This is an author-created, un-copyedited version of an article accepted for publication in Smart Materials and

Structures. IOP Publishing Ltd is not responsible for any errors or omissions in this version of the manuscript or
any version derived from it. The definitive publisher-authenticated version is available online at
doi:10.1088/0964-1726/20/10/105013
S Boisseau et al 2011 Smart Mater. Struct. 20 105013
http://iopscience.iop.org/0964-1726/20/10/105013

Cantilever-based electret energy harvesters

S Boisseau1, G Despesse1, T Ricart1, E Defay1 and A Sylvestre2
1
2

CEA/LETI, 17 avenue des martyrs, Minatec Campus, Grenoble, France

G2ELab, CNRS, 25 avenue des Martyrs, Grenoble, France
E-mail: sboisseau@gmail.com
Abstract. Integration of structures and functions has permitted to reduce electric consumptions of
sensors, actuators and electronic devices. Therefore, it is now possible to imagine low-consumption
devices able to harvest energy in their surrounding environment. One way to proceed is to develop
converters able to turn mechanical energy, such as vibrations, into electricity: this paper focuses on
electrostatic converters using electrets. We develop an accurate analytical model of a simple but efficient
cantilever-based electret energy harvester. Therefore, we prove that with vibrations of 0.1g (~1m/s), it is
theoretically possible to harvest up to 30W per gram of mobile mass. This power corresponds to the
maximum output power of a resonant energy harvester according to the model of William and Yates.
Simulations results are validated by experimental measurements, raising at the same time the large impact
of parasitic capacitances on the output power. Therefore, we only managed to harvest 10W per gram
of mobile mass, but according to our factor of merit, this puts us in the best results of the state of the art.

1. Introduction
Thanks to size reduction, micro-electro-mechanical-systems (MEMS) are consuming less and less
energy, giving them the opportunity to harvest energy in their surrounding environment. This field of
research called Energy Harvesting consists in the development of converters able to turn ambient
energy (light, air flux, variation of temperatures, vibrations) into electricity that is used to power the
microsystem. Many principles of conversion have already been developed: photovoltaic,
thermoelectric, biofuelcells... As for mechanical energy from vibrations, it can be converted by three
main different principles: piezoelectric, electromagnetic and electrostatic conversion. In this study, we
focus on electrostatic converters which are based on a capacitive architecture (two charged electrodes
spaced by an air gap) and connected to a load. Vibrations induce changes in the geometry of the
capacitor and a circulation of charges between electrodes through the electrical load. The electronic
circuit that manages power conversion of standard electrostatic energy harvesters [1] is quite
complicated and induces losses and therefore a decrease of efficiency. To limit the use of a
management electronic circuit, it is possible to use electrets (stable electrically charged dielectrics)
that polarize the capacitance and allow to harvest energy from vibrations without using cycles of
charging and discharging.
Many electret-based energy harvesters have been developed and have proven the interest of such
devices [2-22]. Most of these devices are in-plane structures where the variation of capacitance is
obtained by a variation of surface between patterned electrodes, while the gap is kept constant. These
structures are generally hard to manufacture using elaborate clean room processes and especially
DRIE (Deep Reactive Ion Etching) but give the opportunity to harvest energy when the vibrations of
the ambient environment are not constant because they avoid contacts between electrets and
electrodes. In this paper, we have chosen to study a simpler structure: the cantilever-based electret

energy harvester. This structure does not maximize the output power of the energy harvester if
vibrations are not well defined but is particularly suitable when vibrations are stable in terms of
frequency and amplitude in the time. Moreover, this structure is quite easy to manufacture and
therefore low-cost.
In section 2, we present the theory of vibration energy harvesting and more especially of energy
harvesting using electrets. Then, we develop an accurate analytical model, its implementation (under
Matlab/Simulink) and its validation using FEM (Finite Element Method). Thanks to this model, the
structure can be optimized, as presented in section 4. Finally, in section 5, we present experimental
results and a comparison to simulation results. We finally develop a model taking parasitic
capacitances into account to explain the differences between our first model and experimental results.

