Applied Energy 212 (2018) 362–371

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Micro electrostatic energy harvester with both broad bandwidth and high T
normalized power density
⁎
Yulong Zhanga,b, Tianyang Wanga, Anxin Luoa, Yushen Hua, Xinxin Lic, Fei Wanga,b,d,
a
Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen 518055, China
b
Shenzhen Key Laboratory of 3rd Generation Semiconductor Devices, Shenzhen 518055, China
c
State Key Lab of Transducer Technology, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China
d
Department of Mechanical and Energy Engineering, Southern University of Science and Technology, Shenzhen 518055, China

H I G H L I G H T S G RA P H I C A L AB S T R A C T

• Both broad bandwidth and high nor-

malized power density are achieved
resulting in excellent overall perfor-
mance.
• Energy harvesting from random
sources and multi-devices stack are
feasible for wireless sensing networks.
• Mechanical collision and air damping
effect on electrostatic energy har-
vesting has been investigated.
• MEMS compatible process flow has
been developed for system integration
with sensors and IC.

A R T I C L E I N F O A B S T R A C T

Keywords: In this paper, an electrostatic vibration energy harvester is proposed and fabricated with both broad bandwidth
Energy harvesting and high normalized power density (NPD, harvested power/volume/acceleration2). The device is made up of
Broad bandwidth two parts, top movable plate and bottom fixed plate, both of which are fabricated from silicon wafers using
Normalized power density advanced micro-electro-mechanical system (MEMS) technology. A CYTOP polymer layer is spray coated for an
Electrets
electret material as well as an adhesive layer for low temperature wafer bonding. The air damping effect in
Random vibration source
energy harvesting has been investigated. At a low pressure of 3 Pa in a vacuum chamber, maximum power
Wireless sensor network
output of 4.95 μW has been harvested at low vibration amplitude of 0.09 g, resulting in a bandwidth of 12 Hz and
NPD of 3 mW/cm3/g2, which outperforms most of the previous harvesters. A high harvester effectiveness of
67.9% is therefore achieved. The response on random vibrations is also tested. An average output power of
2.22 μW is harvested when random vibration is applied at a frequency range of 160 ± 12.5 Hz with RMS ac-
celeration of 10.5 m/s2. The excellent overall performance gives promising application for energy harvesting
from random sources and multi-device stack for wireless sensor networks.

1. Introduction attention from both academia and industry for their potential applica-
tions on the personal health monitoring, micro systems, and wireless
In recent years, self-sustainable power sources have attracted much sensor networks [1–3]. The energy harvesting technologies from

⁎
Corresponding author at: Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen 518055, China.
E-mail address: [email protected] (F. Wang).

https://doi.org/10.1016/j.apenergy.2017.12.053
Received 20 September 2017; Received in revised form 6 December 2017; Accepted 9 December 2017
0306-2619/ © 2017 Elsevier Ltd. All rights reserved.
Y. Zhang et al. Applied Energy 212 (2018) 362–371

