Development of PZT and PZN-PT Based Unimorph Actuators for

Micromechanical Flapping Mechanisms

Metin Sitti, Domenico Campolo , Joseph Yan, Ronald S. Fearing
Dept. of EECS, University of California, Berkeley, CA 94720

Scuola Superiore SantAnna, MiTech Laboratory, Pisa, Italy

Tao Su, David Taylor, Timothy D. Sands
Dept. of Materials Science and Engineering, University of California, Berkeley, CA 94720
{sitti,ronf}@eecs.berkeley.edu
Abstract
This paper focuses on the design, fabrication and characterization of unimorph actuators for a microaerial flapping mechanism. PZT-5H and PZN-PT are investigated
as piezoelectric layers in the unimorph actuators. Design issues for microaerial flapping actuators are discussed, and criteria for the optimal dimensions of actuators are determined. For low power consumption actuation, a square wave based electronic driving circuit is
proposed. Fabricated piezoelectric unimorphs are characterized by an optical measurement system in quasi-static
and dynamic mode. Experimental performance of PZT5H and PZN-PT based unimorphs is compared with desired design specifications. A 1 d.o.f. flapping mechanism with a PZT-5H unimorph is constructed, and 180
stroke motion at 95 Hz is achieved. Thus, it is shown
that unimorphs could be promising flapping mechanism
actuators.

Introduction

Actuators are one of the most significant integral parts

of robotic mechanisms. Flapping mechanisms [1], [2],
[3] require actuators with large periodic stroke (rotational) motion (30 150o) at high speed (10-100s of Hz)
with large output forces for overcoming the aerodynamic
damping. Moreover, light weight (10s of mg), high efficiency, long life time, and compact size are important
issues. Piezoelectric actuators with proper design almost
satisfy all of these requirements. There are different types
of piezoelectric actuators such as stack [4] and bending [5]
types, motors, impact type, etc. The stack type enables
very high output forces although it has large size and
weight, and smaller displacement relative to the bending
type. Thus, flexural bending actuators generally generate large deflection with low weight. Therefore, bimorphs
and unimorphs are more suitable for microaerial flapping
applications [6], [3], [7]. These actuators consist of piezoelectric layers bonded to a purely elastic layer. Since they
are easier to fabricate, the unimorph type is selected in
this study.
Unimorph actuators have been studied by many
groups [8], [9], [10], [11], [12], [13]. Its one dimensional

beam theory is well established [8]. Characteristic parameters based on this theory were formulated [13]. Utilizing unimorphs in flapping mechanisms, Cox et al. [7]
developed four-bar and five-bar flexure mechanisms integrated with unimorph actuators for stroke motions upto
30 50o .
Taking the microaerial flapping mechanisms as the target application in this paper, design issues, selection of
the proper actuator, and its fabrication and characterization are investigated. PZT-5H and a recently developed
PZN-PT single crystal material [14] are investigated as
the piezoelectric layers. PZT-5H is a widely used soft
piezoelectric ceramic. Strain levels as high as 1.7% can
be achieved in PZN-PT depending on composition, orientation and the applied electric field. Miniature actuators
are fabricated and characterized to evaluate their performance for flapping actuation.

Unimorph Actuators

A standard rectangular shape unimorph actuator under

activation is illustrated in Figure 1. The actuator consists
of a single piezoelectric layer bonded to a purely elastic
layer. Steel or titanium is usually chosen for the elastic
layer. When a voltage is applied across the thickness of
the piezoelectric layer, longitudinal and transverse strain
develop. The elastic layer opposes the transverse strain
which leads to a bending deformation.
F

+
p
e

i e
l a

o
t i c

l e
l a

c
y

t r i c
e

l a

p
s

Figure 1: Basic cantilevered rectangular shape unimorph actuator structure.

