berlin, New York, 1979) pp. 245–286
61. W.P. Mason, J. Acoust. Soc. Am. 70(6), 1561–1566 (December 1981)
62. P. Curie et, J. Curie, Bull. Soc. Fr. Mineral. Cristallogr. 3, 90–93 (1880);
C. R. Acad. Sci. (Paris) 91, 294, 383 (March 1880)
63. M. Trainer, Eur. J. Phys. 24(5), 535–542 (September 2005)
64. R. Bechmann, et al., in Piezoelectricity (Her Majesty’s Stationery Office,
London, 1957), 369 pp
65. D. Berlincourt, J. Acoust. Soc. Am. 70(6), 1586–1595 (December 1981)
66. T. Yamada, N. Niizeki, H. Toyoda, Japan. J. Appl. Phys. 6(2), 151–155 (1967)
67. A.W. Warner, M. Onoe, G.A. Coquin, J. Acoust. Soc. Am. 42(6), 1223–1231
(December 1967)
68. R.T. Smith, F. S. Welsh, J. Appl. Phys. 42(6), 2219–2230 (May 1971)
69. G.M. Sessler, J. Acoust. Soc. Am. 70(6), 1596–1608 (December 1981)
70. R.C. Smythe, IEEE Intl. Frequency Control Symp. Proc. 761–765 (May 1998)
71. D.C. Malocha, M.P. da Cunha, E. Adler, R.C. Smythe, S. Frederick, M. Chou,
R. Helmbold, Y.S. Zhou, 2000 IEEE/EIA Intl. Freq. Control Symp. Proc.,
pp. 201–205, Kansas City, MO, June 2000
72. B.H.T. Chai, A.N.P. Bustamante, M. C. Chou, “A new class of ordered lang-
asite structure compounds,” 2000 IEEE/EIA Intl. Freq. Control Symp. Proc.,
pp. 163–168, Kansas City, MO, June 2000
73. P.W. Krempl, J. Phys. IV France 126, 95–100 (June 2005)
74. F. Iwasaki, H. Iwasaki, J. Crystal Growth, 237–239, 820–827 (2002)
75. W. Shockley, D.R. Curran, D.J. Koneval, Proc. 17th Ann. Frequency Control
Symp. 88–126 (May 1963)
76. D.R. Curran, D.J. Koneval, Proc. 18th Ann. Freq. Control Symp. 93–119
(May 1964)
77. W.S. Mortley, Wireless World 57, 399–403 (October 1951)
78. W.S. Mortley, Proc. IEE (London) 104B, 239–249 (December 1956)
79. R.D. Mindlin, J. Acoust. Soc. Am. 43(6), 1329–1331 (June 1968)
80. H.F. Tiersten, R.C. Smythe, J. Acoust. Soc. Am. 65(6), 1455–1460 (June 1979)
81. J.R. Vig, J.W. Le Bus, IEEE Trans. Parts, Hybrids, Packaging PHP-12(4),
365–370 (December 1976)
82. J.R. Vig, J.W. LeBus, R.L. Filler, Proc. 31st Ann. Freq. Control Symp. 131–
143 (June 1977)
83. R.M. White, F.W. Voltmer, Appl. Phys. Lett. 7(12), 314–316 (1965)
84. M.B. Schulz, B.J. Matsinger, M.G. Holland, J. Appl. Phys. 41(7), 2755–2765
(1970)
85. R.M. White, Proc. IEEE 58(8), 1238–1276 (August 1970)
86. M.G. Holland, L.T. Claiborne, Proc. IEEE 62(5), 582–611 (May 1974)
87. B.K. Sinha, H.F. Tiersten, Appl. Phys. Lett. 34(12), 817–819 (15 June 1979)
88. N.F. Foster, J. Acoust. Soc. Am. 70(6), 1609–1614 (December 1981)
89. W.E. Newell, Proc. IEEE 52(12), 1603–1607 (December 1964)
90. W.E. Newell, Proc. IEEE 53(6), 575–581 (June 1965)
91. W.E. Newell, Proc. IEEE 53(10), 1305–1308 (October 1965)
92. A. Ballato, T. Lukaszek, Proc. IEEE 51(10), 1495–1496 (October 1973)
2 Basic Material Quartz and Related Innovations 35
93. A. Ballato, H.L. Bertoni, T. Tamir, IEEE Trans. Microwave Theory Tech.
MTT-22(1), 14–25 (January 1974)
94. K.M. Lakin, G.R. Kline, R.S. Ketcham, J.T. Martin, K.T. McCarron, IEEE
Intl. Freq. Control Symp. Proc. 536–543 (May-June 1989)
95. K.M. Lakin, IEEE Intl. Freq. Control Symp. Proc. 201–206 (May 1991)
96. K.M. Lakin, G.R. Kline, K.T. McCarron, IEEE MTT-S Intl. Microwave Symp.
Digest 3, 1517–1520 (June 1993)
97. K.M. Lakin, IEEE Microwave Magazine 4(4), 61–67 (December 2003)
98. A. Ballato, J.G. Gualtieri, IEEE Trans. Ultrason., Ferroelect., Freq. Cont.
41(6), 834–844 (November 1994)
99. H. Iwata, IEICE Electronics Express 1(12), 346–351 (September 2004)
100. K.L. Ekinci, M.L. Roukes, Rev. Sci. Instrum. 76 art. 061101, 12pp. (2005)
101. MEMS: A Practical Guide to Design, Analysis, and Applications, ed. by J.G.
Korvink, O. Paul (Springer-Verlag, Heidelberg 2006), ISBN: 3-540-21117-9
102. N. Takahashi, T. Nakumura, S. Nonaka, H. Yagi, Y. Sinriki, K. Tamanuki,
Electrochemical and Solid State Lett. 6(5), C77–C78 (2003)
103. N. Takahashi, T. Nakumura, Electrochemical and Solid State Lett. 6(11), H25–
H26 (2003)
104. A. Ballato, IEEE Trans. Ultrason., Ferroelect., Freq. Contr., 48(5), 1189–1240
(September 2001)
105. R. A. Heising (ed.), Quartz Crystals for Electrical Circuits: Their Design and
Manufacture (D. Van Nostrand, New York, 1946)
106. W.P. Mason, Piezoelectric Crystals and their Application to Ultrasonics (Van
Nostrand, New York, 1950)
107. A. Ballato, Proc. IEEE 58(1), 149–151 (January 1970)
108. B.A. Auld, Acoustic Fields and Waves in Solids, Vol. I and II, (Robert E.
Krieger Publishing, Malabar, FL, 1990). ISBNs: 0-89874-783-X; 0-89874-782-1
109. W.P. Mason, Proc. IEEE 57(10), 1723–1734 (October 1969)
110. R.D. Mindlin, J. Appl. Phys., 23(1), 83–88 (January 1952)
111. R.D. Mindlin, D.C. Gazis, Proc. Fourth U. S. Natl. Congr. Appl. Math. 305–
310 (1962)
112. R.D. Mindlin, P.C.Y. Lee, Intl. J. Solids and Struct. 2(1), 125–139 (Janu-
ary 1966)
113. R.D. Mindlin, W.J. Spencer, J. Acoust. Soc. Am. 42(6), 1268–1277 (Decem-
ber 1967)
114. B.K. Sinha, IEEE Trans. Ultrason., Ferroelec., Frequency Contr. 48(5), 1162–
1180 (September 2001)
3
The Role of Ferroelectricity for Piezoelectric
Materials
1. Because of the high piezoelectric effect that can be found in these mate-
rials, a high and efficient electromechanical transformation of energy and
signals can be achieved.
2. In general, the remnant electrical polarization that occurs in these ma-
terials can be oriented into a desired direction by applying an external
electrical field: this means by poling, if needed, at elevated tempera-
ture. Therefore it becomes possible to imprint a unipolar direction of
macroscopic preference or anisotropy to a device even after processing it.
Obviously, this unipolarity is mandatory for every piezoelectric action.
3. Therefore, materials that are macroscopically isotropic by nature after the
production process can be also used. This concerns, e.g., ceramics, plastics,
or composites, which can be processed using well-known methods to adapt
and shape the material according to the requirements of the final use.
4. For piezoelectric applications, the group of ceramics based on the so-called
perovskite structure is especially important. These materials are already
being used over a broad field of technological applications such as in capac-
itors and PTC resistors1 . Therefore, experience in the mass production,
reliability, etc. of these materials is already available.
5. Therefore, also from the point of view of production cost, ferroelectric
materials offer good advantages.
Consequently, it makes sense to offer here a short introduction into this class
of materials.
1
PTC: Giant Positive Temperature Coefficient of resistivity within the ferroelectric
Curie-Weiss region.
38 W. Wersing et al.
stresses have to be minimized. The properties of such samples are called in-
trinsic, as long as the domain configuration is not changed by the measuring
signal. A conceptual overview will try to explain the interrelation between
these intrinsic properties on one hand, and their dependence on composition
as well as on preparation and poling procedure on the other.
Section 3.3 finally deals with the same properties in the case that a reori-
entation of the domains becomes possible under the influence of the measuring
signals themselves. We call these properties extrinsic.
It should be emphasized again that the aim of this chapter is only to
provide a conceptual understanding of possibilities and limitations of piezo-
electric material engineering. It should be sufficient for the understanding of
the different applications. A more profound description of the state of the art
of materials and device design is given in Chap. 12.
2. Ferroelectric materials where the lattice cells are not polarized except un-
der the influence of the internal field below the Curie temperature. In this
case, the spontaneous polarization is based on the relative displacement
of sublattices formed by ions and their electron clouds (displacive type).
But we should state already here that in every ferroelectric material both
these mechanisms coexist, even though to different extents. On one hand,
dipoles that are preformed in the paraelectric state grow in size by the influ-
ence of the internal field below the Curie temperature; on the other, purely
ionic bonded crystals cannot undergo a transition to a ferroelectric phase, but
become antiferroelectric owing to reasons of electrostatics. Angular bonds,
which are mandatory for the development of ferroelectricity, cannot be formed
except under the influence of covalent contributions.
From the explanation of ferromagnetism, the thermodynamic statistics of
an order–disorder transition is well known. It can be understood as a conse-
quence of dipole orientation within the internal field.
The case of a displacive ferroelectric transition is more complex, espe-
cially since there does not exist a well-defined borderline between dipoles and
the surrounding lattice. Therefore at this stage a simple introduction to fea-
tures of such transitions could be helpful. A more thorough explanation and
quantitative results cannot be given here, extended outlines are presented in
Chap. 12.
The most representative materials among the displacive ferroelectrics are
the perovskites, of which a prototype is BaTiO3 (BT), the lattice structure
of which is depicted in Fig. 3.1. The core of the essentially cubic lattice cell is
formed by an oxygen octahedron with a 4-valent titanium ion in the center.
Neutrality and steric stability of the lattice is ensured by the 2-valent barium
ions at the corners of the unit cell. The essential part of the cell that con-
cerns the ferroelectric and piezoelectric properties is represented by the TiO6
octahedron. This can be expected not only because of the high permittivity
A Ba, Pb
O O
B Ti, Zr
Fig. 3.1. The perovskite structure ABO3 . BaTiO3 is the prototype ferroelec-
tric that crystallizes in this structure. Other important examples are PbTiO3 and
Pb(Zrx Ti1−x )O3
3 The Role of Ferroelectricity for Piezoelectric Materials 41
found in all titanium oxides themselves, but also because of the fact that the
temperature-independent quotient
T Cε 1 dε ε0 1
= · =− ∼= −10−5 K−1 (3.3)
ε/ε0 ε dT ε C
as represented by the Curie–Weiss law of BT can be found in equal magnitude
within all TiO2 compounds independent of their specific lattice structure [1].
Additionally, the reason for the development of a ferroelectric phase in-
stead of an antiferroelectric one below the Curie temperature can be found
within the TiO6 octahedra themselves and their interaction. For, the elec-
trostatic preference of the mutual saturation of the apparent charges at the
ends of dipole chains is overcompensated by contributions of the homopolar
parts of the Ti–O bonds. It has been shown by X-ray fine structure investi-
gations of BT [2] that already above the Curie temperature a closer binding
between the Ti ion and three of its closest oxygen neighbors exists, causing a
minor deviation from the central position and the formation of a tetrahedron-
like TiO3 configuration. But owing to the minor energy contribution of this
deviation the Ti ion changes the possible partners, continuously oscillating
from one position to the other. The average position is found in the center of
the unit cell.
The preference for a smaller atomic distance between the Ti ion and its
oxygen partners is due to the homopolar part of the Ti–O bond. Furthermore,
this homopolar contribution involves also the tendency of oxygen towards an-
gled valencies. This tendency is well known, e.g., from the water molecule, and
plays an important role in ferroelectricity. For, it favors the parallel orientation
of the polarization in neighboring lattice cells as illustrated in Fig. 3.2. In other
words, it causes the transition to ferroelectricity instead of antiferroelectricity
below the Curie temperature.
The transition from the paraelectric to the ferroelectric state, which we
have concentrated on up to now, is not the only ferroelectric phase tran-
sition occurring in BT. This can be seen, e.g., by the results shown in
Fig. 3.3. Here Fig. 3.3a shows the temperature dependence of permittivity of
Fig. 3.2. The homopolar contribution of the Ti–O bond involving the tendency of
oxygen towards angled valencies, which favors the parallel orientation of the polar-
ization in neighbouring lattice cells
42 W. Wersing et al.
10000
8000
Rhombo- Ortho- Tetragonal
Permittivity hedral rhombic
6000
e T11
4000
2000
e T33
20
Rhombo- Ortho- Tetragonal
16 hedral rhombic
12
Fig. 3.3. (a) Dielectric constants of BaTiO3 vs. temperature. The values of εT 11
and εT 33 in the tetragonal phase refer to single-domain crystals (according to Merz
[3]). (b) Spontaneous polarization of BaTiO3 vs. temperature, measured along the
pseudo-cubic (001) edge, i.e., parallel to the 3-axis in the tetragonal phase (according
to Merz [3]). In the orthorhombic phase or in the rhombohedral phase, the magnitude
√ √
of Ps is obtained by multiplying the value in Fig. 3.3b with 2 or 3, respectively.
The onset of Ps at TC appears to be rather continuous. Although this was found to
be due to crystal imperfections, it shows that the transition is only weakly first order
Fig. 3.4. The four phases of BaTiO3 . The arrows indicate the direction of the
spontaneous polarization related to spontaneous cell deformation
χ 2 1
A = A0 + P1 + P22 + P32 + ξ11 (P14 + P24 + P34 ) + terms of higher order.
2 4
(3.4)
Because of the general relation
∂A ∂ P
= E,
(3.5)
44 W. Wersing et al.
A-A0 A-A0
T>Tc T>Tc
T=Tc T=Tc
P3 P3
T=T0
T<Tc T<T0
a b
ξ11>0 ξ11<0
one gets in the case of high permittivity (ε >> ε0 or D ≈ P) from the Curie–
Weiss law (Eq. 3.1) as valid in the cubic state
T − T0
χ∼
= . (3.6)
ε0 C
Now, in the Devonshire theory the validity of (3.4) is extended to the re-
gion below the Curie temperature. Furthermore, it is assumed that all terms
of higher power in (3.4) do not depend on temperature significantly. Under
these assumptions, (3.4) represents a family of curves as shown in Fig. 3.5,
where A − A0 is plotted versus the polarization for different temperatures. It
should be noted that the Devonshire equation needs only the correct adapta-
tion of three constants for a roughly correct description of the permittivity,
polarization, and phase transitions observed in BT. In this description, the
observed lattice distortions are nothing but the electrostrictive consequence
of the spontaneous polarization (cf. Fig. 3.4).
Let us go back to Fig. 3.5, which describes the neighborhood of the Curie
temperature. In the case of Fig. 3.5a, no secondary minimum exists above the
Curie temperature and the curve of free energy is nearly parabolic with a
curvature corresponding to the Curie–Weiss law. Below TC , the central min-
imum becomes unstable and lateral minima develop in the (100) directions.
This represents a thermodynamic transition to a ferroelectric state of second
order. On the contrary, in case the side minima pre-exist above the Curie tem-
perature, owing to the already mentioned influence of covalency, the diagram
of (A − A0 ) vs. P transforms to the shape in Fig. 3.5b. The central minimum
and the (100) minima coexist within a limited temperature region, whether
they are stable or metastable. The transition from the non-ferroelectric cubic
state to the tetragonal ferroelectric state is of first order.
As mentioned already, on cooling down further two more ferroelectric
phase transitions occur. Both are of first order as shown in the complete
3 The Role of Ferroelectricity for Piezoelectric Materials 45
Arb. units
A0
Free energy A
Atetragonal
Aorthorhomb
Arhombohedral
Fig. 3.6. The free energy of the different ferroelectric phases of BaTiO3 as a function
of temperature (schematic)
energy phase diagram of BT (Fig. 3.6), where we have plotted the free energy
of the possible phases vs. temperature. Of course, the stable phase is repre-
sented by the lowest curve. But owing to the fact that we are dealing with
very small energy differences, especially near the transitions themselves, the
apparent properties often are co-determined by the neighboring metastable
phases.
Generalizing, we may call a phase ferroelectric when it has two or more
polar states (in the absence of an electric field) that can be shifted or switched
from one to another by an electric field. The polar character of these orien-
tation states must represent a completely stable configuration in zero field.
Ferroelectrics are therefore often considered to belong to the subgroup of the
10 polar crystal classes4 that are capable of being switched in some manner.
The possibility of switching usually implies a polar structure characterized
by small distortions from a higher-symmetry non-ferroelectric phase, which
is termed the prototype phase. Although the prototype phase is not necessar-
ily of nonpolar character, it proves to be so for the vast majority of known
ferroelectrics, and in the following we shall make this assumption generally.
As a result of the small structural displacements from the prototype phase,
a typical ferroelectric possesses a spontaneous polarization Ps , which decreases
with increasing temperature T , to disappear continuously, or more often dis-
continuously, at a certain temperature, the so-called Curie temperature Tc .
The phase above the Curie point is usually termed paraelectric. It is not, of
course, necessary that the paraelectric phase be the prototype phase, although
this is quite often the case.
A ferroelectric transition is usually connected with the condensation of
a soft polar or optically active mode of lattice motion at the Brillouin-zone
4
These classes are called polar because they possess a spontaneous polarization or
electric moment per unit volume.
46 W. Wersing et al.
5
Up to 130◦ C, depending on crystal quality.
6
Superscript T denotes “at constant stress”, see also Eqs.: (3.9)–(3.11).
3 The Role of Ferroelectricity for Piezoelectric Materials 47
400
K
cubic
320
T tetragonal
240
orthorhomb.
160
rhombohedral
80
0 0.3 0.6 0.9
u
a b
Fig. 3.8. Phase diagrams of Ba(Mex Ti1−x )O3 and Pb(Mex Ti1−x )O3
48 W. Wersing et al.
Li
Nb, Ta
Fig. 3.9. The ferroelectric structure of LiNbO3 and LiTaO3 . The structure consists
of a sequence of distorted oxygen octahedra joined by their faces along the trigonal
polar axis
character7 . Both compounds are uniaxial at all temperatures, with only a sin-
gle second-order transition from the paraelectric phase (point group R3m) to
the ferroelectric (R3̄m) phase.
Large optical-quality single crystals of LiNbO3 and LiTaO3 can be grown
by the Czochralski method and are commercially available. These materials
are generally used in the single-crystal form. The robustness, good perfor-
mance, and wide availability of these materials make them a popular choice
for commercial nonlinear optical and electro-optic devices, such as modulators
and harmonic generators as well as pyroelectric detectors. Particularly, they
are important for surface acoustic wave (SAW) devices (see Chap. 14).
The soft-mode picture introduced by Cochran [9] during the 1960s is consid-
ered as a key concept to explain the ferroelectric phase transitions in per-
ovskite (ABO3 ) compounds, since soft modes are strong indicators of lattice
instabilities and often precursors of such phase transitions. For a microscopic
understanding, we have to take into account furthermore the interaction be-
tween binding forces (attraction and repulsion) on one hand, and long-range
Coulomb forces on the other. It is now generally agreed that this delicate bal-
ance between short-range and long-range forces is connected to the homopolar
part of the Ti–O bond, i.e., the orbital hybridization between O 2p and Ti
3d states (for a detailed discussion see Chap. 21). This is also the reason for
the preference of a smaller atomic distance between the Ti ion and its oxygen
partners, which has been shown by X-ray fine structure investigations on BT.
7
Order-disorder ferroelectrics such as TGS have Curie constants up to two orders
of magnitude smaller.
50 W. Wersing et al.
A Pb
O O
B Ti
111-direction
Fig. 3.10. Local dipole moments in (111) direction induced by covalent Pb–O and
Ti–O bonds in rhombohedral Pb(Zrx Ti1−x )O3
3 The Role of Ferroelectricity for Piezoelectric Materials 51
8
Though the same letters are used to denote temperature and stress and entropy
and strain, respectively they can be well distinguished because stress and strain
are typed as tensors (bold type) or as tensor components (with indices).
52 W. Wersing et al.
dU = T dS + Tα dSα + Ei dDi
dA = −SdT + Tα dSα + Ei dDi
dG1 = −SdT − Sα dTα + Ei dDi (3.8)
dG2 = −SdT + Tα dSα − Di dEi
dG = −SdT − Sα dTα − Di dEi
and explicitly indicate the three principal variables for each potential.
When, according to the rule given, a particular set of independent variables
is held constant, a system will come to thermodynamic equilibrium in such a
way that the free energy – for which the constrained variables are the principal
ones – is minimized. Thus, for example, in the common experimental situation
for static measurements with T, Tα , and Ei fixed, irreversible processes bring
about an approach to equilibrium, which minimizes the Gibbs free energy
G. If the system undergoes a phase transition from phase A to phase B, the
transition occurs when GA (T, Tα , Ek ) = GB (T, Tα , Ek ). Therefore, this
equation determines the transition temperature, i. e., Tt = Tt (Tα , Ek ).
If we take temperature, stress, and electric field as independent variables,
we get from the Gibbs potential the following relations:
∂G ∂G ∂G
−S = , − Sα = , − Di = . (3.9)
∂T Tα ,Ei ∂Tα T,Ei ∂Ei T,Tα
which, in their isothermal form (∆T = 0), are called basic piezoelectric equa-
tions. In addition, six other linear equations, i.e., three further sets of basic
9
i = 1, 2, 3 and α = 1, . . . , 6 (Voigt notation); Tα Sα ≡ Tα Sα (summation
α
convention).
3 The Role of Ferroelectricity for Piezoelectric Materials 53
Sα = sE
αβ Tβ + Mikα Ei Ek . (3.12)
From the elastic Gibbs potential G1 one can get a similar equation of state
for centrosymmetric crystals:
Sα = sD
αβ Tβ + Qikα Di Dk . (3.13)
From (3.15), the permittivity of the cubic phase can be easily calculated (as
shown below) to be ε = 1/χ. Therefore, the temperature dependence of χ can
be written in the Curie–Weiss form
1
χT = (T − T0 ) = χ0 (T − T0 ) . (3.16)
ε0 C
Minimizing (3.15a) with respect to Di at constant temperature generates the
description of the different possible ferroelectric phases, expressed in terms
of their spontaneous polarization. For ferroelectrics with a cubic prototype
phase, such as BT, beside the cubic phase, three stable phases can exist prin-
cipally: a tetragonal, an orthorhombic, and a trigonal (rhombohedral) phase.
The polar axis in these three phases is (001), (011), and (111) respectively,
and the spontaneous polarization P s and Gibbs potential G1 of each phase
follows as:
s2 −a 4bχ0 (T − T0 )
P = 1± 1− , (3.17)13
2b a2
1 s4 4 s2
G = G0 − P a + bP (3.18)
4 3
where the coefficients a and b can be expressed for the different phases as
follows:
13
This eqn for Ps and the following eqns for material properties hold only for a single
domain or for a single crystal with completely aligned spontaneous polarization.
56 W. Wersing et al.
at = ξ11 ,
tetragonal phase: (3.19)
bt = ς111,
2 (ξ11 + ξ12 ) ,
1
ao =
orthorhombic phase: 1
(3.20)
bo = 4 (ς111 + 3ς112 ) ,
ar = 13 (ξ11 + 2ξ12 ) ,
rhombohedral phase: (3.21)
br = 19 (ς111 + 6ς112 + 2ς123 ) .
The sign of the a coefficient determines the order of the corresponding tran-
sition. It is, e.g., of first order if the corresponding a is negative. Then, the
upper sign in (3.17) holds.
In the case of a second-order transition, the terms of sixth order in (3.15)
are often neglected. This approximation yields a simplified expression for the
spontaneous polarization
χ0 (T0 − T )
P s2 = . (3.22)
a
The transition temperature TC from the nonpolar phase to a ferroelectric
phase in the case of a second-order transition is identical with T0 . In case of
a first-order transition, TC can be calculated by imposing that the Gibbs free
energy (for zero stress and field) of the polar and nonpolar phase are equal,
which leads to
3 a2
TC = T0 + 16 χ b. (3.23)
0
where a and b for the different phases are given in (3.19–3.21). The permit-
tivity components normal to P s are
1 1
for the tetragonal phase: εT T
11 = ε 22 = · , (3.26)
P s2 (ξ12 − ξ11 ) + (ς112 − ς111 ) P s2
1 1
εT
11 = s2 · 1 s2
,
P 2
( ξ 12 − ξ 11 ) + 1
4
(2ς123 − ς112 − ς111 ) P
for the orthorhombic phase: (3.27)
1 1
εT
22 = · ,
P s2 (ξ11 − ξ12 ) + (ς111 − ς112 ) P s2
1 1
for the rhombohedral phase: εT = εT
22 = . (3.28)
11
P s2 23 (ξ11 − ξ12 ) + 49 (ς111 − ς123 ) P s2
Figure 3.13 shows the permittivities calculated with the aid of (3.25–3.28)
in the vicinity of the technologically most interesting phase transition between
the tetragonal and the rhombohedral ferroelectric phase of PZT known as the
morphotropic phase boundary (MPB). The strong peak of the permittivity
components normal to the spontaneous polarization is characteristic of this
boundary (see below).
Applied field Ek and stress Tα are assumed to be small enough to make the
induced displacement ∆Di in (3.11) considerably smaller than spontaneous
polarization. Therefore, (3.24) can be inserted into (3.13) with ∆Di taken
from (3.11) for constant T (∆Tα = Tα and ∆Ek = Ek ) to give:
Sα = Sαs + sD s s T
αβ Tβ + 2Qikα Pk diβ Tβ + 2Qjkα Pk εij Ei . (3.29)
Comparing (3.29) with the first equation of (3.11) for constant T and zero
stress and bearing in mind that the field- and stress-induced strain in the
polar phase ∆Sα = Sα − Sα s , one finds:
diα = 2Qjkα Pks εTij . (3.30)
Inserting (3.30) into (3.29), one finds similarly for Tβ = 0:
sE D s s T
αβ = sαβ + 4Qikα Qjlβ Pk Pl εij . (3.31)
From (3.30) we conclude that the piezoelectric effect of ferroelectrics with a
centrosymmetric prototype phase is the result of electrostriction linearized
by spontaneous polarization. However, the most important result of (3.30)
and (3.31) is the fact that the dielectric anomalies occurring near ferroelectric
phase transitions are directly reflected in the piezoelectric and elastic behav-
ior, too.
Today’s most important piezoelectric materials are based on the solid solu-
tion PZT, which transforms only to two ferroelectric phases; it is tetragonal at
low Zr concentrations x and rhombohedral at high x. The main nonzero prin-
cipal axis components of the piezoelectric matrices for both of these phases
are given below:
58 W. Wersing et al.
15
Principially, also the movement of a phase boundary (e.g. the boundary between
the tetragonal and rhombohedral phase in PZT) driven by an external field am-
plitude can contribute to the extrinsic response.
3 The Role of Ferroelectricity for Piezoelectric Materials 59
Fig. 3.11. Polarization and strain hysteresis loops of tetragonal Pb(Zrx Ti1−x )O3
ceramics (left-hand side). The bold curves are obtained when a field of magnitude
smaller than the coercive field strength Ec is applied. The upper right figure shows
schematically the microstructure (grains and domains) before and after poling. The
highest possible degree of poling results in a remnant polarization Pr = <Ps >
0.86 Ps . The lower right figure shows that any shift of 90◦ (non-180◦ ) domain walls
due to an applied electric field results in a change of polarization and deformation
(extrinsic contribution to the piezoelectric effect)
In this section, we deal with the intrinsic piezoelectric effect mainly of lead-
containing perovskites; the extrinsic effect is discussed in the following section.
When comparing different piezoelectric material classes (Table 3.1), one
finds that all non-ferroelectric piezoelectric materials exhibit only very small
piezoelectric constants diα and small electromechanical coupling constants
kiα . Furthermore, these materials are only piezoelectric in the form of sin-
gle crystals or textured thin films. On the other hand, these materials show
extremely good long-term stability, and in many cases it is possible to find
crystal cuts that also give high temperature stability. Both these properties
make them important for frequency and time standards.
In applications where higher piezoelectric constants or higher coupling
coefficients are required together with suitable high stability, as for example
for broadband frequency filters (see Chap. 14), ferroelectric materials can be
used with high Curie temperatures as compared to the operating temperature
and with a material composition far away from any possible phase transition
60 W. Wersing et al.
–200 3000
ε/ε0
–300 2000 1000 2000
TCε
–400 1000 ε/ε0
TCF
–500 0
0 0
Fig. 3.12. Permittivity ε/ε0 , planar frequency constant NpE , and their temperature
coefficients of Pb(Zrx Ti1−x )O3 + 2% Nd ceramics
We have already seen in Fig. 3.2 that in BT single crystals the permittivity
components perpendicular to the spontaneous polarization approach a strong
maximum when going from the tetragonal to the orthorhombic phase and a
16
In this section we generally consider principal axis components i.e. the primes to
characterize principal axis components have been omitted for simplicity.
3 The Role of Ferroelectricity for Piezoelectric Materials 63
Fig. 3.13. Single-crystal permittivities ε11 T = ε22 T and ε33 T and effective permit-
tivity εT ∗ of unpoled ceramics of Pb(Zrx Ti1−x )O3
similar peak seems to occur when going from the orthorhombic to the rhombo-
hedral phase. On the other hand, the permittivity component εT 33 parallel to
the spontaneous polarization shows almost no change at the phase transitions
and its value is very much smaller than εT 11 = εT 22 . The permittivity max-
ima in Fig. 3.3a would be even more pronounced if instead of the permittivity
components, which refer to the axis system of the cubic prototype phase, the
principal axis components were shown.
A similar strong dielectric anisotropy is observed in the vicinity of the
MPB of PZT, i.e., a pronounced permittivity peak for εT 11 = εT 22 and only
a very flat curve for εT 33 (Fig. 3.13). As a result of the strong maximum of
εT 11 = εT 22 , one obtains together with this maximum also a maximum in
the effective permittivity εT∗ of unpoled ceramics and similarly (not shown in
Fig. 3.13) also in the effective permittivities of poled ceramics.
This important property of the MPB has been well known for a long
time [19]; however, in recent years it has been rediscovered and has initiated
intensive discussions [20]. The large dielectric permittivity perpendicular to
the spontaneous polarization facilitates polarization rotation by electric fields
applied along nonpolar axes [21], i.e., the effect of strong polarization rotation
often discussed in relaxor ferroelectrics such as PbTiO3 –Pb(Mg1/3 Nb2/3 ) is
nothing but the dielectric anisotropy discussed here.
The strong dielectric anisotropy, as well as the permittivity maxima at the
morphotropic phase boundary, is not only of theoretical interest but are also
very important for the tailoring of application-adapted piezoelectric ceramics.
The strong dielectric anisotropy at the MPB can be explained with the aid of
the results already obtained for BT in Sect. 3.1.
64 W. Wersing et al.
17
In case of a first order phase transition (cf. (3.17) and (3.22).
18
Calculated from the lattice parameter ratio [22] and the electrostrictive coeffi-
cients [32].
3 The Role of Ferroelectricity for Piezoelectric Materials 65
Fig. 3.14. Permittivities of (Pb0.83 Ba0.05 Sr0.12 )(Zrx Ti0.98−x Nb0.02 )O3 . Single crys-
tal permittivities ε11 T = ε22 T and ε33 T , effective permttivity εT ∗ of unpoled ceram-
ics, compared to measured permittivity εT Cer of unpoled ceramics
seen that in the investigated ceramic system the domain wall contribution
to permittivity is higher on the rhombehedral side of the MPB than on the
tetragonal side (see Chap. 3.3). This agrees with the experimental result that
the dielectric losses on the rhombohedral side of the MPB are about twice as
large as those on the tetragonal side.
The strong anisotropy of single-crystal permittivities determines also the
piezoelectric and elastic properties of single crystals and hence those of ce-
ramics, thin films, and composites.
From the result obtained for the single-crystal permittivities, the princi-
pal axis components of the piezoelectric constants can be calculated using
(3.32) and (3.33) with the aid of proper electrostrictive coefficients. Usually,
measurement techniques to determine electrostrictive coefficients require ori-
ented single crystals. However, a new, highly sophisticated method has been
developed to measure electrostrictive tensor components of ferroelectric ce-
ramics above their Curie temperature, based on X-ray diffraction using a
position-sensitive detector in combination with a time-resolved registration
technique [30].
Apparently, the maximum of εT11 = εT22 in the vicinity of the MPB causes
the piezoelectric shear modulus d15 = d24 to reach a pronounced maximum
at the same time. On the other hand, the components d31 = d32 and d33
show only a very flat behavior at the phase boundary. For example, Fig. 3.15
shows the piezoelectric constants of single-domain and polycrystalline PZTs,
which have been calculated with the aid of Devonshire’s theory [17] using
electrostrictive coefficients from measurements on PZT ceramics [30].
3 The Role of Ferroelectricity for Piezoelectric Materials 67
Quite similarly, one finds that the maximum of εT11 = εT22 in the vicin-
ity of the MPB causes the shear compliance modulus sE E
44 = s55 to reach a
pronounced maximum, whereas the nonshear compliance moduli again stay
quite flat.
The calculation of the temperature coefficients of the dielectric, elastic, and
piezoelectric single-crystal properties such as T CεT 11 = (1/εT 11 )(dεT 11 /dT )
is also straightforward now with the aid of Devonshire’s theory, i.e., with the
equations given in Sects. 3.1.4.1 and 3.1.4.2 (see also Chap. 7).
Both the tetragonal crystal distortion c/a and the rhombohedral distortion
expressed by the angle α of the rhombohedral cell can be measured directly by
means of X-ray diffraction. Often, the rhombohedral distortion is expressed
by the deviation of the angle α from π/2:
δ = (π/2 − α) . (3.42)
From (3.40b) and (3.41b) one can see that in the hypothetical case of
infinite poling (θ = 0) the remnant strain in a rhombohedral ceramic would
be equivalent to δ. Thus sometimes δ is called the homogeneity parameter (cf.
Chap. 4) and is generally used to characterize the spontaneous cell deformation
along the polar axis. By comparing (3.41a) and (3.41b) one can see that in
the tetragonal phase δ is then given by:
δ= 2
3 (c/a − 1) . (3.43)
22
This deformation is zero in case of a phase transition without change of volume.
23
If the strain components are not referred to principal axis but to the cubic pro-
totype phase’s axis we have instead S4s = π/2 − α.
3 The Role of Ferroelectricity for Piezoelectric Materials 69
cos3 θ
with κ = 2 cos θ − 1.
Figure 3.16 shows the parameter κ as a function of the mean degree of ori-
entation of the spontaneous polarization of all domains relative to the external
poling field. In general, κ decreases with increasing degree of orientation from
1 to 0 (at <cos θ> = 0.5), and then increases again. The exact behavior of
κ depends on which distribution function for the orientation of polarization
was used to calculate the average, or which domain processes were assumed
to be applicable. In the simplest case, the averaging can be performed if one
assumes that, starting at a statistical orientation of polarization directions in
the unpoled ceramic, with increasing degree of poling the polarization direc-
tions continue to be statistically distributed, but in a confined angle range θp
(Fig. 3.17(b)) where θp is a function of poling degree. This leads to a behavior
of κ as described by the solid line in Fig. 3.16. It is more realistic, however,
to assume that with increasing poling field the distribution of polarization
directions transforms from the behavior shown in Fig. 3.17(a) to that shown
in Fig. 3.17(c) because 180◦ and 90◦24 domain walls (Fig. 3.17(c )) with fitting
orientations are shifted. In this case, θ2D increases to 90◦ and a further increase
of poling field leads to Fig. 3.17(d) after 90◦ domain walls (Fig. 3.17(d )) have
been shifted. This behavior of κ is described by the dashed line in Fig. 3.16.
Since the difference between the two described distributions of polarization
24
We use here 90◦ walls synonymously for 71◦ and 109◦ walls in rhombohedral
materials.
70 W. Wersing et al.
Single domain
θ
κ
Ps
Electrodes
Fig. 3.16. κ = 2(<cos3 θ>/< cos θ>) − 1 as function of mean degree of orientation
of spontaneous polarization <cos θ> of all domains in relation to an external pol-
ing field. The two curves correspond to two different distributions of polarization
orientations
directions does not essentially influence the behavior of κ, the simpler first
distribution has been used for the calculation of piezoelectric mean values in
Fig. 3.15.
A review of (3.44) and (3.45) enables the understanding of the piezoelectric
properties of PZT ceramics not only in the vicinity of the morphotropic phase
boundary but also far away from it, e.g., that of PbTiO3 ceramics. Further-
more, the dependence of these properties on the degree of domain orientation,
i.e., on the degree of poling, can be seen.
Let us first consider the piezoelectric properties far away from the mor-
photropic phase boundary, e.g., those of PbTiO3 . Typical properties of mod-
ified PbTiO3 are around d33 /d31 ≈ −5 and d33 /d15 ≈ 2. Therefore, with
certain material modifications it can be achieved that the term (3d31 + d33 −
d15 ) in the first line of (3.45) tends to zero. As a result, at a certain degree of
poling, such that <cos θ> ≈ 0.5, <d31 > will assume very small values. Be-
cause at the same time <d33 > can have quite normal values, the ratio |d33 /d31 |
can achieve very high values (20). This effect is known as anisotropic piezo-
electric effect [32] and is found, e.g., in lead titanate ceramics modified with
Ca [33] and in those modified with Sm [34, 35]. The large anisotropy of the
piezoelectric effect implies that these ceramics possess an especially high hy-
drostatic piezoelectric effect dh = d33 + 2d31 . They are therefore very inter-
esting for underwater transducers, e.g., for hydrophones [36].
Now we consider the piezoelectric properties in the vicinity of the MPB.
As mentioned above, in the vicinity of the MPB the maximum of εT11 =
εT22 causes the piezoelectric shear modulus d15 = d24 to reach a pronounced
maximum as well. On the other hand, the components d31 = d32 and d33
show only a flat behavior at the phase boundary (Fig. 3.15). From (3.45) it
can be seen that for an adequate degree of poling, i.e., κ ≈ 0 (Fig. 3.16), the
large value of d15 drives the effective piezoelectric constants d∗ 31 and d∗ 33
of a polycrystalline material into a pronounced maximum, too. On the other
hand, in those crystallites that are completely oriented parallel to the external
field (θ = 0, κ = 1), e.g., (001)-oriented tetragonal crystallites or (111)-
oriented rhombohedral crystallites (Fig. 3.18), the large d15 values cancel in
(3.44) and (3.45) and do not contribute to the effective constants d∗ 31 and
d∗ 33 . Obviously, the contribution of those crystallites with a Ps direction that
considerably deviates from the 3-axis (poling direction) of the ceramic sample
contribute strongly; particularly crystallites with a Ps direction of 57◦ to the
3-axis of the sample contribute optimally. Furthermore, it is quite interesting
(Fig. 3.18) that just those crystallites that do not contribute much to the
intrinsic piezoelectric effect, i.e., (001)-oriented tetragonal and (111)-oriented
rhombohedral crystallites, optimally contribute to the extrinsic piezoelectric
effect via 90◦ domain wall processes. On the other hand, those crystallites that
optimally contribute to the intrinsic piezoelectric effect, i.e., (111)-oriented
tetragonal and (001)-oriented rhombohedral crystallites, cannot contribute to
the extrinsic effect because the applied driving field will not favor any of the
other possible domain states, and the domain structure remains stable.
72 W. Wersing et al.
tetragonal rhombohedral
Ps
(001)- Ps
Ps Ps
oriented
Ps
(111)- Ps
Ps Ps
oriented
Finally, we would like to stress here that with a similar consideration the
elastic behavior of PZT ceramics in the vicinity of the MPB can be understood.
It can be shown that in polycrystalline PZT non-shear compliance components
∗ E∗
such as sE11 or s33 go through a considerable maximum at the MPB driven
by the large single-crystal shear value sE 44 : a fact which is well known and
widely used for many piezoelectric applications. Furthermore, the temperature
dependence of dielectric and elastic properties of PZT ceramics in the vicinity
of the MPB can be understood similarly, a fact that is very important for the
application of piezoelectric ceramics in mechanical frequency filters [37] (see
also Chap. 7).
which have been poled in (001) direction. On the other hand, the coeffi-
cients measured along the direction of the polar crystal axis were found to
be up to one order of magnitude smaller [38]. These results are listed in
Table 3.2. As can be seen from Fig. 3.18, 180◦ domain wall switching in a
(001)-oriented crystal leads to a domain wall distribution where the spon-
taneous polarization of each domain points in the direction of one of the
four possible cell diagonals (θ = 54.7◦ ), whereby all these four directions
are energetically equivalent. From√ Fig. 3.16, we see that in this case, i.e. for
<cos 54.7◦ > = cos 54.7◦ = 1/ 3, the parameter κ ≈ 0 holds and, thus, the
huge piezomodulus d15 optimally contributes to d33 ∗ . On the other hand, in
case of an applied field parallel to the spontaneous polarization (θ = 0◦ ), the
parameter κ ≈ 1 holds and d15 does not contribute to d33 ∗ .
Similarly, the piezoelectric properties of ferroelectric thin films can be un-
derstood. It has been shown that the intrinsic piezoelectric effect of films ori-
ented along their polar axis must be smaller than that of randomly oriented
films [39].
David et al. investigated rhombohedral PZT films with (001), (111), and
randomly oriented crystallites [40]. In case of an applied field amplitude of
0.1 Vµm−1 , these films showed effective d33 values of 100, 77, and 63 pm V−1 ,
respectively. It should be mentioned that these values have not been corrected
for the clamping caused by the substrate. With increasing field amplitude, the
effective d33 value of the (111)-oriented film increased especially strongly, ob-
viously because of domain wall motions which are possible in this orientation,
as has been discussed in the last section.
Kügeler et al. [41] studied (111)-oriented Pb(Zrx Ti1−x )O3 films for micro-
electromechanical system (MEMS) applications. A maximum of the effective
piezoelectric coefficients d33 (d33 ≈ 100 pm V−1 ) has been found for the film
composition x = 0.45 slightly on the tetragonal side of the MPB. For more
tetragonal compositions, x < 0.45, d33 decreased very rapidly. On the other
hand, on the rhombohedral side of the MPB, the effective piezoelectric module
decreased only very slightly; even at x = 0.6, the d33 value was only less than
10% smaller than the maximum value at x = 0.45. This result fully agrees
74 W. Wersing et al.
with the discussion of the last section. Near the MPB, the intrinsic effect of
tetragonal films reached a maximum owing to the strong contribution of the
shear effect to the effective d33 . Although the intrinsic effect was low at the
rhombohedral side of the MPB, a strong piezoelectric response was found as a
result of 90◦ (72◦ /108◦ ) domain wall movements that could take place over a
wide composition range because of the relatively high amplitude (700 kV m−1 )
of the measuring field.
+
+ – +
– – +
– +
– +
– +
If the domain walls are fixed and no other defects contribute to piezoelec-
tricity in ceramics, then an averaging calculation would predict all the main
intrinsic electromechanical properties when using appropriate values of single-
domain crystals and a simple orientation distribution. This calculation [45]
has to take into account the dielectric, piezoelectric, and elastic interactions
76 W. Wersing et al.
Fig. 3.20. Domain structures in polycrystalline materials: (a) simple lamellar twin-
ning. (b) banded twin structure in ceramics [44]
between the domains, their shape and configuration, and their degree of orien-
tation. Muratake [46] first introduced a method that takes into account some
interaction effects by considering electric and elastic boundary conditions of
the grain. Turik [47] improved the Muratake model by using a laminar grain
consisting of a stack of 90◦ or 180◦ domains. For the laminar grain of BT, the
effective piezoelectric, elastic, and dielectric coefficients were calculated con-
sidering the boundary conditions between the domains. The experimentally
determined piezoelectric constants d31 and d33 of BT ceramics are, however,
much larger than the calculated ones. This deviation between calculated and
measured piezoelectric constants had led Turik to suppose additional “orien-
tational” contributions that may be caused by domain wall motions.
Since the 1980s, Arlt et al. (e.g., [43] and [44]) have systematically stud-
ied the microstructure and dynamics of 90◦ domain wall motion in BT and
PZT ceramics. They could explain the strong increase in the piezoelectric
and dielectric coefficients by shifts and oscillations of 90◦ -type domain walls
(90◦ in the tetragonal, 71◦ and 109◦ in the rhombohedral phase). They used
Fousek’s model [48] to evaluate the dielectric, piezoelectric, and elastic coeffi-
cients referring to a special distribution function and by averaging the domain
contributions in every grain of the ceramic samples. A similar work was done
by some Russian scientists (e.g., [49]). They considered some special distribu-
tions of domain structures in ceramics (the laminar twin structure) to assess
the physical properties of ferroelectric ceramics.
All these models describe the average dynamic behaviour of the domain
wall motion in the single crystal with respect to the different orientations of
polarizations. All these averaging methods have drawbacks and are of limited
value for the calculation of the volume part of contributions to piezoelec-
tric constants of ceramics. The experimentally determined piezoelectric con-
stants d31 and d33 are effectively larger than those calculated by the averaging
3 The Role of Ferroelectricity for Piezoelectric Materials 77
methods. A similar disagreement was found for the dielectric constants, which
in ceramics are much larger than expected from theoretical calculations.
Generally speaking, the electromechanical properties of ferroelectric ce-
ramics are caused by two mechanisms: (1) the intrinsic piezoelectric effect
connected with the deformation of each unit cell of the ferroelectric material
by an electric field, and (2) the extrinsic piezoelectric effect based on domain
wall motion. In PZT this motion refers to non-180◦ domain walls and the
movement of the interphase boundaries between the tetragonal and rhombo-
hedral phase regions. It is believed that the piezoelectric effect in polycrys-
talline ferroelectrics is caused not only by the ionic displacement in connection
with the change of the polarization magnitude but also by the movements of
domain walls and interphase boundaries.
Studies on materials such as BT and PZT [50] have shown that 60–70% of
the piezoelectric coefficient values may originate from extrinsic contributions.
In fact, the performance of many capacitors, transducers, and actuators is
based on the control of the domain structure under the applied electric field.
Microstructural investigations (e.g. [51] and [52]) have also shown that
poled ferroelectric ceramics contain a large number of non-180◦ domain
boundaries which strongly affect their electromechanical behavior. In general,
these domain structures give rise to complex linear and nonlinear behavior,
which is very sensitive to the quality of the sample and its defect concentration
as well as to the external boundary conditions.
Now let us consider a physical model in order to obtain an insight into the
relationship between electromechanical properties and domain wall motions as
well as into the movement of interphase boundaries in ferroelectric ceramics.
The movements of the domain walls are not independent, but are strongly
interacting. The microscopic process for an individual domain wall motion is
rather complicated and not yet well understood. However, in order to describe
the macroscopic effects resulting from the collective motion of domain walls,
the microscopic details of each individual domain may be neglected.
Let us assume that the average displacement of a domain wall in a certain
direction is quasi one-dimensional, i.e., the wall moves as a plane without
bending. Figure 3.21 shows the variation of the wall energy as a function of
the position generated by the interaction of the domain wall with the lattice,
point defects, dislocations, and neighboring walls in a ceramic sample. The
height of the potential barrier (φ1 + φ2 ) is the minimum energy required for
the wall to relocate from a depoling state to a poled state. φw is the height of
the potential barrier for a non-180◦ domain wall to cross one lattice distance.
In a depolarized ceramic sample, all domain walls are randomly oriented
and the distribution of the domain walls possesses the symmetry ∞∞m. Ob-
viously, the domain walls are located at the minima of the potential energy
(z = 0). When an external a.c. field is applied, the domain walls oscillate.
78 W. Wersing et al.
POTENTIAL ENERGY
fw
f1
f2
Z=–a 0 Z=a
Z
Fig. 3.21. Energy levels in the domain wall potential of ferroelectric ceramics
However, as long as the magnitude of the domain wall vibration is not very
large, the macroscopic potential energy for the domain wall motions is sym-
metric with respect to the z-axis, i.e., U (∆z) = U (−∆z).
The situation is different for poled ceramics. During the poling process,
the domain walls overcome neighboring energy barriers and move into new
equilibrium positions. As shown in Fig. 3.21, the new positions are metastable
states (z = a or z = −a). After the field is removed, the orientation distri-
bution of the domain walls becomes conical in the poled samples. The po-
tential for the domain wall motion is no longer symmetric in the poled state,
i.e., U (a − ∆z) = U (a + ∆z). It also has been shown [53] experimentally that
the domain wall motion in poled ceramics is highly nonlinear. Therefore, the
potential energy for the domain wall motion may be expanded as a polynomial
function of the domain wall displacement ∆l [54]:
c1 2 c2 3 c3 4
U = U0 + ∆l + ∆l + ∆l + higher-order terms, (3.46)
2 3 4
where U0 is the residual energy of a domain boundary, which is assumed
to be independent of the domain wall motion. The presence of the third-
power term describes the asymmetric feature of the domain wall motion in
ferroelectric ceramics. The odd terms will be zero if the domain wall vibrations
take place around a potential energy minimum of the domain wall. Therefore,
the differential equation of the forced vibration of a non-180◦ domain wall in
ferroelectric ceramics may be expressed as follows (e.g., [48] and [54]):
∂U ∂WE ∂WM
Am∆¨l + Ab∆l˙ + =− + , (3.47)
∂∆l ∂∆l ∂∆l
where m represents the effective mass of the domain wall. A is the area of the
domain wall, b is the damping constant, and the dots on top of ∆l represent
the derivates with respect to time. The third term is the restoring force.
3 The Role of Ferroelectricity for Piezoelectric Materials 79
DIELECTRIC CONSTANT
SOFT PZT F = 200Hz SOFT PZT
–d31&d33 (10–12 V/M)
500 800
ε11
Fig. 3.22. (a) Depencence of the piezoelectric coefficients d33 , d15 , and d31 on the
a.c. electric field amplitude. (b) dependence of the permittivities ε11 T and ε33 T on
the a.c. electric field amplitude. The two insertions show the P–E loops at low and
medium-strong electric fields, respectively
WE and WM are the energies of the average electric and elastic dipoles induced
by the electric field and the mechanical stress, respectively. In general, the
electric and elastic driving frequencies are different.
Equations (3.46) and (3.47) show that the nonlinearities in ferroelectric ce-
ramics can be evaluated from the non-180◦ domain wall motion. Furthermore,
the domain wall motion can be influenced by pinning effects.
Figure 3.22 shows the a.c. electric field dependence of the piezoelectric and
dielectric coefficients of soft PZT ceramics [54].
T / MPa
Fig. 3.23. (a) strain and (b) depolarization of hard and soft PZT ceramics under
uniaxial stress [57]
are essentially linked, as can be seen from (3.48) and (3.49) and in Fig. 19.7
of Chap. 19; i.e., both hysteresis and nonlinearity are results of the same do-
main wall pinning processes. This link between hysteresis and nonlinearity
is characteristic for soft PZT ceramics in contrast to the behaviour of hard
PZT ceramics in which the response hysteresis cannot be predicted from the
nonlinearity of the piezoelectric coefficient [20].
Also, a large mechanical stress applied parallel to the poling direction
induces a significant domain wall motion. With increasing load, the ferroelastic
domain switching causes an additional strain and depolarization ∆P [56].
Figure 3.23 shows the strain and depolarization of hard and soft PZTs under
uniaxial stress.
0.8
(a) 2
0.6
0.4
0
0.2
0.0
–2
Polarization [C/m2]
0.8
Strain [10–3]
0.6 (b)
2
0.4
0.2
0
0.0
0.8
0.6 (c)
2
0.4
0.2
0
0.0
–0.2
–2 –1 0 1 2 –2 –1 0 1 2
Electric Field [kv/mm]
Fig. 3.24. Macroscopic polarization and strain in soft PZT measured at 10−3 HZ,
after (a) 0, (b) 3 · 106 , and (c) 108 bipolar electric fatigue cycles at 50 Hz,
2 kV mm−1 [57]
small barrier to domain wall motion. Their average effect is of a frictional type.
Thus a viscosity for domain wall motion should be defined. In this regime, the
domain wall motion is essentially smooth at varying velocities yielding slightly
modified dielectric constants. Further increase of defect size alters this smooth
motion into discrete jumps of the domain wall. Large changes in the dielectric
constant as well as offsets arise. Finally, a further increase in defect size may
be sufficient to block the polarization within a certain volume. No further
switching takes place in that volume, the size of which has not yet been
determined [58].
It is generally agreed that a certain degree of clustering of defects develops
during fatigue. Whether such clusters consist of oxygen vacancies, or defect
states at grain boundaries, at domain walls, at microcracks, or at electrode
contacts, is still an open question [59].
PZT is one of the most commonly used materials for electromechanical trans-
ducers and actuators. In these applications, the material is subject to high
electric fields and mechanical stress under which it may fracture. The limited
performance of the components due to cracking still impedes their large-scale
use in highest power applications. Therefore, the fracture behavior of the
material needs to be well understood in order to improve their lifetime and
reliability.
In a cracked PZT component, the high stresses in the vicinity of the crack
tip cause ferroelastic domain switching within a finite zone ahead of the crack
82 W. Wersing et al.
front. Once the crack penetrates the zone, toughening takes place in the region
behind the crack tip, where compressive stresses close the crack and shield the
crack tip from the external load.
A few attempts have been made to experimentally quantify the fracture
behaviour of PZTs. Some of them have been devoted to the distribution of the
shielding stresses in the wake of a growing crack, while others were focused on
the measurement of the domain switching zone using experimental methods
such as atomic force microscopy, interference contrast, and X-ray diffraction
(e.g., [60]).
In polarized ferroelectric ferroelastic materials, the switching mechanism is
accompanied by a considerable change of the electric potential on the specimen
surface. This electric potential can be imaged with the help of liquid crystals.
Ferroelectrics exhibiting fairly large domains have been characterized on a
local scale using this approach. This technique can be extended to visualize
the ferroelastic domain switching zone in front of a crack tip in PZT [60]. Using
this technique, it is further possible to track the evolution of the switching
zone during crack propagation. For a better understanding of the mechanical
behavior of ferroelectric ferroelastic materials, an analysis of the crack tip field
within the switching zone is desirable. As soon as the domain switching takes
place, the local poling condition and thereby the macroscopic piezoelectric
coefficients are altered.
The relation between switching behavior and fracture toughness under
applied electric and elastic fields in ferroelectric materials has not been fully
explored.
3.3.8 Conclusions
the outlines as given in Chaps. 2 and 3, up to now, the following rules can
be deduced for the development and application of these different classes of
piezoelectric materials:
1. Non-ferroelectric materials cannot be piezoelectric, unless they are single
crystalline or at least produced by oriented, e.g., epitaxial, growth. They
exhibit only small piezo-moduli and small permittivities, a typical exam-
ple being quartz. The elastic properties resemble those of other inorganic
materials. In single crystals, all material properties are stable over time
and depend on crystal quality only. Owing to the existing anisotropy, nor-
mally orientations of minor temperature dependence can be found. Thus
specific cuts are widely used as frequency- or time standards (cf. Chap. 2).
In addition, we would like to mention langasite as a material with the same
crystal structure as α-quartz, but applicable up to extreme temperatures,
e.g., for sensors working up to 900◦ C. Epitaxial layers are by far cheaper
than specific cuts of single crystals. Only polar substances should be used
because normally they cannot be deposited as single crystals, but as ori-
ented, texturized layers. Important examples are hexagonal ZnO and AlN.
Both are used for integrated microwave resonators and filters in telecom-
munications as well as increasingly in biosensor arrays.
2. Higher piezoelectric coupling than in these polar substances cannot be
found except in ferroelectric materials. In case the highest stability is
required, normally single-crystalline slices or at least textured epitax-
ial layers are used. They must exhibit a Curie temperature that is
high enough, or more precisely, no phase transition should occur in the
neighbourhood, both with respect to a change of temperature and to a
variation of composition. Important examples in the case of uniaxial ferro-
electrics are LiNbO3 and LiTaO3 , and in the case of polyaxial ferroelectrics
PbTiO3 . Such materials exhibit a piezoelectric coupling that is about 10
times higher than that of quartz. They have medium permittivity and no
anomalies of the elastic properties. In uniaxial ferroelectrics, 90◦ domain
rotation cannot occur, but only single crystals or texturized layers can
be used. Extrinsic effects, which could affect long-term stability, are ex-
cluded. By specific polarization means, e.g., by crystal growth within an
electric field, even single-domain substrates can be produced. Thus, from
the point of view of piezoelectricity, the highest long-term stability can be
warranted, which is mandatory for frequency control and band filtering
as well as for signal transmission or identification (cf. Chap. 14).
3. In case an even higher energy coupling is desired, one has not only to
use ferroelectric substances; additionally we cannot avoid any more the
neighbourhood of ferroelectric phase transitions; such a neighbourhood
can enhance the electromechanical coupling by almost two orders of mag-
nitude. Stability is no more a decisive criterion for technological applica-
bility, it is only a part of the decisive figure of merit.
84 W. Wersing et al.
These concessions to highest stability bring about not only applications where
highest coupling is desired; even more they open the way to the use of poly-
crystalline substances, which means to ceramic production procedures. These
procedures are cheaper than growing single crystals or textured layers by
orders of magnitude. Their technology has been well elaborated for the mass
production of electrical devices such as resistors, capacitors, or magnetic cores.
Their shape can be adapted exactly to the requirements of the user. The well-
known ceramic thin-layer and thick-film technologies can be applied. Double
layers that bend under applied voltage have been produced, stapling technolo-
gies can be used to achieve higher elongation, and so on.
By nature ceramics are isotropic and do not become piezoelectric without a
specific polarization process, which means applying a high voltage at elevated
temperature. On the one hand this causes some restrictions, but on the other
offers also novel opportunities. The final polarization need not necessarily be
homogeneous throughout the whole device. A polarization process in situ can
be adapted to all final requirements.
We recognize immediately that piezoelectric materials engineering is not
restricted to the material itself; it must comprise the whole device or even the
system in which it is used. Thus it is advisable that the designer and the end-
user also have a certain insight into the basic principles of this engineering.
For this reason, the conclusions of this paragraph do not intend to discuss
the materials engineering of piezoelectric devices in detail for any specific
application. This will be done in the chapters to follow. They intend to offer
general rules that are helpful not only for the producer but also for the user,
giving an insight into possibilities, interdependencies, and limitations. They
should promote a fruitful cooperation.
The example of a specific application could be helpful to elucidate the
problems to be solved within such a cooperative engineering procedure. As
already mentioned, concessions have to be made, e.g., as to the temperature
dependence. Nevertheless, it has been possible to take advantage of these
highly piezoelectric materials, e.g., for channel filters in telecommunication.
The characteristic of these high-quality filters is determined by specifically
shaped metal resonators; nevertheless, the piezoelectric transducers are not
allowed to affect this high quality remarkably by the temperature dependence
of their piezoelectric, dielectric, or elastic properties or even by aging. Thus
extrinsic piezoelectric effects must not occur. Any neighbourhood of the Curie
temperature by warming cannot be tolerated, but for high enough energy
conversion the neighbourhood of an MPE is mandatory. Any deviation from
this boundary due to temperature change must be compensated by other
parts of the filter device, but on the contrary, temperature coefficients in the
metallic parts or in the address circuitry can be compensated by the material
choice of the piezoelectric transducer.
For more detailed information we would like to refer to Chap. 7, which
is devoted to this topic. We have only used this example for demonstrating
that in the piezoelectric case materials engineering depends on the design of
3 The Role of Ferroelectricity for Piezoelectric Materials 85
the specific application by far more than in many other fields. It is a must
that the materials engineer understands the requirements of the user. On the
other hand, these remarks should motivate the system designer to understand
more about the possibilities, interdependencies, and also limitations of piezo-
electricity. Only this way can he become an active partner of the materials
engineer.
Concerning such a cooperation, we should not forget that production costs
are a major factor of the figure of merit for applicability. Here the simple
structure and compactness of ceramic piezoelectric transducers offer many
interesting advantages as compared to conventional electromechanical devices.
Returning to materials engineering itself, we have seen that above all,
surroundings of ferroelectric phase boundaries are of interest; but there not
only large piezoelectric effects can be found, but external parameters like
stress or temperature have an increasing influence as well. Therefore, it is of
major technical interest that the phase boundary in question is as vertical as
possible referring to the composition–temperature phase diagram. Only then
the critical distance to the boundary does not change remarkably within an
acceptable temperature range. This is the case in the PZT phase diagram and
explains the extraordinary technical relevance of PZT materials.
Precursors of every ferroelectric phase transition are lattice instabilities
which can be recognized by a corresponding soft mode or as a Curie–Weiss-like
increase of permittivity. Consequently, the transition from one ferroelectric
phase to another is accompanied by an increase of the permittivity perpen-
dicular to the existing spontaneous polarization, whereas only minor changes
are found parallel to it. The resulting pronounced dielectric anisotropy ex-
plains the specific intrinsic anomalies near the phase boundary. It is directly
related to the maxima of the piezoelectric and elastic shear moduli, by which
especially inclined domains can contribute to the intrinsic longitudinal and
transverse piezoelectric effect. This underlines the importance of domain en-
gineering, not only in ceramics.
Obviously, the extrinsic contributions to piezoelectricity are also influenced
by phase transitions. Owing to the small lattice distortions in the neighbour-
hood of phase transitions, 90◦ walls in the tetragonal phase (correspondingly
72◦ or 108◦ walls in the rhombohedral phase) can be moved easily by an exter-
nal field. Only in such a class of materials can major extrinsic contributions be
expected. Interestingly, these contributions become higher the less the crystal-
lite orientation favors the intrinsic contributions and vice versa. Domain wall
movements in a small field are reversible. They become more irreversible with
higher external signals. The nonlinear behavior and hysteresis are connected
with it.
It is possible to pin the domain walls by specific modifications of the mate-
rial and to suppress the movement of 90◦ walls (cf. Chap. 4) omitting mostly
extrinsic contributions and reducing dielectric and elastic losses remarkably.
This is important not only in the already mentioned case of filters but also
for every power application.
86 W. Wersing et al.
References
1. W. Wersing, in Electronic Ceramics, ed. by B.C.H. Steele (Elsevier, London,
New York, 1991)
2. R. Comes, M. Lambert, A. Guinier, Solid State Commun. 6, 715 (1968)
3. W.J. Merz, Phys. Rev. 76, 1221 (1949)
4. A.F. Devonshire, Phil. Mag. 40, 1040 (1949); Phil. Mag. 42, 1065 (1951)
5. V. Dvorak, Phys. Stat. Sol. 45, 147 (1971)
6. W. Wersing, Ferroelectrics 54, 207 (1984)
7. W. Wersing, Ferroelectrics 7, 163 (1973)
8. Y. Saito, H. Takao, T. Tani, T. Nonoyama, K. Takatori, T. Homma, T. Nagaya,
M. Nakamura, Nature 432, 84 (2004)
9. W. Cochran, Phys. Rev. Lett. 3, 412 (1959)
10. H. Thomann, Ferroelectrics 73, 183 (1987)
11. M.J. Haun, Z.Q. Zhuang, E. Furman, S.J. Jang, L.E. Cross, Ferroelectrics 99,
45 (1989)
12. A. von Hippel, R.G. Breckenridge, F.G. Chesley, L. Tisza, Ind. Eng. Chem. 38,
1097 (1946)
13. R.B. Gray, U.S. Patent 2, 486, 560, (Nov. 1, 1949); filed (Sept. 20, 1946)
14. B. Jaffe, R.S. Roth, S. Marzullo, J. Appl. Phys. 25, 809 (1954)
15. D. Damjanovic, M. Demartin, J. Phys. Condens. Matter 9, 4943 (1997)
16. B. Noheda, D.E. Cox, G. Shirane, J.A. Gonzalo, L.E. Cross, S.-E. Park, Appl.
Phys. Lett. 74, 2059 (1999)
17. W. Wersing, G. Zorn, K. Lubitz, J. Mohaupt, Jpn. J. Appl. Phys. 24(Suppl.
24–2), 724 (1985)
18. W. Wersing, Ferroelectrics 37, 611 (1981)
19. K. Carl, K.H. Härdtl, Phys. Stat. Sol. (a) 8, 87 (1971)
20. D. Damjanovic, J. Am. Ceram. Soc. 88, 2663 (2005)
21. L. Bellaiche, A. Garcia, D. Vanderbilt, Phys. Rev. B 64, 060103 (2001)
22. B. Jaffe, W.R. Cook, H. Jaffe, Piezoelectric Ceramics (Academic, London, New
York, 1971)
23. Y. Fetiveau, Thesis, University Lyon, 1969
24. G. Burns, B.A. Scott, Phys. Rev. Letters 25, 1191 (1970)
25. W. Wersing, Proc. 4th Int. Meeting Electro- and Magnetoceramics, Celle, 162–
182 (1981)
26. D.A.G. Bruggemann, Ann. Phys. (Germany) 24, 636 (1935)
3 The Role of Ferroelectricity for Piezoelectric Materials 87
P,C/m2
0.6
0.4
Pr
0.2
0.0
–0.2
–0.4
–0.6
–Ec Ec
–3 –2 –1 0 1 2 3
field strength, kV/mm
as polar. When switching off the electric field after this poling procedure, a
metastable state of domain configurations remains, resulting in a finite rema-
nent polarization Pr .
The dependence of the polarization on the magnitude and direction of
the electric field is described by the ferroelectric hysteresis loop. The specific
processes that determine the shape of the hysteresis loop and the character-
istics such as Pr or the electrical coercive field strength Ec are complicated,
and their nature depends on the material parameters. The shape of a typical
hysteresis loop of a ceramic is shown in Fig. 4.1.
In the remanent state, the polarization Pr is the average dipole moment
of the domains. Owing to the shift and decrease of the number of the domain
walls and the statistical orientation of the domains, a maximum Pr = 0.831 Ps
for the tetragonal phase and Pr = 0.861 Ps for rhombohedral phase (Ps is the
spontaneous polarization of the domains) can be achieved [5]. The polarization
direction of the individual domains is limited to the minor angles Θ around the
direction of the polarization field. The stability of the remanent polarization
depends on the domain configuration. The anisotropy of the physical constants
will also be influenced by the mobility of the non-180◦ domains.
For domain reorientation, an external poling field must be applied on the
sample for a certain duration. Theoretically, most of the domains should be
reoriented when the electric poling field is larger than the coercive field of
the ferroelectric sample. However, a practical poling field must exceed 3 or 4
times the coercive field before maximum piezoelectricity can be obtained. For
a given poling field and poling time, better domain rearrangement results at
higher poling temperatures. It must be pointed out that both the coercive field
and the saturation field decrease when the temperature increases. This is due
92 G. Helke and K. Lubitz
One of the most valuable features of ferroelectricity is the fact that ferro-
electric ceramics can be transformed into polar materials by applying a d.c.
electric field. The reorientation of most domains persists also after switching
off the d.c. electric field, and the existence of the remanent polarized state is
a necessary condition for the piezoelectricity of ferroelectric ceramics.
Ab initio computations of the effective properties of the piezoceramics are
very much limited because of the complicated domain structure.
A basic phenomenological approach (see also Sect. 3.1.4) leads to the fol-
lowing relationship using the tensor form, and adding subscripts to distinguish
induced polarization P3 from remanent polarization Pr
where Q33 ∗ is the electrostriction constant for poled ceramics. Here small-
signal parameters and quasistatic conditions without hysteresis and domain
effects are used.
The term Q33 ∗ Pr 2 represents the remanent strain S3 pr of a polarized
ceramic. The term 2 Q33 ∗ Pr P3 corresponds to the piezoelectric effect as
linearized electrostriction. Q33 ∗ P3 2 describes the electrostriction caused by
the induced polarization, which can be neglected here.
The term Q33 ∗ Pr 2 is the strain resulting from poling. The second term
2Q33 ∗ Pr P3 is the piezoelectric response. Following the conventional practice,
the piezoelectric voltage coefficient is defined via g33 = 2Q33 ∗ Pr , neglecting
higher-order terms. Combining the above equations shows that the piezoelec-
tric charge coefficient d33 is directly related to the remanent polarization Pr
and the dielectric coefficient εT 33 by
E,P Pr
Fig. 4.2. Strain of ferroelectric ceramics (piezoceramics) caused by the field and
remanent strain
The term “hysteresis loop” of ferroelectrics usually means the P (E) depen-
dence. However, the large-signal physical quantities vary with electric field
strength, polarization, and domain reorientation. From the practical point of
view, the most important features are the field-dependent piezoelectric coef-
ficients and, especially, the strain. The extrinsic contributions are described
by effective piezoelectric properties, which consider domain processes and re-
orientations additionally [6, 7].
A simple illustration of the poling process is shown in Fig. 4.2 [1]. The
unpoled sample (left) is elongated by the electric field and shows remanent
strain after field removal.
Non-180◦ wall motion, that is motion of 90◦ walls in the tetragonal phase
and motion of 71◦ and 109◦ walls in the rhombohedral phases, will give rise
to shape changes [6, 7].
In Fig. 4.3, the strain is described both for poling and the repeated electric
field change depicted by the so-called butterfly loop. During repeated field
change without change of direction, the strain shows the behavior represented
in the section Sm .
The coercive field strength Ec is defined by the point P = 0 (see Fig. 4.1).
It should be mentioned that the minimum of the strain does not coincide with
P = 0.
S3’ ‰
6
4 Sm
2
Sr
1
0
–Ec Ec
–3 –2 –1 0 1 2 3
field strength, kV/mm
Fig. 4.3. Strain field strength characteristic of ferroelectric ceramics with perovskite
structure
cubic
400
1 1
350
300
Curie temperature, C
2
250
P5
200
tetragonal
150
100 P5
50
rhombohedral
0
0.48 0.50 0.52 0.54 0.56 0.58
x PbZrO3
Fig. 4.4. Phase transitions of lead zirconate titanate ceramics at the polymorphic
(1) and at the morphotropic (2) phase boundary. Note the extension of the abscissa
4.14
4.12
lattice parameter, A
CT
4.10
4.08 aR
4.06
4.04 aT
4.02
0.48 0.50 0.52 0.54 0.56 0.58
x PbZrO3
Fig. 4.5. Lattice constants of a continuous series of solid solutions of the system
PZT in the neighborhood of the morphotropic phase transition
0.020
0.016
δ = (2/3(c/a–1)/cos α
0.012
0.008
0.004
0.002
0.48 0.50 0.52 0.54 0.56 0.58
x PbZrO3
Fig. 4.6. Homogeneous strain parameter δ within the range of morphotropic phase
transition
Fig. 4.7. Domain configuration within the range of the morphotropic phase
transition
Thus the homogeneous strain parameter is a good additional means for un-
derstanding the properties of ferroelectric ceramics, which depend in a com-
plex manner on composition, fabrication, grain size, and poling conditions.
There are many causes for these grain size effects. One of them is the increase
of the surface-to-volume ratio so that the grain boundaries play a substantial
role when determining the material properties.
According to Fig. 4.8, the dielectric permittivity ε/ε0 before the poling (1)
passes through a smooth maximum within the range of the morphotropic
phase transition. After poling, ε33 T /ε0 (2) reaches a more distinct maximum,
which appears in the more tetragonal section of the morphotropic phase tran-
sition region, where its value increases. This somewhat surprising result can
be explained using the above-mentioned features: owing to the minor energy
difference, the morphotropic phase transition in PZT is very soft resulting in
a broad region of coexistence of both phases changing from grain to grain
4 Piezoelectric PZT Ceramics 99
ε33T/ε0
ε∗/ε0 ε33 T/ε0 ε33T = ε∗
2200
2
1950
1700
1450 1
1200
950
700
0.536
0.510 0.520 0.530 0.540 0.550
x PbZrO3
Fig. 4.8. Dielectric permittivity (1) before and (2) after the polarization
In the tetragonal (titanium-rich) phase of PZT ceramics, both 180◦ and 90◦
domains exist. The 180◦ domains are switched nearly completely by an exter-
nal field and are sufficiently stable after poling. In the rhombohedral phase,
180◦ , 71◦ , and 109◦ domain walls prevail, among which only the 180◦ domains
remain stable, while reorientations of the 71◦ and 109◦ domains are unstable
because of internal stress.
100 G. Helke and K. Lubitz
for the rhombohedral phase (e.g., Pr ∼ = 43.3 µC cm−2 for η = 0.6 and
δ = 0.006).
It must be noted that the ratio of the mean number of domain reorien-
tations in the rhombohedral and tetragonal phases ηR and ηT is equal to
the ratio of the reciprocal values of the strain of the appropriate phases:
ηR /ηT = 0.6/0.2 = 3 ≈ δT /δR [16, 28]. This ratio is consistent with the
measured mobility of non-180◦ domains for a rhombohedral and tetragonal
composition of the considered “PZT-SKN” containing x% PbZrO3 : in the po-
larized state of a rhombohedral composition (x = 56%) 22–25% domains were
switchable, in the tetragonal composition (x = 51%) only 7–8% [18, 22].
In the rhombohedral phase, both the number of the reorientations of non-
180◦ domains and the remanent polarization reach maximum values wherever
δ is minimum. The coercive field Ec , however, decreases continuously with
the transition from a more tetragonal to a more rhombohedral structure. It
reaches a slight minimum on the rhombohedral side (Fig. 4.9).
From Figs 4.8 and 4.9 it follows that the position of the maxima of the
dielectric permittivity ε33 T /ε0 and of the remanent polarization Pr does not
correspond within a composition series; the maxima rather lie near the tetrag-
onal and/or rhombohedral edge of the morphotropic transition range.
Coming from the tetragonal side, the lattice distortion (strain δ) decreases
with increasing PbZrO3 content in the MPB region and further increases
slightly in the rhombohedral PbZrO3 -rich region (Fig. 4.6). This suggests that
the domains near the MPB are aligned most easily by poling owing to the small
spontaneous strain (Fig. 4.6).
4 Piezoelectric PZT Ceramics 101
Pr’
Ec, kVmm–1 Cm–2
3.0 0.40
1
2.5 0.35
2.0 0.30
1.5 0.25
2
1.0
0.20
0.5
0.15
0.0 0.10
FT 0.5
0.51 0.52 0.53 0.54 0.55 0.56 0.57
x PbZrO3
Fig. 4.9. Remanent polarization Pr (1) and coercive field Ec (2) in the range of
morphotropic phase transition
Since the phase composition ratio can also be changed by an electric field,
furthermore, a sample within the range of the morphotropic phase transition
is poled more easily because there are more states available to achieve a higher
degree of polarization alignment. Also, the unidirectional strain Sm of poled
lead zirconate titanate ceramics attains a maximum in samples within the
morphotropic phase transition region (Fig. 4.10).