2. Electret-based energy harvesters using cantilevers

Our electret-based energy harvester is a microsystem able to convert mechanical energy from
vibrations into electricity. It is part of vibration energy harvesters whose general model is presented
hereafter.
2.1. William and Yates general model for vibration energy harvesters
Regardless of the conversion principle (electrostatic, electromagnetic or piezoelectric), resonant
energy harvesters can be modeled as a mobile mass (m) suspended to a support by a spring (k) and
damped by forces (felec and fmec). When a vibration occurs y (t ) = Y sin(t ) , it induces a relative
displacement of the mobile mass x(t ) = X sin(t + ) compared to the frame (figure 1). Part of the
kinetic energy of the moving mass is lost due to mechanical damping ( fmec) while the other part is
converted into electricity, which is modeled by an electrostatic force (felec) in electrostatic energy
harvesters. Ambient vibrations are generally low in amplitude (typically Y=25m) and the use of a
mass-spring structure enables to take advantage of a resonance phenomenon that amplifies the
amplitude of vibrations perceived by the mobile mass and the harvested energy. Newtons second law
gives the differential equation that rules the movement of the mobile mass (1).
m&x& + kx + f elec + f mec = m&y&

(1)
frame

Mobile mass
(m)
felec

x(t)

fmec

y(t)

Figure 1. Mechanical system.

When forces can be modeled as viscous forces, f elec = be x& and f mec = bm x& , where be and bm are
respectively electrical and mechanical damping coefficients, William and Yates [23] have proven that
the maximum output power of a resonant energy harvester subjected to an ambient vibration is reached
when the natural angular frequency ( n ) of the mass-spring structure is tuned to the angular frequency
of ambient vibrations ( ) and when the damping rate e = be (2m n ) of the electrostatic force f elec

is equal to the damping rate m = bm (2mn ) of the mechanical friction force f mec . This maximum

output power PW&Y can be simply expressed with (2), when e = m = .

PW &Y =

mY 2 n
16

(2)

As PW&Y is a good approximation to know the output power of vibration energy harvesters when
forces are modeled as viscous forces, comparing the output power (P) of a resonant energy harvester
to PW&Y gives a legitimate factor of merit W&Y:
P
W &Y =
(3)
PW &Y
Nevertheless, in many studies, the weight of the mobile mass is not given while the surface area of
the electrodes (S) is often provided. Therefore, to compare systems, we had developed, in a previous
study, an other factor of merit, normalized by the active surface S in place of the mass [24]:
P
= 2 3
(4)
Y S
These two factors of merit will be used in the next parts, to compare our system to the state of the
art.
2.2. Cantilever-based electret energy harvesters Principles and Model

The particularity of electret-based energy harvesters is the use of an electret to maintain the
electrostatic converter charged through time. Electrets are dielectrics able to keep an electric field (and
a surface voltage V) for years thanks to charge trapping (figure 2). These materials are in electrostatics,
the equivalent to magnets in magnetostatics.

Figure 2. Electret film.

Electrets are obtained by implanting electric charges into dielectrics. Theoretically, dielectrics do
not conduct electricity; therefore, the implanted charges stay trapped inside. Many techniques exist to
manufacture electrets [25, 26], but the most common is the corona discharge (figure 3). It consists in a
point-grid-plane structure whose point is subjected to a strong electric field: it leads to the creation of a
plasma, made of ions. These ions are projected onto the surface of the sample to charge, and transfer
their charges to its surface. This mechanism results in the implantation of charges at the surface (figure
2), into the bulk or at the interfaces of the material. The grid is used to limit the surface voltage of the
electret to a wanted final value.

Figure 3. Corona discharge used to charge our device.

Nevertheless, dielectrics are not perfect insulators and implanted charges can move inside the
material or can be compensated by other charges or environmental conditions, and finally disappear. A
focal area of research on electrets concerns their stability [15, 27, 28]. Nowadays, many materials are

known as good electrets able to keep their charges for years: for example, Teflon and silicon dioxide
(SiO2) whose stability is estimated to more than 100 years [29-32].
The structure able to turn vibrations into electricity using electrets is introduced in figure 4: the
system is composed of a counter-electrode and an electrode on which is deposited an electret, spaced
by an air gap and connected by an electrical load (here a resistor). The electret has a constant charge
Qi, and, due to electrostatic induction and charges conservation, the sum of charges on the electrode
and on the counter-electrode equals the charge on the electret: Qi = Q1+Q2.

Figure 4. Electrostatic converter using electret.

