IOP PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING

J. Micromech. Microeng. 19 (2009) 094005 (8pp) doi:10.1088/0960-1317/19/9/094005

Vibration energy harvesting with

aluminum nitride-based piezoelectric
devices
R Elfrink, T M Kamel, M Goedbloed, S Matova, D Hohlfeld, Y van Andel
and R van Schaijk
IMEC/Holst Centre, Eindhoven, The Netherlands
E-mail: [email protected]

Received 13 January 2009, in final form 31 March 2009

Published 26 August 2009
Online at stacks.iop.org/JMM/19/094005

Abstract
This paper describes the measurement results of piezoelectric energy harvesters with
aluminum nitride (AlN) as a piezoelectric material. AlN was chosen for its material properties
and for its well-known sputter deposition process. For AlN devices a high optimum load
resistance is required, which is favorable due to the high resulting voltage level. The output
power harvested from mechanical vibrations has been measured on micromachined harvesters
with different geometries. The resonance frequencies ranged from 200 up to 1200 Hz. The
packaged devices had limited output powers and quality factors due to air damping caused by
the package. A maximum output power of 60 μW has been measured on an unpackaged
device at an acceleration of 2.0 g and at a resonance frequency of 572 Hz. The package of the
harvester requires special attention, since air damping can significantly decrease the maximum
power output.
(Some figures in this article are in colour only in the electronic version)

1. Introduction makes a fair comparison of the results difficult. On the

other hand, in the near future wireless sensor systems
Small wireless autonomous sensors become more and need to become smaller, cheaper and sophisticated, which
more standard components of our intelligent environment. will make them applicable in more environments. The
Nowadays, batteries are used to power the wireless sensor simple design of a piezoelectric transducer can help to
nodes, but size limitations prevent long autonomy. For this meet the conditions of a small footprint (<1 cm2) together
reason, a large effort is ongoing to replace batteries with with the use of micromachining as a potentially low-cost
more efficient power sources [1]. The aim is to generate manufacturing method. The design of the piezoelectric
and store power at the micro-scale to improve autonomy energy harvesting devices is similar to the classical design
and reduce the size of the wireless sensor systems. The of accelerometers and consists of a piezoelectric stack on a
choice of the harvesting principle depends on the applications thin cantilever beam with an attached mass. Micromachined
and photovoltaic, vibration, thermal or radiofrequency power piezoelectric energy harvesters generated 3 μW [6] and 40 μW
conversions are investigated [2]. For converting motion [7] for the PZT-based harvester and 2 μW [8] and 13 μW
or vibration, the established transduction mechanisms to (0.25% duty cycle) [9] for AlN-based devices. A comparison
generate electrical power are electrostatic, piezoelectric and is not obvious as dimensions, frequencies and vibration
electromagnetic [3, 4]. There is no clear conclusion conditions vary.
with respect to the preferred transducer technology and In this work, record output powers are presented for
design if power output performance from reported fabricated MEMS-based piezoelectric energy harvesters with AlN as a
devices is compared [5]. Results on energy harvesters piezoelectric material. The power output is in the order of
presented in the literature cover a large range of device tens of μW and makes it possible to use these small harvesters
dimensions, power output and resonance frequencies. This in low-power wireless sensor systems. Devices with different

0960-1317/09/094005+08$30.00 1 © 2009 IOP Publishing Ltd Printed in the UK

J. Micromech. Microeng. 19 (2009) 094005 R Elfrink et al

Table 1. Device dimensions.

Top wafer
1 Device Cantilever beam Mass Cantilever beam
type length (mm) length (mm) width (mm)