The largest values of the unidirectional strain Sm are observed near the
center of the coexistence range of the phases resulting from the increasing
number of domain reorientations within an electric field.
In summary, the number of reorientations of non-180◦ domains in the
coexistence range is not only connected with the spontaneous strain but also
with induced transformations of the coexisting phases by an electric field
[11, 12, 29, 30], especially when strong d.c. electric fields and/or pressure are
applied.
Figure 4.10 shows the unidirectional strain Sm as function of the PbZrO3
content (Zr/Ti ratio) of lead zirconate titanate solid solutions in the range
of the morphotropic phase transition in comparison to the maximum of the
dielectric permittivity ε33 T /ε0 of the remanent polarized ceramics. The max-
imum of the unidirectional strain Sm is not obtained at the MPB itself (com-
position x = 0.538 PbZr03 ) but with samples in which the effective volume
fraction of tetragonal phase is about 60% (FT 50%, Fig. 4.11) [17]. Corre-
sponding results were obtained also for other PZT solid solution systems [18].
The maximum of the unidirectional strain Sm is found already on the
tetragonal side of the coexistence range and corresponds to spontaneous lattice
102 G. Helke and K. Lubitz
ε33T/ε0,max
ε33T/ε0 Sm,max Sm,10–3
2200 1,8
1950 2 1,6
1700 1,4
1450 1,2
1200 1,0
1
950 0,8
700 0,6
0.528 0.535
0.510 0.520 0.530 0.540 0.550
x PbZrO3
Fig. 4.10. Dielectric permittivity ε33 T /ε0 (1) and unidirectional strain Sm (2) at a
maximum field strength of 2 kV/mm
ε33T/ε0,max
Sm,max
FT’ FT 0.5
Vol.%
100
80
71
60
50
40
20
0
0.51 0.52 0.53 0.54 0.55 0.56 0.57
x PbZrO3
Fig. 4.11. Concentration of the tetragonal phase within the range of the mor-
photropic phase transition
The maximum of ε33 T /ε0 is shifted into the more tetragonal section by
about 1 mol% corresponding to x = 0.528 PbZrO3 . According to Fig. 4.11,
this corresponds to FT ≈ 71%.
The dependence in Fig. 4.11 was measured by X-ray structure analysis for
defining the fraction of tetragonal phase in the coexistence field [18, 31, 32].
This typical behavior was found also by X-ray measurements of other similar
lead zirconate titanate compositions [12, 22, 31].
When comparing the two most important ferroelectric characteristics – ε33 T/ε0
and Pr – on one hand with the substantial structure parameter δ and on the
other with the configuration and mobility η of the domains, we can get an
overall insight into the performance of the piezoelectric d and g coefficients
and of the coupling coefficient kp :
- Simple 180◦ domain wall motions do not contribute to the piezoelectric
response d [6]. The behavior of the ε33 T /ε0 (Fig. 4.8) and Pr (Fig. 4.9) can
be used to describe further piezoelectric characteristics (Fig. 4.12).
- The piezoelectric characteristics k and d correlate with the remanent po-
larization and the dielectric permittivities using the (4.5) and (13.22) by
the following relationships [15, 16]:
d ∼ εT Pr , (4.10)
k ∼ (ε )
T 1/2
Pr . (4.11)
Fig. 4.12. Comparison of the dielectric permittivity ε33 T /ε0 (1), the piezoelectric
coefficient d33 (2), the electromechanical coupling coefficient kp (3), and the large-
signal unidirectional strain Sm (4) in the range of the morphotropic phase transition
g33, Pr’
10–3Vm N–1 Cm–2
50 0.425
45 0.400
40 0.375
35 1 0.350
30 0.325
2
25 0.300
20 0.275
0.51 0.52 0.53 0.54 0.55 0.56 0.57
x PbZrO3
Material δ = 2/3(c/a − 1) TC ,◦ C
PZT–NiSb 44 0.0102 263
PZT–NiSb 42 0.0133 268
PZT–SKN 530 0.0149 348
PZT–SKN 525 0.0152 351
PZT–SKN 520 0.0154 352
PZT–SKN 515 0.0160 353
PZT–SKN 510 0.0161 354
PZT 52 0.0187 386
PZT–NiSb 44: Pb[(Zr0.44 Ti0.56 )0.92 (Ni1/3 Sb2/3 )0.08 ]O3 ,
PZT–SKN 530: [Pb0.975 Sr0.02 K0.005 ][(Zr0.53 Ti0.47 )0.985 Nb0.015 ]O3 .
PZT 52: Pb(Zr0.52 Ti0.48 )O3 .
106 G. Helke and K. Lubitz
–1.2 0 0
0.515 FT 0.5 0.545 FT 0.5
0.51 0.52 0.53 0.54 0.55 0.56 0.57 0.510 0.520 0.530 0.540 0.550
x PbZrO3 x PbZrO3
cε,,10–2 cd,,10–2
0 0
–1.0 –0.4
2
1
–2.0 –0.8
–3.0 –1.2
–4.0 –1.6
–5.0 –2.0
0.51 0.52 0.53 0.54 0.55
x PbZrO3
Fig. 4.15. Aging behaviour of the dielectric permittivity εT /εo (1) and the piezo-
electric coefficients d33 (2) within the range of the morphotropic phase transition.
(1000 h after poling)
Within the field of piezoceramics, the by far most used and widespread group is
that derived from PZT. Not the least, this is due to the fact that the properties
of PZT can be adapted to a broad spectrum of different requirements by minor
amounts of oxides. In this context, the competition between the two phases
(tetragonal and rhombohedral) plays a major role.
Many results on the impact of these additives have been found empirically.
Thus, it is not easy to find the relevant correlation of the different factors
such as grain size, lattice defects, domain structure, or actual incorporation
of the additives. Nevertheless, we try to offer a systematic survey on the data
available referring to the basic considerations of Chap. 3.
Research and development of such additives started by the use of iso-
valent substitutions of Pb2+ or Zr4+ respectively by Ti4+ on A- and B-sites.
Then heterovalent substitutions were included as well. Nowadays very complex
ternary and quaternary systems are used including specific dopants.
When discussing these systems, we will follow mainly their historical
development.
The first additives used for the modification of PZT were the isovalent ones
[10]. Possible modifications of lead zirconate titanate ceramics are composi-
tions with isovalent substitutions of the cations Pb2+ by Ba2+ , Ca2+ , and/or
Sr2+ on A-sites and with substitutions of Ti4+ and Zr4+ by Sn4+ on B-sites.
This leads to solid solutions of
The modifying elements are added as oxides or carbonates into the com-
position during mixing the components. Usually, the concentration of the
modifiers is varied from 1 to 5 at. %. The effects of isovalent modifiers are
classified according to the individual characteristics of the modifiers: radii of
the atoms, electron configuration, and nature of chemical bonding.
The first examples for modification of PZT by substitution of isovalent
ions have been the composition series Pb1−x Srx (Zr0.53 Ti0.47 )O3 [39]. The sub-
stitution of strontium ions at lead positions results in an increase of the
room-temperature dielectric permittivities ε33 T /ε0 , a decrease of the Curie
temperature TC , and a maximum electromechanical coupling coefficient kp
for instance with x = 0.10 (Table 4.2). That maximum correlates to the MPB
shifted by the Sr doping. Nevertheless, the piezoelectric parameters are not
high enough for any application.
4 Piezoelectric PZT Ceramics 109
x TC , ◦ C ε33 T /ε0 kp
0 385 544 0.48
0.01 — 584 0.49
0.05 360 1,002 0.50
0.10 328 1,129 0.59
0.125 290 1,237 0.47
0.15 265 1,260 0.43
0.20 242 1,257 0.34
The application of piezoceramics outside of the PZT binary system and doped
with heterovalent ions has been favored for a long time because of their ex-
cellent physical characteristics [1].
Dopants are usually added in concentrations 3 at. %. Actually, two dif-
ferent groups of ions have been identified for doping: ions with higher valencies
as donors, and ions with lower valencies as acceptors (Table 4.3). Donors and
acceptors influence the PZT ceramics antagonistically and are called softeners
or hardeners, respectively. The antagonism can be understood considering the
fact that vacancies in general are the mediators.
The violation of neutrality by the substitution of ions with deviating
valencies is compensated in PZT by creation of charged lead or oxygen
vacancies.
Dopings by 3-valent ions at A2+ -positions replacing Pb2+ resp. 5-valent
or even 6-valent ions at B4+ -positions replacing (Zr4+ 4+
x + Ti1−x ) act as donors
2−
and form double negatively charged (VA )-lead vacancies for compensation
[40, 41].
110 G. Helke and K. Lubitz
Primarily donors and acceptors are antagonists and therefore can cancel this
influence mutually. When the charge concentrations of donors and acceptors in
the perovskite structure are equal, the effects can cancel out [46,47]. A coupled
substitution of heterovalent ions can be used for controlling the properties.
Thus, for example, it will be possible to compensate the charge deficits caused
4 Piezoelectric PZT Ceramics 111
with xAi , xBi the fractions of the ions and nAi , nBi the valencies of the ions.
The stoichiometry condition gives
m
n
xAi = 1, xBi = 1. (4.14)
i=1 i=1
The above conditions can be fulfilled in the case of cation valencies 1, 2, and
3 on the A-site and 1, 2, 3, 5, 6 on the B-site. Since RO is unchanged, the
following conditions have been found experimentally in practice [52]:
RA 0, 90 × 10−10 m
0, 51 × 10−10 m < RB < 1, 1 × 10−10 m
RA RB .
To date, more than 1,000 compositions of complex compounds have been
identified [16, 45, 54] that fulfill the above criteria.
However, the role of vacancies remains an unsolved problem. Also, no
unanimous opinion exists concerning the resulting stoichiometry in case of a
simultaneous acceptor and donor doping. Here it should be noted that the con-
centration and impact of vacancies cannot be determined by the initial chemi-
cal composition alone. Evaporation of Pb and valency changes during sintering
are mostly not known. Of course, the neutrality condition must be fulfilled and
is helpful for this problem. A consistent terminology is still lacking concerning
definitions such as structure with vacancies, cation–anion complexes, defect
structure, non-stoichiometric composition, substructure, and second phases.
Clear definitions would be helpful.
114 G. Helke and K. Lubitz
Table 4.6. Composition of complex compounds [1, 10, 54, 55, 62]
As can be seen from both Tables 4.7 and 4.8, specific materials are gener-
ally available and partially already standardized. Nevertheless, specific adap-
tations to special requirements are often necessary. They should remain within
the given framework and obey the general rules given above.
4 Piezoelectric PZT Ceramics 117
25 20 h
20
15
10
5
0
0 1 2 3 4 5 %
Nd concentration
Fig. 4.16. Grain size as a function of Nd2 O3 concentration for
Pb1.00 Ndx Zr0·53 Ti0.47 O3 at 1280◦ C sintering temperature and different sintering
times
be explained by the assumption that only below this concentration the exist-
ing natural impurities of the raw materials are homogeneously incorporated
on A- and B-sites together with the Nd3+ on the A-sites. Thus they ensure
electro-neutrality by their different valencies without formation of additional
vacancies.
A further example is undoped lead titanate (PT). This material sinters at
700◦ C already forming large grains, but cracks into a powder of single grains
during cooling down because of the large tetragonal anisotropy.
A third example is shown in Fig. 4.17. Extreme grain growth at reduced
sintering temperature can be obtained even in complex doping systems if
the doping concentration is varied empirically until defect equilibrium and
stoichiometric incorporation of all offered cations at sintering temperature
is achieved [71, 72]. In general, this doping concentration is different from
the so-called stoichiometric hypothesis, which in the present case would be
1.0 at.% Mn2+ on B-sites and 2 at.% Nd3+ on A-sites. It must be assumed
that temperature-dependent defect equilibria exist where valency changes of
the cations (for example Mn3+ or Mn4+ ), incorporation of even smaller ions on
A- and B-sites, and the formation of Pb and O vacancies have to be included.
Such a temperature shift of equilibrium may be the reason for the minor
temperature shift of the maxima as found with sintering temperature (cf. the
example of Fig. 4.17).
In Fig. 4.18, the grain size shows a clear maximum for an Fe addition
of 1.5 at.% independent of the sintering temperature. For a compensation of
2 at.% Nd3+ on A-site using Fe2 O3 as a dopant, the maximum expected is
2 at.% Fe. The fact that the point of stoichiometric incorporation and max-
imum grain growth lies at 1.5 at.% Fe shows that Fe is incorporated with
120 G. Helke and K. Lubitz
35
1250C
30 1100C
25
Grain size (µm)
20
15
10
1.5
tg δafter (%)
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0
Mn (atom%)
Fig. 4.17. Grain size and loss angle as a function of Mn concentration for the com-
position Pb1.00 Nd0.02 (Zr0.545 Ti0.455 )1−y Mny O3 for two sintering temperatures [71]
valencies 2+ and 3+. It also suggests that depending on doping the grain
growth inhibition in PZT is caused by the formation of A-site or oxygen va-
cancies and by the diffusion processes by which Pb and/or oxygen leave the
sample.
Further results of the doping variations with Mn and Fe in Figs. 4.17 and 18
show another important behavior. Both dopings are used for the fabrication
of hard PZT (see Chaps. 7, 10 and 11). It is generally agreed that the effect
of these dopings is based on the fixing of the domain structure and on the
reduction of the domain wall mobility. Till now, two different reasons have
been discussed [73, 74]. As a bulk effect, complexes of dopants or vacancies
within the domains or within the domain walls may be oriented and stabilize
the polarization. Charged and polarized layers at the grain boundaries may
have the same effect.
In Fig. 4.19 the mechanical quality factor Q is shown as a function of the
Mn and Fe doping. With Fe doping up to 1.5 at.%, the ceramic shows all
characteristics of a soft PZT. For higher doping concentrations, Fe enrich-
ment at the grain boundaries and a steep Q increase can be found, indicating
a stabilization by internal electric fields. In case of Mn, the quality factor
4 Piezoelectric PZT Ceramics 121
mm
25
20
Grain size
15
10
0
0 0.5 1 1.5 2 2.5%
Fe content
Fig. 4.18. Grain size as a function of Fe concentration for the composition
Pb1.00 Nd0.02 (Zr0.545 Ti0.455 )1−y Fey O3 for two sintering temperatures [71]
1800
1600
1400
Mn
Resonance Q
1200
1000
800
600
Fe
400
200
0
0 0.5 1 1.5 2 2.5 3%
Mn, Fe molar concentration
Fig. 4.19. Mechanical quality factor versus Mn or Fe doping
122 G. Helke and K. Lubitz
Fig. 4.20. Relative permittivity and coupling factor as a function of Zr content [79]
For getting optimum piezoelectric values in the range of the MPB, in the past
a large variety of fabrication methods were examined in order to reach high
homogeneity and reproducibility within tolerable costs. The following meth-
ods should be mentioned especially: the chemical precipitation of all or some
special cations, hydrothermal methods, spray-drying with simultaneous pre-
cipitation from a solution, and the sol–gel method. Seldom, however, within
this investigation was the improvement of the homogeneity verified by the
narrowing the MPB range or by microanalytical methods (see Sect. 13.1). Be-
sides poor reproducibility and, particularly, drastically increased production
costs, the typical PZT problems of second phase formation, incomplete incor-
poration of cations during grain growth, and difficult grain size control have
not yet been solved satisfactorily.
Therefore, for the most applications up to now the above-described mixed-
oxide method is used. It is a compromise between low costs, high repro-
ducibility, and good, but not best, piezoelectric data. Consciously the doping
combinations are adjusted to compromise between moderate grain growth
inhibition and a broader range of MPB. At sintering temperatures between
1, 200–1, 300◦ C, relative densities >96% and typical mean grain size of 4 µm
are obtained. At larger grain sizes, the mechanical toughness is reduced. This
causes drawbacks for the polarization process because of internal mechani-
cal stress between adjacent grains as well as for machining and for long-term
stability under large-signal operation. At a mean grain size <1 µm, the piezo-
electric activity decreases drastically. Different reasons are given: composi-
tional fluctuations, mechanical clamping of the grains and change of domain
structure, and domain wall mobility [19].
124 G. Helke and K. Lubitz
As dopants, preferentially oxides of such cations are used that are not
volatile during sintering, have a well-defined valence state, and are incorpo-
rated at a well-defined lattice site. As to standard PZT, the large rare earth
ions La3+ and Nd3+ on A-site and Nb5+ on B-site are preferred, while for
hard or power PZT, Fe or Mn on B-sites are preferred additionally as out-
lined above.
Often the mixed-oxide method is modified by using precursors instead of
oxides. Such precursors are, e.g., NiNbO2 or MnNbO2 . They are used for
avoiding second phases or the formation of pyrochlore. An especially posi-
tive influence on the homogeneous Zr/Ti distribution is found when using a
Zrx Ti1−x O2 as precursor. It can be produced by separate calcination of TiO2
and ZrO2 or, for high homogeneity and small particle size, by hydrothermal
precipitation [78,79]. A narrower MPB range and best piezoelectric values are
the result (Fig. 4.20).
The pre-reacted or calcined and milled PZT powder is the starting material
for the shaping of the different components. In this case, most traditional
ceramic technologies can be used.
For all larger components such as disks, blocks, plates or calottes, the
PZT powder is granulated under addition of a suitable binder. For small
quantities this is done in a mortar, but for industrial series by spray drying.
For shaping, uniaxial or isostatic pressure is applied. Typical relative green
densities of 60% are achievable. The next steps are debindering and sintering,
often under controlled O2 and PbO partial pressure. The machining to final
dimensions is done by dicing, grinding, lapping, and polishing. Owing to the
considerable softness of the PZT ceramic, in general diamond tools are used.
For metallization, in the simplest case metal pastes (especially Ag, but also Cu,
Ni) are screen printed and burned in at 400–600◦ C. Higher-quality electrodes
are sputtered and may consist of many different layers depending on the type
of further contacting. An example of a solderable metallization is the triple
sputtering of Cr for adhesion, Pt as diffusion barrier when soldering, and
finally Au because of its excellent wetting during soldering.
For optimum adhesion of the metallization, it is recommended to anneal
the components after machining at 950–1,000◦ C. The removal of contami-
nants, the annealing of mechanical damages, and a certain degree of recrys-
tallization of the surface can be achieved.
For the final polarization, there is no uniform choice of the parameters
such as time, temperature, and electric field strength. It depends strongly on
the type of PZT and on the design of the component. Thin ceramic foils of soft
PZT can be polarized completely in air at 2 kV mm−1 within 1 s. For thicker
components, the necessary voltage comes up to the kilovolt range, where pol-
ing is performed in oil. If instead of oil SF6 gas is used, no further cleaning
is necessary. Because the switching of 90◦ domains is connected with internal
mechanical stress, for bigger components in general the poling temperature is
raised. This way the coercive force is decreased (see Chap. 3) and thermally
activated processes for compensation are promoted. Especially for hard PZT,
4 Piezoelectric PZT Ceramics 125
Deflection,
force
80 µm
Inactive
region
Fig. 4.21. Principle of design of a multilayer stack. The inner AgPd electrodes are
gray, the electrical connections are black, and the cross-sections of inactive regions
are white
increased poling temperature and time are necessary. Also, poling by a re-
duced electric field is possible when heating above the Curie temperature and
cooling down afterwards.
For the fabrication of thin piezoelectric components, wet-shaping is used.
The PZT powder is mixed with the binder and an aqueous or organic sol-
vent. With that slurry, green tapes are manufactured by doctor-blading or
tape casting with thicknesses between 20 and 200 µm [80]. Out of these green
tapes, round parts, for example, are punched out, sprayed with ZrO2 , stacked,
sintered, separated, metallized, and poled. Several millions of low-cost buzzers
are produced annually this way worldwide.
For the fabrication of stack actuators and bimorphs, a multilayer technol-
ogy has been adapted to PZT [81]. It was developed initially for multilayer
capacitors of BaTiO3 . Figure 4.21 shows the design of a multilayer stack ac-
tuator with thin ceramic sheets and internal electrodes. The key advantage of
this design is the reduction of the operational voltage. With values <200 V,
an electrical field strength of 2–3 kV mm−1 can be realized, which is necessary
for the large-signal driving of such stacks. The fabrication of PZT multilayer
components comprises the following steps: tape casting, punching of square,
thin plates, screen printing of two different designs with metal pastes (in gen-
eral AgPd 70/30), stacking of unprinted and printed (alternately design 1 or
2) plates, lamination under pressure at elevated temperature, dicing of single
components, debindering, sintering, machining, and finally metallization in
order to connect all internal electrodes of same polarity.
The application of the multilayer technology on PZT requires the solution
of three specific material problems [82]. The sintering temperature must be
126 G. Helke and K. Lubitz
Pd-Pb-O Pd-Pb-O
x>0 Ag loss > Pd loss
x=0 Ag loss = Pd loss
x<0 Ag loss < Pd loss
Fig. 4.22. Schematic reaction of AgPd inner electrodes and PZT ceramic during
sintering of a multilayer actuator
1,0 1,0
0,9 0,9
0,8 0,8
Ag
Ag content [mass%]
Pd content [mass%]
0,7 Pd 0,7
0,6 0,6
0,5 0,5
0,4 0,4
0,3 0,3
0,2 0,2
0,1 0,1
0,0 0,0
0 10 20 30 40 50 60 70 80 90 100
reduced by about 200◦ C when using AgPd 70/30 as the electrode metal. A
PbO excess normally leaving the sample during sintering may form a Pd–Pb
compound at the ceramic–electrode interface (see Fig. 4.22). In addition, Ag
may diffuse from the electrode inside the ceramic (see Fig. 4.23), where it is
incorporated on A-sites of the perovskite with atomicity +1, changing the
stoichiometry, the grain size, and the position of the MPB with all positive
consequences for the piezoelectric data [84]. As mentioned in Sect. 4.3, during
the past years specific recipes have been developed that fulfil all the require-
ments. They allow, e.g., the mass production of multilayer stack actuators for
diesel injection valves, described in detail in Chap. 13.
4 Piezoelectric PZT Ceramics 127
4.6 Conclusion
References
1. Y. Xu, Ferroelectric Materials and Their Applications (North Holland, New
York, 1991)
2. I.S. Jeludev, Ferroelectricity and Symmetry, Solid State Physics, vol. 26
(Academic Press, New York, 1971)
3. A.J. Moulsen, J.M. Herbert, Materials, Properties, Applications (Chapman &
Hall, London, 1996 reprinted)
4. G. Arlt, Bol. Soc. Esp. Ceram. Vidrio. 34(5/6), 267–271 (1995)
5. N. Uchida, T. Ikeda, Jpn. J. Appl. Phys. 6(9), 1079–1088 (1967)
6. L.E. Cross, In Ferroelectric Ceramics (Birkhäuser Verlag, Basel, 1993)
7. L.E. Cross, Jpn. J. Appl. Phys. 34, 2525–2532 (1995)
8. B. Jaffe, W.R. Cook, H. Jaffe, Piezoelectric Ceramics (Academic Press, London,
1971)
9. T. Ikeda, Fundamentals of Piezoelectricity. (Oxford University Press, Oxford,
1990)
10. Don Berlincourt, J. Acoust. Soc. Am. 91(5), 3034–3040 (1992)
11. V.A. Isupov, Y.u.E. Stolypin, J. Phys. Soc. Jpn. 28(Suppl.), 312 (1970)
12. V.A. Isupov, Ferroelectrics 46, 217–225 (1983)
13. G.M. Konstantinov, M.F. Kupriyanov, V.A. Serduli, A.A. Lebedinskij,
S.S. Zaycev, E.G. Fesenko, J. Tech. Phys. 59(7), 80–84
14. M. Kupriyanov, G. Konstantinov, A. Panich, Ferroelectrics 127, 77–82 (1992)
15. E.G. Fesenko, A.Y.A. Dantsiger, L.A. Reznitchenko, M.F. Kupriyanov,
Ferroelectrics 41, 137–142 (1982)
16. E.G. Fesenko, A.Y.A. Dantsiger, O.N. Razumovskaya, New piezoceramic ma-
terials, in Russian: Novye p’ezoelektricheskie materialy. Univ. Rostov on Don
1983
17. U. Lange, Internal Report Fraunhofer ISC, 07/2000 (2000)
18. M.J. Hoffmann, H. Kungl, Curr. Opin. Solid State Mater. Sci. 8, 52–17 (2004)
19. C.A. Randall, N. Kim, J.P. Kucera, W. Cao, T.R. Shrout, J. Am. Ceram. Soc.
81(5), 677–688 (1998)
20. G. Arlt, Ferroelectrics 104, 217–227 (1990)
21. G. Helke, S. Seifert, S.J. Cho, J. Eur. Ceram. Soc. 19, 1265–1268 (1999)
22. J.T.H. Reszat, Dissertation, (University of Karlsruhe, Germany, 2003)
23. G. Helke, A. Schönecker, P. Obenaus, U. Keitel, L. Seffner, T. Scholehwar,
U. Lange, Paper ISAF (2000)
24. H. Jaffe, Proc. IEEE. 5310, 1372–1386 (1965)
25. D.A. Berlincourt, D.R. Curran, H. Jaffe, In Physical Acoustics, vol. IA, ed. by
W.P. Mason, (Academic Press, New York, 1964)
4 Piezoelectric PZT Ceramics 129
L.E. Cross
5.1 Introduction
It is interesting to note that as with almost all initial advances in the study of
ferroelectric oxides the impetus for studies of relaxor ferroelectrics stemmed
from early work on polycrystalline ceramic systems. In the case of the relaxor
ferroelectrics, the advance originated from studies in Professor Smolensky’s
group at the Ioffe Institute in Leningrad (now St. Petersburg) on perovskite
structure electroceramics of complex composition. Originally classified as fer-
roelectrics with diffuse phase transitions, it slowly became clear that the very
high dielectric maximum being highly dispersive could not mark a classical
ferroelectric phase transition. Now, following studies at Penn State University,
the suggested designation as relaxor ferroelectrics has become internationally
adopted. The name is nicely compact and does highlight two key features,
the combination of massive dielectric relaxation with almost classical lower-
temperature ferroelectric response.
Over the last 20 years there has been an explosive growth of interest in the
topic. Not surprising perhaps, as relaxor studies combine a focus on materials
with properties of very strong practical importance, but underpinned also
with challenging fundamental considerations, a nearly ideal mix for current
applied research funding. The current, very high level of interest is nicely
reflected in a recent excellent review by G. Samara “Ferroelectrics Revisited:
Advances in Materials and Physics” [1], which devotes some 30% of the text
to the topic of relaxor ferroelectrics. In this short chapter, the focus will be
upon the relevance of relaxor studies to the basic understanding of elasto-
dielectric transducing functions in oxide ferroelectrics.
Early realization of the importance of relaxor ferroelectrics for transducing
structures came with the exploration of the large, stable electrostrictive strains
that could be controlled in lead magnesium niobate (PbMg1/3 Nb2/3 O3 )
(PMN): a Smolensky group discovery [2] that has became the prototype
for perovskite-structure relaxors. The utility of this ceramic was highlighted
132 L.E. Cross
Three of the key features of this typical relaxor are illustrated in Fig. 5.1.
In the weak-field dielectric response (Fig. 5.1a), the permittivity reaches a
high peak value ∼20,000 typical for a ferroelectric perovskite near Tc , but
5 Relaxor Ferroelectrics 133
Pb (Mg1/3 Nb 2/3)O3
20
Strong dispersion in ’
10
0.10
0.8
0.6
2 0.00 Strong dispersion
–160 –120 –80 –40 0 40 80 in tan δ
T
Hysteretic Slow decay of ferroelectric character
a b c d e
T increasing
no line
splitting
except
α1α2
Birefringence
0.03
0.02 no birefringence
0.01
Fig. 5.1. Characteristic features of the ferroelectric relaxor, lead magnesium nio-
bate, Pb(Mg1/3 Nb2/3 )O3 (PMN). (a) Dielectric dispersion in PMN as a function
of temperature and frequency. (b) Dielectric hysteresis in PMN as a function of
temperature. (c) Optical and X-ray evidence for the absence of a macroscopic phase
change below the dielectric maximum at Tm
the maximum clearly does not mark a phase change to a ferroelectric form as
the maximum temperature Tm moves to higher temperature with increasing
measuring frequency in a manner typical for a relaxation dielectric. There is
also an associated maximum in tan δ, again typical of relaxor response (not
shown). Behavior at lower temperatures under high alternating fields does,
however, exhibit bonafide hysteretic response, a necessary and sufficient con-
dition for ferroelectricity. However, unlike a normal ferroelectric, the hysteretic
response is not lost suddenly at Tm but decays more slowly into just nonlinear
response (Fig. 5.1b).
The third surprising feature is that to longer-coherence-length probing
radiation, samples even at very low temperature show no evidence of optical
anisotropy or X-ray line splitting, which would be characteristic of the domains
in a normal ferroelectric (Fig. 5.1c). The early explanation that this could be
due to a nanoscale structure of polar regions has proven correct, but only
recently has its origin been correctly diagnosed.
134 L.E. Cross
Fig. 5.2. (a) Refractive index of PMN at optical wave length showing the departure
from linear behavior due to the onset of polar nano regions. (b) Thermal contraction
in PMN showing the change from linearity induced by electrostriction in the polar
nano region
Key forward progress came from the realization that even though in such
√ compensating nanoscale polarizations, macro P = 0, RMS
a structure of
polarization P 2 = 0 so that examination of properties that depend on P 2
such as quadratic electro-optic effects or electrostriction could indicate the
onset of nanoscale polarization fluctuations.
The first clear evidence for the presence of nanoscale polar regions came
from refractive index measurement on PMN by Burns and Dacol [3] (Fig. 5.2a).
Normal, linear decrease of index n with temperature is enhanced beginning at
a temperature TB much above Tm through contribution from the electro-optic
constants g33 and g31 of the form
n3 g33 + 2g31
∆n = − o Pd2
2 3
showing the onset of nanoscale polarization that did not perturb optical
isotropy.
5 Relaxor Ferroelectrics 135
from which
S11 (T ) − α(To − T )
Pd2 =
(Q11 + 2Q12 )
A microscopic picture of the polarization situation is now also emerging
from X-ray [5] and neutron diffraction [6], high-resolution transmission elec-
tron microscopy [7], and neutron pair distribution function analysis [8]. The
consensus appears to be that there are nanopolar regions that begin to appear
at ∼618 K, which grow from 2–3 nm in size near TB to ∼10 nm at 160 K, and
that the volume fraction increases to near 30% at low temperature. It does
appear generally agreed that above Tm the polar nanoregions are dynamic
and that they do probably carry a major fraction of the system polarizabil-
ity so that the dynamics of the nanopolar regions must tie in closely with
the massive dielectric relaxation near Tm . Analysis of the major dispersion
reveals departure from Arrhenius behavior, which can be very satisfactorily
described by the empirical Vogel–Fulcher relation [9] that takes the form:
τ −1 = ω = ωo exp(E/k(T − Tf ))
4.2
Tf = 217.7 K
4.0
I/T max × 103
3.8
3.6
3.4
3.2
100 102 104 106 108
ω (HZ)
Fig. 5.3. The non-Arrhenius character of the dielectric relaxation in PMN. Solid
line is a Vogel–Fulcher fit to the data. E = 0.079 eV, Tf = 217.7 K, ωo = 1012 Hz
136 L.E. Cross
DIELECTRIC CONSTANT
16000 16000
PLZT-6 PLZT-6.5
14000 14000
12000 12000
10000 10000
8000 8000
6000 6000
100 150 200 250 100 150 200 250
TEMPERATURE (C) TEMPERATURE (C)
(a) (b)
DIELECTRIC CONSTANT
DIELECTRIC CONSTANT
16000 12000
14000 PLZT-7 10000 PLZT-8
12000 8000
10000 6000
8000 4000
6000 2000
75 125 175 225 0 50 100 150
TEMPERATURE (C) TEMPERATURE (C)
(c) (d)
25
4.0 20 kbar
103 PLZT 6/65/35
3.6 104 12
20 105 Tm
ε’ x 10–3
3.2 Hz
2.8
2.4
tan δ x102
15
ε’ x 10–3
2.0 8
280 320 360 400 440
T (K)
103 Hz
10
104
1 bar
105 4
5 20 kbar
103 Hz 10 kbar
TF-R 1 bar
0 0
250 300 350 400 450 500 550
T (K)
700
PARAELECTRIC
Tburns
600 KINETICALLY DISORDERED (K.D.)
a) polarization
Tperc
TEMPERATURE (K)
500
F.E. K.D.
ROMB. II a) polarization
400 b) orientation
c) elastic
300 Glassy
a) polarization Tmax
b) orientation
200 K.D.
F.E.
ROMB. I a) elastic
Glassy
100 a) polarization
b) orientation
c) elastic Tf
0
0 2 4 6 8 10 12 14
ATOMIC % La
Fig. 5.7. Phase diagram for x/65/35 PLZT over the range of x from 0 to 14%.
Note that the Burns temperature does not change over the whole composition range
and the divergence of Tmax and Tf at 4% Lanthanum with the onset of relaxor
ferroelectric response
Fig. 5.8. (a) Structure of nanoscale domains revealed in PLZT 8.2/65/35 composi-
tion by dark-field TEM. (b) Modification of the domain pattern by beam charging
showing the induction of larger macrodomain stripes
added La. One of the best demonstrations of NPR in PLZT is the TEM study
by Randall [18] showing for an 8.2:65:35 PLZT composition dark field contrast
from NPR (Fig. 5.8a) and by charging from the beam current the evolution
towards more normal continuous ferroelectric macro-domains (Fig. 5.8b).
Fig. 5.9. (a) Temperature dependence of ε for Sr1−1.5x Bix TiO3 with x ranging
from 0 to 4% at 0.1, 1, 10, 100 and 1,000 kHz. Note the log T scale. (b) Temperature
dependence of ε for similar compositions and temperature range
E
τ = τo exp of a relaxor ferroelectric.
k(T − Tf )
20 000
0
–70 10 90 170
Temperature (C)
Fig. 5.10. Absence of aging in 0.9 PMN:0.1PT over the temperature range −60◦ C
to +160◦ C after 1,000 h of aging at 20◦ C
STRAIN 12x10–4
10
8
0.9PMN-0.1PT.
6 Pb(Mg0.3Nb0.6Ti0.1)O3
d33~1500p C/N
AT 3.7 kV/cm 2 PZT=Pb(Zr0.5Ti0.5)O3
20 10 0 10 20
ELECTRIC FIELD (kV/cm)
15000
14000 1. 1KHZ
2. 10 KHZ
K 3. 1000 KHZ
13000
1
2
3
12000
101 102 103 104
TIME (MINS)
Fig. 5.12. Absence of log linear aging in 0.9 PMN:0.1PT ceramic modified by a
small addition of MnO to introduce the aging
20000
NO aging
3 h aging
10 h
Dielectric constant
100 h
1000 h
10000
0
–70 10 90 170
Temperature (C)
Fig. 5.13. Influence of aging at 20◦ C on the dielectric response from −60 to +160◦ C
in the MnO-modified PMN:0.1PT measured at 100 Hz
aging temperature (Fig. 5.13), and markedly reduces the dispersive component
(Fig. 5.14). Most intriguing is the asymmetry about the aging temperature,
where dispersion is lost on heating up to the temperature of the dielectric
maximum but is recovered rapidly to the original virgin condition on cooling.
Since the micropolar structure is disorder related, it is unlikely that
alternative orientations of the NPR will be energetically identical, so it is
probable that the micropolar vector will spend longer time in deeper min-
ima. In this case, mobile defects will probably rearrange to accommodate
144 L.E. Cross
20000
(I) 100 Hz (II) (III) (IIII)
(a)
1000 Hz
10 000 Hz
100000 Hz
Dielectric constant
10000 100 Hz
1000 Hz
10000 Hz
100000 Hz
0
–70 10 90 170
Temperature (C)
this preference, further stabilizing the orientation and reducing the associated
“hopping” dispersion. In this model, the asymmetry between heating and cool-
ing would appear logical, as on heating the whole family of accessible mobile
NPRs is already stabilized and of reduced dispersion, while on cooling new
unstabilized NPRs will appear to rebuild the dispersion. The model that sug-
gests a “breathing mode” of the NPRs as contributing strongly to dispersion
would depend on the stabilization of an annulus around the NPR by defect
migration, in which case either heating or cooling would move the boundary
region and thus restore dispersion.
Although the suggestions are not definitive, aging does suggest the need for
real caution in studies aimed at exploring the dynamics of the NPR. Surface
studies will always favor static models owing to the huge anisotropy induced
by any real surface. Volume studies must clearly be carried forward on age-
free materials or on freshly quenched samples in which the dispersion has not
been attenuated.
1.2
0.8
0.6
Slrain (%)
0.4
–0.2
–0.4
0 10 20 30 40 50 60 70
E (kV/cm)
Q11 show that the induced strain is more than adequate. More recent calcu-
lation [28] using total strain measured for the tetragonal state after switching
to determine Q11 agrees closely with the observed S3 and S1 values, and the
calculated tilt of Ps from (111) before switchover is similar to that measured
in the tilt of the optical indicatrix under (001)-oriented E-field in a rhombo-
hedral single domain [29]. With support also from ab initio calculation [30],
there is now little doubt that the driving mechanism for high d33 and large S3
under (001)-oriented E-field is rotation of the polarization vector away from
the (111) direction in the rhombohedral domain towards the orientation of
the driving E-field.
The basic explanation is quite clear; however, there is much complexity in
the detail. In a relaxor ferroelectric, a poling operation is required to put the
low-temperature form into the ferroelectric phase with long-range order. The
resulting phase may depend on the magnitude, orientation, and temperature
of the poling operation. The original phase diagram is oversimplified, and in
some compositions close to the MPB there is evidence of stable monoclinic
and orthorhombic forms. Unfortunately, much confusion has also been gener-
ated by overemphasis on the role of the monoclinic phase in these crystals.
5 Relaxor Ferroelectrics 147
It is absolutely true that under high (001) E-field the P vector in the domain
is tilted more than 20◦ away from the rhombohedral (111) and the symmetry
is unquestionably monoclinic, but of critical importance is the fact that the
monoclinic form is metastable so that when the field is removed the domain
reverts back to its equilibrium rhombohedral form giving the recoverable an-
hysteretic quasi-linear actuation strain under the E-field. It is perfectly true
that in the monoclinic state Ps may lie anywhere in the mirror plane between
rhombohedral and tetragonal orientation, but it is absolutely untrue that it
is free to rotate in this plane in that for any given composition, temperature,
and pressure, the vector will always occur at the same orientation in one of
the family of 24 permitted orientation states. What is unusual is that the Ps
vector will change its orientation in the mirror as a function of temperature,
giving an unusual component of pyroelectricity orthogonal to the direction
of Ps in the single domain. Clearly, for the type of actuation achieved in the
(001)-poled rhombohedral form of both PMN:PT and PZN:PT, in rhombo-
hedral compositions close to the MPB the last thing one would want is a
stable intervening monoclinic form between tetragonal and rhombohedral, as
this would necessarily introduce massive hysteresis in the strain because the
Ps vector drops into and has then to be pulled out from the stable tilted
orientation. A key basic unresolved problem in the PMN:PT and PZN:PT
systems is why the tilting mechanism is so astoundingly successful in raising
piezoelectric response. Other simpler perovskites such as BaTiO3 and KNbO3
have ferroelectric/ferroelectric phase boundaries, and at temperature close to
the boundary exhibit appropriate high dielectric anisotropy [31, 32]. For sim-
ple domain-average-engineered structures, however, both measured and cal-
culated properties, although significantly enhanced, are almost an order of
magnitude less than in the relaxor ferroelectrics.
For future development based on the relaxor single crystals, it would ap-
pear that a very important question is how the residual disorder and random
fields soften both the dielectric and elastic properties without destroying the
high-temperature reach and the strong remanent polarization in the poled
ferroelectric forms, as discussed further in Sect. 5.
Table 5.1. Ferroelectric species that could occur from the 4/mmm prototypic point
symmetry in Shuvalov notation. Non-zero components of Ps delineated for each
species
Neither pure barium niobate nor pure strontium niobate will form in the
bronze structure; however, in the solid solution they form 0.25Ba/0.75Sr
to 0.75Ba/0.25Sr stable tungsten-bronze structure forms. The most widely
studied is (Sr0.61 Ba0.39 )Nb2 O6 , which is pseudo-congruently melting so that
excellent-quality, large, and striation-free single crystals can be grown. It has
been shown from X-ray analysis by Abrahams [33] that the large Ba2+ ions
are exclusively in the 15-fold A2 -sites, while the strontium ions are mixed
between A1 and A2 sites and the smaller C site is completely empty.
Massive dispersion in the weak-field dielectric permittivity (Fig. 5.18) im-
mediately suggests relaxor ferroelectric behavior, which is supported by the
Vogel–Fulcher analysis (Fig. 5.19), and there is clear evidence from both op-
tical birefringence and thermal contraction studies of the presence of c-axis
micropolar regions up to a Burns temperature some 200◦ C above the dielectric
maximum at Tm . The uniaxial character of the polar behavior permits a very
simple macroscopic demonstration that the polar nanoregions between Tm
60000
(a)
50000
40000
χ’
30000
20000
10000
0
10000
(b)
8000
6000
χ’’
4000
2000
0
–100 0 100 200
Temperature (C)
Fig. 5.18. Massive dielectric dispersion in real and imaginary components of the
weak-field permittivity Cc as a function of temperature and frequency. Measured
frequencies are 2 × 102 , 4 × 102 , 103 , 2 × 103 , 4 × 103 , 104 , 2 × 104 , 4 × 104 , 105 , 2 × 105 ,
4 × 105 , and 106 Hz. Top curve is at the lowest frequency, the bottom at the highest
150 L.E. Cross
Fig. 5.19. Vogel–Fulcher fitting of dielectric data for the more strongly dispersive
(Sr0.75 Ba0.25 )Nb2 O6
and TB , the Burns temperature, are in fact dynamic. As the enhanced ther-
mal contraction above Tm shows, the onset of nano-polarity contributes an
electrostrictive component to the elastic strain of the form S3 = Q33 P2s(3) ; one
notes also that the local strain S3 does not change with the sign of Ps(3) . Now,
if we measure an effective Q33(e) by the induced strain S3(e) vs. total polariza-
tion charge, i.e., S3(e) = Q33e PT 2 , and an increasing part of PT comes from
the switching of the already polarized and therefore non-strain-generating re-
gions, then Q33e < Q33 and will decrease as the fraction of polarization carried
by the NPR switching increases.
The strain S11 induced by P3 through the constant Q31 will be similarly
affected; however, the polarization P1 cannot be influenced by the nanopolar
P3 , and Q11 should not be effected, and this is indeed the case in Fig. 5.20.
Thus, as expected, the kinetics of nanopolar switching may be responsible
for the large dispersion and interaction between the NPR responsible for the
Vogel–Fulcher slowing down evidenced in relaxor behavior.
It has long been known from the admirable symmetry studies of Aizu [34] and
Shuvalov [35] that monoclinic and triclinic phases were feasible ferroelectric
species from the perovskite m3m prototypic point symmetry. Ceramist un-
fortunately are not avid readers of symmetry literature, but the outstanding
properties and clearly evident complexity in the MPB phase region in the
PMN:PT and the PZN:PT solid solution systems helped to promote a closer
5 Relaxor Ferroelectrics 151
0.03
O33
ELECTROSTRICTIVE
+ O13
0.01
+ + +
+
+
+
0.00
80 100 120 140 160 180 200
TEMPERATURE C
ELECTROSTRICTIVE COEFFICIENT
0.04
0.03
O II x m4/ C2
0.02
0.01
0.00
110 120 130 140 150
TEMPERATURE C
Fig. 5.20. Apparent electrostriction constants Q33 , Q13 , and Q11 measured as the
ratio of total induced strain to the square of total electrical polarization
look at the PZT solid solution which underpins all polycrystalline piezoceram-
ics. Much earlier, it had been suspected by the pioneer development group
under Hans Jaffe at Clevite Corporation that there was a low-temperature
phase transition for the MPB composition [36]. A possible candidate was the
rhombohedral I – rhombohedral II (R3m − R3c) phase transition in the rhom-
bohedral region which might cross the MPB at lower temperature. Structure
studies by Amin [37] initiated to check this hypothesis soon disproved the pos-
sibility, but unfortunately at that time neither samples nor equipment were
adequate for more detailed symmetry studies.
It needed Dr. Noheda from Professor Gonzalo’s ferroelectric group in
Madrid working with Dr. Shirane at Brookhaven to uncover the narrow in-
sert of monoclinic phase that interleaves the tetragonal and rhombohedral
phases at lower temperature at the PZT MPB composition [38] (Fig. 5.21).
Truly, however, how this interleaving monoclinicity in the pure PZT relates
to the exciting family of properties available in the modified commercial PZT
composition will need more careful study. In commercial compositions, the
152 L.E. Cross
500
Tc Pcubic
400
300
Temperature [C] AT
FR (HT)
200 MPB
100
AO FR (LT) FT
0
–100 Fmonoclinic
–200
10 20 30 40 50 60 70 80 90
PZ PbTiO3 [mol.%] PT
Fig. 5.21. Revised phase diagram for the PZT composition field showing the nar-
row insert of monoclinic phase between rhombohedral and tetragonal phases at the
morphotropic phase boundary
useful properties are often a complex ensemble of intrinsic domain strains, ex-
trinsic contribution from ferroelastic-ferroelectric domain wall motions, and
possible phase switching. Recent high-precision X-ray studies [39] suggest that
as in the PMN:PT and PZT:PT systems the major intrinsic contribution from
both tetragonal and rhombohedral grains comes largely through tilting of the
Ps vectors. Since through its 24 permitted orientation states the monoclinic
form may provide bridging polarization continuity, one suspects that this will
permit switching of a wider cone of orientation states as evidenced by the
large Pr values. This flatter cone will then expose more suitably tilted polar
vectors to measuring fields applied in the original poling directions.
Since for the single-crystal systems the model of enhanced piezo activity by
off-axis fields tilting the Ps vector is now very widely accepted and well au-
thenticated in the relaxor–ferroelectric MPB compositions, a natural question
is how general this model is and how it will apply to simple classic proper fer-
roelectrics.
Extensive studies of BaTiO3 [40] and KNbO3 [41], for which excellent
single crystals have been grown, certainly confirm the enhancement of d∗
constants by off-axis fields, but in both these systems the level of d values and
coupling constants that can be achieved falls far short of those in the best
relaxor ferroelectric solid solutions.
5 Relaxor Ferroelectrics 153
References
1. G.A. Samara, in Solid State Physics, vol. 56, ed. by H. Ehrenreich (Academic,
New York
2. G.A. Smolensky, A.I. Agranovskaya, Sov. Phys. Solid State 1429, 1959
3. G. Burns, F.H. Dacol, Solid State Comm. 48, 853, 1993; Phys. Rev. B 28,
2527, 1983
4. L.E. Cross, Ferroelectrics 76, 241, 1987
5. N.D. Mathan, E. Husson, G. Calvarin, J.R. Gavarri, A.W. Hewat, A. Morell,
J. Phys. Cond. Matter 3, 8159, 1991
6. A. Naberezhnov, S. Vakhrushev, B. Dorner, D. Stauch, H. Moudden, Eur. Phys.
J. B 11, 13, 1999
7. D. Viehland, M.C. Kien, Z. Xu, J.F. Li, Appl. Phys. Letters 67, 2471, 1995
8. I.K. Jeong, T.W. Darling, J.K. Lee, Th. Proffen, H. Heffner, J.S. Park,
K.S. Hong, W. Dmowski, T. Egami, Phys. Rev. Lett. 94, 147602, 2005
9. D. Viehland, S.J. Jeong, L.E. Cross, J. Appl. Phys. 68, 2916, 1990
10. D. Viehland, J.F. Li, S.J. Jong, L.E. Cross, Phys. Rev. B 43, 8316, 1991
11. E.V. Colla, N.K. Yuskin, D. Viehland, J. Appl. Phys. 83, 3298, 1998
12. N. Setter, L.E. Cross, J. Appl. Phys. 51, 4356, 1980
13. C.G.F. Stenger, F.L. Scholten, A.J. Burggraaf, Solid State Comm. 32, 989, 1979
14. C.G.F. Stenger, A.J. Burggraaf, Phys. Stat. Solidi 61, 653, 1980
15. F. Cho, J.M. Reaney, N. Setter, J. Am. Ceram. Soc. 78(7), 1947, 1995
16. C.A. Randall, A.S. Bhalla, T.R. Shrout, L.E. Cross, J. Mat. Res. 5(4), 829, 1990
17. T. Egami, Ferroelectrics, 267, 101, 2002
18. C.A. Randall, D.J. Barber, R.W. Whatmore, J. Microscopy 145, 275, 1987
19. J.H. Barrett, Phy. Rev. 86, 118, 1952
20. E. Pytte, Phys. Rev. B 5, 3758
21. S. Swartz, T.R. Shrout, Material Res. Bull. 17, 1245, 1980
22. W. Pan, E. Furman, G.O. Dayton, L.E. Cross, J. Mat. Sci. Lett. 5, 647, 1986
23. S. Nomura, H. Arima, F. Kojima, Jpn. J. Appl. Phys. 12, 531, 1973
24. J. Kuwata, K. Uchino, S. Nomura, Ferroelectrics 37, 579, 1981
25. J. Kuwata, K. Uchino, S. Nomura, Jpn. J. Appl. Phys. 21, 1298, 1982
26. S.E. Park, T.R. Shrout, J. Appl. Phys. 82, 1804, 1997
5 Relaxor Ferroelectrics 155
27. L.E. Cross, P. Hana, Ninth US Japan Seminar on Dielectric and Piezoelectric
Ceramics, Okinawa, November 2, 1.1.15, 1999
28. L.E. Cross, Fundamental Physics of Ferroelectrics, Aspen Colorado, AIP Proc
535, 1, 2000
29. A.A. Bokov, Z.-G. Ye, Ceram. Trans. 136, 37, 2003
30. H. Fu, R.E. Cohen, Nature 403, 281, 2000
31. S. Wada, S. Suzuki, T. Noma, T. Suzuki, M. Osada, M. Kakihana, S.E. Park,
L.E. Cross, T.R. Shrout, Jpn. J. Appl. Phys. Pt 1 38, 5505, 1999
32. N. Nakamura, T. Tokiwa, Y. Kawamura, J. Appl. Phys. 91, 9272, 2002
33. P.B. Jamiesan, S.C. Abarahams, J.L. Bernstein, J. Chem. Phys. 48, 5048, 1968
34. K. Aziu, Phys. Rev. 146, 423, 1966
35. L.A. Shuvalov, J. Phys. Soc. Japan 28, 38, 1970
36. H. Jaffe (private communication)
37. A. Amin, Thesis in Solid State Science, Penn State, November 1979
38. J. Noheda, A. Gonzalo, L.E. Cross, R. Guo, S.E. Park, D.E. Cox, G. Shirane,
Phys. Rev. B 61, 8687, 2000
39. M.J. Hoffman, H. Kungl, J-Th. Reszat, S. Wagner, in Polar Oxides Properties,
Characterization, and Imaging, ch. 7, p. 137, ed. by R. Waser, U. Böttger,
S. Tiedke (Wiley-VCH)
40. D. Damjanovic, F. Brem, N. Setter, Appl. Phys. Lett. 80(4) 652, 2002
41. S. Wada, K. Muraoka, H. Kakemoto, T. Tsurumi, H. Kwmagai, Jpn. J. Appl.
Phys. 40(98), 5690, 2000
42. S. Wada, T. Tsurumi, Br. Ceram. Trans. 103, 93, 2004
43. S. Wada, K. Yaho, K. Yakoo, T. Tsurumi, in Proceedings of 12th US:Japan
Seminar on Dielectric and Piezoelectric Ceramics, Annapolis, MD, November
2005, 177
44. Y.M. Jin, Y.U. Wang, A.G. Khachaturyan, J. Appl. Phys. 94, 3629, 2003
45. D. Viehland, J. Appl. Phys. 88, 4794, 2000
46. Z. Kutnjak, J. Petzelt, R. Blinc, Nature 441, 956, 2006
6
Piezoelectric Polymers and Their Applications
6.1 Introduction
The science and technology of piezoelectric polymers has long been dominated
by ferroelectric polymers from the polyvinylidene fluoride (PVDF) family. The
piezoelectricity in this polymer class arises from the strong molecular dipoles
within the polymer chain and from the resulting change of the dipole den-
sity upon application of a mechanical stimulus. Ferroelectric polymers show
moderate piezoelectric coefficients (d33 around 20–30 pC/N) in comparison to
ceramic piezoelectrics, with an acoustic impedance comparable to that of wa-
ter and other liquids. The thermal stability of the piezoelectric effect is limited
to below 100◦ C [1]. Applications of ferroelectric polymers emerged in many
niches [2], examples are hydrophones [3] and clamp-on transducers [4] used
as pressure sensor for diesel injection lines, with selling numbers over 50 mil-
lion pieces per year, as well as piezoelectric ignition systems [5] for measuring
the mechanical and physical state of matter under shock loading. Printable
polymer ceramic composites [6] and nanocomposites [7] further enlarge the
material basis available to application engineers.
A relatively new development are relaxor ferroelectric polymers, based on
electron-irradiated poly(vinylidene fluoride) trifluoroethylene copolymers [8]
or on terpolymers of vinylidene fluoride, trifluoroethylene, and chlorofluo-
roethylene [9]. Relaxor ferroelectric copolymers exhibit strong electrostric-
tion and thus large piezoelectric coefficients, when used under electric dc-bias
fields.
Around 20 years ago, internally charged cellular polymer electrets were
shown to display large piezoelectric d33 coefficients [10]. Recent work demon-
strated close similarities between cellular polymers and ferroelectric materials,
hence they are now commonly referred to as ferroelectrets [11]. Ferroelectrets
display large intrinsic piezoelectric d33 -coefficients well above 100 pC N−1 ,
very small d31 and d32 coefficients and an acoustic impedance well suited for
coupling to air and other gases. The best characterized ferroelectret, based on
158 S. Bauer and F. Bauer
Direct piezoelectricity was first described by Jaques and Pierre Curie in 1880,
at the time of their discovery being 25 and 21-years old at the beginning
of their scientific careers. The discovery was not made by chance, the Curie
brothers have looked for the piezoelectric phenomenon, possibly inspired by
their familiarity with the phenomenon of pyroelectricity. The early experi-
ments of the Curie brothers dealt only with direct piezoelectricity, so the
introduction of the converse piezoelectric effect is credited to the theoret-
ical work of Lippmann on the basis of general thermodynamic arguments.
However, after becoming aware of the theoretical predictions, converse piezo-
electricity was verified experimentally also by the Curie brothers. The early
6 Piezoelectric Polymers and Their Applications 159
The direct and converse piezoelectric d-tensor is defined by dip = ∂Di /∂Tp =
∂Sp /∂Ei , where D is the displacement, E the electric field, S the mechan-
ical strain, and T the mechanical stress, respectively [30]. For S and T the
contracted notion is employed, with values of p ranging from 0 to 6. This is
a commonly adopted practice to represent the third rank piezoelectric ten-
sor in matrix form. There are in total four possibilities to define piezoelectric
tensors, d, e, g, and h each of them being used for specific purposes, d is the
charge coefficient and g = d/ε the voltage coefficient, where ε is the dielectric
“constant” of the material. For hydrophones, the dg product is often used as
a figure of merit (FOM). Further figures of merit are the acoustic impedance
Z, describing the efficiency of coupling sound waves from the piezoelectric
material to adjacent media and vice versa. For air-borne ultrasound, another
application niche of piezoelectric polymers, a FOM based on energy consid-
erations has been introduced recently [31]. In this FOM, the proportion of
the acoustical output and the total input energy delivered to the material is
considered. In the case of a negligible acoustic impedance of the medium (air
or any other gas) in comparison to the acoustic impedance of the piezoelectric
material, the FOM is f = 104 k33 4
/ZP2 . Here k33 is the piezoelectric thickness
extension coupling factor and ZP is the acoustical impedance of the piezo-
electric material. The factor 104 is only used to make numbers for different
materials reasonable in size.
Polymers belong to the orthorhombic mm2 symmetry group. There-
fore, the piezoelectric tensor of polymers is described by five piezoelectric
6 Piezoelectric Polymers and Their Applications 161
-
k1 ----
+ + + ++ + + +
+ ++++
k2
-
----
k1
+ ++++
+ ++ + + + + +
(a) (b) (c)
Fig. 6.1. Scheme of the chain structure in ferroelectric polymers (a), of the most
simple model for piezoelectricity (b) and of the voided structure of ferroelectrets (c)
(after [32])
Fig. 6.2. Stacks of amorphous and crystal lamellae in polyvinylidene fluoride, and
one poled stack with compensation charges at the interface between crystallites and
amorphous regions
Fig. 6.3. Cellular polypropylene ferroelectret and internally charged voided polymer
(after [11])
Table 6.1. Comparison of piezoelectric polymers with quartz and the ceramic PZT
coefficients, due to the softness of the material and the low dielectric constant
of foams. However, in comparison to PVDF, both the d31 and d32 coefficients,
as well as the corresponding g values are negligible in cellular polymers.
Most remarkable is the very large value for the airborne ultrasound
FOM obtained for cellular polymers (160 with an acoustic impedance of
0.028 MRayl), whereas the values for piezoelectric polymers like PVDF are
much smaller (0.2 with an acoustic impedance of 3.9 MRayl) [31]. Only porous
piezoelectric ceramics may compete with cellular polymers, with a FOM for
airborne ultrasound of 40 and an acoustic impedance of only 6.5 MRayl. Bulk
ceramics have values comparable to that of PVDF (0.4) with a very large
acoustic impedance of 25–37 MRayl [31].
164 S. Bauer and F. Bauer
The low responses in traditional ferroelectric materials, such as the low strain
level (∼0.1% strain) and low sensitivity to external mechanical stimuli that
can be achieved limit the performance of devices and systems made of fer-
roelectric polymers. As has been observed in many ferroelectric materials,
by operating near the instability regions such as ferroelectric-paraelectric
(F-P) transitions, these responses can be significantly enhanced [37]. Unfor-
tunately, phase transitions occur over a narrow temperature range and, in
most cases, involve large hysteresis, preventing the enhanced responses to
be used in applications. It has been found that by introducing defects into
P(VDF-TrFE) copolymers, it is possible to convert the polymer from a normal
ferroelectric to a relaxor ferroelectric [8]. A new class of ferroelectric polymers,
i.e., the terpolymers of P(VDF-TrFE-CFE), were developed from the normal
ferroelectric PVDF polymer by employing proper defect modifications, which
eliminate detrimental effects associated with a normal first order F-P tran-
sition while maintaining high material responses [9, 38]. The introduction of
the third monomer into the polymer chain serves to interrupt the ferroelec-
tric domains, thereby reducing their size [39]. Random defect introduction, as
in the irradiated copolymer samples [8], broadens the ferroelectric transition
and reduces the ferroelectric-paraelectric transition temperature. The ran-
dom incorporation of the bulky third monomer into the polymer chains forces
a conformation change from the all-trans (Tm 4) or conformation to the
transgauche (TG) and T3 G conformations [40]. Addition of defects in the form
of chemical monomers, in this case 1,1-chlorofluoroethylene (CFE), which are
copolymerised with VDF-TrFE, eventually favor the TG conformation and
eliminate the normal ferroelectric phase, leading to a relaxor ferroelectric
with electromechanical strain greater than 7% and an elastic energy den-
sity of 0.7 J cm−3 at 150 MV m−1 [41]. It is the change in conformation from
the disordered TG and T3 G to the all trans conformation that leads to the
observed large electrostrictive strain, shown in Fig. 6.4. This high electrostric-
tive strain coupled with a relatively high elastic modulus (>0.3 GPa) makes
the polymers suitable as a structural component in addition to the electroac-
tive functions (actuators and sensors). The relaxor-ferroelectric terpolymer
P(VDF-TrFE-CFE) exhibits a room temperature dielectric constant greater
than 50 [38, 39]. This value is among the highest reported in literature for
polymeric materials [42]. Relaxor ferroelectric polymers show practically no
piezoelectric activity, but they can display very large field-induced piezoelec-
tricity when biased with a suitable large electric dc-field. From Fig. 6.4, a large
field-induced piezoelectric d33 coefficient around 400 pC N−1 is estimated at
a dc-bias field of 100 Vµm−1 . Relaxor ferroelectric polymers, therefore, may
find applications in certain niches, where large piezoelectric coefficients are
desired and where dc-bias operation is not detrimental.
6 Piezoelectric Polymers and Their Applications 165
S3(%) 0
−1
−2
−3
(a) −4
−5
−6
−7
−8
0 50 100 150 200
E (MV/m)
S3
S1
S2
Fig. 6.4. Strain vs. electric field for a relaxor ferroelectric P(VDF-TrFE-CFE) ter-
polymer. The large strain that can be achieved with relaxor ferroelectric polymers
make these materials interesting for inducing large piezoelectric effects by field bi-
asing (after [41])
Piezoelectric laquers have been developed by the swiss company ALGRA [6],
for example. Their piezoelectric laquers are based on the incorporation of
piezoelectric ceramics into laquer matrices. The piezoelectric laquers can be
printed on many different substrates. After printing, the piezoelectric elements
are poled with a not revealed poling process. No details on the fabrication
process are available in open literature, but the technical data information
provided sound impressive with sensitivities reaching 1 nC N−1 in a typical
element. Piezoelectric laquers are certainly an interesting class of materials,
since they combine the advantageous features of polymers (flexibility and low
brittleness) with the advantages of piezoelectric ceramics (large piezoelectric
coefficients).
Piezoelectric laquers belong to the family of composites. Although this
chapter primarily deals with polymers alone, it is interesting to mention that
by combining piezoelectric ceramics with ferroelectric copolymers, true mul-
tifunctionality can be easily achieved [7]. 0–3 composites of lead titanate
nanoparticles in PVDF-TrFE copolymers have been poled in a sequential
manner to make the material piezoelectric but not pyroelectric, or vice versa.
166 S. Bauer and F. Bauer
A schematic view of the composites with the different orientation of the po-
larization vectors is shown in Fig. 6.5. Compensation of effects is possible,
since the piezoelectric coefficients are different in sign in piezoelectric ceram-
ics (intrinsic piezoelectricity) and ferroelectric polymers (dipole density piezo-
electricity), while the pyroelectric coefficients are both negative in polymers
and ceramics. Applications may include piezoelectric transducers insensitive
to temperature changes and drift, and pyroelectric sensors insensitive to mi-
crophony.
Table 6.1 is also useful in assessing the possible application niches for the dif-
ferent materials. PVDF and its copolymers are excellent materials for use in
hydrophones, because of its nearly perfect match with water and other liquids.
Commercially available are hydrophones made of PVDF used for the calibra-
tion of medical ultrasound equipment [3], examples of such hydrophones are
depicted in Fig. 6.6. PVDF-based hydrophones are sold by many companies
worldwide [43], and provide an excellent niche application of this piezoelectric
polymer. Another success story for the use of PVDF in a large scale commer-
cial application, with production rates of several million pieces per year is a
pressure sensor, developed by the Austrian company AVL in Graz for Diesel
injection lines [4], as illustrated in Fig. 6.7. Also noteworthy are the applica-
tions of PVDF in shock wave sensors, work pioneered by Francois Bauer of
ISL, France, R.A. Graham of Sandia National Laboratories and L.M. Lee of
Ktech Corporation, Albuquerque [5,34] (see also Chap. 6.5 for further details).
The flexibility and the ease of production in any kind of form also allow for un-
usually shaped piezoelectric elements, the piezoelectric cable described in [2]
and shown in Fig. 6.8 being an example. Such cables are extremely rugged and
can stand heavy loads, for example in truck axle counting. Application areas
include physical security, proximity sensing, traffic counting, and transducers
for security fences.
6 Piezoelectric Polymers and Their Applications 167
Fig. 6.7. Pressure sensor for Diesel injection lines. Such pressure sensors have a
multimillion pieces per year market (taken from [4])
Fig. 6.8. Piezoelectric cable elements, in the piezoelectric coaxial cable the piezo-
electric polymer acts as a polarized dielectric. Piezoelectric cables are extremely
rugged and can withstand also heavy loads
The soft cellular polymers may only be used for applications in shallow
water, although their dg product is quite large [11]. The reason is the heavy
loading of the cellular material by the liquid in a water environment. Ferro-
electrets are therefore much more promising for air-borne ultrasound applica-
tions, since their FOM is the highest of all materials available. Applications
168 S. Bauer and F. Bauer
Fig. 6.9. Pulse echo location system based on cellular polypropylene ferroelectrets.
The system is comparable to the ultrasound imaging systems of animals, like bats
(after [44])
Fig. 6.10. Flexible keypad made from cellular polypropylene ferroelectrets. The
keypad is vandal proved (taken from [15])
are available on the lab scale, an example being an ultrasound-based echo lo-
cation system shown in Fig. 6.9 [44]. This echo location system is very similar
to that of animals like bats.
The negligible longitudinal piezoelectric coefficients prevents applications
where large deflections are required, an example is the piezoelectric bimorph.
However, it is an advantage when only the thickness change is of interest,
such as in switches. Applications of cellular polymer films are discussed in
depth in a recent review of Wegener and Bauer [28], and on the homepages
of several small and medium-sized companies in Finland [13–15]. Musical key
pads based on cellular polypropylene are for example currently favored by
many musicians. Figure 6.10 shows an example of a fully flexible, vandal
proved keypad, successfully employed in the market [15].
Most noteworthy is the recent demonstration of a cellular polypropylene
microphone by Hillenbrand and Sessler [45] with a sensitivity approaching that
6 Piezoelectric Polymers and Their Applications 169
20
10
Sensitivity [mv/Pa]
2
Single-film microphone
Five-film microphone
1
10 100 1000 10000
Frequency [Hz]
Fig. 6.11. Frequency response of a single layer and a five layer cellular
polypropylene-based microphone (taken from [45])
Fig. 6.12. Flexible keypads of different geometry, realized with piezoelectric laquer
technology (taken from [6])
Polarization (µC/cm²)
i(t) u(t) 6
0.4
Current (µA)
High u(t) 4
Voltage PVDF 0.2
+/- 10 kV 2
i(t)
0 0
u(t)
-0.2 -2
-4
Function
-0.4
Generator
i(t) -6
-0.6
0.1 up to 0.001 Hz Digital GPIB -8
Voltmeter -0.8 - 10
HP3852a -10 -8 -6 -4 -2 0 2 3 6 8 10
Voltage (kV)
Fig. 6.13. Poling equipment for the cyclic Bauer poling process of PVDF, together
with the evolution of the hysteresis curve. Gauges can be produced with a repro-
ducibility of the material properties better than 2% (after [47])
area films for applications that do not require a very high electrical qual-
ity of the samples. In such a large volume production method, quality and
reproducibility of data is not necessary or is it desired.
For applications and materials studies requiring high quality and repro-
ducibility, individual samples must be produced with well-defined electrical
properties in a process, which controls the amplitude, duration, and the
history of the electric field. By controlling the electrical field applied to in-
dividual samples, the electric displacement, displacement current, remanent
polarization, and homogeneity of polarization can be also controlled [46].
The Piezotech poling equipment, depicted in Fig. 6.13a, allows us to produce
extremely well-defined piezoelectric polymer samples, as shown by the evo-
lution of the hysteresis curves in Fig. 6.13b. With this Bauer cyclic poling
process, reproducible remanent polarizations as large as 9 µC cm−2 are rou-
tinely achieved in a commercial process. Because electric fields as high 5 MV
per cm are utilized, inferior films are eliminated in the process. Each PVDF
gauge, fabricated with this process is characterized with an individual poling
history with well-defined electrical properties, better than 2%.
Impactor
(250-3000 m/s)
PVDF gauge
Velocity pins
0.5% accuracy (at impact)
Target
0 -1
-2
-3
-2 -4
-5
0 100 200 0 100 200
Time (ns) Time (ns)
With a two-stage light gas gun, impact velocities between 0.85 and 3 km s−1
are achieved. Symmetric impact, nonsymmetric impact, and reverse impact
techniques have been utilized for investigating the shock response of the PVDF
gauges. Also included in Fig. 6.14 is the current response of the PVDF shock
gauge, and the released charge obtained by integrating the current response.
Since the signal rise times are on the order of few nanoseconds, low inductance,
GHz instrumentation must be used throughout, for example Tektronix TDS
684B digitizers.
From a series of symmetric and nonsymmetric impact experiments, the
charge response of PVDF shock gauges is well documented up to pressure
levels of 35 GPa. Figure 6.15 summarizes the experimentally obtained data
and compares the data with calculations obtained from hydrocode DYNA
(Livermore-Software Two Dimensional Hydrodynamic Finite Element Code)
and SHYLAC (Hydrodynamic Lagrangian Simulation of Shocks) models. The
experimental data show that the electrical charge response of the PVDF
gauges is independent of the loading path [47]. Therefore, PVDF gauges can
be used to study shock wave impacts under a large variety of conditions.
PVDF gauges are extremely versatile for investigating porous explosives
under Hugoniot conditions [47]. An example of an experiment, where shock
wave pressure signals are recorded within a porous H.E. Formex explosive is
shown in Fig. 6.16. A PVDF gauge is mounted between the two H.E., two
6 Piezoelectric Polymers and Their Applications 173
7
6
Charge (µC/sqcm)
5
4
3
2
ISL CALCULATION
1 SANDIA FIT
This work
0
0 10 20 30 40
Pressure (GPa)
Fig. 6.15. Charge response of PVDF vs. pressure up to 35 GPa. The very well-
documented charge response of PVDF allows for measurements of pressures during
shock wave experiments (after [47])
Fig. 6.16. Experimental arrangement for the Hugoniot investigation of porous high
explosives. Several PVDF gauges are used to monitor the pressure during explosion,
even within the explosive (after [47])
other gauges are placed in front and at the end of the H.E.’s. The shock wave
response of the shock gauges in the front and between the H.E. Formex are
shown in Fig. 6.17 for illustration. After the shock pressure jump, a plateau
and a slight collapse is observed until the pressure reaches a maximum value
of 9.7 GPa, following the reaction of the H.E. For the gauge in between the
H.E. samples, a fast superpressure of a few nanoseconds duration is observed
at a level of 14.5 GPa, followed by a pressure release down to a plateau level
of 10.5 GPa and finally a pressure decay. The experiments reveal that precise
pressure measurements are obtained by polymer shock gauges, so these gauges
allow the study of materials and properties.
174 S. Bauer and F. Bauer
10 15
8
PRESSURE (GPa)
PRESSURE (GPa)
10
6
5
4
0
2
TIME (µs)
0 −5
−0.5 0 0.5 1 1.5 −0.05 0.00 0.05 0.10 0.15 0.20
TIME (µs)
Fig. 6.17. Pressure levels during the shock-wave experiment with porous explosives
(left), first PVDF gauge in Fig. 6.16, right second PVDF gauge in Fig. 6.16 (after
[47])
ferroelectret
top electrode
thin film transistor
drain
dielectric
source
Cr/Al/Cr interface
a-Si:H
undoped
a-Si:H
SiNx gate electrode
TFT
Fig. 6.18. Ferroelectret field effect transistor. The high impedance ferroelectret is
directly coupled to an impedance matching thin film transistor, enabling the element
to work as a force activated switch, a pressure sensor and a microphone (after [49])
6.8 Conclusions
Acknowledgment
Financial support of the work in Austria from the Austrian Science Funds
and the European Union (3 Plant) is gratefully acknowledged.
176 S. Bauer and F. Bauer
References
1. H.S. Nalwa (ed.), Ferroelectric Polymers: Chemistry, Physics and Applications
(Marcel Dekker, New York, 1995)
2. http://www.meas-spec.com/myMeas/sensors/piezo.asp; http://piezotech.fr
3. see for example http://www.rp-acoustics.de/english/PVDF e.htm
4. information of the company AVL, Austria, www.avlditest.com
5. F. Bauer, IEEE Trans. Ultrason. Ferroelectr. Freq. Contr., 47, 1448 (2000)
6. http://www.algra.ch/Englisch/keyboards.htm
7. B. Ploss, B. Ploss, F.G. Shin, H.L.W. Chan, C.L. Choy, Appl.Phys. Lett. 76,
2776–2778 (2000)
8. Q.M. Zhang, V. Bharti, X. Zhao, Science 280, 2101 (1998)
9. R.J. Klein, F. Xia, Q.M. Zhang, F. Bauer, J. Appl. Phys. 97, 094105 (2005)
10. A. Savolainen, K. Kirjavainen, J. Macromol. Sci. Chem. A26, 583 (1989)
11. S. Bauer, R. Gerhard-Multhaupt, G.M. Sessler, Phys. Today 57(2), 37–43 (2004)
12. M. Paajanen, J. Lekkala, H. Välimäki, IEEE Trans. Diel. Electr. Insul. 8,
629 (2001)
13. information of the company Emfit Ltd., Finland, www.emfit.com
14. information of the company B-Band, Finland, www.b-band.com
15. information of the company Screentec, Finland, www.screentec.com
16. A. Mellinger, M. Wegener, W. Wirges, R. Gerhard-Multhaupt, Appl. Phys. Lett.
79, 1852–1854 (2001)
17. Z. Hu, H. von Seggern, J. Appl. Phys., 99, 024102 (2006)
18. X. Zhang, J. Hillenbrand, G.M. Sessler, Appl. Phys. A 84, 139–142 (2006)
19. S. Katzir, Arch. Hist. Exact. Sci. 57, 61 (2003) and references therein
20. H. Kawai, Jpn. J. Appl. Phys. 8, 975 (1969)
21. J.G. Bergman, J.H. McFee, G.R. Crane, Appl. Phys. Lett. 18, 203 (1971)
22. K. Nakamura, Y. Wada, J. Polym. Sci. A-29, 161 (1971)
23. S.B. Lang, Phys. Today 58(8), 31–36 (2005)
24. A.J. Lovinger, Science 220, 1115 (1983)
25. B. Chu, X. Zhou, K. Ren, B. Neese, M. Lin, Q. Wang, F. Bauer, Q.M. Zhang,
Science 313, 334–336 (2006)
26. S.B Lang, S. Muensit, Appl. Phys. A 85, 125–134 (2006)
27. R. Gerhard-Multhaupt, IEEE Trans. Diel. Electr. Insul., 9, 850–859 (2002)
28. M. Wegener, S. Bauer, Chem Phys Chem 6, 1014 (2005)
29. S. Bauer, IEEE Trans. Diel. Electr. Insul. 13, 953–962 (2006)
30. C. Z. Rosen, B.V. Hiremath, R. Newnham, Key Papers in Physics: Piezoelec-
tricity (American Institute of Physics, New York, 1992)
31. J. Döring, V. Bovtun, J. Bartusch, A. Erhard, DACH Jahrestagung 2004
Salzburg, http://www.ndt.net/article/dgzfp04/papers/p11/p11.htm
32. M. Lindner, H. Hoislbauer, R. Schwödiauer, S. Bauer-Gogonea, S. Bauer, IEEE
Trans. Diel. Electr. Insul. 11, 255 (2004)
33. G.M. Sessler (ed.), Electrets, Vol. I, 3rd edn., (Laplacian Press, Morgan
Hill, 1999)
34. technical information from the homepage http://www.piezotech.fr/
35. T. Furukawa, Phase Transitions 18, 143 (1989)
36. G.S. Neugschwandtner, R. Schwödiauer, S. Bauer-Gogonea, S. Bauer,
M. Paajanen, J. Lekkala, J. Appl. Phys. 89, 4503–4511 (2001)
37. Q.M. Zhang, J. Shao, T. Shrout, N. Kim, L.E. Cross, A. Amin, B.M. Kulwicki,
J. Appl. Phys. 77, 2549–2555 (1995)
6 Piezoelectric Polymers and Their Applications 177
38. F. Bauer, E. Fousson, Q.M. Zhang, L.M. Fee, IEEE Trans. Diel. Electr. Insul.
11, 293–297 (2004)
39. C. Huang, R. Klein, H. Li, Q.M. Zhang, F. Bauer, Z.Y. Cheng, IEEE Trans.
Diel. Electr. Insul. 11, 299–311 (2004)
40. V. Bobnar, B. Vodopivec, M. Kosec, A. Levstik, B. Hilczer, Q.M. Zhang, Ma-
comolecules 36, 4436–4442 (2003)
41. Q.M. Zhang, C. Huang, F. Xia, J. Su, in Electroactive Polymer Actuators as
Artificial Muscles, Chapter 4, ed. by Y. Bar Cohen, (SPIE Optical Engineering
Press, WA, 2004)
42. M. Rabuffi, G. Picci, IEEE Trans. Plasma Sci. 30, 1939–1945 (2002)
43. R. Pechta, RP-acoustics, private communication, 2006
44. A. Streicher, R. Müller, H. Peremans, M. Kaltenbacher, R. Lerch, 2004 IEEE
Ultrason. Symp., 1142–1145 (2004)
45. J. Hillenbrand, G.M. Sessler, J. Acoust. Soc. Am. 116, 3267 (2004)
46. F. Bauer, R.A. Graham, Ferroelectrics 171, 95–102 (1995) and references therein
47. F. Bauer, Mat. Res. Symp. Proc. 698, 53–64 (2002)
48. R.C.G. Naber, C. Tanase, P.W.M. Blom, G.H. Gelinck, A.W. Marsman,
F.J. Touwslager, S. Setayesh, D.M. de Leeuw, Nat. Mat. 4, 243 (2005)
49. I. Graz, M. Kaltenbrunner, C. Keplinger, R. Schwödiauer, S. Bauer, S. Lacour,
S. Wagner, Appl Phys. Lett. 89, 073501 (2006)
50. S.M. Nakhmanson, M. Buongiorno Nardelli, J. Bernholc, Phys. Rev. Lett. 92,
115504 (2004)
51. H. Su, A. Strachan, W.A. Goddard III, Phys. Rev. B 70, 064101 (2004)
7
Electromechanical Frequency Filters
7.1 Introduction
1
Comité Consultatif Internationale Téléphonique et Télégraphique.