When a vibration occurs, it induces a change in the capacitor geometry (e.g. the counter-electrode
moves away from the electret, changing the air gap and then the electret influence on the counterelectrode) and a reorganization of charges between the electrode and the counter-electrode through the
load. This induces a current across the load R and part of the mechanical energy is then turned into
electricity.
The converter introduced in figure 4 is integrated into a clamped-free beam mechanical structure
(figure 5). The lower face of the beam is metalized and is used as the counter-electrode. The electret
and the electrode are placed under the beam, separated by an air gap and electrically connected by a
load (figure 5(a)). According to William and Yates formula (2), the output power is proportional to
the mobile mass; consequently, to increase the output power, a proof mass m is added at the free end
of the cantilever. Vibrations y(t) induce a relative displacement x(t) of the mobile mass compared to
the electrode. Structure parameters are presented in figure 5(b): h is the beam thickness, w its width, L
the length between the clamping and the gravity center of the mass, 2Lm the mass length, g(t) the
thickness of the air gap between the counter-electrode and the electret, g0 the thickness of this air gap
r
without vibrations and without the weight effects W , V the surface voltage of the electret, C1 the
capacitance of the electret, C2 the capacitance of the air gap and finally the electrode length.

(a)

(b)
Figure 5. (a) Cantilever-based energy harvester using electrets. (b) Energy harvester parameters.
To identify the main parameters of this kind of energy harvesters and to maximize the output power
for a given vibration ( y (t ) = Y .sin(t ) ), it is necessary to find coupled mechanical and electrostatic
equations that rule the energy harvester.

3. Analytical model of the Cantilever-based electret energy harvester

To determine the output power of the energy harvester for a given vibration y(t), it is necessary to
solve the equation of motion and to find the quantity of charge transferred to the output. Therefore, the
goal of section 3 is to develop the analytical model of the cantilever-based electret energy harvester
parameterized in figure 5 for mechanical and electrostatic parts.
3.1. Model of the mechanical system

The clamped-free beam with a mass at the free end can be modeled as a damped mass-spring
r
r
structure as presented in figure 1 and by adding the effect of weight W = mg . The mechanical friction

forces can be modeled as viscous forces ( f mec = bm x& ) and the electrostatic force is the derivative of
the electrostatic energy of the capacitor We with respect to the displacement x. We is equal to the
charge on the upper electrode Q2 squared, divided by two times the capacitance as a function of time
C(t). Thereby, the mechanical system is ruled by (5).
d
d Q2
m&x& + bm x& + kx (We ) mg = m&y& m&x&+ bm .x& + kx 2 mg = m&y&
(5)
dx
dx 2C(t)
To maximize the output power of the energy harvester, the natural angular frequency ( n = k m )
of the mass-spring structure has to be tuned to the angular frequency of the ambient vibrations ( ).
Moreover, according to equations from mechanical structures theory, the spring constant k can be
deduced from the beam geometric parameters as follow:
3EI Ewh 3
2
(6)
k = m n = 3 =
L
4L3
Where E is the Youngs modulus and I the quadratic moment of the beam.
Because of the mass, the behavior of the beam has to be studied on two parts. A drawing of the
structure is presented in figure 6 and shows the deformation of the cantilever (z) as a function of the
position on the cantilever z for a forced deflection x at z=L. The first part ( z [0, L1 = L Lm ] ) does not
have an additional mass: its behavior corresponds to the one of a clamped-free beam whose deflection
x
at the end (x1) is imposed and given by (z) = 1 3 z 2 (3L1 z ) . The second part that has the additional
2L1

mass ( z [L1 , L2 = L + Lm ] ) follows the deflection of part 1: the derivative of the deflection ((z))
with respect to the position (z) for part 2 is constant and equal to the derivative of the deflection of part
1 at z=L1 (7).
d ( z )
d ( z )
3 x1
c=
=
=
with x1 = x cLm
(7)
dz z = L
dz z[L , L ] 2 L1
Therefore, for a given static deflection (x) on the position L of the beam, the deformation of the
beam can be simply expressed as a function of the parameters in both parts:
x

2
(z) = 1 3 z (3L1 z ) [part 1]
(8)
2L1

(z) = c( z L ) + x
[part 2]
1

Figure 6. Deformation of the cantilever for an imposed deflection (x).

Figure 7 presents the beam deformation resulting from equation (8) for a beam of L=30mm and
Lm=2mm and for an imposed static displacement of x=300m compared to the deformation performed
by FEM calculation (Comsol Multiphysics). It proves that our calculations fit with FEM results.

Figure 7. Deformation of the cantilever for an imposed static deflection (x).