2 1 1.31 3.0 3.0

3 1.61 3.0 3.0
5 1.01 5.0 5.0
3 7 1.29 5.0 5.0
9 2.10 7.0 7.0
4

fabrication starts with the deposition of SiO2 and Si3N4

5
isolation layers (step 1). Next step is the deposition and
patterning of a Pt bottom electrode with the Ta adhesion layer
6 (step 2), followed by the deposition and patterning of the AlN
piezoelectric layer (step 3). The AlN films with a thickness
7 of 400 nm were deposited by reactive sputtering from an Al
target. The AlN layer typically consists of a thin amorphous
8 layer at the AlN–Pt interface and a columnar structure on
top. The CMOS compatible and relatively easy processing of
Piezoelectric device wafer Bottom wafer AlN is a major advantage compared with the more commonly
used piezoelectric material PZT. The aluminum top electrode
deposition and patterning finish the piezoelectric capacitor
stack (step 4). Next the trench patterning is done, which
defines the cantilever (step 5). The depth of the trench should
be deep enough to connect front to backside during the release
etch. On the backside of the wafer, the SiO2/Si3N4 is patterned
Packaged harvester as hardmask for KOH etching (step 6). The mass is shaped with
Figure 1. Process flow of a cantilever-type piezoelectric vibration KOH etching, but the etch stop is before complete release (step
energy harvester. 7). The complete release cannot be done with KOH due to bad
etch selectivity toward Al and AlN. The release is done either
with dry etching or wet etching with a tetramethylammonium
dimensions (all below 1 cm2) have been investigated to cover
hydroxide (TMAH) solution (step 8). The silicon cantilever
a broad frequency range (200–1200 Hz) and the design is
beam thickness after KOH etching and release etching is 45 ±
optimized for the maximum power output. AlN was chosen as
3 μm.
a piezoelectric material for its easy processing compared with
2 Glass wafers are used for the top and bottom covers. A
PZT. One can define a power generation figure of merit e31 /εr ,
cavity with a depth of 400 μm is etched into the glass wafers
with e31 the piezoelectric constant and εr the relative dielectric
with HF. In the top wafer contact holes for the wire bonding are
constant. For AlN the values for e31 and εr are −1.1 C m−2 and
fabricated by powder blasting. The process is completed with
10.5 respectively [15]. For PZT these values range from −8 up
a final bonding step to realize a packaged cantilever harvester.
to −12 C m−2 for e31 and 300 up to 1300 for εr [10]. Although
Different devices with different cantilever beams and
the piezoelectric coefficients are lower compared to PZT, the
mass geometries were produced. An overview of the used
power generation figure of merit of AlN is comparable or
dimensions is given in table 1. All results presented in
higher due to the lower dielectric constant. Together with the
this paper are measured on devices with a cantilever beam
well-known standard sputter-deposition techniques for AlN in
thickness of 45 ± 3 μm and an AlN thickness of 800 nm. The
contrast to the more complex deposition of PZT, AlN is a good
harvesters are packaged in between two glass substrates with
choice for piezoelectric energy harvesters. This paper will 400 μm deep cavities to allow mass displacement up to the
present the fabrication method, the experimental measurement cavity depth. Figure 2(left) shows the harvester design and its
method and results. package configuration at the rest position and figure 2(right)
shows the motion of the mass when the harvester is at
2. Harvester fabrication and design resonance.
Both the top and bottom electrodes of the piezoelectric
The piezoelectric harvesters consist of a silicon cantilever capacitor are connected through a small contact opening in the
beam ended with an attached mass and are produced by upper glass substrate. The harvester and a small PCB board
standard micromachining techniques [11]. The capacitor is are glued on a supportive epoxy board. Electrical contacts are
located on top of the cantilever beam and is formed by a made with wire bonds to a small PCB board on which standard
platinum bottom electrode, the piezoelectric AlN layer and an connectors are soldered (figure 3). In order to investigate
aluminum top electrode. the influence of the package, also unpackaged devices were
The process flow of the energy harvester with AlN measured. Therefore, the silicon frame of the harvester was
piezoelectric layers is schematically shown in figure 1. The mounted on 1.0 mm high pillars on the supportive epoxy board.

2
J. Micromech. Microeng. 19 (2009) 094005 R Elfrink et al

Cpiezo bondwire

glass

mass silicon

glass

mass beam
lenght lenght

Figure 2. (Left) Vibration energy harvester packaged in between glass substrates at the rest position. (Right) Motion of the mass of the
harvester.

