Ferroelectric Thin Films for Micro-Sensors and Actuators: A Review

Abstract

This paper reviews deposition, integration, and device fabrication of
ferroelectric PbZrxTi1-xO3 (PZT) films for
applications in microelectromechanical systems. As examples, a piezoelectric
ultrasonic micromotor and pyroelectric infrared detector array are presented.
A summary of the published data on the piezoelectric properties of PZT thin
films is given. The figures of merit for various applications are discussed.
Some considerations and results on operation, reliability, and depolarization
of PZT thin films are presented.

Full text

J. Micromech. Microeng. 10 (2000) 136–146. Printed in the UK PII: S0960-1317(00)09328-1
Ferroelectric thin films for
micro-sensors and actuators: a
review
P Muralt
Ceramics Laboratory, Materials Department, EPFL Swiss Federal Institute of Technology,
CH-1015 Lausanne, Switzerland
E-mail: paul.muralt@epfl.ch
Received 15 December 1999
Abstract. This paper reviews deposition, integration, and device fabrication of ferroelectric
PbZrxTi1−xO3(PZT) films for applications in microelectromechanical systems. As examples,
a piezoelectric ultrasonic micromotor and pyroelectric infrared detector array are presented.
A summary of the published data on the piezoelectric properties of PZT thin films is given.
The figures of merit for various applications are discussed. Some considerations and results
on operation, reliability, and depolarization of PZT thin films are presented.
1. Introduction
During recent years, the study of microelectromechanical
systems (MEMS) has shown significant opportunities for
miniaturized mechanical devices based on thin-film materials
and silicon technology. In mainstream MEMS technology,
materials are restricted to those used in microelectronics in
order to profit from materials and processes that are readily
available. In addition, the same fabrication facilities are often
used for both MEMS and microelectronics, forbidding any
application of materials with fast diffusing ions. However,
in order to cover the whole range of physical phenomena
that are exploitable for sensors and actuators, it is necessary
to add a variety of functional materials to the existing base
materials.
An important family of functional materials are
ferroelectrics or, more generally, polar materials. Their
piezoelectricity can be used in sensors, actuators, and
transducers; their pyroelectricity is employed in infrared
detectors. In this article emphasis is given to lead zirconate
titanate (Pb(ZrxTi1−x)O3or PZT), a solid solution of
ferroelectric PbTiO3and antiferroelectric PbZrO3[1]. As
a bulk material, it is plainly the piezoelectric ceramic,
and is used in various sensor, actuator, and transducer
applications [2]. Its use in MEMS devices was very much
delayed by integration difficulties. During recent years
much progress has been made in this area. The main
impetus for its integration onto silicon was the prospect of
non-volatile, radiation-robust memories. While the early
work on PZT thin-film actuators [3] still suffered from
integration problems, more recently a number of devices
have successfully been fabricated and characterized. Among
the piezoelectric devices, these include cantilever actuators
[4, 5], probes for atomic force microscopy [6], ultrasonic
micromotors [7, 8], micropumps [9], ultrasonic transducers
for medical applications [10], and linear actuators [11]. The
PZT solid-solution system also contains compositions with
fairly good pyroelectric properties (x<0.3). For this reason,
PZT has also served in establishing the first generation
of pyroelectric thin-film sensors. The efforts concentrated
mainly on infrared (IR) detector arrays [12–14] for imaging,
security systems, and gas sensor applications.
2. Thin-film deposition
Most of the existing coating techniques have been
investigated for the deposition of PZT. The early work
was carried out by means of physical techniques such as
ion beam sputtering [15], rf planar magnetron sputtering
[16, 17] or dc magnetron sputtering [18]. Metal-organic-
chemical vapor deposition (MOCVD) was the first chemical
method to follow [19–23]. Chemical solution deposition
(CSD), including sol-gel routes [24, 25] and metal-organic-
decomposition [26, 27], were investigated next, and, finally,
pulsed laser deposition (PLD) was also applied [28–30].
Today there is a clear trend to apply MOCVD or CSD.
The conformal coverage of the first is better suited for
integrated circuit (IC) production, whereas CSD is the
cheapest technique for small-scale production as required in
the sensor industry. CSD techniques need post-annealing
treatments for the crystallization of the film. All of the other
methods allow in situ growth. The main growth phenomena
of PZT can be roughly understood in terms of a few key
features of the PbO–ZrO2–TiO2system, independently of
the deposition method.
(1) Nucleation and growth of the perovskite require a rather
precise stoichiometry, otherwise competing phases with
fluorite and pyrochlore structures nucleate (see, e.g.,
[31]).
0960-1317/00/020136+11$30.00 © 2000 IOP Publishing Ltd

Ferroelectric thin films for micro-sensors and actuators
(2) Lead ions or PbO molecules that are not incorporated
into the perovskite lattice exhibit high diffusivities and
volatility above 500 ◦C. The PbO vapor pressure above
PbO is approximately 100 times larger than above PZT,
and amounts to 1.1 Pa at 600 ◦C [32].
(3) The activation energy for nucleation of the perovskite
(4.4eV/unit cell) is considerably larger than for its
growth (1.1 eV) [33].
PZT thin films can be grown at much lower temperatures
than the typical sintering temperatures of bulk ceramics.
This is due to much smaller diffusion distances needed
with thin-film processing techniques, which provide for
a homogeneous, stoichiometric mixture on the molecular
level. However, second phases compete with the perovskite
structure when deviations from stoichiometry cannot be
compensated for by local diffusion. The fluorite phase
appearing at low temperature mostly in sol-gel deposited
films have the formula Pb2+xTi2−xO7−yand may result
from oxygen deficiency at the initial stage of annealing, at
which residual hydrocarbon groups have to be volatilized by
oxidation. Similar problems arise with the MOCVD. There is
a low-temperature limit for pyrolysis and oxidation reactions,
which can differ greatly for the different species (PbO, TiO2,
ZrO2) [34, 35]. Physical methods may allow to grow films
below 500 ◦C. However, the crystallization is not tolerant
to lead excess, and the oxygen content does not necessarily
adjust to the stoichiometry by oxygen exchange with the
gas phase. At high temperatures, the PbTi3O7pyrochlore
structure is common to all methods. The reason for its
formation is heavy lead loss. There is also a compositional
effect. The activation energies for nucleation and growth
increase with Zr content [33]. All of the processes usually
work with an excess of lead, in order to compensate for
lead loss before perovskite formation. With the in situ
deposition techniques, one observes a self-stabilization of the
lead content at the stoichiometry above a critical temperature,
even for large quantities of excess lead flux. The critical
temperature depends on the deposition method and amounts
to 700 ◦C for PbTiO3grown by MOCVD [36, 37], and was
found to be lower for sputter deposition (550–600 ◦C) due to
plasma effects [38, 39]. In case of post-annealed films, rapid
thermal annealing (RTA) proved to be a good technique to
provide a quick formation of the perovskite, thus reducing
lead loss [40]. Practically all films applied in devices are
presently deposited between 525 ◦C and 700 ◦C.
Lower deposition temperatures would greatly help to
facilitate device fabrication. It seems that some of the
coating methods have the potential to lower this temperature.