Nevertheless, the problem we want to solve is not static but dynamic. Therefore, it is useful to verify
that the beam deformation behavior is the same in dynamic and in static. We have verified this using
FEM : it confirms that the deformation in dynamic and in static can be considered as equivalent. Thus,
we can consider that the deflection in dynamics can be simply expressed with (8) assuming that x is
the imposed deflection on the mass gravity point.
3.2. Modeling of the electrostatic system
The equivalent model of the energy harvester is presented in figure 8, where Q2 is the charge on the
counter-electrode, V the surface voltage of the electret and C(t) the capacitance between the beam and
the electrode. This capacitance corresponds to the serial capacitance formed by the electret dielectric
material capacitance C1 and the air gap capacitance C2(t). Kirchhoffs laws give the differential
equation that governs the electrostatic system (9):

Figure 8. Equivalent electric model of the energy harvester.

dQ 2 V Q 2 1 V Q 2 1
1
=

=
dt
R
R C (t ) R
R C 1 C 2 (t )

Moreover, the electrostatic energy stored in the capacitor is:

1 Q 22 (t )
We =
2 C (t )

(9)

(10)

To solve (9), it is necessary to know the capacitance of the electrostatic converter as a function of
the imposed deflection (x). Knowing the cantilever deformation, and considering a capacitor of
infinitesimal length (dz) (figure 6), one can get the infinitesimal capacitance on both part (dCp1 and
dCp2) for a given x.

0 w.dz
x

with (z) = 1 3 z 2 (3L1 z ) [part 1]

dC p (x) =
d
2L1
g 0 ( z )+

0 w.dz
dC p (x) =
with (z) = c( z L ) + x
[part 2]
d

g 0 (z )+

(11)

By integrating these expressions, the total capacitance between both electrodes is:
d

g 0 + + cLm x
L
0 w
r
dz

ln
+
C ( x) = C p (x) + C p (x) = 0 w
(12)
2

d
c

z
(
3L
z)
d
L
g
+

cL

x
g0 x
+
m
0

2L3
r
r

The integral defining Cp1(x) cannot be analytically calculated and will be numerically computed.
1

This capacitance expression has been compared to a FEM simulation and the curves presented in
figure 9 show that results are in excellent agreement. These results were also compared to the formula
of a simple plane capacitor neglecting fringe effects ( C ( x) = 0 S (g 0 x + d / r ) ), where S is the
surface of the electrodes, g0 the initial gap and x the imposed deflection. With our parameters
(Lm=2mm, L=30mm, g0=505m, d=100m, w=12.33mm, r=2, =10mm), we have found that the
model of the simple plane capacitor overestimate (up to 35%) the maximal capacitance of the energy
harvester.

Figure 9. Capacitance between the electrodes (C) versus forced displacement (x) (Lm=2mm,
L=30mm, g0=505m, d=100m, w=12.33mm, r=2, =10mm).
This accurate value of the capacitance for a given deflection is then applied in the mechanical
system introduced in section 3.1.
3.3. Complete analytical model
In order to get the output power of the energy harvester, mechanical and electrostatic systems have
to be coupled. From (5) and (9), one can find that the system of equations that governs the energy
harvester is (13).

d Q22
mg = m&y&
m&x& + bm .x& + kx
dx 2C(t)

dQ2 V
Q2

dt
R C(t)R

(13)

Nevertheless, it is not possible to get an analytic expression of x and Q2. Therefore, the system is
numerically solved in Matlab/Simulink (figure 10).

Figure 10. Simulink model of the energy harvester.

The deflection (x) given by Matlab and the voltage across the resistor U R = R(dQ2 dt ) versus time
are presented in figure 11(a) and 11(b) for V=1400V, d=127m, g0=1mm, =20mm, R=300M.

(a)

(b)
Figure 11. (a) Example of output voltage and (b) deflection versus time.

Figure 11(a) shows that the output voltage of cantilever-based electret energy harvesters can be
higher than 200V. This can greatly simplify rectification of the output voltage using diode bridges.
Moreover, figure 11(a) shows a particularity of the output voltage of cantilever-based electret energy
harvesters: output voltage presents a discontinuity when it passes from its higher value (the

capacitance is just before its maximum Cmax

) to its lower value (the capacitance is just after its
+
maximum Cmax
) because the current changes direction when the capacitance crosses its maximum.

The current also changes direction when the capacitance crosses its minimum Cmin . But, since the
output voltage equals 0 when the capacitance is minimum, no discontinuity on the output voltage
appears.
In section 3, we have developed the complete analytical model of the energy harvester and its
implementation on Simulink. Thanks to this model, the system can be optimized to give the maximum
output power.

4. Output power and Optimization

The goal of this section is to maximize the average output power of the energy harvester P for a
given vibration y(t). Actually, the power can be simply computed from the derivative of Q2 given by
the Simulink model presented in figure 10 (14). We will determine in section 4.1 the parameters to
optimize before optimizing them in section 4.2.