ambient sources, like mechanical vibration [4,5], wasted heat [6,7] and P
triboelectricity [8–13], provide a promising way to replace the tradi-
tional chemical batteries with micro-scale devices. For instance, based
on the periodic contact/separation between polydimethylsiloxane (c)
(PDMS) film and human skin, peak power density up to 4.8 mW/m2 was
harvested, which is attractive for self-powered and portable devices
[11].
Many vibration-based energy harvesters have been designed based
on the electromagnetic [14–22], piezoelectric [23–37], and electro- (a)
static [38–49] transduction methods for vibration-to-electricity con-
version. The electromagnetic energy harvester is basically designed on
the Faraday’s law of induction. Using an array of alternating north and
south-orientation magnets, Zhang and Kim have harvested 263 mW
from vibration at low frequency, which is high enough to light an in-
(b)
candescent light bulb [15]. The bottleneck for electromagnetic method
is the bulky size since the power output is dependent on the number of
coil turns and the magnet. In piezoelectric energy harvester, some f
certain materials with piezoelectric effect, such as PZT [23–26] and Fig. 1. (a) High NPD with high-Q; (b) broad bandwidth with low-Q; and (c) all-round
PVDF [28,29], are utilized. Jung et al. developed flexible energy har- performance with both decent power output and broad bandwidth can be achieved from a
vesters with PVDF films, which could harvest 0.2 W with power density gap-closing scheme with high-Q from low packaging pressure (without increasing the
of 8 W/m2 for roadway applications [28]. One challenge for this acceleration amplitude).
method is that most of the piezoelectric materials are not compatible to
IC fabrication. favorable for high NPD. But optimal power output can only be achieved
Among the transduction types above, electrostatic energy harvester at the resonant frequency with a narrow bandwidth, as shown in
has become more attractive thanks to its good compatibility to in- Fig. 1(a). This type of device might be applicable to harvest energy from
tegrated circuit (IC) and micro-electro-mechanical system (MEMS) vibration sources with specific resonant frequency. However, broad
processes. It is mainly based on a variable capacitor structure which is bandwidth is necessary to harvest energy from the ambient vibrations
biased by either an external voltage source [50] or pre-charged electret like human motion and wind flow, which generally exhibit random
materials [51–53]. For instance, the capacitance of the device would frequency spectrum. Broad bandwidth can provide another benefit that
change when the gap distance between the two electrodes is changed we can pile up multi devices for high power, which was impractical due
due to the relative movement of proof mass according to vibration to the mismatch of the resonant frequency from the fabrication errors.
source. Based on this out-of-the-plane scheme, a cantilever based en- By reducing the Q-factor of the linear oscillator, a broad bandwidth can
ergy harvest was developed by Boisseau et al. [39]. A maximum power be achieved at the cost of the NPD, as shown in Fig. 1(b).
of 50 µW was harvested from an active chip size of 4.2 cm2. In 2014, IC A few approaches have been recently tried to optimize the fre-
compatible process flow was developed for the electret based energy quency response of the device. A broad-band energy harvester with
harvester, where four-wafer stack was batch fabricated and fully multiple beams was fabricated with a series of separate linear devices at
packaged at wafer scale [42]. Energy harvesting from in-plane vibration different resonant frequencies [62]. It can harvest energy at multiple
can also be accomplished by switching the counter electrode [38,41] or resonant frequencies; however, the efficiency of the overall device is in
varying the overlapping area [46,47]. Based on the advanced MEMS doubt [26]. Other researchers have tried to design double well potential
technology, electrostatic micro energy harvester takes the advantages with bi-stable or tri-stable resonant systems, which can be realized by
of small size, high sensitivity, high energy density and conversion ef- adding external magnets or by buckling a beam with preload [63–67].
ficiency [54]. And thanks to the compatible processes, it is feasible in These methods have been studied in depth, mainly for the piezoelectric
the future to fabricate and package the electrostatic micro power gen- energy harvesters. The bandwidth of electrostatic energy harvesters can
erator together with the wireless sensor into a self-powered system. be broadened by introducing a repulsive electrostatic force [46], gap-
There are a few general considerations for the energy harvesting spacing controlling with a mechanical stopper [40], or using nonlinear
devices. The resonant frequency and the maximum power output of the spring structures [48,68,69]. However, the NPD is generally much
device are of the highest concern, since the size and mass of the device lower than piezoelectric devices.
should be comparable to the conventional battery. A low resonant In this paper, we propose an electret based energy harvester with
frequency around 10–200 Hz is preferred for the application on vibra- both high NPD and broad bandwidth. Comparing to the conventional
tion sources from human motion, wind flow and structure motions in devices at atmosphere, the proposed device works with a gap-closing
our daily life [55,56]. However, the maximum generated power is scheme packaged at a low pressure, which reduces the air damping and
proportional to the cube of the resonant frequency, which significantly increases the Q-factor. Therefore, it is feasible to drive the mass to
limits the output of the energy harvesters [57]. Mechanical frequency- decent amplitude at a low acceleration, resulting in a high NPD as
up converters were designed to harvest energy more effectively from shown in Fig. 1(c). Thanks to the low air damping, high harvester ef-
external vibrations at low frequency [58–61]. fectiveness could also be achieved. From another aspect, the maximum
Furthermore, normalized power density (NPD) and bandwidth are displacement is restricted by the packaging plate. The squeeze-film
two important technical factors to evaluate the efficiency of the mate- damping force (Fd) increases when the gap (g) between the electrodes is
rial and the structure of a vibration based energy harvester. NPD, de- small (Fd ∝ g−3) [70], which would also affect the bandwidth of the
fined as harvested power/volume/acceleration2, is a factor of merit device. By tuning the packaging pressure and the vibration amplitude,
based on the fact that the harvested power is typically in proportion to we have successfully broadened the bandwidth of the device. The broad
the device volume and the square of the vibration amplitude [14]. bandwidth solves two major challenges for the practical application of
FWHM (full width at half maximum) bandwidth of the device is defined vibration energy harvesters. One challenge is due to the difficulty of
as the vibration frequency range where the power harvested from the harvesting from random vibration sources; the other one comes from
device is higher than half of the maximum output power. Generally, the feasibility to stack multi devices for high power output. The re-
there is a trade-off between the NPD and the FWHM bandwidth of an lationship between the output power and the air pressure has been
energy harvester. For linear oscillators, high Q-factor is typically

363
Y. Zhang et al. Applied Energy 212 (2018) 362–371

Fig. 3. The electrostatic transduction method for the energy harvesting from vibration
source. The mechanism of the device includes both the electric domain and the me-
chanical domain.