In situ processing clearly offers an advantage for achieving
low deposition temperatures. Traces of the perovskite
phase can be found down to 200 ◦C in films deposited by
magnetron sputtering [41]. Good pyroelectric thin films of
PbTiO3are reported to grow at 450◦Cbyin situ multi-target
sputtering, albeit with some traces of a PbO second phase
[42]. MOCVD seems to manage such low temperatures on
suitable substrates [21]. Activation by photon and electron
absorption or other impinging plasma species reduces the
thermal activation energies, and thus allows reduction of
the process temperature. This was shown by plasma and/or
UV assisted processing [43, 44]. Also, bombardment with
0
20000
40000
60000
80000
15 20 25 30 35 40 45 50
2-theta (°)
PZT(111)
Pt (111)
1.3µm PZT 45/55
in-situ sputter deposition
x 100
Figure 1. X-ray diffractogram for a 1.3µm thick (111) textured
PZT film grown by sputter deposition and TiO2seeding. No other
peaks than (111) and higher orders of it can be found (Courtesy
S Hiboux).
low-energy oxygen ions was found to be a possible way to
improve PZT thin films [45, 46] grown at lower temperatures.
Inherent to low-temperature processing is the need for precise
lead concentration control, as excess lead does not evaporate.
In plasma assisted CVD of PbTiO3at 500 ◦C, the pure
perovskite forms precisely at Pb stoichiometry only. Lead
deficiency leads to pyrochlore, and lead excess to PbO
second phases [47]. For low-temperature growth, the oxygen
concentration, or in other words, the Pb valence needs to be
tuned, too. This requirement is correlated with the fact that
PbO can be oxidized further at temperatures below about
450 ◦C, yielding Pb3O4to PbO2[48].
The third key feature mentioned above means that the
growth is nucleation controlled [49], i.e. heterogeneous
nucleation is preferred over homogeneous nucleation. This is
very important for sol-gel deposition techniques, as it allows
one to obtain a columnar film microstructure nucleated at
the bottom electrode. Nucleation controlled growth also
permits one to choose the texture of the film by suitable
electrodes with seeding functions [50–52] or to reduce the
deposition temperature [53]. With in situ sputter deposition,
(100) nucleation of PbTiO3works very well on Pt(111),
whereas PZT(100) is more difficult to obtain. However, on
a PbTiO3{100}template PZT grows as well in (100) [38].
A selectivity of the seeding by platinum (111) is obtained
by an additional 2 nm thick, hot-deposited TiO2seed layer,
which yields the (111) orientation for PZT deposited by in situ
sputtering and also by sol-gel deposition [52] (see figure 1).
3. Integration
High-quality PZT films cannot be grown directly on silicon
[54]. Buffer layers are needed to prevent interdiffusion and
oxidation reactions. For most applications, the PZT film has
to be grown on an electrode, which obviously should neither
oxidize nor become insulating. The most often reported
materials include platinum [55], and the metal oxides RuO2
(rutile structure) [56], SrRuO3, and (La,Sr)CoO3(LSCO,
perovskite structure) [57]. Usually, the chemical barrier
function is provided by two or more layers, including the
electrode. PZT/Pt/Ti/SiO2/Si is the most widely applied
sequence, where the Ti is needed as an adhesion layer.
137

P Muralt
Figure 2. A1µm thick PZT thin film deposited by the sol-gel
technique onto a Pt/Ta bottom electrode film (the bottom layer is
Si3N4, on the top is SiO2).
In micromechanics, Si3N4is very often used instead of
the thermal oxide. In this case, a Ta adhesion layer has
proved to be better than Ti (see figure 2) [7]. The adhesion
layers play an important role in the diffusion phenomena that
proceeds during processing. Platinum does not inhibit the
diffusion of Ti to the PZT side, where it reacts with oxygen
and serves as nucleation centers for PZT [58]. Pockets of
TiO2have also been observed between the Pt grains [59].
There is also evidence that oxygen migrates along the grain
boundaries through the platinum film and reacts with the Ti
layer [60]. For stable electrodes, the latter has to be pre-
oxidized. A comparative study of Ti, Zr, and Ta adhesion
layers showed that the most diffusing species is always the
one in the corresponding oxide, i.e., Ti in TiO2, O in ZrO2
and Ta2O5[59]. Barrier schemes with insulating films, such
as SiO2and Si3N4, cannot be applied when a direct electrical
contact to a silicon or metal substrate is required. It would
be convenient to use, for example, a metal membrane also
as a conductor layer to reduce resistive losses in ultrasonic
applications. In this case, an additional barrier layer is needed
that stops oxygen diffusion without getting too resistive.
Possible solutions are nitrides consisting of mixed conductive
and insulating compounds such as TiAlN [61] and TaSiN
[62]. Such films may resist to oxidation up to about 700 ◦C.
Other possibilities have been identified in combination with
RuO2and oxide scale forming oxides such as Cr [63]. This
combination allowed the growth of PZT on very reactive
refractory metals, such as zirconium [64], without oxidizing
the latter.
4. Device examples
4.1. Ultrasonic micromotor
As compared to other techniques, ultrasonic actuation bears
various strong points. These are a flat profile (small
thickness), and a comparatively high torque at low speeds.
Torques of almost 1 µNmat4V
rms,or0.3µNmV
−1
rms have
been achieved [8] with millimeter-sized rotors and PZT thin-
film stators. This is more than has been demonstrated with
electrostatic micromotors (scaled to the same motor size).
The low speeds (typically 5 rpm at 2 Vrms) of ultrasonic
micromotors obliterates the need for gearboxes. These are
very difficult to manufacture in submillimeter dimensions,
and tend to reduce the output torque to zero because of
unsolved friction problems. In some of the applications such
as for wristwatches, the flat profile of ultrasonic micromotors
is an additional advantage. Ultrasonic micromotors have
indeed some very attractive features for turning the date
Si wafer
membrane
spacer
rotor
g
ear
centrin
g
wheel
electrode
fin
Figure 3. Hybrid construction of an elastic fin micromotor (from
[7]). The silicon wafer is upside-down, the rotor moves inside the
etched cavity. The operation principle of the fins is shown in the
inset.
wheel of a wristwatch: flatter profiles than today’s motors,
high enough torques (1 µN m), and low enough supply
voltages for battery operation. The disadvantage of ultrasonic
micromotors lies in the fact that the coupling between the
stator and the rotor is based on frictional forces, and that
there is no constant relation between stator excitation and
the output speed, as with a stepper motor. Complicated
phenomena of gliding, slipping, and sticking of static as
well as dynamic friction may happen. However, most of the
problems can be solved with a closed-loop regulation based
on angular position detection [65, 66].
The example given here deals with ultrasonic stators
designed for a hybrid version of an ‘elastic fin’micromotor, as
depicted in figure 3. The motor type was demonstrated with
bulk PZT, and proposed for down scaling by Kurosawa et al
[67]. The term ‘hybrid’refers to the fabrication principle
and means in this case that the rotor was fabricated and
assembled to the stator by classical means, and that only the
stator was obtained with silicon micromachining techniques.
This stator consisted of a silicon frame keeping a thin silicon
membrane, which was covered by a PZT thin film. The latter
serves to excite standing flexure waves in the membrane.