P=

t2

1
dQ
R 2 dt where t1 and t2 are times taken in the steady state

t 2 t1 t dt

(14)

4.1. Parameters to optimize

We have chosen to consider L, Lm, m, w, k, Y, =2f, V, r, d and as given parameters and the load
R, the electrode length and the initial air gap g0 as parameters to optimize. Figures 12(a), 12(b) and
12(c) present the output power of the energy harvester when one parameter varies while the other are
kept constant. Constant values are: Y=10m, f=50Hz, m=5g, =1/150, V=1400, w=12.3mm, d=127m,
r=2, Lm=2mm, L=30mm and R=1G, =10mm, g0=2mm when one of them varies.

(a)

(b)
(c)
Figure 12. Parameters to optimize and their effect on the average output power: (a) load, (b) length
of the electrode and (c) initial gap.

It is obvious that those three parameters play an important role in the behavior of the energy
harvester and it is also obvious that an optimum is present in each curve signifying that optimal
parameters exist. As these parameters are not independent, they have to be optimized together in a
single loop.
4.2. Optimization and maximum output power
The optimization process uses the fminsearch Matlab function to minimize the inverse of the output
power (1/P) considering (R, and g0) as parameters. Table 1 gives the maximum output powers and
the optimal parameters for different mobile masses for an ambient vibration of (Y, f)=(10m,
50Hz)1ms-2: it proves that the model of William and Yates gives a good approximation of the output
power but is not rigorously exact in cantilever-based electret energy harvesters. Actually, if the
system could be modeled by William and Yates model, W&Y should be equal to 1 whatever the value
of the mobile mass. Moreover, when the electrostatic force is sufficiently high, which is directly linked
to the surface voltage of the electret, output power of the energy harvester is bigger than the output
power determined by William and Yates model. When the surface voltage of the electret is not high
enough to induce a sufficient electrostatic force that can absorb the kinetic energy of the mobile mass,
it cannot permit to obtain the optimal energy with the mobile mass (e.g. when m=10g). For m=5g, a
surface voltage of 1400V should allow to harvest 160W which corresponds to a power density per
mass unit of ~30W/g.
Table 1. Output Power (P) as a function of the mass (m) with V=1400V.
m (g) Ropt (G) opt (mm) g0opt (m) P(W) PW&Y (W) P/m (W/g) W&Y
1
10
6.4
700
36.59
29.07
36.59
1.26
2
7.1
5.9
602
71.7
58.14
35.85
1.23
3
4.36
7.4
600
104
87.21
34.67
1.19
5
2.18
9.6
593
160
145.34
32
1.1
10
0.8
14.30
901
173
290.68
17.3
0.6
The results presented in table 1 are given with an accuracy of 1m for g0 and 10m for . These
precisions will not be easy to obtain. To see the effect of inaccuracies on and g0 on the response of
the system, we have plotted the output power of the system when g0 and range from 100m to their
optimal values (0 corresponds to the optimal value). Results presented in figure 13 prove that the
output power does not vary much near the optimal values of g0 and , but to avoid a contact between

the counter-electrode and the electret, for the prototype, we will choose a value of g0 slightly higher
than the optimal value. Therefore, even with inaccuracies on and g0 (~50m), the output power
should be equal to at least 140W.

Figure 13. Output power in function of and g0 variations around their optimal value.
Similarly, the effect of the frequency and the amplitude of vibrations were evaluated on the optimal
design. As the system is resonant and low damped, f and Y variations induce a large output power
change (figure 14): these parameters are critical but can be adjusted with a good accuracy. Therefore,
it proves that, when the parameters of vibrations are constant, cantilever-based electret energy
harvesters are good energy harvesters. But, if the amplitude of vibrations increases, it can lead to a
contact between the upper electrode and the electret that can damage this latter.

Figure 14. Effect of the vibrations on the optimized system.

The resonant system has been optimized and theoretical results have proven that up to 160W could
be reached with low vibrations (10m@50Hz)1ms-2. These parameters are now tested on a
prototype.

5. Experimental results and new model taking parasitic capacitances into account
The goal of experimental results presented in section 5 is to validate theoretical results that have
been obtained in the previous sections and to see the limits of our model. We conclude this section
with a better analytical model of the energy harvester that takes parasitic capacitances into account.