Rload
harvester
A
vibration

Figure 4. Electrical measurement scheme.

curve. Besides the resonance frequency and maximum output

power, the quality factor is derived from the resonance curve.
We have used a simple harmonic oscillator model to fit the
measured resonance curve to a Lorentz function to obtain the
quality factor [12]. The quality factor is used in this paper as
a parameter to compare results.
Figure 3. Vibration energy harvester mounted on a supportive board The power output is measured as a function of the load
with electrical connectors. resistor. The maximum power output is obtained at a load
resistor matched to the impedance of the harvester. The
In this case the mass can displace unlimited up to pillar height matched load resistor can be calculated by (2):
in open air. Further, also partly opened devices were produced 1
and measured. Here both parts of the silicon frame and the RLoad = |Z| = , (2)
ω·C
glass substrates on the side of the mass tip were removed in
where Z is the harvester impedance, ω is the resonance
such a manner that the mass displacement was again limited
frequency and C is the capacitance of the harvester. Both
by the cavity depth, but no enclosed air cavities above or
ω and C are measured with a standard multimeter and the
underneath the mass were present.
calculated matched load resistor agrees within a few percent
with the measured one.
3. Experimental methods In order to measure the displacement of the mass an optical
microscope in combination with a digital camera was used.
We have investigated the AlN piezoelectric harvesters with an The camera was mounted perpendicular to the direction of
electrodynamic shaker, by applying a sinusoidal oscillation at vibration such that the displacement of the tip of the mass could
varying frequency and acceleration as a mechanical input. To be observed and measured. The calibrated camera images of
indicate the applied input acceleration we use the fractions of g, the moving mass have a measurement error less than 10 μm.
where g is the gravitational acceleration of 9.81 m s−2. When
alternating stress is present in the piezoelectric harvester due
to cantilever beam bending, an alternating current through a 4. Results and discussion
discrete resistive load is generated. The piezoelectric harvester
is represented as a capacitor generating the current which is 4.1. Packaged devices
measured with a current meter (figure 4). An example of a harvester’s power output with a capacitance
The average power P dissipated in the resistive load Rload of 0.6 nF and a resonance frequency of 611 Hz at 0.25 g as a
is calculated according to (1) using the measured root mean function of the discrete load resistor value is given in figure 5.
squared (RMS) value of the ac current IRMS and the used A maximum power of 0.17 μW is dissipated when a load
discrete load resistor value: resistor of 450 k is used. The empirically found optimum
P = IRMS2
· RLoad . (1) load resistor value is in accordance with the estimated value
The power plotted as a function of the frequency at a certain as calculated by equation (2). The curve shows the decrease
load resistor and constant acceleration forms the resonance of output power when a non-optimum load is used.

3
J. Micromech. Microeng. 19 (2009) 094005 R Elfrink et al

0.20 1

type 1
type 3
type 5
0.15
type 7
0.1 type 9

Power[μW]
0.10

0.01
0.05

0.00
1.0E+04 1.0E+05 1.0E+06 1.0E+07 0.001
0 200 400 600 800 1000 1200
Load resistance [Ω] Frequency [Hz]

Figure 5. Load resistor matching for a packaged device type 5 at Figure 6. Resonance curves at 1.0 g for devices with different
0.25 g at the resonance frequency. geometries. They are packaged, which causes considerable air
damping and decreases the output power.
Within the used set of AlN devices, each with its
specific capacitance and resonance frequency, the optimum 100
1.0 g
load resistance was always in the range of 0.1–1.0 M. For
2.0 g
devices based on lead zirconate titanate (PZT) with the same
4.0 g
dimensions, the optimum load resistance will be in the order of 10 8.0 g
a few k due to its higher dielectric constant and thus resulting
higher capacitance. This high optimum load resistance for
Power[μW]