Fig. 2. 3D schematic of the electrostatic energy harvester.

are the damping coefficient and the electrostatic transduction force,
respectively. The initial conditions of displacement and velocity are
investigated in details. To evaluate the overall performance on both the both set to zero x (0) = 0 and ẋ (0) = 0 in (1). When the external ac-
NPD and the bandwidth, a figure of merit (Tech. Perf.) as the product of celeration amplitude is large enough to trigger the collision between the
NPD and FWHM bandwidth is introduced [30]. By broadening the mass and the bottom plate (x + g0 < 0), an unknown colliding force Fc
bandwidth, we have successfully piled up two devices to achieve high must be included in the equation, or the boundary conditions should be
power output which provides sufficient energy for a wireless tem- revised as [71],
perature sensor.
x = −g0, x ̇ = ηx ̇ (tc ) (2)

2. Device design where x ̇ (tc ) is the velocity of the mass at the moment of collision; η is a
coefficient for the velocity change during the collision, which should be
In this paper, a MEMS electrostatic vibration energy harvester is η = −1 if we neglect any energy loss during collision.
designed with pre-charged CYTOP electret material. The device mainly According to the Kirchhoff’s voltage law, a differential equation can
contains a variable capacitor on the basis of out-of-the-plane gap be used to analyze the electrical domain of an electret-based energy
closing scheme, as shown in Fig. 2. The capacitance can be calculated as harvester [72],
C = εrε0A/g, where εr represents the relative permittivity of the di-
dQ1 1 1 ⎤
electric material, ε0 is the permittivity of vacuum, A is the overlap area R = V −Q1 ⎡ +
dt ⎢ C
⎣ 1 ( t ) C2⎥
⎦ (3)
of the electrode and g is the two electrodes’ air gap. The top plate
consists of a proof mass suspending to outer frame through four beams where Q1 is the charge induced on the top electrode (the surface of the
(spring structure), forming a spring-mass resonant system. The bottom proof mass); R represents the external load resistance (R = R0 + R1); V
plate consists of a cavity with certain depth that is used for controlling stands for the surface potential of the pre-charged electret; C1(t) is the
the initial gap of the variable capacitor. Both of the two parallel plates capacitance of the air gap and C2 is the capacitance of the electret di-
are fabricated on silicon substrate with SiO2 insulation layer on the electric material.
surfaces. Metal layers are deposited on each plate as the electrodes of To solve the equations, we have used a Matlab/Simulink model with
the capacitor. CYTOP polymer is coated on the bottom electrode as an the parameters from Table 1. A similar model was proposed with more
electret material, and it can also function as the adhesive layer in low details for a piezoelectric energy harvester with bi-resonant structure
temperature bonding. Bump stoppers are designed on the proof mass to [71]. Fig. 4 shows the calculation results for Device 2 under sinusoidal
avoid the “pull-in” effect of electrostatic force when the proof mass vibration sources (f = 136 Hz) with amplitudes of 2 m/s2 and 10 m/s2,
moves close to the bottom electrode at high vibration acceleration. respectively. Under low vibration amplitude, the velocity of the mass
When driven by an ambient vibration source, the capacitance of the follows a continuous function (very close to a sinusoidal form). When
device changes as the gap distance between the two electrodes varies. the vibration amplitude increases up to 10 m/s2, collision between the
Hence, a current between the two electrodes is generated from the proof mass and the bottom plate occurs, which can be clearly seen from
movement of induced charges by the pre-charged electret. Tuning the the velocity change of the mass. The output voltage and power are also
overlapping area between two electrodes by the in-plane vibration of
the proof mass can also achieve the same goal for another scheme of Table 1
electrostatic energy harvester [38,41]. Parameters of the electret-based energy harvester.
The principle of the energy conversion involves both the electrical
Parameters Device 1 Device 2
and mechanical domains as shown in Fig. 3. When there is no collision
(x + g0 > 0), we can describe the mechanical performance of device as ft Resonant frequency 160.9 Hz 135 Hz
a mass-spring-damper system l Beam length 7 mm 8 mm
w Beam width 200 μm 200 μm
m (¨y + x¨) + cx ̇ + kx + mg−Fele = 0, (x + g0 > 0) (1) h Beam thickness 60 μm 60 μm
H Silicon wafer thickness 400 μm 400 μm
where x and y are the relative displacement of the proof mass and the g Initial gap 360 μm 360 μm
d Electret thickness 20 μm 20 μm
displacement of the external ambient vibration source, respectively; g0
m Proof mass 0.07 g 0.07 g
represents the initial air gap; m is the equivalent mass; g is the grav- V Surface potential of electret −400 V −400 V
itational acceleration; k is the stiffness of the spring structure; c and Fele