2.1

Unloaded Unimorph Equations

For a free deflecting unimorph actuator, DC (low frequency) tip displacement dc , blocking force Fb , resonant

frequency fr , mechanical stiffness Km , and mechanical

quality factor Qa can be written as [12], [13]:
dc

3l2 AB(B + 1)
d31 V
h2p
D

Fb

fr

Qa

Km

2.2

3whp AB(B + 1)
d31 V
4sp l AB + 1
s
i
2i hp Ep h
D
2
4l
3p (BC + 1)(AB + 1)
fr
fr1 fr2
wh3p
Fb
D
=
3
dc
4sp l AB + 1

(1)

where A = sp /ss = Es /Ep , B = hs /hp , C = s /p , D =

A2 B 4 + 2A(2B + 3B 2 + 2B 3 ) + 1. Here, l is the unimorph
length, w is the width, V is the applied voltage, fr1 and
fr2 are the frequencies where the deflection magnitude
drops to 0.707 of its resonance peak value, and d31 is
the transverse piezoelectric coefficient. sp and ss are the
elastic compliances, hp and hs are the thicknesses, Ep
and Es are the Young Moduli, and p and s are the
densities of the piezoelectric and steel layers respectively,
and i is the eigenvalue [11] where i denotes the resonance
mode, i.e. first mode 1 = 1.875, and the second mode
2 = 4.694. For the PZT-5H, PZN-PT and steel layers,
Young Modulus E, density , d31 , coupling factor k31 ,
relative dielectric constant K3T = /0 , and maximum
electric field E3 values are taken as in Table 1.
E (GP a)
(kg/m3 )
d31 (C/N )
k31
K3T
E3 (V /m)

PZT-5H
61
7500
320 1012
0.44
3800
1.5 106

PZN-PT
15
8000
950 1012
0.5
5000
10 106

the actuators are driven unipolarly, i.e. V > 0. Thus,

the wing motion is [0, dc ] at DC, and [dc /2
r , dc /2 + r ] at resonance.

Unimorph Fabrication Process

Commercial piezoelectric ceramics and single crystal

plates were used: (1) Doped P b(Zr, T i)O3 (PZT-5H,
T105-H4E-602 ceramic single sheet, Piezo Systems, Inc.)
with 127 m thickness, and (2) P b(Zn1/3 N b2/3 )O3
P bT iO3 (PZN-PT, TRS Ceramics Inc.) with 136 m
thickness.
Rectangular piezoelectric and steel layers are cut using a high-speed cut-off saw (Accutom 50) with desired
dimensions. For the piezo layers, the as-received thicknesses are used, and polishing is not necessary. Then, the
cut layers are bonded using a thin epoxy glue (M-Bond
610, Measurement Groups Inc.) with a slight offset along
the width, and cured for 24 hours at room temperature.
A conductive epoxy (CW2400, Chemtronics) is applied to
the side offset edge for connecting the piezo lower electrode with the steel layer. Next, wires are soldered to
the steel and piezoelectric electrodes using a soldering
flux (MSF-003-NI, Piezo Systems, Inc.), and the actuator is fixed to a holder for tests. Example prototypes
of the PZT-5H and PZN-PT unimorphs are displayed in
Figure 2. A close side view of the PZT-5H actuator is
also displayed in Figure 3 for showing the piezo, steel and
glue layers.

steel
193
7872

Table 1. PZT-5H, PZN-PT and steel layer properties [1],[5].

For the design considerations, above equtions are converted to the rotational motion of the actuator for simplicity. Assuming the actuator tip deflection is small,
actuator rotation angle dc , output torque a , and rotational stiffness Ka are given as follows:

dc
a
Ka

Figure 2: Photos of the prototype 1630.21 mm3 PZT-5H

(upper) and 510.22 mm3 size PZN-PT (lower) unimorphs.

dc
3l AB(B + 1)
= 2
d31 V
l
hp
D

= Fb l =

3whp AB(B + 1)
d31 V
4sp AB + 1

= Km l 2 =

wh3p
D
.
4sp l AB + 1

(2)

Here, the maximum input voltages are Vmax = E3 hp , e.g.

for hp = 100 m, Vmax = 150 V and Vmax = 1000 V for
PZT-5H and PZN-PT unimorphs respectively. Moreover,

Figure 3: Side view image of the tip of the PZT-5H unimorph

for showing the details of the piezo, steel and glue layers using
the optical microscope.