AlN has a positive effect on the generated voltage level over

1
the load. In the case when the AlN device generates a current
of 4 μA through a load of 500 k, the dissipated power equals
8 μW and the voltage level over the resistive load is 2 V. A PZT-
0.1
based device operating at the same output power will result in
a voltage level of only a few mV. This higher voltage level
of AlN-based devices is favorable for future power processing
circuits, such as rectifiers or battery chargers. In this case 0.01
500 550 600 650 700
rectification can be done by a regular diode bridge which Frequency [Hz]
makes ac–dc conversion easier. The higher output voltage
Figure 7. Resonance curves for a packaged device type 5 at
makes conversion easier toward a battery voltage level [13]. different acceleration forces.
The resonance curves of devices with different
dimensions, as given in table 1, are presented at an acceleration
measurements, the quality factor decreased from 42 to 25
level of 1.0 g in figure 6. It shows that by controlling the
dimensions of the cantilever beam and mass, the resonance for acceleration levels of 1.0 g and 8.0 g respectively. This
frequency can be set. For this set of devices, the cantilever decrease of the quality factor can be seen in the figure as
beam thickness was 45 ± 3 μm yielding a frequency between a broadening of the resonance curves. The degradation of
277 Hz (device type 9) and 1110 Hz (device type 1). The the quality factor is caused by increasing damping at larger
measured resonance frequencies were consistent with the mass displacements. The applied acceleration of 8.0 g is
predicted resonance frequencies obtain by modeling [14]. considered as high, while the power output and quality factor
All devices gave a maximum output power between 0.3 and are considered as low. We assume that the air in the cavities
0.8 μW. The quality factors of the individual resonance curves of the package is compressed by the movement of the mass
are between 18 for the largest device and 87 for the smallest in such a way that the system behaves like a damping piston.
device. We consider this to be low for a vibration cantilever The mechanical damping caused by air is a loss factor which
beam system and interpret it as an indication that the vibration reduces the output power.
of the mass is damped, in the present case most likely by air The maximum mass tip displacement is limited by the
in between the mass and the cavity walls. dimensions of the cavity and is at maximum 400 μm from
Figure 7 shows the resonance curves of a packaged device the initial position. For this device, the displacement at 8.0 g
type 5 at acceleration levels from 1.0 g up to 8.0 g. In was about 250 μm at the resonance frequency. As will be
this case a maximum power output of 17 μW is measured shown later, an unpackaged device with less damping, and
at the acceleration of 8.0 g. The quality factor decreases with consequently higher mass displacement, will lead to higher
acceleration, indicating an increase in damping. For these output power at lower acceleration.

4
J. Micromech. Microeng. 19 (2009) 094005 R Elfrink et al

100 70
packaged
opened package
unpackaged 60

10
50

Power [μW]
40
Power [μW]

1
30

20
0.1

10

0
0.01 550 560 570 580 590
0.90 0.95 1.00 1.05 1.10
Normalized Frequency [Hz]
Frequency [Hz]

Figure 8. Resonance curves at 2.0 g of three type 5 devices with Figure 9. Resonance curve of an unpackaged device type 5
different packaging, resulting in different air damping and thus generating a maximum output power of 60 μW at 2.0 g.
output power and quality factor.
the input displacement and should equal the quality factor.
4.2. Unpackaged devices The displacement at the center of the mass can be calculated
from the device geometry and the measured displacement at
A comparison between three devices of type 5 with different the tip of the mass. For this measurement, the mechanical
packaging and therefore different air damping is visualized amplification factor equals 160. The quality factor of this
in figure 8. The frequency scale on the x-axis is normalized resonance curve obtained from the fit to the Lorentz function is
to compensate for a small frequency shift between the three 170. The difference between the electrically measured quality
devices (572, 592 and 611 Hz respectively). The frequency factor and the mechanical measured amplification factor is
shift is caused by a small difference in cantilever beam considered to be within the measurement accuracy.
thickness due to process variation over the silicon wafer. For this device, the output power as a function of the
The packaged device generated an output power of input acceleration (ainput) at the resonance frequency is shown
2.1 μW. The device for which the package is partly opened in figure 10(left). At low acceleration levels up to about
at the mass tip generated 22 μW. The unpackaged device 1.0 g, the slope of this double logarithmic plot is close to
generated 60 μW. All measurements were performed at an +2, indicating a quadratic relation between the output power
acceleration of 2.0 g. The corresponding quality factors are and the input acceleration. This is according to a general
37, 110 and 170 respectively. The bandwidth increases when model of a velocity damped oscillator for the conversion of
air damping is present. However, the power output is lower mechanical energy of a vibrating mass into electrical power
at all frequencies. The air damping introduces mechanical [4]. At higher acceleration levels, the slope seems to reduce
loss and thus reduces the displacement of the mass. This toward a value of +1. This change from quadratic to linear
especially suppresses the resonance peak, which leads to behavior is the result of increased damping with the input
a lower output power and lower quality factor and should acceleration. This is possible due to the larger cantilever
therefore be prevented. beam bending or due to increased air damping at higher mass
A further increase of output power and quality factor can velocity. This increase in damping is measured as a decrease
be expected when air damping is fully prevented by the use of in quality factor. In this case, the quality factor decreased
vacuum packaged devices. In that case the moving mass is not from 326 down to 160 for accelerations of 0.05 g and 2.0 g
damped by air and the mass will have a larger displacement at respectively. In figure 10(right), the output power is plotted as
lower excitation force, leading to a higher power output and a function of the mass acceleration (amass); the latter is defined
higher quality factor. as follows:
Figure 9 shows once more the resonance curve of the
unpacked device type 5 with power now plotted on a linear amass = ainput · Q. (3)
scale. The resonance curve is slightly asymmetrical, most The slope of the mass acceleration versus output power is
likely due to nonlinear behavior of large mass displacements. now constant +2.0 for the measured range.
The output power at resonance is 60 μW. The input The resonance curves for a set of unpackaged devices at
acceleration of 2.0 g is corresponding to a peak-to-peak an acceleration of 1.0 g are given in figure 11. All unpackaged
input displacement of 3.03 μm at the resonance frequency devices gave a maximum output power between 8 and 35 μW.
of 572 Hz. The maximum peak-to-peak displacement at the The quality factors of the individual resonance curves are
tip of mass was 895 μm, which was measured with an optical between 135 for the largest device and 340 for the smallest
microscope. The mechanical amplification factor is calculated device. In table 2, an overview of both the results of the
by dividing the displacement at the center of the mass by packaged and unpackaged devices is given. Compared with