364
Y. Zhang et al. Applied Energy 212 (2018) 362–371

Fig. 4. The calculated displacement (x), and velocity

(v) of the mass under a vibration amplitude of (a)
2 m/s2; (b) 10 m/s2. The output voltage (V) and
power (P) are also shown in the plots. With large vi-
bration amplitude, collision occurs which changes the
velocity of the mass as shown in the dash circle.

140

120 0.5 g
1.0 g
RMS Output Power ( μW)

100 1.5 g
2.0 g
80 2.5 g

60

40

20

0
100 110 120 130 140 150 160 170
Frequency of vibration source (Hz)
Fig. 5. Simulated relationship between the output power and the frequency of the vi-
bration under different amplitude (Device 2, with an estimated Q-factor of 30).

shown in the plots with a load resistance of 16 MΩ and electret surface

potential of −400 V.
In this model, we have also studied the relationship between the
output power of the electrostatic harvester and the frequency of the
vibration source under different amplitudes, as shown in Fig. 5 for
Device 2. A Q-factor of 30 is assumed for normal air pressure of 1 at-
mosphere. When vibration amplitude lower than 10 m/s2 is applied
with frequency increasing from 100 Hz to 170 Hz, a maximum output
power can be generated at 134 Hz, which is exactly the resonant fre-
quency of the device. With higher vibration amplitude, the bottom plate
will limit the displacement of the proof mass; therefore, the frequency
spectrum of the device can be tuned. At an amplitude of 2.5 g (24.5 m/
s2), the bandwidth of the harvester can be broadened to 21 Hz, which is
about 16% of the resonant frequency.

3. Fabrication

As mentioned in Fig. 2, the electrostatic energy harvester consists of

two parallel plates, namely, top plate and bottom plate. Both of the two
plates are fabricated in 4-inch silicon wafer using the standard MEMS
techniques. Fig. 6 shows the fabrication process for each plate which
will be described in details below.

3.1. Top plate

Fig. 6. Fabrication process flow for (T1–T4) top plate and (B1–B3) bottom plate of the
energy harvester. The two components are fabricated individually and then bonded to-
The top plate with a spring-mass resonant system includes a proof gether using the adhesive CYTOP polymer.
mass suspended in the center of frame by four fixed beams. It is the
movable component in the variable capacitor structure of the device.
bottom plate. The fabrication of top plate starts with a (100) silicon
On the surface of the mass, we have designed a few bump stoppers in
wafer of 400 μm-thick.
case of the electrostatic collapse of “pull-in” effect, when the vibration
T1: as shown in Fig. 6, thermal oxidation process is applied to grow
acceleration is high enough to make the proof mass collide with the

365
Y. Zhang et al. Applied Energy 212 (2018) 362–371

-540

-520

Surface Potential (Volt)

-500

-480

-460

-440

-420

-400
0 5 10 15 20 25 30 35 40 45 50
Time (hour)
Fig. 8. The decay test at 25 °C and 50%RH for the surface potential in the CYTOP layer
Fig. 7. SEM image for the bump stoppers on the top plate.
with negative charge.

SiO2 layers of 2 μm on both sides of the silicon wafer. Afterwards,

photolithography is applied on the top of the wafer, and the bottom
SiO2 layer is protected by photoresist. Then SiO2 layer on the top side is
etched by buffered oxide etchant (BOE).
T2: with the SiO2 as masking layer, wet etching of the silicon wafer
is performed in a 40% KOH solution at 50 °C to form a deep cavity with
a depth of 280 μm.
T3: photolithography is used on the bottom side to pattern the SiO2
layer, and then reactive ion etching (RIE) process is applied to pattern
the SiO2 layer with photoresist as masking layer.
T4: the silicon wafer is wet etched again in the 40% KOH solution at
50 °C to release the bump structures and the beam structures. The final
beam thickness and bump height will be about 60 μm. Fig. 7 shows that
the width at the top of the bump is about 110 μm. Finally, sputtering is
used to deposit a metal multilayer of Cr/Al (15 nm/100 nm) on the
surface of the mass.