The amplitude of this vibration is proportional to e31,f U,
where Uis the applied voltage, independent of the thickness
of the piezoelectric thin film. The piezoelectric coefficient
e31,f =eSST
31 is introduced as an effective coefficient by
taking care of the fact that the film is free to change its
thickness, but clamped in the plane [68] (the ordinary e31 is
defined for clamping in all directions: eSSS
31 , see below). The
deflections of the membrane are converted into a rotation by
a rectifying rotor, which has tilted elastic legs or fins. This
rectification works in principle as follows.
When the membrane moves towards the elastic fins, the
latter are compressed and bent. In this phase the fins do
not slip because of frictional forces. So the rotor turns in
order to release the compression. When the membrane moves
away from the fins (decompression) the frictional forces are
minimal and the fins glide forward, following the body of
the rotor. The 3.5 mm diameter rotor was cut out by laser
from a metallic foil. The desired tilt angle of the legs was
achieved by molding in a pressing machine. A steel axle
of 0.25 mm diameter is clamped into the rotor and centered
by a centering wheel resting on spacers. On the top of the
138

Ferroelectric thin films for micro-sensors and actuators
Silicon (100)Si3N4/SiO2
SiO2
PZT
Pt/Ta
Au/Cr
Au/Cr
etched cavity
Figure 4. Schematic diagram of the structure and major
fabrication steps of the stator with bottom electrode contact, top
electrode, and low-dielectric SiO2layer (according to [8]).
axle, a wheel was fixed which serves as a load to increase the
friction and the moment of inertia for torque measurements.
The normal force between the rotor and the stator was varied
with weights on the gear.
The membrane of the stator was obtained by silicon bulk
micromachining in hot KOH (see figures 6 and 7 below).
A stress compensated bilayer of thermal oxide (SiO2) and
LPCVD nitride (Si3N4) was applied as masking layer. The
first lithographic step consisted in opening this masking layer
by dry etching (CF4) for the definition of the membranes.
Pt/Ta bottom electrode and PZT deposition followed as the
next steps. A sacrificial Au/Cr top electrode pattern serving
as an etch stop and a PZT protection coating were deposited
and patterned. A silicon dioxide thin film was then deposited
by sputtering (see figures 2 and 4). This film reduces the
parasitic capacities of the conductor lines and contact pads.
The silicon dioxide was patterned with dry etching (CF4)
to liberate the top electrode contacts. The sacrificial Au/Cr
contacts were needed in this step to avoid deterioration of the
PZT surface, i.e. the formation of a low-dielectric fluorinated
film. This sacrificial layer was removed and the complete
top electrode system, including conductors and pads was
deposited.
The electrode system (see figure 5) was designed to
optimally excite the most effective vibration mode B10. This
mode exhibits one circular node. The outer ring-shaped
maximum was placed to coincide with the fin position. The
motor was run with either a single ac supply on the annular
electrode, or with two ac voltages having a relative phase
shift of 180◦to the annular and central electrodes. With the
second version a larger coupling coefficient is achieved.
Stators with 1 µm thick sol-gel deposited PZT 45/55
of (111) orientation have been investigated for various
thicknesses (15–100 µm) of the passive silicon part. The
piezoelectric coefficient typically amounted to 6 cm−2for
Figure 5. View of top electrodes (darker parts) and contact lines
and pads (bright). The central and outer electrodes have two
connections each. The bottom electrode connection is in the lower
left corner of the image (from [8]).
0
20
40
60
80
100
120
0 5 10 15 20 25 30
PZT 0V dc bias
PZT 2V dc bias
PZT 4V dc bias
AlN
angular velocity [rad/s]
ac-voltage [Vrms ]
Figure 6. The measured angular velocity of the elastic thin-film
micromotor for a PZT stator (33 kHz, 1 µm PZT 53/47) and an
AlN stator (38 kHz, 2 µm AlN). Various dc bias voltages were
applied in case of the PZT stator [69].
these films. In figure 6 the speed against applied ac voltage
is displayed for a PZT and AlN stator. The application
of an additional dc bias increases the polarization and the
piezoelectric constant, and thus the rotation speed of the
motor. There is a threshold voltage of 0.5–1.0V
rms below
which there is not enough amplitude to initiate the rotation.
Above the threshold, the speed increases linearly with the
applied ac voltage, and thus the deflection amplitude. The
torque at zero speed saturates at some level, depending on the
applied normal force: at 0.4–0.44 µN m with 3–4V
rms and
230 mgf (2.3 mN) normal force; at 0.9µNmat4V
rms and
670 mgf (6.7 mN) normal force (see figure 7). These values
are equivalent to the frictional torque [67] for the applied
normal force. While it is difficult to describe in detail the
transmission of force by the elastic fins, the experiments show
a very simple relation between the output power and the stator
power. In a wide range of frequency (20–100 kHz), the output
power is simply a constant fraction of the stator power. The
frequency was varied through thinning down, in steps, the
silicon thickness.
The output power of the motor was typically measured
as 1–2µWV
−1
rms, independently of the silicon thickness.
139

P Muralt
0
0.2
0.4
0.6
0.8
1
01234567
torque M0 (µNm)
normal force (mN)
operatin
g
conditions:
4 Vrms, - 2 V bias
Figure 7. Zero-speed output torque of the micromotor as a
function of the normal force (after [8]).
This is for the thinnest membrane (i.e. 15 µm thick silicon)
4×10−4times the reactive power delivered to the capacitance,
and hence also smaller than the dielectric loss power
(tan δ=0.03). The square of the coupling constant k2,
which determines the mechanical power in the stator, was
determined by calculations to be around 2 ×10−3. The
transmission efficiency between the stator and the rotor
therefore seems to be rather small (20%), and the overall
efficiency turned out to be less than 1%.
4.2. Infrared detector array
Pyroelectricity is one of the best performing principles for
the detection of temperature changes. Bulk crystals and
ceramics have therefore been used for many years to fabricate
thermal IR detectors. They were, and still are, applied
for contactless temperature measurement, security detectors
(intruder alarms) and human presence sensors. With respect
to semiconductor devices, thermal detectors are competitive
in the important wavelength interval from 8–12 µm. Their
special attraction lies in the fact that they do not need cooling.
Thermal detectors are too slow to be used in IR imaging
with scanning mirrors. However, they are fast enough if
two-dimensional (2D) arrays can be realized. In this case
the read-out rate for each pixel is identical to the frame rate
(30–50 Hz). Surface micromachining techniques combined
with the thin-film deposition of pyroelectric thin films
allowed the realization of such 2D arrays, in a monolithic
way, directly on the read-out chip [13]. In this work, we
focus on a simpler structure, a 1 ×64 (or 1 ×50) linear array
for an IR spectrometer [70, 71].
With their small thermal capacity (H), thin films have
a considerable advantage over bulk pyroelectric detectors.
The reason is that the response at frequencies above the
inverse thermal time constant H/G, where Gis the thermal
conduction, is proportional to 1/H . However, Gmust be
enough small so that the the inverse thermal time constant is
indeed lower or equal to the operation frequency (normally
the chopper frequency). For operation at a few tens of
hertz, the heat conductivity must be already quite small to
achieve this condition. The role of silicon micromachining
techniques is thus to provide good a thermal insulation.