Actuator Design Issues

For a flapping mechanism with a wing load on it, design

parameters such as unimorph dimensions, output torque,
resonant frequency, required transmission ratio, quality
factor, weight, etc. are to be selected for optimal performance.
Ba

:1

a
J w
Ka

Figure 4: Linear dynamic model of piezo actuator, lossless

transmission, and wing.

In our proposed flapping mechanism, a four-bar mechanism is coupled with the unimorph actuator for the
stroke amplification [6], [3]. Assuming there is a wing
as a load with inertia Jw and damping Bw , and a fourbar based transmission mechanism with stiffness Kt and
stroke amplification (transmission ratio) T , a linear approximate dynamic modeling of Figure 4 gives:
K

a
a
+
K
(3)
Jw + Bw +
=
t
T2
T
where is the flapping stroke angle, and Ka is the actuator rotational stiffness. Here, the actuator damping Ba
and inertia Ja are assumed to be negligible with respect
to the load damping and inertia.

3.1

Selection of Actuator Dimensions

For a given load power requirement, the actuator dimensions are to be chosen for ease of mechanical drive, fabrication, and drive voltage requirements. Considering a micromechanical insect modelled after a blowfly, with mass
m = 0.1 g, wing beat of = 2150 rad/s, and wing
stroke amplitude r = 70 at resonance, the net wing lift
force must match the insect weight of 103 N . Although
in the quasi-steady state lift and drag forces are generally proportional to the square of velocity, we choose a
linear damper with a force at peak wing velocity equals
to the weight of the micromechanical flying insect (MFI)
as an upper bound. (Note that the linear damper overestimates the damping force for all wing velocities less than
the peak velocity). Hence the wing damping Bw (at the
wing hinge) can be estimated from:
mglw
Bw =
(4)
r
where m = 0.1 g, g = 9.81 m/s2 , lw is the length of the
wing center of pressure, and is the wing beat frequency.
For lw = 10 mm, Bw = 8.65 109 N sm.
The Q of a resonant system is defined as the ratio
of stored energy to energy dissipated per radian. With
proper actuator and transmission design, energy dissipation for the MFI is work done on moving air, i.e. useful work. A high Q hence implies large internal stored

energy, and poor controllability of wing amplitude and

phase due to actuator saturation. As shown in the data
by Sotavalta [15], blowflies have a relatively low Q, estimated on the order of 1-3. For the MFI, we choose the
quality factor of the wing and thorax as Qw = 2.5, as a
higher Qw system requires a lower transmission ratio and
less actuator motion at DC.
To have a low Qw , i.e. maneuverable wing, the wing
inertia is:
Qw Bw
= 2.26 1011 kg m2 .
(5)
Jw =

The actuator stiffness, as seen at the wing hinge, must

resonate at , hence:
K1 = Ka /T 2 + Kt = Jw 2 = 2.0 105 N m
(6)
The four-bar transmission converts the small rotation of
the actuator to the wing rotation by a transmission
ratio T . At DC, the displacement of the wing is just
T a
2r
dc =
=
= T dc
(7)
Qw
T K1
Then, for a given T and the desired wing flapping amplitude r at resonance frequency :
a
dc

=T
=

2K1 r
dc K1 T
=
T Qw
T
2r
dc
=
T
T Qw

(8)

For a given hs , hp and V , l and w can be computed

as
l

h2p D
dc
3d31 AB(B + 1)V
4sp
AB + 1
a
3d31 hp V AB(B + 1)

(9)