5
J. Micromech. Microeng. 19 (2009) 094005 R Elfrink et al

100 100

10 10
Power [μW]

Power [μW]
1 1

0.1 0.1
0.01 0.1 1 10 10 100 1000
Acceleration/g Mass acceleration/g

Figure 10. (Left) Power output as a function of the input acceleration at the resonance frequency. (Right) Power output as a function of the
mass acceleration.

Table 2. Comparison of packaged and unpackaged devices at 1.0 g.

Device Power (μW) Quality factor
Type Packaged Unpackaged Packaged Unpackaged Packaged Unpackaged
1 1111 1110 0.42 8.2 87 337
3 999 1002 0.38 10.4 73 426
5 611 573 0.62 30.0 42 252
7 523 528 0.70 22.4 39 216
9 277 294 0.72 32.2 18 135

the set of packaged devices, both the output power and the
100
quality factor are considerably higher for the unpackaged ones. type 1
This effect is explained by the reduced air damping compared type 3
type 5
with the high air damping in the cavities of the packaged type 7
10
devices. One should consider that the contribution of air type 9
damping to the total damping is dependent on the cantilever
Power [μW]

beam and mass size. A device with a large surface area

will have a higher air damping contribution compared with 1

a device with a small surface area. As mentioned before,

a further increase in the power output and quality factor is
expected for vacuum packaged devices where no air damping 0.1
is present. The difference in resonance frequency between
the same device types is maximum 6% and again explained
by a small difference in cantilever beam thickness caused by 0.01
process variation over the silicon wafer. The difference in 0 200 400 600 800 1000 1200
Frequency [Hz]
resonance frequency due to the packaging, caused by different
damping, has a minor influence on the resonance frequency Figure 11. Resonance curves at 1.0 g of unpackaged devices
and is not considered here. showing a higher output power compared to packaged devices.

4.3. Displacement and electrical damping relationship of displacement and current is measured, one
can choose to use the current measurement to obtain the
With the use of the optical microscope, the displacement of displacement of the mass.
the tip of the moving mass has been measured. At constant Harvesting energy can be visualized as a reduction in
load resistance, the measured current should be linear with mass displacement due to electrical damping. This is shown in
the displacement of the mass [4]. This linearity is shown by figure 13 for an unpackaged device with reduced air damping.
simultaneous measurement of displacement and current while The mass tip displacement is measured as a function of load
the frequency or the acceleration is varied. Figure 12 shows resistor. When the load resistor has a very low (short circuit)
an example for an unpackaged device type 5 at a constant or very high (open circuit) value, the energy dissipation in
frequency while the acceleration was varied. Once this linear the load is low (see figure 5). At the optimum load, resistor