5.1. Prototype and expected output power

To insure a good flatness, the beam is made in silicon and attached to the frame at one end. The
mobile mass attached to the other end is made in tungsten that has a high density (d=17) to limit its
size. The electret is made in Teflon FEP (Fluorinated Ethylene Propylene) that is well known for being
a good electret [32]. In this study, the electret is charged to 1400V. Our prototype design is presented
in figure 15.

Figure 15. Cantilever-based electret energy harvester Schema.

The natural frequency of the structure, computed with FEM, is actually 48.77Hz. The difference
with our model is due to the simplification in (6). To fit the vibrations imposed by the environment
(50Hz), the beam width is slightly changed to 13mm. Therefore, the output power of the energy
harvester should be 140W and the voltage across the resistance versus time should look like the one
presented on figure 16(a) given by simulation (figure 16(b) is a zoom of figure 16(a)).

(a)

(b)
Figure 16. (a) Output voltage of the energy harvester versus time. (b) Zoom of figure 15(a) (m=5g,
V=1400V, R=2.18G, =9.6mm, g0=593m, Y=10m, f=50Hz).

The prototype was made on a glass support (figure 17(a)) to limit parasitic capacitances. A mass of
5g in tungsten was added at the free end of the cantilever. The electret is obtained by evaporating a
300nm-thick layer of aluminum on the rear face of a Teflon FEP film. It is glued on a sheet of copper
to ensure the flatness of the electret during charging. The electret is charged using a standard corona
discharge as presented in figure 3 with a point voltage Vp of 10kV and a grid voltage Vg of 1400V. It is
placed in an oven at 175C and cooled to the ambient temperature while charging to improve stability.
The long-term stability was not studied but the short-term (some days) experiments showed low
charge losses (-0.21% in 8 days) (figure 17(b)).

Figure 17. (a)Cantilever-based electret energy harvester prototype. (b) Stability of Teflon FEP
electret charged at 1400V
5.2. Output Power, comparison to the theory and limits
We present hereafter the experimental results we got on our prototype. The optimized parameters
were applied to the prototype introduced in figure 17(a). The output power of our prototype is
presented in figure 18(a). Experimental and theoretical curves do not fit and the output power is much
lower than expected. These differences are due to parasitic capacitances that become important when
using loads of high values. In those cases, the model given in figures 5(b) and 8 should be modified to
take a parasitic capacitance Cpar in parallel with the load into account.

(a)

(b)
Figure 18. Experimental output voltages (a) for R=2,2G and (b) for R=300M.

To avoid these phenomena, the value of the load was constraint to 300M (chosen after some
experimental measurements in order to limit the parasitic capacitances induced by the load) and the
optimization process was restarted on g0 and . New optimized values are presented in table 2.
Table 2. Parameters and values.
Designation
Parameter
Value
Material of the beam
Mbeam
Silicon
Youngs Modulus of Silicon
E
160 GPa
distance between the clamping and the centre of gravity of the mass
L
30 mm
Thickness of the beam
h
300 m
Width of the beam / Width of the electret
w
13 mm
Length of the mobile mass
2Lm
4 mm
Mobile mass
m
5g
Natural angular frequency/Angular frequency of vibrations
50 2 rad/s
n=
Mechanical quality factor of the structure
Qm=(2m)-1
75
Material of the electret
Melectret
FEP
Dielectric constant of the electret
2
r
Thickness of the electret
d
127 m
Surface voltage of the electret
V
1400 V
Thickness of the initial air gap
g0
700 m
Length of the electrode

22.8 mm
Load
R
300 M
The structure was tested again with the new load and its output voltage is presented in figure 18(b).
Experimental and theoretical curves fit, except for negative voltages. This is once again due to
parasitic capacitances that clip the signal in its negative part. The mean output power of the energy
harvester is 50W when it is submitted to vibrations of 10m@50Hz (1ms-2) (our simulation predicted
80W).