The average power at the wing is also another important parameter for the design which can be computed
from
a2 Bw
(mglw )2 Bw
Pw =
=
(10)
8T 2 (Bw + Ba /T 2 )2
2(Bw + Ba /T 2)2
where Ba = Ka /(Qa ), Qa = 20 and Ka = T 2 (K1 Kt ).
Furthermore, the mass of the actuator ma is limited for
enabling a total flying robot mass of m = 0.1 g. Therefore, ma = (p hp + s hs )lw should be also checked.
On the other hand, the thickness ratio of PZT and
elastic layers is important to select for maximum displacement and force. For a given hp = 127 m, the effect
of changing hs on the normalized values of dc , Fb and
fr are shown in Figure 5. In the figure, depending on
the design criteria, hs could be selected to maximize dc
(hs = 35 m case), or maximize Fb (hs is as large as
possible case). In our case, since both dc and Fb are
to be maximized, the following relation is used [13] for
choosing hs :
r
ss
hp
(11)
hs =
sp
In the figure, this corresponds to hs = 76 m (dashed
line). Morever, following constraints exist for hs selection:

switching stage to the wing load which is simply modeled as a linear system with inertia (Lwing ) and damping (Rdamp ) (although this is a crude approximation it
will not affect what will follow). In the figure, Cstif f
represents the piezoelectric mechanical stiffness. For a
PZT unimorph with parameters as in Table 1, N = T ,
Rloss = 12 M , Cl = 12 nF , Tpzt = 6.75 104 ,
Cstif f = 1/Km = 0.0065 F , Rdamp = 6(Tpzt /T )2 M ,
2
and Lwing = 1/(T 2 2 Cstif
f ) for a given T .

1
0.9

Blocking Force
(x90 mN)

Normalized Values

0.8
0.7

Deflection
(x200 um)

0.6

Resonant
Frequency
(x1182 Hz)

0.5
0.4
0.3
0.2
0.1
0
0

25

50

75

100

125

150

175

200

225

250

Steel Thickness (m)

Figure 5: The effect of elastic layer thickness hs on normalized values of the deflection dc , the blocking force Fb ,
and the resonant frequency fr for a PZT-5H unimorph with
hp = 127 m and steel elastic layer.
The resonant frequency of the actuator fr should be
around ten times larger than the desired wing beat
frequency of 150 Hz in order to have a nonsignificant
actuator damping at resonance, and enabling square
wave driving for minimum power loss.
Applied voltage is increased for a thicker actuator.
During fabrication, polishing ceramic is troublesome. Therefore very thin layers are not desirable.
Using the above specifications, the MFI piezo parameters are computed as given in Table 2. Here, Jw = 2.26
1011 kgm2 , Bw = 8.65 109 N sm, = 2150 rad/s,
r = 70 , Qw = 2.5, and Kt = 5.3 106 N m/rad are
taken. Considering the available piezoelectric plates we
have, V = 150 V and hp = 127 m, and V = 250 V
and hp = 136 m are fixed for the PZT-5H and PZNTPT layers respectively. For these values, Pw = 4.7 mW .
Then, for given T and hs , l, w, Fb , dc , Pw , fr and ma
values are computed. From the values, it can be seen that
16 3 0.21 mm2 size PZT-5H and 5 1.3 0.22 mm2
size PZN-PT would enable the desired 140 wing flapping
at 150 Hz with relatively low masses. For PZT-5H unimorph to fly, V should be increased to 250 V , lowering
actuator mass to 26 mg per wing.
T
type
PZT
PZN-PT
PZT
PZN-PT

44
36
39
28

hs
(m)
76
76
50
50

lw
(mm2 )
16 2.9
5 1.4
16 3.6
5 1.3

Fb
(mN)
54
142
49
109

dc
(m)
354
135
393
176

fr
(Hz)
464
3032
406
2548

ma
(mg)
74
12
78
10

Table 2. Selected MFI unimorph actuator parameters for

different T and hs values and unimorph types.

4
4.1

Actuator Driving Issues

Electromechanical Model

Figure 6 represents the simplified electrical equivalent

of the whole system, ranging from power supply and

Figure 6: Electrical equivalent of the whole system where the

hysteretic capacitor is the only nonlinear element.