6
J. Micromech. Microeng. 19 (2009) 094005 R Elfrink et al

Table 3. Measured resonance frequency, power and quality factor at

12
0.25 g for various load resistor values.
Rload (k) Resonance Power Quality
10
frequency (Hz) (μW) factor

8
28 571.94 0.95 554
200 572.09 3.47 427
Current [μA]

447 572.28 3.73 389

6 1000 572.51 3.50 421
7100 572.56 1.04 539
4

4.0
2 Ropt (447 kΩ)
0.5*Ropt (200 kΩ)
2*Ropt (1.0 MΩ)
0 Rshort (28 kΩ)
0 200 400 600 800 100 0 3.0 Ropen (7.1 MΩ)
mass tip displacement pk-pk [μm]

Power [μW]
Figure 12. Current through the used load resistor versus the
peak-to-peak displacement of the tip of the mass while the 2.0
acceleration was varied between 0.05 g and 2.0 g.

380 572.8 1.0

mass tip displacement
resonance frequency

340 572.6
Mass tip displacement [μm]

Resonance frequency [Hz]

0.0
568 569 570 571 572 573 574 575 576
Frequency [Hz]
300 572.4

Figure 14. Resonance curves at Ropt/16, Ropt/2, Ropt, 2 × Ropt and

16 × Ropt showing the lower power output, higher quality factor and
260 572.2
frequency shift for the non-optimum load resistor case.

220 572.0
maximum power output is obtained at the optimum load of
447 k. As mentioned above, the electrical damping is then
maximal resulting in a quality factor of 389. The two curves
180 571.8 at about half (200 k) and about twice (1.0 M) the optimum
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
Loadresistance [Ω]
load show a lower power output and a higher quality factor
of about 425. The shift in resonance frequency can be clearly
Figure 13. Load resistor matching for an unpackaged device type 5 distinguished. For the two dotted lines, a low load resistor
at 0.25 g at resonance.
value of 28 k and a high load resistor value of 7.11 M have
been used. With respect to the optimal load at the resonance
value maximum power is transferred from the harvester to the frequency, the use of these resistor values can be considered
load. Energy dissipation in the load (harvesting) is considered as short circuit and open circuit situations, while the power
as electrically damping the mechanical displacement of the output is still at a measurable high level of about 1.0 μW.
mass. In figure 13, one can observe that both in the open and Again the quality factor has increased to about 550, indicating
short circuit cases the electrical damping is minimal and the a further decrease in electrical damping, thus less energy is
measured mass displacement of about 320 μm is maximal. At harvested. Compared with the curve at the optimum load,
the optimum load resistor value of 450 k, the measured mass the output power is significantly lower, the quality factor is
displacement of 200 μm is minimal and thus the harvesting, higher and the shift in frequency is maximal. At optimum load
electrical damping, is at its maximum. resistor, both the power and the bandwidth are maximized. An
Another observation is the change in resonance frequency overview of the measured results is presented in table 3.
with the load resistor value. At low load resistor values, the
resonance frequency is about 0.6 Hz lower compared with the
situation where high load resistors are applied. This change in 5. Conclusions
resonance frequency is described by Renaud et al and can be
used to calculate the generalized electromechanical coupling We presented measurement results of AlN-based piezoelectric
factor [15]. It is noticed that at the optimum load resistor energy harvesters. AlN was chosen as a piezoelectric material
value, the resonance frequency can be found in between the for its material properties and for its well-known sputter
open and short circuit frequencies. deposition process. For AlN devices, a higher optimum load
The resonance curves at a constant acceleration of 0.25 g resistance is required compared with the lower optimum load
and various load resistor values are given in figure 14. The resistance for PZT-based devices. This favors the use of AlN