3.2. Bottom plate

The bottom plate of the energy harvester contains a cavity that is

used for controlling the maximum displacement of the proof mass and
the initial gap of the variable capacitor. The bottom plate is also fab-
ricated with a (100) silicon wafer of 400 μm-thick, as described below.
B1: similar to the top plate, thermal oxidation is used to grow SiO2
layers with thickness of 2 μm on both sides of the wafer. Then litho-
graphy and wet etching with BOE are used to pattern the top side SiO2.
B2: with the SiO2 as masking layer, wet etching is performed in the
40% KOH solution at 50 °C to form a cavity (12.5 × 12.5 mm2) with an
etch depth of 300 μm. The etch depth here is used for controlling the
initial gap of the variable capacitor.
Fig. 9. (a) The final device after bonding. (b) Overall 16 energy harvesters can be si-
B3: finally, a metal multilayer of Cr/Al (15 nm/100 nm) for elec-
multaneously fabricated in a single 4-in. silicon wafer.
trode is deposited on the cavity by sputtering, and a polymer layer of
CYTOP is spray coated on the electrode surface, which acts as both an
electret material and an adhesive layer for low-temperature wafer 4. Characterization and discussion
bonding [73].
A shaker setup which mimics the external vibration source is used to
characterize the performance of the micro energy harvester. The device
3.3. Corona charging and final bonding
is fixed on a mechanical shaker with an accelerometer to monitor the
acceleration during the measurement. The shaker is driven by an ex-
CYTOP polymer coated on the bottom plate is charged in a custom-
citation signal generated from a signal generator (Brüel&Kjær, LAN-XI
built corona charging setup. The setup consists of a grounded wafer
3160) and a power amplifier (Brüel&Kjær, 2719). The test circuit dia-
stage, a metal mesh grid (Vg = −1500 V) and a high-voltage probe tip
gram is shown in Fig. 3. The voltage across the fixed resistance R0 is
(VH = −6 kV). After several hours for charge stabilization at 25 °C and
measured and recorded using PULSE FFT system (Brüel&Kjær, M1-
50%RH, the electret surface potential stabilized to about −400 V
7770), and the output power of the device can be calculated with the
(Fig. 8). As shown in Fig. 6, the two plates are finally bonded at 130 °C
series circuit [74]. Here, the energy harvester was tested at low fre-
for 30 min to avoid severe loss of surface charge in the electret. Fig. 9(a)
quency (less than 200 Hz). The output power of the energy harvester is
shows the final packaged device. The size of the final device is about 1.3
dependent on the external load resistance. Firstly, the optimal load
(L) × 1.8 (W) × 0.08 (H) cm3. Fig. 9(b) shows that in a 4-inch wafer,
R0 + R1 of the energy harvesters (Device 2) was tested by tuning the
overall 16 energy harvesting devices can be fabricated simultaneously
variable resistance R1, when the external sinusoidal vibration source
with the advance MEMS technology.

366
Y. Zhang et al. Applied Energy 212 (2018) 362–371

Fig. 10. The RMS power output when the external load increases from 1 MΩ to 26 MΩ at
Fig. 13. Power output of the device at a chamber pressure of 3 Pa. The device can gen-
the RMS acceleration of 3.8 m/s2 where the external sinusoidal vibration frequency is
erate 1.8 μW at a very low vibration of 0.06 g. The output power saturates at 4.9 μW,
135 Hz.
while the bandwidth can be broadened to 12 Hz at vibration of 0.09 g. Comparing to the
measurement under 1 atm (shown in Fig. 11), much lower excitation level is needed when
5 the device works at lower pressure.
0.63 g
4.5
0.94 g
4 1.25 g
1.56 g
3.5 1.72 g
3 2.19 g
2.50 g
2.5

1.5

0.5

0
100 110 120 130 140 150 160 170
Frequency (Hz)
Fig. 11. Power output of the device under air pressure of 1 atm. At low vibration am-
plitude of 0.63 g, a typical frequency spectrum is achieved with a bandwidth of about
10 Hz, which is increased to about 33 Hz at vibration amplitude of 2.5 g.

Fig. 14. The displacement of the proof mass (Device 2) at 3 Pa with driven vibration of
(a) 0.09 g; (b) 0.12 g. Unstable movement in (b) indicates the collision between the proof
mass and the bottom plate.