7 Electromechanical Frequency Filters 183
80
130 kHz Channel filter
70 0.5 23c
dB dB 60c
60
50 0.3
Fig. 7.1. Tolerance scheme for 50 and 130 kHz-channel filters as developed for tele-
phone systems (1/20 of CCITT1 -specification). It shows the extremely high perfor-
mance required, i.e. a very wide bandwidth, very high attenuation and steep flanks
in the stop band, and very low ripple in the pass band (straight lines). In addition,
the response of a mechanical 130 kHz-filter is shown at 23◦ C (full line) and at 60◦ C
(broken line)
Fig. 7.3. Mechanical channel filters: 50 kHz-filter above and 130 kHz-filter below
For the development of the first filter generation for 50 kHz channel modems, it
was necessary to take into consideration the properties of suitable piezoelectric
ceramics available at that time. On the basis of this the material specifications
for 50 kHz-filters have been established (Table 7.1). Special emphasis was laid
on the positive temperature coefficient of the frequency constant, this being
required to compensate the temperature coefficient of the solder. Otherwise,
the data of acceptor doped (hard) PZT ceramics as used for ceramic AM and
FM radio intermediate frequency (IF) filters have been adopted. For com-
parison reasons, ceramic specifications for radio IF-filters are also shown in
Table 7.1.
Still more stringent specifications had to be met by piezoelectric ceramics
for channel filters for a new generation of 130 kHz channel modems. When the
decision was reached to develop the new generation channel modems, it was
clear that the goal to reduce the size of the channel filters by 70% without
any release in performance (Fig. 7.1) could only be reached by a close mutual
cooperation between ceramic materials experts, filter development engineers,
and system experts.
7 Electromechanical Frequency Filters 185
From Fig. 7.2, it can be seen that a considerable reduction in size can be
achieved by avoiding the coils at the filter input and output. The function of
these coils was to increase the bandwidth of the input and output transduc-
ers, which can otherwise only be achieved with a higher piezoelectric coupling
factor of the transducer ceramic. However, the required coupling factor was
estimated to be much higher than that achievable at best by any ceramic de-
sign. Thus, it was up to the system and filter designers to considerably reduce
the demand on the coupling factor (to the value shown in Table 7.1). They
proposed the following measures which have been realized the following way:
1. The center frequency was increased to 130 kHz to reduce the required rel-
ative bandwidth of the transducers (Fig. 7.1)
2. Directly driven transducer resonators utilizing the thickness coupling factor
k33 2 of the ceramic were used
The directly driven transducer resonators consist of two iron–nickel cylin-
ders with two piezoelectric ceramic platelets of opposite polarity soldered in
between (Fig. 7.3). The elimination of the input and output coils has become
possible because of a higher mechanical coupling factor of the piezoelectric
ceramic. This considerably reduces the length of the filter. The increased cen-
ter frequency also reduces the resonator bolts’ length and therefore the filter
width, at least by a factor of two. Furthermore, by designing the new channel
2
k33 is approximately two times larger than k31 utilized in transducer resonators
of the 50 kHz-filters.
186 W. Wersing and K. Lubitz
filter with attenuation poles on each side of its passband, it was possible to
manage with fewer resonators, thereby again reducing the length of the filter
and, above all, shortening its group delay [2]. Altogether, a bulk size reduction
by almost 70% (Fig. 7.3) has been achieved.
To realize sufficiently low input and output impedances to allow the use
of fully integrated amplifiers without impedance-matching transformers, the
permittivity of the transducer ceramic had to be increased to about 1,350 and
its temperature coefficient decreased to TCε < 2, 500 ppm K−1 .
More stringent specifications had to be met by the piezoelectric ceramic
for the narrowband signal filters of the 130 kHz channel modems as well. Es-
pecially a very high resonance Q > 1, 800 in combination with a very low
temperature coefficient TCQ (|TCQ/Q| 5 ppm K−1 ) was required to keep
the insertion loss sufficiently independent of temperature (Sect. 7.4.5).
Finally, it should be stressed that not only improved ceramic properties
but improved properties together with much lower fabrication tolerances had
to be met for the 130 kHz-modems (Table 7.1).
% V/mm V/mm
AF Ei Ec
0.6 600 1500
0 0
0 1 2 3 4 mole% 5
Mn concentration
Fig. 7.4. Coercive field Ec , internal field Ei and aging rate of Mn-doped PZT
ceramics at the morphotropic phase boundary vs Mn-Concentration
7 Electromechanical Frequency Filters 187
It has been found that the resonance Q of ceramics doped exclusively with
acceptors such as Cr, Mn, and Fe is limited to about 800–1,000. This limita-
tion results at least partly from the electrical conductivity3 of these ceramics
(10−10 − 10−9 (Ωm)−1 ), which is far higher than that of ceramics doped with
donors or modified with complex oxides (see later). The former sometimes
show anomalously large dielectric losses [3] because of inhomogeneous high
conductivities (10−5 − 10−4 (Ωm)−1 ) at grain boundaries and as a result also
decreased Q values of about 400.
Especially ceramics doped with Cr or Mn were found to have a frequency
constant of very high time stability (Fig. 7.4) and thus already in the sixties
the temperature stability of their frequency constant was also investigated to
see whether they can be used for frequency filters [4].
The Cr or Mn doping stabilizes the domain configuration in the ceramic
grains and thereby also the ferroelectric phase distribution within ceram-
ics near the morphotropic phase boundary. Otherwise these ceramics would
be generally unsuitable for filter applications because phase transitions in-
duced by temperature variations would result in a frequency constant having
a markedly nonlinear temperature coefficient with a superposed hysteresis
loop. This is best demonstrated with the aid of samples having a composition
fairly close to the phase boundary (Fig. 7.5). In the sample doped with only
0.4% Mn, the transition from the rhombohedral phase with a highly negative
TCF to the tetragonal phase with a highly positive TCF is well pronounced,
whereas in the sample doped with 4% Mn, the transition is seriously inhibited.
.2
% 4%
Mn
0
−.2
.4
%
f/f
.2
0.4%
0 Mn
−.2
−.4
−.6
−50 0 50 100C
T
3
Because of the electromechanical coupling, the mechanical losses are coupled to
the dielectric losses and, therefore, depend also on the conductivity of the ce-
ramic [5].
188 W. Wersing and K. Lubitz
For ceramics with a composition more in the tetragonal phase, e.g., ceramics
with peak permittivity (poled ceramics) a purely positive TCF with nearly
no temperature hysteresis is obtained.
The stabilization of the domain configuration of these ceramics becomes
obvious from there hysteresis loops. Nonpolarized samples exhibit constricted
and polarized samples exhibit displaced hysteresis loops (Fig. 7.4). This effect
is described usually by introducing an internal stabilization field Ei (Fig. 7.4).
This internal field Ei develops according to a logarithmic time law [6]. A sim-
ilar logarithmic time law is experimentally observed for the aging of materials
parameters such as permittivity, loss tangent, frequency constant, and reso-
nance Q. This time law can be substantiated theoretically using a model that
traces back the aging of ceramic parameters to a thermally activated tran-
sition from a domain configuration that is not in thermal equilibrium to a
thermally balanced configuration. All symptoms described – high resonance
Q, linearization of TCF, reduction of aging rate, appearance of an internal
field – appear to have the same physical origin. Numerous models have been
discussed in the literature for this, including bulk effects, domain wall effects,
and grain boundary effects. In all these models, oxygen vacancies appear to
be of critical significance, and it has been shown [6] that each of these effects
has its justification, i.e., stabilization cannot be attributed to a single effect
in the case of all materials.
On the basis of these basic phenomena important for filter ceramics, it was
possible to develop PZT ceramics solely doped with Mn that were able to
fulfil the requirements for the first generation of channel modems (Table 7.1).
We will now show how the properties of filter ceramics can be considerably
improved by the combination of additives. As the mole fraction of an ad-
ditive in PZT is increased, a point is soon reached, depending on the limit
of the solubility of the additive, beyond which the ceramic properties can-
not be further improved or even a marked degradation of the properties
will set in. This situation was soon to lead to the use of coupled substitu-
tions with ions of different valences, in other words, substitutions with com-
plex oxides (Chap. 4). For filter ceramics various ternary ceramic systems
such as PbTiO3 -PbZrO3 -Pb(Mg1/3 Nb2/3 )O3 have been investigated, which
had to be doped additionally with a stabilizer such as Mn [7]. Takahashi
et al. have shown that it is possible to use Mn directly as a constituent
of a complex oxide [8]. They reported the presence of ceramics with fa-
vorable coupling factors and very high mechanical Q values in the system
PbTiO3 -PbZrO3 -Pb(Mn1/3 Nb2/3 )O3 that recommend this system for filter
ceramics [8].
Therefore, we studied this system in the vicinity of the morphotropic phase
boundary varying the amount of the complex additive (Mn1/3 Nb2/3 ) between
2 and 20%. The principal results – permittivity of poled samples, the planar
7 Electromechanical Frequency Filters 189
0.20
0.15
Z
ε/ε0
0.10
0.05
0
0.35 0.4 0.45 0.5 X 0.55 0.6
Fig. 7.6. Permittivity of poled Pb[Zrx Tiy (Mn1/3 Nb2/3 )z ]O3 ceramics (x+y +z = 1)
The strong correlation between the temperature coefficients TCε and TCF
as well as their dependence on composition can be understood as a result
of the strong dielectric anisotropy in the vicinity of the morphotropic phase
boundary, as discussed in Chap. 3. This can be shown simply by calculating
TCε and TCF of ceramic samples (slim bars) from the temperature coefficients
of the single crystal permittivity and compliance components as derived with
the aid of Devonshire’s theory in Chap. 3.1. Because TCε and TCF anomalies
in the vicinity of the morphotropic phase boundary likewise appear in unpoled
samples (see Fig. 7.1 of [9]), one can calculate the temperature coefficients
of unpoled ceramic samples, i.e., one can calculate them from the effective
permittivity εT∗ (Equation 3.35 of Chap. 3.2) and from the isotropic fraction
190 W. Wersing and K. Lubitz
0.20
0.15
kp Z
0.10
0.05
0
X
0.35 0.4 0.45 0.5 0.55 0.6
Fig. 7.7. Planar coupling factor of Pb[Zrx Tiy (Mn1/3 Nb2/3 )z ]O3 ceramics
(x + y + z = 1)
0.20
0.15
TCε
Z
0.10
0.05
0
X
0.35 0.4 0.45 0.5 0.55 0.6
Fig. 7.8. Temperature coefficient of permittivity of Pb[Zrx Tiy (Mn1/3 Nb2/3 )z ]O3
ceramics (x + y + z = 1)
7 Electromechanical Frequency Filters 191
0.2
0.15
Z
0.1
0.05
0
0.35 0.4 0.45 0.5 X 0.55 0.6
of the compliance matrix sαβ E [10]. The advantage of this procedure is that
the calculation of effective properties of unpoled ceramics is much simpler
than that of poled ceramics. Of course, this procedure can only be used for
single crystal properties, which do not depend on the direction of spontaneous
polarization, i.e., for those that are described by tensors of even rank.
Let us first consider the temperature coefficient TCF. Because the TCF
of discs is nearly identical to that of slim bars4 , we consider here the simpler
case, i.e. we calculate the TCF of slim (unpoled) ceramic bars given by:
TCF = − 12 · TCsD∗
11 − α1 (7.1)5
Calculating s11 D∗ from the isotropic fraction of the compliance matrix sαβ E
D 2 E 1 E
11 ≈ s11 =
sD∗ 1
5sE E E E E E
11 + s22 + s33 + 15 s12 + s23 + s13 + 15 s44 + s55 + s66
E
(7.2)
and using the form (3.31) for the sαβ E components one obtains for the effective
compliance s11 D∗
D
sD∗ s2 T
11 = s11 + κ1 P ε11 + κ3 P ε33
s2 T
(7.3)
4
This is found experimentally and can also be derived theoretically.
5
The thermal expansion coefficient α1 vertical to Pr is typically 5–10 ppm K−1 and
can be neglected for the following discussion.
192 W. Wersing and K. Lubitz
boundary is caused by the strong dielectric anisotropy as can be seen from the
second term of (7.3). As can be further seen from (7.3), the TCF is mainly
controlled by the temperature coefficients TC(ε11 T · P s2 ) and TC(ε33 T · P s2 )
which are calculated from (3.25)–(3.28) to be:
TC εT 33 P
s2
= 8χ0 bεT2
33 P
s4
(7.6)
tetragonal phase: TC εT 33 (ζ112 − ζ111 ) ε11 P
s2
11 P = 2χ0 εT T s4
(7.7)
rhombohedral phase: TC εT 33 9 (ζ111 − ζ123 ) ε11 P
s2
11 P = 2χ0 εT 4 T s4
(7.8)6
the other hand, the third term in (7.3) has no effect at the rhombohedral side
of the phase boundary because here κ3 is even smaller than κ1 . Having in
mind that at the morphotropic phase boundary both the tetragonal and the
rhombohedral phase are coexisting, the jump of TCF from positive to large
negative values at the phase boundary as obtained from (7.7) and (7.8) is
smeared out and the experimentally found behavior is well described as can
be seen in Fig. 12 of Chap. 3 and in Fig. 7.9. However, in Fig 7.9, the negative
TCF peak is not fully visible within the plotted data set. The correlation be-
tween the temperature coefficient of permittivity and frequency constant can
be seen easily from (7.10). The first term in (7.10) simply shifts the typical
dependence of TCs11 D∗ on composition x mainly controlled by TC(ε11 T P s2 )
to positiver values.
The theoretically found jump of the temperature coefficients at the phase
boundary shows clearly that the measurement of TCF or TCε (or both) in the
vicinity of the morphotropic phase is very useful to determine the exact phase
boundary and the coexistence region by means of electrical measurements.
Therefore, it can be noted from Figs. 7.8 and 7.9 that the phase boundary is
shifted to lower Zr contents x, and the mixed phase region of the rhombohedral
and tetragonal phases grows broader as the mole fraction of Mn1/3 Nb2/3 is
increased. In addition, however, a strong levelling off of the TCF maximum in
the tetragonal phase is observed, where even the TCF for z 0.15 is no longer
positive at all. Furthermore, this lowering of the TCF peak is accompanied by
its broadening and, above all, its displacement away from the phase boundary.
It is interesting to compare this behavior with that of dopants such as Nd
which substitute Pb. Such dopants also lead to significant lowering of the TCF
peak, while any large mole fraction of such dopants will even drive the TCF
peak to negative values. Since, however, they will not result in any notable
displacement of the peak away from the phase boundary, they can be used to
lower the TCF peak of a ceramic system to a desired value without danger
of any serious reduction of the coupling coefficient. Such an example is given
in [9] where a complex ceramic system modified with Ni1/3 Nb2/3 and with 4%
Mn1/3 Nb2/3 was additionally doped with Nd. Without any Nd doping a TCF
peak of almost 300 ppm K−1 was obtained. With the increased addition of Nd,
the TCF peak was sharply reduced and already assumes negative values as of
about 4% Nd. Of course, the Nd additive causes a displacement of the phase
boundary, but without any notable displacement of the TCF peak relative to
the phase boundary.
The possibility to adjust the TCF peak to a value that is requested by the
filter design is technologically of enormous significance. Only for a material
close to the TCF peak, the TCF variation with composition is minimal and,
therefore, its sensitivity to fabrication tolerances is low, i.e., it is possible to
achieve a highly reproducible mass production of ceramics with tight TCF
tolerances.
194 W. Wersing and K. Lubitz
On the basis of the ceramic system studied so far we find a TCF value
of about 80 ppm K−1 as required for channel filters for a composition with
x = 0.43 and z = 0.1 (cf. Fig. 7.9). With this ceramic composition a TCF
tolerance of ±10 ppm K−1 is readily reproducible. Since the permittivity and
coupling factor of this composition also vary only at a low rate, the too are
easily reproducible. Furthermore, since their absolute values are higher than
those required for 50 kHz channel filters, these ceramics were ideally suitable
for the production of first-generation 50 kHz channel modems. However, the
permittivity and coupling factor values of these ceramics could not fulfil the
requirements for 130 kHz modems.
In the ceramic system Pb[Zrx Ti0.9-x (Mn1/3 Nb2/3 )0.1 ]O3 and especially for
the composition with x = 0.43, we obtained permittivity and coupling factor
values of 1,050 and 0.46, respectively, considerably lower than required (for
130 kHz modems). However, all other important parameters (cf. Table 7.1)
were met excellently. Therefore, it would have been obvious to improve per-
mittivity and coupling factor simply by adding of a softener such as Nd or Nb
because the addition of a softener to PZT is known to improve these parame-
ters considerably (Chap. 4). However, since on the one hand, practically all
additives lower the Curie temperature and, on the other hand, the Curie point
of filter ceramics has to exceed 300◦ C substantially to avoid a degradation of
the ceramics by soldering sufficiently small, the total mole fraction of addi-
tives has to be kept as small as possible. Thus it just was logical to induce a
softening effect in the system Pb[ZrTi(Mn1/3 Nb2/3 )]O3 by reducing the ratio
of Mn to Nb. This procedure was found to allow the realization of a continu-
ous series of ceramics extending from hard to soft (exclusive Nb doping). It is
obvious that this knowledge has become crucial for many other applications,
too. Ceramics with optimally high permittivity and coupling factor on the one
hand, and high resonance Q and low aging rates on the other, were obtained
by choosing a Mn/Nb ratio of approximately 1/5. Figure 7.10 shows the prin-
cipal material parameters of such ceramics vs. the Zr content. Ceramics with
particularly favorable material parameters with respect to the requirements
of 130 kHz modems (cf. Table 7.1) were realized with Zr concentrations x
between 0.43 and 0.44.
We would like to emphasize that the noncompensating-valence substitu-
tion influences the TCF peak similarly like three-valent dopants such as Nd
substituting Pb, as discussed earlier. With this substitution, the TCF peak
was considerably reduced, but without any notable displacement of the TCF
peak away from the phase boundary, i.e., the TCF peak remains quite close
to the permittivity peak (cf. Fig. 7.10).
7 Electromechanical Frequency Filters 195
TCF
ε/ε0
ε/ε0 Kp TCF Q
Fig. 7.10. Properties of PZT ceramics modified by Mn and Nb in the ratio 1/5
50
30
20
1000/Q
10
5
3
2
0.5
0.3
−200 −100 0 T 100
References
1. H. Albert, Feinwerktechnik 72, 244 (1968)
2. A.E. Günther, H. Albsmeier, K. Traub, Proc. IEEE 67, 102 (1979)
3. W. Wersing, Ferroelectrics 22, 813 (1978)
4. H. Banno, T. Tsunooka, Jpn. J. Appl. Phys. 6, 954 (1967)
5. R. Truell, C. Elbaum, B.B. Chick, Ultrasonic methods in Solid State Physics
(Academic Press, New York, 1969) pp. 341–343
6. K. Carl, K.H. Härdtl, Ferroelectrics 17, 473 (1978)
7. H. Schichl, NTZ 4, 299 (1976)
8. M. Takahashi, N. Tsubouchi, M. Yonezawa, T. Ohno, T. Akashi, Nec Res.
Develop. Jpn. 35, 57 (1974)
9. W. Wersing, Ferroelectrics 37, 611 (1981)
10. I.M. Lifschitz, L.N. Rosenzweig, J. Exp. Theoret. Phys. 16, 967 (1946)
11. G. Zorn, W. Wersing, H. Göbel, Jpn. J. Appl. Phys. 24 suppl. 24–2, 721 (1985)
12. J.O. Genter, P. Gerthsen, N.A. Schmidt, R.E. Send, J. Appl. Phys. 49,
4485 (1978)
13. H.-J. Hagemann, J. Phys. C. Solid State Phys. 11, 3333 (1978)
14. K.H. Härdtl, Ferroelectrics 24, 75 (1980)
15. K. Möhring, H. Schichl, Ber. Deutsch. Keram. Ges. 53, 200 (1976)
8
Ultrasonic Imaging
8.1 Introduction
Ultrasonic medical imaging provides a low cost and non-invasive tool to look
inside the human body. It is one of the most popular imaging tools, second only
to conventional X-ray imaging in the number of clinical procedures performed.
It has various advantages over conventional X-ray, computed tomography and
magnetic resonance imaging, because it does not utilise harmful electromag-
netic radiation, is non-invasive, easy to perform, cost effective, capable of
providing images in real time, and provides good soft tissue differentiation [1].
Ultrasonic imagers for medical diagnosis are to be categorized as piezo-
electric sonar sensors. These are sensors that emit ultrasound and detect
ultrasound after it has been reflected from objects or passed through media.
They are used for the identification of object structures, for the measurement
of distances or for the measurement of medium or object properties. Ultrasonic
transducer for sonar sensors differ considerably, depending on whether sound
propagation happens in gas, liquid or bulk. For measurements of the amount
of a liquid and its flow speed, special sonar techniques have been developed.
Ultrasonic imaging technology has come a long way from its beginnings in
the 1970s. The image quality of medical ultrasound has improved vastly with
linear and phased array transducer designs replacing the mechanical scanners
used 2–3 decades ago. Further improvements have been achieved with the
optimisation of the relevant properties of piezoelectric materials and array
processing techniques. These advanced features are mandatory for modern
imaging modalities, whether this is compounded imaging, second harmonic
imaging, contrast agents, colour flow Doppler or 3D image reconstruction to
enhance the image quality. Modern ultrasonic imaging has been and continues
to be a field of intense research and development where many innovations from
first ideas to final commercial products are achieved in rather short periods.
New imaging modalities are continually developed and adapted to existing
imaging systems. Many of those are recognized not only as useful but often as
indispensable diagnostic aids. These improvements require the development of
200 W. Wersing and R. Lerch
imaging transducers with specific properties and of high performance being ca-
pable of operating in a wide range of modes. These performance requirements
have a strong impact on the complexity of transducer fabrication processes
and piezoelectric materials’ performance.
Z = ρv, (8.1)
where ρ is the density and v is the sound velocity of the medium. When the
acoustic waves generated by the piezoelectric transducer encounter an im-
pedance discontinuity, some of the energy is reflected and the remaining is
transmitted ahead (see Fig. 8.1). When reaching the transducer, the reflected
echo generates a voltage that is proportional to the acoustic impedance mis-
match at the interface between the tissues. The transmitted part of the sound
wave can serve to detect other impedance discontinuities that may be present
at farther distances (Fig. 8.1). The relative small impedance mismatch be-
tween soft tissues allows for the ultrasonic waves to propagate across several
interfaces, providing imaging capability up to large depth. The penetration
depth depends on the tissue type and the ultrasound frequency f and reaches
typically 10–15 cm at a frequency of 5 MHz. The lateral resolution is approx-
imately three times the wavelength, that is at 5 MHz about 1 mm.
Fig. 8.1. Principle of ultrasonic imaging systems. Mean values for soft biologi-
cal tissue are sound velocity v = 1, 540 m s−1 , characteristic acoustic impedance
Z = ρv = 1.5 × 106 Pa s m−1 , sound wavelength at f = 3.5 MHz λ = 0.44 mm
8 Ultrasonic Imaging 201
The received voltage pulses deliver two kinds of information: first their
amplitudes are proportional to the acoustic impedance mismatch at the po-
sition of reflection and second, the position itself can be calculated exactly
from the delay time between pulse transmission and reception, provided the
sound wave velocity is known. Because of the fact that biological tissue mainly
consists of water, only longitudinal sound waves with a typical sound wave
velocity between 1,500 and 1, 600 m s−1 can propagate in biological tissue.
The simplest imaging mode is named A-scan (after amplitude). Here the
amplitudes of the received echo pulses are simply displayed vs. the propagation
path, that is penetration distance.
If the amplitudes of the echo pulses are transformed into brightness sig-
nals, the echo positions along the propagation path or penetration distance
d can be depicted directly as brightness modulated dots alongside the pic-
ture line, usually vertical on the screen. To obtain a complete picture, many
(at least 100) of these picture lines are lined up scanning along the body
(span s) using mechanical or electronic scan techniques. By this means a sec-
tional grey scale image (d vs. s) of the investigated tissue is obtained. This
allows anatomy to be observed in a wide range of organs, for example in ob-
stetrical or cardiological imaging. The brightness modulated sectional view is
usually named B-scan (after brightness). Today, B-scan imaging techniques
completely outweigh other imaging modalities.
There exists a third imaging mode, which is called M-scan (after motion).
To obtain this mode the picture lines are not lined up from different transducer
positions s but from the same position at different times t. Thus, the obtained
grey scale image (d vs. t) shows the motion of the sounded tissue. On the
contrary, this explains that for sectional grey scale imaging, it is essential to
guarantee good real time capability to avoid artefacts caused by movements
of the organs.
However, the movement of blood can also be used for estimating its flow
velocity with high local resolution. This estimate is then superposed to the
grey scale image using colour coding. The flow measurement itself is based on
the physical effect, discovered by Doppler, that the sound frequency of a source
is shifted due to its movement. In the case of blood flow measurement, the
reflecting blood acts as sound source. The frequency f of the received signal
backscattered from the blood particles is shifted in relation to the transmitter
frequency f0 by the Doppler frequency
vblood
∆f = f − f0 = 2f0 cos ϕ, (8.2)
v
with v the sound velocity, vblood the velocity of the reflecting blood parti-
cles and ϕ the angle between sound direction and direction of blood flow.
The detection of the weak backscattered signals from the blood particles puts
forward highest demands upon the dynamics of signal processing. These per-
formance requirements have direct consequences on those of the piezoelectric
transducer materials.
202 W. Wersing and R. Lerch
There are many possibilities to scan and to focus ultrasonic images us-
ing specific ultrasonic transducers often also named ultrasonic antennas, as
discussed in the following section.
Fig. 8.2. Single element transducers, the classical disc transducer with fixed focus
defined by the lens form (left) and annular array transducer consisting of acoustically
isolated piezoelectric rings (right). By individually controlling the signal phases of
the different rings, it is possible to vary the transducer’s focus electronically. Match-
ing layers and backing are not shown in the figure
8 Ultrasonic Imaging 203
are integrated into one layer or layer package. On the back side of the trans-
ducer, a thick backing layer is attached usually. It completely absorbs the
ultrasonic power emitted in this direction so that no energy can be radiated
back to the active layer, which would produce parasitic echo signals.
The image quality at a given ultrasound frequency1 strongly depends on
the axial and lateral resolution. Usually, because of a good acoustic matching,
the axial resolution is better than the lateral one. Thus, improving of the image
quality requires the improving of the lateral resolution. The lateral resolution
of a single-element transducer is correlated with the beam profile and depends
on the diameter of the transducer and the lens curvature (curvature radius
R). Furthermore, it depends upon the degree of focusing, that is the ratio
between the active transducer diameter 2a (aperture) and the focal distance
z0 ≈ R. With increasing degree of focusing, the resolution increases within the
focus range (Fig. 8.3); however, at the same time, the depth of focus decreases.
Therefore, an optimal resolution requires for each focal distance z a different
lens curvature. Furthermore, to keep the lateral resolution constant within the
whole image depth, it would be necessary to increase the transducer aperture
with increasing distance z so that z/2a ≈ constant.
This clearly shows that classical disc transducers cannot be focussed op-
timally, a compromise must be found between the degree of focusing and the
depth of focus. To overcome this problem annular array transducer have been
developed (Fig. 8.2). Such a transducer functions quite similarly to a classical
single element disc transducer, merely the transducer disc is divided into con-
centric piezoelectric ring transducers (Figs. 8.2 and 8.4), each of which can be
individually driven by a signal pulse. Focusing can be achieved by applying
delays to the driving pulses in such a way that the acoustic pressure gener-
ated by each ring transducer arrives at the focal point at the same time, that
is the driving signals applied to the inner rings must be more delayed than
those applied to the outer rings. For a transmitted pulse the delays are fixed
to a certain focal distance according to the desired image depth. However, for
received echo signals, the delays can be adapted continuously to focus on the
depths at which the echo signals were produced. This method is referred to
d
2a
z
2.4λ ×z0
d=
z0 2a
Fig. 8.3. Lateral resolution and depth of focus for a classical single element trans-
ducer. The beam profiles for two transducers with 2a = 15 mm and 2a = 40 mm are
compared, the focal distance is z0 = 60 mm
1
The ultrasound frequency is usually limited by the necessary penetration depth.
204 W. Wersing and R. Lerch
Fig. 8.4. Ultrasonic annular array transducer, top view (left) and side view of an
array cut through (right)
Scan direction
1.2.3.4.5. . . . . Element Transmitter
delay lines
τ1
1
τ2 2
3 Sound
τ3
. ...
4 beam
τ4 .. . s = (τ1– τn)v
n
Array
τn
Fig. 8.5. Principle function of ultrasonic electronic array scanners, linear array (left)
and phased array (right)
2
It shows details much better than modern linear arrays.
206 W. Wersing and R. Lerch
Fig. 8.6. View of the piezoelectric transducer elements (without matching layer)
of a linear ultrasonic array of the first generation (top) and array with transducer
elements divided into sub-elements (bottom)
3.5 MHz Convex array 6.5 MHz Prostata probe 13 MHz 1.5 D array antenna
Fig. 8.9. Modern ultrasonic imaging system together with imaging probes for
different applications in medical diagnostics
Fig. 8.10. Cutting of a 2D ultrasonic phased array antenna. The transducer ele-
ments have a base area of 150 × 150 µm2 and a height of about 1 mm (matching
layers included)
In this section, general aspects concerning materials and array fabrication are
explained, whereas in Sect. 8.5 we discuss all features related to both the state
of the art piezoelectric ceramics used in ultrasonic imagers (or imaging equip-
ment) and to piezoelectric materials specially designed for this application.
The fabrication of standard 1D linear and phased arrays usually starts
with thin ceramic strips of length l and wideness w according to the array
dimensions, typically 1 cm wide and 1–10 cm long. The thickness t of those
strips is λ/23 . This corresponds to a thickness between 0.15 and 1 mm for ar-
rays with a frequency between 15 and 2.5 MHz, respectively. These stripes cut
from sintered ceramic blocks are processed very similar to the filter ceramic
platelets discussed in Chap. 7. This concerns similarly the deposition of elec-
trode systems, which usually consist of a thin adhesion layer, a thin diffusion
barrier layer and a gold electrode. These ceramic stripes are connected with
the backing block on the one side (rear side) and with the matching layers
(up to three λ/4 layers) and a protection layer including the lens on the other
side (front side) (Fig. 8.7). The sides of the backing usually serve to contact
the transducer elements via fixed or deposited transmission lines. Contacting
of the transducer elements is achieved using soldering or glueing.
The backing on the rear face of the transducer array is usually the thickest
layer. It serves as the mechanical support of the antenna beside its absorber
3
λ is the wave length of sound in the piezoelectric ceramic.
210 W. Wersing and R. Lerch
function (acoustic swamp). For this layer, a composite made of a high loss
polymer (high visco-elastic attenuation) and embedded heavy metal particles
such as tungsten particles [4] is used. Via the concentration and grain size
of the embedded heavy particles, this composite with (0–3) connectivity4 can
be optimised concerning acoustic impedance and sound attenuation scatter-
ing, respectively. Very high attenuations together with a well defined acoustic
impedance can be achieved using a mixture of different particle types. Addi-
tionally, the thickness of the backing can be reduced considerably to realize
miniaturized antennas.
Matching layers are also fabricated from particle loaded polymer compos-
ites. However, their attenuation must be kept as low as possible. Therefore,
a single type of well defined particles is added in an amount to achieve the
desired acoustic impedance.
thickness) the relevant coupling factor is k33t ≡ kt 5 < k33 and for small area
elements of arrays (e.g. bars or rods with lateral dimensions thickness as in
Fig. 8.10) it is k33 . For transducer elements of phased arrays (Figs. 8.7 and 8.8)
with one lateral dimension < thickness and the other lateral dimension
d33
FVS = , (8.3)
CE + Cline + Cinput
with CE ∼ ε33 T
d33
and FSNR = . (8.4)
T
ε33 tan δ
From (8.3) and (8.4) it can be seen that in the receiving mode the trans-
ducer material should have a high piezoelectric coefficient d33 , a low permit-
tivity ε33 T and a low loss factor tan δ. The general requirement of a high
coupling factor k33 or kt is usually in line with a high piezoelectric coef-
ficient d33 . However, the requirement of a high d33 and a low permittivity
ε33 T is contradictory for MPB6 -piezoelectrics and thus a suitable compro-
mise must be found. Using √ the relation between k33 and d33 shown √ in (18.21)
of Chap. 18 d33 = k33 (ε33 T s33 E ), it is recognized that d33 ∼ CE . Thus,
both the voltage sensitivity and FSV increase with increasing CE (respec-
tively, ε33 T ), they assume a maximum value at CE = Cline + Cinput (electrical
matching) and then decrease again. Assuming a transmission line of length
1 m, the line capacitance7 Cline of a line with Zline = 80 Ω is typically 88 pF.
To achieve electrical matching with a standard phased array transducer el-
ement of wideness w = 0.17 mm (Fig. 8.7), length l = 10 mm and thickness
t = 0.55 mm (3.5 MHz), a ceramic permittivity ε33 T /ε0 > 6, 500 would be
required. Usually a soft PZT ceramic is chosen with k33 as high as possible
and with 2,000 < ε33 T < 3,000. But, of course, moderate dielectric and me-
chanical losses (≈0.02) must be accepted with these kind of materials. Some
mechanical losses can even be useful to damp internal mechanical resonances
of the active layer. However, it should be noted that within complex PZT-
based ceramic systems doped with Mn, it is possible to obtain materials with
high d33 , medium ε33 T and low losses (Chaps. 4, 7, and 9).
For high frequency devices and for arrays, that is in all cases where fine
sawing techniques are required for structuring, fine-grained ceramics without
any large pores or other defects (such as second phases like ZrO2 ) are nec-
essary. Furthermore, sawing parameters must be optimally adapted to the
piezoelectric material, for example when benefiting from ferroelastic proper-
ties to avoid any flaw.
6
PZT ceramics with a Zr/Ti ratio in the vicinity of the morphotropic phase bound-
ary (MPB), see Chap. 3.
7
The typical capacitance lining of a low capacitance transmission line with Zline =
150 Ω is 27 pF m−1 .
212 W. Wersing and R. Lerch
8
New processing techniques for composites are shown in Chap. 16, regarding the
connectivity definitions, see also Chap. 16.
8 Ultrasonic Imaging 213
distribution of the active ceramic phase [9, 10]. Different approaches based
on traditional ceramic processing techniques such as tape casting, injection
moulding, dicing and such have been proven in making composites with dif-
ferent but simple connectivity. However, these techniques do not permit the
fabrication of structures with complex internal hierarchy and symmetry due
to their rectilinear mode of operation. To overcome this limitation, a new
manufacturing/prototyping technique the so-called Solid Freeform Fabrica-
tion (SFF) is emerging. It provides very high design flexibility due to an in-
tegrated way of manufacturing 3D components directly from computer-aided
design (CAD) files (see Chap. 16).
It is interesting to notice that linear and phased arrays constructed from
alternating layers of piezoelectric ceramic and polymer can be considered
as ceramic–polymer composite with 2–2 connectivity, manufactured with the
traditional dicing and filling technique. Highly porous ceramics can also be
considered as composites, ceramic–air-composites with 3–0 and 3–3 connec-
tivity. They can be fabricated quite inexpensively using different processing
methods and show rather interesting properties [11]. Recently, a further more
traditional technique has been developed for ceramic–polymer-composites
with 1–3 connectivity, well suited for ultrasonic imaging. It is based on a fine
structured piezoelectric ceramic (Fig. 8.11) obtained with a lost wax method
in combination with a precision slip-casting process [12]. A high thickness cou-
pling factor kt , close to the ceramic’s k33 , can be obtained even for relatively
low ceramic contents. So that it is possible to reduce the acoustic impedance
Zcomposite to about Zc /4.
400 m
Fig. 8.11. Fine structured piezoelectric ceramics for 1–3 composites, fabricated by
a lost wax method in combination with a precision slip-casting process
214 W. Wersing and R. Lerch
Fig. 8.12. Properties of thickness transducers made from bulk ceramics (top) and
1–3 composites fabricated by a lost wax method (bottom)
The advantage of such 1–3 composites can be seen in Fig. 8.12, where
a thickness transducer (resonator) made of a bulk ceramic is compared to
one made of such a 1–3 composite. From the admittance (amplitude and
phase) vs. frequency measurements, shown in Fig. 8.12, the effective coupling
factors and resonance Q values have been obtained. The advantage of the
composite transducer is clearly to be seen. Its kt is about 95% of the ceramic’s
k33 , Zcomposite ≈ 0.4Zc , and now spurious resonance artefacts due to lateral
overtones are observed.
Multilayer
transducer
Backing
layer
Bottom
electrode
contact
Fig. 8.16. Ultrasonic image of a foetus (tenth week pregnancy) taken with a me-
chanically moved 1D linear multilayer antenna
k · (r0 − r), the following non-linear relation between displacement and applied
voltage can be derived
2k √
V = r r0 − r. (8.6)
ε0 A
When V increases r decreases until V (r) reaches its maximum at r = 2r0 /3.
Then a collapse of the membrane occurs and the gap closes abruptly. A linear
small signal response of the membrane can be achieved by applying a bias
voltage to the membrane. If the bias voltage is chosen to be just below the
collapse voltage (pull down voltage), a very high response can be achieved,
equivalent to nearly 1,000 times the d33 coefficient of a typical piezoelectric
ceramic. Of course, the dynamic is then very low, that is there is a trade-off
between sensitivity and dynamics. A decisive advantage of c-MUTs is that
they can be manufactured by surface micro-machining using standard CMOS
technologies (Fig. 8.17).
In p-MUTs, the actuation is based on a piezoelectric bending transducer,
which is obtained by depositing a thin (typically 1 µm) piezoelectric layer
together with the necessary electrodes on top of a membrane (e.g., poly-
Si) thermally matched to silicon. The p-MUT technology offers the following
advantages:
• The thin piezoelectric layers, although very small, have a low electrical
impedance and, therefore, can be easily matched to the 50 Ω coaxial trans-
mission lines usually used.
Bond pads
100mm.100mm.
Fifo register
(first in first out)
and digital memory
Array antenna
30×30 pixels
Imager
50–80Ω
Qs Vinput
d33 An
Vinput = p, (8.7)9
CE + Cline + Cinput
where CE is the capacitance of the transducer element, Cline that of the trans-
mission line and Cinput the input capacitance of the processing electronic.
From (8.7) the voltage sensitivity of the antenna element is obtained to be
Vinput d33 An
SVS = = . (8.8)
p CE + Cline + Cinput
The characterization of sensor response is insufficient without analyzing
also its noise generation, which limits the smallest input signal that can be
detected. Therefore, we evaluate the total noise measured at the output of the
read-out amplifier. It is related to the noise generated in the matching layers,
the transducer element and the read-out pre-amplifier. However, since a gen-
eral discussion of the pre-amplifier noise is out of the scope of this chapter,
only the noise is considered that can be measured at the input of the signal
processing unit (sensitivity is also related to this input). The main noise volt-
age at this input results from a fluctuating current <In 2 > within a certain
frequency band ∆f = f2 − f1 often denoted as noise bandwidth
f2
2
In = wI (f ) df , (8.9)
f1
d33
FVS = , (8.13)
CE + Cline + Cinput
with CE ∼ ε33 T
d33
and FSNR = . (8.14)
T
ε33 tan δ
References
1. K.K. Shung, IEEE Eng. Med. Biol. 15, 18 (1996)
2. C.S. Desilets, J.D. Fraser, G.S. Kino, Ultrasonics 25, 115 (1978)
3. J. Borburgh, I. Feigt, P. Hini, V. Zurinski, Siemens Forschungs-u. Entwickl. Ber.
9, 116 (1980)
4. T.N. Nguyen, M. Lethiecq, F. Levassort, L. Pourcelot, IEEE Trans. Ultrason.
Ferroelect. Freq. Contr. 43, 640 (1996)
5. T. Kobayashi, S. Shimanuki, S. Saitoh, Y. Yamashita, Jpn. J. Appl. Phys. Part
1R 36, 6035 (1997)
6. C.G. Oakley, M.J. Zipparo, IEEE Ultrason. Symp. Proc. 1, 1157 (2000)
7. S. Saitoh, T. Takeuchi, T. Kobayashi, K. Harada, S. Shimanuki, Y. Yamashita,
IEEE Trans. Ultrason. Ferroelect. Freq. Contr. 46, 414 (1999)
8. R.E. Newnham, D.P. Skinner, L.E. Cross, Mat. Res. Bull. 13, 525 (1978)
9. W. Wersing, in Proceedings of 6th IEEE International Symposium on Applica-
tion of Ferroelectrics, p. 212, 1986
10. W.A. Smith, Proc. SPIE 1733, 2 (1992)
11. W. Wersing, K. Lubitz, J. Mohaupt, Ferroelectrics 68, 77 (1986)
12. U. Bast, D. Cramer, A. Wolff, in Proceedings of 7th CIMTEC International
Meeting on Modern Ceramic Technology, p. 2005, 1990
13. R.L. Goldberg, S.W. Smith, IEEE Trans. Ultrason. Ferroelect. Freq. Contr. 41,
761 (1994)
14. G. Percin, A. Atalar, F.L. Degertekin, B.T. Khuri-Yakub, App. Phys. Lett. 72,
1397 (1998)
15. W. Wersing, R. Bruchhaus, in Ferroelectric Film Devices, ed. by M.H.
Francombe, Handbook of Thin Film Devices, vol. 5, (Academic Press, New York,
2000) p. 143
9
High Effective Lead Perovskite Ceramics
and Single Crystals for Ultrasonic Imaging
9.1 Introduction
Some complex lead perovskites, namely Pb(B B )O3 (B = Mg, Zn, Ni, Fe,
Sc, In, and Yb; B = Nb, Ta, and W), are relaxors with the rhombohedral
or monoclinic structure and tetragonal structure with lead titanate PbTiO3
(PT); these perovskites exhibit a morphotropic phase boundary (MPB) and
large electromechanical coupling factor and dielectric constants near the MPB
[1–8]. Table 9.1 shows the Curie temperature (Tc , ), maximum dielectric con-
stant (K), crystal structure, ferroelectric (F) or antiferroelectric (AF) phase,
PbTiO3 mol(%) at the MPB, and Tc at the MPB for various relaxor materials,
along with their abbreviations. Figure 9.1 shows the MPB (I) in PZT and the
MPB (II) in Pb(B B )O3 -PbTiO3 and the MPB (III) in Pb(B B )O3 -PZT.
The locations of the MPB (II) and (III) vary when different B or B ions
are selected as shown in Fig. 9.1. As regards practical piezoelectric appli-
cations, little attention was paid to MPB (II) and (III) compositions until
1980, due to the poor piezoelectric properties of the ceramics, notably their
low electromechanical coupling factors compared to those of PZT ceramics,
and the high cost of the raw materials. Initially, lead magnesium niobate,
Pb(Mg1/3 Nb2/3 )O3 (PMN), lead nickel niobate, Pb(Ni1/3 Nb2/3 )O3 (PNN),
and lead zinc niobate, Pb(Zn1/3 Nb2/3 )O3 (PZN), were considered to be the
three best relaxors.
The present authors have reported large electromechanical coupling factors
of kp > 0.7, k33 > 0.75, and kt > 0.5 near the MPB in Pb{(Sc1/2 Nb1/2 )0.58
Ti0.42 }O3 (PSNT 58/42; in accordance with the standard expression for
Pb(ZrTi)O3 (PZT) composition) ceramic samples [6, 7]. We speculated that
the selection of B and B ions, which tend to have a disordered structure, and
a large PT content in the MPB composition would be necessary to realize a
large kp [8]. We also proposed a new model, which focused on the correlation
of the electromechanical coupling factor and the molecular mass of B-site ions
in Pb(B, Ti)O3 -type perovskite [9,10]. In this model, a low molecular mass for
B-site ions, such as those of Sc, Mg, and Ni, is effective for obtaining materials
224 Y.(J.) Yamashita and Y. Hosono
Table 9.1. Relaxor materials, high Tc materials, and their MPBs with lead titanate
Others
∗
PbZrO3 PZ 513 (240) 3,000 O AF 47 633 (360)
(Pb, La)(Zr, Ti)O3 PLZT <623 (< 350) 30,000 R, T F, AF 35–47 <623
(<350)
BiScO3 BS >673 (>350) <1, 000 R F 64 723 (450)
BiInO3 BIn >973 (>700) <1, 000 R F? 70 843 (570)
PbTiO3 PT 763 (490) 9,000 T F
(Me=Mg,Zn,In,Yb and Sc
Pb (Me,Nb) O3
MPB-II
PZNT 91/09
Rhombohedral Tetragonal
Pb ZrO3 Pb TiO3
MPB-I
Fig. 9.1. Location of morphotropic phase boundary (MPB)-I, II, and III of relaxor-
PT system
4.05
QP region
BaTiO3
4.00
PbTiO3
3.95
(Bik)TiO3
(LaK)TiO3
3.90 SrTiO3
(BiAg)TiO3
ap / A
(LaNa)TiO3 (BiNa)TiO3
(NdNa)TiO3
3.85
CaTiO3
(EuNa)TiO3 (TINa)TiO3
3.80 (TbNa)TiO3
CdTiO3
(DyNa)TiO3
(YNa)TiO3
3.75 (LuNa)TiO3
FE region
3.70
50 100 150 200
Average mass of A ions/a.u.
Fig. 9.2. Effect of ionic size and molecular mass of A-site ions of various ATiO3 on
the ferroelectric properties (after Nakamura et al. [14])
kp (%)
90
Confirmed
Estimated
80
L4
PSNT58/42
PZT53/47
70
8PSN-16PMN-26
PNN-13.5PZ-36. PNNT70/30
5PT-3.5Sr
PMNT68/32 PZNT90/10
60 PSnT56/44
PSTT55/45
PINT63/37
PYbNT50/50
50
PHf T53/47
40
20 40 60 80 100 120
(B-site molecular mass)
Fig. 9.3. Electromechanical coupling factor kp and B-site molecular mass of several
lead perovskite compounds near the MPB [9]
228 Y.(J.) Yamashita and Y. Hosono
35000
B-site modified
Pb [(MeNb),ZrTi]O3 Pb(ZrTi)O3
30000
20000
A-site modified
15000 (Pb,La)(ZrTi)O3
10000
5000
0
0 50 100 150 200 250 300 350 400
Temperature (C)
Fig. 9.4. High-dielectric-constant piezoelectric materials design for A-site and B-site
modification of PZT
Table 9.3. Electrical properties of low and high B-site molecular mass perovskite
compounds of MPB composition [9]
rials used in this study. Three relaxor compositions are selected: PSN, PMN,
and PNN have low molecular mass, from 68.93 to 81.52. In Table 9.2, the
Tc = −150◦ C of PNN was taken from a recent report [13]. Table 9.3 shows low
molecular mass perovskite compounds of the MPB composition and their Tc ’s.
In addition, the MPB composition of PNNT was determined as follows. Many
researchers reported that the 0.50PNN-0.34PT-0.16PZ composition showed
the highest electromechanical coupling factor kp = 70%. If the composition is
regarded as a combination of MPBs of PZT 53/47 and PNNT ##/##, the
MPB of PNNT is calculated as PNNT 71.4/28.6. Then, the MPB of PNNT
71.4/28.6 was applied to further calculations.
Figure 9.5 shows MPB composition lines in the PSN-(PMN-PNN)-PZ-
PT system. In this quaternary pyramid system, a combination of Pb{(Mg1/3
Nb2/3 )0.68 Ti0.32 }O3 (PMNT 68/32, Tc = 150◦ C) and Pb{(Ni1/3 Nb2/3 )0.714
Ti0.286 }O3 (PNNT 71.4/28.6, Tc = 33◦ C) is indicated as Pb{((MgNi)1/3
Nb2/3 )0.70 Ti0.30 }O3 (PMNNT 70/30, Tc = 92◦ C) to locate four lead per-
ovskite compounds in the corner of the pyramid. The MPB composition lines
of the ternary and quartery systems are determined by the binary system MPB
compositions. A plane in the pyramid shows an MPB composition. The MPB
composition in Fig. 9.5 is an electromechanical coupling factor kp peak compo-
sition. A composition corresponding to a peak in dielectric constant is located
at about 1.5 mol% in the TiO2 -rich tetragonal side of the electromechanical
coupling factor kp -peak composition as shown in previous papers [5, 7, 11].
230 Y.(J.) Yamashita and Y. Hosono
PSNT58/42
Tc=260C PMNNT70/30
Tc=92C
Pb(Sc1/2Nb1/2)O3 Pb((Mg,Ni)1/3Nb2/3)O3
Tc=90C Tc=(–10+(–150))/2=–70C
PSNT58/42 PMNNT70/30
Tc=260C Tc=92C
PbTiO3
Tc=490C
Fig. 9.5. MPB composition lines in Pb(Sc1/2 Nb1/2 )O3 -Pb(MgNi1/3 Nb2/3 )O3 -
PbZrO3 -PbTiO3 (PSN-(PMN-PNN)-PZ-PT) system
PSNT 72/28
Tc=200C PMNNT 29/29/42
Tc=200C
Pb(Sc1/2Nb1/2)O3 Pb((Mg,Ni)1/3Nb2/3)O3
Tc=90C Tc=(–10+(–150))/2=–70C
PbTiO3
Tc=490C
PbTiO3
Tc=490C
PZT 53/47
Tc=352C PbZrO3
Tc=240C
PSNT 58/42
Tc=260C PMNNT 29/29/42
Tc=200C
PSNT 72/28
Tc=200C PMNNT 70/30
Tc=92C
Pb(Sc1/2Nb1/2)O3 Pb(Mg,Ni)1/3Nb2/3)O3
Tc=90C Tc=(–10+(–150))/2=–70C
Fig. 9.7. Curie temperature (Tc ) = 200◦ C and MPB composition lines and planes in
Pb(Sc1/2 Nb1/2 )O3 -Pb(MgNi1/3 Nb2/3 )O3 -PbZrO3 -PbTiO3 (PSN-(PMN-PNN)-PZ-
PT) system. Note: the line that cross over two planes is the MPB with Tc = 200◦ C
composition in the multicomponent system. The Tc = 155◦ C plane is obtained when
4.8 mol% Ba and Sr are substituted for A-site Pb
All ceramic specimens showed high densities of more than 8.00 g cm−3 , more
than 98% of the theoretical densities, with an average grain size of 5 µm.
The fired PSN-PMN-PNN-PBSZT samples have a complete single-phase per-
ovskite structure with a mixture of rhombohedral and tetragonal phases. One
9 High Effective Lead Perovskite Ceramics and Single Crystals 233
80.0
75.0
70.0
EMCF kp(%)
65.0
60.0
55.0
50.0
0.33 0.334 0.338 0.342 0.346 0.35 0.354 0.358 0.362 0.366 0.37
TiO 2(x)
6000
After Poling
5000
Dielectric constant
4000
Before Poling
3000
2000
1000
0
0.33 0.334 0.338 0.342 0.346 0.35 0.354 0.358 0.362 0.366 0.37
TiO 2 (x)
Fig. 9.9. Dielectric constant before and after poling as a function of TiO2 (x) con-
tent of (Pb0.962 , Ba0.024 Sr0.024 ){(Sc1/2 Nb1/2 )0.16 (Mg1/3 Nb2/3 )0.17 (Ni1/3 Nb2/3 )0.17
Zr0.5−x Tix }O3 ceramics
Nb2/3 )0.17 (Ni1/3 Nb2/3 )0.17 Zr0.13 Ti0.37 }O3 ceramics. A peak of a planar-mode
kp = 73.6% was found at x = 0.36, which is very close to the calculated value
of x = 0.362. The kp is very large for low-Tc piezoelectric materials.
Figure 9.9 shows the dielectric constant before and after poling as a
function of the TiO2 (x) content of PSN-PMN-PNN-PBSZT ceramics. For
x < 0.35, the dielectric constant after poling is low and for x > 0.35, it is
high. The same result has also been reported for PZT [1] and PSNT [7].
9 High Effective Lead Perovskite Ceramics and Single Crystals 235
900
800
Piezoelectric constant d33 (pC/N)
700
600
500
400
300
200
100
0
0.33 0.334 0.338 0.342 0.346 0.35 0.354 0.358 0.362 0.366 0.37
TiO 2 (x)
Tc=153C
35000
Dielectric constants after poling
30000
25000
20000
15000
10000
5000
0
0 50 100 150 200 250
Temperature(C)
Figure 9.10 shows piezoelectric constants d33 measured by the d33 meter as
a function of the TiO2 (x) content of (Pb0.962 , Ba0.024 Sr0.024 ){(Sc1/2 Nb1/2 )0.16
(Mg1/3 Nb2/3 )0.17 (Ni1/3 Nb2/3 )0.17 Zr0.13 Ti0.37 }O3 ceramics. Piezoelectric cha-
rge constants of more than 800 pC N−1 were obtained when x = 0.36–0.37 for
the PSN-PMN-PNN-PBSZT ceramics.
236 Y.(J.) Yamashita and Y. Hosono
The first large-size PZNT piezoelectric single crystal by flux growth for med-
ical transducer applications was reported by Kobayashi in 1997 [18]. However,
in recent years, single crystal growth has been performed by the solution
Bridgman or direct Bridgman process. Using these processes, it is possible to
control evaporation of PbO at high temperature during crystal growth. The
single crystal growth by the Bridgman process started at about the same time
in Japan, the USA, and China [19]. The Bridgman process furnace has a steep
vertical temperature gradient, more than 20◦ C cm−1 , at around solid/liquid
interfaces determined by the phase diagram. Figure 9.12 shows the large size,
80 mm in diameter PMN-PIN-PT single crystal boule produced by the Bridg-
man process [20]. PMNT and PZNT single crystals of almost the same size
have been successfully grown by other research teams [21–23].
One of the disadvantages of PZNT and PMNT single crystal is that they
have a low Curie temperature Tc and a low rhombohedral to tetragonal phase
change temperature (Trt ) compared to those of PZT ceramics (Table 9.4).
Unlike PZT ceramics, PZNT and PMNT exhibit a phase change temperature
Trt of around 50–120◦ C. Although the Trt values shift depending on poling
conditions, fairly large decreases of k33 and d33 are observed above the Trt [19].
Therefore, in practice, the operating temperature range of these single crys-
tal devices is limited by Trt , not by Tc . The Tc ’s and Trt ’s of PZNT 94/6 to
PZNT 89/11 lie in the ranges of Tc = 161–178◦ C and Trt = 107–47◦ C, and
those of PMNT 70/30 to PMNT 67/33 in the ranges of Tc = 140–155◦ C and
9 High Effective Lead Perovskite Ceramics and Single Crystals 237
Trt = 87–47◦ C. These Trt and Tc ’s are sufficiently high for naval or medical
applications where the operating temperatures are usually below 60◦ C. How-
ever, they are unsuitable for high-temperature applications such as actua-
tors for outdoor use. Therefore, new material systems with Tc > 180◦ C and
Trt > 100◦ C are required for high-temperature applications. Using the binary
system of xPb(In1/2 Nb1/2 )O3 -yPbTiO3 (PINT 100x/100y) and the ternary
system of xPb(In1/2 Nb1/2 )O3 -yPb(Mg1/3 Nb2/3 )O3 -zPbTiO3 (PIN-PMN-PT
100x/100y/100z), attempts have been made to grow single crystal of sufficient
size. Yasuda et al. reported that PINT single crystals of fairly good quality,
more than 10 × 10 × 3 mm3 in size, were obtained by the flux method. The
PINT single crystal shows a Tc of 260◦ C, which is 80◦ C higher than that of the
PZNT 91/9 single crystal. A Trt is observed near 100–150◦ C. The PSN-PMN-
PT single crystal shows a peak dielectric constant maximum at Tc = 200◦ C
at 1 kHz, which is 50◦ C higher than that of the PMNT 67/33 single crystal.
A small peak of dielectric constant is observed near 120◦ C, which is consid-
ered to be the Trt from rhombohedral to tetragonal phase. The PINT shows
excellent temperature dependence of piezoelectric properties [19].
Recently, the present authors reported piezoelectric properties of xPb(In1/2
Nb1/2 )O3 -yPb(Mg1/3 Nb2/3 )O3 -zPbTiO3 (PIN-PMN-PT 100x/100y/100z)
single crystal [21]. Figure 9.13 shows the dielectric constants before poling
of the PIN-PMN-PT 16/51/33 and 24/51/33 single crystals. Small peaks of
the dielectric constant are observed at 50–90◦ C, corresponding to the phase-
Table 9.4. Comparison of several relaxor-PT single crystals and PZT type ceramics
for transducer applications
Materials Ti Tc (◦ C) K33 k33 (%) k31 (%) d33 d31 g33 Size
(mol%) (pC/N) (pC/N) (10−3 Vm N−1 ) (mm)
PZNT 5–10 150–175 4,000 95 81 2,400 −1, 700 60 50
PMNT 28–33 135–155 4,500 93 61 2,500 −1, 000 50 70
PSNT 40–44 240–260 2,000 (94a ) (80a ) (1,500) (−1,000) (>40) <10
PINT 33–38 300–320 2,000 (85a ) (>50) (1,000) (−700) (>40) <15
PIN- 30–35 180–250 3,100 90 60 2,200 −1,000 60 25
PMN-PT
PZT 5Hb 40–47 210 3,200 75 40 700 −350 20 >100
Relaxor- 35–38 135–155 5000–7200 78 42 800–900 −380 20 >100
PZT high
Kb
a
Predicted value
b
Ceramics
40000
Dielectric constants ε33T/ε 0
30000
PIMNT 24/42/34
20000
10000
PIMNT 16/51/33
0
0 50 100 150 200 250
Temperature (C)
Table 9.5. Electrical properties of the obtained and ideal single crystal for medical
array transducers
The Bridgman process and top seeded solution growth (TSSG) process are well
established single crystal growth processes [20–23]. However, these processes
have several shortcomings, such as Pt crucible cost and composition homo-
geneity. To overcome these drawbacks, a new approach has been studied. Solid
state crystal growth (SSCG) and templated grain growth (TGG) processes
have many advantages. The reasonable size and quality of SSCG process for
lead-relaxor materials was reported by researchers in Korea and USA [19].
Since these processes have many advantages compared to the conventional
single crystal process, they are expected to lead to the reduction in the cost
of relaxor single crystals for mass production in the near future.
9.6 Applications
Medical echo diagnostic ultrasound systems use ultrasonic array transducers
for their applications. Frequencies used for these applications are 2–20 MHz.
100
Coupling factorsk 33′ (%)
90 PIMNT24/42/34
Trt=89C
80
70
PIMNT16/51/33
60 Trt =52C
50
0 20 40 60 80 100 120
Temperature (C)
The typical cardiac transducer operates at a center frequency of 2–4 MHz and
makes use of a piezoelectric material plate with a size of 15 × 25 mm2 . Re-
cently, medical echo ultrasound has been greatly improved. The color Doppler
technique and tissue harmonic imaging are frequently used for functional di-
agnosis as well as for ultrasonic tomography (B mode). However, clearer im-
ages are required for early and reliable diagnosis of human organs with small
tumors. This requires a wider bandwidth of probes as well as improved sen-
sitivity. The key for achieving this is the performance of the piezoelectric
material which transmits and receives the echo ultrasound. High dielectric
constant type PZT5H ceramics have been mostly used for this application
in recent decades. However, recent research has focused on obtaining greater
sensitivity and broader bandwidth, and new transducers such as piezoelec-
tric composites have been studied. Single crystals of PZNT 91/9 and PMNT
70/30 have a large d33 (>2,000 pC N−1 ) and k33 (>92%), k33 (>80%), and
g33 (>60 × 10−3 Vm N−1 ), values far larger than those obtained for any PZT
ceramic specimens. This is one reason that intensive research on medical trans-
ducers has been conducted in various parts of the world during the last 15
years. After the initial reports and presentations by Toshiba in the early 1990s,
several organizations in Korea and the USA entered the field [25, 26]. Saitoh
et al. reported the first images of B-mode and Doppler-mode obtained by us-
ing PZNT 91/09 single crystal for cardiac transducers operating at 3.5 MHz
with 96 channels [27]. The echo amplitude of the PZNT probe is about 6 dB
higher than that of the PZT probe, and the frictional bandwidth is 30%
wider. This means that both the penetration and the resolution of the PZNT
probe are superior to those of conventional PZT probes. Large coupling factor
k33 (>82%), dielectric constant after poling (K > 3,000), and low acoustic
impedance Z33 < 24 × 106 kg m−2 s−1 make PZNT 91/9 single crystals an
excellent transducer material for medical diagnostic applications. At present,
the size of PZNT and PMNT single crystal is sufficient for various applica-
tions. However, the quality and uniformity of these crystals within and be-
tween the wafers, as well as among lots and among manufacturers, are still
insufficient [19]. Although it is usually easy to grow PZNT and PMNT sin-
gle crystals in [111] or [110] directions, it is very difficult to grow them in
[001] direction [28]. Since [001] wafers are used for medical transducers, there
is always some TiO2 distribution within wafers. This is one of the causes
of large scattering of k33 , d33 , and dielectric constant within wafer. In ad-
dition, the manufacture of high-frequency single crystal transducers, that is,
more than 7 MHz, requiring thin plate (less than 150 µm), is very difficult
because of the low-frequency constant and low mechanical strength of single
crystal. The present cost of single crystal wafers is more than 10 times that of
PZT ceramics. Other drawbacks are temperature instability, poor mechanical
properties, chipping bydicing, nonuniform bonding strengths between match-
ing layers, electrodes, and crystal surfaces, low clamped dielectric constants,
low Ec , depoling effects, low production yield, etc. So, at present, although
single crystals have properties which are very attractive compared to those of
242 Y.(J.) Yamashita and Y. Hosono
PZT ceramics, the supply of single crystal medical array transducers is limited
to only a few probe manufacturers [25, 26].
Figure 9.15 shows the first commercialized 2.6 MHz PMNT single crystal
medical probe manufactured by Humanscan Co. of Korea. The 6 dB band-
width of the single crystal probe is about 100%, which is larger than that of
conventional PZT probe of 80% [25].
9.7 Conclusions
1. A smart material composition design for realizing superior dielectric and
piezoelectric properties in lead-based perovskite compounds has been pro-
posed. The key to the composition design is the selection of Tc and the
exact MPB composition of TiO2 content in the complex lead–perovskite
system ceramics Pb(Sc1/2 Nb1/2 ))O3 -Pb(Mg1/3 Nb2/3 )O3 -Pb(Ni1/3 Nb2/3 )
O3 -(Pb, Ba, Sr)(Zr, Ti)O3 . On the basis of this composition design, a
proper composition of complex lead–perovskite system ceramics PSN-
PMN-PNN-PBSZT near the MPB was proposed. A small amount of Ba
and Sr, less than 5 mol%, for the Pb site was replaced to stabilize the per-
ovskite structure of these ceramics. The (Pb0.962 , Ba0.024 Sr0.024 ) {(Sc1/2
Nb1/2 )0.16 (Mg1/3 Nb2/3 )0.17 Pb(Ni1/3 Nb2/3 )0.17 Zr0.13 Ti0.37 }O3 ceramics
showed a pure perovskite structure with an average grain size of 5 µm
and a density of 8.00 g cm−3 . A dielectric constants of ε33 T /ε0 = 5, 500,
a planar-mode electromechanical coupling factor of kp = 72%, d33 =
850 pC N−1 (measured using a d33 meter), and Tc = 153◦ C were obtained.
Thus, an effective approach for the smart material composition design of
the complex lead–perovskite structure for high-performance piezoelectric
materials was verified.
Fig. 9.15. The first commercialized 2.6 MHz PMNT single crystal medical probe
(Courtesy of Humanscan Co. of Korea, see [25])
9 High Effective Lead Perovskite Ceramics and Single Crystals 243
2. Sufficiently large PZNT and PMNT piezoelectric single crystals, that is,
25–80 mm in diameter, for various applications, such as medical transduc-
ers, actuators, sonar, and hydrophones, are available.
3. Process capabilities, such as polishing, electroding, dicing, lead-attaching,
and the assembling of 2–4 MHz medical cardiac single crystal transducers,
are almost established. High quality images on screens of medical echo
systems have been confirmed by several echo equipment manufacturers.
4. Quality and uniformity level between and within wafers, lot to lot and
among manufacturers of single crystal made by the (solution) Bridgman
process are still inferior to those of PZT ceramics.
5. Design and manufacture of high-frequency single crystal array transduc-
ers, that is, higher than 7 MHz, is very difficult due to the low frequency
constant and mechanical strength of single crystal.
6. New high Tc single crystals of In-doped PMNT, PIN-PMN-PT 24/42/34,
would be very promising for the ideal single crystal if single crystals of
sufficient size and quality were grown.
7. Some other drawbacks of single crystal array transducers have been
pointed out. So, at present, single crystals have yet to win acceptance
in the mass production of medical array transducers except at a few
probe manufactures. Nevertheless, it is quite possible that single crystals
and/or single crystal-like ceramics may become the predominant materi-
als for sophisticated, performance-oriented piezoelectric products in the
near future.
8. Regarding future tasks, research on methods of evaluating single crystal
wafer quality and uniformity using nondestructive processes is urgently
required. Improvement in the uniformity of capacitance, electromechanical
coupling factors k33 , and piezoelectric constant d33 within and between
wafers is also a major task for crystal growers. Detailed basic research
to determine the mechanisms accounting for the scattering of electrical
and acoustical properties should be performed from the viewpoints of
composition, domain structure, and defect chemistry. Finally, research on
new single crystal materials and single crystal growth technology to solve
all the present problems is essential. From the viewpoint of applications, a
new dicing process capable of preventing cracking of fragile single crystal
transducer is also necessary to make fine pitch array transducers.
References
1. B. Jaffe, Piezoelectric Ceramics (Academic Press, London, 1971) p. 200
2. J. Kuwata, K. Uchino, S. Nomura, Jpn. J. Appl. Phys. 21(9), 1298 (1982)
3. S.-E. Park, T. R. Shrout, IEEE Trans. Ultrason. Ferroelect. Freq. Contr. 44,
1140 (1997)
4. H. Wang, B. Jiang, T. R. Shrout, W. Cao, IEEE Trans. Ultrason. Ferroelect.
Freq. Contr. 51, 907 (2004)
244 Y.(J.) Yamashita and Y. Hosono
D. Cathignol
10.1 Introduction
The usual models used to predict the field radiated by the active surface of
a transducer propose the hypothesis that the latter vibrates in a homoge-
neous fashion, i.e., the amplitude and phase of displacement are the same at
any point of the emitting surface. This hypothesis is not valid in the case
of a piezoceramic. In fact the vibrating behaviour in thickness mode of such
transducers is disturbed by the existence of the plate and the surface waves
propagating from the edges of the ceramic layer. For example, Fig. 10.1 shows
that the on- axis pressure distribution differs greatly from the prediction of the
Rayleigh integral in the case of a spherical shell made of piezoceramic material.
In contrast, it is very close to the theoretical curve when the piezocomposite
transducer is used. Chapelon et al. in [5] demonstrated the predictability of
the radiation pattern emitted by a truncated annular array-type piezocom-
posite transducer where the acoustic field on the axis was measured for several
electronic focusing configurations and compared to a theoretical simulation.
Despite the probable effects of several factors that were not taken into ac-
count in the simulation (nonlinear effects of the propagation, directivity of
the hydrophone, etc.), good radiation predictability is evident, allowing for
the development of transducers with complex geometries.
1
0.8
P/Pmax
0.6
0.4
0.2
0
(a) 0 20 40 60 80 100 120
1
0.8
P/Pmax
0.6
0.4
0.2
0
(b) 0 20 40 60 80 100 120
1
0.8
P/Pmax
0.6
0.4
0.2
0
0 20 40 60 80 100 120
(c) Propagating axis (mm)
Plate
waves
Focus
Additional
signals radiated
by plate
waves
(d)
Fig. 10.1. Comparison of the pressure distribution along the axis between piezoce-
ramic (a) and piezocomposite (b). The former differs greatly from the theoretical
value (c). Plate waves and surface waves propagating from the edges are responsible
for the grating lobes in a
10 High-Power Ultrasound Transducers for Therapeutic Applications 249
intensity of 4.6 W cm−2 . The intensity limitation due to the cavitation thresh-
old is thus relaxed, and high-intensity focused ultrasound treatments can be
performed with greater precision in terms of aim and better protection of the
surrounding tissues [7].
(a)
(b)
Fig. 10.2. Piezoceramic and piezocomposite transducers after aging tests. Piezoce-
ramic brakes in several pieces while piezocomposite are destroyed by piercing
Xr2
Backing Load
Xr1
Pulse at
t=to
Pulse at
t=to+e/c
pressures because of the reduced significance of the radial modes, the maxi-
mum pressure does not exceed ∼3 MPa for long ageing tests. However,these
values still appear to be too weak for the creation of a shock-wave genera-
tor of very small diameter. For this reason, another method was investigated:
stacking the transducers [11]. When excited at particular time intervals, the
pressure waves generated by each element of the stacked transducers reinforce
each other at the transducer-to-load interface. For example, in Fig. 10.3, two
stacked piezoelectric transducers are shown. If the first transducer is excited
at time t0 , the pressure wave reaches the transducer-to-load interface at time
10 High-Power Ultrasound Transducers for Therapeutic Applications 251
An alternative with the same aim of increasing the transducers’ surface pres-
sure was proposed, called prestrain. When a piezoelectric material is subjected
to an electrical field, its thickness varies. When the field is in the same or op-
posite direction to the polarisation, a compressive or an extended wave is
generated in the adjacent medium. For shock-wave generation, because a very
strong compression wave has to be generated, the piezoelectric material is
subjected to a major extension. Piezoelectric material obtained by sintering,
which constitutes the active piezocomposite plots, withstands compression
very well but extension very poorly. The principle of prestraining the mater-
ial [13] is as follows (Fig. 10.4). First, an electric field in the opposite direction
to that of the polarisation vector is applied, causing the material to contract,
and second, a transient field in the same direction as the polarisation vector
is superimposed, generating the compression wave. The resulting extension,
when the transient field is applied, compared with what would have been
obtained using conventional excitation, is reduced by the value of the com-
pression induced by the first field applied. The prestrain may be applied in
three different manners as illustrated in Figs. 10.5a–c. Applying an electrical
field in the opposite direction of the polarisation vector leads to a depolarisa-
tion of the material, even in hard materials. Thirdly, the voltage step function
is symmetric with respect to the zero voltage and returns to the base line with
a large time constant. This part should contribute to repole the transducer
and to allow an increase of the electrical prestrain. The compressive wave
252 D. Cathignol
P
(a)
–U
(b)
–U
(c)
C +HT –HT C
S2 S1
R2 R1 R
L
(d) t3 Xr
t2
t0 t
1
(e) t4
Fig. 10.4. Principle of the electrical prestrain (a–c). The electrical circuit (d) and
the chronogram (e)
250 ms
Time
0
(1)
Prestrain (2)
(3)
(a)
3
Pressure values (MPa)
2.5
2
1.5
Mode 1
1
Mode 2
0.5
Mode 3
0
0 1000 2000 3000 4000 5000 6000 7000
(b) Voltage (Volts)
Fig. 10.5. The three different modalities tested (a). Only the third one does not
depolarise the material. Maximum pressure values obtained for the three different
modalities (b)
focus (measured with an optical fibre 100 µm in diameter and with a band-
width of 30 MHz) is 60 MPa. The focal zone measured at −3 dB is 15 mm long
in the propagating axis and ±1.5 mm in the off- axis. The disintegration effi-
cacy was tested on standard kidney stone-mimicking plaster balls, 15 mm in
diameter, made with calcium sulphate with hollow glass microspheres 100 µm
in diameter (HMT, Kreuzlingen, Switzerland). Similar to the case of an elec-
tohydraulic generator, for a focal pressure of 60 MPa, the number of shocks
to disintegrate the plaster balls is about 150 which is considered to be the
golden standard.