Piezoelectric internal inertia and damping turn out

to be negligible compared to the wing parameters. For
the piezoelectric actuator, a model is introduced which
is valid for frequencies up to resonance. More complex
models which take the linear dynamics at high frequencies into account are also available. In [10], tip deflection , volume displacement and electrode charge q of
the piezoelectric actuator are dynamically related to the
moment and force Fb at the tip, uniform body pressure,
and electrode voltage V by means of a 4 4 frequencyvarying symmetric matrix. Since electrode voltage and
current (or charge) and tip force and displacement are
important in our case, only the 2 2 submatrix can be
utilized:
h i h E F ih F i
b
=
(12)
q
F G
V

The Above matrix describes a 2-port electromechanical

system whose electrical variables are voltage and charge
while the mechanical parameters are force and displacement. An equivalent 2-port electrical system can be obtained by letting the force be represented by a voltage
and displacement by a charge (i.e. mechanical speed is
equivalent to current). The electrical equivalent to the
2 2 electromechanical matrix is shown in Figure 7.

Figure 7: Electrical 2-port circuit equivalent to the electromechanical system(N = T ).

The circuit in Figure 7 completely describes the linear behavior of the piezo actuator at all frequencies, and
since it is described by a linear partial differential equation (acoustic waves in piezoelectric material), Z1 , Z2
cannot be expected to be equivalent to simple devices
such as capacitors, resistors and inductors (or networks
with a finite number of them) because otherwise the system in Eq. (12) could be simply described by an ordinary

differential equation. Since we are operating at low frequencies (even 10 times lower than the first mode of the
actuator) the following aproximations can be used:
N = T is constant (this is valid even at frequencies
higher than the first mode),
Z1 is a linear capacitor (the parasitic capacitance of
piezoelectric materials),
Z2 is a RLC series impedance where R (resistor) represents damping, L (inductor) represents equivalent
mass of the actuator, and C (capacitor) is related to
the mechanical stiffness.
In Figure 6, the elements related to the PZT actuator
are almost the same as the circuit in Figure 7. The difference is that, at the electrical port of the piezoelectric
actuator two elements have been added: a resistor Rloss
to deal with DC leakage at high fields and a nonlinear
capacitor responsible for the hysteretic behavior. In [4]
it has been pointed out experimentally why such hysteresis is rate independent (described by a capacitor whose
voltage-charge relation is a nonlinear hysteretic function
independent of frequency), and how to practically model
it.

4.2

Square Wave Driving

In Figure 6, on the left, the power supply and switching

stage are shown. Since power is a first concern in flying mechanisms, switching has to be used (linear drive
would lead to high power consumption) together with
techniques to reduce dissipation in switching capacitors
[16], i.e. in this case, the parasitic capacitance of the
piezoelectric actuator is the main issue. Assuming that
perfect square waves are obtained from the switching
stage, these waves are directly applied to the capacitor
(both the linear and the hysteretic one). By connecting
discharged capacitors to a power supply, high currents
are produced which charge capacitors rapidly and independently from the rest of the circuit (which is mainly
inductive, i.e slow to follow sudden voltage variations,
and behaves as if disconnected). Also, since capacitors
charge up (or even discharge in the second half of the
square wave) much faster than any other voltage variation in the rest of the circuit, square waves can be assumed on the capacitors. Therefore, a square wave source
just after the piezoelectric transformer (Tpzt is assumed
constant similarly to the transformer ratio in Figure 7)
is used. Here, Tpzt is constant over frequencies ranging from 0 to frequencies higher than the first mode, i.e.
more than 10 times higher than the first harmonic of our
square wave. This means that for the first 10 harmonics
of the square wave Tpzt is constant, and distortion will
only affect harmonics after the 10th one. Such a distortion will not modify the square wave since most energy
is distributed over the first 10 harmonics.
Next, Thevenins theorem is applied to the circuit in
Figure 8a, and an equivalent circuit is obtained in Figure 8b by neglecting PZT damping and equivalent mass.

Applying a square wave to this RLC circuit, and behavior is observed as given in Figure 9. Although we are
in the linear case, waveforms are distorted sinusoids. In
the linear case, the output can be thought as a square
wave filtered by the second order system. The filter is
a resonant one, and Lwing and N 2 Cstif f are designed to
resonate at 150 Hz . The outputs most relevant harmonic is the first (a sine at 150Hz, tuned with the resonant frequency of the filter) while the others will be attenuated. Therefore, higher harmonics are present with
a small amount of energy, and their contribution leads
to a distorted sine. By choosing a square wave with a
different duty cycle, it is possible to have an input with
different energy distributions among harmonics so that
can be reshaped.