Silicon is a too good thermal conductor and thus a ceramic
Figure 8. Electro-chemically deposited black platinum grown on
aCr–Au top electrode on a PZT/Pt/Ta/Si3N4/SiO2layer stack [74].
Top electrode
IR absorbing layer
Pb(Ti,Zr)O3
Si3N4/SiO2
Pt
etched silicon cavity
silicon
Figure 9. Typical cross section structure of the pyroelectric
elements of a linear array on a thin membrane fabricated by means
of micromachining (from [75]). The elements are contacted to
pads in the other direction than the one seen in this cross section.
membrane was chosen to carry the pyroelectric elements.
The membrane consisted of a LPCVD Si3N4film grown
on a thermal oxide (SiO2). The thicknesses were chosen
to compensate for the mechanical stresses of these films
[72]. As these films were also coated the back side of the
double sided polished wafer, they also served as a mask
for back-side etching in KOH. The bottom electrode and
pyroelectric film (PZT15/85) are deposited by sputtering and
sol-gel, respectively. The top electrode is deposited and
patterned by a lift-off technique before a quartz layer is sputter
deposited for reduction of the parasitic capacity below the
contact pads [8]. Windows to access the top electrodes are
opened by a CF4reactive ion etching. The PZT elements
on the membrane part are etched free in a HCl:F solution,
leaving only narrow bridges between the elements and the
bulk silicon part, as required for separation of bottom and
top conductor. The platinum bottom electrode is removed
between the elements by electrochemical etching. This
etching technique does not attack the membrane material.
After deposition and patterning of the conductor lines, pads
(Au/Cr), and absorbing layer (see figure 8), the silicon is
removed below the elements by back-side etching, as defined
by a window in the back-side nitride layer, in order to obtain
the result shown in figures 9 and 10. Inherent to thermal
detectors is the need to absorb infrared radiation. A black
platinum film was utilized as absorber. Black Pt exhibits
a dendritic morphology and grows at some given current
densities and concentrations in an electro- chemical bath [73].
The 0.9µm thick membrane with a specific conductivity
of 2 W m−1K−1gives fairly good thermal insulation.
The PZT15/85 thin films typically yielded a pyroelectric
coefficient of 170 µCm
−2K−1and a relative dielectric
constant of 220. A first series of arrays with 12 elements of
140

Ferroelectric thin films for micro-sensors and actuators
Figure 10. Top view on a 50-element array with 200 µm period
obtained with bulk micromachining; the membrane size is
2×11 mm. The black platinum absorbers, the Cr–Au contact
lines, the membrane layers between the elements, and the SiO2
layer for reduction of parasitic capacitance are clearly visible
(from [12]).
0.36 mm2gave good voltage responses at 1 Hz of 800 V W−1
in air and 3000 V W−1in vacuum. Rather long thermal
time constants of 28 ms in air and 104 ms in vacuum were
obtained [74]. The much larger heat conduction for the
operation in air was due to the heat transfer in air between
the membrane and the device socket (0.4 mm distance). At
higher frequencies, current measurement is preferred. At
30 Hz, the current response amounted to 15–20 µAW
−1
with only small changes as a function of air pressure [74].
The latter is due the fact that above the inverse thermal time
constant, the current response is determined by the inverse
heat capacity, and no longer by the thermal conductivity.
The smaller elements (0.125 mm2) of the larger arrays
showed a smaller voltage response at low frequencies
(460 V W−1). The current response (16 µAW
−1) was about
the same. The current detection at a 10 Hz chopper frequency
was chosen as the operation mode for the gas spectrometer.
The IR source was a simple hot filament. The measurement
of absorption spectra for CO2and CO was demonstrated, see
figure 11. The low-noise equivalent power of 1 nW Hz−1/2
allows the detection of a few ppm of CO2, provided that the
electronics do not increase the noise level [70].
5. Piezoelectric coefficients
In most of the structures applied in MEMS, the piezoelectric
film is part of a composite structure, i.e. the piezoelectric
film is clamped to another elastic body. A rigorous treatment
of this problem requires the solution of the equations of
state with two piezoelectric and several elastic coefficients.
The latter are, however, usually not known precisely. A
more pragmatic way is to consider effective piezoelectric
coefficients of films clamped to a rigid substrate. d33,f
describes the thickness change as a function of the applied
field, i.e. the longitudinal effect; e31,f is the in-plane stress as
a function of the applied field, i.e. the transverse effect. The
film is clamped in the film plane (coordinates 1,2). In the off-
plane direction (coordinate 3), the film is free to move (see
0.7
0.8
0.9
1
1.1
3.8 4 4.2 4.4 4.6 4.8 5
CO
2
Ab
sorpt
i
on
350 ppm
233 ppm
156 ppm
104 ppm
69 ppm
46 ppm
31 ppm
Wavelength (µm)
Figure 11. The CO2absorption spectrum measured by means of a
thin-film pyroelectric array (from [76]). (From bottom curve, 350
ppm to top curve, 31 ppm.)
(This figure is in colour only in the electronic version, see www.iop.org)
piezoelectric
film
U
strain x1
char
g
e
D(x1, x2, σ3,E )
stress σ3
+++++++++++++
---------------------
tp
hpassive
material
σ1(U)
neutral plane
electrodes
d
x3
Actuators Sensors
Figure 12. Schematic description of the geometry and the
working principle of the piezoelectric film applied in actuators and
sensors.
figure 12). This corresponds to a mixed boundary condition.
The directly measured piezoelectric coefficients of thin films
on substrates are therefore functions of standard piezoelectric
coefficients and elastic constants. These effective coefficients
are related to the ordinary coefficients by the following
relations [68, 77]:
e31,f =d31
sE
11 +sE
12
=e31 −cE
13
cE
33
e33 |e31,f |>|e31|
d33,f =e33
cE
33
=d33 −2sE
13
sE
11 +sE
12
d31 <d
33
e31,f is determined either by substrate bending (variation of
x1and x2at σ3=0 and E3=0) and collecting the developed
charges that are related to the in-plane strains as
D3=e31,f (x1+x2)
or by applying a field and measuring the deflection of the
substrate which is governed by the in-plane stresses
σ1,2=e31,f E3.
Note that e31,f is always larger than the bulk coefficient e31.
This originates from the fact that larger piezoelectric stresses
can be developed in the transverse directions if the sample is
free to move in the longitudinal direction.
Most of the potential applications are based on the
transverse coefficient e31,f . Bending of beams and
141

P Muralt
Tab l e 1. Various figures of merit for the different materials. The
PZT thin-film data are evaluated for 1 µm thick sol-gel films
[81–83]. The AlN data are from [84] and the bulk ceramics data
are for typical PZT ceramics [85].