Fig. 12. The measurement setup for the air damping effect.

was applied at 135 Hz with root mean square (RMS) amplitude of

3.8 m/s2. Fig. 10 shows the relationship between the external load re-
sistance and the RMS output power. With an optimal load of about
16 MΩ, maximum RMS power of 0.2 μW for the Device 2 was harvested
from the vibration of low amplitude. In the following sections, the ex-
ternal load resistance was set to 16 MΩ in all of the measurements Fig. 15. Comparison of energy harvesters (Device 2) regarding to the normalized power
unless otherwise noted. density and the bandwidth.

367
Y. Zhang et al. Applied Energy 212 (2018) 362–371

Table 2
Detailed comparison of the vibration based energy harvesters.

f0 (Hz) Acceleration (m/s2) Output Power (μW) Mass (gram) Volume (cm3) Bandwidth Δf (Hz) NPD (mW/cm3/g2) Tech. Perf. (Δf * NPD)

EM [18] 8.5 5 2090.16 – 40.18 3 0.208 0.624

EM [19] 60 19 0.4 0.209 2.7 40 0.00004 0.0016
EM [20] 102 9.8 290 1.5 8.6 5 0.031 0.155
EM [21] 26 9.8 210 – 68 12 0.0031 0.0372
EM [22] 55 14.9 0.61 0.031 0.13 5 0.002 0.01
PE [30] 263 1 0.15 0.026 0.0121 4.2 1.215 5.103
PE [30] 80.1 1 0.78 0.217 0.0121 2.5 6.309 15.7725
PE [31] 572 20 60 – 0.1 2 0.15 0.3
PE [32] 235 9.8 13.98 – 0.05 8 0.28 2.24
PE [33] 461.2 19.6 2.15 0.001 0.00065 1.2 0.82 0.984
PE [34] 16 10 22 4.6 1.67 12 0.01317 0.15804
PE [35] 36 10 0.09 0.023 0.016 17 0.00534 0.09078
PE [36] 385/387/398 0.5 52.9 11.7 6.5 16 3.2 51.2
PE [37] 155 9.8 95 – 0.027 13 3.5 45.5
ES [41] 179 0.3 0.03 0.06 0.15 5.5 0.2407 1.32385
ES [42] 96 9.8 0.15 0.07 0.29 20 0.00052 0.0104
ES [43] 110 20 20.7 0.267 0.484 22 0.0107 0.2354
ES [44] 150 10 2.2 0.066 0.042 40 0.052 2.08
ES [45] 109 2.5 1.3 0.066 0.038 6 0.547 3.282
ES [46] 63 20 1 0.1 0.305 27 0.00082 0.02214
ES [47] 28 5 1.5 0.5 0.305 12 0.0197 0.2364
ES [48] 95 4.2 0.95 0.02 0.14 3.7 0.0374 0.13838
ES [49] 139 9.8 54 0.088 0.8 13 0.0675 0.8775
This work 136 0.9 4.95 0.104 0.187 12 2.9952 35.9424
This work 136 3.1 4.94 0.104 0.187 13 0.2695 3.5035
This work 136 6.1 5.56 0.104 0.187 41 0.0758 3.1078

EM – Electromagnetic method; PE – Piezoelectric method; ES – Electrostatic method.

4.1. Test results at atmosphere

We have characterized the output performance at atmosphere.

Fig. 11 plots the RMS power outputs versus the frequencies of the
shaker at atmosphere. A typical spectrum similar to Fig. 1(b) is ob-
tained with power output of 0.7 μW at an acceleration amplitude of
0.63 g. This device demonstrates a bandwidth of about 10 Hz and an
overall estimated Q-factor of ∼13. The output power saturates to
4.3 μW when the vibration increases to 1.56 g, above which damping
force and the collision between the mass and the bottom plate will limit
the maximum displacement of the proof mass. The theoretical max-
imum power output can be estimated as,

P = mωn3 YZmax /2 (4)

Fig. 16. The output power of the energy harvester (Device 2) at resonance with chamber where Y and Zmax are the amplitude of the vibration source and the
pressure increased from 4 Pa to atmosphere. maximum displacement of the proof mass, respectively [2,54,75].
Based on this calculation, the power output effectiveness of our device
at 1.56 g is about 3.5%.
A broader bandwidth up to 35 Hz can be observed when higher
acceleration is applied. It should be noted that the maximum power
output of the device is about 27 times lower than the simulation shown
in Fig. 5, which is mainly due to the parasitic capacitance of the device
and the testing system [72].