10.5 Conclusion
The evolution of ultrasonic therapy and the growing interest that it arouses
make it increasingly necessary to take into account the specific nature of
each application. The applications of piezocomposite technology, which have
been discussed, show the significance of this technology. This is particularly
obvious if we consider all the needs that must be satisfied in the design of a
transducer for therapeutic applications. The specific properties of composite
materials and the combination of these properties bring to transducer design
a flexibility and a degree of optimisation which are often lacking in designs
based on ceramic materials. This flexibility and this degree of optimisation
form the basis for new solutions, thus responding to the diverse requirements
of therapeutic ultrasound.
References
1. W.A. Smith, B.A. Auld. IEEE Trans. Ultrason. Ferr. 38(1), 40–47 (1991)
2. J.A. Hossack, G. Hayward,IEEE Ultrason. Ferr. 38(6), 618–629 (1991)
3. W. Cao, M. Zhang, L.E. Cross, J. Appl. Phys. 72(12), 5814–5821 (1992)
4. J.Y. Chapelon, D. Cathignol, C. Cain, E. Ebbini, J.U. Kluiwstra,
O.A. Sapozhnikov, G. Fleury, R. Berriet, Ultrasound Med. Biol. 26(1), 153–
159 (2000)
5. J.Y. Chapelon, F. Prat, A. Arefiev, D. Cathignol, R. Souchon, Y. Theillère,
IEEE Ultrason. Symp. Proc. 1265–1268 (1996)
6. F. Dupenloup, J.Y. Chapelon, D. Cathignol, O.A. Sapozhnikov, IEEE Ultrason.
Ferr. 43(6), 991–998 (1996)
7. J.Y. Chapelon, F. Dupenloup, H. Cohen, P. Lenz, IEEE Ultrason. Ferr. 43(4),
623–625 (1996)
8. P.H. Thibault. Ann. Urol. 20, 20–25 (1986)
9. J.P. Sferruzza, A. Birer, A. Matias, Y. Theillère, D. Cathignol, Sens. Actuators
A Phys. 88, 146–155 (2001)
10. E. Closset, P. Trompette, A. Birer, D. Cathignol, IEEE Ultrason Ferr. (to be
published)
11. J.P. Sferruzza, A. Birer, D. Cathignol, Ultrasonics 38, 965–968 (2000)
10 High-Power Ultrasound Transducers for Therapeutic Applications 255
K. Uchino
11.1 Introduction
Piezoelectric ceramics forms a new field between electronic and structural
ceramics [1–4]. Application fields are classified into three categories: position-
ers, motors, and vibration suppressors. From the market research result for
80 Japanese component industries in 1992, tiny motors in the range of 5–
8 mm are required in large numbers for office and portable equipment; the
conventional electromagnetic (EM) motors are rather difficult to produce in
this size with sufficient energy efficiency, while Silicon MEMS actuators are
too small to be used in practice. Piezoelectric ultrasonic motors whose effi-
ciency is insensitive to size are superior in the millimeter motor area. The
manufacturing precision of optical instruments such as lasers and cameras,
and the positioning accuracy for fabricating semiconductor chips are of the
order of 0.1 µm which is much smaller than the backlash of the EM motors.
Vibration suppression in space structures and military vehicles also require
compact but mighty piezoelectric actuators.
On the other hand, one of the bulkiest components in information process-
ing equipment (such as laptop computers) is the power supply component,
especially the electromagnetic transformer used in the power supply. Losses
such as skin effect, thin wire and core losses of the electromagnetic trans-
former increase rapidly as the size is reduced. Thus, high efficiency, small size,
and absence of electromagnetic noise are some of the attractive features of
piezoelectric transformers (PT), making them more suitable for miniaturized
power inverter components used in lighting up the cold cathode fluorescent
lamp behind a color liquid crystal display or generating high voltage for air-
cleaners [5].
The advantages of piezoelectric devices over electromagnetic types are
summarized (see Fig. 11.1):
(a) More suitable to miniaturization – As the stored energy density is larger
than that of an EM type, is 1/10 smaller in volume, and weight can be
achieved
258 K. Uchino
90 %
Piezoelectric
device
30 %
Efficiency
10 % Electromagnetic
device
3%
30 mW 30 W 30 kW
Power
Fig. 11.1. Efficiency vs. power relation for electromagnetic and piezoelectric motors
6.4
PZT(100) 5.6
5.2
e31eff
4.8 [C/m2]
PZT(111)
4.4
Fig. 11.2. Experimental results for the piezoelectric e31 constants in epitaxially
grown PZT thin films with a rhombohedral composition 70/30 [11]
been used for more than 10 years until high k materials have gained atten-
tion in medical acoustics. These data have been reconfirmed, and improved
data were obtained recently, aiming at medical acoustic applications [8, 9].
The strains as large as 1.7% can be induced practically for the PZN-PT solid
solution single crystals. It is notable that the highest values are observed for
a rhombohedral composition only when the single crystal is poled along the
perovskite [001] axis, not along the [111] spontaneous polarization axis.
A series of theoretical calculations made on perovskite type ferroelectric
crystals suggests that large d and k values in similar magnitudes to PZN-PT
can also be expected in PZT. Crystal orientation dependence of piezoelectric
properties was phenomenologically calculated for compositions around the
morphotropic phase boundary of PZT [10]. The maximum longitudinal piezo-
electric constant d33 (4–5 times enhancement) and electromechanical coupling
factor k33 (more than 90%) in the rhombohedral composition were found to
be at an angle of around 57◦ , canted from the spontaneous polarization direc-
tion [111], which corresponds roughly to the perovskite [100] axis. Figure 11.2
shows the experimental results for the piezoelectric e31 constants in epitaxi-
ally grown PZT thin films [11]. Note that the maximum e31 constant can be
obtained in the rhombohedral phase near the morphotropic phase boundary,
and in the [100] specimen (rather than in the [111] specimen). Damjanovic
et al. reported similar supporting results [12].
high Curie temperature piezoelectric materials. Shrout et al. reported the solid
solutions of BiScO3 , BiInO3 and BiYbO3 with PbTiO3 with the morphotropic
phase boundary Curie temperature more than 450◦ C [13].
To the contrary, cryogenic actuator materials are required for space ap-
plications such as “Hubble” telescope adjustment. Mulvihill et al. reported
PZT-based multilayer actuators for this purpose [14].
critical vibration level, which is the limit of the output vibration energy for
that particular PZT. Even if the input electrical energy is increased further,
the additional energy will be converted only into heat.
At present, the highest value 1 m s−1 of the maximum vibration ve-
locity (defined at 20◦ C rise from room temperature) can be obtained in
Pb(Zr, Ti)O3 -Pb(Mn1/3 X2/3 )O3 (X = Sb, Nb)-doped with rare-earth ions
(Yb, Eu) [19, 20]. This velocity is more than three times larger than that in
commercially available hard PZTs, leading to a power density capability more
than ten times. The loss mechanisms in piezoelectrics have been gradually clar-
ified [21]. The heat generation under off-resonance is basically originated from
intensive dielectric loss, while under resonance it is caused by intensive me-
chanical loss, where both intensive (experimentally obtained apparent) dielec-
tric and mechanical losses are inter-related with extensive (material’s proper)
dielectric and mechanical losses. Further, the resonance loss Qm −1 at a small
vibration velocity is mainly determined by the extensive mechanical loss, and
with increasing vibration velocity, the extensive dielectric loss contribution
significantly increases. Microscopically, under a small electric field the origin
of loss will be primarily non-180◦ domain wall motion (extensive mechanical
loss) and the contribution of 180◦ domain wall motion (extensive dielectric
loss) will significantly increase above a certain vibration level.
Relating to the hysteresis in piezoelectrics, Smith proposed a computer
simulation to evaluate the polarization and strain hysteresis under an electric
field cycle, one example of which is shown in Fig. 11.3 [22].
Fig. 11.3. Comparison among the model calculation (dashed ) and experimental
polarization (solid ) as a function of applied electric field for PZT 5A [22]
262 K. Uchino
others
6%
13%
Tungsten bronze
type compound 34%
Bi-layered type
compound
(Na,K)Nbo3 type
compound
20%
(Bi1/2Na1/2)Tio3 type
compound
27%
Fig. 11.5. A photo-driven micro walking device, designed to begin moving by light
illumination. It is simple in structure, having neither lead wires nor electric circuitry
plate. When a violet light is irradiated to one side of the PLZT bimorph, a
photovoltage of 1 kV mm−1 is generated, causing a bending motion. The tip
displacement of a 20 mm bimorph 0.4 mm in thickness was 150 µm, with a
response time of 1 s.
A photodriven micro walking device, designed to begin moving by light
illumination, has been developed (shown in Fig. 11.5) [26]. It is simple in
structure, having neither lead wires nor electric circuitry, with two bimorph
legs fixed to a plastic board. When the legs are irradiated alternately with
light, the device moves like an inchworm with a speed of 100 µm min−1 .
In pursuit of thick film type photostrictive actuators for space structure
applications, in collaboration with Jet Propulsion Laboratory, Penn State
investigated the optimal range of sample thickness and surface roughness de-
pendence of photostriction. 30 µm thick PLZT films exhibit the maximum
photovoltaic phenomenon [27].
Two of the most popular actuator designs are multilayers and bimorphs.
The multilayer, in which roughly 100 thin piezoelectric/electrostrictive ce-
ramic sheets are stacked together, has advantages such as low driving volt-
age (100 V), quick response (10 µs), high generative force (100 kgf) and high
electromechanical coupling. But the displacement in the range of 10 µm is not
sufficient for some applications. This contrasts with the bimorph, consisting
of multiple piezoelectric and elastic plates bonded together to generate a large
bending displacement of several hundred µm, but the response (1 ms) and the
generative force (100 gf) are low.
A multilayer actuator with interdigital internal electrodes has been devel-
oped by Tokin [28]. In contrast to the conventional electrode configuration,
line electrodes are printed on piezoelectric green sheets, and are stacked so
that alternating electrode lines are displaced by one-half pitch. This actuator
generates motions at right angles to the stacking direction using the longi-
tudinal piezoelectric effect. Long ceramic actuators up to 74 mm in length
were manufactured, which could generate the displacement up to 55 µm. A
three-dimensional positioning actuator with a stacked structure has been pro-
posed by PI Ceramic, in which shear strain is utilized to generate x and y
displacements [29].
A monomorph device has been developed to replace the conventional bi-
morphs, with simpler structure and manufacturing process. The principle is a
superposed effect of piezoelectricity and semiconductivity [30]. A monomorph
plate with 30 mm in length and 0.5 mm in thickness can generate 200 µm tip
displacement, in equal magnitude of that of the conventional bimorphs. The
“rainbow” actuator by Aura Ceramics [31] is a modification of the above-
mentioned semiconductive piezoelectric monomorphs, where half of the piezo-
electric plate is reduced so as to make a thick semiconductive electrode to
cause a bend.
A market research held in 1992 revealed that the customers’ demands for
actuators can be satisfied neither with multilayers nor bimorphs. The demands
are 100 µm in stroke, 100 N in force with 100 µs in response speed. A composite
actuator structure called the “cymbal” has been developed at Penn State to
provide characteristics intermediate between the multilayer and bimorph actu-
ators; this transducer exhibits an order of magnitude larger displacement than
the multilayer, and much larger generative force with quicker response than
the bimorph [32, 33]. A Cymbal with a thickness 2 mm and a diameter 12 mm
can generate a displacement up to 100 µm. The Cymbal has been applied to
make a miniaturized laser beam scanner. Ring-morphs are another alternative
to a displacement amplification mechanism [34]. By eliminating unnecessary
ceramic center portion of the bimorph/unimorph, 30–40% enhancement of the
field-induced displacement was realized.
11 Piezoelectric Motors and Transformers 265
Modified from the original Sashida’s surface wave type, Canon and Seiko In-
struments have successfully commercialized camera automatic focusing mech-
anisms and date change mechanisms in a wrist watch, respectively.
However, a significant problem in miniaturizing this sort of traveling wave
motor is found in the further miniaturization process down to 8 mm φ; without
providing a sufficient buffer gap between the adjacent electrodes, the electrical
poling process easily initiates cracks on the electrode gap due to the resid-
ual stress concentration. Uchino et al. proposed the following concepts for
designing compact ultrasonic motors (a) simplification of the structure and
reduction in the number of components, (b) the use of simple (i.e., uniform)
poling configuration (i.e., gentle to PZT ), and (c) the use of standing-wave
type for reducing the number of drive circuit components.
A Windmill motor design with basically a flat and wide configuration
(pancake type), using a metal–ceramic composite structure is the first good
example [35], followed by a thin and long configuration, a metal hollow tube
motor (needle type). It is composed of a metal tube and two rectangular PZT
plates bonded on it (see Fig. 11.6) [36, 37]. When one of the PZT plates is
driven by a sinusoidal voltage, the metal tube exhibits “wobbling” motion,
like a hula-hoop or Chinese dish spinning. A no-load speed of 1,800 rpm and
an output torque of more than 1 mNm were obtained at 80 Vpeak for both
directions for a motor 2.4 mmφ in diameter and 12 mm in length. This signif-
icantly high torque was obtained due the dual stator configuration and the
high pressing force between the stator and rotors made of metal.
Finite Element Analysis (FEA) is a useful tool to obtain the vibration
mode simulation in a stator. The Penn State used an ATILA software code
(developed by ISEN, and distributed by Micromechatronics, Inc., PA) for the
motor stator calculation in linear and rotary motors, and found the resonance
Y
Elastic hollow
cylinder
x' Plate X
y'
Plate Y
(a) (b)
Fig. 11.6. “Metal tube” motor using a metal tube and two rectangular PZT plates.
(a) Schematic structure and (b) photo of a world-smallest motor (1.5 mmφ) [37]
266 K. Uchino
Top Bottom
Electrode PZT IDTs Electrode
Sio2/Si
Inlet Inlet
Glass
wafer outlet
frequency discrepancy to be less than 5% [38]. The design optimization trial for
the ultrasonic motors using ATILA, in conjunction with the so-called genetic
algorism, seems to suggest a future direction of the design algorism [39].
PZT thin films are deposited on a silicon wafer, which is then micro-machined
to leave a membrane for fabricating micro actuators and sensors, i.e., micro-
electromechanical systems (MEMS). Figure 11.7 illustrates a blood tester de-
veloped by The Penn State in collaboration with OMRON Corporation in
Japan [40]. When a voltage is applied to two surface interdigital electrodes,
the surface PZT film generates surface membrane waves, which soak up the
blood and the test chemical from the two inlets, then mix them in the cen-
ter part, and send the mixture to the monitor part through the outlet. FEA
simulation was conducted to evaluate the flow rate of the liquid by changing
the thickness of the PZT or of the Si membrane, inlet and outlet nozzle size,
cavity thickness.
The rise time adjustment is also very important from the heat generation
viewpoint. The temperature rise was examined in consideration of the pulse
drive using a trapezoidal wave, as in the applications to diesel engine injec-
tion systems. The trapezoidal wave with a maximum voltage of 100 V, 60 Hz,
and a duty ratio of 50% was applied to a commercialized multilayer actuator
(product by Ceramtech). The temperature rise is surprisingly dependent on
the step rise time, and this, in turn, is basically due to the actuator’s vibration
overshoot and ringing [unpublished].
P [W]
100W STEP-UP STEP-DOWN
TRANSONER•
LOW POWER HIGH POWER
30W
20W
Ultrasonic
motor
Chopper
Rectifier
Battery Piezo-
Low transformer High Piezoelectric
voltage voltage
input output actuator
(a) (b)
Fig. 11.10. (a) Commercialized Shinsei motor and its drive circuit and (b) piezo-
transformer integrated ultrasonic motor (Penn State) [45]
11 Piezoelectric Motors and Transformers 269
300VDC
S1 D5
D1 D2
C L
Driving
Battery TIMER D3 D4 Circuit S2
D6
24 555
DCV (70kHz)
Helicopter
Ground Floating
High Power RECTIFIER INVERTER ACTUATOR
OSCILLATOR Piezoelectric
Transformer
Fig. 11.11. Compact drive system for piezoelectric actuator control [46]
Fig. 11.12. Piezoelectric windmill (a) for driving an LEC traffic light array system
(b) (courtesy by NEC-Tokin)
11.5 Applications
Commercialized products seem to come 5–7 years after the basic key technol-
ogy is developed in the case of piezoelectric devices, which can be found in
Table 11.1.
MLChips
Upper electrode
PZT
Lower electrode 0.5 mm
Vibration plate
Chamber plate
Communication plate
Ink pass
Adhesive layer
Ink supply hole
Stainless plate
Ink chamber
Nozzle
Fuel inlet
Piezoelectric actuator
Needle group
Fig. 11.14. Common rail type diesel injection valve with a piezoelectric multilayer
actuator (News release by Bosch)
Fig. 11.15. A piezoelectric pump water cooling module (left) for a fuel cell in a
laptop computer (Courtesy by NEC)
Bolt coupled
Horn type L-L converter Langevin transduser
Driving face
w PZT
Support
Radiating face
Fig. 11.16. L–L coupler horn for a washing machine application (courtesy by Honda
Electronics)
11 Piezoelectric Motors and Transformers 273
Output End
Bearing Surface ro
ro
dz
Lateral
Normal
(dz)
Eccentuator
Input Rotation
Ultrasonic Motor
Fig. 11.17. Wing shape control with an ultrasonic motor and an eccentuator mech-
anism (courtesy by AFRL)
Shinsei ultrasonic motors have been adopted for the US Air Force Smart
Wing Program. Coupled with an eccentuator mechanism, the USM functions
to control the wing shape (morphing) (Fig. 11.17).
Seiko micro motors have already been used in wrist watches for silent
alarm and perpetual function. The most recent application can be found in a
microaerial vehicle. Figure 11.18 shows a micro flying robot, a helicopter of
the size 13 cmφ, 7 cm height and 8.9 g weight, in which the floating force is
obtained by two piezoelectric motors and additional two motors provide the
position/cant control.
The Penn State metal tube motors have been utilized in the world’s small-
est 4WD vehicle (Fig. 11.19) and the camera auto zooming and focusing mech-
anism in a Samsung cellular phone (Fig. 11.20).
Future research trends will be divided into two ways: up-sizing in space struc-
tures and down-sizing in office equipment. Further down-sizing will also be
required in medical diagnostic applications such as blood test kits and surgi-
cal catheters. The development keywords will be integration and hybridization.
Piezoelectric thin films compatible with silicon technology will be much
focused in MEMS systems. An ultrasonic rotary motor as tiny as 2 mm in
274 K. Uchino
Motor 1 Motor 2
Fig. 11.19. World-smallest vehicle with two micro metal tube USMs. The size is
7 × 7 × 7 mm3
Fig. 11.20. Camera auto zooming/focusing mechanism with two metal tube USM’s
in a Samsung cellular phone
References
1. K. Uchino, Piezoelectric Actuators and Ultrasonic Motors (Kluwer Academic,
Norwell, MA, 1996)
2. K. Uchino, Ferroelectric Devices (Marcel Dekker, New York, 2000)
3. K. Uchino, MRS Bull. 18(4), 42 (1993)
276 K. Uchino
36. B. Koc, S. Cagatay, K. Uchino, IEEE Ultrason. Ferr. Trans. 49(4), 495–
500 (2002)
37. S. Cagatay, B. Koc, K. Uchino, IEEE Trans. UFFC, 50(7), 782–786 (2003)
38. P. Bouchilloux, K. Uchino, in Proceedings of the SPIE: The International Society
for Optical Engineering, USA, vol. 4693, 2002
39. P. Bouchilloux, K. Uchino, J. Intell. Mater. Syst. Struct. 14(10), 657–667 (2002)
40. S. Kalpat, Ph.D. Thesis, Penn State University, Fall, 2001
41. K. Uchino, J. Giniewicz, Micromechatronics (Marcel Dekker, New York, 2003)
42. S. Kawashima, O. Ohnishi, H. Hakamata, S. Tagami, A. Fukuoka, T. Inoue,
S. Hirose, in Proceedings of IEEE International Ultrasonic Symposium 1994,
France, 1994
43. B. Koc, Y. Gao, K. Uchino, Jpn. J. Appl. Phys. 42(2A), 509–514 (2003)
44. P. Laoratanakul, A.V. Carazo, P. Bouchilloux, K. Uchino, Jpn. J. Appl. Phys.
41, 1446–1450 (2002)
45. S. Manuspiya, P. Laoratanakul, K. Uchino, Ultrasonics 41(2), 83–87 (2003)
46. A.V. Carazo, K. Uchino, J. Electroceramics 7, 197–210 (2001)
47. K. Uchino, T. Ishii, J. Jpn. Ceram. Soc. 96(8), 863–867 (1988)
48. H.-W. Kim, A. Batra, S. Priya, K. Uchino, D. Markley, R.E. Newnham, H.F.
Hofmann, Jpn. J. Appl. Phys. 43(9A) 6178–83 (2004)
49. N. Kurashima, in Proceedings of Machine Technical Institute Seminar (MITI,
Tsukuba, Japan, 1999)
50. A.M. Flyn, L.S. Tavrow, S.F. Bart, R.A. Brooks, D.J. Ehrlich, K.R. Udayaku-
mar, L.E. Cross, J. Microelectromech. Syst. 1, 44 (1992)
51. H. Saigo, in 15th Symposium on Ultrasonic Electronics, No.PB-46, Nov. 1994,
p. 253
52. K. Uchino, J. Ind. Educ. Soc. Jpn. 40, 28 (1992)
53. H. Aburatani, K. Uchino, A. Furuta, Y. Fuda, in Proceedings of 9th International
Symposium on Application Ferroelectrics, 1995, p. 750
54. H. Aburatani, K. Uchino, Jpn. J. Appl. Phys. 37(Part 1, No. 1), 204 (1998)
12
Piezoelectric Positioning
12.1 Introduction
Piezoelectric materials change shape under the influence of electric fields.
Known as the inverse piezoelectric effect, this phenomenon is the principle
upon which a large class of motion-producing devices is based.
The piezoceramic elements in today’s piezo actuators are optimized to
offer sufficient displacement to make them suitable for certain motion tasks.
They are used in high-frequency oscillators in ultrasonic wave generators, in
ultrasonic motors and as valve actuators in vehicles. Piezo actuators, however,
can also be designed for ultra-precise motion and positioning of objects in
quasi-static operation, i.e., for nanopositioning.
The displacement of a piezo actuator upon exposure to an electric field
is caused by ionic shifts and ferroelectric reorientations in the piezoceramics.
The motion is virtually friction free and the resolution practically unlimited.
Piezo actuators are especially well suited for positioning tasks requiring accu-
racies in the one-nanometer range over travel ranges of a few tens of microns
(Table 12.1).
Modern nanopositioning systems offer motion in six degrees of freedom
(X, Y, Z and the three corresponding rotation axes) with several hundred
micron travel and resolutions in the realm of one nanometer. Ideally, they are
designed as parallel-kinematics systems with non-contacting position sensors
for all motion directions, each measuring the moved platform against the sta-
tionary baseplate directly, and the whole connected to digital signal processing
and servocontrol electronics.
While in the early 1970s, positioning with such accuracies was reserved for
exotic applications in university research labs, by the beginning of the twenty-
first century, nanopositioning had found its way into a number of key industrial
processes – processes which would be virtually impossible without piezoelec-
tric technology. In commercial high-resolution optical, confocal and scanning-
probe microscopes, piezo actuators have become standard and are playing an
important role in research and quality-assurance inspection systems [1].
280 S. Arnold et al.
Piezo actuators are usually operated well below their resonant frequencies.
Keeping a position will led to high DC electric field loads over long time.
Therefore actuator lifetime is largely a function of the actuator’s resistance to
electric DC-fields, temperature and humidity. Another important positioning
related property is the thermal expansion behavior.
Shear actuators have the highest displacement coefficients. The effective value
of d15 with a field strength of 0.5 kV mm−1 can be up to 1, 200 pm V−1 ,
almost double the longitudinal coefficient at the same field strength. A further
advantage of shear piezos is their suitability for bipolar operation, whereby the
mid-position corresponds to a drive voltage of 0 V. In shear mode, unlike in
the other modes, the electric field is applied perpendicular to the polarization
direction (see Fig. 12.2). To prevent repolarization, the nominal field is limited
to 0.5 kV mm−1 [2].
Fig. 12.3. Linear thermal expansion of different actuator types. The negative ther-
mal expansion coefficient results from thermal depolarization of the piezoceramic
different processes can occur. Among others Ag+ ions can move as Ag(OH)
from the anode through cracks, voids, dissolved secondary phases at the grain
boundaries and, especially, on the surface to the cathode. There, it is recon-
verted to metallic silver. As a result, silver dendrites grow from the cathode
to the anode. First they lower the resistance, later they can cause breakdown.
Dissociated protons can also destroy the material directly. That’s why DC
humidity degradation is not just seen with Ag/Pd internal electrodes; even
actuators with platinum electrodes tend to degrade at high fields [4].
Active volume
Ceramic insulation
layer
Slots (preventing
stress accumulation)
Contact stripe
(a) (b)
Ceramic
insulation
layer
(c)
Fig. 12.4. Ceramic-insulated design, side view (a) non-termination side, (b) termi-
nation side, (c) insulation layer
12 Piezoelectric Positioning 285
The mechanical coupling between the load and the piezo actuator is of great
importance. It is ideally designed to compensate for the undesirable features
of piezo actuators: low pulling force capacity limiting dynamic operation, low
lateral load capacity and limited displacement. In addition to compensating
for these properties, the most modern mechanics designs include elements of
parallel kinematics, which is superior to the conventional serial arrangements
in many respects.
Fig. 12.6. To avoid damage, pulling forces are only applied to preloaded actuators;
the preload is either applied externally, or integrated in the actuator housing
For sine wave motion, the maximum forces can be estimated as follows:
∆L
Fdyn = ±4π meff
2
f 2, (12.2)
2
where Fdyn = dynamic force (N), meff = effective mass: ∼1/3 the weight the
piezoceramics plus the moved mass (g), ∆L = displacement (peak–peak) (m),
f = drive frequency (Hz).
The maximum permissible forces must be considered when choosing an
operating frequency.
One way to extend the dynamic operation capacity is the use of a preload,
for example by compressing the ceramic with a spring (Fig. 12.6). However, it
must not be forgotten that when operating an actuator against a spring load,
a part of the travel range is lost due to the elasticity of the piezo element.
Ball tips or flexures can be used to decouple lateral and bending forces
from the actuator (Fig. 12.7).
12.3.2 Flexures
Decoupling a piezo actuator from its environment with flexures does more
than just protect the piezoceramic from undesirable outside forces. Flexures
can also serve as guidance elements, preventing motion components perpen-
dicular to the actuator’s displacement direction from being transferred to the
moved mass.
12 Piezoelectric Positioning 287
Fig. 12.7. Ball tips or flexures to decouple lateral and bending forces
Fig. 12.8. Basic parallelogram flexure guiding system with motion amplification.
The amplification r (transmission ratio) is given by (a + b)/a. The system displace-
ment, ∆L, equals the lever amplification, r, times the displacement of the actuator
where r = lever ratio, ∆L0 = travel of the primary drive element, ∆Lsys =
travel of the lever-amplified system, ksys = stiffness of the lever-amplified sys-
tem, k0 = stiffness of the primary drive system (actuator stack and joints),
fres−sys = resonant frequency of the amplified system, fres−0 = resonant fre-
quency of the primary drive system (actuator stack and joints) (Fig. 12.8).
In practice, a balance between mass, stiffness and cost must be found.
The coupling between the actuator stack and the lever system is crucial. It
must be very stiff in the motion direction but should be soft in all other degrees
of freedom to avoid damage to the ceramics. If, for example, the stiffness of
each of the two lever interfaces is only as high as that of the actuator stack
alone, the overall system stiffness will be reduced by a factor of three. In many
piezo-driven systems, the stiffness of the ceramic is thus not the limiting factor
in determining the stiffness of the mechanism as a whole.
There are two basic ways to design multiaxis positioning systems: serial kine-
matics and parallel kinematics. Serial kinematics (Fig. 12.10) are easier to
design and build and can be operated with simpler controllers. In a multiaxis,
serial kinematics system, each actuator is assigned to exactly one degree of
freedom. If there are integrated position sensors, they are also each assigned
to one drive and measure only the motion caused by that drive and in its
direction of motion. All undesired motion (guiding error) in the other degrees
(a) (b)
Fig. 12.15. Flatness (Z-axis deviation) of a 100 × 100 µm, XY-plane scan with
a six-axis nanopositioning system with active trajectory control. The moving por-
tion of this parallel-metrology positioner is equipped with precise, parallel-metrology
capacitive sensors for all six degrees of freedom. The sensors are continually mea-
suring the actual platform position against the stationary reference. Any deviations
that occur in the four out-of-plane directions are corrected in real-time through the
corresponding actuators
12 Piezoelectric Positioning 293
Fig. 12.17. Same scan as before but using dynamic digital linearization, a highly
effective preshaping algorithm. Target and actual data can hardly be discerned. The
tracking error has been reduced to a few nanometers
12 Piezoelectric Positioning 295
Inertial Piezo Drives also offer theoretically unlimited travel. They use
a special controller and exploit the difference between static and dynamic
friction to push a runner in the motion direction. Because of the role of friction
between the runner and stator, the resulting force generation capacity and
system stiffness is limited. The range over which high-resolution motion can
be obtained is a few tens of microns; the speed up to a few tens of microns
per second.
The third possibility is a drive in which multiple piezo elements execute
a complicated, stepping motion sequence. Here too, the travel range is prac-
tically unlimited. The resolution, however, is not limited to the step size,
because an analog mode allows high-resolution motion over the range of one
step – a few tens of microns – and that with excellent dynamics. Speeds of up
to several millimeters per second can be obtained.
The step-mode motion sequence (Fig. 12.18) includes phases in which the
piezo element lifts off the runner and others where it drives the runner in
the direction of motion. The clamp/unclamp and drive motions can be imple-
mented with bender actuators. For applications with large moved masses, or
where the dynamics require extremely high stiffness of the drive, the piezo el-
ements can consist of two differently polarized segments: one for the clamping
motion and one for the shear motion. The shear elements are responsible for
the driving force, while the longitudinally polarized clamping elements control
contact with the runner in the different step phases.
In analog mode, where only the shear segments are active, motion with
subnanometer resolution can be carried out over a travel range of about one
step (Fig. 12.19). Analog mode exhibits excellent dynamic performance, mak-
ing these drives suitable for tasks as demanding as active vibration control.
Piezoelectric nanopositioners with stepping drives can generate forces of up
to 1,000 N and offer resolutions of better than 100 pm (Fig. 12.20). High-
resolution incremental encoders, which provide very high accuracy over ranges
of many millimeters, are used as position control sensors.
Specially designed electronics ensures the coordination of the shear and
clamping motion of the individual piezos. Stepping drives represent a
Fig. 12.18. Motion sequence in a piezo stepping drive with shear and clamping
element segments. While one set of piezo elements clamps and drives, the other set
takes a new bite
296 S. Arnold et al.
Fig. 12.19. In analog mode only the shear segments are active and both sets of
elements are clamped to the runner
or rapid scanning of CCD arrays at the pixel level to increase the effec-
tive resolution (dithering). The widening use of such systems in digital
cameras designed for long-distance aerial photography, in astronomy or in
microscopy-based inspection systems and medical equipment will continue to
pose new challenges for nanopositioning technology.
While taking a position within a few nanometers and holding it for a cer-
tain time is the central claim of nanopositioning, the ancillary conditions will
become more and more demanding. Longer and longer travel ranges with no
degradation in resolution or stiffness will be required. Harsh environments,
such as vacuum, helium atmosphere or intense magnetic fields will challenge
reliability. The manufacturers of nanopositioning equipment will find it in-
creasingly necessary to examine new technologies like stepping drives. They
will also have to look for more efficient manufacturing processes because in-
dustrial applications are generally cost-sensitive and the use of piezoelectric
drives in an industrial process must not reduce the throughput, but rather in-
crease it. Not all limitations have their solution in improved actuator design or
other hardware changes. Dynamic signal reshaping for faster settling or better
trajectory control, filtering techniques or new types of control algorithms are
also part of the development of nanopositioning technology. Nonetheless, one
important direction is the research into new active materials and technologies,
with the object of obtaining greater induced mechanical displacement, lower
hysteresis and improved linearity.
References
1. S. O. R. Moheimani, Rev. Sci. Instr., 79(7), pp. 071101 (2008)
2. P. Pertsch, D. Rössger, E. Hennig, W. Plötner, A. Bauer, in Proceedings of 8th
ACTUATOR 02, Bremen, June 2002, Axon, pp. 45–48
3. P. Pertsch, H. Marth, P. Anger, A. Feltz, in Proceedings of 7th ACTUATOR 00,
Bremen, June 2000, Axon, pp. 41–44
4. P. Pertsch, S. Richter, D. Kopsch, N. Krämer, J. Pogodzik, E. Hennig, in Pro-
ceedings of 10th ACTUATOR 06, Bremen June 2006, Axon, pp. 527–530
5. R. Glöss, P. Pertsch, Mechatronik F&M 107(9), pp. 64–68 (1999)
6. K.-D. Müller, H. Marth, P. Pertsch, R. Glöss, X. Zhao, in Proceedings of 10th
ACTUATOR 06, Bremen, June 2006, Axon, pp. 149–153
13
Piezoelectric Injection Systems
Fig. 13.2. PCR Injector: pressure distribution in the closed state (upper view) and
in the open state (lower view) of the servo-hydraulic injector drive
ring-shaped gap is closed again so that via the two throttles the full sys-
tem pressure is built up again inside the control chamber. As soon as the full
pressure is reached, the resulting force acting on the nozzle and control piston
again points towards the seat. The nozzle moves into its stationary position
and the nozzle outlets are closed.
Via the diameter ratio of the two throttles, the opening and closing veloci-
ties of the nozzle can be adapted to the requirements of the respective engine.
That such a servo principle is used instead of, for instance, a direct drive
is due to the fact that the mechanical work that has to be applied in or-
der to open the nozzle exceeds the capabilities of the actual piezo actuator
by one order of magnitude. This circumstance is discussed in more detail in
Chap. 13.3.1.
A fuel atomisation, the quality of which being sufficient for a good combus-
tion, is reached in the case of gasoline at pressures of about 200 bar. As a
consequence, it is feasible to realise gasoline high pressure injectors actuating
the nozzle by a direct drive (Fig. 13.3).
In the case of the injector for which series production has been started
in 2006 [5–8], the functional unit of piezo actuator, the nozzle and hydraulic
compensator is mounted into a double-walled, cylindric housing. The fuel is
fed to the nozzle via the gap between the inner and the outer tubes, thus
acting as a coolant for the inner injector parts, mainly the actuator unit.
The top cap of the piezo actuator is supported by the hydraulic compen-
sator, whereas the bottom cap is in contact with the valve needle. During
the mounting procedure, the needle is loaded by a spring which generates a
force of about 200 N by which the needle is pressed into the seat. After the
installation and the application of the electric contacting to the piezo actu-
ator, the hydraulic compensator unit is mounted into the injector housing
and loaded by a force of about 100 N against the piezo actuator. Thereby the
force between the needle and the seat is reduced by just this number so that
100 N remains as a resulting force which retains the nozzle in its seat against
a maximum fuel pressure of 200 bar.
The fuel enters the gap between the needle and the cartridge at the lower
end of the injector housing across two boreholes in the inner tube of the hous-
ing. To prevent the piezoactuator from being wetted by the fuel, a bellow
between the upper end of the cartridge and the needle acts as a flexible seal
in the axial direction. Its effective pressure-loaded cross section area is di-
mensioned such that the hydraulic forces acting on the bellow and the needle
compensate in essence.
As the piezo actuator is electrically charged, it expands against the forces
by the compensator and the needle. As the compensator is conceived such
that it forms a hard bearing for a fast switching motion, the valve lifts off
13 Piezoelectric Injection Systems 303
fuel inlet
hydraulic
compensator
piezo
actuator
needle
group
Fig. 13.3. Photographic view and CAD cross section of a piezoelectric high pressure
gasoline injector
due to the electric excitation of the actuator after the closing force has been
overcome. A ring-shaped gap opens between the needle and the seat through
which the fuel is injected and atomised.
As soon as the desired fuel quantity is reached, the piezo actuator is elec-
trically discharged and as a consequence contracts. The preload spring then
pulls the needle back to the seat. The ring-shaped gap is closed, and the
injection cycle is complete.
Considering the CO2 problem and the global warming, gaseous fuels increas-
ingly become the focus of attention for developers. They either exhibit a
304 R. Mock and K. Lubitz
greatly improved CO2 balance than gasoline or Diesel, or like hydrogen they
can completely be generated by means of regenerative energy sources and
accordingly are CO2 neutral.
At present, the development focuses on injectors which take into account
the specific requirements of the injection of gaseous fuels [9]. The fundamen-
tal difference to liquid fuels is the completely different flow characteristics of
gases. To dose a gaseous fuel which contains the same amount of energy as
liquid fuels, one has to inject a considerably larger volume. At a pressure of,
e.g. 20 bar the injector has to deliver a throughput quantity which exceeds
that of gasoline by a factor of fifty in order to obtain the same engine power.
The injection timing and the gas pressure, however, are defined by system
requirements so that the larger volume throughput has to be realised by an
increased cross section of the orifice. As the diameter of the employed nozzle
can be increased only to a limited amount, there remains only an increase
of the needle lift to a value of 300 µm to guarantee the throughput quantity.
However, using the present piezo actuators, a stroke of 50 µm cannot be ex-
ceeded. Therefore a hydraulic transformer is mounted between the actuator
and the nozzle which at the same time acts as a thermal compensator. The
working principle of this injector including the hydraulic transformer element
is shown in Fig. 13.4.
gas inlet
piezo
actuator
hydraulic
transformer
needle
group
Fig. 13.4. Photographic view and CAD cross section of a piezoelectric CNG injector
13 Piezoelectric Injection Systems 305
Due to the different working principles of the three injector concepts presented
before, the technical requirements which have to be established for a suitable
piezo actuator greatly differ. Nevertheless, the basic procedure by which the
mechanical and electric characteristics are specified is the same in all the three
cases. For the PCR, this procedure will be explained in some detail.
By the functional principle of the PCR, it becomes obvious that the open-
ing and closing event only occurs when the pressure inside the control chamber
has reached its respective end value. Particularly during the opening event,
the pressure drop inside that chamber has to be effected sufficiently fast so
that the movement of the control piston and the valve needle begins after a
minimum time delay. As this pressure drop is substantially influenced by the
flow characteristic of the servo valve, the piezo actuator has to fulfil two tasks
in this case:
1. It has to provide sufficient stroke so that the servo valve is fully unthrottled
in the opened condition (i.e. only the fixed throttles in the high pressure
feed determine the flow rate)
2. It has to generate enough force to open the servo valve sufficiently fast
against the pressure in the control chamber
306 R. Mock and K. Lubitz
To guarantee the first requirement, the flow cross section area released by
it has to be significantly larger than that of the two fixed throttles which act
in series with it. In case of the PCR the flow cross section at a stroke of 35 µm
is roughly twice as large as that of the largest fixed throttle – therefore just
sufficient to fulfil the requirement.
The force on the servo valve built up by the chamber pressure can be
calculated from the pressure loaded surface of the valve at a given pressure.
Taking into account also the force of the closing spring, the actuator has to
generate a force of about 400 N at maximum system pressure before the servo
valve is moved at all. Detailed simulations show that in order to obtain a
sufficient dynamics of this process one indeed needs about 1,000 N!
The piezo actuator’s maximum force, however, is not available during the
complete actuation process. Due to the limited stiffness of the mechanical
connections (weldings, etc.), part of the mechanical energy generated by the
actuator is consumed by the mechanical work needed to expand the housing
and its components. Provided that the stiffness values of housing and me-
chanical connections are optimised, the actuator has to exhibit a blocking
force of about 2,100 N – at a blind lift of 35 µm this corresponds to a spring
stiffness of 60 N µm−1 – to guarantee a net blocking force of 1,100 N for the
switching cycle.
From the electric point of view, the second requirement means that the
actuator has to be electrically energised with a rise and fall time of 100 µs to
the maximum field strength of 2 kV mm−1 . Regarding technical boundaries
and system costs, these electric characteristics have to be realised at moderate
voltages (<200 V) and currents (<30 A). The consequence of this is that the
piezo actuator has to be conceived as a multilayer stack with an appropriate
number of piezoelectric layers.
Motor injection systems in the past were realised solely with electromagnetic
drives. In order to replace them by the completely new concept of a piezo-
electric drive, the latter must have comparable production costs, life time and
reliability and in addition a remarkably better functionality. It is described in
the following section how a specific material design, i.e. matching of construc-
tion form, material composition and fabrication technology can lead to such
a solution.
If we consider the functional principles of Diesel injection valves shown
in Sect. 13.2.1, we can summarise the specifications for the piezoelectric drive
itself by the following conditions: Drive voltage of up to 200 V DC, drive
frequency of up to 100 Hz, piezoelectric elongation of 40 µm, blocking force of
about 2,000 N, pulse length without rise and fall times 50–100 µs, operating
temperatures from −40 to +160◦ C, where the maximum temperature is given
by the external motor heating and by self-heating, and finally a minimum life
time of >109 cycles.
13 Piezoelectric Injection Systems 307
driving condition, in this dynamic mode during the falling pulses mechanical
tensile stresses occur which by undercritical and critical crack growth destroy
the stack very fast. Therefore a constant compressive stress must act on the
stack in such a way that on no account tensile stresses can occur within the
sample. That will be realised by tube springs in which the stack is welded
under a compressive stress of 10 MPa [14].
The polarisation takes place within the tube spring. The superimposed
compressive stress remarkably reduces the remanent relative elongation com-
pared with a free poling. The number of domains oriented parallel to the pol-
ing field is lowered due to the ferroelastic influence of the mechanical stress.
During driving and under an electrical field of 2 kV mm−1 many domains
are switched in field direction and the stack elongation reaches the necessary
amount.
This kind of polarisation has a further beneficial effect. In Fig. 13.5, the
thermal expansion coefficients are plotted of unpoled and poled piezoceram-
ics parallel and perpendicular to the poling direction [15]. In poling direction,
against all normal experience the coefficient is negative. This is due to a
decrease of the spontaneous polarisation and of the anisotropic lattice distor-
tion with increasing temperature. The stack polarisation under mechanical
compression stress creates a domain structure similar to that of an unpoled
sample. Thus the thermal expansion coefficient is positive and comparable to
that of the surrounding metallic tube spring.
The most important point which at all has made possible the application
of piezo stacks as injection valve drives and which guarantees life time and
reliability, is the application of quite specific electrical contacts. They must
control and eliminate the detrimental influence of in principle originating
∆I/I 4 x10–4 Pr
2 x10–4
α = +2 x10–6 K–1 unpoled
1 x10–4
0
20 30 40 50 60 70
–1 x10–4 temperature [C]
–2 x10–4
α = –5 x10–6 K–1
–3 x10–4 Pr
–4 x10–4
Fig. 13.5. Thermal expansion coefficients of unpoled and poled samples parallel
and perpendicular to the remanent polarisation
13 Piezoelectric Injection Systems 309
3
S [10–3]
0
0 1 2 inactive contact zone
E [kV/mm]
Fig. 13.6. Formation of delamination cracks during poling inside the inactive con-
tact areas
cracks. Within the contact areas of the stacks (see Fig. 4.21, Chap. 4.5) even
during polarisation and beginning operation tensile stresses arise although
the active stack area is under compression stress by the tube spring. These
stresses can be calculated, are highest in the middle of the stack and lead
to delamination cracks in the contact area in general between PZT and the
electrode. In Fig. 13.6, the evolution of such cracks during poling is measured
by sound emission analysis and also their position is shown.
These fissures act as mechanical stress relievers so that their number after
short operation time remains constant, depending on design, technology and
especially on the strength of the interface PZT – inner electrodes. The crack
tips stop permanently within the transition region to the active stack area
which is under compression stress. Also some design variations are known
where a series of artificial slits within the contact areas of the stack is incor-
porated during the fabrication having the same stress relief function [16, 17].
During operation in the region of the relief cracks a large stroke can be ob-
served corresponding to the total elongation of the adjacent undamaged stack
segment. Locally relative elongations of more than 100% may occur which,
after short operation time, unravel the originally continuous metallic contact
strip. In order to provide all these single stack elements with electrical energy,
a multiple contact design is necessary. That can be realised for example by a
wire harp or by soldered metal wool.
With the sum of the described measures, it succeeds to manufacture reli-
able stacks with life times of >109 cycles. Meanwhile, they are in use in cars
a millionfold. It is remarkable that the absolute stack length as well as the
elongation remain constant within narrow limits during life time. This means
that both the remanent domain structure and the number and kind of do-
mains switched up to saturation during each pulse remain unchanged. This
also shows the future potential of piezoceramic stacks in the area of actuator
applications if material and technology are modified specifically.
310 R. Mock and K. Lubitz
References
1. D. Baranowski, W. Kluegl, D. Schoeppe, in Simulation and Design Optimization
of a Common Rail Piezo Injector for Passenger Car DI Diesel Engines Fuel
Injection System, London, GB, Dec 1–2, 1999 (IMechE Seminar Publication,
London, 1999)
2. K. Egger, D. Schoeppe, Diesel Common Rail II – Einspritztechnologie fuer die
Herausforderungen der Zukunft Fortschritt-Berichte VDI Reihe 12, Nr. 348,
Band 2
3. R. Mock, P.-C. Eccardt, B. Gottlieb, A. Kappel, K. Niederer, in Modellierung der
dynamischen und akustischen Eigenschaften eines piezoelektrischen Common-
Rail-Diesel-Hochdruckinjektors mit ANSYS XXVI. FEM-Kongress, Baden-
Baden, D, 15–16 Nov. 1999
4. K. Egger, J. Warga, W. Klügl, Motortechnische Zeitschrift. 63(9), 14–17
(English), 696–704 (German) (2002)
5. E. Achleitner, S. Berger, H. Frenzel, M. Klepatsch, V. Warnecke, Benzin-
Direkteinspritzsystem mit Piezo-Injektor für strahlgeführte Brennverfahren
MTZ (Motortechnische Zeitschrift) 05/2004
6. E. Müller, R. Otte, Potenzial der Doppeleinspritzung beim DI-Ottomotor MTZ
(Motortechnische Zeitschrift) 09/2006
7. A.E. Funaioli, E. Achleitner, H. Baecker, in Direct Injection Systems for Otto
Engines SAE World Congress and Exhibition, April 2007, Detroit, MI, USA,
Session: Direct Injection SI Engine Technology (Part 3 of 3), 2007-01-1416
8. W. Mährle, C. Luttermann, M. Krauss, N. Klauer, High-Precision Injection in
Combination with the New BMW Twin-Turbo Petrol Engine MTZ Worldwide
03/2007
9. A. Koch, G. Lugert, E. Magori, G. Bachmaier, S. Eisen, in Key Components for
Gaseous Fuels: Actual Demands and Future Concepts 1st International Sympo-
sium on Hydrogen Internal Combustion Engines, University of Graz (Austria)
September 2006
10. K. Lubitz, H. Hellebrand, D. Cramer, I. Probst, in Euro-Ceramics II, Proceed-
ings ECerS’91, September 11–14 (1991), Augsburg, 1995
11. S. Takahashi, A. Ochi, M. Yonezawa, T. Yano, T. Hamatsuki, I. Fukui, Ferro-
electrics 50, 181 (1983)
12. C. Schuh, K. Lubitz, T. Steinkopff, A. Wolff, in Piezoelectric Materials in Ad-
vances in Science, Technology and Applications, ed. by C. Galassi et al. (Kluwer
Academic, Dordrecht, 2000), pp. 391–399
13. K. Lubitz, C. Schuh, T. Steinkopff, A. Wolff, in Proceedings of the 7th Interna-
tional Conference on Actuators, Actuator 2000, June 19–21, Bremen, Germany,
pp. 58–61 (2000)
14. Patent EP 1053568 B1, 2000
15. K. Lubitz, C. Schuh, T. Steinkopff, A. Wolff, in Piezoelectric Materials in De-
vices, ed. by N. Setter (edited and published, Lausanne, 2002), pp. 183–194
16. Patent US 6307306 B1, 2001
17. Patent EP 0844678 B1, 2002
14
Advanced RF Signal Processing with Surface
Acoustic Waves on Piezoelectric Single Crystal
Substrates
14.1 Introduction
Today’s rapidly growing proliferation of wireless services such as radio and TV
broadcasting, cordless phones, mobile communication, wide and short range
data communication, satellite based navigation systems and the growing num-
ber of users require an unprecedented increase of information that needs to
be transmitted. Frequencies which are predominantly used in actual wireless
transmission systems are limited to the radio frequency (RF) range from about
100 MHz to a few GHz. Today a big portion of the EM spectrum available for
transmission is already allocated (Fig. 14.1, EM spectrum allocation) and to-
day’s services already require an efficient utilization of the allocated channels.
In the future, with more services to be provided, the efficiency of utilization
has to be further improved.
According to Shannon’s Theory of Communication [1], the information
capacity of transmission channels is proportional to the product of their band-
width and signal to noise ratio (in dB). Thus, in order to provide more in-
formation capacity, the allocated frequency spectrum has to be continuously
expanded to higher frequencies, any free space between allocated frequency
bands has to be minimized and wider bandwidth channels have to be pro-
vided for higher data rate services. However, as information capacity depends
on bandwidth as much as on signal to noise ratio, this signal to noise ratio
must be maintained at its theoretical upper limit set by the signal power and
the thermal noise present in the channel. Therefore filters are needed that se-
lect signals within the desired channel and prevent inter-channel interference
from increasing the noise level.
The growing need for efficient utilization of the EM spectrum for wireless
transmission poses challenging requirements on filter technology. Filters must
operate at higher frequencies up to the GHz region, with higher bandwidth and
low losses in the pass band, high out of band rejection and narrow transition
bandwidth between pass band and stop band. As mobility is required for
many services, miniaturization and low power consumption becomes another
312 R. Veith et al.
VIII UL VIII DL
VI UL VI DL
V UL V DL
820 830 840 850 860 870 880 890 900 910 920 930 940 950 960
X UL X DL
IX UL IX DL
IV UL II UL II DL IV DL
III UL III DL I UL I DL
1710 1750 1800 1850 1900 1950 2000 2050 2100 2150 2200
Fig. 14.1. EM spectrum allocation: Part of the spectrum allocated for GSM and
WCDMA mobile communication in the 900 and 2,000 MHz regions (UL: uplink, DL:
downlink)
challenging requirement for filters. Last but not least high volume consumer
applications require reproducibility at high volume and economical cost, while
others such as automotive applications require highest reliability in harsh
environments.
With piezoelectric materials, mechanical waves in solids can be employed
to perform filter functions for electrical signals. The low propagation velocity
of mechanical waves provides miniaturization by a factor of 10−5 compared to
realization of filters with EM waves in the free space. Further more, mechanical
waves provide low propagation losses and high stability. Bulk wave resonators
made up of thin quartz plates have been used since long for frequency control
and are still the component of choice for precisely controlling frequencies in
the range of 1–10 MHz. Their application to filtering has been limited as this
technology is bound to a low frequency range and does not render itself easily
to integration of multiple coupled resonators.
With the advent of the Interdigital Transducer (IDT), an electrode struc-
ture that can efficiently launch waves propagating along the surface of piezo-
electric materials, higher frequencies in the range 10 MHz to 10 GHz became
accessible together with almost unlimited design capabilities in conjuncture
with submicron photolithographic planar patterning technology.
Though basically powerful for application to filter design, this technol-
ogy required the identification of optimized piezoelectric materials in order to
meet stringent system requirements. Single crystals proved to be most suit-
able because of their inherent homogeneous and highly reproducible structure.
14 Advanced RF Signal Processing with SAW 313
Figure 14.2 shows the basic building blocks of a transceiver circuit, so des-
ignated as it combines Tx and Rx functions. Actual transmitters and receivers
in any wireless communication system may consist of more or less complex ar-
chitectures but they are essentially made up of the same basic building blocks.
On the Tx path the Tx signals formed in the base band processor are up con-
verted to the designated Tx channel carrier frequency by mixing with the
carrier frequency generated in the corresponding RF local oscillator, passed
through an RF Tx interstage filter, amplified in the RF power amplifier to
the required output power level and finally passed through an RF Tx front
end filter before reaching the antenna. On the Rx path, RF signals received
from the antenna are first passed through an RF Rx front end filter, ampli-
fied in the low noise amplifier (LNA), passed through an RF Rx interstage
filter, down converted to an intermediate frequency (IF) by mixing with the
frequency generated in the corresponding RF local oscillator, passed through
an IF filter, amplified and finally down converted to the base band by mixing
with the IF frequency generated in the corresponding local oscillator.
The basic transceiver structure of Fig. 14.2 uses three different categories
of filters. The first category, RF front end filters directly connected to the
antenna, are used in the Tx path to remove any spurious components of
the transmitted signal spectrum outside the allocated frequency band and in
the Rx path to reject any unwanted signal outside the allocated frequency
band in order to prevent interference with the wanted receive signal. Depend-
ing on the system and allocated frequencies, RF filter center frequencies and
bandwidths typically range from 400 MHz to 5 GHz respectively from several
100 KHz to 100 MHz. RF front end filters in particular have to meet two criti-
cal requirements. On the one hand, front end filters need to provide minimum
insertion attenuation in the pass band. Any loss of the wanted receive signal
power would degrade the signal to noise ration at the LNA input and thus
reduce the receiver sensitivity. Likewise any loss of the transmit signal power
would be a waste of output power from the power amplifier and, at least for
14 Advanced RF Signal Processing with SAW 315
Historically filters and resonators have been first realized with lumped ele-
ment inductors (L) and capacitors (C) with one LC pair forming a resonator
and several resonators forming higher order filters with steeper skirts and
wider bandwidth. While such filters had been widely used at frequencies be-
low 100 MHz, the lower quality factors (Q) of inductors and capacitors at
higher frequencies were typically not sufficient to meet system requirements.
In addition LC filters and frequency control elements were lacking the repro-
ducibility and stability required for today’s wireless communication systems
and they were by far too bulky to be considered for mobile devices.
At higher frequencies, beyond 1 GHz, electromagnetic waves resonating
in metal enclosed cavities provided the necessary Q values to implement high
performance resonators and filters with excellent reproducibility and stability.
However, with the dimensions of these devices ranging from half a wavelength
to several wavelengths, the propagation velocity of EM waves in free space
of 3 × 108 m s−1 and the corresponding wavelength at 1 GHz amounting to
0.3 m, these cavity filters were even more bulky than their LC counterparts and
their application thus limited to large size Microwave and Radar transceiver
installations.
The next step in miniaturization was to use electromagnetic waves in di-
electric materials in order to slow down the propagation velocity by a factor
14 Advanced RF Signal Processing with SAW 317
waves. The variation of the propagation velocity with direction can also be
used to minimize diffraction of a narrow beam of surface acoustic waves. Pro-
vided the variation of propagation velocity in the propagation plane around
the main propagation direction is symmetrical and the propagation velocity in
the main propagation direction is at minimum, diffraction effects are reduced
and the SAW beam stays more focused in the main direction as compared
to propagation in an isotropic material. On the other hand, in the general
case with the variation of propagation velocity around the main propagation
direction being asymmetrical, the direction of power flow will not coincide
with the direction of phase velocity and the SAW beam propagation will be
diverted from phase velocity vector by the power flow angle. This effect of
beam steering has then to be considered and accounted for in the design of
SAW devices.
Fig. 14.3. Different types of IDTs. (a) Single electrode type, (b) double electrode
type, (c) λ/6-finger-type
(a) (b)
Fig. 14.4. Different types of SPUDT cells. (a) EWC SPUDT; (b) Hanma-
Hunsinger-SPUDT
fingers is λ/4. Consequently, this type of IDT is used, whenever there are strin-
gent requirements on the group delay characteristics of the SAW component.
Owing to their symmetry, the IDTs described above generate surface waves
of equal amplitude in both the forward and reverse directions, so that it is
called bidirectional. If an IDT can excite a SAW predominantly in one di-
rection only, we call it a unidirectional IDT (UDT). Unidirectionality can
be realized in three different ways: (1) using multi-phase electrical inputs,
e.g., three-phase input signals with a relative phase shift of 120◦ ; (2) using
single-phase electrical input signals applied to comparatively complex, spa-
tially asymmetric IDT cells, which rely on internal reflection to shift the
“center of gravity” of excitation in the cell relative to the “center of grav-
ity” of reflection of the cell such that unidirectionality is achieved; this trans-
ducer type is called single-phase unidirectional trandsducer (SPUDT); (3)
using substrates with so-called natural unidirectionality, for which – due to
crystallographic asymmetry – the center of reflection is displaced from the
geometrical center of the electrode, where the center of excitation is located;
this form is generally named NSPUDT.
Of all these different forms, only the SPUDTs have gained practical impor-
tance. There is a significant number of different realizations of SPUDTs: prob-
ably the most-used is the electrode-width-controlled SPUDT (EWC-SPUDT)
and the Hanma-Hunsinger-SPUDT (HH-SPUDT) of Fig. 14.4, which consists
of cells with three or four electrodes of unequal widths (Fig. 14.5).
Placing two IDTs next to each other and using one as the input transducer,
which converts the electrical signal into an acoustic wave, the other one as
the output transducer, which regenerates an electrical signal from received
14 Advanced RF Signal Processing with SAW 321
acoustic waves, the most simple filter configuration is obtained (Fig. 14.5).
The best achievable insertion loss and the fractional 3 dB-bandwith are both
directly proportional to the electro-acoustic coupling factor of the substrate
material.
applied on or grooves etched into the substrate can be used. For economic
reasons and ease of processing, however, metallic strips are usually preferred.
The choice of materials has important impact on resonators’ performance.
The reflectivity of each strip and, consequently, the width of the stop band are
determined by the material properties of both the piezoelectric substrate and
those of the strips as well as their geometry. Reflectivity is increased (1) by
increased piezoelectric coupling of the substrate, (2) by increased mass density
of the electrode material, and (3) increased thickness of the electrodes.
From the application’s point of view, on one hand a high strip reflectivity
is desirable, because using highly reflecting electrodes, less strips and conse-
quently less die area is needed to attain total reflection. On the other hand a
high strip reflectivity results in a higher sensitivity of frequency with respect
to fabrication tolerances.
Finally, the maybe most relevant of the secondary properties of a resonator
is its frequency behavior with respect to temperature. It is defined by the
thermal expansion of the substrate α as well as by the variation of surface
acoustic wave phase velocity with temperature TCVp : TCF = TCVp − α.
To first order, TCF is defined by the substrate crystal alone. To second order,
also the material and height of electrodes influence TCF considerably.
L1
C0 C1
R1
deliberately mismatched so that all signals are reflected at the input port and
the transmission is minimized. Since impedances of acoustic resonators have
strong frequency dependence and can be controlled within certain limits by
choice of materials and design measures, the objective was to use them instead
of discrete reactance elements.
There are several methods from network synthesis that aim at finding a
circuit that implements a given transfer characteristic. Nearly all these meth-
ods result in a net-list consisting of ideal inductance (L) and capacitance
(C) elements. Since a SAW one-port resonator’s electrical characteristics can
be well approximated by an equivalent circuit consisting of L and C elements
only, it is possible to construct a filter circuit with SAW resonators as building
blocks, if the network topology is adequate (so-called ladder type topology)
and the element values are realizable.
The equivalent circuit of a one-port SAW resonator consisting of an IDT
and two shorted reflectors is shown in Fig. 14.6. It comprises of four elements:
a static capacitance C0 , the dynamic capacitance C1 , the dynamic inductance
L1 , and a small series resistor R1 . The resistor models the losses that exist in
a realistic SAW resonator, and can be neglected for simplified considerations.
Figure 14.7 shows a plot of the admittance function of an on-port resonator.
There are two characteristic frequencies:
• The (series) resonance frequency fs , where the admittance has a pro-
nounced maximum (“pole”)
• The parallel resonance frequency fp or anti-resonance frequency fa , where
the admittance function approaches zero
With fixed metallization height and ratio, the series resonance frequency
of an IDT is inversely proportional to its finger period. Off the resonances,
the admittance function is dominated by the static capacitance C0 .
Figure 14.8 shows the basic section of a ladder-type filter. It consists of
a series-resonator and a shunt resonator, whose anti-resonance frequency fa,2
coincides with the series resonance frequency fs,1 of the series resonator. This
basic section has a band-pass characteristic: At f = fs,2 , the parallel resonator
provides a short circuit for the input signal, at f = fa,1 , the series resonator
324 R. Veith et al.
Fig. 14.7. Absolute value (red solid line) and real part (dashed blue line) of admit-
tance Y11 of a one-port resonator in logarithmic scale
provides an open circuit for the input signal. So in both situations, the trans-
mission approaches zero. At f = fs,1 = fa,2 , the parallel resonator provides
an open circuit and the series resonator a short circuit, so transmission from
input to output port is maximal (see Fig. 14.9).
The selectivity of a single basic section is rather low. For practical appli-
cations, therefore typically up to nine basic sections are cascaded in order to
form a filter characteristic with high selectivity, on the cost of some additional
insertion attenuation.
Whereas in the so-called image parameter design, identical basic sections
and their mirror images are combined, in most state-of-the-art filter designs
the individual sections are tuned individually to reach optimum electrical
performance. The common name of this filter structure is LTF) or Impedance
Element Filter (IEF).
14 Advanced RF Signal Processing with SAW 325
In Out
In the race for SAW filters with wide pass band, minimum insertion loss
and narrow transition bandwidth, the ideal piezoelectric material would have
a high electromechanical coupling coefficient K 2 and a zero temperature co-
efficient of delay TCD. As such a material does not exist as a single crystal,
SAW devices, depending on the requirements for specific applications, use a
number of different piezoelectric crystals. The vast majority of today’s SAW
devices are thus using any of the following three piezoelectric single crystals,
quartz, LT, and LN, of which quartz offers the highest temperature stability
with lowest K 2 , LT offers moderate temperature stability and K 2 and LN
combines the highest K 2 with the lowest temperature stability. Other piezo-
electric crystals such as Li2 B4 O7 , Bi12 Ge4 O20 and La3 Ga5 SiO14 have been
tried for special applications but not found widespread usage so far.
Piezoelectric crystals are highly anisotropic and this fact has been used to
search for and identify optimum planes and directions for SAW propagation
offering high electromechanical coupling K 2 and low temperature coefficient
of delay TCD. In combination with the boundary conditions at the surface,
basically the metallization employed, further optimizations have been achieved
by identifying planes and directions for low SAW propagation losses, reduced
SAW diffraction and reduced excitation of spurious bulk waves.
The identification of optimum SAW propagation directions depends on
the type of wave to be employed. The Rayleigh wave, having the slowest
propagation velocity and the lowest propagation loss, offers the best potential
for miniaturization of high performance SAW devices but electromechanical
coupling K 2 is limited to about 5% within the range of most frequently used
piezoelectric crystals (Table 14.1). Rayleigh waves are therefore employed for
frequency control SAW resonators, narrow band low loss front end filters and
highly selective IF band pass filters of all bandwidth ranges.
Rayleigh waves on ST-cut quartz, propagating in the X-direction are be-
ing used for frequency control devices and narrow band filters. ST-cut quartz
Table 14.2. Piezoelectric single crystals for leaky surface acoustic wave (LSAW)
devices
Crystal Crystal cut SAW Velocity K2 TCD
propagation (m s−1 ) (%) (ppm K−1 )
LiTaO3 36◦ rot Y X 4,100–4,212 4.7–7.6 28–35
LiTaO3 42◦ rot Y X 4,100–4,212 4.7–7.6 28–35
LiNbO3 64◦ rot Y X 4,478 11.3 81
LiNbO3 41◦ rot Y X 4,379 17.2 80
TCD ranging from about 5 to 17%, respectively 32–81 ppm K−1 (Table 14.2).
The critical issue for employing LSAW, however, is to find planes and direc-
tions for LSAW propagation with minimum propagation loss. For LT rotated
Y -cut X-propagating LSAW, leaky propagation losses exhibit a parabolic de-
pendence on both cut-angle and metallization height. Thus minimum leaky
propagation loss is observed for zero metallization height on 36◦ rot Y LT
and for a more practical Al metallization height h/λ of about 10% on 42◦
rot Y LT. The majority of mobile phone front end filters with relative band-
width about 5% are made on 42◦ rot Y LT substrates. For wider bandwidths,
LN has to be used, which offers a range of rotated Y -cuts with low LSAW
propagation losses and K 2 of 11–17%. However, higher temperature shifts of
frequency due to the TCD of about 80 ppm K−1 has to be taken into account
in the design of SAW filters on these substrates.
Before SAW devices could find their widespread application in wireless
communication, much efforts had to be spent for the development of crys-
tal growth and wafer fabrication processes for the economical production of
piezoelectric single crystal substrates with the necessary low defect densities,
compositional variances and dimensional tolerances.
Quartz crystals are grown in a hydrothermal process in autoclaves at tem-
peratures between 300 and 400◦ C and pressures between 700 and 1,500 bar
by immersing seed plates into aqueous solutions of quartz with sodium hy-
droxide or carbonate acting to increase quartz solubility [14]. Growth rates
are mainly controlled by the temperature difference between quartz grow-
ing and quartz supply regions and are typically ranging between 0.4 and
1 mm day−1 (Z-direction). Critical defects for SAW devices are voids and
dislocations that can originate from the seed solution interface and propagate
from there through the crystal in the course of crystal growth. X-ray topog-
raphy has been used to analyze such defects and provide valuable hints about
their origin as well as effective ways to prevent them [15]. With careful control
of seed surface quality and growth parameters, especially in the initial growth
phase, state of the art growth processes allow to grow quartz crystals almost
free of such defects (Fig. 14.12).
LT and LN crystals are grown from the melt using the Czochralski method
by first dipping a crystal seed rod into the melt and then pulling the rotating
seed crystal rod upward. With precise control of temperature gradients, rate
330 R. Veith et al.
Fig. 14.12. X-ray topography of 4 in. quartz wafers with (a) and without (b) crystal
defects originating from the seed interface [16]
Fig. 14.13. X-ray topography of LiNbO3 boule cross sections with defects originat-
ing from the seed and the cone segment [16]
With I-line 5× projection printing wafer steppers 300 nm line width res-
olution can be achieved with the photolithographic micro patterning process
corresponding to an upper limit of 3 GHz for the center frequency of LSAW RF
filters on LT. With proper control of all relevant process parameters standard
deviations of critical dimensions can be reduced to 20 nm for the line width and
to 0.5% for the relative metallization height. The sensitivity of SAW devices,
mainly their frequency characteristics, to such dimensional deviations depends
on many factors such as piezoelectric material, propagation direction, type of
wave, metallization material, line width to gap ratio, metallization height as
a fraction of the acoustic wavelength and even the metallization edge pro-
file [21]. In case of a 1.9 GHz LSAW filter with 150 nm Al-Cu metallization
on LT, 20 nm line width variations correspond to about 75 ppm and 450 ppm
variation of the lower and higher frequency filter skirt frequencies, respec-
tively, and 0.5% variation of relative metallization height correspond to about
290 ppm variation of center frequency. In another example of a 500 MHz SAW
filter with 120 nm Al metallization on ST-cut quartz the dimensional standard
deviations mentioned above translate into about 30 ppm standard deviation
of center frequency for each of the line width and relative metallization height
variations. Together with about 30 ppm standard deviation caused by subse-
quent processes including mechanical stresses introduced during packaging,
the total center frequency standard deviation of the completed device thus
amounts to 52 ppm or 26 KHz.
After the metallization of the acoustically active areas, subsequent metal-
lization processes are needed to enhance the metallization height of bus bars
and bond pads for improved electrical conductivity and bonding ability and
to provide the multilayer under bump metallization, typically a Ti, Ni, Au
sequence, necessary for forming reliable solder ball joints in flip chip device
assembly. Passivation of SAW devices for improved reliability, i.e., protection
against corrosion and electrical shorts by conductive particles, can be achieved
by applying thin layers of insulating SiO2 or SiN, electrolytic enhancement of
the Al oxide layer or the fabrication of micro cavities on top of the acoustically
active areas consisting of two photolithographically structured organic layers.
Further wafer level processes comprise the application of organic acoustic ab-
sorbers for SAW IF filter devices and the application of solder bumps for
subsequent flip chip bonding to ceramic substrates.
Singulation of wafers into individual chips is performed by diamond blade
dicing for wafers between 500 and 250 nm thick. If thinner chips are required
for low profile packages, a dicing before the grinding process has to be applied
whereby a thicker wafer of, e.g., 350 nm is first half-diced from the top surface
to a depth of 130 nm and then ground from the back surface to achieve the
desired 130 nm chip thickness.
Encapsulation of SAW chips is to protect against environmental influ-
ences and to provide electrical interconnection between bond pads on the
chip surface and corresponding contact areas on the application circuit board.
Packages have to conform to the technology used for mounting components on
334 R. Veith et al.
the application circuit board, i.e., through the hole (TTH) or surface mount
technology (SMT). The development of special packaging technologies has
been the main driver for the miniaturization of SAW devices.
For application in TV sets or tuners, lead frame packages for TTH mount-
ing occupying the minimum of precious real estate on the circuit board are
required. A five pin single in line package has thus become the standard for
SAW TV-IF filters. In the package shown in Fig. 14.14 the SAW chip is bonded
to the lead frame with organic glue, serving also as an absorber for acoustic
bulk waves radiated towards the back side of the chip, and electrically con-
nected by wire bonding. The acoustic active area of the chip is protected by
an organic two layer structure forming a cavity allowing the SAW filter to
be encapsulated by molding with a standard thermosetting mold compound.
The outer dimensions of this package are 13.7 × 4.8 × 2.4 mm3 . World wide
production volume for polymer encapsulated lead frame packaged SAW filters
amounts to about 500 million pieces per year.
For mobile applications, low profile SMT packages are required. First mo-
bile phones used SAW filters packaged in leadless ceramic chip carriers with
wire bonded SAW chips and soldered or seam-welded metallic lids (Fig. 14.15).
The outer dimensions of these packages range from 19 × 6.5 × 1.6 mm3 to
3 × 3 × 1.1 mm3 . While today’s mobile phones require much smaller filter
packages, the hermetically sealed ceramic packages are still in use for high
reliability applications such as automotive keyless entry systems and for SAW
IF filters in mobile digital radios and TV tuners. World wide production vol-
ume for ceramic SMT packaged SAW devices amounts to about 200 million
pieces per year.
Further miniaturization as required by more complex multifunctional mo-
bile phones was achieved with chip-sized packages (CSP) utilizing SAW chips
flip chip bonded to a high temperature co-fired alumina ceramic substrate and
covered with a polymer laminate (Fig. 14.16). The outer dimensions of CSP
14 Advanced RF Signal Processing with SAW 335
SAW filters range from 2 × 2.5 × 0.74 mm3 to 1.4 × 1.1 × 0.4 mm3 . Functionally
integrated devices with more then one filter function may need larger pack-
ages. With multilayer low temperature co-fired ceramic (LTTC) substrates
high Q matching elements, as required for duplexers, can be integrated into
the package. With 1 billion mobile phones produced per year the world wide
production volume for CSP SAW devices amounts to about 4 billion pieces
per year.
The last step in miniaturization could be the wafer level packaging with
two organic or inorganic layers on top of the piezoelectric crystal chip forming
a cavity over the active area, a ball grid array of solder bumps on top of the
cavity and three-dimensional interconnect lines connecting bond pads on the
336 R. Veith et al.
chip surface with the corresponding under bump metallization on top of the
cavity. The outer dimensions of single RF SAW filters could thus shrink to
0.8 × 0.6 × 0.35 mm3 .
TV-IF filters have been the first application of SAW devices in a consumer
mass market. Highly selective SAW TV-IF filters with precisely shaped pass
band characteristics have become a necessary prerequisite for adjacent TV
channel occupation in cable and terrestrial TV as well as for higher resolution
pictures on flat panel displays.
World wide TV programs are broadcast in frequency channels spaced 8, 7
or 6 MHz apart, depending on the standard, over various very high frequency
(VHF) and ultrahigh high frequency (UHF) bands covering the frequency
range from 54 to 862 MHz. In classical analog TV broadcasting the trans-
mitted spectrum of a TV program comprises picture, color and sound carrier
frequencies modulated with the respective signals (Fig. 14.17). The TV-IF fil-
ter required in the receiver to select the wanted program and reject all others
has three distinct features. A Nyquist slope at the picture carrier, i.e., a filter
skirt perfectly symmetrical about the picture carrier required to completely
restore one side band after demodulation from the vestigial side band trans-
mitted, a sound shelf to suppress the sound carrier preventing interference
with the other carriers and to provide a flat pass band in the vicinity of the
6.000
20.00
40.00
45.00
50.00
Fig. 14.17. Analog TV broadcasting signal frequency spectrum and IF filter char-
acteristic (BG standard spectrum, BG intercarrier filter)
14 Advanced RF Signal Processing with SAW 337
sound carrier for good sound quality and deep traps to reject strong adjacent
channel carrier frequencies. In analog TV broadcasting any distortion in the
TV-IF filter pass band or any insufficient rejection of neighboring channels
will cause visible and audible deteriorations of picture and sound quality and
SAW TV-IF filters therefore have to fulfill tight specifications.
Digital TV broadcasting allows for a substantially more efficient utilization
of the EM spectrum. While analog TV needs one full 8–6 MHz wide channel
to broadcast one program, digital modulation allows transmitting one HDTV
program with six times the resolution or the equivalent number of standard
programs simultaneously over the same channel. Over the next decade, analog
TV broadcasting will therefore be replaced by digital broadcasting. In digital
TV broadcasting, the transmitted spectrum is evenly distributed over the
channel bandwidth (Fig. 14.18). The characteristic features of the IF filter are
a flat pass band, steep skirts and high rejection of adjacent channel signals.
In digital TV broadcasting, any distortion in the TV-IF filter pass band or
any insufficient rejection of neighboring channels will cause deterioration of
the bit error rate and channel information capacity and SAW TV-IF filters
for digital applications therefore have to fulfill similar tight specifications as
their analog counterparts, though sensitivity of degradation to certain parts
of the specifications may be different in both cases.