Ceramic
PZT
derived PZT PZT
from bulk (111) (100)
Figure of 45/55 53/47
merits ZnO AlN A B film film
Force, current −0.7 −1.0 −14.7 −14.7 −8.5 −12
response: e31,f
(C m−2)
Voltage −7.2 −10.3 −1.4 −1.0 −1.2 −1.4
response:
(e31,f /ε0ε33)
(GV m−1)
Coupling 0.06 0.11 0.22 0.27 0.11 0.19
coefficient
(kp,f )2on Si
deflections of membranes are much more suited principles
for obtaining large responses or large excursions. For this
reason, this coefficient is discussed in more detail below. In
terms of piezoelectric coefficients, PZT is clearly the leader
among the above materials. This translates into a superior
performance in force, torque, and output power of actuators
and motors, and also of sensors with current detection. This
fact is revealed by the difference in speed per voltage of
an ultrasonic micromotor, i.e. AlN stator and PZT stator
(figure 6). The motor speed is proportional to the vibration
amplitude, which is proportional to the piezoelectric bending
moment, i.e. proportional to e31,f U. However, when voltages
are detected, when the dielectric noise current limits the
signal-to-noise ratio, and when the coupling coefficient is
important (power consumption, power yield and transducer
response), the dielectric constant and the dielectric losses
also have to be considered. In these cases, PZT is no longer
so brilliant because of its high dielectric constant. AlN and
ZnO are more suited for voltage detection (see table 1). The
coupling coefficient in thin-film composite structures needs
to be considered in a different way than in homogeneous bulk
materials. The stiffness of the structure usually depends more
on the passive part, i.e. silicon, thermal oxide, silicon nitride,
etc, than on the PZT itself. On silicon structures, the optimal
coupling coefficients are obtained for a thickness of the
passive layers that is somewhat larger than the PZT thickness
[78, 79]. This means that one should rather consider the
compliance of the substrate than the one of PZT. In analogy
with the planar coupling coefficient kp, the following material
figure of merit for the coupling factor is therefore considered:
k2
p,f =2e2
31,f
ε0ε33,f 1−ν
YSi
.
The data given in table 1 show that the texture of the PZT
thin films is quite important for the piezoelectric properties.
PZT(100) films yield much superior properties as compared
to the (111)-textured films and approach those of optimized,
i.e. doped, PZT ceramics. In fact, PZT(100) films yield better
results than the undoped PZT as published by Berlincourt et al
0
1 104
2 104
3 104
4 104
5 104
0
0.02
0.04
0.06
0.08
0.1
0 5 10 15 20 25
Resonance frequency (Hz)
Coupling constant k
Silicon thickness (µm)
k
f10
no stress
40 MPa
Figure 13. The calculated coupling factor and resonance
frequency, as a function of silicon thickness, for a round disk of
silicon covered by a layer stack, including 1 µm of PZT
(e31,f =6Cm
−2), as discussed in the text (the calculations base
on the analytical model given in [8]). The calculations are shown
for a stress-free and a tensile stressed layer stack.
in 1960 [80], which yield a e31,f of −9.6Cm
−2. The same
table also shows the values for the frequently used ZnO and
the semiconductor compatible AlN. Replacement of these
materials by the optimized PZT thin film allows a gain of
factor 12 in force, and factor two in coupling coefficient k2
p,f .
In thin-film structures, the coupling coefficient not only
depends on the material parameters, but film stresses also play
a role. Film stresses are hardly avoidable. In spite of efforts
to reduce or to compensate for such stresses, there will be
a residual value between 10–100 MPa. Such stresses give a
pre-strain, or a pre-curvature to micromechanical structures.
Poling of PZT thin films may lead to a change of the residual
stress in PZT thin films. In some cases, this stress has to
be taken into account in the design phase of the device. In
very thin membranes, tensile stresses increase the resonance
frequency and reduce the coupling coefficient, as illustrated
in figure 13 for a PZT/Si3N4/SiO2/Si structure. In this case,
the stress of the 200 nm thick nitride was compensated for
by the stress of the 650 nm thick SiO2(originally used
for pyroelectric detectors [72]). In thin-film diaphragms
subjected to tensile stress, a transition from disk behavior
(resonance frequencies depend on the rigidity of the plate)
to membrane behavior (resonance frequencies depend on
the stretching forces) is observed when thinning down the
diaphragm [68].
6. Operation of piezo-electric thin films, poling,
and reliability issues
PZT bulk ceramics and PZT thin films differ in two major
properties: thin films exhibit much higher coercive fields
(typically 50–100 kV cm−1) and higher breakdown voltages
(200–400 kV cm−1). It is therefore possible to drive thin-film
actuators with higher fields in order to compensate partially
for the smaller thickness. Depolarization takes place when
the operation field is too large compared to the coercive
field. A dc field superimposed on the ac field helps in this
case to maintain a good polarization. This is well seen in
142

Ferroelectric thin films for micro-sensors and actuators
figure 6, where the motor speed is very much increased
by a bias of only 2 V. Operation with unipolar fields (as,
e.g., E(1 + sin ωt )) yield stable operating performance and
also proved to be applicable during longer tests (100 h, see
[65, 86]). For some applications, such a dc bias might be
an undesirable technical complication. In such cases it is
favorable to select a Ti-rich PZT composition with a larger
coercive field.
When choosing Ti-rich compositions, poling becomes
an issue for piezoelectric as well as pyroelectric applications
[87, 88]. The very Ti-rich films require hot poling. Films
nearer to the morphotropic phase boundary may be poled also
with UV-light assistance [89]. Poling is not yet understood
in its whole complexity. It is related to a phenomenon that
is presently intensively studied for memory applications:
imprint. Charge injection, defect dipole alignment, and
defect migration are involved in building-up internal fields.
A further important point of performance is stability
during operation and with time. Depolarization (fatigue)
may occur and, if integration is not mastered, delamination
of the PZT film or the electrodes may occur. From an
industrial point of view, the evaluation of ageing and fatigue
is certainly an important task. However, only a few studies
have been reported so far. The motor described above was
subjected to a test lasting 100 h with a unipolar ac field of
20 kHz. Apart of a slight increase of the revolution speed,
no deterioration was observed [86]. The same test was
performed with a stator alone while measuring the vibration
amplitude. A 5–10% decrease of the amplitude was observed
[86]. Most likely this was due to depolarization. Some of the
deposition methods yield films exhibiting an internal field that
gives preference for one direction of polarization. When the
film is poled in this preferential direction, the piezoelectric
properties are more stable with time than when poled on the
opposite direction [90]. With unipolar operation, or operation
below the switching threshold, three different processes can
be identified in fatiguing. The first is depolarization by
180◦domain back switching; which should be completely
reversible and avoidable with a superimposed dc field. The
second mechanism is based on elastic domains such as 90◦
domains. The walls of such domains may migrate in order
to reduce the mechanical stresses built up during poling.
This process might also affect polarization, but should be
mainly reversible. The third category includes irreversible
phenomena such as delamination and cracking. On search
of delamination—which was not found—the second type
of processes was recently evidenced by high-resolution
x-ray diffraction of the silicon interface region of a Si(100)
cantilever coated with PZT/Pt/TiO2/SiO2. After poling the
PZT thin film, a broadening of the Si(400) reflection was
found. This broadening disappeared during a fatigue test
with a unipolar ac field of 100 kV cm−1at 1.2 kHz and
1 day duration. The piezoelectric coefficient, however, did
not decrease. This is explained by stress relaxation due to
mechanical or electrical ac excitation [91].