4.2. Air damping effect on energy harvesting

From Eq. (1), it can be seen that the squeeze air damping force (cẋ)
would affect the mechanical performance of the proof mass and
therefore influence the energy harvesting. It is known that when the air
pressure is near the atmospheric pressure, the coefficient of air damping
is nearly constant. However, when the air is pumped down to a pressure
level well below, the air damping could be significantly reduced, either
Fig. 17. The root mean square (RMS) output power of the Device 1 when vibration source
according to the Energy Transfer Model or the Christian’s Model [70].
with RMS amplitude of 10.5 m/s2 (∼1 g) is applied at random frequency through a
bandpass filter (f = 160 ± 12.5 Hz). Thus, it is important to study how packaging pressure affects the per-
formance of the energy harvesting device. The measurement setup is
shown in Fig. 12. The shaker is placed in a chamber where the pressure
can be controlled with a pump and monitored with an air pressure

368
Y. Zhang et al. Applied Energy 212 (2018) 362–371

The inset plot shows the close-up view of the output voltage at
136 Hz. The device bandwidth is broadened to 12 Hz at vibration of
0.09 g. When the driven amplitude increases to 0.12 g, unstable vibra-
tion of the mass is observed (Fig. 14(b)), which might due to the col-
lision between the proof mass and the bottom plate. With decent power
output at low vibration amplitude, we have achieved a high normalized
power density (NPD) of 3 mW/cm3/g2, which provides an excellent all-
round performance as shown in Fig. 15. A more detailed comparison
among different types of vibration based energy harvesters is listed in
Table 2.
The air damping effect can be more clearly seen in Fig. 16. With a
fixed vibration amplitude of 1.2 m/s2, the RMS power of the device at
resonance decreases gradually when the chamber pressure increases
from 4 Pa to the atmosphere. The changes of the output power with the
packaging pressure of the MEMS process have been investigated here,
Fig. 18. Power output of a two-devices stack at chamber pressure of 600 Pa. When in-
creasing the driven acceleration, the two separated resonant peaks merge together and
which could be useful for all the other type of vibration based energy
therefore, the overall power is improved. The top-left inset shows the output voltage after harvester.
rectifying.

4.3. Energy harvesting from random vibration source

Random vibration sources can be generated by a noise signal and

specific filter in the signal generator for the shaker. During this test, we
have applied a random source through a bandpass filter with a fre-
quency range of 160 ± 12.5 Hz. The energy harvester is tested under
the random vibration for 300 cycles with 6.4 s per cycle at the average
acceleration of 10.5 m/s2 (∼1g). As shown in Fig. 17, an average power
of 2.22 μW can be harvested by the energy harvester, which demon-
strates very promising application for random sources in our daily life.

4.4. Two device stacks

The broad bandwidth of the energy harvester makes it feasible to

pile up multi-devices for higher power output, which was actually im-
practical due to the mismatch of the resonant frequencies of each de-
Fig. 19. A wireless temperature sensor network powered by electrostatic vibration energy vice. In Fig. 18, we have measured the output of two devices, whose
harvesters (e-VEHs). The system consists of power management circuit with rectifier, outputs are rectified and connected in series. At low vibration of 0.44 g,
storage capacitors, micro controller unit (MCU) with temperature sensor and radio two separate resonant peaks are observed which merge together when
transmitter. The wireless signal can be detected and processed by a receiver and a the vibration amplitude increases to 1.4 g. An overall power of 13.5 μW
Raspberry Pi. is generated and the benefit from the stack of multi-devices can be
clearly seen.
sensor. The displacement of the proof mass is measured by a laser de-
tector during the vibration. By tuning the air pressure, we can char-
5. Application for self-powered wireless temperature sensing
acterize the device performance under various damping conditions.
Fig. 13 shows the RMS power output versus the vibration frequency
We have applied the above stack of two energy harvesters for a self-
at a chamber pressure of 3 Pa. At 10 times lower acceleration (0.06 g)
powered wireless temperature sensor network. Fig. 19 shows the
than the test at atmosphere, more than twice power (1.8 μW) is har-
wireless temperature sensor network with all the components, and a
vested which gives an overall Q-factor of ∼70. With the same estimate
live demo of the wireless sensing application can be more clearly seen
from Eq. (4), a high effectiveness of 67.9% has been achieved.
from the supplementary video

Fig. 20. (a) With the energy harvested from vibration

source, a wireless temperature sensor node can be
self-powered and the temperature signal can be de-
livered to the internet, which is shown on the web-
page of our group. (b) The temperature signal can be
accessed anywhere by scanning a barcode with a
smart phone or a laptop.