Both analog and digital TV broadcasting need IF filters with 10–20% rela-
tive bandwidth, precise pass band shaping, narrow transition bandwidth and
high out of band rejection. The exact relative bandwidth required depends
3.000
8 MHz
40.00
50.00
PC (B) SC (B)
31.9 MHz 40.4 MHz
PC (G) SC (G)
30.9 MHz 41.4 MHz
Fig. 14.18. Digital TV broadcasting signal frequency spectrum and IF filter char-
acteristic (DVBT spectrum with analogue adjacent channels, DVBT filter)
338 R. Veith et al.
Fig. 14.19. Typical SAW TV-IF filters in five-pin single in line lead frame packages
14 Advanced RF Signal Processing with SAW 339
0.05
–20
0.04
magnitude [dB]
–40 0.03
0.02
–60
0.01
–80
433.00 433.25 433.50 433.75 434.00 434.25 434.50 434.75
frequency [MHZ]
Fig. 14.20. SAW devices for remote keyless car entry systems (R 900 SAW
resonator: frequency response of conductance: B3575 SAW filter: frequency
response S21 )
340 R. Veith et al.
Fig. 14.21. SAW resonator and SAW filter for remote keyless car entry system
(Photo R900, B3575)
One major application of SAW filters are front end duplexers in frequency di-
vision duplex (FDD) and CDMA/WCDMA cellular phones. A duplexer allows
simultaneous receiving of signals from an antenna and transmitting signals to
it, with very high isolation between the Tx and Rx ports [22]. For this, two
RF-filters, with well-separated pass bands, are integrated on a ceramic sub-
strate (low temperature co-fired ceramics, LTCC, or high temperature co-fired
ceramics, HTCC). The transmit signal is passed through the Tx filter, while
signals entering the antenna are received by the Rx filter, whose pass band has
a higher frequency than that of the Tx filter. Both pass bands are separated
by a frequency gap (duplex spacing). For US-PCS and UMTS cellular phones,
both operating at about 2 GHz, the pass bandwidths are 60 MHz. The duplex
spacing in the case of the US-PCS amounts to only 20 MHz, while for the
UMTS case it is 130 MHz.
The requirements for these filters are very stringent. Besides low insertion
attenuation, a high rejection in the passband of the respective other filter
is required. In particular, high Tx–Rx isolation is needed. In order to pro-
vide sufficient frequency selection sufficient steepness of the passband edges
is necessary, in particular at the frequency gap, where temperature shifts and
fabrication tolerances limit the available space for filter roll-off. A further im-
portant requirement is the high power durability of the Tx filter. The choice
of material system and design technique to fulfill these requirements is exem-
plified now in US-PCS and UMTS 2 GHz filters.
14 Advanced RF Signal Processing with SAW 341
Within the last 20 years a rapid growth of the field of wirelessly readable
RFIDs has taken place [28]. This is mainly due to the cheap and compre-
hensive availability of complementary metal-oxide silicon integrated circuits
(CMOSIC). But also SAW devices gained an important sector for reliable
industrial applications of RFID and remote sensors. Here unique wirelessly
operating systems were developed, based on special piezoelectric devices.
All RFID systems work with a two-way (up–down) communications pro-
tocol: An interrogator (reader) transmits an RF request signal over an an-
tenna to the RFID device. After processing this request in the transponder it
answers with its ID and/or sensor measurement value (code). Two types of
transponders are usually distinguished: The passive one gets its energy for the
response signal from the reader by RF continuous wave (CW) or the request
signal. The active tag has a power source on board. Since the standardized
limit of the reader signal is rather low (10–30 mW), the achievable interroga-
tion distance for passive transponders is much smaller than for active tags.
On the other hand, passive tags have advantages in cost, reliability, lifetime,
weight and volume. So, for low-range RFID systems (e.g., admission control
for persons) passive silicon systems are used. However, for systems requiring
higher range and mechanical and jamming resistance (industry and traffic),
the special integrating signal processing possibilities of piezoelectric SAW de-
vices compensate for the cost advantage of silicon RFIDs.
The first SAW RFID system was invented by NYSEN, SKEIE and ARM-
STRONG [29] and implemented for cattle flock counting. As sketched in
Fig. 14.24a, the RFID tag is a beam substrate of an appropriate piezoelectric
material (LN, LT, etc.) with crystallographic orientation for SAW propaga-
tion parallel to its axis. The simplified sketch of Fig. 14.24a shows an exciting
interdigital transducer IDT0, on the upper surface of the substrate connected
to the tag antenna. We explain the operation of this SAW RFID assum-
ing an interrogation by a short RF burst at the centre frequency fm of the
IDTs [30,31]. This excites a SAW burst propagation towards the other reflect-
ing IDT1, IDT2, IDT4, etc. Each of the IDTs in the SAW path on the surface
of the tag reflects the SAW burst back to the antenna IDT0. This radiates
at the burst frequency fm to the interrogator, so that we get the response
pattern of Fig. 14.24b. The time intervals are given by twice the tag’s length
divided by the SAW velocity vs . Typically, we have 1 µs time between request
344 R. Veith et al.
a)
b)
c)
d)
Fig. 14.24. Sketch of SAW RFID structure and LFM read out signal scheme. (a)
Substrate with reflective IDTs. Antenna is connected to IDT0. External electrical
impedance Z is connected to IDT2. (b) RFID response requested by a short burst.
(c) Frequency dependence of response signal due to a LFM request chirp. (d) Audio
frequencies originating by self interference of LFM request chirp with response chirps
and first response impulse for switching the reader between transmission Tx
and reception Rx. This duplexer works at request impulse powers of typically
10 mW compared to many kilowatts of a radar Tx impulse. Here we should
note cautiously that special IDTs (e.g. single phase unidirectional) with shifted
locations are necessary to achieve such distinct responses as in Fig. 14.24b.
The location of the reflecting IDTs determines the tag’s binary identity
number (Fig 14.24b shows the code 1,1,0,1,0,1). For a road toll system in
Norway, a 22 bit code was used providing 220 (= 1,0242 > 1 million) possible
14 Advanced RF Signal Processing with SAW 345
identity numbers plus a start and a stop bit [32]. In this car toll system the
SAW tag together with its antenna is mounted behind the car’s windshield. No
electrical connection is necessary from the tag to the battery. The interrogat-
ing reader antenna is mounted on a steel frame above the highway. A pioneer
of SAW technology, Clinton Hartmann, has carefully investigated a possible
global SAW ID tag system in a 40 MHz range of the 2.4 GHz ISM band [33].
Making use of the signal processing capabilities given by the precise position
of the reflecting IDTs and corresponding possibilities of pulsed signal matched
filtering, the design of tag devices with capacities up to 256 bits is described,
showing high signal accuracy and low losses. For usual tagging systems 25–30
bits can be estimated to suffice, so that a large redundancy may be exploited
for error correction, as done in communications.
In the sketch of the passive ID tag signal response Fig. 14.24b we observe
long time intervals between the response bursts, determined by the tag geom-
etry only. This allows low noise signal processing after decay of the request
burst. So, additionally to response burst amplitude detection, applied to most
silicon ID tags, using RF interrogation bursts with stabilized burst period TB
and RF frequency fm, also phase measurements are possible for each response
burst of the SAW ID tag. This is accomplished using a quartz stabilized mas-
ter clock for TB and fm over many periods TB and integrating over this long
time the difference in amplitude and phase of each burst. Therefore very ac-
curate measurements of small shifts of the response bursts are feasible. Since
SAW tags at constant temperature and constant mechanical clamping are
crystallographically and electrically stable, no difference in burst response is
observed for these. But changes of tag temperature and/or mechanical strain
(e.g., acceleration) cause observable substrate length effects on the measure-
ment results, i.e. mainly the relative phase of the response burst. These simple
wirelessly readable ID tag strain and temperature sensors could be realized
only with piezoelectric single crystal SAW substrates.
Additionally to this remote mechanical strain and temperature sensors the
SAW tag of Fig. 14.24 together with accurate signal processing creates the
possibility of wirelessly readable sensors for electrical impedance. As shown in
Fig. 14.24a an electrical impedance Z can be connected as electrical load to one
of the reflecting IDTs. Here the Z-loaded IDT2 is situated between the shorted
IDT1 and IDT4, which serve as reference of SAW amplitude and phase in the
response signal. As described by Hashimoto [34], change of electrical load Z
of an IDT causes a definite change of its acoustical reflectivity in amplitude
and phase. Thus, by observing the response of a loaded SAW-RFID we can
remotely measure resistance R, inductance L and capacity C contained in Z.
Two examples may be mentioned: (1) As changing capacitive load the
abrasion of a piston bearing ring of a working natural gas pump was measured
[35]. (2) Magnetic field and/or electric current were measured wirelessly by
using a giant magneto impedance effect wire as load Z [36]. Its impedance
is very sensitive to a magnetic field yielding great effect in the corresponding
SAW sensor response burst [37].
346 R. Veith et al.
losses, steepness of filter skirts and fabrication tolerances affecting the center
frequency. To this end, first of all, processes have to be developed for the
deposition of highly reproducible top layers and electrode gratings with low
acoustic and electric losses. Second, the sensitivity of layered structure devices
against dimensional variations of heavy electrodes, both electrode height and
electrode width, will become much larger than in the case of relatively light
Al electrodes on piezoelectric single crystal surfaces. As the dimensional con-
trol of Al electrodes of SAW filters on piezoelectric single crystal materials
has already reached its physical limits, same or lower fabrication tolerances
for layered structure devices can only be achieved with appropriate trimming
processes that are still yet to be developed. Innovative approaches may be-
come necessary in order to provide means for trimming structures on buried
inner surfaces.
With these new challenges being met, signal processing with acoustic waves
propagating on or in piezoelectric materials will continue to play an impor-
tant part in further developments of wireless communication systems. While
new materials, material combinations and processes will have to be developed
in order to meet growing requirements resulting from the growing need for
efficient utilization of available RF spectrum, the state of the art achieved so
far for SAW signal processing on piezoelectric single crystals does provide a
solid base on which these new technologies can be built.
References
1. C.E. Shannon, Bell Syst. Tech. J. 27, 379–423, 623–656 (1948)
2. R.M. White, F.W. Voltmer, Appl. Phys. Lett. 7, 314–316 (1965)
3. B.A. Auld, Acoustic Waves and Fields in Solids, vol. 1 (Wiley, New York, 1973)
4. K.-ya Hashimoto, SAW Devices in Telecommunications (Springer, Berlin 2000),
pp. 1–22
5. Lord Rayleigh, Proc. Lond. Math. Soc. 7, 4–11 (1885)
6. T. Makkonen, Appl. Phys. Lett. 82, 3351–3353 (2003)
7. W.S. Mortley, British Patent 988 102, May 1963
8. J.H. Rowen, U.S. Patent 3 289 114, Dec. 1963
9. W. R. Shreve, in Proc. 1977 IEEE Ultrasonic Symposium, 1977, pp. 857–861
10. T. Nishihara, H. Uchishiba, O. Ikata, Y. Satoh, Jpn. J. Appl. Phys. 34(Part 1,
5B), 2688–2692 (1995)
11. A. Sakurai, H. Nakanishi, Y. Yoshino, Jpn. J. Appl. Phys. 34(Part 1, 5B), 2674–
2677 (1995)
12. Y. Shimizu, M. Tanaka, T. Watanabe, in Proc. 1985 IEEE Ultrasonic Sympo-
sium, 1985, pp. 233–236
13. C.S. Lam et al., First Symposium on Piezoelectric Acoustic Waves and De-
vice Applications, Dec. 14–17, 2004, Ningbo, Zhejiang, China, download from
http://www.txc.com.tw/en/d support/03.html
14. G. Johnson, J. Foise, Encyclopedia Appl. Phys. 15, 365 (1996)
15. H.R. Stocker, R. Veith, E. Willibald, G. Riha, in Proc. 1981 IEEE Ultrasonic
Symposium, 1981, 78–82
350 R. Veith et al.
P. Muralt
15.1 Introduction
Microelectromechanical systems (MEMS) were born as a new technological
discipline during the 1980s (for an introductory textbook, see, for instance [1]).
The idea of the pioneers was to enlarge capabilities of integrated circuits based
on silicon beyond pure electronics by adding mechanical elements, which were
made of silicon and further materials of semiconductor technology. The ad-
dition of mechanics extended the application range of silicon technology to
motion sensors, pressure and force sensors, small actuators, and a number of
acoustic and ultrasonic devices, most importantly resonators for signal treat-
ment. In order to profit from the symbiosis with electronics, those mechanical
elements should, of course, be controlled by electronic signals. Evidently, this
new silicon technology makes sense only for small, miniaturized devices. The
technical advantage comes from the fact that powerful thin-film deposition
and patterning techniques as used for semiconductor fabrication allow un-
precedented precision of mechanics in the nano- to micrometer range. As a
large number of devices are produced in parallel on the same wafer (batch
processing), the cost level is acceptable in spite of expensive fabrication tools,
at least at high production volumes. Concerning processing, the chemistry of
silicon turned out to be very helpful: high etching rates of anisotropic wet
etching in a base solution (as, e.g., KOH) and anisotropic deep silicon etch-
ing in a plasma reactor are crucial issues in efficiently tailoring silicon. Over
the last 20 years, MEMS technology has became a proven and mature tech-
nology with many applications. While “MEMS” is still taken as a standing
brand name for the field, the actual MEMS field has become much wider than
stipulated by the notion of electromechanics, including thermal, optical, mag-
netic, chemical, biochemical, and further functional properties. Also, the main
material of the device is not necessarily silicon, but may be glass or plastics,
especially for biomedical applications.
It is expected that piezoelectric materials with their intrinsic electro-
mechanical coupling should and will play an important role in MEMS. Indeed,
352 P. Muralt
piezoelectric thin film devices based on ZnO were among the first ones to be
demonstrated [2]. In piezoelectric materials, a deformation occurs upon ap-
plication of an electric field (converse effect) whereby the obtained strains
depend, to first approximation, linearly on the electric field. The opposite
effect works as well: upon application of a stress, charges develop on the
surfaces of the piezoelectric body (direct effect). The most frequently ap-
plied bulk material is PZT, a solid solution with formula PbZrx Ti1−x O3 ,
which exhibits outstanding piezolectric properties at the morphotropic phase
boundary at x = 0.53. Considering the applications of bulk materials (in-
cluding thick films), we identify many typical features reflecting their unique
characteristics, which may be also unique on the micro- or nanoscale and
also useful for applications at small dimensions. The use of piezoelectrics in
scanning microprobe microscopy is justified by the high rigidity prevailing
during deformation and enabling sub-nanometer precision in all directions.
Piezoelectrics are used to generate and sense ultrasonic waves as needed in
ultrasonic imaging at frequencies of 1–50 MHz and in nondestructive testing
at similar frequencies by virtue of the large efficiency in energy conversion
between electrical and mechanical energy. In signal filtering, piezoelectric res-
onators play a dominant role in the range 0.1–10 GHz thanks to their high
quality factors, outperforming by far LC resonators. In injection valves for
combustion motors, piezoelectrics are applied thanks to large forces and high
operation speeds. Without discussing dimensional details, one can summarize
the advantages as follows:
• Strong forces or, alternatively, large excursions in bending structures
• Low voltage because of high dielectric constants
• High efficiency in energy conversion, and equivalently low noise detection
• Possibility of high-speed and high-frequency operation
• High acoustic quality of many piezoelectrics (not true in multidomain fer-
roelectrics)
• Last but not least, a more or less linear behavior.
For thin-film applications, the same principles remain valid; however, one
has to keep in mind that smaller dimensions lead, on the one hand, to smaller
voltages and, on the other, to weaker forces, smaller excursions, and higher
resonance frequencies. Given the large number of degrees of freedom in me-
chanics, the outstanding features of piezoelectricity will lead to a plethora of
applications and devices with application frequencies in a large range between
a few hertz and some gigahertz.
Piezoelectricity is not the only possible choice, though. There are also other
electromechanical phenomena available in the micro world, and not necessar-
ily the same as in the macro world. There is first of all electrostatic interaction
across an air or vacuum gap of a capacitor [3,4] or between comb fingers [5,6].
Even with the disadvantage of nonlinearity, electrostatic interaction can well
be used for actuators and, with a bias voltage, also for sensors and transducers.
Not so practical in the macro world (very high voltages needed), electrostatics
15 Piezoelectric Films for Innovations 353
growth on an electrical conductor. The nuclei grow first in all directions until
they touch each other. Afterwards, the grains grow in columns perpendicular
to the film plane (z-direction), as indicated in Fig. 15.1a. Each column is a
single crystal. If one specific crystalline axis stays perpendicular to the film in
each grain, meaning that each column grows in the direction of this specific
crystalline axis, one deals with a so-called textured film. A (001)-textured
film thus means that the (001) planes are parallel to the film plane, and
that the direction of the columns (the z-direction) coincides with the c-axis
of the crystalline lattice (denoting the indices in the film system by i, and
the one in the grain-crystal system by i , we can state 3 = 3 , and random
relation between 1 and 1 ). The control of the film texture is the only means to
control the orientation of the anisotropic piezoelectric materials. It follows that
only polar materials can be used as thin-film piezoelectrics (which excludes
quartz). In case of ferroelectric materials, a net polarization can be obtained
by reorienting domains in an external electric field (poling process). In case
of non-ferroelectric polar materials, such as the hexagonal wurtzites ZnO and
AlN, the growth process must not only provide for a film texture, but in
addition also for alignment of the polar direction (c-axis in this case). This is
indeed possible. In most cases films are deposited in (001) texture, or tilted
(001) texture (see later). Considering (001)-textured, polycrystalline columnar
film slab as a free body, we would find a thickness mode d33 that is equal to the
d3 3 of the single-crystal column. The in-plane ordering is random, resulting
in a cylindrical symmetry. This gives the same symmetry of the piezoelectric
tensor as for hexagonal or tetragonal crystal symmetry. It follows that in case
of wurtzites, d31 = d32 would be equal to d3 1 , and that the shear mode
coefficients d15 or d24 coincide as well. In perfectly (001)-textured films of
ZnO or AlN, as schematically drawn in Fig. 15.1a, we thus find single-crystal
properties again for the relevant piezoelectric coefficients.
In ferroelectrics, the situation is in general more complicated. Uniaxial
ferroelectrics such as the tungsten-bronze BaSrNb2 O5 are research topics [11,
12] but not (yet) established as thin-film materials. Much better established
are ferroelastic perovskite structures such as PZT. The polarization direction
is not uniquely defined, but may point along different crystallographic axes,
subdividing the material into domains. The poling of such a film leads to
an average polarization direction along the external electric field, which in
the case of a planar capacitor will lead to a situation as shown in Fig. 15.1a,
where the arrows now indicate the average polarization direction. We obtain
the same symmetry as a ceramic, i.e., cylindrical symmetry with one infinite
rotation axis, usually denoted as axis 3, and find again that the free film slab
should show the three coefficients d33 , d31 , and d15 as in a ceramic slab. In
PZT thin films, no trivial link to single-crystal values is possible. First, the
single-crystal values are not as well known, and second, domain sizes, grain
sizes, and textures are different from those of bulk ceramics. A single column
of a PZT film does not possess the average property of a PZT ceramic, simply
because it is a monocrystal, usually small compared to a ceramic grain, and
15 Piezoelectric Films for Innovations 355
1, 2
(a)
E E
(b)
‘P’
(c)
(d)
Fig. 15.1. Electrode systems for driving piezoelectric films: (a) Planar capacitor
structure with top and bottom electrodes. (b) Electric field with interdigital elec-
trodes grown on insulator. (c) Case of non-ferroelectric, polar film with uniform
perpendicular polarization. (d) Schematic drawing of strains induced by d15 be-
tween the electrodes, and d33 , d31 below the electrodes for the situation in (c)
high voltages are needed (actuator) or delivered (sensor) and that the film
is in compression instead of tension when working at high fields parallel to
the polarization. Such structures make very much sense in energy harvest-
ing applications [15], which require higher voltages than for sensors. Actuator
and transducer structures were recently investigated [16], showing promising
results. Theoretically, the effective piezoelectric coefficients should be larger
than in case of a parallel plate structure [16,17]. When combining the ID elec-
trodes with a continuous bottom electrode, the transverse effect (d31 ) is manly
employed in non-ferroelectric polar materials. The field direction changes sign
between neighboring electrodes, and can be used again for SAW excitation
with wurtzites. In the case of ferroelectrics, this structure makes no sense
without prior homogeneous poling of the film. This might be again a poling
by the deposition process, arc poling, or an extended top electrode that is
removed after poling.
Another important difference between thin-film and bulk materials lies in
the fact that thin films are used in a composite structure where the total
elastic properties are often dominated by the passive part of the structure.
This passive part may be a silicon cantilever, or a silicon oxide or nitride
membrane, for instance. The interaction with the substrate is very anisotropic.
At the interface along the in-plane directions (indices 1, 2), the piezoelectric
thin film and the substrate have identical strains. Perpendicular to the film
plane, the thin film is usually free to move, i.e., T3 = 0. As a consequence,
the complete equation of state needs to be analyzed. For illustration, a thin
film clamped to a much thicker substrate is considered (i.e., a piezoelectric
laminated plate). When a field is applied, the strains in the plane stay at zero
(S1 = S2 = 0); in-plane stresses and an out-of-plane strain are developed.
The equation of state using compliance tensor and dij coefficients reads now
as (for the average polar direction parallel to E-field: no shear components):
S1 = (sE E
11 + s12 )T1 + d31 E3 = 0,
(15.1)
S3 = 2sE
13 T1 + d33 E3 .
The first line of (15.3) allows derivation of the in-plane stresses, that is, an
effective e31,f = −T1 /E3 . The second line leads to the definition of an effective
d33,f = S3 /E3 :
d31
e31,f = E ,
s11 + sE12
(15.2)
2sE
d33,f = d33 − E 13
d31 .
s11 + sE 12
These two coefficients can be used to describe piezoelectric laminated
structures using a mixed variable set of (S1 , S2 , T3 ) in cases where the shear
strains can be neglected [18]. The clamping of the ferroelectric film on a sub-
strate affects piezoelectric properties in two ways. There is first the reduction
of d33,f as given by (15.2). A second problem arises from the elastic clamping of
15 Piezoelectric Films for Innovations 357
Top electrode
Bottom electrode
Q3, E3
External load
or displacement
Strain or stress
z, 3
y, 2
x, 1
Fig. 15.2. Input (E3 , load) and output parameters (displacement, Q3 ) in actuator
and sensor applications for piezoelectric laminated plates
The thin-film materials are in principle the same as the known bulk materials.
However, as they have to be synthesized on the device, various limitations exist
related to thermal and chemical compatibility. Such limitations are certainly
a problem for ferroelectrics with their many fast diffusing and volatile species
such as K, Na, Pb, etc. For this reason, integration needs to be studied as
intensively as the growth itself.
358 P. Muralt
15.3.1 AlN
There is at least one material that can be synthesized better as a thin film
than as a bulk single crystal: the wurtzite structure AlN (see Fig. 15.3). It
appears to be very difficult to synthesize single crystals with good piezoelec-
tric properties [28], whereas excellent piezoelectric properties are obtained at
films processed by reactive sputter deposition at very moderate temperatures
of 300◦ C or even below [29]. The growth mode leading to high piezoelec-
tric values can be ascribed to enhanced mobility of atoms (adatoms) at the
growing surface promoted by ion bombardment, and is typical for sputter-
deposited films under certain temperature–pressure/RF-bias conditions (Zone
T growth mode according to Thornton [30]). Since ion bombardment leads also
to compressive stress, the process needs to be carefully tuned to balance the
piezoelectric coefficient and stress [31]. A summary of properties is given in
Table 15.1.
N, O Al, Zn
Using a beam tool, the optimal PZT composition and orientation were
identified as the morphotropic composition (53/47 for Zr/Ti) with (100) ori-
entation [45] (Fig. 15.5). This is in fact the same composition as is used for
standard PZT ceramics. This composition excels with a large amount of do-
main activities as well as peaking of the dielectric constant and all piezoelectric
coefficients, especially d15 . It was observed that sol–gel films exhibit a compo-
sitional gradient that does not allow a very precise hitting of the maximum.
This problem could recently be corrected by a preventive variation of the
solution composition [47].
A good indicator of the progress is the published value of the effective
in-plane transverse piezoelectric coefficient e31,f , which is very relevant for
MEMS applications and indicates at the same time a good density, and is
conveniently measured in the direct [20–22, 48] and converse mode [26]. In
Fig. 15.6, the reported e31,f is shown as a function of the year of publication.
A constant progress is observed in the case of PZT thin films integrated on
silicon. The major integration and texture control issues were solved in the
1990s. It is instructive to note that now integrated films on silicon outperform
the epitaxial films, which were better initially, since in their case no integration
and seeding issues had to be solved. The lead-based epitaxial films show always
the same values in the range of −11 to −13 C m−2 , whether it is a PZT or a
relaxor PMN-PT (Pb(Mg, Nb)O3 -PbTiO3 ).
362 P. Muralt
Fig. 15.6. The history of the transverse piezoelectric coefficient e31,f as an in-
dicator of the advancement in the field. The films are in the 0.5–2 µm thickness
range. PZT thin films integrated on silicon (data from [20, 21, 26, 45–47, 49])
are compared with several types of lead-based epitaxial films (PZT [50, 51]),
BiScO3 –PbTiO3 [52], Pb(MgNb)O3 –PbTiO3 [53–55], relaxor type integrated thin
films such as Pb(Yb, Nb)O3 –PbTiO3 [56–58]. The values derived from ceramics data
refer to early PZT ceramics (Jaffe), standard hard PZT 4 (product of Morgan Elec-
troceramics), and modern optimized PZT ceramics, the 3203 of Motorola. The small
numbers aside of the points are the film thickness in micrometers (from [17])
electrode: Pt as used for integrated films is a true metal that can screen (or
compensate) the polarization charges on a very short range by free electrons.
This is quite different when using an oxide electrode. Charge compensation
involves a deeper zone, and might even involve oxygen vacancies. Furthermore,
the nature of chemical bonds across the interface is different. Stress-driven Pt
diffusion at the interface is known to happen at the growth temperature [63].
All such effects could lead to differences in dislocation densities and strain
relaxation close to the interface, and thus eventually to stronger ferroelastic
domain wall pinning near oxide electrodes, which would impede domain wall
contributions to piezoelectricity at the small-signal response level.
Relaxor materials do not seem to give any advantage over PZT to date.
This can be explained by the following: (1) Relaxors are more difficult to nucle-
ate and to grow, thus requiring higher growth temperatures. This complicates
integration tasks; (2) The still relatively low growth temperatures in compar-
ison to bulk materials do not allow for a very homogeneous distributions of
the disordered B-site ions, leading to a much lower dielectric constant (see
discussion in [17,64]); (3) The piezoelectric eij coefficients express mechanical
stress per applied electric field, and are thus proportional to the stiffness. The
high strain response of relaxors does not equally translate in stress response
because relaxors are softer.
Ignoring domain effects, well-oriented or textured films should exhibit su-
perior properties than ceramics. This is indeed achieved comparing with pure
PZT ceramics (such as the old material published by Jaffe and Berlincourt
[65]) and with standard PZT 4 (Morgan Electroceramics). However, there are
modern ceramics with elaborate substitutions and dopants that are still su-
perior to the best PZT thin films (Motorola’s 3203 in Fig. 15.6). One can thus
argue that improvements are still possible.
(a) (b)
Fig. 15.7. (a) Ferroelastic domains in a (111)-oriented, tetragonal PZT film grown
as epitaxial film by reactive sputter deposition on a SrTiO3 (111) single crystal
(process described in [66]), as revealed by dark field TEM images. (b) Schematic
drawing of domain patterns corresponding to the observed ones
for the properties. On the one hand they may reduce properties because of an-
tiparallel alignments; on the other, the strain changes by domain wall motions
in an electric field may exceed the intrinsic effect of ionic displacements in the
crystal lattice [67]. Furthermore, higher-order terms have to be added to the
linear piezoelectric effect when domain wall motions become relevant [68].
In PZT ceramics, about 40–60% [67, 69] of the piezoelectric small-signal
response is due to ferroelastic domain wall motions. In an electric field, do-
mains change volumes depending on whether their potential energy −E
P
is
increased or reduced. In the switched volumes, the direction of the sponta-
neous strains change, yielding a net strain change in the ferroelectric body. In
thin films, ferroelastic domains adapt to strain or stress imposed by the sub-
strate in order to reduce the elastic energy [70]. In very thin films, such strains
are due to epitaxial mismatch [71, 72]; in thicker films, the cause is the ther-
mal mismatch upon cooling down from the growth temperature. Considering
a tetragonal {100}-oriented film, a domain pattern is formed by compensating
tensile stress by switching the c-axis into the plane (forming an a-domain),
and compensating compressive stress by switching the c-axis out of plane
(c-domain). The laminar c/a/c/a domain pattern is a possible solution to the
elastic problem [73–75]. The ratio of a- to c-domain volume fraction is given by
substrate clamping and thermal mismatch history. However, the domain walls
may still be free to move in an electric field compensating the increase in elas-
tic energy. Analytical model calculations of the extrinsic piezoelectric effect
(d33,f ) in thin films are available for an ideal c/a/c/a domain pattern [74].
For many years not much evidence of such piezoelectric domain contribu-
tions was found in thin films. It was thought that ferroelastic domain walls
were very much pinned by defects. This idea was supported by hot-poling
experiments of tetragonal {100} PZT15/85 films [76]. It was observed that
hot poling helped to reduce the a-domain faction. This was attributed to un-
pinning from defects, mainly oxygen vacancies, since oxygen vacancies may
move at the applied temperature of 150–200◦ C. But after such poling, domains
15 Piezoelectric Films for Innovations 365
were even more fixed than before (which is advantageous for pyroelectric de-
tectors). Later, it was observed that more than 3 µm thick polycrystalline films
deliver significant contributions from ferroelastic domain motions as indicated
by considerable nonlinear effects in piezoelectricity [59, 60].
In recent years, much progress in hunting domain phenomena has been
made thanks to piezoelectric-sensitive atomic force microscopy (AFM) mea-
surements [77–79]. Local piezoelectricity can be measured, and even individ-
ual domains can be observed. When patterning a 200-nm thick PZT 40/60
film into dots with widths in the 100–200 nm range, therefore smaller than the
height, an increasing piezo-response with increasing aspect ratio was found [66]
(Fig. 15.8). As unclamping can account for at maximum a doubling of the re-
sponse, the observed fourfold increase was attributed to unpinning of 90◦
domains and to the major domain reconfiguration taking place in addition
to simple unclamping. Surprisingly, such a drastic increase took place just
by patterning, i.e., without “additional help” such as, for example, anneal-
ing or poling, indicating that ferroelastic domain walls must be nevertheless
quite mobile once the clamping is removed. Recently, a fivefold increase of the
piezoresponse close to the corners of patterned features was observed when the
film region was additionally subjected to poling before the measurement [80].
In a very recent paper, it was reported that the response at a 90◦ domain
800
600
Amplitude of piezoresponse (a.u.)
100 nm
600
Amplitude of piezoresponse (a.u.)
-400 200
-600 100
-800 0
-20 -10 0 10 20 0 200 400 600 800 1000 1200
DC bias to tip (V) Lateral size (nm)
wall boundary can be detected even on a continuous film [81]. It can be con-
cluded that the mechanical clamping to a substrate leads to a very significant
stiffening of ferroelastic domain wall motions.
Fig. 15.9. Schematic drawing of an ink jet printing device based on piezoelectric
thin films (after [106])
15 Piezoelectric Films for Innovations 367
0.004
K2 = 5.3 % a1-df–res1-20dc
Re
K = 23 %
0.003 Q = 135
C0 = 90 pF
Admittance(Ω–1)
0.002 Cpara =0
0.001
lm
–0.001
SOI,4.5 mm Si
1.2 mm SiO2, 2 mm PZT
–0.002
730 735 740 745 750 755 760 765 770
(a) (b) Frequency (kHz)
BAW devices are known for their efficient transformation of electrical power
into mechanical power and vice versa. A longitudinal ultrasound wave is
trapped in an electroded, half-wavelength-thick slab (thickness tp ) of piezo-
electric material having its polarization along the wave propagation (see
Fig. 15.11). In the bulk world this principle is very successfully applied in
ultrasound probes for medical imaging and nondestructive testing at a few
MHz. The antiresonance frequency fa ≈ vs /(2tp ) is inversely proportional to
the thickness tp . Thin-film BAW resonators (TFBARs) resonate at frequen-
cies of a few gigahertz. The high sound velocity vs of AlN of 11,000 m s−1
yields roughly 2 GHz at 2 µm (in reality somewhat less due to mass loading
by electrodes). The typical film thickness range is thus ideal for RF filters
in mobile communication. (The US-CDMA standard requires, for instance, a
duplex filter centered at frequencies 1960 and 1880 MHz for the receive and
the transmit line, respectively.) The elastic energy trapped in the TFBAR
constitutes an energy tank of high quality as needed in RF signal process-
ing. TFBARs have been investigated since the early 1980s, initially mainly
motivated by applications in television filters. The required frequencies were,
however, too small, meaning that the required film thickness was too large
to be meaningful. The competing SAW devices were more easily produced
at these low frequencies. In addition, sputter deposition was not yet the op-
timized technique with outstanding uniformity and reproducibility as it is
today. The advent of mobile communication and the ever-increasing process
control in semiconductor technology, together with the momentum of MEMS
research, have led to the convergence of all requirements for a new product
called TFBAR RF filter. Today, it is one of the most successful MEMS prod-
ucts. The present market of around 100 million dollars is believed to extend
to above 1 billion dollars.
There are two key properties for the choice of the piezoelectric material in
TFBARs: the coupling coefficient kt2 of the thickness mode and the material
quality factor Qm :
e2 1
kt2 = D 33 S , Qm = . (15.3)
c33 ε0 ε33 ωτ1
Table 15.2. Relevant materials parameters for TFBAR applications for the two
most used wurtzite thin films, as derived from standard literature data [89, 90]
Fig. 15.12. Setup of the FBAR biosensor: (a) top view of the resonator and the
electrical leads; (b) electron microscope picture of c-axis inclined ZnO; (c) schematic
picture of the lateral structure comprising the resonator with two electrodes solidly
mounted on an acoustic Bragg mirror (from [102])
1000
Rey-Mermet
et al.
Gabl et al.
100
BAW
Harding and Du
Log(S) [m2/kg]
10 SAW
1
Kösslinger et al.
Fig. 15.13. overview on sensitivity results obtained with quartz micro balances
(QCM) [97], SAW devices [98], and BAW devices [99, 101]
With the advent of TFBARs, one can imagine the realization of miniaturized
biomedical, chemical, and environmental sensors with enhanced functionality,
as their integration into MEMS fluidic systems follows quite naturally and
the smaller size due to the higher frequency would additionally allow for large
arrays with strong screening capabilities. Work in this field is already quite
advanced (see Fig. 15.12).
Such sensors were first realized in versions operating with longitudinal
waves [99, 100]. The increase of sensitivity with frequency is indeed observed,
as shown in Fig. 15.13, which depicts the sensitivity reported in some selected
15 Piezoelectric Films for Innovations 371
works [101]. The BAW numbers given in this figure correspond to longitudinal
modes. Such devices could be useful as gas detectors. A thin polymer layer on
the resonator would serve as the gas-sensitive layer changing its mass and its
elastic constant with absorption of the gas.
Detection in a liquid, as required for biomedical sensors, requires the use of
shear mode BAWs. Longitudinal BAWs emit too much power into the liquid,
causing excessive damping. Ideally, shear modes have no amplitude perpendi-
cular to the film plane, and thus excite no emission in the liquid. The resonator
is less damped and a good resolution is still possible [102–104]. The excitation
of a shear wave requires an electric field that is inclined with respect to the
polar axis of the film (S5 = d15 E1 ). This requires either the growth of polar
films with a tilted polar axis in combination with a usual planar capacitor,
or the use of ID electrodes. The growth of tilted grains requires first the
nucleation of grains that are not (001) oriented, and furthermore an oblique
flux of the arriving atoms from the sputter target in order to give a growth
advantage for non-(001)-oriented grains [103, 105]. It is not evident to achieve
a homogeneous film on a large wafer. By tilting the wafer away from the usual
face target situation, the deposition uniformity becomes very bad.
In general, the sensitivity S of such a device is defined as the relative
frequency change ∆f /f0 of the resonance (originally at f0 ) upon loading with
a surface mass density µ (kg m−2 ):
1 |∆f |
S= lim . (15.4)
f0 µ−>0 µ
If we consider the simple case when loading with a material of identi-
cal acoustic properties as the piezoelectric (same acoustic impedance, same
sound velocity) and identical density ρ, one obtains the frequency f (µ) and
sensitivity S as:
2
vs 1 vs 1 2
f (µ) = , S= = f0 . (15.5)
2 t0 + µ/ρ 2f0 ρ tp vs ρ
k2 k2
D≈ ∝
πτ1 vs ρ τ1 c1/2 ρ1/2
Of course, there are many other parameters influencing losses and noise
levels, certainly the electronics. A real thin-film resonator is not of a homoge-
nous material, but contains electrodes and passive layers. A simple figure of
merit as D is only part of the story. In Weber et al. [102], detectivities were
determined by experiments. They showed that their shear-mode TFBAR was
indeed better than a quartz microbalance, but not by a factor 100 as the
sensitivity would suggest, but only by a factor 4. The reason lies in the in-
crease of noise level with frequency. Since TFBAR sensors are new, one can
nevertheless expect that some improvements are still possible.
To conclude this chapter, it is certainly well justified to recognize the large
progress in piezoelectric thin-film MEMS over the last two decades. Materials
have been very much improved, property–process relations are now much bet-
ter understood, and integration in micromachined structures has been well
mastered. Whereas AlN TFBARs have achieved industrial validation, PZT
and other perovskite materials still need to prove their reliability in industrial
or consumer applications.
References
1. N. Malouf, Artech House 265 (2000)
2. K. Uozumi, K. Ohsone, R.M. White, Appl. Phys. Lett. 43, 917–919 (1983)
3. H.T. Soh, I. Ladabaum, A. Atalar, C.F. Quate, B.T. Khuri-Yakub, Appl. Phys.
Lett. 69, 3674–3676 (1996)
4. A.S. Ergun, G.G. Yaralioglu, B.T. Khuri-Yakub, J. Aerospace Eng. 16, 76–84
(2003)
5. V.P. Jaecklin, C. Linder, N.F. de Rooij, J.-M. Moret, Sensors Actuators A 39,
83–89 (1993)
6. P. Cheung, R. Horowitz, R.T. Howe, ASME Micormech. Sens. Actuators Syst.
32, 269–278 (1992)
7. E. Quandt, A. Ludwig, Sens. Actuators 81, 275–280 (2000)
8. J.F. Scott, C.A. Paz de Araujo, Science 246, 1400–1405 (1989)
9. O. Auciello, J.F. Scott, R. Ramesh, Phys. Today 51, 22–27 (1998)
10. H. Raeder, F. Tyholdt, W. Booij, F. Calame, N.P. Ostbo, R. Bredesen,
K. Prume, G. Rijnders, P. Muralt, J. Electroceram. 19, 357–362 (2007)
11. A. Infortuna, P. Muralt, M. Cantoni, N. Setter, J. Appl. Phys. 100(10),
104110 (2006)
15 Piezoelectric Films for Innovations 373
16.1 Introduction
Piezoelectrics and electrostrictors are indispensable materials for use in trans-
ducer technology, as they inherently possess both direct (sensing) and con-
verse (actuation) effects. A piezoelectric/electrostrictive sensor converts a
mechanical input (displacement or force) into a measurable electrical out-
put through piezoelectric/electrostrictive energy conversion. In the case of a
piezoelectric, an applied mechanical force (stress) induces a voltage across
the terminals of the transducer. On the other hand, an applied mechanical
force induces a change in the capacitance of an electrostrictive transducer
that could be electrically detected. Hence, the mechanical to electrical en-
ergy conversion is accomplished directly when a piezoelectric is used, while
the same is obtained indirectly if the electroactive material of choice is an
electrostrictor. Conversely, both piezoelectric and electrostrictive materials
develop an elastic strain under an applied electric field. The said elastic strain
is linearly proportional to the applied field in a piezoelectric, whereas elec-
trostrictive coupling involves the second-order (quadratic) coupling of electric
field with elastic strain. While piezoelectricity is possible only in noncentrosy-
metric point groups, electrostriction is observed in all solids, which make it a
much more general solid-state phenomenon. Sensing and actuation functions
can coexist in a given transducer by the intelligent use of such materials.
Piezoelectrics and electrostrictors, therefore, constitute the backbone of mod-
ern transducer technology, as mechanical to electric energy (and vice versa)
conversion can be accomplished with great efficiency in a way that is second
to none among all phenomena known to date [1, 2].
The direct piezoelectric effect enables a transducer to function as a passive
sound receiver or as a pickup by converting sound pressure into an electrical
signal. Applications in which such electromechanical transducers are used in-
clude hydrophones for underwater low-frequency noise detection and micro-
phones, among others. On the other hand, the converse piezoelectric effect
permits a transducer to act as an active sound transmitter or a loudspeaker.
378 A. Safari and E.K. Akdoğan
When used in the pulse-echo mode, a transducer can perform both active
and passive functions in a quasi-simultaneous fashion. This mode of opera-
tion entails the emission of an acoustic wave into the medium of interest, and
sensing the reflection (echo) back onto it. Such echoes are produced when
sound waves are reflected back from an interface (boundary) between two
substances of different acoustic impedances [3]. The magnitude of the echo is
proportional to the acoustic impedance mismatch between the two materials
forming the interface and the medium in which the wave travels. Electrostric-
tive materials, on the other hand, can be used as a high-displacement actuator
since the strain is coupled to the applied field quadratically. Such materials
have also been used as sensors where the change in permittivity with applied
stress constitutes the mechanism for sensing.
The design and fabrication of composite materials opens new avenues to
optimizing the electrical, magnetic, and mechanical properties for specific ap-
plication. Properties that are otherwise impossible to achieve with constituent
materials in single-phase fashion can be synthesized by the use of several con-
nectivity patterns (see Sect. 16.3). Piezoelectric composites originated from
this line of reasoning to tailor the input and output responses of an electro-
mechanical transducer to a specific application. For instance, one may wish to
maximize the piezoelectric sensitivity, minimize the density to obtain a good
acoustic matching with water, and at the same time, make the transducer
mechanically flexible to conform to a curved surface [3]. These properties are
mutually exclusive; and Mother Nature does not provide us with single-phase
materials that simultaneously satisfy such requirements. Thus, in many ap-
plications, one might meet conflicting design requirements by combining the
most useful properties of two or more phases through the effective use of black
and white symmetry – the essence of the composite approach.
What is elaborated herein appertains to the development of piezoelectric
ceramic–polymer composites and novel actuators by solid freeform fabrication
(SFF). This advanced fabrication method brings unique degrees of freedom
to composite making, which is likely to set the stage for the prototyping of
next-generation piezoelectric composites in the twenty-first century.
Fig. 16.2. Examples of various composites with reference to their connectivity pattern
381
382 A. Safari and E.K. Akdoğan
Filament
Rollers
Heater
Filament Nozzle
Spool
Slice Thickness
Platform
system. Third, the so-called stl file is converted into cross-sectional slices,
or a slice file, in which each slice can be uniquely defined about its build
strategy by varying the tool path. The slices collectively define the shape
of the part. Finally, the information related to each slice (i.e., toolpath) is
then transmitted in a layer-by-layer fashion to the SFF machine. In Fig. 16.3
a schematic representation of the FDC process is shown. The lowering of
the fixtureless platform by one slice thickness in the z-direction follows the
completion of each layer. Subsequent layers are built on top of the preced-
ing one in a sequential manner, and the process is repeated until the whole
part is finished. Each layer fuses onto each other, forming a solid contin-
uum thereby. The process typically uses a filament feedstock made out of a
thermoplastic polymer in which is dispersed a ceramic or metal, using the
principles of colloidal chemistry. The sequential or layered approach of man-
ufacturing a 3D object of arbitrary shape is indeed the quintessential feature
of SFF. In other words, the way SFF works is to mimic Mother Nature’s
method of putting together a 3D crystal structure, which is one atomic plane
at a time.
In the fused deposition of multiple materials (FDMM) process, the same
approach is pursued as in the FDC process; however, a multihead assembly is
used instead of a single-head assembly. The multihead assembly enables one to
build 3D structures comprising two or more materials in a sequential fashion.
16 Piezoelectric Composites by Solid Freeform Fabrication 383
While the FDMM process offers great flexibility in the design and fabrication
of multimaterial performs, co-densification is a major challenge owing to the
differential shrinkage of each material. However, judicious materials selection
as well as the fine tuning of the solids loading in the filament feedstock can
circumvent this difficulty [6, 21, 22].
SFF techniques provide many advantages for the manufacturing of ad-
vanced functional components over traditional ones. A part normally takes
weeks or months to fabricate because of the time spent in fabrication of the
mold and tools, machining operations, and so on. In contrast, the same part
can be prototyped in a few days’ time using SFF, providing reduced lead times
and costs in the development of new functional parts thereby. This technology
is advantageous to use in cost-effective iterative design to produce components
with optimum shape and desired properties.
The feasibility of the FDC process in making piezoelectric composites
was demonstrated in the fabrication of various PZT/polymer composites with
different connectivity patterns. The FDC process was used via two distinct
processing routes: (1) direct FDC and (2) indirect technique or the lost mold
process where polymeric molds were manufactured by FDM. In the direct
processing route, FDC is used with a PZT-powder-loaded polymer filament
as the feed material for a direct-layered manufacturing of the 3D green ce-
ramic structure. The indirect route was used in fabricating a polymer mold
in which the negative of the desired structure is formed via the FDM process
using the Stratasys commercial filament (ICW-04) and Sanders prototyping.
The mold is infiltrated with the PZT slurry and then placed in a vacuum oven
to ensure complete filling of the voids. The samples are then dried in an am-
bient atmosphere (∼2 h), followed by drying at higher temperatures (∼70◦ C,
2 h). The initial slow drying step has to be included to reduce the possibility
of cracking in the slip. The thermoplastic polymer mold is evaporated during
the early stages of a specifically designed binder burnout cycle.
The optimum solids loading used in the production of filaments are typi-
cally 50–55 vol.%, which utilizes a four-component thermoplastic binder sys-
tem containing a polymer, tackifier, wax, and plasticizer. It was found that
powder pretreatment with an organic surfactant before compounding is crit-
ical in obtaining feedstock materials that can be extruded. Filaments of
1.78 mm in diameter are typically used in the process. Fused deposition of
these filaments is accomplished using the 3D Modeler (Stratasys, Inc.) or an
in-house-built FDMM system.
Previous research which harnessed the design guidelines provided by the
connectivity pattern formalism has unequivocally shown that composites fab-
ricated by SFF exhibit superior piezoelectric properties in comparison to
single-phase piezoelectric ceramics as well as composites made by traditional
methods. In the following section, we will present some examples of composites
and actuators which were fabricated by SFF.
384 A. Safari and E.K. Akdoğan
–5
Amplitude (dB)
–10
Regular
Gaussian gradient
–15 Linear gradient
Exponential gradient
0 2 4 6 8
Distance (mm)
Fig. 16.4. Variation of acoustic amplitude with distance from the center in a volume
fraction gradient (2–2) composite as a function of gradient type
and moderately low values of the kp . The high kt in such composites was
attributed to the high aspect ratio of the ceramic elements (>6). The longitu-
dinal piezocharge coefficient (d33 ) of such composites was found to be greater
than 400 pC N−1 (see Table 16.1).
The vibration profiles of the (2–2) composites with and without VFGs
are shown in Fig. 16.4 as a function of the distance from the center. The
pressure output of the VFG designs were compared with a regular (2–2)
composite, which was obtained from a diced PZT-5H ceramic (wall width:
220 µm; Kerf: 440 µm). The regular design had ∼33 vol. % ceramic with a d33
and kt of 340 pC N−1 and 63%, respectively. As expected, the amplitude of
the vibration for this composite did not vary much with the distance from the
center. The small decrease in the amplitude at the edges was attributed to
the clamping of the edges. On the other hand, all VFG distributions showed a
decrease in the vibration amplitude as distance from the center increased (see
Fig. 16.4). The Gaussian VFG was designed to have a very slow decrease in ce-
ramic volume fraction with distance from the center. The vibration amplitude
386 A. Safari and E.K. Akdoğan
observed in Fig. 16.4 followed a similar trend, initially remaining constant and
then dropping by only −4 dB at a distance of 5 mm. Hence, this distribution
was found to be very different from the regular (2–2) composite at the same
distance from the center. The linear gradient showed a very smooth decline
in the vibration output with distance (see Fig. 16.4), linearly falling to about
−8 dB at a distance of 5 mm. The exponential distribution had a vibration
profile that was slightly lower in amplitude than the linear distribution. It
should be noted that although the ratio of the total ceramic content in the
center and the edges is constant for all VFG distributions, the pressure output
ratios are not similar, which was attributed to the limitation in the measure-
ment techniques [22]. The precise manipulation of the volume fraction within
the same composite allowed for the control of the vibration amplitude profile
to obtain the desired pressure output as seen in Fig. 16.5. The profile for the
Gaussian vibration amplitude did not fall as fast, and was very similar to
the regular distribution. The linear or the exponential gradient showed a very
smooth decrease in the vibration from the center to the edges.
0
–10 Regular 2-2
Signal Amplitude (dB)
Gaussian VFG
–20 Linear VFG
–30
–40
–50
–60
–70
–80
–90 –70 –50 –30 –10 10 30 50 70 90
Angle (degrees)
Fig. 16.5. Variation of signal amplitude with angle in a volume fraction gradient
(2–2) composite as a function of gradient type
Turcu et al. [30, 31] studied oriented (2–2) soft PZT-Epoxy composites
that were fabricated by FDC, which are depicted in Fig. 16.6. The inspiration
for this study came from the modeling work of Nan et al. [32, 33], who the-
oretically investigated the effects of orientation of the ceramic phase relative
to the poling directions in (0–3) and (1–3) piezocomposites using a Green’s
function formalism. Nan et al.’s modeling results showed that the dielectric
and piezoelectric properties of the piezocomposites with (1–3) connectivity de-
creased when the poling direction deviated from the orientation of the ceramic
phase. In Turcu’s work, the orientation angle of the ceramic phase relative to
the poling direction varied in the range of 0–75◦ with 15◦ increments (see
16 Piezoelectric Composites by Solid Freeform Fabrication 387
Fig. 16.6. Cross-sectional photographs of oriented (2–2) and (3–3) composites with
various orientation angles
200
150
100
d31 (x 10–12 C/N)
50
0
0 15 30 45 60 75 90
– 50
Fig. 16.6) [30, 31]. The volume fraction of the ceramic phase in both com-
posites was fixed at 0.30. It was found that the effective dielectric constant
and effective d33 coefficients of both composites decreased with increasing ori-
entation angle. However, the effective d31 coefficient became positive at 45◦ ,
reaching a value of 150 pC N−1 as shown in Fig. 16.7. The data showed a max-
−1
imum in dh gh at ∼45◦ as well, reaching a magnitude of 15,000 × 10−15 m2 N
−1
as depicted in Fig. 16.8. On the other hand, the d33 was ∼430 pC N up
to 45◦ , followed by a sharp decrease to 100 pC N−1 at 75◦ . Turcu et al. also
studied oriented (3–3) soft PZT–epoxy composites by FDC (see Fig. 16.6).
These composites exhibit an 85% increase in piezoelectric voltage coefficient
388 A. Safari and E.K. Akdoğan
60000
2-2 composite
50000
3-3 composite
dhgh(10–15m2/N)
40000
30000
20000
10000
0 15 30 45 60 75 90
degrees
Fig. 16.8. Orientation angle dependence of the hydrostatic figure of merit for ori-
ented (2–2) and (3–3) composites
g33 at an orientation angle of 45◦ . The FOM of the oriented (3–3) composite
−1
reached an extremely high value of ∼50,000 × 10−15 m2 N , while is much
higher than the obtained in (2–2) composites as shown in Fig. 16.8.
A composite with (3–3) connectivity could also be envisioned as two phases
mutually penetrating each other, thereby forming two 3D self-connected net-
works in intimate contact. The feasibility of the FDC process for making
piezoelectric composites was demonstrated in the fabrication of PZT/polymer
3D honeycomb and ladder composites with (3–3) connectivity patterns [6,10].
The ladder structures were built by using a raster-fill strategy with a fixed
inter-road spacing. The consecutive layers were built at 90◦ to each other.
The volume fraction of the ceramic phase in the structure is approximately
70%, and can be varied by changing the width and spacing between the ce-
ramic roads. In this particular design, the ceramic roads are ∼300 µm wide
with ∼800 µm center-to-center spacing. The mesoscopic structures created by
FDC are very uniform with excellent unit cell repeatability, unequivocally
demonstrating the superiority of the FDC process over traditional methods
of composite making [6]. The indirect method was also used to fabricate 3D
honeycomb composites. The scanning electron micrograph in Fig. 16.9 depicts
a structure of a sintered ceramic prior to epoxy infiltration. The structure
consists of a 3D lattice composed of “air pipes” defining the 3D piezoelectric
skeleton (matrix). The “air pipes” were created upon burning out the sac-
rificial polymer. By virtue of the flexibility offered by the FDC process, the
diameter of the holes, the spacing, and the volume fraction of the ceramic
phase can be varied between the holes. In the example given in Fig. 16.9,
the diameter of the holes is ∼200 µm, with ∼350 µm center-to-center uniform
spacing, and the PZT content is ∼25% by volume [6, 34–36].
16 Piezoelectric Composites by Solid Freeform Fabrication 389
Z [001]
X [100]
Y [010]
Fig. 16.9. Scanning electron micrograph of an air-pipe composite with (3–3) con-
nectivity showing the ceramic phase prior to infiltration with polymer
Monolithic ceramic tubes, like their planar counterparts, suffer from simi-
lar deficiencies such as high weight, high capacitance, and low figure of merit
dh gh . Until recently, the fabrication of functional piezoelectric composites with
radially oriented connectivity has been difficult or impossible using conven-
tional processing. Radial composites with four different connectivity patterns
have been fabricated using FDC and the SP processes. These consist of (1–3),
(2–2), (3–1), and (3–2) type connectivity as shown in Fig. 16.10 [37]. For the
purpose of simplicity, cylindrical coordinate systems are being used to desig-
nate the connectivity of each composite design. For the (1–3) type composite,
the ceramic phase is continuous in only the r direction, while the polymer
phase is continuous in the diameter (r), height (z), and sweep angle (θ) di-
rections (see Fig. 16.10). For (2–2) type composites, both the ceramic phase
and the polymer phases are continuous in the r and z directions. In the case
of (3–1) type composites, the ceramic phase is continuous in all three direc-
tions, while the polymer is continuous in only the z direction. Finally, for
the (3–2) type composites, the ceramic phase is continuous in all three di-
rections, while the polymer phase is continuous in the z and θ directions. In
this structure, the inner and outer ceramic rings are approximately 1 mm in
thickness, and the ceramic sheets connecting the inner and outer rings are
approximately 1 mm by 1 mm.
SFF has been used to fabricate novel actuators as well [38–46]. It has been
demonstrated that improved electromechanical properties can be achieved by
tailoring the geometry and dimensions of an actuator, which we call form
factor engineering. Processing and properties of a few rapidly prototyped ac-
tuators will be reviewed with emphasis on the recent results on unimorph and
multimaterial spirals and bending actuators.
390 A. Safari and E.K. Akdoğan
a)
b)
Fig. 16.10. Photographs of radial composites with (a) 3–1 and (b) 3–2 connectivity
1
Displacement (mm)
0.8
0.6
0.4
0.2
0
0 100 200 300 400 500 600 700 800
Voltage (V)
1.0
5:1
4:1 6:1
0.8
0.6
δ/L
3:1
0.4
1:1
0.2
1:3
Piezoelectric
Bar
0.0
0.0 0.2 0.4 0.6 0.8 1.0
tPiezo/ttotal
Fig. 16.13. Dependence of tip deflection (δ/L) per unit actuator length on
piezoelectric–electrostrictive layer ratio in PMN–PT monolithic bilayer bending
actuators
394 A. Safari and E.K. Akdoğan
P 2 ES 2
∆x1 = dP P P
31 E3 + Q12 E3 − QES
12 E3 , (16.1)
Deflection (h)
−∆x31
Electrostrictor E3
cantilever
Electrostrictor
cantilever
E3
Ps Piezoelectric Ps Piezoelectric
Deflection (h)
∆x31
(a) (b)
|xi| |xi|
Piezoelectric Effect
(Low field) ∝dE
Ps E
Electrostrictive ∝QE 2
Effect (Low field)
Ej Ej
0
(c) (d)
Fig. 16.14. Schematic showing the concept of asymmetric with respect to the po-
larity of the applied electric field in piezoelectric–electrostrictive bilayer bending
actuators
16 Piezoelectric Composites by Solid Freeform Fabrication 395
an unprecedented fashion [53]. The advantage of using SFF emerges here once
again, as such complex transducers can be prototyped within days, instead of
weeks and months by using ceramic extrusion technology.
The use of SFF in the development of novel piezoelectric composites was elab-
orated on; its effectiveness in developing composites with unmatched struc-
tural complexity was advocated. Through a series of examples, such as the
“ladder” and “3D Honeycomb” composites with (3–3) connectivity as well
as the oriented (2–2) and VFG (2–2) piezocomposites, it was shown that
the properties of composites can be “engineered” on the basis of symmetry
principles, thereby mimicking Mother Nature’s way of forming natural crys-
tals. In addition, it was demonstrated that composites with controlled phase
periodicity, volume fractions, and micro- and macro hierarchy could be ob-
tained in such structures, which are otherwise extremely difficult, if not im-
possible, to fabricate with traditional techniques. It was also shown that SFF
can be used to make various novel actuators.
The flexibility of SFF allows for the rapid fabrication of parts, whereby
one can go through a multitude of iterations and make frequent changes to the
dimensions of such actuators to accomplish full-scale optimization. Therefore,
SFF is ideally suited for what we have come to call “form factor engineering.”
As in the case of spiral actuators, the large tangential displacements (albeit
moderate blocking force) are a consequence of the form factor used, onto which
is superimposed the asymmetric deflection of the monolithic piezoelectric–
electrostrictive PMN–PT bilayer. SFF is also very useful in accomplishing
monolithic integration of various electroactive materials, which provides for
multifunctionality as well as better reliability than layered structures made
by the “glue-on” route of fabrication.
Earlier, we mentioned that SFF is a nature-inspired fabrication technique.
A quick glance to the mode by which crystals form would indicate that growth
takes place in an atomic-layer-by-atomic-layer progression. Of course, the scale
of such a growth mode is much finer, which results in the macroscopic smooth-
ness of crystal surfaces. Hence, one might visualize the vast range of material
properties as consequence of the interaction among such atomic layers in a
given crystal. Following this line of reasoning, SFF in principle should be able
to provide the means to create manmade macro- and meso-crystals made out
of different materials, enabling one to induce across the interfaces. Such an
approach seems to have great potential in obtaining cross-coupling of cou-
pling phenomena in a composite structure: e.g., ferroelectric–ferromagnetic
coupling, magentostrictive–piezoelectric coupling [54–56], among others. Such
novel designs based on the intelligent use of materials would surely supercede
the traditional composite properties, since one can possibly induce interac-
tions that are otherwise absent.
16 Piezoelectric Composites by Solid Freeform Fabrication 397
Acknowledgement
The authors are thankful to the Office of Naval Research (ONR), the De-
fense Advanced Research Projects Agency (DARPA), New Jersey Commis-
sion on Science and Technology (NJCTS), and the Glenn Howatt Foundation
at Rutgers University for the generous funding provided. The authors thank
past graduate students and colleagues A. Bandhophadhyay, R.K. Panda,
T.F. McNulty, S. Turcu, G. Lous, F. Mohammadi, B. Jadidian, A. Hall and
M. Allahverdi.
References
1. K. Uchino, Piezoelectric Actuators and Unltrasonic Motors (Kluwer Academic,
New York, 1996)
2. X. Yu, Ferroelectric Materials and Their Applications (North Holland, Amster-
dam, 1991)
3. E.K. Akdogan, M. Allahverdi, A. Safari, IEEE Trans. Ultrason. Ferr. 52(5),
746–775 (2005)
4. D.P. Skinner, R.E. Newnham, Mater. Res. Bull. 13, 599–607 (1978)
5. S.M. Pilgrim, R.E. Newnham, L.L. Rohlfing, Mater. Res. Bull. 13, 677–
684 (1987)
6. R.K. Panda, Ph.D. Thesis, Rutgers University, New Brunswick, NJ, 1998
7. A. Safari, Ph.D. Thesis, The Pennsylvania State University, State College,
PA, 1983
8. W. Schulze, Ferroelectrics 50, 33–37 (1983)
9. T.R. Shrout, L.J. Bowen, W.A. Schulze, Mater. Res. Bull. 15, 1371–1379 (1980)
10. A. Safari, Mater. Res. Bull. 17, 301–304 (1982)
11. J.W. Halloran, Br. Ceram. Trans. 98, 299–303 (1999)
12. H.L. Marcus, J.J. Beaman, J.W. Barlow, D.L. Bourell, R.H. Crawford, Rapid
Prototyping Using FDM: A Fast, Precise, Safe Technology. Paper presented at
Solid Freeform Fabrication Symposium, Austin, TX, 1992
398 A. Safari and E.K. Akdoğan
13. H.L. Marcus, J.J. Beamen, J.W. Barlow, D.L. Bourell, R.H. Crawford, in Rapid
Prototyping Using FDM: A Fast, Precise, Safe Technology, Solid Freeform Fab-
rication Proceedings, Austin, TX, 1992
14. H.L. Marcus, J.J. Beamen, J.W. Barlow, D.L. Bourell, R.H. Crawford, in Fast,
Precise, Safe Prototypes With FDM, Solid Freeform Fabrication Proceedings,
University of Texas at Austin, Austin, TX, 1991
15. H.L. Marcus, D.L. Bourell, Adv. Mater. Proc. 9, 677–684 (1993)
16. H.L. Marcus, Mech. Eng. 117, 62–66 (1995)
17. D.L. Bourell, J.J. Beamen, H.L. Marcus, J.W. Barlow, Solid Freeform Fabrica-
tion, An Advanced Manufacturing Approach. Paper presented at Solid Freeform
Fabrication Proceedings, Austin, TX, 1990
18. P.F. Jacobs, Stereolithography and Other RP & M Technologies from Rapid
Prototyping to Rapid Tooling (Society of Manufacturing Engineers, Dearborn,
MI, 1995)
19. P.F. Jacobs, Rapid Prototyping and Manufacturing: Fundamentals of Stereo
Lithography (Society of Manufacturing Engineers, Dearborn, MI, 1992)
20. M.A. Jafari, W. Han, F. Mohammadi, A. Safari, S.C. Danforth, N. Langrana,
J. Rapid Prototyping 6(3), 161–175 (2000)
21. T.F. McNulty, F. Mohammadi, A. Bandyopadhyay, D.J. Shanefield,
S.C. Danforth, A. Safari, Development of a Binder Formulation for Fused De-
position of Ceramics, 4(4), 144–150 (2001)
22. M. Feygin, B. Hsieh, in Proceedings of the Solid Freeform Fabrication Sympo-
sium, 1991, pp. 123–125
23. M. Feygin, U.S. Patent No. 5354414, 1994
24. J. Cesarano, T.A. Baer, P. Calvert, in Proceedings of the Solid Freeform Fabri-
cation Symposium, vol. 8, 1997, 25–27
25. J. Cesarano, B.H. King, H.B. Denham, in Proceedings of Solid Freeform Fabri-
cation Symposium, vol. 9, 1998, 697–699
26. J. Cesarano, P. Calvert, U.S. Patent No. 6027326, 2000
27. E.M. Sachs, M.J. Cima, P. Williams, D. Brancazio, J. Cornie, J. Eng. Ind. 114,
481–485 (1992)
28. A. Bandyopadhyay, R.K. Panda, V.F. Janas, M.K. Agarwala, S.C. Danforth,
A. Safari, J. Am. Ceram. Soc. 80(6), 1366–1371 (1997)
29. A. Safari, S.C. Danforth, R.K. Panda, B. Jadidian, T.F. McNulty, G. Lous, in
Proceedings of 8th U.S.-Japan Conference on Composite Materials, Baltimore,
MD, 1998, pp. 526–535
30. S. Turcu, M.S. Thesis, Rutgers University, New Brunswick, NJ, 2002
31. S. Turcu, B. Jadidian, S.C. Danforth, A. Safari, J. Electroceram. 9, 165–
171 (2002)
32. C.W. Nan, L. Liu, L. Li, J. Phys. D 33, 2988–2991 (2000)
33. C.W. Nan, G.W. Weng, J. Appl. Phys. 88, 416–420 (2000)
34. A. Bandyopadhyay, R.K. Panda, V.E. Janas, M.K. Agarwala, S.C. Danforth,
A. Safari, J. Am. Ceram. Soc. 80, 1366–1372 (1997)
35. A. Safari, V.F. Janas, R.K. Panda, Fabrication of Fine Scale 1–3 PZT Ceramic/
Polymer Composites Using a Modified Lost Mold Method. Paper presented at
SPIE Symposium on Smart Structures and Materials, San Diego, CA (1996)
36. A. Safari, J. Cesarano, P.G. Clem, B. Bender, Fabrication of Advanced Func-
tional Electroceramic Components by Layered Manufacturing (LM) Methods.
Paper presented at Proceedings of the 13th IEEE International Symposium on
Applications of Ferroelectronics, Nara, Japan, 2002
16 Piezoelectric Composites by Solid Freeform Fabrication 399
37. T.F. McNulty, Ph.D. Thesis, Rutgers University, Piscstaway, New Jersey, USA,
1999
38. F. Mohammadi, A. Kholkin, B. Jadidian, A. Safari, Appl. Phys. Lett. 75,
2488 (1999)
39. Q.M. Wang, Q. Zhang, B. Xu, R. Liu, L.E. Cross, J. Appl. Phys. 86, 3352 (1999)
40. A. Safari, S.C. Danforth, M.A. Jafari, M. Allahverdi, B. Jadidian,
F. Mohammadi, N. Vankataraman, S. Rangarajan, in Proceedings of the 9th Eu-
ropean Conference on Rapid Prototyping and Manufacturing, 2000, pp. 247–249
41. A. Safari, S.C. Danforth, in Proceedings of the 11th IEEE International Sympo-
sium on the Applications of Ferroelectrics, 1998, pp. 229–231
42. A. Bandyopadhyay, R.K. Panda, T.F. McNulty, F. Mohammadi, S.C. Danforth,
A. Safari, Rapid Prototyping J. 4, 37–42 (1998)
43. A. Safari, M. Allahverdi, Ceram. Eng. Sci. Proc. 22, 473–475 (2001)
44. F. Mohammadi, Ph.D. Thesis, Department of Ceramic and Materials Engineer-
ing, Rutgers University, New Jersey, USA, 2001
45. F. Mohammadi, A. Kholkin, S.C. Danforth, A. Safari, in Proceedings of the
11th IEEE International Symposium on the Applications of Ferroelectrics, 1998,
pp. 273–275
46. S.A. Wise, Sens. Actuators A69, 33–37 (1998)
47. Q.M. Zhang, H. Wang, L.E. Cross, J. Mater. Sci. 28, 3962–3965 (1993)
48. Y. Sugawara, K. Onitsuka, S. Yoshikawa, Q.C. Xu, R.E. Newnham, K. Uchino,
J. Am. Ceram. Soc. 75, 996–998 (1992)
49. A. Hall, Ph.D. Thesis, Rutgers University, Piscataway, New Jersey, USA, 2006
50. A. Hall, E.K. Akdogan, M. Allahverdi, A. Safari, J. Eur. Ceram. Soc. 25, 2991–
2997 (2005)
51. A. Hall, E.K. Akdogan, M. Allahverdi, A. Safari, J. Electroceram. 15(2), 143–
150 (2005)
52. E.K. Akdogan, A. Hall, A. Safari (unpublished)
53. R.E. Newnham, D.C. Markley, R.J. Meyer Jr., W.J. Hughes, A.-C. Hladky-
Hennion, J.K. Cochran Jr., Ceram. Trans. 150, 427–443 (2004)
54. S. Dong, J.-F. Li, D. Viehland, IEEE Trans. Ultrason. Ferr. 50(10), 1253–
1261 (2003)
55. S. Dong, K. Uchino, L. Li, D. Viehland, IEEE Trans. Ultrason. Ferr. 54(6),
1240–1249 (2007)
56. S. Dong, J.-F. Li, D. Viehland, IEEE Trans. Ultrason. Ferr. 54(6), 1240–
1249 (2007)
17
Microstructural Analysis Based on Microscopy
and X-Ray Diffraction
17.1.1.1 BaTiO3
17.1.1.2 KNN
1500
1400 1400
Liquid
Temperature, C (Note change of scale at high temp.)
1300
1200 + Soild
Liquid Pc
700
1000 Soild
600
PTM
500 Pc 500
2
P TM
3
P TM
400 POM 400
FT1 A0
300 FT2 FTM 300
200 200
A0 + FOM
FOM2 FOM1
100 FO1 FO2 100
Rising
temp.
0 Fallin Temp 0
g tem . hyst
p. eresis
–100 FR FMONO –100
FMONO
0 10 20 30 40 50 60 70 80 90 92 94 96 98 100
KNbO3 Mole % NaNbO3 (Note change of scale at 90 %) NaNbO3
17.1.1.3 PZT
For the solid solution system PbZrO3 −PbTiO3 the Curie temperature ranges
from 495◦ C for lead titanate to 250◦ C for lead zirconate. The crystal structure
for all compositions above the Curie temperature is cubic with space group
P m3m. Below Tc , the materials with high Ti content are tetragonal with
the space group P 4mm (Fig. 17.2 left side). Materials with very high Zr con-
tent are transformed to an anti-ferroelectric orthorhombic structure (Pbam)
when cooling. A large part of the Zr-rich compositions, however, adopts a
rhombohedral structure below Tc (Fig. 17.2 right side). Space groups of the
rhombohedral materials are R3c and R3m depending on the composition. The
compositional range with Zr contents between 50 and 60 mol% Zr is denoted
as the MPB. The structural feature of these materials has been considered for
many years as the coexistence of the tetragonal and rhombohedral phases [2].
However, the nature of the crystallographic structure and phase composition
of these materials have been the subjects of ongoing scientific discussion.
Approaches to estimate lattice constants can be based on {200} reflections
within a pseudocubic setting. For the tetragonal phase materials, the reflec-
tions split into the (002) and (200) reflections. The lattice distortion, given in
terms of δt = (c/a) − 1 = (d002 /d200 ) − 1 (Fig. 17.2 right side), ranges from 6%
for lead titanate to about 2% for tetragonal compositions close to the MPB.
According to Randall et al. [54], lattice parameters are grain-size dependent,
showing a decreasing lattice distortion δt with lower grain size. With increasing
temperature, the c/a ratio diminishes by increasing a and decreasing c lattice
spacings, thus approaching the cubic state. Rhombohedral lattice parameters
can also be obtained from {200} reflections. Information on the rhombohedral
lattice distortion are derived from the (111) and (−111) reflections and can
be quantified in terms of δr = (9/8)(d111 /d−111 −1) (Fig. 17.2 right side). The
lattice distortion for the rhombohedral phase of approximately 0.7% is only
slightly sensitive to composition and corresponds to a small angular distortion
Fig. 17.2. Crystal structure and lattice distortion for the tetragonal (left) and the
rhombohedral (right) phase of the PZT solid solution system
17 Microstructural Analysis Based on Microscopy and X-Ray Diffraction 407
from cubic symmetry of 89.7◦ . For doped PZT (<5 mole%), there are only mi-
nor shifts with respect to the phase relations. Qualitatively similar dependence
of phases and structures from composition prevail also for many Pb(B1,B2)–
PT relaxor ferroelectric solid solutions.
Fig. 17.3. Asymmetry of (002) and (200) reflection intensity between peaks in
tetragonal PZT 50/50 1La2Sr
408 M.J. Hoffmann et al.
(a)
Polarization,
(b)
Strain 2
0.0
Polarization
A=1.25
1.5
–5 –4 –3 –2 –1 0 1 2 3 4 5
x / x180
(I200,t + I002,t )
χt = .
(I200,t + I002,r + I002,t )
Fig. 17.5. Overlapping of the tetragonal (t) and rhombohedral (r) peaks in mor-
photropic PZT
410 M.J. Hoffmann et al.
4.150
4.125 c
4.075
4.050
4.025 at
4.000
3.975
3.950
35 40 45 50 55 60 65 70
mol% Zr
Fig. 17.6. Lattice constants in single phase and in morphotropic PZT 1La2Sr
1.250
HWB001,t’ HWB100,r’ HWB100,t [8]
1.125
1.000
0.875
0.750
0.625
0.500 (200)r
0.375 (002)t
0.250
0.125
(200)t
0.000
35 40 45 50 55 60 65 70
mol% Zr
Fig. 17.7. Full width half-maximum (FWHM) in single phase and in morphotropic
PZT 1La2Sr
been suggested for t- and r-phase material. Therefore, the monoclinic phase
was considered as a structural bridge providing polarization rotation. Similar
patterns were found in the MPB compositions of several relaxor ferroelectrics.
In PZN–PT the monoclinic distortion and attribution to the monoclinic space
group depend on the strain state [23, 24]. Differences were found between the
surface and the inner parts of the material. From further investigations on
undoped PZT 52/48 it turned out that the materials on the tetragonal side
close to the MPB exhibited an average structure that appeared monoclinic
in the diffraction patterns at room temperature [12]. Locally distorted crys-
talline regions in a tetragonal matrix structure were suggested to be the origin
of the monoclinic reflections [25, 26]. The origin of the locally distorted re-
gions is still under discussion. Several approaches have been presented. Local
17 Microstructural Analysis Based on Microscopy and X-Ray Diffraction 411
Along with electric fields, two types of microstructural mechanisms are acti-
vated in ferroelectric ceramics. Strain is provided by the piezoelectric effect
and domain switching. Both effects are coupled to elastic strains by stress–
strain interactions of neighboring grains [38, 50]. In a similar way, mechanical
pressure results in an elastic strain and strain by ferroelastic domain switching,
both of them also involve stress–strain interaction. The microscopic features of
the strain mechanisms are changes in lattice spacings for the piezoelectric ef-
fect and elastic strain, whereas ferroelectric and ferroelastic domain switching
412 M.J. Hoffmann et al.
no field
2 kV/mm
intensity
2 theta
changes the orientation of the polarization axis. The effects on the diffrac-
tion patterns are as follows: (1) Piezoelectric and elastic strain results in a
shift of the peak positions; (2) Strain from 90◦ , 71◦ and 109◦ domain switch-
ing can be analyzed from changes in the intensities of particular diffraction
peaks (Fig. 17.8). XRD therefore provides an option for the analysis of the mi-
crostructural mechanisms of strain in ferroelectrics during poling, with electric
fields and mechanical loads. An experimental setup for an in situ experiment
with electric field is sketched in Fig. 17.9. An experimental precondition when
measuring in situ with an electric field or mechanical loads in reflection is
that the sample is focused for the complete experiment. Parallel beam setup
and suitable detectors as available in synchrotron beamlines are required for
detection of the peak shift [32, 33]. Careful choice of electrode materials and
balancing thickness of electrode layers are necessary. On the one hand ab-
sorption from the electrode material and generation of additional peaks has
to be minimized; on the other, sufficient conductivity is required in order to
17 Microstructural Analysis Based on Microscopy and X-Ray Diffraction 413
0.30 4.5
0.20 4.0
3.5
0.10
strain [%]
I002/I200
3.0
0.00
2.5
–0.10
2.0
–0.20 1.5
–0.30 1.0
–3 –2 –1 0 1 2 3 –3 –2 –1 0 1 2 3
electric field [kV/mm] electric field [kV/mm]
Fig. 17.10. Macroscopic strain and intensity ratio I002,t /I200,t during bipolar elec-
trical cycling of donor-doped morphotropic PZT
17.2 Microscopy
17.2.1 SEM
Microstructures of PZT, BaTiO3 , or KNN in terms of grain size, grain size dis-
tribution, porosity, and the presence of secondary phases can be analyzed us-
ing scanning electron microscopy (SEM). Furthermore, the type of secondary
phases can be analyzed if the SEM is equipped with an energy dispersed X-ray
(EDX) detector. The SEM technique requires polished and etched samples
and, because the samples are insulators, sputtering with a thin conductive
layer such as Au or Au/Pd is necessary. There exist two types of etching
processes. The first one is a thermal etching at temperatures 100–150◦ C be-
low the sintering temperature. Here the grain boundaries are mostly affected.