Table 3. Comparison to 7 prototypes among the most recent electret energy harvesters in the state of the art.
Electret
Active
Output
Figure of
Author
Ref
Vibrations
Potential
Surface (S)
Power (P)
merit ()
(V)
-2
Suzuki
[14]
1mm@37Hz (54.0ms )
2.33 cm
450V
0.28W
9.5610-5
Halvorsen
[16] 2.8m@596Hz (39.2ms-2)
0.48 cm
1W
5.0610-2
-2
Kloub
[17] 0.08m@1740Hz (9.6ms )
0.42 cm
25V
5W
14.2
9 cm
40W
3.5810-2
Naruse
[18]
25mm@2Hz (3.9ms-2)
Edamoto
[19]
500m@21Hz (8.7ms-2)
3 cm
600 V
12W
6.9710-2
-2
Miki
[20]
100m@63Hz (15.7ms )
3 cm
180V
1W
5.3710-3
-2
0.01 cm
52V
90 pW
3.3210-5
Honzumi
[21]
9.35m@500Hz (92ms )
-2
2
This work (th.)
10m@50Hz (1.0ms )
4.16cm
1400V
152W
117.84
This work (exp.)
10m@50Hz (1.0ms-2)
4.16cm2
1400V
50W
38.75
Our experimental results correspond to a factor of merit W&Y equals to 34% and to a factor of merit
equals to 38.75, putting us in the best results of the state of the art (table 3), yet, our experimental
results are quite different from theoretical results.

5.3. Model taking parasitic capacitances into account

In order to explain the differences between our theory and our experimental results, we have
developed a new model that takes parasitic capacitances into account. The parasitic capacitance of the
whole system is modeled as a capacitor Cpar in parallel with the energy harvester and the load, as
presented in figure 19. U is the voltage across the resistor, the parasitic capacitance and the electret
energy harvester.

Figure 19. Equivalent electric model of the energy harvester taking parasitic capacitance into
account.
In order to model the behavior of the energy harvester taking parasitic capacitances into account, the
equation that rules the electrostatic part is modified as follow (15) (obtained using Kirchhoffs laws)
while the equation that rules the mechanical part is the same as in (5) and (13).

V
C

dQ2
1
Q2 1 par2 dC (t )
=

dt C par R
RC (t ) C (t ) dt
1 +

C (t )

(15)

And the instantaneous harvested power is given by (16):

U2 1
Q
p (t ) =
= V 2
R
R
C (t )

(16)

Our Simulink model was modified to take these changes into account. Our experimental results
where then compared to our theoretical results taking parasitic capacitances into account (figures 20(a)
and 20(b)) where parasitic capacitances with the 300M load are estimated to 5pF and to 10pF with
the 2.2G load.

(a)
(b)
Figure 20. Experimental output voltages (a) for R=2,2G and (b) for R=300M and comparison to
theory taking parasitic capacitances into account.
Figures 20(a) and 20(b) show that theoretical and experimental results fit perfectly and validate our
new model. Therefore, it appears that parasitic capacitances have a large impact on the behavior of the
energy harvester, decreasing the harvested power, especially when using high-value resistors.
Unfortunately, as parasitic capacitances greatly depend on the load, restarting an optimization process
taking parasitic capacitances into account would be difficult. Moreover, it would have a limited
interest since parasitic capacitances can change a lot with the use of management electronic circuits.
Therefore, to limit their effects, the load should be chosen so as not to exceed Z par = 1
, the
C par
impedance of the parasitic capacitances which is roughly equal to Zpar=500M in our case.

6. Conclusion and perspectives

We developed an analytical model of cantilever-based electret energy harvester that is in
agreement with FEM results. The optimization process has shown that the power harvested by these
structures are in the same magnitude as theoretical output powers developed by William and Yates as
soon as the surface voltage of the electret is sufficient to absorb the kinetic energy of the mobile mass.
Finally, we validated our model with experimental results which reach up to 10W per gram of mobile
mass for low ambient vibrations of 0.1g (1ms-2), using a resonant system.
Cantilever-based energy harvester can be a good low-cost solution to harvest energy when
vibrations are constant in frequency and amplitude. The output power meets the magnitude of powers
reached by piezoelectric or electromagnetic solutions.
References
[1] Despesse G et al. 2005 High damping electrostatic system for vibration energy scavenging Proc.
Joint sOc-EUSAI conference pp 283-6
[2] Tada Y 1992 Experimental characteristics of electret generator using polymer film electrets Japan.
J. Appl. Phys. 31 pp 84651
[3] Genda T, Tanaka S and Esashi M 2003 High power electrostatic motor and generator using
electrets Proc. Transducers 03 pp 4925
[4] Boland J S, Messenger J D M, Lo H W and Tai Y C 2005 Arrayed liquid rotor electret power
generator systems Proc. MEMS05 pp 61821
[5] Peano F and Tambosso T 2005 Design and optimization of a MEMS electret-based capacitive
energy scavenger Journal of MEMS 14 pp 429-35
[6] Tsutsumino T, Suzuki Y, Kasagi N and Sakane Y 2006 Seismic power generator using highperformance polymer electret Proc. MEMS06 pp 98101
[7] Lo H-W, Whang R and Tai Y-C 2007 A simple micro electret power generator Technical Digest
MEMS2007 pp 85962