7. Summary and conclusions
The growth of good quality PZT thin films still needs some
effort. Yet, there is much more experience available than five
years ago. Reproducible film quality is certainly possible
if an industrial approach is adopted. The electrodes below
the PZT thin film play a very important role for seeding the
correct phase and the film texture. Without reproducible
electrode quality, no reproducible PZT quality is achieved.
Good piezoelectric properties have only been obtained
for deposition temperatures higher than 550–600 ◦C. In
principle, these temperatures can be well handled. However,
they are problematic if direct integration of PZT thin films
onto integrated circuits is the goal. The crucial problem lies
in the aluminum metallization that hardly survives the 500 ◦C
post-processing temperatures. The aluminum metallization
step thus has to be performed after PZT deposition. However,
this increases the number of process steps, and thus costs,
after standard CMOS processing. In the case of pyroelectric
applications, the quality of low-temperature grown films is
more acceptable, as there is no domain wall contribution for
pyroelectricity. Recently it was reported that a pyroelectric
array was successfully integrated by back-end processing at
500 ◦C onto a completed CMOS read-out integrated circuit
[13]. It is too early to give a final statement on the
piezoelectric properties of PZT thin films. According to
table 1, one might say that we have arrived at 80% of
the theoretical values of e31,f . Better values may still be
achieved. One problem of judging piezoelectric performance
is the fact that we do not know exactly how to derive
thin-film properties from known bulk ceramic properties.
Microstructural differences, defects due to lower growth
temperature, and interface effects at the electrode certainly
play a role. Lattice and domain wall contributions need to
be considered differently. Single crystal values for PZT are
not available. The theoretical understanding is restricted to
simpler systems, such as epitaxial {100}-oriented PbTiO3.If
there would be lattice contributions only, in bulk as well as
in thin films, the thin-film values could simply be calculated
as in the case of the non-ferroelectric AlN, where the values
predicted from single crystal data correspond well with the
experimental values for correctly textured thin films. The
domain wall contributions in ferroelectrics, however, are
very much influenced by film texture, by grain size [92, 93],
and by defects. Clamping of specially oriented grains may
change domain configuration, and may even impose other
symmetries [94]. For a final statement, there is also not
enough experience with doping of PZT thin films. In
bulk ceramics, piezoelectric properties can be very much
improved through doping. The question is open whether this
also works with thin films. Another field where experience is
still missing is that of thin films of relaxor-type ferroelectrics
such as Pb(Mg1/3Nb2/3)O3. These substances show strong
electrostrictive coefficients due to large dielectric constants.
With the application of a dc field of 150 kV cm−1,an
equivalent d33,f of 60 pm V−1was obtained from the small-
signal ac response [95]. Although this is inferior to the
values achieved with PZT thin films (100–120 pm V−1), as
afirst result it is promising. Relaxor materials have the
advantage that they do not exhibit hysteresis phenomena.
This allows a better linearity of positioning systems in micro-
and nanotechnology.
The overall estimation of performance is best seen in
device applications. In actuators, piezoelectricity competes
143

P Muralt
with other principles such as the application of electrostatic
forces. The comparison of the above micromotor with an
electrostatic motor reveals the advantages and disadvantages
of PZT thin-film driven ultrasonic actuators. A millimeter-
sized electrostatic wobble micromotor has recently been
described in the literature [96]. Similar maximal torques
were obtained with both motors (0.1–1µN m). However,
the ultrasonic motor needed about 30 times less voltage
for the same torque. The electrostatic motor does not
support any normal force—the torque is reduced to zero at
0.6mN—whereas the ultrasonic motor increases output force
with increasing normal load (up to, probably, 10 mN). The
ultrasonic motor is thus more easily connected to a gear. The
speed-per-voltage ratio is about the same for both motors.
The ultrasonic motor can be run at lower speeds (60 rpm),
the electrostatic motor runs faster (4000 rpm). The main
advantage of electrostatic motors seems to be the excellent
efficiency of the motor. In open-loop operation, 80% of
the consumed power is transformed to mechanical energy
of the rotor. With ultrasonic motors, this is much more
difficult to achieve. While electrostatic micromotors need
high voltages and low currents, PZT thin-film ultrasonic
micromotors need low voltages and high currents. The
dielectric loss of the PZT thin films (3–5%) was found to be a
major reason for power dissipation. The friction mechanism
applied for the transfer between the stator and the rotor is
also too dissipative, since 80% of the stator energy is lost.
The latter was too small in relation to the reactive power in
the capacitor. The clamping of the membrane at its borders
greatly reduces the coupling factor. Calculations show that a
free membrane, fixed in the center, would have a five times
higher coupling coefficient k2. The first improvement of
this motor would consist in integrating the new (100) films,
allowing for a quadruplication of speed and power at a given
voltage. Whereas this would lead to an excellent torque-per-
voltage performance, the power consumption problem would
still remain. In order to reduce power dissipation, it is also
necessary to improve the vibrator structure, and the transfer
to the rotor. The more suited applications for PZT thin
films are probably those without frictional interaction, such
as linear actuators based on cantilevers as used in scanning
probe techniques. PZT-laminated cantilevers can be excited
to resonance and allow sensing of topography and forces [97].
As to pyroelectric applications, important demonstrators
have been achieved with PZT thin films. PZT is currently
being replaced by better pyroelectric materials of the same
family, such as Ca- and La-substituted PbTiO3. The matter
is not only to increase the pyroelectric coefficient, but also
to decrease the dielectric constant [98]. Both of these will
enable the increase of detectivity by a factor three to six with
respect to today’s PZT versions.
In conclusion, PZT thin-film solutions in micromechan-
ical sensors and actuators are very competitive when cur-
rent signal and force or power outputs are demanded. When
voltage signal, intrinsic signal-to-noise ratio, or power effi-
ciency is an issue, the non-ferroelectric AlN is a competitive
material. The reduction of the dielectric losses of PZT is
desirable. The example of the ultrasonic motor shows that
improvements in the design are also important in order to
adapt to the reduced geometrical possibilities with microma-
chined structures. Although pyroelectric IR detectors with
PZT meet many of the present specifications, PZT will be
replaced by more sensitive materials.
Acknowledgments
The author wishes to thank his colleagues for many useful
discussions. The work was supported by the Swiss Priority
Programs on Materials Research (PPM) and Micro- and
Nano Systems Technology (MINAST), and the Program
‘Microsystems and Microtechnique’of EPFL.
References
[1] Jaffe B, Cook W R and Jaffe H 1971 Piezoelectric Ceramics
(London: Academic)
[2] Haertling G H 1986 Ceramic Materials for Electronics:
Processing, Properties, and Applications
ed R C Buchanan (New York: Dekker)
[3] Flynn A M et al 1992 Piezoelectric micromotors for
microrobots J. Microelectromech. Syst. 144–51
[4] Watanabe S, Suzuki M and Fujiu T 1995 J. Vac. Sci. Technol.