More information is obtained by chemical etching at room temperature with
a solution consisting of 95 ml distilled H2 O, 5 ml 0.1 molar HCl, and five drops
of HF. Hereby the etching time depends strongly on the composition varying
between 5 s and 4 min. Chemical etching affects not only the grain boundaries
but also 90◦ domain walls can be visualized, because the etching attack of
the domains depends on the polarization direction. On the contrary, 180◦ do-
mains provide no contrast. Typical domain structures are lamellar patterns
for tetragonal and wedge-shaped structures for rhombohedral PZT as shown
in Fig. 17.11.
17 Microstructural Analysis Based on Microscopy and X-Ray Diffraction 415
Fig. 17.11. Domain structures of tetragonal (left) and of rhombohedral (right) PZT
as observed by SEM
17.2.2 TEM
Imaging of the domain structure by TEM bright field shows lamellar domains
for the tetragonal ferroelectric materials (Fig. 17.12 left side). The domain
configurations of rhombohedral PZT are described as domain bands, wedge
shaped, or herringbone structures [51] (Fig. 17.12 right side). Owing to many
variable operating modes, high resolution, and the combination of imaging
with electron diffraction, additional details and characteristics can be an-
alyzed and crystallographic orientations can be attributed to the domains.
The correct representation of the microstructures in bulk materials by the
microstructures observed in the thin specimens is, however, a point of ongo-
ing discussion. Care has to be taken with respect to preparation of the samples
416 M.J. Hoffmann et al.
Fig. 17.12. The lamellar domain structure of a tetragonal PZT sample with the
composition Pb(Zr0.52 Ti0.48 )O3 observed in the [010] direction (left) [52], and the
herringbone structure of a rhombohedral PZT sample (right) with the composition
Pb(Zr0.60 Ti0.40 )O3 observed along a [−111] direction (Courtesy R. Schierholz). The
CBED pattern (bottom right) shows a mirror plane, but no 3-fold symmetry as
expected for the [111] direction. The defocused CBED pattern (top right) gives a
bright-field image (central spot) of the observed area, each diffraction spot containing
the corresponding dark-field image. All domain walls which are on-edge (are within
the crystallographic zone) can be attributed directly to the corresponding reflection
Domain Size
Stacking Configuration
Varying grain size not only affects the domain width but also the stack-
ing configuration of domains. As predicted from an energy minimizing ap-
proach considering elastic strain and interface domain wall energy [55], the
domain configuration changes from a three-dimensional pattern to a two-
dimensional structure with decreasing grain size. Reduction in the number
of domain variants was found for grain sizes <1 µm in PZT 52/48 by Cao and
Randall [56]. The critical grain size is specific for each material depending on
lattice distortion, domain wall energy, and elastic modulus.
Nanodomains
Grain Boundaries
HRTEM investigation of the local ordering in mixed crystal systems has be-
come possible with the development of exit wave reconstruction software, but
is often limited by practical aspects. The technique reconstructs the exit wave
function on the basis of a series of images with varying defocus if the kine-
matic approximation is satisfied. In the case of lead zirconate titanate, the
heavy lead atoms restrict the sample thickness to approximately two elemen-
tary cells along the beam direction. Other systems, e.g., KNaNbO3 , are more
promising candidates for this kind of investigation. The imaging of domain
walls is principally limited by the same restrictions. But still a number of
questions can be answered using even thicker samples. The tilt of adjacent
domains can be measured and correlated to the lattice parameters, the width
of domain walls (the strain field) can be estimated, and the morphology of
the domain wall can be seen (Fig. 17.13).
17 Microstructural Analysis Based on Microscopy and X-Ray Diffraction 419
References
1. P.W. Forsbergh, Phys. Rev. 76, 1187–1201 (1949)
2. B. Jaffee, W.R. Cook, H. Jaffee, Piezoelectric Ceramics (R.A.N. Publishers,
Marietta, OH, 1971)
3. M.T. Buscaglia, V. Buscaglia, M. Vivani, J. Petzelt, M. Savinov, L. Mitioseriu,
A. Testino, P. Nanni, C. Harnagea, Z. Zhao, M. Nygren, Nanotechnology 15,
1113–1117 (2004)
4. Y. Saito, H. Takao, T. Tani, T. Nonoyama, K. Takatori, T. Homma, T. Nagaya,
M. Nakamura, Nature 432, 84–87 (2004)
5. G. Zorn, W. Wersing, H. Goebel, Jpn. J. Appl. Phys. 24(Suppl.2), 721–
723 (1985)
6. H. Boysen, Zeitschrift für Kristallographie 20, 726–734 (2005)
7. W. Cao, L.E. Cross, Phys. Rev. B 44, 5–12 (1991)
8. D.A. Bruce, J. Phys. C Solid State Phys. 14, 5195–5214 (1981)
9. H. Boysen, Phase Transit. 55, 1–16 (1995)
10. N. Floquet, C.M. Valot, M.T. Mesnier, J.C. Niepce, L. Normand, A. Thorel,
R. Kilaas, J. Phys. III 7, 1105–1128 (1997)
11. C.M. Valot, N. Floquet, P. Perriat, M.T. Mesnier, J.C. Niepce, Ferroelectrics
172, 235–241 (1995)
12. J. Frantti, J. Lappalainen, S. Eriksson, V. Lantto, S. Nishio, M. Kakihana,
S. Ivanov, H. Rundlöf, Jpn. J. Appl. Phys. 39, 5697–5703 (2002)
13. D.J. Corker, A.M. Glazer, R.W. Whatmore, A. Stallard, F. Fauth, J. Phys.
Condens. Matter 10, 6251–6269 (1998)
14. P.W. Stevens, J. Appl. Crystallogr. 27, 462 (1999)
420 M.J. Hoffmann et al.
15. L. Hanh, K. Uchino, S. Nomura, Jpn. J. Appl. Phys. 39, 5697–5703 (1978)
16. W. Cao, L.E. Cross, Jpn. J. Appl. Phys. 31, 1399–1402 (1992)
17. W. Wersing, W. Rossner, G. Eckstein, G. Tomandl, Silicates Industriels 3–4,
41–46 (1985)
18. M.J. Hoffmann, H. Kungl, Curr. Opin. Solid State Mater. Sci. 8, 51–57 (2004)
19. K.H. Härdtl, D. Hennings, Science of Ceramics 6, VII/1 (1973)
20. B. Noheda, D.E. Cox, G. Shirane, R. Guo, B. Jones, E. Cross, Phys. Rev. B 63,
14103-1–14103-9 (2000)
21. B. Noheda, J.A. Gonzalo, E. Cross, R. Guo, S.E. Park, D.E. Cox, G. Shirane,
Phys. Rev. B 61, 8687–8695 (2000)
22. R. Guo, E. Cross, S.E. Park, B. Noheda, D.E. Cox, G. Shirane, Phys. Rev. Lett.
84, 5423–5426 (2000)
23. K. Ohwada, K. Hirota, P.W. Rehrig, P. Gehring, B. Noheda, Y. Fujii, S.E. Park,
G. Shirane, J. Phys. Soc. Jpn. 70, 2778–2783 (2001)
24. M.K. Durbin, C.J. Hicks, S.E. Park, T.R. Shrout, J. Appl. Phys. 87, 8159–8164
(2000)
25. J. Frantti, S. Ivanov, J. Lappalainen, S. Eriksson, V. Lantto, S. Nishio,
M. Kakihana, H. Rundlöf, Ferroelectrics 266, 71–90 (2002)
26. J. Frantti, S. Ivanov, S. Eriksson, H. Rundlöf, V. Lantto, J. Lappalainen,
M. Kakihana, Phys. Rev. B 66, 64108-1–64108-15 (2002)
27. D.I. Woodward, J. Knudsen, I.M. Reaney, Phys. Rev. B 72, 104110-1–104110-
8 (2005)
28. K.A. Schönau, L. Schmitt, M. Knapp, H. Fuess, R. Eichel, H. Kungl, M.J.
Hoffmann, Phys. Rev. B 75, 184117 (2007)
29. Y.M. Jin, Y.U. Wang, A.G. Khachaturian, J.F. Li, D. Viehland, Phys. Rev.
Lett. 91, 197601-1–197601-4 (2003)
30. Y.M. Jin, Y.U. Wang, A.G. Khachaturian, J.F. Li, D. Viehland, J. Appl. Phys.
94, 3629–3640 (2003)
31. M. Hammer, M.J. Hoffmann, J. Am. Ceram. Soc. 81, 3277–3284 (1998)
32. M. Hammer, C. Monty, A. Endriss, M.J. Hoffmann, J. Am. Ceram. Soc. 81,
721–724 (1998)
33. M. Knapp, C. Baehtz, H. Ehrenberg, H. Fuess, J. Synchrotron Radiat. 11, 328–
334 (2004)
34. Y. Saito, Jpn. J. Appl. Phys. 36, 5963–5969 (1997)
35. T. Ogawa, Jpn. J. Appl. Phys. 39, 5538–5541 (2000)
36. M.J. Hoffmann, M. Hammer, A. Endriss, D.C. Lupascu, Acta Materialia 49,
1301–1310 (2001)
37. M.J. Hoffmann, H. Kungl, J.-Th. Reszat, S. Wagner, Polar Oxides – Properties,
Characterization and Imaging, ed. by R. Waser et. al., Wiley (2005)
38. J.T. Reszat, A.E. Glazounov, M.J. Hoffmann, J. Eur. Ceram. Soc. 21, 1349–
1352 (2001)
39. D.A. Hall, A. Steuwer, B. Cherdirunkorn, T. Mori, P.J. Withers, J. Appl. Phys.
96, 4245 (2004)
40. D.A. Hall, A. Steuwer, B. Cherdirunkorn, P.J. Withers, T. Mori, Mater. Sci.
Eng. 409, 206–210 (2005)
41. D.A. Hall, A. Steuwer, B. Cherdirunkorn, T. Mori, P.J. Withers, Acta Materialia
54, 3075–3083 (2005)
42. S. Wan, K. Bowman, J. Mater. Res. 16, 2306–2313 (2001)
43. J.L. Jones, E.B. Slamovic, K. Bowman, J. Appl. Phys. 97, 034113-1–034113-
6 (2005)
17 Microstructural Analysis Based on Microscopy and X-Ray Diffraction 421
44. W. Chang, A.H. King, K.J. Bowman, Appl. Phys. Lett. 88, 242901 (2006)
45. R. Rogan, E. Üstündag, B. Clausen, M.R. Daymond, J. Appl. Phys. 93, 4104–
4111 (2003)
46. J.Y. Li, R. Rogan, E. Üstündag, K. Bhattacharya, Nat. Mater. 4, 776–781 (2005)
47. W. Wersing, K. Lubitz, J. Mohaupt, IEEE Trans. Ultrason. Piezoelectr. Freq.
Control 36, 424–433 (1989)
48. J.L. Jones, M. Hoffmann, J.E. Daniels, A.J. Studer, J. Appl. Phys. 98, 092901-
1–092901-3 (2006)
49. H. Kungl, Dehnungsverhalten von morphotropem PZT Dissertation, Universität
Karlsruhe (2005)
50. H. Kungl, R. Theissmann, C. Baehtz, M. Knapp, H. Fuess, S. Wagner, T. Fett,
M.J. Hoffmann, Acta Materialia 55, 5780–5791 (2007)
51. J. Ricote, R.W. Whatmore, D.J. Barber, J. Phys. Condens. Matter 12, 323–
337 (2000)
52. R. Theissmann, L.A. Schmitt, J. Kling, R. Schierholz, K.A. Schönau, H. Fuess,
M. Knapp, H. Kungl, M.J. Hoffmann, J. Appl. Phys. 102, 024111 (2007)
53. L. Schmitt, K.A. Schönau, R. Theissmann, H. Fuess, H. Kungl, M.J. Hoffmann,
J. Appl. Phys. 101, 074107 (2007)
54. C.A. Randall, N. Kim, J. Kucera, W. Cao, T.R. Shrout, J. Am. Ceram. Soc.
81, 677–688 (1998)
55. G. Arlt, J. Mater. Sci. 25, 2655–2666 (1990)
56. W. Cao, C.A. Randall, J. Phys. Chem. Solids 57, 1499–1505 (1996)
57. I. Mac Laren, L. Schmitt, H. Fuess, H. Kungl, M.J. Hoffmann, J. Appl. Phys.
97, 094102-1–094102-8 (2004)
58. N.A. Pertsev, G. Arlt, Ferroelectrics 132, 27 (1992)
59. P.G. Lucuta, V. Teodorescu, Appl. Phys. A: Solid Surf. 37, 237 (1985)
60. X. Tan, J.K. Shang, Philos. Mag. A 82, 1463–1478 (2002)
61. A. Glazer, P.A. Thomas, K.Z. Baba-Kishi, G.K.K. Pang, C.W. Tai, Phys. Rev.
B 70, 184123-1–184123-9 (2004)
18
Small Signal Resonance Methods
W. Wersing
18.1 Introduction
The properties of a piezoelectric material are defined by the elastic, dielectric,
and piezoelectric tensor components which will be introduced in Sect. 18.2.
Because of the inherent asymmetrical nature of piezoelectric materials and
the fact that we want to predict different response directions, a tensor-based
description of the properties of these materials is unavoidable. Generally these
components are by far not constant, they depend on temperature, applied me-
chanical stress and electric field strength. Furthermore, they are amplitude-
dependent and become non-linear or even non-reversible when the applied
signal stress or signal field amplitudes exceed limits characteristic of the ma-
terial. The dielectric, elastic and piezoelectric coefficients have not only effec-
tive (real, in-phase) components but also dissipative (imaginary, out-of-phase)
terms which are of particular importance at high amplitudes.
The fact that we have to deal with the directionality of the response of
piezoelectric materials and consequently with tensors complicates engineering
to some extent, however, on the other hand, it offers a great design potential
for engineers in the development of piezoelectric devices. Because, today, the
design of new devices is done more or less by using computational tools, it
is decisive for the engineer to have access to a complete set of the property
coefficients (tensor components) of the materials applied.
Therefore, in this subchapter we will describe methods to precisely deter-
mine dielectric, elastic, and piezoelectric coefficients under small signal condi-
tions. The determination of these properties under large signal conditions is
discussed in Chap. 19. Because we consider small signal conditions, we need
not take into account the non-linearities in the constants and the equations
of state relating the electric and elastic variables can be written in the lin-
ear form derived in Sect. 3.1 from the thermodynamic potentials. Under small
signal conditions the contributions to the piezoelectric response in ferroelec-
tric materials are due to the intrinsic and the reversible extrinsic piezoelectric
effect caused by the reversible movement of domain walls.
424 W. Wersing
In Sect. 3.1, we derived from the Gibbs free energy the equations (3.11):
E
∆Sα = αα ∆T + sEαβ ∆Tβ + diα ∆Ei ,
(18.1)1
∆Di = pT
i ∆T + d T
iα ∆Tα + εik ∆Ek ,
E
where αα are the thermal expansion coefficients, sE αβ the elastic compliances
at constant electric field, diα the piezoelectric constants, pT i the pyroelectric
coefficients, and εT ik the permittivities at constant stress. In addition, six
other linear equations, i.e., three further sets of equations can be derived
by use of the other thermodynamic potentials. The interrelationships among
the mechanical, electrical, and thermal properties mediated by the different
material constants can be well illustrated using a construct called a Heckmann
diagram [4] shown in Fig. 18.1. The circles of the outer triangle represent
the intensive variables (forces), the circles of the inner triangle represent the
extensive variables (displacements).
In their isothermal form (∆T = 0) these four sets of equations are called
basic piezoelectric equations or piezoelectric constitutive equations:
Tα = cD
αβ Sβ + hiα Di ,
(18.2)
Ei = −hiβ Sβ + βik
S
Dk ,
Sα = sD
αβ Tβ + giα Di ,
(18.3)
Ei = −giβ Tβ + βik
T
Dk ,
αβ Sβ − eiα Ei ,
Tα = cE
(18.4)
Di = eiβ Sβ + εSik Ek ,
Sα = sE
αβ Tβ + diα Ei ,
(18.5)
Di = diβ Tβ + εT
ik Ek .
1
Notations are according to the IRE Standards [2, 3].
2
i, k = 1, 2, 3 and α, β = 1, 2, . . . , 6.
18 Small Signal Resonance Methods 425
EK
Permittivity
εik
Piezoelectricity Pyroelectricity
Piezoelectric effect Di Electrocaloric effect
dkα pk –πk
(converse) –ekβ Heat of polarization
Piezoelectric effect Pyroelectric effect
(direct) diβ pi
eiα πi
Heat capacity
τα
Elasticity sαβ Sα S ρ⋅c/T
αβ –τβ αα
Tβ T
Heat of Thermal
deformation expansion
Piezocaloric effect Thermoelasticity
Fig. 18.1. Heckmann diagram showing the interrelationship among the mechani-
cal, electrical, and thermal properties of materials mediated by the different material
constants. The circles of the outer triangle represent the intensive variables (tem-
perature T , stress components Tβ , and electric field components Ek ), the circles of
the inner triangle represent the extensive variables (entropy S, strain components
Sα , and displacement components Di )
Sij = sE
ijkl Tkl + dkij Ek ,
(18.7)
Di = dikl Tkl + εTik Ek ,
where all variables and material properties are described by tensors of first
(Di , Ek ) to fourth rank (sijkl ) and again the summation convention applies.
The property matrices used in (18.2)–(18.6) have the following meaning:
cE and cD are the closed-circuit (constant E) and open-circuit (constant
D) elastic stiffnesses,
sE and sD are the closed-circuit (constant E) and open-circuit (constant
D) elastic compliances, εT and εS are the free (constant stress) and clamped
(constant strain) permittivities, β T and β S are the free (constant stress) and
clamped (constant strain) impermeabilities.
426 W. Wersing
d = esE = εT g
e = dcE = εS h
(18.9)
g = hsD = β T d
h = gcD = β S e
sE − sD = dt g = g t d
(18.10)
cD − cE = et h = ht e
εT − εS = edt = det
(18.11)
β S − β T = hg t = ght
Crystal symmetry imposes restrictions on the form of the property tensors
and consequently on the property matrices. In many cases the full pattern of
symmetries imposed by each of the 32 point groups has been worked out [5]
(see Table 8 of [6]). We show here in (18.12) as an example the detailed form
of the matrices of (18.5) of poled ceramics3 (class ∞mm) identical to those of
crystals belonging to the hexagonal class 6 mm.
⎛ ⎞ ⎛ E E E ⎞⎛ ⎞ ⎛ ⎞
S1 s11 s12 s13 0 0 0 T1 0 0 d31
⎜ S2 ⎟ ⎜ sE sE E
11 s13 0 0 0 ⎟⎜ T2 ⎟ ⎜ 0 0 d31 ⎟⎛ ⎞
⎜ ⎟ ⎜ 12 ⎟⎜ ⎟ ⎜ ⎟
⎜ S3 ⎟ ⎜ s13 s13 sE
E E
0 0 0 ⎟⎜ T3 ⎟ ⎜ 0 0 d33 ⎟ E1
⎜ ⎟=⎜ 33 ⎟⎜ ⎟ ⎜ ⎟⎝ ⎠
⎜ S4 ⎟ ⎜ 0 0 0 sE 44 0 0 ⎟⎜ T4 ⎟+⎜ 0 d15 0 ⎟ E2
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ S5 ⎠ ⎝ 0 0 0 0 sE 44 0 ⎠⎝ T5 ⎠ ⎝ d15 0 0 ⎠ E3
S6 0 0 0 0 0 2 sE − sE
11 ⎛ 12⎞ T6 0 0 0
T1
⎛ ⎞ ⎛ ⎞ ⎜ T2 ⎟ ⎛ T ⎞⎛ ⎞
D1 0 0 0 0 d15 0 ⎜ ⎟
⎜ T3 ⎟ ε11 0 0 E1
⎝ D2 ⎠ = ⎝ 0 0 0 d15 0 0 ⎠ ⎜ ⎟ + ⎝ 0 εT11 0 ⎠ ⎝ E2 ⎠
⎜ T4 ⎟
D3 d31 d31 d33 0 0 0 ⎜ ⎟
⎝ T5 ⎠ 0 0 εT33 E3
T6
(18.12)
3
The 3-axis is usually chosen to be the polar axis (parallel to the poling direction).
18 Small Signal Resonance Methods 427
another property set is often used, especially in the case where small signal
resonance methods are applied for characterizing material properties. This
property set are the electro mechanical coupling factors klm which describe
the strength of the electromechanical response. They are dimensionless mea-
sures of the strength of the piezoelectric effects and therefore, are very useful
as a figure of merit. The coupling factors may also be considered as a measure
of the efficiency of the conversation of electrical to mechanical or mechani-
cal to electrical energy, i.e., for example when an electrical field is applied,
k 2 measures the fraction of the electrical energy converted into mechanical
energy. Therefore, in terms of k 2 we have:
electrical energy converted to mechanical energy
k2 = (18.19a)
input electrical energy
or
mechanical energy converted to electrical energy
k2 = . (18.19b)
input mechanical energy
Since the energy conversion is always incomplete, k 2 (and thus also k) is
always <1.
Figure 18.2 shows as an example the conversion of mechanical to electrical
energy. In the first step of the conversion cycle (A ⇒ B), the mechanical
energy is applied to the transducer material by stressing it, e.g., in Fig. 18.2 in
the direction of the 3-axis of the material (e.g., a ceramic cylinder poled in the
3-direction). During this stress (pressure) increase up to −Tmax the material
sample is short-circuit, i.e., E3 = 0 and the dielectric displacement growth to
−D3 . Then the short-circuit is opened and the stress is removed (second step
Transducer
E3 E3 E3
of the cycle B ⇒ C). Thereby the electric field across the transducer increases
to Tmax · d33 · ε33 T . In the third step of the cycle (C ⇒ A) the electrical energy
stored in the transducer material is transferred to the external load and the
transducer returns to the initial state. The mechanical input energy, i.e., the
energy mechanically applied to the transducer in the step A ⇒ B is
mech
WAB = 12 sE 2
33 Tmax ,
the energy mechanically released from the transducer in the second step
B ⇒ C is
mech 1
WBC = sD T2 ,
2 33 max
and thus, the electrical output energy transferred to the load reads
1 T 2 1 d233 2
elec
WCA mech
= WAB − WBC
mech
= ε33 Emax = Tmax .
2 2 εT
33
the simplest case of the highest possible symmetry (class 6 mm) as considered
for (18.12), besides the coupling factor k33 , the coupling factors
2 d231
k31 = (18.22)
ε33 sE
T
11
and
2 d215
k15 = (18.23)
ε11 sE
T
55
are required.
T1 = T1 (x1 ), T2 = T3 = T4 = T5 = T6 = 0,
(18.24)
E1 = E2 = 0, ∂E3 /∂x1 =∂E3 /∂x2 =∂E3 /∂x3 =0,
i.e., we only have inertia stress along x1 and an electric field component along
x3 which does not depend on x1 , x2 , and x3 .6
The equation of motion
∂ 2 ui ∂
ρ = fi = Tik (18.25)
∂t2 ∂xk
l
2 E
t
w
1
Fig. 18.3. Piezoelectric bar resonator. The faces vertical to the 3-axis are electroded
6
This can be shown with the aid of Maxwell’s equations for the quasi static case:
= 0, divD = 0.
rotE
18 Small Signal Resonance Methods 431
∂ 2 u1 (x1 , t) 2
2 ∂ u1 (x1 , t) 1
= v s with vs = . (18.26)
∂t2 ∂x21 ρsE
11
to solve (18.26) for periodical functions one finds the following solution for
the free vibrating bar (T1 (x1 = ±l/2) = 0), symmetrically clamped at x1 = 0
(v1 = ∂u1 /∂t = 0 at x1 = 0):
1 Ū 1 1
ū1 x1 = l = d31 tan E
ωl ρs11 (18.28)9
2 t ω ρsE11
2
where U (t) = Re Ū eiωt is the voltage applied on the resonator bar.
Calculating D3 (x1 ) using the second constitutive equation (18.7), integrat-
ing over x1 and forming the time derivative gives the electrical current (that
is calculated in (18.29))
¯ d31 l l
I = iω · w · 2 · E · ū1 + ε33 · · Ū
S
(18.29)
s11 2 t
flowing through the resonator bar and with the aid of (18.28) the admittance
of the free vibrating resonator bar
I¯ wl S 2w d231 1 1
Ȳ = = iω ε33 + i tan E
ωl ρs11 . (18.30)
Ū t t sE11 ρsE
11
2
Admittance Impedance
f– Rp
Im[Y] Im[Z]
f
f
fm fr fa
fn 1
wsC0 fs fs
fp wpC0 fm fp Re[Z]
fa fr fn
Re[Y]
f+
1/Rs
Fig. 18.4. Complex admittance and impedance circles in the vicinity of the char-
acteristic frequencies f (1) and f (2) of a real lossy piezoelectric resonator. In a lossy
resonator, there are three frequencies of interest near the admittance maximum (se-
rial resonance) and, similarly, near the impedance maximum (parallel resonance).
The three frequencies fm , fs , fr associated to f (1) and fn , fp , fa associated to f (2)
correspond to maximum absolute admittance (impedance), maximum conductance
(resistance), and zero susceptance (reactance), respectively
18 Small Signal Resonance Methods 433
with ων identical to ων (1) as given in (18.32). However, one has to note that
due to a complex s11 E the frequencies ων (1) are now also complex. In order
(1)2
to split ων into its real and imaginary part we must, according to (18.32),
split 1/s11 E into its real and imaginary part. Using (18.31) we can write:
1 1 1 1 1 1 1
E
= E + i E = E + iω E , (18.36)
s11 s11 s11 Q(ω) s11 s11 Q(ωλ )ωλ
10
This method and thus also equations identical or equivalent to (18.35) can be
found in many textbooks on the theory of complex analytic functions, e.g., in [7].
We assume Re(ων ) ∼
11 (1)2 (1)
= (Re(ων ))2 which is already for Q > 5 a good approxi-
mation.
434 W. Wersing
Cvm Lm Rvm
• • •
• • •
C2m Lm R2m
I 1: N C1m Lm R1m
V R0 C0
Fig. 18.5. Equivalent lumped-element circuit for the piezoelectrically excited bar
resonator. The circuit elements Cνm , Lm , and Rνm symbolize mechanical resonator
properties; they are proportional to compliance, mass of the resonator bar, and
mechanical losses. The transformer symbolizes the transformation of the mechanical
circuit elements on the secondary side of the transformer to real electrical circuit
elements on the primary side via the piezoelectric effect. The electrical capacitor C0
gives the clamped capacitance of the resonator
(1)
with Qν ≡ Q ων ,
wl S
C0 = ε , (18.39)
t 33
1
Lm = ρwlt, (18.40)
8
and
d31
N =w . (18.41)
sE
11
It can be easily verified that the equivalent circuit in Fig. 18.5 exactly
describes the resonator admittance given in (18.38). Inspection of (18.38) for
(1)
ω → 0 (ω ω1 ) shows that the mechanical capacitors Cνm can be identified
to be
1 8 l E
Cνm = = 2 s . (18.42)
m
L ων
(1)2 π (2ν − 1)2 wt 11
∞
Inserting (18.42) into (18.38) for ω → 0 leads with 1 (2ν − 1)2 =
ν=1
π2 8 to
∞
wl d231 wl
Ȳ (ω → 0) = iω C0 + N 2 m
Cν = iω ε33 + E = iω εT
S
= iωC.
ν=1
t s11 t 33
(18.43)
T
Equation (18.43) shows that ε33 can be obtained by measuring the bar ca-
pacitance at a frequency much lower than the lowest resonance frequency of
the bar.
18 Small Signal Resonance Methods 435
(1)
Inspection of (18.38) in the vicinity of the resonance frequencies ων shows
that the motional resistors Rν m symbolizing the resonator losses can be iden-
tified to be
(1)
ων Lm 1 Lm
Rνm = = . (18.44)
Qν Qν Cνm
Furthermore, from (18.42), it can be seen that
1
ων(1)2 ≡ ωsν
2
= , (18.45)
Lm Cνm
ων(2)2 ≡ ωpν
2
, (18.46)
i.e., the real parts of the characteristic frequencies f (1) and f (2) are noth-
ing but the associated serial resonance frequencies fs and parallel resonance
frequencies fp , respectively.
For the determination of material property coefficients, piezoelectrically
excited resonators are measured in the vicinity of isolated resonances, usu-
ally near the fundamental resonance. In this case, it is convenient to apply
the traditional Butterworth-von Dyke equivalent circuit, a lumped-parameter
equivalent circuit in the simplest form shown in Fig. 18.6. The electrical cir-
cuit elements C1 , L1 , and R1 are obtained from the mechanical elements
C1m , Lm , R1 m (Fig. 18.5) by the transformation via the piezoelectric trans-
former in Fig. 18.5:
2
8 k31
C1 = N 2 C1m = 2 C0 , (18.47)
π 1 − k31
2
1
L1 = 2 Lm , (18.48)
N
1 m ωs1 L1 1 L1
R1 = 2 R1 = = = Rs . (18.49)
N Q1 Q1 C1
C1 L1 R1 =Rs
C0
In terms of the circuit elements of Fig. 18.6 the resonance frequencies read:
2 1 1 (1)2 1 π
ωs1 = ωs2 = = m m = ω1 = E , (18.50)
L1 C1 L C1 ρs11 l
2
2 2 1 1 1 (2)2 2 8 k31
ωp1 = ωp = + = ω1 = ωs 1 + 2 . (18.51)12
L1 C0 C1 π 1 − k31
2
For the derivation of the lumped element equivalent circuits, we have as-
sumed that only sE 11 in the tangent of (18.30) is complex. This approximation
is, at least for piezoelectric ceramics, quite reasonable because many of these
ceramics have dielectric, elastic, and piezoelectric losses in the range of 1–2%
and, therefore, the imaginary part of the most left term in (18.34) cancels
more or less leaving k31 real.
As can be seen from (18.50) and (18.51), the material coefficients s11 E
and k31 can be determined by measuring the serial and parallel resonance
frequencies of a slim bar resonator. Measuring the bar capacitance at a fre-
quency considerably lower than the resonance frequencies allows in addition
the determination of ε33 T and d31 . Similarly, it is possible to determine the
material coefficients s33 D and k33 from the serial and parallel resonance fre-
quencies of a slim bar resonator with the electric field applied parallel to the
length of the bar resonator. Together with ε33 T the coefficients s33 E and d33
can then also be determined.
2
2
12 k31
This equation is equal to (18.34) within the approximation 1 + π42 1−k 2 =
2
31
8 k31
1 + π2 1−k2 . This is excellently fulfilled even for the highest observed k31 values
31
of k31 0.35.
18 Small Signal Resonance Methods 437
3 3
t
Length–extensional 2
l E
modes of bars Pr E l Pr
2
w t
1 w
1
Boundary l > 10w(3w), w > 2t; l > 10w(3w), w ≈ t;
conditions E3 ≠ 0; T1 ≠ 0 D3 ≠ 0; T3 ≠ 0
Measurements C(1kHz) fs fp fs fp
r r
(1) (3) (2) (2) (3)
s33D
Calculations
(4)
(5) (5)
Results
e33T s11E d31 k31 k33 d33 s33E
S ample/mode
3 3 3
t 3 t
2
l E l l Pr
Pr Pr E t E E
2
2 d 2
w w t 1 Pr
1 1 w
1
L ength-extens ional Radial Thickness Thickness shear
E q. w ⋅l T π ⋅ d2 w ⋅l
(1) C = e T33 ⋅ C = e 33 ⋅ C = e T11 ⋅
t 4t t
2
k 33 = k 231 1–k312
= k 2p ≅ k t2 = 2
k 15 =
2
π ∆f ∆f 5 ∆f ∆f π ∆f π ∆f
E q.
π fs f
π p π π fs π fs
(2) = tan = tan ≅ – = tan = tan
2 fp 2 fp 2 fs 2 fp 2 fp fp 2 fp 2 fp 2 fp 2 fp
E 1–s E
2
= D 1
s 11 c 44 = =
E q. 1 = 4r f ⋅ l 2 1
= 4r fs ⋅ l
2
= 4r fp ⋅ t
2 E
s 44
h12
D
p c 33
(3) s D E
s 11 = 2
33 2 4r ⋅ fp ⋅ t
π2r fs ⋅ d
D E E
E q. E
D
s 33 s 11 = c 33 = c 44 = 1 s 44
E
=
s 33 = 2
(4)
1– k233 = E
s 11 1– k31 = D
c 33 1– kt2 D 1– k 2
= c 44 15
2 2
k 2p =
d 33 d 31 2 2 2
E q. 2 2 2d 31 e 33 d 15
(5) k 33 = k 31 = = k t2 = 2
k 15 =
E T E T
s 33 ⋅ e 33 s 11 ⋅ e 33 E
s 11 E
+ s12 T
e 33
D
c 33 ⋅ S
e 33 E
s 44 ⋅ e T11
2 2
k p2 = ⋅ k2 k t2 = k33 – a ⋅ kp
2
E
s 13 =
E q. 1 – s E 31 E q. E q.
sE 1 1– a2 1– k2p
(6) (7) = 1 s E + s E E (8)
sE = 12 2 11 12 s 33 – 2
E
s 11
E
c 33 a2 = 2 ⋅ s 13
E E sE + sE
s 33 11 12
compliance and stiffness coefficients to calculate s13 E from Eq. (7, Fig. 18.8),
or measuring the thickness coupling factor kt ≡ k33t and calculating s13 E
from Eq. (8, Fig. 18.8). Again it should be checked that both the output
values of s13 E in Fig. 18.9 are sufficiently equal. It should be mentioned that
in case of piezoelectric ceramics, a fully electroded and poled ceramic disk
is a very attractive test specimen, because it is easily prepared and shows
a strongly excited fundamental radial resonance mode, free from interfering
modes. However, one must make sure that the ceramic disk’s microstructure
and poling state are identical to those of all other resonator specimens used.
In case of single crystals, it is not convenient to prepare disk specimens. In
this case, one can use a thin rectangular plate to determine the coefficients
s12 E and s13 E from the thickness extensional resonance frequencies, as shown
in Fig. 18.10. To calculate the coefficients s12 E and s13 E from (6–8, Fig. 18.8),
kp in (8, Fig. 18.8) is substituted using (6, Fig. 18.8). Then the coefficients
s12 E and s13 E can be calculated from ((7) and (8), Fig. 18.8).
18 Small Signal Resonance Methods 439
Measurements C(1kHz) fs fp fs fp
ρ ρ
Calculations (1) (3) (2) kp (2) kt (3)
D
k33 c33
E
Input s33 (4)
k31 E
E c33
s11
(6) (6) (8) (7)
=? =?
ε33T E E
Results s12 s13
Fig. 18.9. Determination of elasticity coefficients s12 E and s13 E using radial modes
and thickness extensional modes of a circular disk resonator. Again ε33 T can be ob-
tained from measuring the resonator’s capacitance at 1 kHz. The numbers in brackets
denote the applied formulas shown in Fig. 18.8
2
Thickness-extensional l
modes of a plate Pr E
t
1 w
Boundary
conditions l ≈ w >10t; D3 ≠ 0; S3 ≠ 0
Measurements C(1kHz) fs fp
ρ
Calculations (1) (2) (3) D
kt c33
k33
k31
Input E (4)
s33 E
c33
E
s11
(6)+(7)+(8)
Results E E
ε33T s12 s13
Fig. 18.10. Determination of elasticity coefficients s12 E and s13 E using the thickness
extensional mode of a plate resonator. Again ε33 T can be obtained from measuring
the resonator’s capacitance at 1 kHz. The numbers in brackets denote the applied
formulas shown in Fig. 18.8
440 W. Wersing
3
t
Thickness-shear
l
mode of a plate Pr E
2
w
1
Boundary l ≈ w >10t; D1 ≠ 0; S5 ≠ 0
conditions
Measurements C(1kHz) fs fp
ρ
Fig. 18.11. Determination of shear coefficients s44 E and d15 using the thickness
shear mode of a plate resonator. ε11 T can be obtained from measuring the resonator’s
capacitance at 1 kHz. The numbers in brackets denote the applied formulas shown
in Fig. 18.8
14
The notation for designating the orientation of crystalline bars and plates can be
found in [6].
442 W. Wersing
0dB
Signal 40Ω Level Meter
Generator p
Sample
fs fp
RM
Frequency 10Ω 10Ω 1kΩ
Counter
0 dB fm
Qload =
Pm 3 dB f+ – f–
Qload
Q =
P Pm
20 dB
1 – 10
f– fm f+ fn f
Fig. 18.12. Measuring equipment for the application of scalar small signal resonance
methods
2
fp2 − fs2 C1
keff = 2
= (18.54)
fp C0 + C1
Test sample
Ra
Synthesizer
Synthesizer
Rm 1kΩ Rm
A
Analyzer
Analyzer
B
UA = UAeiϕA
U
PB-A = 20log B UB = UB eiϕB 50kΩ
UA
ϕ = ϕB − ϕA 1 UA
Z= = Rm (cosϕ − i . sinϕ)−1
Y UB
Fig. 18.13. Small signal measuring circuit using a vector level meter. Rm and Ra
should be chosen to fulfil the conditions: Rm Rs (test sample’s impedance at the
series resonance) and Rm + Ra = 50 Ω (characteristic wave guide impedance)
sample constant when the frequency is varied. Therefore, the circuitry used
to excite the sample must have voltage source characteristics. This means the
output impedance should be zero, or at least low with respect to the test
sample’s impedance at the series resonance Rs , i.e., Rm Rs . Depending on
the sample geometry, Rm < 1 Ω could be required. In critical cases, it can
happen that the voltages to be measured by the level meter are so small that
problems with noise occur. Then it is possible to increase Rm (Rm ≈ Rs ) and
to control the voltage across the sample throughout the whole measurement
by means of software tools.
A decisive advantage of computer controlled vector analyzer systems is
their ability to perform piezoelectric measurements automatically with high
precision and usually also with high speed. These features make small signal
resonance methods very attractive for testing piezoelectric materials, trans-
ducers and even devices in development and production. Thereby, the effective
piezoelectric performance of transducers is usually measured via the effective
coupling factor as defined in (18.54).
Finally it should be emphasized that with modern computer controlled
vector analyzer systems one is by far not limited to determine the mater-
ial coefficients solely via the resonance frequencies. One measures the whole
admittance or impedance cycle of a test sample and adjusts the related per-
mittivity, elasticity, and piezoelectric coefficients so that the calculated im-
pedance cycle fits exactly to the measured one. In other words one can use
much more frequencies along the admittance or impedance cycle to determine
the material coefficients with much higher accuracy. In this way, it is also
possible not only to determine the real parts of the material coefficients but
also all their imaginary parts with high precision [12].
444 W. Wersing
References
1. W.P. Mason, H. Jaffe, Proc. IRE 42, 921 (1954)
2. IRE Standards on Piezoelectric Crystals, Proc. IRE 37, 1378 (1949)
3. IRE Standards on Piezoelectric Crystals, Proc. IRE 46, 764 (1958)
4. G. Heckmann, Ergeb. Exakt. Naturwiss 4, 100 (1925)
5. J.F. Nye, Physical Properties of Crystals (Clarendon Press, Oxford, 1964)
6. IEEE Standard on Piezoelectricity, ANSI/IEEE Std. 176 (1987)
7. A. Hurwitz, R. Courant, Allgemeine Funktionentheorie und elliptische Funktio-
nen (Springer, Berlin, 1964)
8. IRE Standards on Piezoelectric Crystals, Proc. IRE 49, 1169 (1961)
9. W.E. Köstler, Diploma Thesis, Technical University Munich, 1989
10. A.W. Warner, M. Onoe, G.A. Coquin, J. Acoust. Soc. Am. 42, 1223 (1967)
11. IRE Standards on Piezoelectric Crystals, Proc. IRE 45, 353 (1957)
12. J.G. Smits, IEEE Trans. Sonics Ultrason. SU23, 393 (1976)
19
Large Signal Resonance and Laser Dilatometer
Methods
A. Wolff
19.1 Introduction
In many piezoelectric devices maximum deformations are desirable. They
can only be achieved under high field strengths (1–2 kV mm−1 ) and cause
large mechanical stress amplitudes in the piezoelectric material. This calls for
piezoelectric materials with a strong piezoelectric effect that are capable of
withstanding both high electric field strengths and large mechanical stresses.
Under these large signal conditions, the properties of piezoelectric materials
are considerably non-linear and show hysteretic behaviour. As discussed in
Chap. 3.3, the non-linear behaviour can be attributed to the motion of non
180◦ domain walls. Therefore, the non-linearity of the piezoelectric response
and the accompanying hysteresis can be described with the aid of Rayleigh’s
law (Equations (3.48) and (3.49) in Chap. 3.3). The most important aspect
of the Rayleigh-like behaviour is the fact that hysteresis and non-linearity in
soft piezoelectric ceramics are essentially linked, as can be seen in Fig. 19.7,
that is, both hysteresis and non-linearity are results of the same-domain wall
pinning processes. Since the hysteretic response is closely related to the elec-
trical and mechanical losses occurring in piezoelectric ceramics or ferroelectric
perovskites, in general, the performance of power ceramics can be tested sub-
stantially by measuring either the dielectric losses at large field amplitudes
or the mechanical losses at large mechanical stress amplitudes. However, to
perform these measurements, it is not sufficient to extend simply the stan-
dard small-signal capacitance/loss tangent techniques according to IRE or
IEEE standards [1] or the impedance measurement techniques as described in
Chap. 18 to higher-signal amplitudes. For then the measurement results would
be considerably falsified by an overheating of the test samples. With this in
mind, pulsed test signals are mostly used [2].
Out of several possible large signal measurement methods, two are de-
scribed in detail in the following. They have been selected because of their
usefulness in the case of soft piezoelectric ceramics.
446 A. Wolff
100
50
oscillation signal [%]
–50
–100
The clamp holders, the force sensor, and the mechanical load have a cen-
tral opening through which the measuring beams can enter and exit. The
measuring beams focused directly on the actuator surface are reflected there,
allowing for a contactless, and thus, inertia-free distance measurement with a
time resolution in the microsecond range and a precision of 0.1 µm. A suitable
laser measuring system was chosen that is marked by its very high measuring
speeds of up to 10 m s−1 maximum velocity of the actuator movement [3].
The main improvement of the measuring system is the introduction of a
servo piezoelectric actuator – as opposed to a mechanical spring – that takes
over the role of the mechanical load (see Fig. 19.2). This has a number of
advantages:
• The stiffness of the load can be altered electrically, without requiring the
measurement holders to be rearranged. This means that a constant sta-
tic pre-stress of the actuator can be maintained throughout the measure-
ment cycle.
• In contrast to real mechanical loads that show non-linear properties due to
friction effects, etc., ideal linear loads (Hookean behaviour) can be realized
with the servo piezoelectric actuator.
• The stiffness of the load can be set independently of the static pre-load.
• Realization of load stiffness settings ranging from negative to even infinite
values that cannot be achieved with real passive loads.
Fig. 19.2. Schematic set-up of a force-deflection measuring rig with a servo piezo-
electric actuator as an active load
448 A. Wolff
7
2500
6 working characteristics @ 2 kV/mm
2000 5
4
force [N]
1500
3
1000 2
500
0 1
–20 –10 0 10 20 30
deflection [µm]
Fig. 19.3. Force-deflection diagram with load stiffness values of 0, 20, 40, 80, and
160 N µm−1 at a field strength of 2 kV mm−1 at room temperature
Points 3–7 are on the force–distance state curve for 2 kV mm−1 trigger field
strength, which, during the rise of force regimen throughout the measurement,
can be very closely described by a straight line. The slope of this line, whose
absolute value depicts the actuator’s stiffness, is −43 N µm−1 . The idle travel
of 41.3 µm and the extrapolated blocking force of 1,800 N can also be seen in
the force–distance diagram in Fig. 19.3.
It is worth noting the difference between the above aforementioned slope
of the regression line of −43 N µm−1 and the slope described by a line through
Points 1 and 2 of −59 N µm−1 . For a linear behaviour, the same stiffness values
would be expected for both instances. Yet the difference can be explained by
the different field strengths of 2 kV mm−1 in the first and 0 kV mm−1 in
the second instance. Piezoelectric ceramics at a field strengths of 2 kV mm−1
show non-linear large signal properties that differ substantially from those at
0 kV mm−1 .
For this method [4], the piezo-ceramic sample is triggered quasi-statically with
a uni-polar sinus pulse or a series of single sinusoidal pulses. The energy loss
Wloss can be calculated from the measured voltage and current curves U (t)
and I(t) (see Fig. 19.4) using the equation
T
Wloss = U (t )I(t )dt ,
0
where T denotes the pulse length.
450 A. Wolff
200
100
50
50
current [mA]
–50
Fig. 19.4. Voltage–time and current–time diagram with load stiffness values of 0,
20, 40, 80, and 160 N µm−1 at a field strength of 2 kV mm−1 at room temperature
100
90 20 N/um, 2000 N
80 20 N/um, 1000 N
20 N/um, 100 N
70
80 N/um, 2000 N
work [mWs]
60 80 N/um, 1000 N
50 80 N/um, 0N
40
30 electrical
20
10
0 mechanical
–10
–1 0 1 2 3
t [ms]
Fig. 19.5. Electrical energy–time curves for different mechanical load conditions of
a piezoelectric actuator
t
Wel (t) = U (t )I(t )dt
0
12
deflection [µm]
10
4 pulse duration
1 ms
10 ms
2 100 ms
1s
10 s
0
0 20 40 60 80
voltage [V]
Fig. 19.6. Deflection-voltage diagram for different pulse durations (sinusoidal wave-
form, pulse amplitude 80 V, load condition 900 N, stiffness of load 20 N µm−1 )
452 A. Wolff
strain [10–4]
deflection [µm] 20
10
18
16
8
14
12
6
10
dynamic d33 of 500 pm/V
8 4
at 80 V
6
4 2
2
0 0
0 20 40 60 80 100 120 140
voltage [V]
30
15
10
–5
slope of linear fit 0,147 µm/K
–10
–15
–20
–20 0 20 40 60 80 100 120 140 160 180
T [°C]
Fig. 19.9. Thermal expansion curve for a piezoelectric ceramic with thermal depo-
larisation (first thermal cycle)
30
2nd Cycle with slight Depolarisation
25 linear fit of 2nd Cycle
1st Cycle with moderate Depolarisation
20
thermal expansion [µm]
15
10
–5
slope of linear fit 0,194 µm/K
–10
–15
–20
–20 0 20 40 60 80 100 120 140 160 180
T [°C]
Fig. 19.10. Thermal expansion curve for a piezoelectric ceramic with slight depo-
larisation (second thermal cycle)
References
1. IEEE standard on piezoelectricity, E-ISBN: 0-7381-2411-7, (1988)
2. K. Lubitz, W. Wersing, Ferroelectrics, 40, 237 (1982)
3. A. Wolff, D. Cramer, H. Hellebrand, I. Probst, K. Lubitz, in Proceedings ISAF
’94, Pennsylvania State University, University Park, PA USA, 7–10 August 1994,
p. 755
4. A. Wolff, D. Cramer, H. Hellebrand, C. Schuh, T. Steinkopff, K. Lubitz, in Pro-
ceedings ISAF ’96, Rutgers University, East Brunswick, NJ, USA, 18–21 August
1996, p. 317
5. H. Walcher, Winkel- und Wegmessung im Maschinenbau (VDI-Verlag, Düsseldorf
Germany, 1985)
6. R. Schödel, Proc. SPIE 5858, (2005)
7. Sios Messtechnik GmbH, Anwenderdokumentation zum Miniaturinterferometer
mitPlanspiegelreflektor, Ilmenau Germany, (1999)
20
Ferroelastic Characterization of Piezoelectrics
T. Fett
Ti
Si = Siel + Sipl = + Sipl , (20.1)
Y
where Y is not necessarily a constant, but may depend on the amount of
nonelastic strains, as will be shown later. The methods for the determination
of the Young’s moduli, well-known from conventional ceramics and also ap-
plicable to unpoled piezoelectric materials, are much more complicated for
nonisotropic materials due to the tensorial character of the modulus and
the special electric boundary conditions [1]. The conditions of a constant
electrical field, Ez ≡ E3 = const, or a constant dielectric displacement,
Dz ≡ D3 = const, influence the elastic properties. This effect may be il-
lustrated for a uniaxial stress state with Tx ≡ T1 = 0 exclusively. In the case
of a piezoelectric material, we are interested in the components
S1 = sE
11 T1 + d31 E3 ,
(20.2)
D3 = d31 T1 + εT33 E3 .
T1
x,1
W
x
z,3
E3=0 D3=0
=
∆p ∆p
Pr Pr
z
(a) T1 (b)
Fig. 20.1. Ceramic specimen, poled vertically to the applied stress T1 , a with short-
circuited and insulated electrodes, b specimen, poled vertically to specimen’s length
axis, under bending load
w
U= Ez dz = 0, (20.3)
0
S1 = s11 T1 ,
(20.4)
D3 = d31 T1 .
E T1
Y11 = = 1/s11 , (20.5)
S1
where superscript E denotes a constant electric field (especially E = 0).
Let us now consider the case of insulated electrodes. In Fig. 20.1a, this
electrical condition is symbolized by an open switch, which prevents any flow
of current. The real charges on the electrodes are constant. Consequently, Dz
is constant, too. For Dz = 0, (2) yields
d31 T1 T1 1
E3 = − ⇒ S1 = E 2 , (20.6)
T
ε33 s11 1 − k31
with the coupling factor
|d31 |
k31 = . (20.7)
sE T
11 ε33
20 Ferroelastic Characterization of Piezoelectrics 459
D 1 1 E 1
Y11 = = Y11 2 . (20.8)
s11 1 − k31
E 2 1 − k31
In tests where the electrodes are not connected, we trivially obtain from
D
the measurement Y11 .
In bending tests, the electrical boundary conditions are much more com-
plicated. First, the bending bar is studied in flat position with the electrodes
at the tensile and compression surfaces (see Fig. 20.1b). The change in po-
larization due to the piezoelectric effect (symbol ∆ dropped in the following
considerations) is
P3 = d31 T1 = d31 T1b 2z/w (20.9)
500 500
T1 T3
400 400
T1⊥Pr T3||Pr
(MPa) (MPa)
300 300
200 200
60 GPa
52 GPa
100 100
140 GPa 150 GPa
0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
(a) S1 (%) (b) S3 (%)
Fig. 20.2. Compression tests on soft PZT ceramics. a and b Cao and Evans [2]
(poled perpendicular ⊥ and parallel to the specimen length axis)
60 60 60
unpoled
40 40 40
0 50 100 150 200 0 50 100 150 200 0 50 100 150 200
–T1 (MPa) –T1 (MPa) –T3 (MPa)
Fig. 20.3. Young’s modulus determined from partial unloading tests for the three
poling states during loading up to 150 MPa and during unloading
Fig. 20.4a vs. the applied stress for unpoled material. Starting from σ ≈ 0.39,
the Poisson’s ratio decreases and approaches an asymptotic value of about
σ → 0.3. During unloading, σ is nearly the same as under increasing load and
reaches σ ≈ 0.37–0.38 after total unloading. The poled PZT ceramics show
a more pronounced hysteresis effect. Figure 20.4b represents the results for
the -poled material modification. In the case of the ⊥-poled material, the
transverse strain S2 was measured in the poling direction (resulting in σ⊥ )
and perpendicularly to the poling direction (σ ). In Fig. 20.4c, the average
value σaver = 1/2(σ + σ⊥ ) responsible for the volume change is plotted. Also
this average value tends towards 0.3.
20 Ferroelastic Characterization of Piezoelectrics 461
σ σ σaver
||-poled ⊥-poled
unpoled
0.5 0.5 0.5
loading
loading loading
unloading
unloading unloading
0.4 0.4 0.4
Fig. 20.4. Poisson’s ratio obtained from a partial unloading test a unpoled PIC
151 (two specimens), b ||-poled PZT, c ⊥-poled PZT average ν aver = 1/2(ν || + ν⊥ )
50 50
T1 Tension
fracture T1
40 30
(MPa) 20
30 unpoled
(MPa) unpoled
10 ⊥ poled
20
0.40
10 5
⊥-poled
1
3
0 0.5 1 1.5 2 0.01 0.03 0.1 0.03 1
(a) S1 (%°) (b) S1 PI (%°)
Fig. 20.5. Tensile tests performed on unpoled and poled PZT PIC 151. Fractured
specimens marked by circles, a stress vs. total strain, b stress vs. plastic strain
100
⏐T1⏐
⏐T1⏐ 100
80 compression
(MPa) (MPa)
50
60 40 compression
30
40
} tension 20
tension 0.4
20
10 1
0.01 0.1 1
0
(a) 0 1 2 3 (b) ⏐S1PI⏐(%°)
⏐S1⏐(%°)
Fig. 20.6. Stress–strain curves (unpoled PZT): a linear plot T1 vs. S1 ; b represented
as log(T1 ) − log(S1 pl ) diagram
500
150
–T1,3 unpoled
⏐T1⏐ 400
⊥−poled (MPa)
100 300 ⊥ poled
⊥-
(MPa) compression
200 ||-poled
50
} tension 100
0 0
0 1 2 3 0 2 4 6 8 10
(a) ⏐S1⏐(%°) (b) –S1,3 (%°)
Fig. 20.7. Linear stress–strain curves a ⊥-poled PZT b compression tests on PIC
151, influence of the poling state
the unpoled and in Fig. 20.7a for the poled specimens. The tensile specimens
show larger strains at the same stress level for unpoled and ⊥-poled PZT. In
the case of -poled PZT, the compressive strains are significantly increased.
The plastic strains obtained from the data of Fig. 20.5a are plotted in
Fig. 20.5b vs. the stress in a logarithmic representation. The experimental
results, represented by the lines, can be expressed for Sipl < 1% as
Ti = A+ Sipln , (20.10)
with A+ = 635 MPa and n = 0.40 for the unpoled material and with A+ =
545 MPa and n = 0.43 for the poled material.
20 Ferroelastic Characterization of Piezoelectrics 463
Ti = A− |Sipl |n , (20.11)
with n = 0.4 and A− = −940 MPa for the unpoled and n = 0.43 and A− =
−975 MPa for the ⊥-poled PZT.
Apart from the poling state, the applied electrical field will also influence
the stress–strain curves. Results for PZT PIC 151 can be seen in [8]. As a
singular result, the compression strain is shown in Fig. 20.8a for the unpoled
material under a variable transverse electric field.
The nonsymmetry in the deformation behavior must have consequences
on the stress distribution in nonhomogeneously loaded components, for ex-
ample, in a bending bar. The information given in Fig. 20.6a for the unpoled
material is sufficient to compute the stresses in a bending bar. A result is
given in Fig. 20.8b for the unpoled material, where the stress distribution is
shown for various elastically computed bending stresses T1el . The resulting
stresses become strongly nonlinear. A significant stress reduction at the ten-
sile surface is visible. While the stresses are reduced at the tensile surface,
the compressive stresses increase at the compressive surface. This deviation
from the linear-elastic behavior affects the strength [9] and toughness data
obtained in bending tests.
Finally, it may be concluded that:
• PZT shows a nonsymmetric stress–strain behavior under tension and
compression.
T1 T1el
150 80 MPa
1.0 0 50 70 MPa
60 MPa
–T1
E=1.2 kV/mm 0.5
(MPa)
100 tension
0
(MPa)
50
−50
0.75
compression
0 −100
0 1 2 3 4 −1 0 1
(a) –S1 (‰) (b) z
• Plastic strains of unpoled PZT at the same stress in tension exceed those
in compression by a factor of about 2.
• PZT, poled perpendicularly to the external load, shows a stronger plastic
deformation than unpoled material.
60
⏐T1⏐,τ compression 20
T2
50 tension 10
(MPa)
40 0
(MPa)
30 torsion
–10
20 –20
10 –30
–40
0 –40 –20 0 20
0 0.5 1 1.5 2
(a) ⏐S1⏐,(%°) (b) T1 (MPa)
Fig. 20.9. a Stress vs. plastic strain curves for PIC 151 under uniaxial and biaxial
loading, stresses plotted vs. S1 (tension, torsion) and −S2 for compression, b T1 −T2
representation of the onset of plastic deformation (defined as an effective strain
S0 = 0.01%). Tensile and compression test: filled circles, torsion test: open circles,
curve: (15)
20 Ferroelastic Characterization of Piezoelectrics 465
√
For torsion tests with S2 = −S1 and S3 = 0, this yields S0 = 2S1 / 3.
This plastic strain seems to be sufficiently small when compared to the elas-
tic strains. The related tensile and compression stresses are shown in Fig. 20.9b
as filled circles. The torsion results are given as open circles.
It is a clear a priori that a Tresca or von Mises yield criterion can never
describe the experimental results. Application of a Drucker–Prager yield cri-
terion [12] seems to be much more promising. Here, the stress sign enters via
the hydrostatic stress Th . For T3 = 0, it reads
Th = 13 (T1 + T2 ). (20.13)
In terms of the first invariant J1 of the stress tensor and the second in-
variant J 2 of the stress deviator, the yield condition is given by
αJ1 + J2 = κ, (20.14)
with m = 0.27, n = 1.52, and B = 2.7 × 10−4 (for T in MPa, t in s), where
the exponent n results from the plot of the creep strain at t = 120 s vs. the
applied stress. Similar creep effects are also visible in compression loading and
under electric fields (see, e.g., [14, 15]).
466 T. Fett
1.2 1.2
1 under load 1
S1 S1 after unloading
0.8 S1c 0.8
(‰) (‰)
0.6 0.6
0.2 0.2
0 0
0 10000 20000 30000 0 200 400 600 800 1000
(a) time (s) (b) time (s)
Fig. 20.10. Tensile creep curve of PZT PIC 151 at T1 = 25 MPa, a deformations
under load, b back deformations after unloading (unpoled material)
(‰)
c
0.2 0.1
S1
0.1
(‰)
0.05
25 MPa 0.05 1.52
0.02 T1 21.3 MPa
15.4 MPa 1
10.7 MPa
10 100 1000 10000 10 20 30
While constant load is applied in the creep tests, the load depends on time
in most practical applications. Under such conditions, the creep strains can be
obtained from the so-called hardening rules, similar to the high-temperature
creep of conventional ceramics. Two well-known rules are the time and strain
hardening rules (see, e.g., [7]).
References
1. T. Fett, D. Munz, J. Test. Evaluat 28, S.27 (2000)
2. H. Cao, A.G. Evans, J. Am. Ceram. Soc. 76, 890 (1993)
3. A. Schäufele, Ferroelastische Eigenschaften von Blei-Zirkonat-Titanat-
Keramiken, Fortschrittsberichte VDI, Reihe 5, Nr. 445, (VDI-Verlag
Düsseldorf, 1996)
20 Ferroelastic Characterization of Piezoelectrics 467
R.E. Cohen
potential model, which can then be used to study the effects of temperature
and simulate disordered materials.
A number of major advances in our understanding of piezoelectric materi-
als are due to first-principles studies. First is the importance of hybridization.
Ferroelectricity is due to the competition between long-range forces, which fa-
vor off-centering, and short-range forces that favor the high-symmetry centric
phase [5]. Unlike the venerable Slater rattling ion model [16], the key element
in oxide ferroelectrics is covalency, or hybridization between the cation and
its oxygen neighbors that allows the cation to move off-center. This concept
is now widely used in experiments and for development of new piezoelectric
materials. Second is the concept of polarization rotation, which is responsible
for the giant electromechanical coupling seen in relaxor ferroelectrics such as
Pb(Mg1/3 , Nb2/3 )O6 (PMN) − PbTiO3 (PT) [17]. Third is the relationship
between cation ordering and polar nanoregions in relaxors [18, 19]. Fourth is
the prediction of a morphotropic phase boundary in pure PT at high pressure,
with huge electromechanical coupling in the transition region, indicating that
the main effect of the relaxor PMN or PZN, for example, is to tune the transi-
tion to zero pressure, rather than something intrinsic to relaxor behavior [11].
Fifth is the discovery of re-entrant ferroelectricity, with ferroelectricity reap-
pearing at very high pressures, indicating the possibility of whole new classes
of ferroelectric materials [20].
First-principles studies of ferroelectrics have also given rise to major
advances in theoretical methods. The development of the modern theory of
polarization [21–23] was motivated entirely to understand piezoelectrics, and
has other broader implications as well [24, 25]. The theory of insulators under
applied electric fields was also developed to understand piezoelectric materi-
als [26, 27].
The goal of computational research on piezoelectrics is threefold: to help
understand experiments, to help guide experiments, and to make predictions
for new materials. An example of piezoelectric materials by design is exem-
plified by the predictions of interesting and tunable properties of ferroelectric
superlattices [28–31].
by an applied electric field, and generally have a phase transition with in-
creased temperature to a nonpolar, paraelectric state. Many perovskites are
ferroelectric, and this review will mainly consider the oxide perovskites. Ferro-
electricity in the LiNbO3 structure is very similar in origin, but the structure
allows uniaxial ferroelectricity only (the soft mode is nondegenerate and the
polarization must be along the c-axis), which significantly changes the dy-
namical behavior.
Ferroelectrics research was a very hot topic in the 1960s and earlier, driven
by experimental studies and phenomenological models. Starting in the 1990s,
the field was rejuvenated with the introduction of first-principles methods
along with a new generation of fundamental experimental studies.
computations are then used to obtain the parameters of the Hamiltonian, and
energy minimization, Monte Carlo, and/or molecular dynamics can then be
used to obtain materials properties [39–41].
Fig. 21.1. Energy vs. phonon displacement for (a) PbTiO3 and (b) BaTiO3 show-
ing the multiple-well potential surfaces that underlie ferroelectric behavior. Also
illustrated is the great sensitivity to volume and shear strain (from [5])
476 R.E. Cohen
Fig. 21.2. Energy vs. mode coordinate for PbTiO3 for the unstrained, cubic lattice
(dotted), and relaxed tetragonal lattice (solid). Note that the energy is a minimum
along the (111), rhombohedral, direction for the unstrained lattice and along the
(001), tetragonal, direction for the strained lattice. The great energetic importance
of the lattice strain is also evident. This behavior contrasts with that of BaTiO3 ,
which would show (111) always most stable, which is why BaTiO3 has a series of
three-phase transitions at zero pressures, whereas PbTiO3 has only 1 (from [41])
A1 A1 A1
E B1
30
Soft TO mode frequency (cm–1)
200 Cubic 6
20
Rhombohedral
Orthorhombic
(meV)
(THz)
Tetragonal
100 F1u 3
10
E
B2
0 500 1000
Temperature (K)
Fig. 21.3. Soft-mode behavior in BaTiO3 . One mode hardens at each transition,
and the mode frequencies do not go to zero. Only the orthorhombic to rhombohe-
dral transition looks similar to a conventional soft mode, even though it should be
understood as having a large order–disorder character (from [43])
478 R.E. Cohen
BaTiO3
900
2’
600 LO
1/λ (cm–1)
TO
LO 15
25’
300 25
12’
TO 25
LO 15
0 A
TO
300i
Γ Χ Μ Γ R
PbTiO3
900
2’
600 LO
1/λ (cm–1)
TO
LO 12’
25’
300 15
25
TO
LO
0 A
15
25
TO
300i
Γ Χ Μ Γ R
PbZrO3
900 2’
LO
12’
600 TO
1/λ (cm–1)
300 LO 25’
15
TO
LO
25
0 A
15
TO
25
300i
Γ Χ Μ Γ R
Fig. 21.4. Computed phonon dispersion curves for cubic BaTiO3 , PbTiO3 , and
PbZrO3 (from [47])
∗ V ∂Pα
Zs,αβ = ,
e ∂us,β
are greatly enhanced in oxide ferroelectrics, and the large coupling to electric
fields is partly due to the softness of the potential surface, and partly due to
the enhanced effective charges, since the field E couples with displacements d
via Sum Z ∗ d.
In ferroelectrics, the effective charges are often much larger than their
nominal values, contrary to chemical intuition that they would be lower than
their nominal charges due to covalency, as would be static Mulliken charges, for
example. Tables 21.1 and 21.2 show computed effective charges for PbTiO3 [8]
and rhombohedrally ordered PMN [54]. The effective charges can be greatly
21 First-Principles Theories of Piezoelectric Materials 481
Table 21.2. Computed effective charges for 1:2 ordered PMN with symmetry C2
(from [54])
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
Atom WP Zxx Zyy Zzz Zxy Zxz Zyx Zyz Zzx Zzy
type
Pb 1a 3.34 3.93 3.35 −0.29 −0.24 −0.50 −0.50 −0.23 −0.28
Pb 2c 3.57 3.79 3.50 −0.13 0.00 0.09 0.58 0.24 0.53
Mg 1b 2.89 2.87 2.88 0.04 0.14 0.02 0.02 0.14 0.04
Nb 2c 6.39 6.41 5.70 0.03 −0.39 0.11 −0.26 −0.13 −0.08
O1 2c −3.39 −2.74 −2.32 −0.11 0.43 0.05 0.12 0.43 0.03
O2 2c −1.95 −2.19 −4.03 −0.06 −0.19 −0.04 0.03 −0.19 −0.10
O3 2c −2.39 −3.69 −2.34 −0.20 0.16 −0.35 0.19 0.15 0.33
O4 2c −2.08 −2.27 −4.79 −0.03 −0.56 −0.10 −0.19 −0.14 0.12
O5 1a −2.09 −5.37 −2.07 −0.13 0.31 0.20 0.19 0.31 −0.15
WP Wyckoff position
enhanced, so that Z ∗ for oxygen can reach magnitudes of above 5. The fact
that these enhanced charges come from hybridization was shown in the same
way that the importance of hybridization was shown for the energies. When
the d-states are removed from the basis, the effective charges regain nominal
values [55].
Enhanced effective charges can also be considered to arise from atomic
polarization, as can be seen from the shell model. The shell model is a me-
chanical model for materials, where atoms are modeled as charged “cores”
and charged “shells” connected by springs. Atomic polarization is modeled by
displacement of an atom’s shell from its core. Shell models can reproduce large
effective charges through the “knock-on” effect. For example, when an oxygen
core is displaced, its shell is displaced. The shell interacts with other ionic
shells through its short-range interactions (springs) and charge, and the other
shells move as well. This can give a much greater change in polarization of the
crystal than that would occur in an isolated ion, and thus an enhanced effective
charge. Effective charges can also be determined experimentally, from LO–TO
splitting and from oscillator strengths. In fact, the enhancement in effective
482 R.E. Cohen
Table 21.3. Piezoelectric (in C/m2 ) and elastic (in GPa) constants for PbTiO3
at the experimental lattice computed using density functional perturbation theory
(DFPT) and frozen strains (FS) [11] compared with experiment (from [11])
Method e31 e33 e15 c11 c12 c13 c33 c44 c66
DFPT 2.06 4.41 6.63 230 96.2 65.2 41.9 46.6 98.8
FS 2.07 4.48 6.66 229 95.6 64.3 41.2 47.2 98.6
Experiment 2.1 5.0 4.4 237 90 70 66 69 104
charges over their nominal charges was observed first experimentally [37, 38],
although its significance was not widely recognized.
Piezoelectric constants are derivatives of the polarization with respect to
strain:
∂Pα dusµ
∗
eαβδ = = Zs,αµ ..
∂εβδ sµ
dεβδ
0
R3m
H - Hcubic (meV)
0.2
H - HP4mm (meV)
0.0
–20 Cm
–0.2
–40 –0.4
20 22 0.0
–60 P4mm
R3m
–80 –0.2
(a) 0 8 16 24 (b) 8 9 10 11 12
P4mm Cm R3m cubic
12
Elastic constants (GPa)
60 c44 ω2(E)
c11-c12
ω2 (103 cm–2)
ω2(A1)
8
40 ω2(F1u)
20 4
ω2(E)
0 0
(c) 0 4 8 12 16 20 24
12,000
Dielectric susceptibility
χ11 = χ22
9,000 χ33
6,000
3,000
0 4 8 12 16 20 24
(d) LDA pressure (GPa)
Fig. 21.5. Stability of PbTiO3 phases and properties vs. pressure. (a) Enthalpy
difference with respect to the cubic (C) phase for the tetragonal (T ) and rhombohe-
dral (R) phases. (b) Enthalpy difference with respect to the T phase for the R and
monoclinic (M ) phases. (c) Elastic constants for the T and R phases, and square of
the lowest optical phonon frequencies ω 2 for the T, R, and C phases. (d) Dielectric
susceptibility (from [11])
have been given much attention since the discovery of huge electromechani-
cal coupling in single crystals such as PbMg1/3 Nb2/3 O3 –PbTiO3 (PMN–PT)
and PbZn1/3 Nb2/3 O3 –PbTiO3 (PZN–PT) [56]. These crystals are rhombohe-
dral at the relaxor end of the phase diagram and tetragonal at the PbTiO3
side, with a morphotropic phase boundary and intermediate monoclinic or or-
thorhombic phase in-between. The largest piezoelectric response occurs near
the morphotropic phase boundary in the rhombohedral phase, with the po-
larization along the [111] directions, and an applied electric field along the
[001] direction. The large response has been understood as due to polariza-
tion rotation [57].
484 R.E. Cohen
60 T M R
e15
e (C/m2)
40 e33
20
(a) 0
0 4 8 12 16 20
9000
d15
6000
d (pC/N)
d33
3000
0
0 4 8 12 16 20
(b) LDA pressure (GPa)
The polarization rotation effect can give much higher coupling than the
normal collinear effect, in which the polarization and electric field are parallel
to each other. The polarization along [111] is compatible with a rhombohedral
structure, along [001] with tetragonal, and along [011] with orthorhombic. In-
termediate directions like [xxz] or [x0z] are monoclinic. When the polarization
is rotated with an applied field, the response is like that of a field-driven phase
transition (Figs. 21.7 and 21.8). The large response in PMN–PT, for example,
can be understood by considering pure PT. PT has a large tetragonal strain of
6%. If PT had a rhombohedral phase at zero pressure, one could apply a field
to the rhombohedral phase driving the structure toward tetragonal. The net
result would be a huge 6% strain. However, PT does not have a rhombohedral
phase at zero pressure, but a solid solution with PMN, PZN, or PZ does have
a rhombohedral phase that allows this rotation effect. The resulting strains
are smaller than in pure PT, but still much larger than the collinear effect.
The huge electromechanical coupling predicted in pure PT under pressure is
due to polarization rotation as well. Ahart et al. recently confirmed these the-
oretical predictions using cryogenic in situ microRaman and high resolution
synchrotron X-ray experiments [57B].
The polarization rotation effect has been verified now by numerous experi-
ments. It leads to intermediate low-symmetry monoclinic or orthorhombic
phases between rhombohedral and tetragonal phases on the phase diagram
[58–63]. Rotation has also been seen in situ measurements under applied elec-
tric fields [64–66].
21 First-Principles Theories of Piezoelectric Materials 485
Fig. 21.7. Polarization rotation effect. An applied field along the cubic [001] direc-
tion rotates the polarization from [111], the rhombohedral direction, toward [001],
through intermediate monoclinic states (from [57])
in a nonpolar matrix, but some data suggest that it might be better to re-
gard the polarization correlation, which would show long tails rather than a
clear separation between polar and nonpolar regions. Theoretical models for
disordered relaxors will be considered further below.
Relaxors can display two types of disorder. One type of disorder is the
polarization disorder mentioned above. Another is chemical disorder. PMN
is not generally found in a long-range ordered state, but there is evidence
for small chemically ordered domains [68–70]. Other relaxors such as PST
(Pb(Sc0.5 Ta0.5 )O3 ) can be chemically ordered by annealing, so that the effects
of chemical order on their relaxor properties can be studied experimentally. In
the ordered state, PST has a normal ferroelectric phase transition but shows
relaxor behavior when chemically disordered [71, 72]. Statistical mechanical
calculations show that the PST ground state is ordered [73–75]; the disordered
state occurs from lack of equilibrium in the B-cation distribution caused by
the slow diffusion times of the highly charged cations. The question remains
if the chemical ordered state will always be ferroelectric in a relaxor, but it
seems probable. Tinte et al. have shown in simulations that polar nanoregions
form around chemical ordered nanoregions, and this leads to many of the
observed phenomena of relaxor ferroelectrics [19].
First-principles calculations for PMN do show a polar, ordered ground
state with monoclinic C2 symmetry with 1:2 stacking of Mg and Nb along
(111) planes [76] or 1:1 symmetry [77, 78]. If cubic PMN did form nanoscale
monoclinic domains, it would be difficult to determine definitively using dif-
fraction data since it is very difficult to deconvolute the diffraction data from
a sample with 12 twin variants. Nanodomains of the twin variants would pro-
vide static random fields that would contribute further to the disorder of the
material and possibly some of the frequency-dependent relaxor behavior.
The effective charges in relaxors may contribute to the low-symmetry
ground state found for PMN. Computations of effective charges in PMN show
them to be very anisotropic. Local fields in the polar phase will lead to oxygen
displacements in oblique directions, giving rise to a monoclinic ground state.
Fig. 21.9. Computed energy surface for LiTaO3 for displacing Li only, O only, or
Li and O along the soft-mode coordinate. Clearly, the instability is not due to Li
alone, and the concerted motion of Li and O is important. The results for LiNbO3
are very similar (from [42])
a triple well potential surface, which is not consistent with the first-principles
results. Frozen phonon calculations showed that the double well is largely due
to hybridization between the d0 Nb5+ and Ta5+ ions with oxygen, much as
for the oxide perovskite, whereas previous work had considered the ferroelec-
tricity to be mainly due to hopping of the Li. The Li does hop, but mainly in
response to the changing fields from the off-centering of the Nb or Ta. Linear
response computations gave the phonon frequencies of the ground state fer-
roelectric state [79]. Modeling of the double well potential as an anharmonic
oscillator gave good agreement with experiments. One outstanding problem
is a first-principles understanding of the difference of Tc of LiNbO3 (1,480 K)
and LiTaO3 (950 K) since they have almost identical potential surfaces.
this study, including the importance of strain coupling, although the words
“ferroelectric” or “piezoelectric” do not appear in the paper. GeTe remains
a material with the simplest possible structure for a ferroelectric phase tran-
sition. There are only two atoms per unit cell and the paraelectric phase is
rocksalt structured. Waghmare et al. [80] revisited GeTe and other chalco-
genides, using first-principles techniques developed in the last 20 years that
were unavailable in the 1980s, including linear response and Barry’s phase
techniques. They found large effective charges and large optical dielectric
constants, and showed that the ferroelectric instabilities are partly due to
lone-pair activity.