[8] Sterken T, Fiorini P, Altena G, Van Hoof C and Puers R 2007 Harvesting energy from vibrations
by a micromachined electret generator Int. Conf. Solid-State Sensors, Actuators and Microsystems pp
12932
[9] Sterken T, Altena G, Fiorini P and Puers R 2007 Characterization of an electrostatic vibration
harvester DTIP of MEMS and MOEMS
[10] Tvedt L G W, Blystad L C J and Halvorsen E 2008 Simulation of an electrostatic energy
harvester at large amplitude narrow and wide band vibrations Symp. on DTIP of MEMS/MOEMS pp
296-301
[11] Lo H W and Tai Y C 2008 Parylene-HT-based electret rotor generator Proc. MEMS08 pp 9847
[12] Zhang J and Lv Z 2008 A fruit jelly mems electret power generator Proc. PowerMEMS08 pp
2858
[13] Yang Z, Wang J and Zhang J 2008 A micro power generator using PECVD SiO2/Si3N4 doublelayer as electret Proc. PowerMEMS08 pp 31720
[14] Suzuki Y, Edamoto M, Kasagi N, Kashwagi K and Morizawa Y 2008 Micro electrets energy
harvesting device with analogue impedance conversion circuit Proc. PowerMEMS08 pp 710
[15] Sakane Y et al. 2008 The development of a high-performance perfluorinated polymer electret and
its application to micro power generation J. Micromech. Microeng. 18 104011
[16] Halvorsen E, Westby E R, Husa1 S, Vogl A, stb N P, Leonov V, Sterken T and Kvistery T
2009 An electrostatic Energy harvester with electret bias Proc. Transducers09 pp 138184
[17] Kloub H, Hoffmann D, Folkmer B and Manoli Y 2009 A micro capacitive vibration Energy
harvester for low power electronics Proc. PowerMEMS09 pp 1658
[18] Naruse Y, Matsubara N, Mabuchi K, Izumi M and Suzuki S 2009 Electrostatic micro power
generation from low-frequency vibration such as human motion Journal of Micromechanics and
Microengineering 19 094002
[19] Edamoto M et al. 2009 Low-resonant-frequency micro electret generator for energy harvesting
application Proc. MEMS09 pp 105962
[20] Miki D, Honzumi M, Suzuki Y and Kasagi N 2010 Large-amplitude MEMS electret generator
with nonlinear spring Proc. MEMS2010 pp 1769
[21] Honzumi M et al. 2010 Soft-X-Ray-charged vertical electrets and its application to electrostatic
transducers Proc. MEMS2010 pp 6358
[22] Yamashita K et al. 2010 Vibration-driven MEMS energy harvester with vertical electrets Proc.
PowerMEMS2010 pp 165-8
[23] Williams C B and Yates R B 1996 Analysis of a micro-electric generator for microsystems
Sensors Actuators A 52 pp 811
[24] Boisseau S, Despesse G and Sylvestre A 2010 Optimization of an electret-based energy harvester
Smart Materials and Structures 19 075015
[25] Giacometti J A, Fedosov S and Costa M M 1999 Corona charging of polymers: recent advances
on constant current charging Brazilian Journal of Physics 29 pp 269-79
[26] Ikezaki K, Miki M and Tamura J I 1981 Thermally stimulated currents from ion-injected teflonFEP film electrets Japanese Journal of Applied Physics 20 pp 1741-7
[27] Honzumi M et al. 2010 High-speed electret charging method using vacuum UV irradiation Proc.
PowerMEMS2010 pp 173-6
[28] Chen Q 2002 PTFE electret negative charge stability after RF plasma treatment J. Phys. D: Appl.
Phys. 35 2939
[29] Amjadi H 1999 Charge storage in double layers of thermally grown silicon dioxide and APCVD
silicon nitride IEEE Transactions on Dielectrics and Electrical Insulation 6 pp 852-7
[30] Leonov V, Fiorini P and Van Hoof C.U. 2006 Stabilization of positive charge in SiO2/Si3N4
electrets IEEE transactions on Dielectrics and Electrical Insulation 13 pp 1049-56
[31] Kressmann R, Sessler G and Gunther P 1996 Space-charge electrets IEEE Transactions on
Dielectrics and Electrical Insulation 3 pp 607-23
[32] Kotrappa P 2008 Long term stability of electrets used in electret ion chambers Journal of
Electrostatics 66 pp 407-9