Figure 8: (a) Thevenins theorem for linear networks is applied to the part of network on the left of the dashed line
and so it is independent on the load whether it is linear or
not. b) After applying Thevenins theorem, only the capacitor is left since PZT damping and equivalent mass (resistor
and inductor) are negligible compared to the load damping
and inertia.

Figure 9: Simulated wing stroke angle (solid line) and

angular speed (dashed line) shapes using a square wave

drive.

Piezoelectric actuators exhibit parasitic capacitance

and many problematic power issues arise from charging
or discharging such a capacitor by means of a constant
voltage source and a switching stage. In [16], Athas et al
showed that capacitors can be charged efficiently with a
voltage ramp (i.e. a constant current source) but circuits
for generating such an input tend to dissipate too much
energy in the power supply itself, even without a load.
The basic operation of a simple charge recovery system
is shown in Figure 10. This circuit introduces an inductor into the system to take advantage of the oscillatory
nature of LC circuits. When the switch is closed, the voltage across the capacitor is reversed. A proof-of-concept
experiment using low voltage and a real capacitor, rather
than a piezo, is shown in Figure 10d.

Actuator Characterization

During the experiments, unipolar input voltages are applied to diminish hysteresis and depolarization problems.

Figure 11: Optical detection setup with the laser and photodiode detector for actuator deflection measurements.

Figure 10: (a) LC oscillating circuit when current is positive; (b) inductor current and capacitor voltage vs. time; (c)
implemented circuit; and (d) acquired waveforms from circuit.
The unimorph tip deflection is measured using an optical
detection system as shown in Figure 11. In this setup,
a horizontal helium neon laser beam is focused onto a
perpendicular 1-D photo diode array (SL5-2, UDT Sensors Inc.). Then, the unimorph tip is positioned in between so that the tip motion is measured by the motion of its shadow on the sensor. A dynamic signal analyzer (HP-3562A) is utilized for frequency response characteristic measurements. For quasi-static measurements
a PC-based real-time (10 15 KHz) ADC and DAC
board (MultiQ data acquisition board, Quanser Consulting Inc.) is used in a Simulink programming environment.
The blocking force is measured with a semiconductor
strain gauge (Entran Inc., ESB-020) glued on the base of
a rigid brass beam where the setup is shown in Figure 12.
The unimorph tip is contacted to the beam end, and the
perpendicular force at the beam is measured as F1 . The
actuator output force also bends the flexible brass beam
with 1 displacement, and this bending is measured by a
side view optical microscope. Thus, measuring the free
deflection dc previously, Fb is computed from
Fb =

F1
1 1 /dc

(13)

For prototype 16 3 0.21 mm3 PZT-5H and 5

1 0.22 mm3 PZN-PT unimorphs, all parameters are
measured, and compared with the theoretical results as
given in Table 3. Here, superscript t denotes the theoretical values, and fr is measured at V = 12 V . For
each unimorph, hs = 76 m.The experimental resonant
frequency and blocking forces are lower than the theoretical ones, due to the nonlinearities at high voltages. On
the other hand, although dc of the PZN-PT unimorph
is around 4 times less than the PZT-5H one, it rotates

Figure 12: The photo of the blocking force measurement

setup with a rigid brass beam and a strain gauge on it.
1.25 times more since it is 3 times shorter. Thus, if the
proper T is selected as shown in Table 2, PZN-PT and
PZT unimorphs could have a similar flapping actuation
performance while the PZN-PT unimorph is around 8
times less in size and weight.
The electrical hysteresis of a 10 1 0.15 mm3 size
PZT-5H unimorph is measured using the circuit with a
series resistor and sinusoidal input as illustrated in Figure
13a. Here, the charge in the piezo actuator is integrated
from the measured current. Resulting charge hysteresis is
displayed in Figure 13b. The hysteretic area corresponds
to the dissipated energy by the hysteretic capacitor during a charging and discharging cycle, and it is computed
as 0.6 mW at 150 Hz. This value is small compared to
the theoretical 10 mW total output power on the load.
The effect of increasing the electric field on the mechanical hysteresis measured at 1 Hz is shown in Figure
14. As can be seen in the figure, mechanical hysteresis, i.e. the structural damping, increases by the higher
voltage.
For the dynamic response characteristic, the frequency
response plots (amplitude and phase) are illustrated in
Figure 15, and from the plot, fr and Q can be computed.
As can be seen from the figures, fr and Q are lowered by
increasing the input voltage. This effect could be due to