21.1.2.5 KDP
21.1.2.6 Multiferroics
21.1.3 Summary
Acknowledgments
References
1. A.F. Devonshire, Philos. Mag. 3(10), 85 (1954)
2. D. Vanderbilt, M.H. Cohen, Phys. Rev. B 63, 94108 (2001)
3. L.L. Boyer et al., Ferroelectrics 111, 1 (1990)
4. R.E. Cohen, H. Krakauer, Phys. Rev. B 42(10), 6416 (1990)
5. R.E. Cohen, Nature 358, 136 (1992)
6. R.E. Cohen, Nature 362, 213 (1993)
7. K.M. Rabe, J.D. Joannopoulos, Phys. Rev. B 36(12), 6631 (1987)
8. G. Saghi-Szabo, R.E. Cohen, H. Krakauer, Phys. Rev. Lett. 80(19), 4321
(1998)
9. G. Saghi-Szabo, R.E. Cohen, H. Krakauer, Phys. Rev. B 59, 12771 (1999)
10. Z. Wu et al., Phys. Rev. Lett. 94(6), 069901 (2005)
11. Z. Wu, R.E. Cohen, Phys. Rev. Lett. 95, 037601 (2005)
12. R. Resta, Model. Simul. Mater. Sci. Eng. 11(4), R69 (2003)
13. K.M. Rabe, P. Ghosez, in Physics of Ferroelectrics: A Modern Perspective, ed.
by C.H. Ahn, K.M. Rabe, J.M. Triscone (Springer, New York, 2006)
14. C. Fiolhais, F. Nogueira, M. Marques (eds.), A Primer in Density Functional
Theory. Lecture Notes in Physics, vol. 620 (Springer, New York, 2003), p. 256
15. W. Kohn, L.J. Sham, Phys. Rev. A 140, 1133 (1965)
21 First-Principles Theories of Piezoelectric Materials 491
22.1 Introduction
The classic work of Devonshire [1, 2] in developing the phenomenological
theory of barium titanate represented a significant advance in the study of
the physical properties of piezoelectric crystals and ceramics based on the
perovskite-structured ferroelectric oxides. Although various theories of fer-
roelectricity in BaTiO3 had been put forward by other workers during the
period, the phenomenological theory as formulated by Devonshire provided
the first unified description of its changes in crystal form and of the variations
of the thermal, dielectric, and piezoelectric properties observed on cooling
through the sequence of cubic → tetragonal → orthorhombic → rhombohedral
ferroelectric phase transitions. By taking account of the multicomponent na-
ture of the ferroelectric polarization in the perovskite oxides, the Devonshire
theory allowed the full matrix of tensor components describing the properties
of the monodomain single-crystalline state to be assembled, even within the
constraints of a comparatively limited set of available experimental data. The
influence of the Devonshire theory on contemporary workers in the field is ev-
ident in the early investigations of Slater [3] into the origins of ferroelectricity
in BaTiO3 , in that of Kittel [4] who predicted the possibility of antiferroelec-
tricity in crystals of the perovskite type before the effect had been experimen-
tally verified, and in that of Cross [5] in his treatment of ferroelectricity and
antiferroelectricity in NaNbO3 −KNbO3 and related perovskite systems.
Though it has been more than 50 years since the Devonshire papers
appeared, phenomenological theories continue to play an important role in
connection with the study of piezoelectric materials, and there are several rea-
sons why this has been the case. Nearly all of the materials of technological
importance for piezoelectric transducer applications are based on perovskite-
structured crystals. The particular form of the phenomenological theory as put
forward by Devonshire for BaTiO3 has therefore naturally been extended in a
more or less straightforward way to the study of these other perovskite materi-
als as well. The phenomenological theory is a useful tool for experimentalists
494 G.A. Rossetti, Jr.
22.2 Phenomenology
The phenomenological theory as applied by Devonshire to barium titanate
crystals is a particular form of the Ginzburg–Landau theory of phase transi-
tions. The use of the Devonshire theory in describing the thermal, dielectric,
and piezoelectric properties of BaTiO3 and related perovskite materials has
been described by many authors, and is treated in detail, e.g., in the text of
Jona and Shirane [6]. More general considerations associated with the foun-
dations of the phenomenological theory and its application to a variety of
other physical phenomena as observed in ferroelectric crystals are elaborated
in considerable depth in the monograph of Strukov and Levanyuk [7] and the
treatment given by them will be followed here. However, here, attention will
be restricted to the specific case of the ferroelectric perovskites and focused
only on those aspects that are especially relevant to describing the pressure–
temperature or composition–temperature phase diagrams of these materials
as the essential starting point for enabling an understanding of the physical
properties that they may display.
For this purpose, it is sufficient to begin by asserting without further justifi-
cation that the behavior of a ferroelectric crystal in the vicinity of a structural
phase transition may be described by some thermodynamic potential function
Φ = Φ(θ, p, η), where Φ is the scalar quantity of energy density (energy per
unit volume), θ is the absolute temperature, and p is the pressure. The vari-
able η represents the order parameter of the structural phase transition(s).
In the high-symmetry or prototype phase η = 0, and in the lower-symmetry
phase or phase(s) η = 0. The thermodynamic potential describes both the
prototype phase and the lower-symmetry phase(s) derived from it and so the
form of the dependence Φ on η is restricted by the condition that Φ must
be invariant with respect to any transformations of the high-symmetry phase.
Specifically, this requires that
Φ(η) = Φ(η ∗ ), (22.1)
where η ∗ is the order parameter transformed by the symmetry operations of
the prototype phase. Close to the temperature θc at which the phase transi-
tion occurs, η is vanishingly small and the thermodynamic potential may be
expressed as a series expansion in powers of the order parameter η:
1
Φ(η) = Φ(0) + Φ η + Φ η 2 + · · · , (22.2)
2
where Φ(η) depends on the external parameters of temperature θ and
pressure p. Consequently, the exact form of the thermodynamic potential
496 G.A. Rossetti, Jr.
where at the pressure p = pc the coefficients A(θ), B(θ), and C(θ) in the
vicinity of θc are expressed as series expansions in powers of (θ − θc ):
1
A(θ) = A(θc ) + A (θ − θc ) + A (θ − θc )2 + · · · ,
2
1
B(θ) = B(θc ) + B (θ − θc ) + B (θ − θc )2 + · · · , (22.4)
2
1
C(θ) = C(θc ) + C (θ − θc ) + C (θ − θc )2 + · · · .
2
For the situation where B(θ) > 0 and C(θ) 0, the transition I ↔ II then oc-
curs as A(θ) passes continuously through zero. In the lowest order approxi-
mation, (22.4) gives A > 0, B(θc ) > 0, C(θc ) 0, and A(θc ) = B = B =
C = C = 0. The thermodynamic potential (22.3) becomes simply
1 1 1
Φ(θ, η) = Φ0 (θ) + α(θ − θc )η 2 + βη 4 + γη 6 , (22.5)
2 4 6
where α = 2A , β = 4B(θc ), and γ = 6C(θc ). It follows that the condition
A(pc , θc ) = 0 defines the line of the second-order transitions in the p−θ plane
between phases I and II and so describes the pressure–temperature phase di-
agram of the crystal. The transition between phases I and II of course need
not be of second order. First-order transitions are described by (22.3) if for
the coefficient of quartic term we have B(p, θ) < 0 and for the sextic term
C(p, θ) > 0. Alternatively, it may happen that a line of first-order transitions
crosses over to a line of second-order transitions. Such a situation occurs, for
example, in crystals placed under hydrostatic pressure, or in some instances,
when the chemical composition of the host crystal is altered by ion substitu-
tion. The point at which this crossover occurs is termed a tricritical point and
22 Thermodynamic Theory 497
corresponds the case where the coefficient of the quartic term changes sign,
viz., when B(p, θ) = 0.
Application of the equilibrium condition to (22.5) requires that
Φ(θ) = Φ(θ, η0 (θ)) (22.6)
with the equilibrium value of the order parameter ηeq = η0 being deter-
mined from
∂Φ ∂2Φ
= 0, > 0. (22.7)
∂η ∂η 2
Generalizing from the above considerations, the thermodynamic potential
for a crystal undergoing a structural phase transition may be written in the
general form
Φ(p, θ, . . . , N, η1 , . . . , ηn ) = Φ0 (p, θ, N ) + Φ1 (p, θ, . . . , N, η1 , . . . , ηn ), (22.8)
where N is some additional external parameter other than pressure or tem-
perature, ηi are variables representing the order parameters of the system,
and the function Φ0 (p, θ, N ) represents that part of the free energy den-
sity that does not depend on these variables. The form of the function of
Φ1 (p, θ, . . . , N, η1 , . . . , ηn ) is determined by the transformation properties of
the order parameter subject to the symmetry operations of the prototype
phase and by the temperature, pressure, or other dependencies of its resulting
nonzero coefficients. Minimizing Φ(p, θ, . . . , N, η1 , . . . , ηn ) with respect to ηi
∂Φ ∂Φ
= 0, . . . , =0 (22.9)
∂η1 ∂ηn
gives a set of values η0 (p, θ, . . . , N ) which when substituted into (22.8) gives
the equilibrium thermodynamic function:
In the case of the perovskite crystals, the cubic prototype phase belongs
to the centrosymmetric symmetry group m3m. As such, it is paraelectric and
nonpiezoelectric. The order parameter η for the transition from the cubic pro-
totype phase into one or more lower-symmetry ferroelectric phases is identified
with the polarization vector P and is related to the amplitude of a transverse
optic soft mode that condenses at θc . The Curie principle asserts that the
“proper” ferroelectric phase(s) that may appear in the crystal are those that
are subgroups of the prototype phase and so must contain all symmetry el-
ements common to the symmetry group of the prototype and the symmetry
group of the order parameter. The order parameter P is a polar vector quan-
tity and its symmetry group is ∞m. Shuvalov [8] has determined all such
possible ferroelectric phases and these are listed in Table 22.1. As shown in
the table, the cubic m3m prototype phase of the perovskite crystals admits
seven proper lower-symmetry ferroelectric phases: one tetragonal (4mm), one
rhombohedral (3m), one orthorhombic (mm2), three monoclinic (m), and one
triclinic (1).
498 G.A. Rossetti, Jr.
Table 22.1. Possible proper ferroelectric phase transitions from symmetry group
m3m (adapted from [8])
ηi , ηi ηj , ηi ηj ηk , ηi ηj ηk ηl , ηi ηj ηk ηl ηm , . . . , (22.11)
to be the exception rather than the rule. For example, in the perovskite
compounds KNbO3 , BaTiO3 , and PbTiO3 , all of which exhibit the transi-
tion m3m → 4mm, in each case the transition is found to be first order [6].
To illustrate how the parameters of the thermodynamic potential may be
determined for some particular crystal of interest, we now consider the simple
case of the single first-order transition in PbTiO3 that occurs, at atmospheric
pressure, at a temperature ≈760 K [10]. In analyzing the experimental data,
it is convenient to define the order parameter η = nη, where n is a unit
vector in the direction of the polarization and η is its absolute value. In
the tetragonal 4mm phase, the polar vector appears along one of the cube
axes of the prototype phase with n = {0, 0, 1}. At atmospheric pressure,
thermodynamic potential then assumes the form
1 1! "
Φ(θ, η, n) = Φ0 (θ) + α(θ − θc )η 2 + β + β (n41 + n42 + n43 ) η 4
2 4 (22.16)
1! "
+ γ + γ (n61 + n62 + n63 ) + γ (n21 n22 n23 ) η 6 .
6
Because there is only one transition corresponding to the spontaneous polar-
ization vector directed parallel to the directions 001, the basic features of
the first-order phase transition will be captured even when the anisotropic
contributions of the sextic invariants in (22.16) are neglected. Making this
approximation, the expansion of the thermodynamic potential begins with
1 1 ! " 1
Φ(θ, η, n) = Φ0 (θ) + α(θ − θc )η 2 + b0 1 − λ(n41 + n42 + n43 ) η 4 + γη 6 ,
2 4 6
(22.17)
where the constants α, b0 , γ > 0. Substituting n = {0, 0, 1}, the entire
quartic coefficient is β0 = b0 (1 − λ) and it is evident that the transition is first
order for λ > 1. From (22.13),
# 1/2 $
β0 4α(θ − θc )γ
η0 = −
2
1+ 1− . (22.18)
2γ β02
From this, it is seen that the highest temperature to which the ferroelectric
phase may exist as a metastable phase in the absence of an applied external
field is given by
β2
θ ∗ = θc + 0 , (22.19)
4αγ
Differentiating (22.17) with respect to temperature, the entropy and specific
heat at constant pressure are
∂Φ ∂σ
σ=− , Cp = θ . (22.20)
∂θ p ∂θ p
Following from (22.17) and (22.20), the temperature dependence of the “ex-
cess” entropy ∆σ of the phase change is
22 Thermodynamic Theory 501
1
∆σ = σ − σ 0 = − αη02 (22.21)
2
and the temperature dependence of the “excess” specific heat ∆Cp of the
phase change is
−2 −2
∆Cp Cp − C 0 4[β02 − 4α(θ0 − θc )] 16γ
= = + 3 (θ0 − θ), (22.22)
θ θ α4 α
where θ0 is the temperature at which the cubic and tetragonal phases are in
equilibrium and is defined by the condition Φ(η = 0) = Φ(η = η0 ).
Under mechanically free conditions, the strain associated with the distor-
tion of the cubic crystal lattice is of the same order of smallness as the square
of the order parameter, Sij ∝ η 2 with
where the fourth-rank tensor qijkl is symmetrical with respect to any permu-
tations i ↔ j, k ↔ l, and (ij) ↔ (kl). Since n = {0, 0, 1}, the relation
between strain and polarization in the tetragonal phase is then
⎛ ⎞
q12 0 0
Sij = ⎝ 0 q12 0 ⎠ η 2 . (22.24)
0 0 q11
From (22.24), the crystal lattice parameters ct and at of the tetragonal ferro-
electric phase are given by
a1 = a2 = ac (1 + S11 ) = ac (1 + q12 η 2 ) = at ,
(22.25)
a3 = ac (1 + S33 ) = ac (1 + q11 η 2 ) = ct ,
Fig. 22.1. Temperature dependence of the excess specific heat, excess entropy, and
order parameter of PbTiO3 at atmospheric pressure (adapted from [11, 12])
Fig. 22.2. Temperature dependence of the crystal lattice parameters (ac of the
cubic phase and at and ct of the tetragonal phase) and the quantity (∆Cp /θ)−2 of
PbTiO3 at atmospheric pressure (adapted from [11, 12])
extends the utility of the analysis of such transitions as occur, e.g., in the
PbZrO3 −PbTiO3 MPB system considered below, where the single proper
ferroelectric transitions m3m → 4mm and m3m → 3m occur over most of
the composition–temperature diagram.
To take account of the effects of pressure on the phase transition, the elastic
energy must be added to the thermodynamic potential. The linear stress-
induced elastic strain is Sij = sijkl Tkl , and for the cubic crystals there are
three nonzero components of the elastic compliance tensor, s11 , s12 , and s44 in
Voigt notation. If the applied pressure is hydrostatic, there are three nonzero
components of the stress tensor −T11 = −T22 = −T33 = −p and following
from (22.25), elastic contributions to thermodynamic potential become
3
Φelas (p, η) = − (s11 + 2s12 )p2 − (q11 + 2q12 )η 2 p. (22.26)
2
504 G.A. Rossetti, Jr.
Fig. 22.3. Dependence of the excess entropy on the square of the order parameter
for PbTiO3 at atmospheric pressure (adapted from [11, 12])
that best agree with all known measured properties can be determined. Using
these coefficients, it is then possible to investigate the relative phase stabilities
and physical properties of the crystal for some other particular set of thermal,
mechanical, or electrical boundary conditions of interest.
Thus, one of the main practical advantages of the phenomenological ap-
proach is that when properly chosen and analyzed, a relatively small set of
experimental data can be used to infer in a self-consistent way a great deal
of information about the thermodynamic behavior and physical properties of
the crystal of interest. Viewed from this perspective, a thermodynamic po-
tential function properly constructed from the phenomenological theory then
provides a compact description of the properties of the crystal. In the case
of ferroelectric solid solutions, relative stabilities of the ferroelectric phases
and the transitions among them will of course depend not only on temper-
ature and pressure but also on the composition variable c. Following from
the most general form of the thermodynamic potential given by (22.8), it is
possible to extend the above analysis to describe some of the features of the
temperature–composition (θ − c) phase diagram. The application of such an
analysis to the specific case of the PbZrO3 −PbTiO3 solid solution is given
in Sect. 22.3 and some thermodynamic considerations that arise in connection
with their piezoelectric properties are discussed.
Fig. 22.6. Discontinuity in the unit cell volume Ω(θc ) and tetragonal lattice distor-
tion δ(θc ) as a function of composition in the PbZrO3 :PbTiO3 system at atmospheric
pressure (adapted from [23])
As shown in the figure, the vanishing values of Ω(θc ) and δ(θc ) indicate
the existence of two tricritical points: one located at c∗1 ≈ 0.22 and another
located at c∗2 ≈ 0.55. The possible existence of such points in the c−θ diagram
of ferroelectric solutions has important consequences that follow directly from
the phenomenological theory. First, as discussed earlier, a tricritical point cor-
responds to the condition that the quartic term(s) in the thermodynamic po-
tential change their sign(s). Consequently, there is a close connection between
the location of a tricritical point and the topology of the c − θ diagram. Such
a connection has in fact been observed. In the phase diagram correspond-
ing to Fig. 22.6, Eremkin et al. [23] found that the morphotropic boundary
separating tetragonal and rhombohedral phases was not vertical as shown in
Fig. 22.5, but instead showed pronounced temperature dependence with the
boundary “bending” sharply toward more Zr-rich compositions and intersect-
ing the line of paraelectric to ferroelectric transitions at the point (c∗1 , θ∗ )
with θ∗ ≈ 610 K. Second, the domain wall energy on either side of a tricritical
transition is expected to display a different dependence on temperature below
the θc . Salje [28] shows that the domain wall energy at θ < θc for a tricritical
transition increases linearly with (θ − θc ), while in a second-order transition
the domain wall energy increases as |(θ − θc )|3/2 . As a result, the domain wall
energy is always greater for the tricritical transition compared with that for
the second-order transition and so domain wall formation in the latter case is
much more strongly favored. On changing composition in a narrow region in
22 Thermodynamic Theory 509
dmkl = ε0 χθ,T
mj qijkl ηi , (22.30)
literature for data collected near room temperature give values that range
from essentially 0 to as much as 15 mole % [32, 33].
The origins of phase coexistence in morphotropic phase systems have long
been controversial and Isupov [34] has collected much the literature on this
topic. Based on considerations of the thermodynamics of mixing in the ap-
proximation of the theory of regular solutions, it is possible, however, to easily
establish some limiting cases. Following from (22.8), the thermodynamic po-
tential can be written as
where the first term on the right-hand side of the equation represents the
free energy of mixing and the second term is the Landau potential. Adopt-
ing the conventional regular solution model for the mixing free energy and
using the form (22.17) for the η-dependent part of the thermodynamic poten-
tial gives
1
Φ(c, θ, η, n) = W c(1 − c) + kb θ [c ln c + (1 − c) ln(1 − c)]
2
1 1
+ α(c)[θ − θc (c)]η 2 + (c − cm ) (22.32)
2 4
! " 1
×b0 1 − λ(n41 + n42 + n43 ) η 4 + γη 6 ,
6
where kb is a Boltzmann’s constant, W is the atomic mixing parameter,
and cm = 0.5 is the nominal composition of the diffusionless tetragonal–
rhombohedral MPB. Based on measurements of the enthalpies of formation,
Rane et al. [35] have shown that the regular solution model is approximately
obeyed for PbZrO3 −PbTiO3 and that the system displays a positive enthalpy
of mixing, W > 0. To make some computations using (22.32), we may take
b0 , γ, λ > 0 to be constants, independent of c, and give α(c) and θ(c) lin-
ear dependencies between the known values for PbZrO3 and √ PbTiO
√ 3 .√Noting
that n = {0, 0, 1} for the tetragonal phase and n = {1/ 3, 1/ 3, 1/ 3} for
the rhombohedral phase and applying the minimum principle (22.9) gives the
equilibrium phase diagram for different values of W [36]. The results of such
an analysis are shown in Fig. 22.8.
The diagram in the panel at left was computed by finding the locus of
points in the c − θ plane where the free energy density curves of the tetrag-
onal and rhombohedral phases are equal. This diagram represents the diffu-
sionless limit. The remaining two diagrams were computed in the conventional
way by finding the points of common tangency on the free energy density vs.
composition curves. The diagram at center represents the situation where the
PbZrO3 −PbTiO3 system behaves as an ideal solution (W = 0) and the dia-
gram at right represents the situation where the systems behaves as regular
solution with a positive enthalpy of mixing (W > 0) constrained to be suffi-
ciently small such that no decomposition may occur in the cubic phase. If dif-
fusional processes are operative, the diagrams in Fig. 22.8 indicate that the
512 G.A. Rossetti, Jr.
800
Temperature (K) PC+FT
Pc
700 Pc Pc
PC+FR FT FT
600
FT
500 FR PC+FR
FR FR 0.498 0.5 PC+FT
400 FR+FT FR+FT
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(a) c,mol fraction PbTiO3 (b) c,mol fraction PbTiO3 (c) c,mol fraction PbTiO3
Fig. 22.9. Changes in Madelung energy (∆EM ) and nonelectrostatic energy (∆EN )
contributions to the change in lattice energy in the PbZrO3 :PbTiO3 perovskite sys-
tem as functions of the degree of deviation from the ideally close-packed cubic struc-
ture (∆ac ) (adapted from [27])
22.4 Summary
The phenomenological theory provides a convenient framework within which
the phase transitions and physical properties of ferroelectric crystals can be in-
terpreted. Because the parameters of the theory are directly related to macro-
scopically observed quantities, they can be determined in a self-consistent way
from a small number of experimental measurements. The general observation
514 G.A. Rossetti, Jr.
that for ferroelectric perovskite crystals the range of validity of the theory
away from the Curie temperature is wider than that would be expected ex-
tends its utility in describing the physical properties of these technologically
important materials. Deviations of measured properties from those predicted
by the parameters of theory thus provide the logical starting point for inves-
tigations into more complex phenomena and for inquiries into their possible
microscopic origins.
Formal extension of the theory to include composition as a parameter of
the thermodynamic potential can be used to compute possible topologies
of the composition–temperature phase diagram and to explain the existence
of spatially inhomogeneous states that may be more complex than those asso-
ciated only with the conventional symmetry twin structures. Considerations
predicated on the possible topologies of these diagrams at equilibrium may
provide insights useful in constructing hypothesis-driven experiments, leading
to better control of their piezoelectric and other structure-sensitive proper-
ties. Knowledge of the mixing properties of such systems further provides a
possible empirical means to correlate the topologies of such phase diagrams
with crystallochemical parameters.
References
1. A.F. Devonshire, Philos. Mag. 40, 1040 (1949)
2. A.F. Devonshire, Philos. Mag. 42, 1065 (1951)
3. J.C. Slater, Phys. Rev. 78, 748 (1950)
4. C. Kittel, Phys. Rev. 82, 729 (1951)
5. L.E. Cross, Philos. Mag. Ser. 8, 1, 76 (1956)
6. F. Jona, G. Shirane, Ferroelectric Crystals (Pergamon, New York, 1962)
7. B.A. Strukov, A.P. Levanyuk, Ferroelectric Phenomena in Crystals (Springer,
Berlin, 1998)
8. L.A. Shuvalov, J. Phys. Soc. Jpn 28, 38 (1970)
9. D. Vanderbilt, M.H. Cohen, Phys. Rev. B 63, 094108 (2001)
10. K. Wójcik, Ferroelectrics 99, 5 (1989)
11. G.A. Rossetti Jr., J.P. Cline, A. Navrotsky, J. Mater. Res. 11, 3179 (1998)
12. G.A. Rossetti Jr., N. Maffei, J. Phys.: Condens. Matter 17, 3953 (2005)
13. G.A. Samara, Ferroelectrics 2, 277 (1971)
14. R. Ramı́rez, M.F. Lapeña, J.A. Gonzalo, Phys. Rev. B 42, 2604 (1990)
15. A. Sani, M. Hanfland, D. Levy, J. Solid State Chem. 167, 446 (2002)
16. B. Jaffe, W.R. Cook, H. Jaffe, Piezoelectric Ceramics (Academic, London, 1971)
17. A.M. Glazer, P.A. Thomas, K.Z. Baba-Kishi, G.K.H. Pang, C.W. Tai, Phys.
Rev. B 70, 184123 (2004)
18. B. Noheda, D.E. Cox, G. Shirane, J.A. Gonzalo, L.E. Cross, S.-E. Park, Appl.
Phys. Lett. 74, 2059 (1999)
19. D.I. Woodward, J. Knudsen, I.M. Reaney, Phys. Rev. B 72, 104110 (2005)
20. B. Noheda, Curr. Opin. Solid State Mater. Sci. 6, 27 (2002)
21. A.J. Bell, J. Mater. Sci. 41, 13 (2006)
22. R.W. Whatmore, R. Clarke, A.M. Glazer, J. Phys. C: Solid State Phys. 11,
3089 (1978)
22 Thermodynamic Theory 515
23. V.V. Eremkin, V.G. Smotrakov, E.G. Fesenko, Ferroelectrics 110, 137 (1990)
24. B. Noheda, N. Cereceda, T. Iglesias, G. Lifante, J.A. Gonzalo, H.T. Chen, Y.L.
Wang, Phys. Rev. B 51, 16388 (1995)
25. S.K. Mishra, A.P. Singh, D. Pandey, Philos. Mag. B 76, 213 (1997)
26. G.A. Rossetti Jr., A. Navrotsky, J. Solid State Chem. 144, 188 (1999)
27. G.A. Rossetti Jr., J.P. Cline, A. Navrotsky, Y.-M. Chiang, J. Phys.: Condens.
Matter 14, 8131 (2002)
28. E.K.H. Salje, Phase Transitions 34, 25 (1991)
29. A. Amin, M.J. Haun, B. Badger, H. McKinstry, L.E. Cross, Ferroelectrics 65,
107 (1985)
30. M.J. Haun, Z.Q. Zhuang, E. Furman, S.J. Jang, L.E. Cross, Ferroelectrics 65,
45 (1989)
31. X.L. Zhang, Z.X. Chen, L.E. Cross, W.A. Schluze, J. Mater. Sci. 18, 968 (1983)
32. B. Noheda, D.E. Cox, G. Shirane, R. Guo, B. Jones, L.E. Cross, Phys. Rev. B
63, 014103 (2000)
33. T. Yamazaki, A. Onodera, H. Fujishita, Y. Ishikawa, M. Tanaka, Ferroelectrics
266, 139 (2002)
34. V.A. Isupov, Ferroelectrics 266, 91 (2002)
35. M.V. Rane, A. Navrotsky, G.A. Rossetti Jr., J. Solid State Chem. 161, 402
(2001)
36. G.A. Rossetti Jr., W. Zhang, A.G. Khachaturyan, Appl. Phys. Lett. 88, 072912
(2006)
37. E. Takayama-Muromachi, A. Navrotsky, J. Solid State Chem. 72, 244 (1988)
38. P.K. Davies, A. Navrotsky, J. Solid State Chem. 46, 1 (1983)
23
Effective Medium Theories
W. Kreher
23.1 Introduction
Predicting macroscopic properties of materials involves two tasks:
1. Projection of macroscopic loading conditions onto the microstructural
level with sufficient and adjusted details (resulting in fluctuating inter-
nal fields)
2. Averaging the local reaction of material to obtain the macroscopic re-
sponse to the loadings
For piezoelectric/ferroelectric polycrystalline materials, the relevant micro-
structural level is presented by special domain configurations within grains
which have different orientations. The local material response comprises the
elastic, dielectric, and intrinsic piezoelectric effect as well as the restructuring
of domain configuration by movement of domain walls or complete reorienta-
tion of the spontaneous polarization, i.e., domain switching. Other processes
as moving free charges and/or defect dipoles are normally not considered in
present micro–macro models. They appear in the form of already averaged
local properties, e.g., as an effective frictional force or activation energy for
domain wall movement.
Obviously, exact three-dimensional solutions for the internal fields and
processes are impossible, so several simplifications are introduced in the mod-
els depending on the aim (i.e., whether linear, nonlinear, or complete repolar-
ization behavior is considered). Typical approximations are:
• Simplification of domain configuration either by a particular deterministic
domain arrangement (e.g., stack of plate-like domains) or by assumption
of a set of noninteracting representative domains
• Simplification of the interactions between neighboring grains (e.g., assump-
tion of homogeneous fields, smoothing by the effective medium approach)
The present chapter describes these modeling approaches in some detail,
putting emphasis on partially exact solutions and practical applicability.
518 W. Kreher
Section 23.2 describes the relevant modeling tools, namely the exact solutions
for a lamellar domain stack and for an ellipsoidal inclusion in a homogeneous
matrix. On this basis, the effective medium approach for a polycrystalline
solid is introduced. Section 23.3 extends the theory to the extrinsic properties
which are determined by the (reversible) movement of domain walls. Finally,
Sect. 23.4 gives a short overview about modeling the nonlinear behavior which
is connected with extensive domain reconfiguration.
To represent the theoretical background in a compact and comprehensible
form, it is necessary to introduce a short form notation for the piezoelec-
tric material laws. These so-called constitutive equations, relating mechanical
stress and electric displacement to strain and electric field, are expressed by
pq Sq − elp El + Tp ,
Tp = cE E
p, q = 1, 2, . . . , 6, k, l = 1, 2, 3. (23.1)
Dk = ekq Sq + εSkl El + DkS ,
(We make use of the abbreviated Voigt notation and the Einstein summa-
tion convention.) Here, TpE is the stress state which arises if a solid, which
is clamped (Sq = 0) and short circuited (El = 0), undergoes a ferroelec-
tric/ferroelastic transformation, and DkS is the electric displacement observed
under the same conditions. Of course, these remnant properties may be ex-
pressed by the corresponding quantities under stress-free conditions (i.e.,
transformation strain and spontaneous polarization); yet the representation
adopted here is especially suited for the derivation of material properties
within the effective medium approximation.
Though we do not explicitly consider pyroelectric and thermoelastic ef-
fects, it should be mentioned that these effects are comprised by T E and
DS , too. Thus, any theoretical result shown below can be utilized to make
predictions also for these thermal properties.
Combining mechanical and electrical field quantities into 9-dim vectors
gives the compact notation:
with
Tp cE t
pq epl Sq TpE
j= , L= , f= , js = . (23.3)
Dk ekq −εSkl −El DkS
Here, the superscript t means the transposition of the 3×6 matrix of piezoelec-
tric constants. The matrix L represents the complete set of the linear elastic,
piezoelectric, and dielectric constants; it is a symmetric 9 × 9 matrix, i.e.,
L = Lt . (23.4)
These equations are valid for each lamella, i.e., the field components appearing
in (23.7) and (23.8) must be homogeneous throughout the stack.
Now, we consider an individual lamella. Then, (23.7) and (23.8) together
with the local material law (23.2) form nine conditions for nine unknowns
(f1 , . . . , f9 ). However, since we only know f¯, which is prescribed, we do not
have the field components j̄3 , j̄4 , j̄5 , j̄9 necessary to utilize (23.8). The missing
four additional global equations are obtained if the average field condition is
taken into account:
fi = f¯i , i = 3, 4, 5, 9. (23.9)
Here, · means averaging over all lamellas. Of course, (23.9) is valid also for
i = 1, 2, 6, 7, 8, but since these field components do not fluctuate within
the stack (cf. (23.7)), the equations are trivially fulfilled.
Thus, (23.7)–(23.9) represent a complete set of equations to determine
each lamella field, which can be solved analytically. Due to the linearity of the
problem, the stack behaves like a homogeneous material described by effective
material constants:
j̄ = j = Lstack f¯ + j s,stack . (23.10)
The effective constants Lstack and j s,stack depend on the properties of domains
and their respective volume fraction. Typically, the lamella properties do not
fluctuate arbitrarily; instead, one has a limited number of domain types (in-
clined at fixed angles to the stack orientation). Then, the internal fields are
23 Effective Medium Theories 521
identical for domains belonging to a certain type, and the averaging is done by
means of volume fractions of domain types. In Sect. 23.3, a domain stack con-
sisting of two types of domains forming 90◦ domain walls will be considered.
where I is the 9 × 9 unit matrix. The Eshelby matrix in (23.12) and (23.13)
depends on the piezoelectric properties of the matrix and on the geometry of
the inclusion, i.e.,
522 W. Kreher
1.0
0.8
99
0.6
Eshelby tensor Sij
91
77
57
0.4
0.2
93
39
0.0
19
75
–0.2
0.01 0.10 1.00 10.00 100.00
Aspect ratio a3 /(a1=a2)
1.0
33
0.8
11
66
0.6
Eshelby tensor Sij
44
0.4
0.2 13
12
0.0
31
–0.2
0.01 0.10 1.00 10.00 100.00
Aspect ratio a3 /(a1=a2)
These properties can be fed into (23.17) and (23.18) in place of L and j s ,
respectively. Together with matrix properties supposed to be known, (23.12)
yields f which is now the average inclusion field. It is not necessary to deter-
mine the residual stress and electric field outside the inclusion since rigorous
expressions are available which relate the potential energy of the solid to
the field inside the inclusion (see [16, 17]). Here, potential energy means the
internal field energy including the potential of the loading forces. For exam-
ple, under isothermal conditions and for prescribed strain and electric field,
the potential energy is a combination of free elastic energy and free electric
enthalpy.
Figure 23.2 shows a typical result demonstrating the equilibrium position
of domain walls. Without an external field, both types of domains have equal
volume fractions, whereas under the action of the electric field the favorably
oriented domains grow on the expense of the other domains. Such calculations
take into account the complete mechanical and electrical interaction fields
(except for the small end regions of the domains). They can be performed for
any orientation of the grain and for any properties of the surrounding medium
provided the matrix S of Eshelby tensors is known.
2000
–5
1500
0
1
3
1000
5
500
0.0 0.2 0.4 0.6 0.8 1.0
Local domain volume fraction
Fig. 23.2. Potential energy wp per volume of a specific inclusion as a function of the
volume fraction ξ of the nonfavorably oriented domain. The inclusion is embedded in
a matrix which has isotropic properties (i.e., a nonpoled material). The far field f 0
prescribes a varying electric field under the condition of a constant strain (after [21])
526 W. Kreher
The results according to (23.19) may be used within the EMA, but now allow-
ing for small reversible shifts of domain walls. Here, all grains are described
by a lamellar stack of 90◦ domain walls where only the orientation Ω and the
domain volume fraction ξ vary. The domain volume fraction is determined by
the minimum of the potential energy as described in Sect. 23.3.1. Of course,
the local equilibrium value depends on the applied field and on the local
orientation:
ξ = ξ(f¯, Ω). (23.20)
This together with (23.19) gives the grain properties which in turn can be
fed into the effective medium formalism (23.17) and (23.18). The resulting
effective properties now become dependent on the applied field. Though in
general this relationship is nonlinear, it can be linearized for small fields. To
this end, we consider the incremental form of (23.15):
In first order, the increments dL∗ and dj ∗ will be proportional to the applied
field increment df¯. Therefore, we obtain
where the new material property matrix L̂∗ is the sum of the already known
effective intrinsic properties and the effective extrinsic contributions due to
the variations dL∗ and dj ∗ caused by the moving domain walls:
300
d15 ∗
d15 intr. + extrinsic
Piezoelectric constants d ij / 10–12 m / v
250
∗
200 d33 d33 intr. + extrinsic
150 3
∗
d15 intrinsic
100
∗
d33 intrinsic
50
qp
0
∗
d31 intrinsic
–50
d31 ∗
–100 d31 intr. + extrinsic
Fig. 23.3. Predictions for intrinsic and extrinsic piezoelectric constants vs. the
polarization state for barium titanate (after [22]). The polycrystalline material
consists of spherical grains with an orientation distribution according to the pol-
ing state described by the poling angle θp . Each grain is modeled by a lamellar
stack of 90◦ domains. The piezoelectric constants of a single domain are given by
d31 = −33.4 pm V−1 , d33 = 90.0 pm V−1 , d15 = 540 pm V−1
0.13
0.12
0.10
0.09
0.08
0.0 0.5 1.0
Electric field [kV/mm]
Fig. 23.4. Hysteretic behavior of electric displacement of BaTiO3 vs. applied electric
field (after [22]). Dissipation constant D0 = 100 kJ m−3
d d
wp = wd = ±D0 for ξ˙ >
< 0. (23.25)
dξ dξ
This condition ensures that the dissipated work must be supplied by the
release of potential energy.
If this scheme is invoked into the formalism, one obtains a threshold and
a hysteresis describing the loss properties due to frictional movement of do-
mains. An example is shown in Fig. 23.4. Thus, macroscopic loss measurements
can be correlated to the microscopic dissipation constant D0 .
The basic idea behind the model is to adopt the viscous concept known to
be successfully applicable in modeling polycrystal plasticity. For ferroelectric/
ferroelastic materials, it means that all possible domain orientations may si-
multaneously transform into any other possible orientation. Only the transi-
tion rate differs as a function of the driving force which is given by the gain
of free enthalpy. Thus, one can describe a gradual change of the polarization
state of the individual crystals and avoids the problematic sharp switching
transition, which is not realistic for a multidomain crystal and, additionally,
causes problems in the numerical evaluation. Moreover, it is easily possible to
include rate- and temperature-dependent behavior.
Let us denote the set of internal state variables by ξr with r = 1, 2, . . . , n,
where n is the number of states (one can find models with n = 2, 6, 8, 20, 42, . . .).
It is meaningful to restrict the variation range of ξ in the sense of weighting
factors (so an interpretation as volume fractions of certain special domains is
possible though not necessary for the model):
n
ξr = 1, 0 ξr 1. (23.26)
r=1
which the transition rates become large (in principal, these parameters may
depend on the transition type r–s). Furthermore, two empirical constants m
and α must be adjusted to experimental data. The second power term takes
into account the saturation of the transition if the weight of the starting state
becomes zero.
It is important to note that (23.30) is a unique function of state, so that in
numerical simulations no iteration is necessary when a certain load increment
is prescribed. Instead, one directly obtains the new domain configuration of
internal state variables by summing all transition rates:
n
ξ˙r = (+vsr − vrs ). (23.31)
s=1
4.0
0.01 Hz
3.5
3.0
0.1 Hz
STRAIN (x10–3)
2.5 1.0 Hz
2.0
1.5
1.0
0.5
0.0
–2 –1 0 1 2
E (MV/m)
4.0 0.01 Hz
0.1 Hz
3.5
3.0
1.0 Hz
STRAIN (x10–3)
2.5
2.0
1.5 1
1.0
0.5
0.0
–2 –1 0 1 2
E (MV/m)
Fig. 23.5. Simulated strain hysteresis curves under sinusoidal loading for soft PZT
(after [4]). Model parameters: n = 6, m = 8, α = 3, wc0 = 0. Comparison with
experimental results after [24]
532 W. Kreher
What we still need is an expression for the potential energy. Here, the most
simple assumption is provided by a linear combination of the potential energies
(r)
wp of certain domains, where r stands now for a particularly orientated
domain:
n
wp (f¯, ξ1 , . . . , ξn ) = pξr w(r) (f¯) (23.32)
r=1
23.5 Summary
Effective medium theories may be used to calculate internal fields on different
structural levels (domains, grains) so that the linear and non-linear material
behavior of piezoelectric materials can be modeled. Some applications have
been presented in this chapter and more examples may be found in the liter-
ature. Below some open problems and possible applications are listed:
• Pyroelectric/thermoelastic properties
• Evaluation for time-dependent properties
• Conduction problems
• Effect of free charges (screening of polarizations, modification of interac-
tion energy)
• Phase mixtures
• Relaxor behavior (diffuse domains)
References
1. G. Arlt, J. Mater. Sci. 25, 2655 (1990)
2. G. Arlt, Ferroelectrics 189, 91 (1996)
3. A.T. Bartic, D.J. Wouters, H.E. Maes, J.T. Rickes, R.M. Waser, J. Appl. Phys.
89(6), 3420 (2001)
4. A.Y. Belov, W.S. Kreher, J. Europ. Ceram. Soc. 25(12), 2567 (2005)
5. M.L. Dunn, M. Taya, J. Am. Ceram. Soc. 76(7), 1697 (1993)
6. M.L. Dunn, M. Taya, Proc. R. Soc. Lond. A 443, 265 (1993)
7. D.A. Hall, J. Mater. Sci. 36(19), 4575 (2001)
8. J.H. Huang, Int. J. Solids Struct. 33(20), 1439 (1996)
9. J.H. Huang, J.S. Yu, Compos. Eng. 4, 1169 (1994)
10. J.E. Huber, N.A. Fleck, J. Mech. Phys. Solids 49(4), 785 (2001)
11. J.E. Huber, N.A. Fleck, C.M. Landis, R.M. McMeeking, J. Mech. Phys. Solids
47(8), 1663 (1999)
23 Effective Medium Theories 533
T. Steinkopff
Nowadays, the most used commercial Finite Element software for solving
piezoelectric problems is based on the linear description of materials behav-
iour. The developers shrink from implementing the complex nonlinear be-
haviour which is essentially characterised by the ferroelectric and ferroelastic
hystereses. At the other end, the user which is interested in nonlinear analyses
is forced to program so-called user elements or to apply iterative algorithms.
The two lead to disproportionately excessive costs.
With the FE software COMSOL an innovative approach is given to take
the nonlinear behaviour into the numerical consideration. Examples of both
linear and nonlinear piezoelectric analyses will also be given here.
S = sE · T + d · E = (cE )−1 · T + d · E,
∼ P,
D = d · T + εT · E =
where cE are elastic coefficients (for constant electric field), εT are dielectric
coefficients (for constant mechanical stress), and d are piezoelectric coefficients
(so-called charge constants).
Generally, the coefficients are functions of the loading conditions, loading
direction, loading rate, and frequency:
E T ∂T ∂E
c , d, ε = f T, E, , ,... .
∂t ∂t
In cases of transient analyses, the variable t corresponds to the real time. In
cases of static analyses, it should be understood as pseudo-time, first of all to
describe the sequence and direction of load changes ∆E and ∆T .
24 Finite-Element Modelling of Piezoelectric Actuators 537
1,6
Polarisation / a.u.
1,5
∆P
1,4
∆E
Fig. 24.3. Dependence of coercive pressure on the parallel applied electric field [1]
Fig. 24.4. Strain asymmetry between tension and compression for a ferroelastic
material without grain-to-grain interaction. The total strain dependence on uniaxial
stress is shown. Experimental data from [1]
In the recent few years, a lot of micromechanical models have been de-
veloped. Models without grain-to-grain interaction are worth to estimate the
maximum values of strain and polarisation attainable by switching. Apply-
ing the ferroelectric considerations of [2] onto the ferroelastic behaviour, the
strain asymmetry between tension and compression of originally untextured
(unpoled) ceramics can be shown (Fig. 24.4). In the case of uniaxial tension
√ −1
and compression,
maximum ferroelastic strain values of 4π · 2 · Sspont and
√ −1
−4π · 1 − 2 · Sspont can be reached, respectively.
Obviously, these maximum values for both polarisation and strain cannot
be reached because of grain-to-grain, domain-to-domain interaction, and other
deformation mechanisms. Finite-element methods recommend by their selves
to take electromechanical interactions into account. Figure 24.5 shows results
of finite-element simulations based on ferroelastic domain switching [3] which
are compared to uniaxial data derived from bending experiments [4]. Because
of mechanical interaction, the strain asymmetry occurs not until stress values
of 40 MPa for this case of a soft PZT material with low coercive stress.
It is worth to mention that using FE methods for modelling domain switch-
ing effects, complex phenomena like auto-catalysis arise in a natural way [5].
From the point of engineering view, micromechanical models are only useful in
cases of simplified considerations [6]. For more complex simulations of piezo-
electric devices, the materials behaviour must be described phenomenologi-
cally [7]. Analogous to the classic elasto-plastic theory yield and hardening
parameters can be well defined for piezoelectric ceramics according to experi-
mental data [8]. Generalised yield stresses should be related to coercive stresses
and fields. Kinematic hardening is useful to describe both the ferroelectric and
ferroelastic hysteresis.
540 T. Steinkopff
80 From bending
experiment
60
40
20
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Strain / ‰
Electrostatics:
− ρ = 0,
div D (24.4)
with
= D(
D
E(ϕ)),
(24.5)
·
n = σ,
D (24.6)
Π(
a) = 12
aTu · K ·
au +
aTu · F
⇒ K ·
au = F
,
ϕ) = 12
aTϕ · C ·
aϕ +
aTϕ · Q
Π(
⇒ C ·
aϕ = Q.
542 T. Steinkopff
Piezoceramic
layer
(electrically driven) Thickness
direction
Neutral middle
layer
Piezoceramic layer
(short-circuit)
Fig. 24.6. Geometry, loading, and bending deformation of the trimorph actuator.
Colours correspond to the axial stress levels (red for tension, blue for compression
stresses)
trimorphs for which the deformation behaviour has been analysed by means
of finite-element methods (Fig. 24.6).
In the following, the static and dynamic behaviour of trimorph bending actu-
ators consisting of two active layers (top and bottom layer) and an inactive
middle layer (Fa. Argillon) will be discussed. Aspects of reliability are in the
foreground where piezoelectric actuators are used permanently for instance
in textile machines. There, a lot of several thousand bending actuators are
used under bi-directional operation conditions. Long-term investigations for
different driving conditions (Fig. 24.7) [10] show a non-monotonic dependence
of the mean-times-to-failure on the driving voltage: bending actuators driven
with 275 V (magenta) show a higher lifetime than actuators driven with 225
and 250 V (orange, blue). By means of finite-element analyses, this unexpected
behaviour can be understood.
Starting with static analysis, Fig. 24.8 shows typical deformation plots for
both the free elongation and the blocking state of a trimorph actuator carried
out with the FE software ANSYS c
[11]. Note that there is no significant
difference in the axial stress values between the free end (FE) and near to
the clamped region (CR) although failure is observed to be a result of under-
critical crack growth near the clamping region (CR).
A better understanding can be reached by considering the deformation and
stress over time behaviour corresponding to Figs. 24.9 and 24.10. In Fig. 24.9,
snapshots give an idea of the transient deformation behaviour. In textile ap-
plications, the deflection direction alternates with frequencies of about 50 Hz
by changing the piezoelectrically active layer (top or bottom layer). In the left
544 T. Steinkopff
0
without failure
ln ln 1/(1-F)
–1
–2
Weibull parameter @ driving voltage
**
m=0.92, N0=1.88e9 @ 200 V
–3 1.71 9.57e7 225 V
1.84 6.49e7 250 V
1.71 1.36e8 275 V
1.22 9.56e7 300 V
–4 1.85 5.32e7 325 V
1.86 2.17e7 350 V
2.16 5.99e6 400 V
2.05 6.51e6 450 V
–5
Fig. 24.7. Weibull plots for bending actuators which were driven with voltages be-
tween ±200 and ±450 V which corresponds to electric field strength up to 2 kV mm−1
ANSYS ANSYS
Clamping
FE MX
CR
Z
Y X
MN
Deflection
Fig. 24.8. Deflection behaviour of a bending actuator. Left: free actuator. Right:
clamped actuator. Keep in mind the different scaling for deflection
and right column, the transient behaviour of the bending actuator is shown for
the case that the piezoelectric top and bottom layer is driven, respectively. Af-
ter switching (driving changes from the top to the bottom layer at t = 20 ms),
inertial effects have a strong influence on the deformation behaviour and the
internal loading conditions. Dependent on driving frequency, in-phase and
out-of-phase deflection behaviour of the actuator can be observed.
For inappropriate switching moments, high dynamic tension stresses which
are of the order of the static stresses are observed near the clamping region
(Fig. 24.10). As a consequence for different operation conditions (voltage, fre-
quency, strikes yes/no), totally different stress levels and mean-times-to-failure
seem to be clear. Here, finite-element calculations are very helpful to analyse
24 Finite-Element Modelling of Piezoelectric Actuators 545
t [ms] t [ms]
0,2 19,4
1,0 20,2
1,8 21,0
2,6 21,8
3,4 22,6
4,2 23,4
5,0 24,2
5,8 25,0
6,6 25,8
7,4 26,8
8,2 27,4
9,0 28,2
9,8 29,8
10,6
..
.
σx,B(t)
[MPa] 30
20
10
0
x-Position
−10 11 m m
20 m m
−20 47 m m
0 5 10 t [ms]
Fig. 24.10. Signal and stress over time for different x positions in the active layer
of the trimorph actuator
546 T. Steinkopff
In the following, simple examples are given which demonstrates the influ-
ence of nonlinear elastic behaviour on the strain and stress distribution of a
trimorph actuator.
Starting point is an elastic stiffness Y that may depend linearly from stress
according to ferroelastic softening under tension stresses:
T
Y (T ) = Y0 − ∆Y · .
T0
To compare piezoelectric materials with different elastic behaviour, various
scenarios can be considered because of the relationship between the piezoelec-
tric charge coefficient e and voltage coefficient d:
e = Y · d ⇒ e(T ) = Y (T ) · d(T ).
For reasons of simplicity, two cases are considered here: d = constant (case A)
and e = constant (case B) materials. That means in case A, materials are com-
pared which have the same piezoelectric d coefficient but different nonlinear
elastic behaviour. In case B, the same procedure is applied for materials with
the same piezoelectric e coefficient. Figure 24.11 shows the axial stress Txx
and electrical field strength Ey distributions in dependence of the coordinate
in thickness direction (left: case A, right: case B).
In case A where materials with same d coefficient are compared, the influ-
ence of softening is much higher than that in case B (e = constant).
In Fig. 24.12, the influence of the softening parameter Ysoft = ∆Y /T0 on
the deformation behaviour is shown.
While the nonlinear materials behaviour considered here is very sim-
plified, the results give a first impression of its influence on the complex
electromechanical behaviour in bending actuators. For case B materials (e =
constant), the influence of the nonlinear elastic behaviour is much smaller
than that for case A materials (d = constant).
24.6 Conclusions
It could be shown that finite-element analyses are helpful for solving more
complex piezoelectric problems. To get initial estimates about the maximum
stress levels, linear finite-element calculations seem to be sufficient.
In the next years, one key aspect of finite-element activities in the field of
piezoelectric applications will be concentrated on finding practicable formu-
lations of the nonlinear materials behaviour not only in regard to the electro-
mechanical loading conditions but also in regard to history, frequency, and
time. From the author’s point of view, there is no way to find a unified mate-
rials law that can incorporate the influence of all these parameters.
40 20
548
20 Y-soft / Y-soft /
GPa 10 GPa
0 −8 0
−6 4
−5 6
−20 0 0 8
4 10
6
T. Steinkopff
−40 12
8
10
−60 12 −10
2,5 1,5
2,0 Y-soft / 1,0 Y-soft /
1,5 GPa GPa
−8 0,5 0
1,0 −6 4
−5 0,0 6
0,5 0 8
0,0 4 10
6 −0,5
12
−0,5 8
10 −1,0
−1,0 12
Bottom layer Middle layer Top layer −1,5 Bottom layer Middle layer Top layer
−1,5
Electric field strength (Ey) / [kV/mm]
Fig. 24.11. Distributions of axial stress (top) and electric field strength (bottom) for cases A and B materials (left and right, respectively)
with different nonlinear elastic behaviour
24 Finite-Element Modelling of Piezoelectric Actuators 549
1,6 0,9
1,4
1,2 0,8
1,0
0,7
Strain / 1e-3
Strain / 1e-3
0,8
0,6 0,6
0,4
0,5
0,2
0,0
0,4
−0,2
−0,4 0,3
−10 −5 0 5 10 15 −1 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Y-soft / GPa Y-soft / GPa
Fig. 24.12. Deformation behaviour of trimorph actuators with cases A and B ma-
terials behaviour (left and right, respectively). Keep in mind the different scaling for
strain
References
1. A. Schäuffele, K.H. Härdtl, J. Am. Ceram. Soc. 79, 2637 (1996)
2. T. Michelitsch, W.S. Kreher, Acta Mater. 46, 5085 (1998)
3. T. Steinkopff, Ferroelectrics 222, 383 (1999)
4. H. Weitzing, G.A. Schneider, Private communications, 1998
5. S.C. Hwang, J.E. Huber, R.M. McMeeking, N.A. Fleck, J. Appl. Phys. 84,
1530 (1998)
6. M. Kamlah, Continuum Mech. Themodyn. 13, 219 (2001)
7. C.M. Landis, J. Mech. Phys. Solids 50, 127 (2002)
8. T. Fett, D. Munz, Adv. Eng. Mater. 5, 718 (2003)
9. D.L. DeVoe, A.P. Pisano, J. Microelectromech. Syst. 6, 266 (1997)
10. M.P. Denzler, Ph.D. Thesis, Cuvillier, Göttingen, 2004
11. ANSYS, Inc., Southpointe, 275 Technology Drive, Canonsburg, PA 15317
12. COMSOL AB, Tegnérgatan 23, SE-III 40 Stockholm
25
Trends in Ferroelectric/Piezoelectric Ceramics
N. Setter
25.1 Introduction
Ferroelectrics and piezoelectrics are bound to keep their place as important
engineering ceramics. Capacitors and piezoelectric transducers are the most
widely spread applications of these group of materials. In thin-film form, ferro-
electrics and, more widely, polar materials have now been used for several years
in rf devices and in non-volatile memories. Components based on piezoelectric
films are also being developed and entered the market of various micro-sensor
and micro-actuator applications.
Barium titanate (BaTiO3 −BTO) and lead–zirconate–titanate (Pb(Zr, Ti)
O3 −PZT) continue to be the most important ferroelectric and piezoelectric
materials in commercial devices. New materials have been introduced con-
stantly. While during 1970s–1980s, those were the relaxors (with giant elec-
trostriction), the 1990s saw the introduction of relaxor ferroelectrics such
as Pb(Zn1/3 Nb2/3 )O3 –PbTiO3 (PZN–PT) increasing the coupling coefficient
from ≈0.7 of PZT to ≈0.90 and over [1]. In the last decade, AIN became
an important piezoelectric material due to its use in piezoelectric electronic
components (bulk acoustic wave devices) for mobile communications [2].
SrBi2 Ta2 O9 (SBT) competed with PZT in ferroelectric non-volatile memory
(FRAM) applications with a superior fatigue resistance [3]. Barium–strontium
titanate (BST) was investigated for DRAM applications and finally seems to
have found market in tunable ferroelectric capacitors for microwave compo-
nents. In 2004, Toyota published results showing that a lead-free material,
modified (K, Na)NbO3 (KNN) can potentially reach properties of the best
lead-containing ferroelectrics [4].
The processing technology of piezoeoelectrics/ferroelectrics has been ada-
pting itself to new needs too: finer powders became available [5], thick film
technologies for piezoelectrics have been developed [6], and textured growth
(the growth of oriented ceramics) gained importance [7]. Following the ex-
cellent properties of piezoelectric single crystals, a method of transforming
554 N. Setter
piezoelectric ceramics into crystals has been reported with expected improve-
ment in composition homogeneity control [8]. Processing of thin films by both
physical and chemical methods reached a certain level of maturity with a
rather tight control of composition (a challenge in particular in lead-containing
compounds), stress, and orientation [9]. Recently, the production of one-
dimensional (1D) nano-structures of ferroelectric and piezoelectric materials
has been reported [10].
Trends that point to directions of future developments in the field of piezo-
electrics and ferroelectrics are summarized below.
Fig. 25.1. Effect of compressive stress σ3 = 0 (dashed line) and −1.79 GPa at
T = 300 K on anisotropic piezoelectric enhancement in PbTiO3 . At σ3 = 0 Pa,
values for d∗33 (dashed line) are multiplied by 10. For σ3 close to the coercive stress
(approximately −1.9 GPa), values of d∗33 are strongly sensitive to input parameters
and vary between 103 and 104 [21]
The increase in operation frequency for wireless communication and the in-
terest in miniaturization and integration motivated the recent replacement of
many of the discrete surface acoustic wave components by bulk acoustic wave
ones based on piezoelectric thin films. This has been discussed in Chap. 15.
At present, the thin films are mostly AlN. In spite of the limited value of its
coupling coefficient, AlN has advantages of a high sound velocity and ease
of process integration. It is clear that search of new materials will continue:
one possible direction is a further exploration of ZnO. This material ‘lost’ the
competition with AlN due to its sensitivity to defects. Very recently, a spec-
tacular improvement of its piezoelectric d33 coefficient was reported through
appropriate doping [23]. If this will turn to influence also the coupling coeffi-
cient, it may be well an interesting material to pursue. At present the origin
of this improvement is obscure. Another direction awaiting further work is the
development of tunable filters by the use of paraelectric materials under bias
field [24].
25 Trends in Ferroelectric/Piezoelectric Ceramics 557
Fig. 25.2. Penetration of the ionic polarization from the ferroelectric into the metal–
oxide electrode. (The gradient in shading represents the concentration of free charge
carriers, while the solid line represents the absolute value of polarization.) First-
principles investigation explains this way the improved remanent polarization of
ferroelectrics capacitors with oxide electrodes relative to metallic (non-polarizable)
electrodes [38]
all, better insight into behaviour at surfaces and interfaces and behaviour of
systems containing piezoelectrics could result in accelerated development of
piezoelectrics and ferroelectrics.
Fig. 25.3. KNN crystal grown from dense ceramics using template-assisted Solid
State Single Crystal Growth (SSCG). (Courtesy of J. Fisher, A. Bencan and
M. Kosec, Jozef Stefan Institute, Ljubljana)
Increase thickness
Second
Second layer
layer
First layer
Fig. 25.4. Full strain relaxation and dislocation confinement during two-step growth
of STO. The inset is a TEM micrograph of the film cross section showing the dislo-
cation confinement to the first layer [46]
and co-workers obtained BaTiO3 films with Curie temperature of the order of
400–600◦ C (while Tc of BaTiO3 crystals is 120◦ C) which enhanced the room
temperature ferroelectric properties [44]. The properties were similar to those
previously predicted by LGD calculations [45]. (2) In another work, the strain
in epitaxial SrTiO3 (STO) films was controlled by a dislocation relaxation
mechanism using a two-step growth process, where the crystallization of a
pre-deposited very thin amorphous layer of identical composition acted as an
absorption layer for the misfit strain (Fig. 25.4). Two-step grown films had a
25 Trends in Ferroelectric/Piezoelectric Ceramics 561
25.4.3 Self-Assembly
Composite (predicted)
0.4
(measured)
0.2
100µm
Pure
ferroelectric
(a) 0.1
0 100 200 300 400 500
Fig. 25.5. (a) Optical micrograph of a fabricated composite structure. (b) Compar-
ison between the relation between permittivity and tunability in pure ferroelectric
and in parallel-connected composite (calculated and measured) [62]
25 Trends in Ferroelectric/Piezoelectric Ceramics 563
300
150
[nm]
0 200
0 100
Fig. 25.6. 3D visualization of the piezoelectric activity of KNbO3 nano-rod and the
existence of ferroelectric domains with piezoelectric activity [84]
nano-wires in an up-scalable way [74]. BTO and STO nano-tubes were pro-
duced by reacting titania nano-tubes with barium and strontium hydrox-
ides [75]. A hydrothermal growth of potassium niobate (KNO) nano-wires
was demonstrated [76]. Hydrothermal method was used also for the fabrica-
tion of arrays of BTO nano-tubes on titanium substrate [77]. Very little has
been reported so far on the growth mechanism of piezoelectric nano-tubes and
nano-wires: while KNO seems to grow by a nucleation and growth mechanism
in sub-saturated liquid [78].
The ferroelectric to paraelectric transition was recently reported on BTO
nano-wires [79] but the crystalline structure of ferroelectric 1D structures has
been only slightly explored [80,81]. Domain walls formation and disappearance
upon cycling around the ferroelectric phase transition were demonstrated in
PZT nano-wires [82]. Rare are yet the detailed reports on the piezoelectric
properties of nano-tubes and nano-wires but switching of the ferroelectric
polarization and local piezoelectric activity have been detected using force
microscopy (Fig. 25.6) [83, 84]. The field of 1D small piezoelectrics is at its
infancy.
by scanning an AFM tip over the wires and the generated mechanical energy
was converted into electrical energy by the piezoelectric effect. The estimated
electrical energy produced was ≈10 pW µm−2 . The collection of the energy
is done in an elegant way taking advantage of the semiconductor nature of
the nano-wires used: the nano-wire/conductor contact acts as a diode and
the charge generated piezoelectrically is collected in a unidirectional way. The
current leading idea is to convert mechanical energy that is available contin-
uously such as body movement into electrical energy, so that self-powering
can be used for wireless low power consumption devices. Development in this
direction was shown recently by Zhong Lin Wang et al. [88] who created a
textile structure that contains the ZnO nano-wires and metal wires for the
current collection.
An inherent exploitation of nano-technology can be obtained when the phe-
nomena at stake are coupled at the same scale. This was shown recently [89]
in utilization of piezoelectric semiconductors to build cantilevers in which the
width of the depletion region (hence the piezoelectric active region) was ma-
nipulated by an electric field, creating positive or negative deflection according
to the extent of the field applied, which was translated into an increased vibra-
tion amplitude relative to a stationary piezoelectric cantilever. It is a similar
effect to that obtained by bi-layer actuators, but on a nano-meter scale, tun-
able, and processed via standard integrated semiconductor technology, where
the electrodes and the piezoelectric active regions are all based on variation
of the doping in a GaAs cantilever. This mechanism functions only when
the dimension of the cross section of the cantilever is of the same order of
magnitude as that of the depletion region, namely, sub-µm. The authors fur-
ther demonstrate various devices based on piezoelectric GaAs nano-structures
such as nano-electromechanical XOR (exclusive-or) logic gate and resonance-
frequency control.
It is reasonable to expect that piezo-NEMS (nano-electromechanical sys-
tems) will develop further with possible applications in sensors, transducers
and computing applications. The interest of piezoelectric semiconductors is
also likely to intensify due to the small dimensions of the components.
25.7 Conclusion
Major trends in materials and processes in the area of ferroelectrics and piezo-
electrics were outlined above. Emphasized directions were new materials, in
particular lead-free and high-temperature ones as well as materials for com-
munication technologies and MEMS–NEMS. The increased role of ab initio
calculations in prediction of new material systems, properties and functioning
of components and devices was pointed out. Trends in processing include ‘do-
main engineering’, nano-composites, self-assembly of small elements on sub-
strates, and new 1D structures.
566 N. Setter
The question of new applications was not discussed although it is clear that
the entire field is driven by applications. Intensification of use of piezoelectrics
and ferroelectrics can be foreseen in several areas: with the present strong
development of polymer electronics, it is likely that ferroelectric polymers will
grow in importance. Increased investments into medical diagnostic tools may
be projected as an opportunity for a further increased use of piezoelectrics in
this domain. Certainly, highly performing lead-free materials will be welcome
here. Increased automation and a more stringent control on efficient energy use
and on clean environment are indicative for an intensified use of piezoelectric
sensors and actuators for these applications.
Acknowledgement
References
1. S.-E. Park, T.R. Shrout, J. Appl. Phys. 82, 1804 (1997)
2. M.-A. Dubois, P. Muralt, Appl. Phys. Lett. 74, 3032 (1999)
3. J.F. Scott, C. Araujo, Science 246, 1400 (1992)
4. Y. Saito, H. Takao, T. Tani, T. Nonoyama, K. Takatori, T. Homma, T. Nagaya,
M. Nakamura, Nature 432, 84 (2004)
5. Y. Sakabe, Y. Yamashita, H. Yamamoto, J. Eur. Ceram. Soc. 25, 2739 (2005)
6. P. Tran-Huu-Hue, F. Levassort, F. Vander Meulen, J. Holc, M. Kosec,
M. Lethiecq, J. Eur. Ceram. Soc. 21, 1445 (2001)
7. T. Takeuchi, T. Tani, Y. Saito, Jpn. J. Appl. Phys. 1, 9B, 5553 (1999)
8. H.Y. Lee, in Piezoelectric Single Crystals, ed. by S. Trolier-McKinstry,
L.E. Cross, Y. Yamashita p. 160, (2004)
9. S. Trolier-McKinstry, P. Muralt, in Electroceramics Based MEMS, ed. by
N. Setter (Springer, Berlin, 2005), pp. 199–215, Chap. 10
10. J.J. Urban, W.S. Yun, Q. Gu, H. Park, J. Am. Chem. Soc. 124, 1186 (2002)
11. E. Hollenstein, M. Davis, D. Damjanovic, N. Setter, Appl. Phys. Lett. 87,
182905 (2005)
12. G.-Z. Zang, J.-F. Wang, H.-C. Chen, W.-B. Su, C.-M. Wang, P. Qi, B.-Q. Ming,
J. Du, L.-M. Zheng, S. Zhang, T.R. Shrout, Appl. Phys. Lett. 88, 212908 (2006)
13. T.R. Shrout, S.J. Zhang, J. Electroceram. 19, 111 (2007)
14. B. Jaffe, R.S. Roth, S. Marzullo, J. Appl. Phys. 25, 809 (1954)
15. T.R. Shrout, R. Eitel, C. Randall, in Piezoelectric Materials in Devices, ed. by
N. Setter (2002), pp. 413–432
16. N. Klein, D. Damjanovic et al. J. Appl. Phys. 102, 014112 (2007)
17. K. Nakamura, T. Tokiwa, Y. Kawamura, J. Appl. Phys. 91, 9272 (2002)
18. S. Wada, S. Suzuki, T. Noma, T. Suzuki, M. Osada, M. Kakihana, S.-E. Park,
L.E. Cross, T.R. Shrout, Jpn. J. Appl. Phys. 1 38, 5505 (1999)
19. X. Du, U. Belegundu, K. Uchino, Jpn. J. Appl. Phys. 1, 36, 5580 (1997)
25 Trends in Ferroelectric/Piezoelectric Ceramics 567
80. A. Magrez, E. Vasco, J.-W. Seo, C. Dieker, N. Setter, L. Forro, J. Phys. Chem.
B 110, 58 (2006)
81. G. Suyal, E. Colla, R. Gysel, N. Setter, Nanoletters 4, 1339 (2004)
82. J. Wang et al., Appl. Phys. Lett. 90, 133107 (2007)
83. W.S. Yun, J.J. Urban, Q. Gu, H. Park, Nanoletters 2, 447 (2002)
84. G. Suyal, E. Colla, R. Gysel, N. Setter, Nanoletters 4, 1339 (2004)
85. Z. Wang, J. Hu, A.P. Suryavanshi, K. Yum, M.-F. Yu, Nanoletters 7, 2966 (2007)
86. Zhong Lin Wang, Jinhui Song, Science 312, 242 (2006)
87. W.S. Su, Y.F. Chen, C.L. Hsiao, L.W. Tu, Appl. Phys. Lett. 90, 063110 (2007)
88. Xudong Wang, Jinhui Song, Jin Liu, Zhong Lin Wang, Science 316, 102 (2007)
89. S.C. Masmanidis, R.B. Karabalin, I. De Vlaminck, G. Borghs, M.-R. Freeman,
M.L. Roukes, Science 317, 780 (2007)
21
First-Principles Theories of Piezoelectric
Materials
R.E. Cohen
potential model, which can then be used to study the effects of temperature
and simulate disordered materials.
A number of major advances in our understanding of piezoelectric materi-
als are due to first-principles studies. First is the importance of hybridization.
Ferroelectricity is due to the competition between long-range forces, which fa-
vor off-centering, and short-range forces that favor the high-symmetry centric
phase [5]. Unlike the venerable Slater rattling ion model [16], the key element
in oxide ferroelectrics is covalency, or hybridization between the cation and
its oxygen neighbors that allows the cation to move off-center. This concept
is now widely used in experiments and for development of new piezoelectric
materials. Second is the concept of polarization rotation, which is responsible
for the giant electromechanical coupling seen in relaxor ferroelectrics such as
Pb(Mg1/3 , Nb2/3 )O6 (PMN) − PbTiO3 (PT) [17]. Third is the relationship
between cation ordering and polar nanoregions in relaxors [18, 19]. Fourth is
the prediction of a morphotropic phase boundary in pure PT at high pressure,
with huge electromechanical coupling in the transition region, indicating that
the main effect of the relaxor PMN or PZN, for example, is to tune the transi-
tion to zero pressure, rather than something intrinsic to relaxor behavior [11].
Fifth is the discovery of re-entrant ferroelectricity, with ferroelectricity reap-
pearing at very high pressures, indicating the possibility of whole new classes
of ferroelectric materials [20].
First-principles studies of ferroelectrics have also given rise to major
advances in theoretical methods. The development of the modern theory of
polarization [21–23] was motivated entirely to understand piezoelectrics, and
has other broader implications as well [24, 25]. The theory of insulators under
applied electric fields was also developed to understand piezoelectric materi-
als [26, 27].
The goal of computational research on piezoelectrics is threefold: to help
understand experiments, to help guide experiments, and to make predictions
for new materials. An example of piezoelectric materials by design is exem-
plified by the predictions of interesting and tunable properties of ferroelectric
superlattices [28–31].
by an applied electric field, and generally have a phase transition with in-
creased temperature to a nonpolar, paraelectric state. Many perovskites are
ferroelectric, and this review will mainly consider the oxide perovskites. Ferro-
electricity in the LiNbO3 structure is very similar in origin, but the structure
allows uniaxial ferroelectricity only (the soft mode is nondegenerate and the
polarization must be along the c-axis), which significantly changes the dy-
namical behavior.
Ferroelectrics research was a very hot topic in the 1960s and earlier, driven
by experimental studies and phenomenological models. Starting in the 1990s,
the field was rejuvenated with the introduction of first-principles methods
along with a new generation of fundamental experimental studies.
computations are then used to obtain the parameters of the Hamiltonian, and
energy minimization, Monte Carlo, and/or molecular dynamics can then be
used to obtain materials properties [39–41].
Fig. 21.1. Energy vs. phonon displacement for (a) PbTiO3 and (b) BaTiO3 show-
ing the multiple-well potential surfaces that underlie ferroelectric behavior. Also
illustrated is the great sensitivity to volume and shear strain (from [5])
476 R.E. Cohen
Fig. 21.2. Energy vs. mode coordinate for PbTiO3 for the unstrained, cubic lattice
(dotted), and relaxed tetragonal lattice (solid). Note that the energy is a minimum
along the (111), rhombohedral, direction for the unstrained lattice and along the
(001), tetragonal, direction for the strained lattice. The great energetic importance
of the lattice strain is also evident. This behavior contrasts with that of BaTiO3 ,
which would show (111) always most stable, which is why BaTiO3 has a series of
three-phase transitions at zero pressures, whereas PbTiO3 has only 1 (from [41])
A1 A1 A1
E B1
30
Soft TO mode frequency (cm–1)
200 Cubic 6
20
Rhombohedral
Orthorhombic
(meV)
(THz)
Tetragonal
100 F1u 3
10
E
B2
0 500 1000
Temperature (K)
Fig. 21.3. Soft-mode behavior in BaTiO3 . One mode hardens at each transition,
and the mode frequencies do not go to zero. Only the orthorhombic to rhombohe-
dral transition looks similar to a conventional soft mode, even though it should be
understood as having a large order–disorder character (from [43])
478 R.E. Cohen
BaTiO3
900
2’
600 LO
1/λ (cm–1)
TO
LO 15
25’
300 25
12’
TO 25
LO 15
0 A
TO
300i
Γ Χ Μ Γ R
PbTiO3
900
2’
600 LO
1/λ (cm–1)
TO
LO 12’
25’
300 15
25
TO
LO
0 A
15
25
TO
300i
Γ Χ Μ Γ R
PbZrO3
900 2’
LO
12’
600 TO
1/λ (cm–1)
300 LO 25’
15
TO
LO
25
0 A
15
TO
25
300i
Γ Χ Μ Γ R
Fig. 21.4. Computed phonon dispersion curves for cubic BaTiO3 , PbTiO3 , and
PbZrO3 (from [47])
∗ V ∂Pα
Zs,αβ = ,
e ∂us,β
are greatly enhanced in oxide ferroelectrics, and the large coupling to electric
fields is partly due to the softness of the potential surface, and partly due to
the enhanced effective charges, since the field E couples with displacements d
via Sum Z ∗ d.
In ferroelectrics, the effective charges are often much larger than their
nominal values, contrary to chemical intuition that they would be lower than
their nominal charges due to covalency, as would be static Mulliken charges, for
example. Tables 21.1 and 21.2 show computed effective charges for PbTiO3 [8]
and rhombohedrally ordered PMN [54]. The effective charges can be greatly
21 First-Principles Theories of Piezoelectric Materials 481
Table 21.2. Computed effective charges for 1:2 ordered PMN with symmetry C2
(from [54])
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
Atom WP Zxx Zyy Zzz Zxy Zxz Zyx Zyz Zzx Zzy
type
Pb 1a 3.34 3.93 3.35 −0.29 −0.24 −0.50 −0.50 −0.23 −0.28
Pb 2c 3.57 3.79 3.50 −0.13 0.00 0.09 0.58 0.24 0.53
Mg 1b 2.89 2.87 2.88 0.04 0.14 0.02 0.02 0.14 0.04
Nb 2c 6.39 6.41 5.70 0.03 −0.39 0.11 −0.26 −0.13 −0.08
O1 2c −3.39 −2.74 −2.32 −0.11 0.43 0.05 0.12 0.43 0.03
O2 2c −1.95 −2.19 −4.03 −0.06 −0.19 −0.04 0.03 −0.19 −0.10
O3 2c −2.39 −3.69 −2.34 −0.20 0.16 −0.35 0.19 0.15 0.33
O4 2c −2.08 −2.27 −4.79 −0.03 −0.56 −0.10 −0.19 −0.14 0.12
O5 1a −2.09 −5.37 −2.07 −0.13 0.31 0.20 0.19 0.31 −0.15
WP Wyckoff position
enhanced, so that Z ∗ for oxygen can reach magnitudes of above 5. The fact
that these enhanced charges come from hybridization was shown in the same
way that the importance of hybridization was shown for the energies. When
the d-states are removed from the basis, the effective charges regain nominal
values [55].
Enhanced effective charges can also be considered to arise from atomic
polarization, as can be seen from the shell model. The shell model is a me-
chanical model for materials, where atoms are modeled as charged “cores”
and charged “shells” connected by springs. Atomic polarization is modeled by
displacement of an atom’s shell from its core. Shell models can reproduce large
effective charges through the “knock-on” effect. For example, when an oxygen
core is displaced, its shell is displaced. The shell interacts with other ionic
shells through its short-range interactions (springs) and charge, and the other
shells move as well. This can give a much greater change in polarization of the
crystal than that would occur in an isolated ion, and thus an enhanced effective
charge. Effective charges can also be determined experimentally, from LO–TO
splitting and from oscillator strengths. In fact, the enhancement in effective
482 R.E. Cohen
Table 21.3. Piezoelectric (in C/m2 ) and elastic (in GPa) constants for PbTiO3
at the experimental lattice computed using density functional perturbation theory
(DFPT) and frozen strains (FS) [11] compared with experiment (from [11])
Method e31 e33 e15 c11 c12 c13 c33 c44 c66
DFPT 2.06 4.41 6.63 230 96.2 65.2 41.9 46.6 98.8
FS 2.07 4.48 6.66 229 95.6 64.3 41.2 47.2 98.6
Experiment 2.1 5.0 4.4 237 90 70 66 69 104
charges over their nominal charges was observed first experimentally [37, 38],
although its significance was not widely recognized.