B13 1119
[5] Lugienbuhl P et al 1996 Piezoelectric cantilever beams
actuated by PZT sol-gel film Sensors Actuators A53
530–5
[6] Miyahara Y et al 1999 PZT cantilever for non-contact
atomic force microscopy Appl. Surf. Sci. 140 428–31
[7] Muralt P et al 1995 Fabrication and characterization of PZT
thin-film vibrators for micromotors Sensors Actuators A
48 157–65
[8] Dubois M-A and Muralt P 1998 PZT thin film actuated
elastic fin micromotor IEEE Trans. Ultrasonics,
Ferroelectrics, Frequency Control 45 1169–77
[9] Lugienbuhl P et al 1997 Microfabricated Lamb wave device
based on PZT sol-gel thin film for mechanical transport of
solid particles and liquids J. Micromech. Syst. 6337–46
[10] Bernstein J J et al 1997 Micromachined high frequency
ferroelectric sonar transducers IEEE Trans. UFFC 44
960–9
[11] Riza N A et al 1993 Electron. Lett. 29 1606–8
[12] Willing B et al 1998 Gas spectrometry based on pyroelectric
thin film arrays integrated on silicon Sensors Actuators A
66 109–13
[13] Beratan H R et al 1998 International Symposium on
Integrated Ferroelectrics (ISIF) (Monterey, CA)
[14] Bruchhaus R et al 1997 5th Infrared Sensors and Systems
vol 4 (Dresden: Dresden University Press) pp 79–84
[15] Castellano R N and Feinstein L G 1979 Ion-beam deposition
of thin films of ferroelectric PZT J. Appl. Phys. 50
4406–11
[16] Krupanidhi S B et al 1983 RF planar magnetron sputtering
and characterization of ferroelectric PZT films J. Appl.
Phys. 54 6601–9
[17] Okuyama M et al 1979 Japan. J. Appl. Phys. 28 15
[18] Sreenivas K, Sayer M and Garrett P 1989 Properties of dc
magnetron-sputtered PZT thin films Thin Solid Films 172
251–67
[19] Kwak B S, Boyd E P and Erbil A 1988 MOCVD of PbTiO3
thin films Appl. Phys. Lett. 53 1702–4
[20] Kojima M et al 1983 Chemical vapor deposition of PbTiO3
films Japan. J. Appl. Phys. 22 14–17
[21] Swartz S L et al 1989 Characterization of MOCVD PbTiO3
thin films Ferroelectrics 93 37–43
[22] Brierley C J et al 1989 The growth of ferroelectric oxides by
MOCVD Ferroelectrics 91 181–92
[23] Okada M et al 1989 Preparation of c-axis oriented PbTiO3
thin films by MOCVD under reduced pressure Japan. J.
Appl. Phys. 28 1030–4
144

Ferroelectric thin films for micro-sensors and actuators
[24] Budd K D, Dey S K and Payne D A 1985 Sol-gel processing
of PT, PZ, PZT and PLZT thin films Br. Ceram. Proc. 36
107
[25] Dey S K, Budd K D and Payne D A 1988 Thin film
ferroelectrics of PZT by sol-gel processing IEEE Trans.
UFFC 35 80–1
[26] SpieringsGAMCet al 1991 Preparation and ferroelectric
properties of PZT thin films by spin coating and
metalorganic decomposition J. Appl. Phys. 70 2290–8
[27] Tuttle B A et al 1992 Microstructural evolution of PZT thin
films prepared by hybrid metallo-organic decomposition
J. Mater. Res. 71876–82
[28] Otsubo S et al 1990 Preparation of PZT films by laser
ablation Japan. J. Appl. Phys. 29 L133–6
[29] Buhay H et al 1991 Pulsed laser deposition and ferroelectric
characterization of bismuth titanate films Appl. Phys. Lett.
58 1470–2
[30] Roy D, Krupanidhi S B and Dougherty J P 1991 Excimer
laser ablated lead zirconate titanate thin films J. Appl.
Phys. 69 7930–2
[31] Klissurska R D et al 1995 Effect of Nb Doping on the
microstructure of sol-gel-derived PZT thin films J. Am.
Ceram. Soc. 78 1513–20
[32] H¨
artl K H and Rau H 1969 PbO vapour pressure in the
Pb(Zr,Ti)O3system Solid State Commun. 741–5
[33] Kwok C K and Desu S B 1994 Formation kinetics of PZT
thin films J. Mater. Res. 91728–33
[34] Okada M et al 1990 MOCVD of c-axis oriented PZT thin
films Japan. J. Appl. Phys. 29 718–22
[35] Okada M and Tominaga K 1992 Preparation and properties
of PLZT thin films by MOCVD J. Appl. Phys. 71 1955–9
[36] de Keijser M and DormansGJM1995 Modelling of
organometallic chemical vapor deposition of lead titanate
J. Cryst. Growth 123 537–44
[37] DormansGJM,vanVeldhoven P J and Keijser M D 1992
Composition-controlled growth of PbTiO3on SrTiO3by
MOCVD J. Cryst. Growth 123 537–44
[38] Maeder T et al 1995 Pb(Zr,Ti)O3thin films by in situ
reactive sputtering on micromachined membranes for
micromechanical applications Br. Ceram. Proc. 54
206–18
[39] Maeder T and Muralt P 1994 In situ thin film growth of
PbTiO3by multi target sputtering Mater. Res. Soc. Symp.
Proc. 341 361–6
[40] KumarCVRVet al 1991 PZT films by rapid thermal
processing Appl. Phys. Lett. 58 1161–3
[41] Li X et al 1993 Low-temperature in situ preparation of
ferroelectric PZT thin films by reactive sputtering Appl.
Phys. Lett. 63 2345–7
[42] Bruchhaus R et al 1992 Deposition of ferroelectric PZT thin
films by planar multi-target sputtering Ferroelectrics 127
137–42
[43] Hayashi S, Iijima K and Hirao T 1991 Preparation and
characterization of Pb-based ferroelectric thin films
Japan. J. Appl. Phys. 30 2186–8
[44] Tabata H et al 1991 Preparation of PbTiO3thin films at low
temperature by an excimer laser ablation technique Appl.
Phys. Lett 59 2354–6
[45] Hu H and Krupanidhi S B 1992 Property modification of
ferroelectric PZT thin films by low-energy oxygen ion
bombardment during film growth Appl. Phys. Lett. 61
1246–8
[46] Roy D, Krupanidhi S B and Dougherty J P 1992 Excimer
laser ablation of ferroelectric PZT thin films with low
pressure direct-current glow discharge J. Vac. Sci.
Technol. A10 1827–31
[47] Mihara T et al 1992 Relationship between crystal structure
and chemical composition of PbTiO3thin films prepared
by sputter-assisted plasma CVD Japan. J. Appl. Phys. 31
1872–3
[48] Brown H E (ed) 1985 Lead Oxide: Properties and
Applications (New York: International Lead Zinc
Research Organization)
[49] Chen K C and Mackenzie J D 1990 Mater. Res. Soc. Symp.
Proc. 180 663–8
[50] Aoki K et al 1995 Effects of titanium buffer layer on PZT
crystallization process in sol-gel deposition technique
Japan. J. Appl. Phys. 34 192–5
[51] Muralt P et al 1998 Texture control of PbTiO3and PZT thin
films with TiO2seeding J. Appl. Phys. 83 3835–41
[52] Muralt P et al 1996 In situ sputter deposition of PbTiO3thin
films: nucleation on textured platinum films Le Vide 279
45–7
[53] Kwok C K and Desu S B 1993 Low temperature perovskite
formation of PZT thin films by a seeding process J. Mater.