7
J. Micromech. Microeng. 19 (2009) 094005 R Elfrink et al

due to the higher resulting voltage level over the load which [2] Hudak N S and Amatucci G G 2008 Small-scale energy
makes ac–dc conversion easier. harvesting through thermoelectric, vibration, and
radiofrequency power conversion J. Appl.
By controlling the dimension of the cantilever beam and
Phys. 103 101301
mass, the resonance frequencies were set within the range [3] Beeby S P, Tudor M J and White N M 2006 Energy harvesting
of 200 up to 1200 Hz. The devices were packaged with vibration sources for microsystems applications Meas. Sci.
top and bottom glass substrates with cavities to allow mass Technol. 17 R175
displacement up to 400 μm. The packaged devices have [4] Roundy S, Wright P K and Rabaey J M 2003 Energy
Scavenging for Wireless Sensor Networks (Dordrecht:
limited output powers and quality factors due to air damping
Kluwer)
caused by air inside the package. We have shown that [5] Mitcheson P D, Reilly E K, Toh T, Wright P K and Yeatman
unpackaged devices with reduced air damping generate more E M 2007 Performance limits of the three MEMS inertial
power at all frequencies. Air damping is a loss factor which energy generator transduction types J. Micromech.
leads to lower power output and quality factor and should be Microeng. 17 S211
[6] Glynne-Jones P, Beeby S P, James E P and White N M 2001
minimized. For the unpacked devices, a maximum output
The modeling of piezoelectric vibration powered generator
power of 60 μW has been measured for the device type 5 at for microsystems Int. Conf. on Solid-State Sensors,
an acceleration of 2.0 g at a resonance frequency of 572 Hz. Actuators and Microsystems, Transducers p 46
It is expected that higher output power and quality factor can [7] Renaud M, Sterken T, Schmitz A, Fiorini P, Van Hoof C and
be obtained with a vacuum package design. Puers R 2007 Piezoelectric harvesters and MEMS
technology: fabrication, modeling and measurements Int.
A linear relationship between current through the load
Conf. on Solid-State Sensors, Actuators and Microsystems,
resistor and the displacement of the mass could be measured Transducers p 891
when frequency or acceleration is varied. This relationship [8] Marzencki M, Ammar Y and Basrour S 2007 Integrated power
can be used to obtain the mass displacement from the current harvesting system including a MEMS generator and a
measurement. The process of harvesting, i.e. dissipating power management circuit Int. Conf. on Solid-State
Sensors, Actuators and Microsystems, Transducers p 887
power in the load resistor, can be recognized from a reduction
[9] Duggirala R, Polacawich R G, Dubey M and Lal A 2008
of mass displacement, which is due to electrical damping. Radioisotope thin-film fueled microfabricated reciprocating
We observed that the optimum load resonance frequency electromechanical power generator J. Microelectromech.
lies in between the short circuit and open circuit resonance Syst. 17 837
frequencies. A maximum power output and bandwidth are [10] Trollier-McKinsky S and Muralt P 2004 Thin film
piezoelectrics for MEMS J. Electroceramics 12 7–17
obtained at the optimum load resistor.
[11] van Schaijk R, Elfrink R, Kamel T M and Goedbloed M 2008
Piezoelectric AlN energy harvester for wireless autonomous
transducer solutions IEEE Sensors Conf. p 45
Acknowledgment [12] Salapaka M V, Bergh H S, Lai J, Majumdar A and McFarland
E 1997 Multi-mode noise analysis of cantilevers for
We would like to thank Dr Devrez M Karabacak for his helpful scanning probe microscopy J. Appl. Phys. 81 2480–7
discussions regarding data analysis. [13] Doms I, Merken P, Mertens R P and van Hoof C 2008
Capacitive power-management circuit for micropower
thermoelectric generators with a 2.1 μW controller IEEE
Int. Solid-State Circuits Conf. p 300
References [14] Matova S, Elfrink R and van Schaijk R 2008 Modelling and
validation of AlN piezoelectric harvesters Eurosensors pp
[1] Cook-Chennault K A, Thambi N and Sastry A M 2008 1482–5
Powering MEMS portable devices—a review of [15] Renaud M, Karakaya K, Sterken T, Fiorini P, van Hoof C and
non-regenerative and regenerative power supply systems Puers R 2008 Fabrication, modelling and characterization
with special emphasis on piezoelectric energy harvesting of MEMS piezoelectric vibration harvesters Sensors
systems Smart Mater. Struct. 17 043001 Actuators A 145–146 380