Piezoelectric constants are derivatives of the polarization with respect to
strain:
∂Pα dusµ
∗
eαβδ = = Zs,αµ ..
∂εβδ sµ
dεβδ
0
R3m
H - Hcubic (meV)
0.2
H - HP4mm (meV)
0.0
–20 Cm
–0.2
–40 –0.4
20 22 0.0
–60 P4mm
R3m
–80 –0.2
(a) 0 8 16 24 (b) 8 9 10 11 12
P4mm Cm R3m cubic
12
Elastic constants (GPa)
60 c44 ω2(E)
c11-c12
ω2 (103 cm–2)
ω2(A1)
8
40 ω2(F1u)
20 4
ω2(E)
0 0
(c) 0 4 8 12 16 20 24
12,000
Dielectric susceptibility
χ11 = χ22
9,000 χ33
6,000
3,000
0 4 8 12 16 20 24
(d) LDA pressure (GPa)
Fig. 21.5. Stability of PbTiO3 phases and properties vs. pressure. (a) Enthalpy
difference with respect to the cubic (C) phase for the tetragonal (T ) and rhombohe-
dral (R) phases. (b) Enthalpy difference with respect to the T phase for the R and
monoclinic (M ) phases. (c) Elastic constants for the T and R phases, and square of
the lowest optical phonon frequencies ω 2 for the T, R, and C phases. (d) Dielectric
susceptibility (from [11])
have been given much attention since the discovery of huge electromechani-
cal coupling in single crystals such as PbMg1/3 Nb2/3 O3 –PbTiO3 (PMN–PT)
and PbZn1/3 Nb2/3 O3 –PbTiO3 (PZN–PT) [56]. These crystals are rhombohe-
dral at the relaxor end of the phase diagram and tetragonal at the PbTiO3
side, with a morphotropic phase boundary and intermediate monoclinic or or-
thorhombic phase in-between. The largest piezoelectric response occurs near
the morphotropic phase boundary in the rhombohedral phase, with the po-
larization along the [111] directions, and an applied electric field along the
[001] direction. The large response has been understood as due to polariza-
tion rotation [57].
484 R.E. Cohen
60 T M R
e15
e (C/m2)
40 e33
20
(a) 0
0 4 8 12 16 20
9000
d15
6000
d (pC/N)
d33
3000
0
0 4 8 12 16 20
(b) LDA pressure (GPa)
The polarization rotation effect can give much higher coupling than the
normal collinear effect, in which the polarization and electric field are parallel
to each other. The polarization along [111] is compatible with a rhombohedral
structure, along [001] with tetragonal, and along [011] with orthorhombic. In-
termediate directions like [xxz] or [x0z] are monoclinic. When the polarization
is rotated with an applied field, the response is like that of a field-driven phase
transition (Figs. 21.7 and 21.8). The large response in PMN–PT, for example,
can be understood by considering pure PT. PT has a large tetragonal strain of
6%. If PT had a rhombohedral phase at zero pressure, one could apply a field
to the rhombohedral phase driving the structure toward tetragonal. The net
result would be a huge 6% strain. However, PT does not have a rhombohedral
phase at zero pressure, but a solid solution with PMN, PZN, or PZ does have
a rhombohedral phase that allows this rotation effect. The resulting strains
are smaller than in pure PT, but still much larger than the collinear effect.
The huge electromechanical coupling predicted in pure PT under pressure is
due to polarization rotation as well. Ahart et al. recently confirmed these the-
oretical predictions using cryogenic in situ microRaman and high resolution
synchrotron X-ray experiments [57B].
The polarization rotation effect has been verified now by numerous experi-
ments. It leads to intermediate low-symmetry monoclinic or orthorhombic
phases between rhombohedral and tetragonal phases on the phase diagram
[58–63]. Rotation has also been seen in situ measurements under applied elec-
tric fields [64–66].
21 First-Principles Theories of Piezoelectric Materials 485
Fig. 21.7. Polarization rotation effect. An applied field along the cubic [001] direc-
tion rotates the polarization from [111], the rhombohedral direction, toward [001],
through intermediate monoclinic states (from [57])
in a nonpolar matrix, but some data suggest that it might be better to re-
gard the polarization correlation, which would show long tails rather than a
clear separation between polar and nonpolar regions. Theoretical models for
disordered relaxors will be considered further below.
Relaxors can display two types of disorder. One type of disorder is the
polarization disorder mentioned above. Another is chemical disorder. PMN
is not generally found in a long-range ordered state, but there is evidence
for small chemically ordered domains [68–70]. Other relaxors such as PST
(Pb(Sc0.5 Ta0.5 )O3 ) can be chemically ordered by annealing, so that the effects
of chemical order on their relaxor properties can be studied experimentally. In
the ordered state, PST has a normal ferroelectric phase transition but shows
relaxor behavior when chemically disordered [71, 72]. Statistical mechanical
calculations show that the PST ground state is ordered [73–75]; the disordered
state occurs from lack of equilibrium in the B-cation distribution caused by
the slow diffusion times of the highly charged cations. The question remains
if the chemical ordered state will always be ferroelectric in a relaxor, but it
seems probable. Tinte et al. have shown in simulations that polar nanoregions
form around chemical ordered nanoregions, and this leads to many of the
observed phenomena of relaxor ferroelectrics [19].
First-principles calculations for PMN do show a polar, ordered ground
state with monoclinic C2 symmetry with 1:2 stacking of Mg and Nb along
(111) planes [76] or 1:1 symmetry [77, 78]. If cubic PMN did form nanoscale
monoclinic domains, it would be difficult to determine definitively using dif-
fraction data since it is very difficult to deconvolute the diffraction data from
a sample with 12 twin variants. Nanodomains of the twin variants would pro-
vide static random fields that would contribute further to the disorder of the
material and possibly some of the frequency-dependent relaxor behavior.
The effective charges in relaxors may contribute to the low-symmetry
ground state found for PMN. Computations of effective charges in PMN show
them to be very anisotropic. Local fields in the polar phase will lead to oxygen
displacements in oblique directions, giving rise to a monoclinic ground state.
Fig. 21.9. Computed energy surface for LiTaO3 for displacing Li only, O only, or
Li and O along the soft-mode coordinate. Clearly, the instability is not due to Li
alone, and the concerted motion of Li and O is important. The results for LiNbO3
are very similar (from [42])
a triple well potential surface, which is not consistent with the first-principles
results. Frozen phonon calculations showed that the double well is largely due
to hybridization between the d0 Nb5+ and Ta5+ ions with oxygen, much as
for the oxide perovskite, whereas previous work had considered the ferroelec-
tricity to be mainly due to hopping of the Li. The Li does hop, but mainly in
response to the changing fields from the off-centering of the Nb or Ta. Linear
response computations gave the phonon frequencies of the ground state fer-
roelectric state [79]. Modeling of the double well potential as an anharmonic
oscillator gave good agreement with experiments. One outstanding problem
is a first-principles understanding of the difference of Tc of LiNbO3 (1,480 K)
and LiTaO3 (950 K) since they have almost identical potential surfaces.
this study, including the importance of strain coupling, although the words
“ferroelectric” or “piezoelectric” do not appear in the paper. GeTe remains
a material with the simplest possible structure for a ferroelectric phase tran-
sition. There are only two atoms per unit cell and the paraelectric phase is
rocksalt structured. Waghmare et al. [80] revisited GeTe and other chalco-
genides, using first-principles techniques developed in the last 20 years that
were unavailable in the 1980s, including linear response and Barry’s phase
techniques. They found large effective charges and large optical dielectric
constants, and showed that the ferroelectric instabilities are partly due to
lone-pair activity.
21.1.2.5 KDP
21.1.2.6 Multiferroics
21.1.3 Summary
Acknowledgments
References
1. A.F. Devonshire, Philos. Mag. 3(10), 85 (1954)
2. D. Vanderbilt, M.H. Cohen, Phys. Rev. B 63, 94108 (2001)
3. L.L. Boyer et al., Ferroelectrics 111, 1 (1990)
4. R.E. Cohen, H. Krakauer, Phys. Rev. B 42(10), 6416 (1990)
5. R.E. Cohen, Nature 358, 136 (1992)
6. R.E. Cohen, Nature 362, 213 (1993)
7. K.M. Rabe, J.D. Joannopoulos, Phys. Rev. B 36(12), 6631 (1987)
8. G. Saghi-Szabo, R.E. Cohen, H. Krakauer, Phys. Rev. Lett. 80(19), 4321
(1998)
9. G. Saghi-Szabo, R.E. Cohen, H. Krakauer, Phys. Rev. B 59, 12771 (1999)
10. Z. Wu et al., Phys. Rev. Lett. 94(6), 069901 (2005)
11. Z. Wu, R.E. Cohen, Phys. Rev. Lett. 95, 037601 (2005)
12. R. Resta, Model. Simul. Mater. Sci. Eng. 11(4), R69 (2003)
13. K.M. Rabe, P. Ghosez, in Physics of Ferroelectrics: A Modern Perspective, ed.
by C.H. Ahn, K.M. Rabe, J.M. Triscone (Springer, New York, 2006)
14. C. Fiolhais, F. Nogueira, M. Marques (eds.), A Primer in Density Functional
Theory. Lecture Notes in Physics, vol. 620 (Springer, New York, 2003), p. 256
15. W. Kohn, L.J. Sham, Phys. Rev. A 140, 1133 (1965)
21 First-Principles Theories of Piezoelectric Materials 491
22.1 Introduction
The classic work of Devonshire [1, 2] in developing the phenomenological
theory of barium titanate represented a significant advance in the study of
the physical properties of piezoelectric crystals and ceramics based on the
perovskite-structured ferroelectric oxides. Although various theories of fer-
roelectricity in BaTiO3 had been put forward by other workers during the
period, the phenomenological theory as formulated by Devonshire provided
the first unified description of its changes in crystal form and of the variations
of the thermal, dielectric, and piezoelectric properties observed on cooling
through the sequence of cubic → tetragonal → orthorhombic → rhombohedral
ferroelectric phase transitions. By taking account of the multicomponent na-
ture of the ferroelectric polarization in the perovskite oxides, the Devonshire
theory allowed the full matrix of tensor components describing the properties
of the monodomain single-crystalline state to be assembled, even within the
constraints of a comparatively limited set of available experimental data. The
influence of the Devonshire theory on contemporary workers in the field is ev-
ident in the early investigations of Slater [3] into the origins of ferroelectricity
in BaTiO3 , in that of Kittel [4] who predicted the possibility of antiferroelec-
tricity in crystals of the perovskite type before the effect had been experimen-
tally verified, and in that of Cross [5] in his treatment of ferroelectricity and
antiferroelectricity in NaNbO3 −KNbO3 and related perovskite systems.
Though it has been more than 50 years since the Devonshire papers
appeared, phenomenological theories continue to play an important role in
connection with the study of piezoelectric materials, and there are several rea-
sons why this has been the case. Nearly all of the materials of technological
importance for piezoelectric transducer applications are based on perovskite-
structured crystals. The particular form of the phenomenological theory as put
forward by Devonshire for BaTiO3 has therefore naturally been extended in a
more or less straightforward way to the study of these other perovskite materi-
als as well. The phenomenological theory is a useful tool for experimentalists
494 G.A. Rossetti, Jr.
22.2 Phenomenology
The phenomenological theory as applied by Devonshire to barium titanate
crystals is a particular form of the Ginzburg–Landau theory of phase transi-
tions. The use of the Devonshire theory in describing the thermal, dielectric,
and piezoelectric properties of BaTiO3 and related perovskite materials has
been described by many authors, and is treated in detail, e.g., in the text of
Jona and Shirane [6]. More general considerations associated with the foun-
dations of the phenomenological theory and its application to a variety of
other physical phenomena as observed in ferroelectric crystals are elaborated
in considerable depth in the monograph of Strukov and Levanyuk [7] and the
treatment given by them will be followed here. However, here, attention will
be restricted to the specific case of the ferroelectric perovskites and focused
only on those aspects that are especially relevant to describing the pressure–
temperature or composition–temperature phase diagrams of these materials
as the essential starting point for enabling an understanding of the physical
properties that they may display.
For this purpose, it is sufficient to begin by asserting without further justifi-
cation that the behavior of a ferroelectric crystal in the vicinity of a structural
phase transition may be described by some thermodynamic potential function
Φ = Φ(θ, p, η), where Φ is the scalar quantity of energy density (energy per
unit volume), θ is the absolute temperature, and p is the pressure. The vari-
able η represents the order parameter of the structural phase transition(s).
In the high-symmetry or prototype phase η = 0, and in the lower-symmetry
phase or phase(s) η = 0. The thermodynamic potential describes both the
prototype phase and the lower-symmetry phase(s) derived from it and so the
form of the dependence Φ on η is restricted by the condition that Φ must
be invariant with respect to any transformations of the high-symmetry phase.
Specifically, this requires that
Φ(η) = Φ(η ∗ ), (22.1)
where η ∗ is the order parameter transformed by the symmetry operations of
the prototype phase. Close to the temperature θc at which the phase transi-
tion occurs, η is vanishingly small and the thermodynamic potential may be
expressed as a series expansion in powers of the order parameter η:
1
Φ(η) = Φ(0) + Φ η + Φ η 2 + · · · , (22.2)
2
where Φ(η) depends on the external parameters of temperature θ and
pressure p. Consequently, the exact form of the thermodynamic potential
496 G.A. Rossetti, Jr.
where at the pressure p = pc the coefficients A(θ), B(θ), and C(θ) in the
vicinity of θc are expressed as series expansions in powers of (θ − θc ):
1
A(θ) = A(θc ) + A (θ − θc ) + A (θ − θc )2 + · · · ,
2
1
B(θ) = B(θc ) + B (θ − θc ) + B (θ − θc )2 + · · · , (22.4)
2
1
C(θ) = C(θc ) + C (θ − θc ) + C (θ − θc )2 + · · · .
2
For the situation where B(θ) > 0 and C(θ) 0, the transition I ↔ II then oc-
curs as A(θ) passes continuously through zero. In the lowest order approxi-
mation, (22.4) gives A > 0, B(θc ) > 0, C(θc ) 0, and A(θc ) = B = B =
C = C = 0. The thermodynamic potential (22.3) becomes simply
1 1 1
Φ(θ, η) = Φ0 (θ) + α(θ − θc )η 2 + βη 4 + γη 6 , (22.5)
2 4 6
where α = 2A , β = 4B(θc ), and γ = 6C(θc ). It follows that the condition
A(pc , θc ) = 0 defines the line of the second-order transitions in the p−θ plane
between phases I and II and so describes the pressure–temperature phase di-
agram of the crystal. The transition between phases I and II of course need
not be of second order. First-order transitions are described by (22.3) if for
the coefficient of quartic term we have B(p, θ) < 0 and for the sextic term
C(p, θ) > 0. Alternatively, it may happen that a line of first-order transitions
crosses over to a line of second-order transitions. Such a situation occurs, for
example, in crystals placed under hydrostatic pressure, or in some instances,
when the chemical composition of the host crystal is altered by ion substitu-
tion. The point at which this crossover occurs is termed a tricritical point and
22 Thermodynamic Theory 497
corresponds the case where the coefficient of the quartic term changes sign,
viz., when B(p, θ) = 0.
Application of the equilibrium condition to (22.5) requires that
Φ(θ) = Φ(θ, η0 (θ)) (22.6)
with the equilibrium value of the order parameter ηeq = η0 being deter-
mined from
∂Φ ∂2Φ
= 0, > 0. (22.7)
∂η ∂η 2
Generalizing from the above considerations, the thermodynamic potential
for a crystal undergoing a structural phase transition may be written in the
general form
Φ(p, θ, . . . , N, η1 , . . . , ηn ) = Φ0 (p, θ, N ) + Φ1 (p, θ, . . . , N, η1 , . . . , ηn ), (22.8)
where N is some additional external parameter other than pressure or tem-
perature, ηi are variables representing the order parameters of the system,
and the function Φ0 (p, θ, N ) represents that part of the free energy den-
sity that does not depend on these variables. The form of the function of
Φ1 (p, θ, . . . , N, η1 , . . . , ηn ) is determined by the transformation properties of
the order parameter subject to the symmetry operations of the prototype
phase and by the temperature, pressure, or other dependencies of its resulting
nonzero coefficients. Minimizing Φ(p, θ, . . . , N, η1 , . . . , ηn ) with respect to ηi
∂Φ ∂Φ
= 0, . . . , =0 (22.9)
∂η1 ∂ηn
gives a set of values η0 (p, θ, . . . , N ) which when substituted into (22.8) gives
the equilibrium thermodynamic function:
In the case of the perovskite crystals, the cubic prototype phase belongs
to the centrosymmetric symmetry group m3m. As such, it is paraelectric and
nonpiezoelectric. The order parameter η for the transition from the cubic pro-
totype phase into one or more lower-symmetry ferroelectric phases is identified
with the polarization vector P and is related to the amplitude of a transverse
optic soft mode that condenses at θc . The Curie principle asserts that the
“proper” ferroelectric phase(s) that may appear in the crystal are those that
are subgroups of the prototype phase and so must contain all symmetry el-
ements common to the symmetry group of the prototype and the symmetry
group of the order parameter. The order parameter P is a polar vector quan-
tity and its symmetry group is ∞m. Shuvalov [8] has determined all such
possible ferroelectric phases and these are listed in Table 22.1. As shown in
the table, the cubic m3m prototype phase of the perovskite crystals admits
seven proper lower-symmetry ferroelectric phases: one tetragonal (4mm), one
rhombohedral (3m), one orthorhombic (mm2), three monoclinic (m), and one
triclinic (1).
498 G.A. Rossetti, Jr.
Table 22.1. Possible proper ferroelectric phase transitions from symmetry group
m3m (adapted from [8])
ηi , ηi ηj , ηi ηj ηk , ηi ηj ηk ηl , ηi ηj ηk ηl ηm , . . . , (22.11)
to be the exception rather than the rule. For example, in the perovskite
compounds KNbO3 , BaTiO3 , and PbTiO3 , all of which exhibit the transi-
tion m3m → 4mm, in each case the transition is found to be first order [6].
To illustrate how the parameters of the thermodynamic potential may be
determined for some particular crystal of interest, we now consider the simple
case of the single first-order transition in PbTiO3 that occurs, at atmospheric
pressure, at a temperature ≈760 K [10]. In analyzing the experimental data,
it is convenient to define the order parameter η = nη, where n is a unit
vector in the direction of the polarization and η is its absolute value. In
the tetragonal 4mm phase, the polar vector appears along one of the cube
axes of the prototype phase with n = {0, 0, 1}. At atmospheric pressure,
thermodynamic potential then assumes the form
1 1! "
Φ(θ, η, n) = Φ0 (θ) + α(θ − θc )η 2 + β + β (n41 + n42 + n43 ) η 4
2 4 (22.16)
1! "
+ γ + γ (n61 + n62 + n63 ) + γ (n21 n22 n23 ) η 6 .
6
Because there is only one transition corresponding to the spontaneous polar-
ization vector directed parallel to the directions 001, the basic features of
the first-order phase transition will be captured even when the anisotropic
contributions of the sextic invariants in (22.16) are neglected. Making this
approximation, the expansion of the thermodynamic potential begins with
1 1 ! " 1
Φ(θ, η, n) = Φ0 (θ) + α(θ − θc )η 2 + b0 1 − λ(n41 + n42 + n43 ) η 4 + γη 6 ,
2 4 6
(22.17)
where the constants α, b0 , γ > 0. Substituting n = {0, 0, 1}, the entire
quartic coefficient is β0 = b0 (1 − λ) and it is evident that the transition is first
order for λ > 1. From (22.13),
# 1/2 $
β0 4α(θ − θc )γ
η0 = −
2
1+ 1− . (22.18)
2γ β02
From this, it is seen that the highest temperature to which the ferroelectric
phase may exist as a metastable phase in the absence of an applied external
field is given by
β2
θ ∗ = θc + 0 , (22.19)
4αγ
Differentiating (22.17) with respect to temperature, the entropy and specific
heat at constant pressure are
∂Φ ∂σ
σ=− , Cp = θ . (22.20)
∂θ p ∂θ p
Following from (22.17) and (22.20), the temperature dependence of the “ex-
cess” entropy ∆σ of the phase change is
22 Thermodynamic Theory 501
1
∆σ = σ − σ 0 = − αη02 (22.21)
2
and the temperature dependence of the “excess” specific heat ∆Cp of the
phase change is
−2 −2
∆Cp Cp − C 0 4[β02 − 4α(θ0 − θc )] 16γ
= = + 3 (θ0 − θ), (22.22)
θ θ α4 α
where θ0 is the temperature at which the cubic and tetragonal phases are in
equilibrium and is defined by the condition Φ(η = 0) = Φ(η = η0 ).
Under mechanically free conditions, the strain associated with the distor-
tion of the cubic crystal lattice is of the same order of smallness as the square
of the order parameter, Sij ∝ η 2 with
where the fourth-rank tensor qijkl is symmetrical with respect to any permu-
tations i ↔ j, k ↔ l, and (ij) ↔ (kl). Since n = {0, 0, 1}, the relation
between strain and polarization in the tetragonal phase is then
⎛ ⎞
q12 0 0
Sij = ⎝ 0 q12 0 ⎠ η 2 . (22.24)
0 0 q11
From (22.24), the crystal lattice parameters ct and at of the tetragonal ferro-
electric phase are given by
a1 = a2 = ac (1 + S11 ) = ac (1 + q12 η 2 ) = at ,
(22.25)
a3 = ac (1 + S33 ) = ac (1 + q11 η 2 ) = ct ,
Fig. 22.1. Temperature dependence of the excess specific heat, excess entropy, and
order parameter of PbTiO3 at atmospheric pressure (adapted from [11, 12])
Fig. 22.2. Temperature dependence of the crystal lattice parameters (ac of the
cubic phase and at and ct of the tetragonal phase) and the quantity (∆Cp /θ)−2 of
PbTiO3 at atmospheric pressure (adapted from [11, 12])
extends the utility of the analysis of such transitions as occur, e.g., in the
PbZrO3 −PbTiO3 MPB system considered below, where the single proper
ferroelectric transitions m3m → 4mm and m3m → 3m occur over most of
the composition–temperature diagram.
To take account of the effects of pressure on the phase transition, the elastic
energy must be added to the thermodynamic potential. The linear stress-
induced elastic strain is Sij = sijkl Tkl , and for the cubic crystals there are
three nonzero components of the elastic compliance tensor, s11 , s12 , and s44 in
Voigt notation. If the applied pressure is hydrostatic, there are three nonzero
components of the stress tensor −T11 = −T22 = −T33 = −p and following
from (22.25), elastic contributions to thermodynamic potential become
3
Φelas (p, η) = − (s11 + 2s12 )p2 − (q11 + 2q12 )η 2 p. (22.26)
2
504 G.A. Rossetti, Jr.
Fig. 22.3. Dependence of the excess entropy on the square of the order parameter
for PbTiO3 at atmospheric pressure (adapted from [11, 12])
that best agree with all known measured properties can be determined. Using
these coefficients, it is then possible to investigate the relative phase stabilities
and physical properties of the crystal for some other particular set of thermal,
mechanical, or electrical boundary conditions of interest.
Thus, one of the main practical advantages of the phenomenological ap-
proach is that when properly chosen and analyzed, a relatively small set of
experimental data can be used to infer in a self-consistent way a great deal
of information about the thermodynamic behavior and physical properties of
the crystal of interest. Viewed from this perspective, a thermodynamic po-
tential function properly constructed from the phenomenological theory then
provides a compact description of the properties of the crystal. In the case
of ferroelectric solid solutions, relative stabilities of the ferroelectric phases
and the transitions among them will of course depend not only on temper-
ature and pressure but also on the composition variable c. Following from
the most general form of the thermodynamic potential given by (22.8), it is
possible to extend the above analysis to describe some of the features of the
temperature–composition (θ − c) phase diagram. The application of such an
analysis to the specific case of the PbZrO3 −PbTiO3 solid solution is given
in Sect. 22.3 and some thermodynamic considerations that arise in connection
with their piezoelectric properties are discussed.
Fig. 22.6. Discontinuity in the unit cell volume Ω(θc ) and tetragonal lattice distor-
tion δ(θc ) as a function of composition in the PbZrO3 :PbTiO3 system at atmospheric
pressure (adapted from [23])
As shown in the figure, the vanishing values of Ω(θc ) and δ(θc ) indicate
the existence of two tricritical points: one located at c∗1 ≈ 0.22 and another
located at c∗2 ≈ 0.55. The possible existence of such points in the c−θ diagram
of ferroelectric solutions has important consequences that follow directly from
the phenomenological theory. First, as discussed earlier, a tricritical point cor-
responds to the condition that the quartic term(s) in the thermodynamic po-
tential change their sign(s). Consequently, there is a close connection between
the location of a tricritical point and the topology of the c − θ diagram. Such
a connection has in fact been observed. In the phase diagram correspond-
ing to Fig. 22.6, Eremkin et al. [23] found that the morphotropic boundary
separating tetragonal and rhombohedral phases was not vertical as shown in
Fig. 22.5, but instead showed pronounced temperature dependence with the
boundary “bending” sharply toward more Zr-rich compositions and intersect-
ing the line of paraelectric to ferroelectric transitions at the point (c∗1 , θ∗ )
with θ∗ ≈ 610 K. Second, the domain wall energy on either side of a tricritical
transition is expected to display a different dependence on temperature below
the θc . Salje [28] shows that the domain wall energy at θ < θc for a tricritical
transition increases linearly with (θ − θc ), while in a second-order transition
the domain wall energy increases as |(θ − θc )|3/2 . As a result, the domain wall
energy is always greater for the tricritical transition compared with that for
the second-order transition and so domain wall formation in the latter case is
much more strongly favored. On changing composition in a narrow region in
22 Thermodynamic Theory 509
dmkl = ε0 χθ,T
mj qijkl ηi , (22.30)
literature for data collected near room temperature give values that range
from essentially 0 to as much as 15 mole % [32, 33].
The origins of phase coexistence in morphotropic phase systems have long
been controversial and Isupov [34] has collected much the literature on this
topic. Based on considerations of the thermodynamics of mixing in the ap-
proximation of the theory of regular solutions, it is possible, however, to easily
establish some limiting cases. Following from (22.8), the thermodynamic po-
tential can be written as
where the first term on the right-hand side of the equation represents the
free energy of mixing and the second term is the Landau potential. Adopt-
ing the conventional regular solution model for the mixing free energy and
using the form (22.17) for the η-dependent part of the thermodynamic poten-
tial gives
1
Φ(c, θ, η, n) = W c(1 − c) + kb θ [c ln c + (1 − c) ln(1 − c)]
2
1 1
+ α(c)[θ − θc (c)]η 2 + (c − cm ) (22.32)
2 4
! " 1
×b0 1 − λ(n41 + n42 + n43 ) η 4 + γη 6 ,
6
where kb is a Boltzmann’s constant, W is the atomic mixing parameter,
and cm = 0.5 is the nominal composition of the diffusionless tetragonal–
rhombohedral MPB. Based on measurements of the enthalpies of formation,
Rane et al. [35] have shown that the regular solution model is approximately
obeyed for PbZrO3 −PbTiO3 and that the system displays a positive enthalpy
of mixing, W > 0. To make some computations using (22.32), we may take
b0 , γ, λ > 0 to be constants, independent of c, and give α(c) and θ(c) lin-
ear dependencies between the known values for PbZrO3 and √ PbTiO
√ 3 .√Noting
that n = {0, 0, 1} for the tetragonal phase and n = {1/ 3, 1/ 3, 1/ 3} for
the rhombohedral phase and applying the minimum principle (22.9) gives the
equilibrium phase diagram for different values of W [36]. The results of such
an analysis are shown in Fig. 22.8.
The diagram in the panel at left was computed by finding the locus of
points in the c − θ plane where the free energy density curves of the tetrag-
onal and rhombohedral phases are equal. This diagram represents the diffu-
sionless limit. The remaining two diagrams were computed in the conventional
way by finding the points of common tangency on the free energy density vs.
composition curves. The diagram at center represents the situation where the
PbZrO3 −PbTiO3 system behaves as an ideal solution (W = 0) and the dia-
gram at right represents the situation where the systems behaves as regular
solution with a positive enthalpy of mixing (W > 0) constrained to be suffi-
ciently small such that no decomposition may occur in the cubic phase. If dif-
fusional processes are operative, the diagrams in Fig. 22.8 indicate that the
512 G.A. Rossetti, Jr.
800
Temperature (K) PC+FT
Pc
700 Pc Pc
PC+FR FT FT
600
FT
500 FR PC+FR
FR FR 0.498 0.5 PC+FT
400 FR+FT FR+FT
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(a) c,mol fraction PbTiO3 (b) c,mol fraction PbTiO3 (c) c,mol fraction PbTiO3
Fig. 22.9. Changes in Madelung energy (∆EM ) and nonelectrostatic energy (∆EN )
contributions to the change in lattice energy in the PbZrO3 :PbTiO3 perovskite sys-
tem as functions of the degree of deviation from the ideally close-packed cubic struc-
ture (∆ac ) (adapted from [27])
22.4 Summary
The phenomenological theory provides a convenient framework within which
the phase transitions and physical properties of ferroelectric crystals can be in-
terpreted. Because the parameters of the theory are directly related to macro-
scopically observed quantities, they can be determined in a self-consistent way
from a small number of experimental measurements. The general observation
514 G.A. Rossetti, Jr.
that for ferroelectric perovskite crystals the range of validity of the theory
away from the Curie temperature is wider than that would be expected ex-
tends its utility in describing the physical properties of these technologically
important materials. Deviations of measured properties from those predicted
by the parameters of theory thus provide the logical starting point for inves-
tigations into more complex phenomena and for inquiries into their possible
microscopic origins.
Formal extension of the theory to include composition as a parameter of
the thermodynamic potential can be used to compute possible topologies
of the composition–temperature phase diagram and to explain the existence
of spatially inhomogeneous states that may be more complex than those asso-
ciated only with the conventional symmetry twin structures. Considerations
predicated on the possible topologies of these diagrams at equilibrium may
provide insights useful in constructing hypothesis-driven experiments, leading
to better control of their piezoelectric and other structure-sensitive proper-
ties. Knowledge of the mixing properties of such systems further provides a
possible empirical means to correlate the topologies of such phase diagrams
with crystallochemical parameters.
References
1. A.F. Devonshire, Philos. Mag. 40, 1040 (1949)
2. A.F. Devonshire, Philos. Mag. 42, 1065 (1951)
3. J.C. Slater, Phys. Rev. 78, 748 (1950)
4. C. Kittel, Phys. Rev. 82, 729 (1951)
5. L.E. Cross, Philos. Mag. Ser. 8, 1, 76 (1956)
6. F. Jona, G. Shirane, Ferroelectric Crystals (Pergamon, New York, 1962)
7. B.A. Strukov, A.P. Levanyuk, Ferroelectric Phenomena in Crystals (Springer,
Berlin, 1998)
8. L.A. Shuvalov, J. Phys. Soc. Jpn 28, 38 (1970)
9. D. Vanderbilt, M.H. Cohen, Phys. Rev. B 63, 094108 (2001)
10. K. Wójcik, Ferroelectrics 99, 5 (1989)
11. G.A. Rossetti Jr., J.P. Cline, A. Navrotsky, J. Mater. Res. 11, 3179 (1998)
12. G.A. Rossetti Jr., N. Maffei, J. Phys.: Condens. Matter 17, 3953 (2005)
13. G.A. Samara, Ferroelectrics 2, 277 (1971)
14. R. Ramı́rez, M.F. Lapeña, J.A. Gonzalo, Phys. Rev. B 42, 2604 (1990)
15. A. Sani, M. Hanfland, D. Levy, J. Solid State Chem. 167, 446 (2002)
16. B. Jaffe, W.R. Cook, H. Jaffe, Piezoelectric Ceramics (Academic, London, 1971)
17. A.M. Glazer, P.A. Thomas, K.Z. Baba-Kishi, G.K.H. Pang, C.W. Tai, Phys.
Rev. B 70, 184123 (2004)
18. B. Noheda, D.E. Cox, G. Shirane, J.A. Gonzalo, L.E. Cross, S.-E. Park, Appl.
Phys. Lett. 74, 2059 (1999)
19. D.I. Woodward, J. Knudsen, I.M. Reaney, Phys. Rev. B 72, 104110 (2005)
20. B. Noheda, Curr. Opin. Solid State Mater. Sci. 6, 27 (2002)
21. A.J. Bell, J. Mater. Sci. 41, 13 (2006)
22. R.W. Whatmore, R. Clarke, A.M. Glazer, J. Phys. C: Solid State Phys. 11,
3089 (1978)
22 Thermodynamic Theory 515
23. V.V. Eremkin, V.G. Smotrakov, E.G. Fesenko, Ferroelectrics 110, 137 (1990)
24. B. Noheda, N. Cereceda, T. Iglesias, G. Lifante, J.A. Gonzalo, H.T. Chen, Y.L.
Wang, Phys. Rev. B 51, 16388 (1995)
25. S.K. Mishra, A.P. Singh, D. Pandey, Philos. Mag. B 76, 213 (1997)
26. G.A. Rossetti Jr., A. Navrotsky, J. Solid State Chem. 144, 188 (1999)
27. G.A. Rossetti Jr., J.P. Cline, A. Navrotsky, Y.-M. Chiang, J. Phys.: Condens.
Matter 14, 8131 (2002)
28. E.K.H. Salje, Phase Transitions 34, 25 (1991)
29. A. Amin, M.J. Haun, B. Badger, H. McKinstry, L.E. Cross, Ferroelectrics 65,
107 (1985)
30. M.J. Haun, Z.Q. Zhuang, E. Furman, S.J. Jang, L.E. Cross, Ferroelectrics 65,
45 (1989)
31. X.L. Zhang, Z.X. Chen, L.E. Cross, W.A. Schluze, J. Mater. Sci. 18, 968 (1983)
32. B. Noheda, D.E. Cox, G. Shirane, R. Guo, B. Jones, L.E. Cross, Phys. Rev. B
63, 014103 (2000)
33. T. Yamazaki, A. Onodera, H. Fujishita, Y. Ishikawa, M. Tanaka, Ferroelectrics
266, 139 (2002)
34. V.A. Isupov, Ferroelectrics 266, 91 (2002)
35. M.V. Rane, A. Navrotsky, G.A. Rossetti Jr., J. Solid State Chem. 161, 402
(2001)
36. G.A. Rossetti Jr., W. Zhang, A.G. Khachaturyan, Appl. Phys. Lett. 88, 072912
(2006)
37. E. Takayama-Muromachi, A. Navrotsky, J. Solid State Chem. 72, 244 (1988)
38. P.K. Davies, A. Navrotsky, J. Solid State Chem. 46, 1 (1983)
23
Effective Medium Theories
W. Kreher
23.1 Introduction
Predicting macroscopic properties of materials involves two tasks:
1. Projection of macroscopic loading conditions onto the microstructural
level with sufficient and adjusted details (resulting in fluctuating inter-
nal fields)
2. Averaging the local reaction of material to obtain the macroscopic re-
sponse to the loadings
For piezoelectric/ferroelectric polycrystalline materials, the relevant micro-
structural level is presented by special domain configurations within grains
which have different orientations. The local material response comprises the
elastic, dielectric, and intrinsic piezoelectric effect as well as the restructuring
of domain configuration by movement of domain walls or complete reorienta-
tion of the spontaneous polarization, i.e., domain switching. Other processes
as moving free charges and/or defect dipoles are normally not considered in
present micro–macro models. They appear in the form of already averaged
local properties, e.g., as an effective frictional force or activation energy for
domain wall movement.
Obviously, exact three-dimensional solutions for the internal fields and
processes are impossible, so several simplifications are introduced in the mod-
els depending on the aim (i.e., whether linear, nonlinear, or complete repolar-
ization behavior is considered). Typical approximations are:
• Simplification of domain configuration either by a particular deterministic
domain arrangement (e.g., stack of plate-like domains) or by assumption
of a set of noninteracting representative domains
• Simplification of the interactions between neighboring grains (e.g., assump-
tion of homogeneous fields, smoothing by the effective medium approach)
The present chapter describes these modeling approaches in some detail,
putting emphasis on partially exact solutions and practical applicability.
518 W. Kreher
Section 23.2 describes the relevant modeling tools, namely the exact solutions
for a lamellar domain stack and for an ellipsoidal inclusion in a homogeneous
matrix. On this basis, the effective medium approach for a polycrystalline
solid is introduced. Section 23.3 extends the theory to the extrinsic properties
which are determined by the (reversible) movement of domain walls. Finally,
Sect. 23.4 gives a short overview about modeling the nonlinear behavior which
is connected with extensive domain reconfiguration.
To represent the theoretical background in a compact and comprehensible
form, it is necessary to introduce a short form notation for the piezoelec-
tric material laws. These so-called constitutive equations, relating mechanical
stress and electric displacement to strain and electric field, are expressed by
pq Sq − elp El + Tp ,
Tp = cE E
p, q = 1, 2, . . . , 6, k, l = 1, 2, 3. (23.1)
Dk = ekq Sq + εSkl El + DkS ,
(We make use of the abbreviated Voigt notation and the Einstein summa-
tion convention.) Here, TpE is the stress state which arises if a solid, which
is clamped (Sq = 0) and short circuited (El = 0), undergoes a ferroelec-
tric/ferroelastic transformation, and DkS is the electric displacement observed
under the same conditions. Of course, these remnant properties may be ex-
pressed by the corresponding quantities under stress-free conditions (i.e.,
transformation strain and spontaneous polarization); yet the representation
adopted here is especially suited for the derivation of material properties
within the effective medium approximation.
Though we do not explicitly consider pyroelectric and thermoelastic ef-
fects, it should be mentioned that these effects are comprised by T E and
DS , too. Thus, any theoretical result shown below can be utilized to make
predictions also for these thermal properties.
Combining mechanical and electrical field quantities into 9-dim vectors
gives the compact notation:
with
Tp cE t
pq epl Sq TpE
j= , L= , f= , js = . (23.3)
Dk ekq −εSkl −El DkS
Here, the superscript t means the transposition of the 3×6 matrix of piezoelec-
tric constants. The matrix L represents the complete set of the linear elastic,
piezoelectric, and dielectric constants; it is a symmetric 9 × 9 matrix, i.e.,
L = Lt . (23.4)
These equations are valid for each lamella, i.e., the field components appearing
in (23.7) and (23.8) must be homogeneous throughout the stack.
Now, we consider an individual lamella. Then, (23.7) and (23.8) together
with the local material law (23.2) form nine conditions for nine unknowns
(f1 , . . . , f9 ). However, since we only know f¯, which is prescribed, we do not
have the field components j̄3 , j̄4 , j̄5 , j̄9 necessary to utilize (23.8). The missing
four additional global equations are obtained if the average field condition is
taken into account:
fi = f¯i , i = 3, 4, 5, 9. (23.9)
Here, · means averaging over all lamellas. Of course, (23.9) is valid also for
i = 1, 2, 6, 7, 8, but since these field components do not fluctuate within
the stack (cf. (23.7)), the equations are trivially fulfilled.
Thus, (23.7)–(23.9) represent a complete set of equations to determine
each lamella field, which can be solved analytically. Due to the linearity of the
problem, the stack behaves like a homogeneous material described by effective
material constants:
j̄ = j = Lstack f¯ + j s,stack . (23.10)
The effective constants Lstack and j s,stack depend on the properties of domains
and their respective volume fraction. Typically, the lamella properties do not
fluctuate arbitrarily; instead, one has a limited number of domain types (in-
clined at fixed angles to the stack orientation). Then, the internal fields are
23 Effective Medium Theories 521
identical for domains belonging to a certain type, and the averaging is done by
means of volume fractions of domain types. In Sect. 23.3, a domain stack con-
sisting of two types of domains forming 90◦ domain walls will be considered.
where I is the 9 × 9 unit matrix. The Eshelby matrix in (23.12) and (23.13)
depends on the piezoelectric properties of the matrix and on the geometry of
the inclusion, i.e.,
522 W. Kreher
1.0
0.8
99
0.6
Eshelby tensor Sij
91
77
57
0.4
0.2
93
39
0.0
19
75
–0.2
0.01 0.10 1.00 10.00 100.00
Aspect ratio a3 /(a1=a2)
1.0
33
0.8
11
66
0.6
Eshelby tensor Sij
44
0.4
0.2 13
12
0.0
31
–0.2
0.01 0.10 1.00 10.00 100.00
Aspect ratio a3 /(a1=a2)
These properties can be fed into (23.17) and (23.18) in place of L and j s ,
respectively. Together with matrix properties supposed to be known, (23.12)
yields f which is now the average inclusion field. It is not necessary to deter-
mine the residual stress and electric field outside the inclusion since rigorous
expressions are available which relate the potential energy of the solid to
the field inside the inclusion (see [16, 17]). Here, potential energy means the
internal field energy including the potential of the loading forces. For exam-
ple, under isothermal conditions and for prescribed strain and electric field,
the potential energy is a combination of free elastic energy and free electric
enthalpy.
Figure 23.2 shows a typical result demonstrating the equilibrium position
of domain walls. Without an external field, both types of domains have equal
volume fractions, whereas under the action of the electric field the favorably
oriented domains grow on the expense of the other domains. Such calculations
take into account the complete mechanical and electrical interaction fields
(except for the small end regions of the domains). They can be performed for
any orientation of the grain and for any properties of the surrounding medium
provided the matrix S of Eshelby tensors is known.
2000
–5
1500
0
1
3
1000
5
500
0.0 0.2 0.4 0.6 0.8 1.0
Local domain volume fraction
Fig. 23.2. Potential energy wp per volume of a specific inclusion as a function of the
volume fraction ξ of the nonfavorably oriented domain. The inclusion is embedded in
a matrix which has isotropic properties (i.e., a nonpoled material). The far field f 0
prescribes a varying electric field under the condition of a constant strain (after [21])
526 W. Kreher
The results according to (23.19) may be used within the EMA, but now allow-
ing for small reversible shifts of domain walls. Here, all grains are described
by a lamellar stack of 90◦ domain walls where only the orientation Ω and the
domain volume fraction ξ vary. The domain volume fraction is determined by
the minimum of the potential energy as described in Sect. 23.3.1. Of course,
the local equilibrium value depends on the applied field and on the local
orientation:
ξ = ξ(f¯, Ω). (23.20)
This together with (23.19) gives the grain properties which in turn can be
fed into the effective medium formalism (23.17) and (23.18). The resulting
effective properties now become dependent on the applied field. Though in
general this relationship is nonlinear, it can be linearized for small fields. To
this end, we consider the incremental form of (23.15):
In first order, the increments dL∗ and dj ∗ will be proportional to the applied
field increment df¯. Therefore, we obtain
where the new material property matrix L̂∗ is the sum of the already known
effective intrinsic properties and the effective extrinsic contributions due to
the variations dL∗ and dj ∗ caused by the moving domain walls:
300
d15 ∗
d15 intr. + extrinsic
Piezoelectric constants d ij / 10–12 m / v
250
∗
200 d33 d33 intr. + extrinsic
150 3
∗
d15 intrinsic
100
∗
d33 intrinsic
50
qp
0
∗
d31 intrinsic
–50
d31 ∗
–100 d31 intr. + extrinsic
Fig. 23.3. Predictions for intrinsic and extrinsic piezoelectric constants vs. the
polarization state for barium titanate (after [22]). The polycrystalline material
consists of spherical grains with an orientation distribution according to the pol-
ing state described by the poling angle θp . Each grain is modeled by a lamellar
stack of 90◦ domains. The piezoelectric constants of a single domain are given by
d31 = −33.4 pm V−1 , d33 = 90.0 pm V−1 , d15 = 540 pm V−1
0.13
0.12
0.10
0.09
0.08
0.0 0.5 1.0
Electric field [kV/mm]
Fig. 23.4. Hysteretic behavior of electric displacement of BaTiO3 vs. applied electric
field (after [22]). Dissipation constant D0 = 100 kJ m−3
d d
wp = wd = ±D0 for ξ˙ >
< 0. (23.25)
dξ dξ
This condition ensures that the dissipated work must be supplied by the
release of potential energy.
If this scheme is invoked into the formalism, one obtains a threshold and
a hysteresis describing the loss properties due to frictional movement of do-
mains. An example is shown in Fig. 23.4. Thus, macroscopic loss measurements
can be correlated to the microscopic dissipation constant D0 .
The basic idea behind the model is to adopt the viscous concept known to
be successfully applicable in modeling polycrystal plasticity. For ferroelectric/
ferroelastic materials, it means that all possible domain orientations may si-
multaneously transform into any other possible orientation. Only the transi-
tion rate differs as a function of the driving force which is given by the gain
of free enthalpy. Thus, one can describe a gradual change of the polarization
state of the individual crystals and avoids the problematic sharp switching
transition, which is not realistic for a multidomain crystal and, additionally,
causes problems in the numerical evaluation. Moreover, it is easily possible to
include rate- and temperature-dependent behavior.
Let us denote the set of internal state variables by ξr with r = 1, 2, . . . , n,
where n is the number of states (one can find models with n = 2, 6, 8, 20, 42, . . .).
It is meaningful to restrict the variation range of ξ in the sense of weighting
factors (so an interpretation as volume fractions of certain special domains is
possible though not necessary for the model):
n
ξr = 1, 0 ξr 1. (23.26)
r=1
which the transition rates become large (in principal, these parameters may
depend on the transition type r–s). Furthermore, two empirical constants m
and α must be adjusted to experimental data. The second power term takes
into account the saturation of the transition if the weight of the starting state
becomes zero.
It is important to note that (23.30) is a unique function of state, so that in
numerical simulations no iteration is necessary when a certain load increment
is prescribed. Instead, one directly obtains the new domain configuration of
internal state variables by summing all transition rates:
n
ξ˙r = (+vsr − vrs ). (23.31)
s=1
4.0
0.01 Hz
3.5
3.0
0.1 Hz
STRAIN (x10–3)
2.5 1.0 Hz
2.0
1.5
1.0
0.5
0.0
–2 –1 0 1 2
E (MV/m)
4.0 0.01 Hz
0.1 Hz
3.5
3.0
1.0 Hz
STRAIN (x10–3)
2.5
2.0
1.5 1
1.0
0.5
0.0
–2 –1 0 1 2
E (MV/m)
Fig. 23.5. Simulated strain hysteresis curves under sinusoidal loading for soft PZT
(after [4]). Model parameters: n = 6, m = 8, α = 3, wc0 = 0. Comparison with
experimental results after [24]
532 W. Kreher
What we still need is an expression for the potential energy. Here, the most
simple assumption is provided by a linear combination of the potential energies
(r)
wp of certain domains, where r stands now for a particularly orientated
domain:
n
wp (f¯, ξ1 , . . . , ξn ) = pξr w(r) (f¯) (23.32)
r=1
23.5 Summary
Effective medium theories may be used to calculate internal fields on different
structural levels (domains, grains) so that the linear and non-linear material
behavior of piezoelectric materials can be modeled. Some applications have
been presented in this chapter and more examples may be found in the liter-
ature. Below some open problems and possible applications are listed:
• Pyroelectric/thermoelastic properties
• Evaluation for time-dependent properties
• Conduction problems
• Effect of free charges (screening of polarizations, modification of interac-
tion energy)
• Phase mixtures
• Relaxor behavior (diffuse domains)
References
1. G. Arlt, J. Mater. Sci. 25, 2655 (1990)
2. G. Arlt, Ferroelectrics 189, 91 (1996)
3. A.T. Bartic, D.J. Wouters, H.E. Maes, J.T. Rickes, R.M. Waser, J. Appl. Phys.
89(6), 3420 (2001)
4. A.Y. Belov, W.S. Kreher, J. Europ. Ceram. Soc. 25(12), 2567 (2005)
5. M.L. Dunn, M. Taya, J. Am. Ceram. Soc. 76(7), 1697 (1993)
6. M.L. Dunn, M. Taya, Proc. R. Soc. Lond. A 443, 265 (1993)
7. D.A. Hall, J. Mater. Sci. 36(19), 4575 (2001)
8. J.H. Huang, Int. J. Solids Struct. 33(20), 1439 (1996)
9. J.H. Huang, J.S. Yu, Compos. Eng. 4, 1169 (1994)
10. J.E. Huber, N.A. Fleck, J. Mech. Phys. Solids 49(4), 785 (2001)
11. J.E. Huber, N.A. Fleck, C.M. Landis, R.M. McMeeking, J. Mech. Phys. Solids
47(8), 1663 (1999)
23 Effective Medium Theories 533
T. Steinkopff
Nowadays, the most used commercial Finite Element software for solving
piezoelectric problems is based on the linear description of materials behav-
iour. The developers shrink from implementing the complex nonlinear be-
haviour which is essentially characterised by the ferroelectric and ferroelastic
hystereses. At the other end, the user which is interested in nonlinear analyses
is forced to program so-called user elements or to apply iterative algorithms.
The two lead to disproportionately excessive costs.
With the FE software COMSOL an innovative approach is given to take
the nonlinear behaviour into the numerical consideration. Examples of both
linear and nonlinear piezoelectric analyses will also be given here.
S = sE · T + d · E = (cE )−1 · T + d · E,
∼ P,
D = d · T + εT · E =
where cE are elastic coefficients (for constant electric field), εT are dielectric
coefficients (for constant mechanical stress), and d are piezoelectric coefficients
(so-called charge constants).
Generally, the coefficients are functions of the loading conditions, loading
direction, loading rate, and frequency:
E T ∂T ∂E
c , d, ε = f T, E, , ,... .
∂t ∂t
In cases of transient analyses, the variable t corresponds to the real time. In
cases of static analyses, it should be understood as pseudo-time, first of all to
describe the sequence and direction of load changes ∆E and ∆T .
24 Finite-Element Modelling of Piezoelectric Actuators 537
1,6
Polarisation / a.u.
1,5
∆P
1,4
∆E
Fig. 24.3. Dependence of coercive pressure on the parallel applied electric field [1]
Fig. 24.4. Strain asymmetry between tension and compression for a ferroelastic
material without grain-to-grain interaction. The total strain dependence on uniaxial
stress is shown. Experimental data from [1]
In the recent few years, a lot of micromechanical models have been de-
veloped. Models without grain-to-grain interaction are worth to estimate the
maximum values of strain and polarisation attainable by switching. Apply-
ing the ferroelectric considerations of [2] onto the ferroelastic behaviour, the
strain asymmetry between tension and compression of originally untextured
(unpoled) ceramics can be shown (Fig. 24.4). In the case of uniaxial tension
√ −1
and compression,
maximum ferroelastic strain values of 4π · 2 · Sspont and
√ −1
−4π · 1 − 2 · Sspont can be reached, respectively.
Obviously, these maximum values for both polarisation and strain cannot
be reached because of grain-to-grain, domain-to-domain interaction, and other
deformation mechanisms. Finite-element methods recommend by their selves
to take electromechanical interactions into account. Figure 24.5 shows results
of finite-element simulations based on ferroelastic domain switching [3] which
are compared to uniaxial data derived from bending experiments [4]. Because
of mechanical interaction, the strain asymmetry occurs not until stress values
of 40 MPa for this case of a soft PZT material with low coercive stress.
It is worth to mention that using FE methods for modelling domain switch-
ing effects, complex phenomena like auto-catalysis arise in a natural way [5].
From the point of engineering view, micromechanical models are only useful in
cases of simplified considerations [6]. For more complex simulations of piezo-
electric devices, the materials behaviour must be described phenomenologi-
cally [7]. Analogous to the classic elasto-plastic theory yield and hardening
parameters can be well defined for piezoelectric ceramics according to experi-
mental data [8]. Generalised yield stresses should be related to coercive stresses
and fields. Kinematic hardening is useful to describe both the ferroelectric and
ferroelastic hysteresis.
540 T. Steinkopff
80 From bending
experiment
60
40
20
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Strain / ‰
Electrostatics:
− ρ = 0,
div D (24.4)
with
= D(
D
E(ϕ)),
(24.5)
·
n = σ,
D (24.6)
Π(
a) = 12
aTu · K ·
au +
aTu · F
⇒ K ·
au = F
,
ϕ) = 12
aTϕ · C ·
aϕ +
aTϕ · Q
Π(
⇒ C ·
aϕ = Q.
542 T. Steinkopff
Piezoceramic
layer
(electrically driven) Thickness
direction
Neutral middle
layer
Piezoceramic layer
(short-circuit)
Fig. 24.6. Geometry, loading, and bending deformation of the trimorph actuator.
Colours correspond to the axial stress levels (red for tension, blue for compression
stresses)
trimorphs for which the deformation behaviour has been analysed by means
of finite-element methods (Fig. 24.6).
In the following, the static and dynamic behaviour of trimorph bending actu-
ators consisting of two active layers (top and bottom layer) and an inactive
middle layer (Fa. Argillon) will be discussed. Aspects of reliability are in the
foreground where piezoelectric actuators are used permanently for instance
in textile machines. There, a lot of several thousand bending actuators are
used under bi-directional operation conditions. Long-term investigations for
different driving conditions (Fig. 24.7) [10] show a non-monotonic dependence
of the mean-times-to-failure on the driving voltage: bending actuators driven
with 275 V (magenta) show a higher lifetime than actuators driven with 225
and 250 V (orange, blue). By means of finite-element analyses, this unexpected
behaviour can be understood.
Starting with static analysis, Fig. 24.8 shows typical deformation plots for
both the free elongation and the blocking state of a trimorph actuator carried
out with the FE software ANSYS c
[11]. Note that there is no significant
difference in the axial stress values between the free end (FE) and near to
the clamped region (CR) although failure is observed to be a result of under-
critical crack growth near the clamping region (CR).
A better understanding can be reached by considering the deformation and
stress over time behaviour corresponding to Figs. 24.9 and 24.10. In Fig. 24.9,
snapshots give an idea of the transient deformation behaviour. In textile ap-
plications, the deflection direction alternates with frequencies of about 50 Hz
by changing the piezoelectrically active layer (top or bottom layer). In the left
544 T. Steinkopff
0
without failure
ln ln 1/(1-F)
–1
–2
Weibull parameter @ driving voltage
**
m=0.92, N0=1.88e9 @ 200 V
–3 1.71 9.57e7 225 V
1.84 6.49e7 250 V
1.71 1.36e8 275 V
1.22 9.56e7 300 V
–4 1.85 5.32e7 325 V
1.86 2.17e7 350 V
2.16 5.99e6 400 V
2.05 6.51e6 450 V
–5
Fig. 24.7. Weibull plots for bending actuators which were driven with voltages be-
tween ±200 and ±450 V which corresponds to electric field strength up to 2 kV mm−1
ANSYS ANSYS
Clamping
FE MX
CR
Z
Y X
MN
Deflection
Fig. 24.8. Deflection behaviour of a bending actuator. Left: free actuator. Right:
clamped actuator. Keep in mind the different scaling for deflection
and right column, the transient behaviour of the bending actuator is shown for
the case that the piezoelectric top and bottom layer is driven, respectively. Af-
ter switching (driving changes from the top to the bottom layer at t = 20 ms),
inertial effects have a strong influence on the deformation behaviour and the
internal loading conditions. Dependent on driving frequency, in-phase and
out-of-phase deflection behaviour of the actuator can be observed.
For inappropriate switching moments, high dynamic tension stresses which
are of the order of the static stresses are observed near the clamping region
(Fig. 24.10). As a consequence for different operation conditions (voltage, fre-
quency, strikes yes/no), totally different stress levels and mean-times-to-failure
seem to be clear. Here, finite-element calculations are very helpful to analyse
24 Finite-Element Modelling of Piezoelectric Actuators 545
t [ms] t [ms]
0,2 19,4
1,0 20,2
1,8 21,0
2,6 21,8
3,4 22,6
4,2 23,4
5,0 24,2
5,8 25,0
6,6 25,8
7,4 26,8
8,2 27,4
9,0 28,2
9,8 29,8
10,6
..
.
σx,B(t)
[MPa] 30
20
10
0
x-Position
−10 11 m m
20 m m
−20 47 m m
0 5 10 t [ms]
Fig. 24.10. Signal and stress over time for different x positions in the active layer
of the trimorph actuator
546 T. Steinkopff
In the following, simple examples are given which demonstrates the influ-
ence of nonlinear elastic behaviour on the strain and stress distribution of a
trimorph actuator.
Starting point is an elastic stiffness Y that may depend linearly from stress
according to ferroelastic softening under tension stresses:
T
Y (T ) = Y0 − ∆Y · .
T0
To compare piezoelectric materials with different elastic behaviour, various
scenarios can be considered because of the relationship between the piezoelec-
tric charge coefficient e and voltage coefficient d:
e = Y · d ⇒ e(T ) = Y (T ) · d(T ).
For reasons of simplicity, two cases are considered here: d = constant (case A)
and e = constant (case B) materials. That means in case A, materials are com-
pared which have the same piezoelectric d coefficient but different nonlinear
elastic behaviour. In case B, the same procedure is applied for materials with
the same piezoelectric e coefficient. Figure 24.11 shows the axial stress Txx
and electrical field strength Ey distributions in dependence of the coordinate
in thickness direction (left: case A, right: case B).
In case A where materials with same d coefficient are compared, the influ-
ence of softening is much higher than that in case B (e = constant).
In Fig. 24.12, the influence of the softening parameter Ysoft = ∆Y /T0 on
the deformation behaviour is shown.
While the nonlinear materials behaviour considered here is very sim-
plified, the results give a first impression of its influence on the complex
electromechanical behaviour in bending actuators. For case B materials (e =
constant), the influence of the nonlinear elastic behaviour is much smaller
than that for case A materials (d = constant).
24.6 Conclusions
It could be shown that finite-element analyses are helpful for solving more
complex piezoelectric problems. To get initial estimates about the maximum
stress levels, linear finite-element calculations seem to be sufficient.
In the next years, one key aspect of finite-element activities in the field of
piezoelectric applications will be concentrated on finding practicable formu-
lations of the nonlinear materials behaviour not only in regard to the electro-
mechanical loading conditions but also in regard to history, frequency, and
time. From the author’s point of view, there is no way to find a unified mate-
rials law that can incorporate the influence of all these parameters.
40 20
548
20 Y-soft / Y-soft /
GPa 10 GPa
0 −8 0
−6 4
−5 6
−20 0 0 8
4 10
6
T. Steinkopff
−40 12
8
10
−60 12 −10
2,5 1,5
2,0 Y-soft / 1,0 Y-soft /
1,5 GPa GPa
−8 0,5 0
1,0 −6 4
−5 0,0 6
0,5 0 8
0,0 4 10
6 −0,5
12
−0,5 8
10 −1,0
−1,0 12
Bottom layer Middle layer Top layer −1,5 Bottom layer Middle layer Top layer
−1,5
Electric field strength (Ey) / [kV/mm]
Fig. 24.11. Distributions of axial stress (top) and electric field strength (bottom) for cases A and B materials (left and right, respectively)
with different nonlinear elastic behaviour
24 Finite-Element Modelling of Piezoelectric Actuators 549
1,6 0,9
1,4
1,2 0,8
1,0
0,7
Strain / 1e-3
Strain / 1e-3
0,8
0,6 0,6
0,4
0,5
0,2
0,0
0,4
−0,2
−0,4 0,3
−10 −5 0 5 10 15 −1 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Y-soft / GPa Y-soft / GPa
Fig. 24.12. Deformation behaviour of trimorph actuators with cases A and B ma-
terials behaviour (left and right, respectively). Keep in mind the different scaling for
strain
References
1. A. Schäuffele, K.H. Härdtl, J. Am. Ceram. Soc. 79, 2637 (1996)
2. T. Michelitsch, W.S. Kreher, Acta Mater. 46, 5085 (1998)
3. T. Steinkopff, Ferroelectrics 222, 383 (1999)
4. H. Weitzing, G.A. Schneider, Private communications, 1998
5. S.C. Hwang, J.E. Huber, R.M. McMeeking, N.A. Fleck, J. Appl. Phys. 84,
1530 (1998)
6. M. Kamlah, Continuum Mech. Themodyn. 13, 219 (2001)
7. C.M. Landis, J. Mech. Phys. Solids 50, 127 (2002)
8. T. Fett, D. Munz, Adv. Eng. Mater. 5, 718 (2003)
9. D.L. DeVoe, A.P. Pisano, J. Microelectromech. Syst. 6, 266 (1997)
10. M.P. Denzler, Ph.D. Thesis, Cuvillier, Göttingen, 2004
11. ANSYS, Inc., Southpointe, 275 Technology Drive, Canonsburg, PA 15317
12. COMSOL AB, Tegnérgatan 23, SE-III 40 Stockholm
25
Trends in Ferroelectric/Piezoelectric Ceramics
N. Setter
25.1 Introduction
Ferroelectrics and piezoelectrics are bound to keep their place as important
engineering ceramics. Capacitors and piezoelectric transducers are the most
widely spread applications of these group of materials. In thin-film form, ferro-
electrics and, more widely, polar materials have now been used for several years
in rf devices and in non-volatile memories. Components based on piezoelectric
films are also being developed and entered the market of various micro-sensor
and micro-actuator applications.
Barium titanate (BaTiO3 −BTO) and lead–zirconate–titanate (Pb(Zr, Ti)
O3 −PZT) continue to be the most important ferroelectric and piezoelectric
materials in commercial devices. New materials have been introduced con-
stantly. While during 1970s–1980s, those were the relaxors (with giant elec-
trostriction), the 1990s saw the introduction of relaxor ferroelectrics such
as Pb(Zn1/3 Nb2/3 )O3 –PbTiO3 (PZN–PT) increasing the coupling coefficient
from ≈0.7 of PZT to ≈0.90 and over [1]. In the last decade, AIN became
an important piezoelectric material due to its use in piezoelectric electronic
components (bulk acoustic wave devices) for mobile communications [2].
SrBi2 Ta2 O9 (SBT) competed with PZT in ferroelectric non-volatile memory
(FRAM) applications with a superior fatigue resistance [3]. Barium–strontium
titanate (BST) was investigated for DRAM applications and finally seems to
have found market in tunable ferroelectric capacitors for microwave compo-
nents. In 2004, Toyota published results showing that a lead-free material,
modified (K, Na)NbO3 (KNN) can potentially reach properties of the best
lead-containing ferroelectrics [4].
The processing technology of piezoeoelectrics/ferroelectrics has been ada-
pting itself to new needs too: finer powders became available [5], thick film
technologies for piezoelectrics have been developed [6], and textured growth
(the growth of oriented ceramics) gained importance [7]. Following the ex-
cellent properties of piezoelectric single crystals, a method of transforming
554 N. Setter
piezoelectric ceramics into crystals has been reported with expected improve-
ment in composition homogeneity control [8]. Processing of thin films by both
physical and chemical methods reached a certain level of maturity with a
rather tight control of composition (a challenge in particular in lead-containing
compounds), stress, and orientation [9]. Recently, the production of one-
dimensional (1D) nano-structures of ferroelectric and piezoelectric materials
has been reported [10].
Trends that point to directions of future developments in the field of piezo-
electrics and ferroelectrics are summarized below.
Fig. 25.1. Effect of compressive stress σ3 = 0 (dashed line) and −1.79 GPa at
T = 300 K on anisotropic piezoelectric enhancement in PbTiO3 . At σ3 = 0 Pa,
values for d∗33 (dashed line) are multiplied by 10. For σ3 close to the coercive stress
(approximately −1.9 GPa), values of d∗33 are strongly sensitive to input parameters
and vary between 103 and 104 [21]
The increase in operation frequency for wireless communication and the in-
terest in miniaturization and integration motivated the recent replacement of
many of the discrete surface acoustic wave components by bulk acoustic wave
ones based on piezoelectric thin films. This has been discussed in Chap. 15.
At present, the thin films are mostly AlN. In spite of the limited value of its
coupling coefficient, AlN has advantages of a high sound velocity and ease
of process integration. It is clear that search of new materials will continue:
one possible direction is a further exploration of ZnO. This material ‘lost’ the
competition with AlN due to its sensitivity to defects. Very recently, a spec-
tacular improvement of its piezoelectric d33 coefficient was reported through
appropriate doping [23]. If this will turn to influence also the coupling coeffi-
cient, it may be well an interesting material to pursue. At present the origin
of this improvement is obscure. Another direction awaiting further work is the
development of tunable filters by the use of paraelectric materials under bias
field [24].
25 Trends in Ferroelectric/Piezoelectric Ceramics 557
Fig. 25.2. Penetration of the ionic polarization from the ferroelectric into the metal–
oxide electrode. (The gradient in shading represents the concentration of free charge
carriers, while the solid line represents the absolute value of polarization.) First-
principles investigation explains this way the improved remanent polarization of
ferroelectrics capacitors with oxide electrodes relative to metallic (non-polarizable)
electrodes [38]
all, better insight into behaviour at surfaces and interfaces and behaviour of
systems containing piezoelectrics could result in accelerated development of
piezoelectrics and ferroelectrics.
Fig. 25.3. KNN crystal grown from dense ceramics using template-assisted Solid
State Single Crystal Growth (SSCG). (Courtesy of J. Fisher, A. Bencan and
M. Kosec, Jozef Stefan Institute, Ljubljana)
Increase thickness
Second
Second layer
layer
First layer
Fig. 25.4. Full strain relaxation and dislocation confinement during two-step growth
of STO. The inset is a TEM micrograph of the film cross section showing the dislo-
cation confinement to the first layer [46]
and co-workers obtained BaTiO3 films with Curie temperature of the order of
400–600◦ C (while Tc of BaTiO3 crystals is 120◦ C) which enhanced the room
temperature ferroelectric properties [44]. The properties were similar to those
previously predicted by LGD calculations [45]. (2) In another work, the strain
in epitaxial SrTiO3 (STO) films was controlled by a dislocation relaxation
mechanism using a two-step growth process, where the crystallization of a
pre-deposited very thin amorphous layer of identical composition acted as an
absorption layer for the misfit strain (Fig. 25.4). Two-step grown films had a
25 Trends in Ferroelectric/Piezoelectric Ceramics 561
25.4.3 Self-Assembly
Composite (predicted)
0.4
(measured)
0.2
100µm
Pure
ferroelectric
(a) 0.1
0 100 200 300 400 500
Fig. 25.5. (a) Optical micrograph of a fabricated composite structure. (b) Compar-
ison between the relation between permittivity and tunability in pure ferroelectric
and in parallel-connected composite (calculated and measured) [62]
25 Trends in Ferroelectric/Piezoelectric Ceramics 563
300
150
[nm]
0 200
0 100
Fig. 25.6. 3D visualization of the piezoelectric activity of KNbO3 nano-rod and the
existence of ferroelectric domains with piezoelectric activity [84]
nano-wires in an up-scalable way [74]. BTO and STO nano-tubes were pro-
duced by reacting titania nano-tubes with barium and strontium hydrox-
ides [75]. A hydrothermal growth of potassium niobate (KNO) nano-wires
was demonstrated [76]. Hydrothermal method was used also for the fabrica-
tion of arrays of BTO nano-tubes on titanium substrate [77]. Very little has
been reported so far on the growth mechanism of piezoelectric nano-tubes and
nano-wires: while KNO seems to grow by a nucleation and growth mechanism
in sub-saturated liquid [78].
The ferroelectric to paraelectric transition was recently reported on BTO
nano-wires [79] but the crystalline structure of ferroelectric 1D structures has
been only slightly explored [80,81]. Domain walls formation and disappearance
upon cycling around the ferroelectric phase transition were demonstrated in
PZT nano-wires [82]. Rare are yet the detailed reports on the piezoelectric
properties of nano-tubes and nano-wires but switching of the ferroelectric
polarization and local piezoelectric activity have been detected using force
microscopy (Fig. 25.6) [83, 84]. The field of 1D small piezoelectrics is at its
infancy.
by scanning an AFM tip over the wires and the generated mechanical energy
was converted into electrical energy by the piezoelectric effect. The estimated
electrical energy produced was ≈10 pW µm−2 . The collection of the energy
is done in an elegant way taking advantage of the semiconductor nature of
the nano-wires used: the nano-wire/conductor contact acts as a diode and
the charge generated piezoelectrically is collected in a unidirectional way. The
current leading idea is to convert mechanical energy that is available contin-
uously such as body movement into electrical energy, so that self-powering
can be used for wireless low power consumption devices. Development in this
direction was shown recently by Zhong Lin Wang et al. [88] who created a
textile structure that contains the ZnO nano-wires and metal wires for the
current collection.
An inherent exploitation of nano-technology can be obtained when the phe-
nomena at stake are coupled at the same scale. This was shown recently [89]
in utilization of piezoelectric semiconductors to build cantilevers in which the
width of the depletion region (hence the piezoelectric active region) was ma-
nipulated by an electric field, creating positive or negative deflection according
to the extent of the field applied, which was translated into an increased vibra-
tion amplitude relative to a stationary piezoelectric cantilever. It is a similar
effect to that obtained by bi-layer actuators, but on a nano-meter scale, tun-
able, and processed via standard integrated semiconductor technology, where
the electrodes and the piezoelectric active regions are all based on variation
of the doping in a GaAs cantilever. This mechanism functions only when
the dimension of the cross section of the cantilever is of the same order of
magnitude as that of the depletion region, namely, sub-µm. The authors fur-
ther demonstrate various devices based on piezoelectric GaAs nano-structures
such as nano-electromechanical XOR (exclusive-or) logic gate and resonance-
frequency control.
It is reasonable to expect that piezo-NEMS (nano-electromechanical sys-
tems) will develop further with possible applications in sensors, transducers
and computing applications. The interest of piezoelectric semiconductors is
also likely to intensify due to the small dimensions of the components.
25.7 Conclusion
Major trends in materials and processes in the area of ferroelectrics and piezo-
electrics were outlined above. Emphasized directions were new materials, in
particular lead-free and high-temperature ones as well as materials for com-
munication technologies and MEMS–NEMS. The increased role of ab initio
calculations in prediction of new material systems, properties and functioning
of components and devices was pointed out. Trends in processing include ‘do-
main engineering’, nano-composites, self-assembly of small elements on sub-
strates, and new 1D structures.
566 N. Setter
The question of new applications was not discussed although it is clear that
the entire field is driven by applications. Intensification of use of piezoelectrics
and ferroelectrics can be foreseen in several areas: with the present strong
development of polymer electronics, it is likely that ferroelectric polymers will
grow in importance. Increased investments into medical diagnostic tools may
be projected as an opportunity for a further increased use of piezoelectrics in
this domain. Certainly, highly performing lead-free materials will be welcome
here. Increased automation and a more stringent control on efficient energy use
and on clean environment are indicative for an intensified use of piezoelectric
sensors and actuators for these applications.
Acknowledgement
References
1. S.-E. Park, T.R. Shrout, J. Appl. Phys. 82, 1804 (1997)
2. M.-A. Dubois, P. Muralt, Appl. Phys. Lett. 74, 3032 (1999)
3. J.F. Scott, C. Araujo, Science 246, 1400 (1992)
4. Y. Saito, H. Takao, T. Tani, T. Nonoyama, K. Takatori, T. Homma, T. Nagaya,
M. Nakamura, Nature 432, 84 (2004)
5. Y. Sakabe, Y. Yamashita, H. Yamamoto, J. Eur. Ceram. Soc. 25, 2739 (2005)
6. P. Tran-Huu-Hue, F. Levassort, F. Vander Meulen, J. Holc, M. Kosec,
M. Lethiecq, J. Eur. Ceram. Soc. 21, 1445 (2001)
7. T. Takeuchi, T. Tani, Y. Saito, Jpn. J. Appl. Phys. 1, 9B, 5553 (1999)
8. H.Y. Lee, in Piezoelectric Single Crystals, ed. by S. Trolier-McKinstry,
L.E. Cross, Y. Yamashita p. 160, (2004)
9. S. Trolier-McKinstry, P. Muralt, in Electroceramics Based MEMS, ed. by
N. Setter (Springer, Berlin, 2005), pp. 199–215, Chap. 10
10. J.J. Urban, W.S. Yun, Q. Gu, H. Park, J. Am. Chem. Soc. 124, 1186 (2002)
11. E. Hollenstein, M. Davis, D. Damjanovic, N. Setter, Appl. Phys. Lett. 87,
182905 (2005)
12. G.-Z. Zang, J.-F. Wang, H.-C. Chen, W.-B. Su, C.-M. Wang, P. Qi, B.-Q. Ming,
J. Du, L.-M. Zheng, S. Zhang, T.R. Shrout, Appl. Phys. Lett. 88, 212908 (2006)
13. T.R. Shrout, S.J. Zhang, J. Electroceram. 19, 111 (2007)
14. B. Jaffe, R.S. Roth, S. Marzullo, J. Appl. Phys. 25, 809 (1954)
15. T.R. Shrout, R. Eitel, C. Randall, in Piezoelectric Materials in Devices, ed. by
N. Setter (2002), pp. 413–432
16. N. Klein, D. Damjanovic et al. J. Appl. Phys. 102, 014112 (2007)
17. K. Nakamura, T. Tokiwa, Y. Kawamura, J. Appl. Phys. 91, 9272 (2002)
18. S. Wada, S. Suzuki, T. Noma, T. Suzuki, M. Osada, M. Kakihana, S.-E. Park,
L.E. Cross, T.R. Shrout, Jpn. J. Appl. Phys. 1 38, 5505 (1999)
19. X. Du, U. Belegundu, K. Uchino, Jpn. J. Appl. Phys. 1, 36, 5580 (1997)
25 Trends in Ferroelectric/Piezoelectric Ceramics 567
80. A. Magrez, E. Vasco, J.-W. Seo, C. Dieker, N. Setter, L. Forro, J. Phys. Chem.
B 110, 58 (2006)
81. G. Suyal, E. Colla, R. Gysel, N. Setter, Nanoletters 4, 1339 (2004)
82. J. Wang et al., Appl. Phys. Lett. 90, 133107 (2007)
83. W.S. Yun, J.J. Urban, Q. Gu, H. Park, Nanoletters 2, 447 (2002)
84. G. Suyal, E. Colla, R. Gysel, N. Setter, Nanoletters 4, 1339 (2004)
85. Z. Wang, J. Hu, A.P. Suryavanshi, K. Yum, M.-F. Yu, Nanoletters 7, 2966 (2007)
86. Zhong Lin Wang, Jinhui Song, Science 312, 242 (2006)
87. W.S. Su, Y.F. Chen, C.L. Hsiao, L.W. Tu, Appl. Phys. Lett. 90, 063110 (2007)
88. Xudong Wang, Jinhui Song, Jin Liu, Zhong Lin Wang, Science 316, 102 (2007)
89. S.C. Masmanidis, R.B. Karabalin, I. De Vlaminck, G. Borghs, M.-R. Freeman,
M.L. Roukes, Science 317, 780 (2007)
Index