Res. 8339–44
[54] Revesz P et al 1991 Mater. Res. Symp. Proc. 243 101–6
[55] Al-Shareef H N et al 1993 Electrodes for ferroelectric thin
films Integrated Ferroelectrics 3321–32
[56] Vijay D P and Desu S B 1993 Electrodes for PZT
ferroelectric thin films J. Electrochem. Soc. 140
2640–4
[57] Ramesh R et al 1993 Ferroelectric LSCO/PZT/LSCO
heterostructures on silicon via template growth Appl.
Phys. Lett. 63 3592–4
[58] Willems G J et al 1997 Nucleation and orientation of sol-gel
PZT thin films on Pt electrodes Integrated Ferroelectrics
15 19–28
[59] Maeder T, Sagalowicz L and Muralt P 1998 Stabilized
platinum electrodes for PZT thin film deposition using Ti,
Zr and Ta adhesion layers Japan. J. Appl. Phys. 37
2007–12
[60] Sreenivas K et al 1994 Investigation of Pt/Ti bilayer
metallization on silicon for ferroelectric thin film
integration J. Appl. Phys. 76 466–76
[61] Hara T et al 1997 Japan. J. Appl. Phys. 36 L893
[62] Grill A, Jahnes C and Cabral C 1999 Layered TaSiN as an
oxidation resistant electrically conductive barrier J. Mater.
Res. 14 1604–9
[63] Maeder T et al 1999 Conducting barriers for direct contact of
PZT thin films on reactive substrates J. Electrochem. Soc.
146 3393–7
[64] Maeder T et al 1996 PZT thin films on Zr membranes for
micromechanical applications Appl. Phys. Lett. 68 776–8
[65] Racine G-A, Muralt P and Dubois M-A 1998
Flexural-standing-waves elastic force motor using ZnO
and PZT thin films on micromachined silicon membranes
for wristwatch applications Smart Mater. Struct. 7
404–16
[66] Racine G-A et al 1994 IEEE Micro Electro Mechanical Syst.
124–9
[67] Kurosawa M et al 1992 IEEE Ultrasonics Symp. (Tucson,
AZ) pp 893–6
[68] Muralt P et al 1996 Piezoelectric actuation of PZT thin film
diaphragms at static and resonant conditions Sensors
Actuators A53 397–403
[69] Dubois M-A 1999 PhD Thesis Materials Department, EPFL,
Lausanne, Switzerland
[70] Willing B et al 1999 Transducers ’99 (Sendai, Japan)
pp 402–5
[71] Willing B et al 1998 Thin film pyroelectric array as a
detector for and infrared gas spectrometer Infrared Phys.
Technol. 39 443–9
[72] Bell A et al 1995 A thin film pyroelectric detector Integrated
Ferroelectrics 6231–40
[73] Lang W, K¨
uhl K and Sandmeier H 1992 Absorbing layers for
thermal infrared detectors Sensors Actuators A34
243–8
[74] Kohli M et al 1997 Pyroelectric thin film sensor array
Sensors Actuators A60 147–53
[75] Muralt P et al 1995 ECASIA (Montreux, Switzerland)
(New York: Wiley) pp 115–22
[76] Willing B et al 1997 Transducers—Int. Conf. on Solid-State
Sensors and Actuators (Chicago, IL) pp 1245–8
145

P Muralt
[77] Lefki K and DormansGJM1994 Measurement of
piezoelectric coefficients of ferroelectric thin films
J. Appl. Phys. 76 1764–7
[78] Pavlicek H and Wachutcha G 1994 CAD tools for MEMS
UETP MEMs Course Swiss Foundation for Research in
Microtechnology, Neuchˆ
atel, Switzerland
[79] Muralt P 1997 Piezoelectric thin films for MEMS Integrated
Ferroelectrics 17 297–307
[80] Berlincourt D A, Cmolik C and Jaffe H 1960 Proc. IRE 48
222–9
[81] Dubois M-A et al 1998 Which PZT thin films for
piezoelectric microactuator applications? Integrated
Ferroelectrics 22 535–43
[82] Ledermann N, Seifert A and Muralt P 1999 Effective
transverse piezoelectric coefficient e31 of (100)/(001)
textured PZT thin films Integrated Ferroelectrics 224
235–24
[83] Seifert A et al 2000 Study of the transverse piezoelectric
coefficient of PZT thin films as a function of texture and
composition Mater. Res. Soc. Proc. 596 at press
[84] Dubois M-A and Muralt P 1999 Properties of AlN thin films
for piezoelectric transducers and microwave filter
applications Appl. Phys. Lett. 74 3032–4
[85] Moulson A J and Herbert J M 1990 Electroceramics
(London: Chapman and Hall)
[86] Muralt P et al 1995 Fabrication and characterization of PZT
thin film micromotor Integrated Ferroelectrics 11 213–20
[87] Seifert A et al 1999 Microstructural evolution of dense and
porous pyroelectric PbCaTiO3thin films J. Mater. Res. 14
2012–22
[88] Kohli M, Seifert A and Muralt P 1998 Poling of pyroelectric
thin films J. Integrated Ferroelectrics 22 453–63
[89] Kholkin A L and Setter N 1997 Photo induced poling of PZT
thin films Appl. Phys. Lett. 71 2854–6
[90] Kholkin A et al 1995 Interferometric study of piezoelectric
degradation in ferroelectric thin films Microelectron. Eng.
29 261–4
[91] Ledermann N, Muralt P and Dommann A 1999 Accurate
microscopic method to investigate the aging of
PZT/silicon heteromorphs cantilevers Proc. 2nd Int. Conf.
on Integrated Micro/Nanotechnology for Space
Applications MNT’99 (Pasadena, April 1999) (The
Aerospace Corporation) pp 250–257
[92] Arlt G, Hennings D and de Witt G 1985 Dielectric properties
of fine-grained barium titanate ceramics J. Appl. Phys. 58
1619–25
[93] Arlt G 1989 Microstructure and domain effects in
ferroelectric ceramics Ferroelectrics 91 3–7
[94] Pertsev N A, Zembilgotov A G and Tagantsev A K 1998
Effect of mechanical boundary conditions on phase
diagrams of epitaxial ferroelectric thin films Phys. Rev.
Lett. 80 1988–91
[95] Kighelman Z et al 1998 Relaxor behavior and
electromechanical properties of Pb(Mg,Nb)O3thin films
Appl. Phys. Lett. 73 2281–3
[96] Samper W D et al 1999 Torque evaluation of a LIGA
fabricated electrostatique micromotor IEEE J. MEMS 8
115–23
[97] Lee C, Itoh T and Suga T 1999 Self-excited piezoelectric
PZT microcantilevers for dynamic SFM—with inherent
sensing and actuating capabilities Sensors Actuators A72
179–88
[98] Seifert A et al 1999 Microstructural evolution of dense and
porous (PbCa)TiO3thin films for pyroelectric applications
J. Mater. Res. 14 2012–22
146