PZT-5H
PZN-PT

V
(V )
150
200

dc
(m)
408
96

t
dc
(m)
354
118

Fb
(mN )
51
41

Ft
b
(mN )
55
77

fr
(Hz)
395
2004

t
fr
(Hz)
463
2897

Q
21
41

Table 3. Measured and theoretical values for prototype

PZT-5H and PZN-PT unimorphs.

Deflection (m)

1600

(a)

1200
2

800

400
0
320

340

360

380

400

420

440

460

480

440

460

480

Frequency (Hz)

Phase (degree)

0
50
2

100

150
200
320

340

360

380

400

420

Frequency (Hz)

the nonlinear behavior of the actuator where the piezoelectric layer stiffness is decreased by the increased field
[12]. This stiffness change is also observed for our unimorphs as displayed in Figure 16. The PZN-PT becomes
relatively more compliant at the increased electric field
with respect to the PZT-5H unimorph.

Conclusion

Development of the unimorph actuators for a microaerial

flapping mechanism is discussed. PZT-5H and PZN-PT
are investigated as the piezoelectric layers of the unimorph actuator. Design issues for microaerial flapping
actuators, and actuator dimensions are determined for
biomimetic wing actuation. 16 3 0.21 mm3 PZT and
5 1 0.22 mm3 PZN-PT unimorphs are fabricated,
600

Deflection (dc) (m)

500

400

300

200

100

0
0

20

40

60

80

100

120

140

160

180

200

Input Voltage (V)

Figure 14: Mechanical hysteresis curves of the PZT-5H unimorph at quasi-static mode (1 Hz) for different voltages.

300
1

200
100
0
1940

1970

2000

2030

2060

2090

2120

2090

2120

Frequency (Hz)

Phase (degree)

(b)
Figure 13: Measured electrical hysteresis behavior of a 10
1 0.15 mm3 size PZT-5H unimorph.

Deflection (m)

400

80
120
2

160
1

200
240
1940

1970

2000

2030

2060

Frequency (Hz)

Figure 15: Shift in the resonant frequencies of the PZT (upper) and PZN-PT (lower) unimorphs depending on the applied voltage (1: V = 12 V , 2: V = 30 V ).
and their performances are tested. Nonlinear behavior of
the piezos are observed at high electric fields. A 1 d.o.f.
wing flapping mechanism is actuated by the PZT-5H unimorph with V = 85 V , and 180 stroke angle is achieved
at 95 Hz which shows the success of the unimorph design
(Figure 17).
In order to increase the displacement and output
torque performance of the unimorphs with smaller dimensions, prestressing could be a possible solution which
is future work. Moreover, switching-based low-loss driving electronics for the actuators will be fabricated as
an on-board module using microelectronics technologies.
Then, a compact and light weight microaerial robot actuation mechanism would become possible.
Acknowledgements
This work was funded by ONR MURI N00014-98-1-0671,
ONR DURIP N00014-99-1-0720 and DARPA. Authors would
like to thank to Robert Wood for creating the force measure-

600
PZT5H
PZNPT

Stiffness (Km) (N/m)

500

400

300

200

100

0
60

80

100

120

140

160

180

200

Input Volt (V)

Figure 16: The nonlinear drop in the mechanical stiffnesses

of PZT-5H and PZN-PT unimorphs by the increasing input
voltage.

ment setup, Srinath Avadhanula for constructing the 1 d.o.f.

four-bar wing structure, and Eric Park for actuator construction.

References
[1] R. Fearing, K. Chiang, M. Dickinson, D. Pick, M. Sitti,
and J. Yan, Transmission mechanism for a micromechanical flying insect, in Proc. of the IEEE Int. Conf.
on Robotics and Automation, San Francisco, USA, Apr.
2000.
[2] E. Precht and S. Hall, Design of a high efficiency,
large stroke, electromechanical actuator, Smart Mater.
Struct., vol. 8, pp. 1330, 1999.
[3] J. Yan, R. Wood, S. Avadhanula, M. Sitti, and R. Fearing, Towards flapping wing control for a micromechanical flying insect, in Proc. of the IEEE Int. Conf. on
Robotics and Automation, Korea, 2001 (to appear).
[4] M. Goldfarb and N. Celanovic, Modelling piezoelectric
stack actuator for control of micromanipulation, IEEE
Control System Magazine, vol. 17, no. 3, pp. 6979, 1997.
[5] Q. Wang, X. Du, B. Xu, and L. Cross, Theoretical analysis of the sensor effect of cantilever piezoelectric benders, J. of Aplied Physics, vol. 85, pp. 17021712, 1
Feb. 1999.
[6] M. Sitti, PZT actuated four-bar mechanism with two
flexible links for micromechanical flying insect thorax,
in Proc. of the IEEE Int. Conf. on Robotics and Automation, Korea, 2001 (to appear).
[7] A. Cox, D. Monopoli, and M. Goldfarb, Development
of piezoelectrically actuated elastodynamic flapping microaerial devices, in ASME Adaptive Structures and
Material Systems, pp. 257262, 1999.
[8] J. Smits and W. Choi, The constituent equations of
piezoelectric heterogenous bimorphs, IEEE Tran. on
Ultrasonics, Ferroelectrics, and Freq. Control, vol. 38,
pp. 256270, May 1991.
[9] M. Weinberg, Working equations for piezoelectric actuators and sensors, Journal of Microelectromechanical
Systems, vol. 8, pp. 529533, Dec. 1999.

Figure 17: A 1 d.o.f. flapping wing experimental result using

the PZT-5H unimorph actuators: up (upper image) and down
(lower image) strokes of the wing spar where 180 stroke angle
is achieved at 95 Hz.

[10] J. Smits and A. Ballato, Dynamic admittance matrix of

piezoelectric cantilever bimorphs, Journal of Microelectromechanical Systems, vol. 3, pp. 105111, Sept. 1994.
[11] J. Smits, W. Choi, and A. Ballato, Resonance and antiresonance of symmetric and asymmetric cantilevered
piezoelectric flexors, IEEE Trans. on Ultrasonics, Ferr.
and Freq. Control, vol. 44, pp. 250258, March 1997.
[12] Q. Wang, Q. Zhang, B. Xu, R. Liu, and L. Cross, Nonlinear piezoelectric behavior of ceramic bending mode actuators under strong electric fields, Journal of Applied
Physics, vol. 86, pp. 33523360, 15 Sept. 1999.
[13] Q. Wang, X. Du, B. Xu, and L. Cross, Electromechanical coupling and output efficiency of piezoelectric bending actuators, IEEE Trans. on Ultrasonics, Ferro. and
Freq. Control, vol. 46, pp. 638646, May 1999.
[14] S. Liu, S. Park, T. Shrout, and L. Cross, Electric field
dependence of piezoelectric properties for rhombohedral
0.955P b(Zn1/3 /N b2/3 /)O3 / 0.045P bT iO3 / single crystals, Journal of Applied Physics, vol. 85, no. 5, pp. 2810
14, 1999.
[15] O. Sotavalta, The wing stroke frequency of insects
in wing mutilation and loading experiments at subatmospheric pressure, Ann. Zool. Soc., vol. 15, no. 2,
pp. 167, 1952.
[16] W. Athas, J.G.Koller, and L. Svensson, An energy efficient CMOS line driver using adiabatic switching, in
Proc. of the Fourth Great Lakes Symp. on VLSI. Design
Automation of High Performance VLSI Systems GLSV,
p. 1994.