369
Y. Zhang et al. Applied Energy 212 (2018) 362–371

2016A030306042) and Guangdong Special Support Program (Project

No.: 2015TQ01X555).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the

online version, at http://dx.doi.org/10.1016/j.apenergy.2017.12.053.

References

[1] Sudevalayam S, Kulkarni P. Energy harvesting sensor nodes: survey and implica-
tions. IEEE Commun Surv Tut 2011;13(3):443–61.
[2] Sue CY, Tsai NC. Human powered MEMS-based energy harvest devices. Appl Energy
May 2012;93:390–403.
[3] http://www.microgensystems.com/.
[4] Zhang Z, et al. Design, modelling and practical tests on a high-voltage kinetic energy
Video 1. . As schemed in Fig. S1, the wireless sensor node consists of harvesting (EH) system for a renewable road tunnel based on linear alternators.
power management circuit, temperature sensor and radio transmitter, Appl Energy 2016;164:152–61.
[5] Roshani H, Dessouky S, Montoya A, Papag AT. Energy harvesting from asphalt
which are fully powered by the electrostatic vibration energy harvesters pavement roadways vehicle-induced stresses: a feasibility study. Appl Energy
(e-VEHs). Two capacitors (100 μF for each) are charged by the har- 2016;182:210–8.
vesters for energy storage and power management. Once the voltage of [6] Hyland M, Hunter H, Liu J, Veety E, Vashaee D. Wearable thermoelectric generators
for human body heat harvesting. Appl Energy 2016;182:518–24.
the capacitor reaches 4.32 V, a trigger signal is sent to a micro con- [7] Bell LE. Cooling, heating, generating power, and recovering waste heat with ther-
troller unit (MCU), which will acquire and send out a temperature moelectric systems. Science 2008;321(5895):1457–61.
signal by the transmitter. This wireless signal is detected and processed [8] Wang S, Mu X, Yang Y, Sun C, Gu AY, Wang ZL. Flow-driven triboelectric generator
for directly powering a wireless sensor node. Adv Mater 2015;27:240–8.
by a radio receiver and a Raspberry Pi, which delivers the sensing signal
[9] Dhakar L, Tay F, Lee C. Development of a broadband triboelectric energy harvester
to internet terminals. Fig. S2 shows the function algorithm for the with SU-8 micropillars. J Microelectromech Syst 2015;24(1):91.
event-driven temperature sensing system. [10] Dhakar L, Tay F, Lee C. Investigation of contact electrification based broadband
energy harvesting mechanism using elastic PDMS microstructures. J Micromech
A vibration source with amplitude of 1.32 g has been applied to
Microeng 2014;24(10):104002.
mimic the vibrations from environment (for instance, the roadways). [11] Rasel M, Park J. A sandpaper assisted micro-structured polydimethylsiloxane fab-
During our experiment, the storage capacitor releases energy every rication for human skin based triboelectric energy harvesting application. Appl
696 s to send a wireless temperature signal (Fig. S3). As shown in Energy 2017;206:150–8.
[12] Shi Q, Wang H, Wu H, Lee C. Self-powered triboelectric nanogenerator buoy ball for
Fig. 20, the real-time temperature information is posted on the webpage applications ranging from environment monitoring to water wave energy farm.
of our group, which could be accessed anywhere by scanning a barcode Nano Energy 2017;40:203–13.
with a smart phone. [13] Gupta R, Shi Q, Dhakar L, Wang T, Heng C, Lee C. Broadband energy harvester
using non-linear polymer spring and electromagnetic/triboelectric hybrid me-
chanism. Sci Rep 2017;7:41396.
6. Conclusion [14] Beeby SP, et al. A micro electromagnetic generator for vibration energy harvesting.
J Micromech Microeng Jul. 2007;17(7):1257–65.
[15] Zhang Q, Kim ES. Vibration energy harvesting based on magnet and coil arrays for
Using advanced MEMS technology, an electrostatic vibration energy watt-level handheld power source. Proc IEEE 2014;102(11):1747–62.
harvester has been designed and fabricated with both high NPD and [16] Constantinou P, Mellor PH, Wilcox PD. A magnetically sprung generator for energy
harvesting applications. IEEE/ASME Trans Mechatron 2012;17(3):415–24.
broad bandwidth. Based on the out-of-the-plane gap closing scheme,
[17] Liu H, Koh K, Lee C. Ultra-wide frequency broadening mechanism for micro-scale
the energy harvester contains a spring-mass structure which converts electromagnetic energy harvester. Appl Phys Lett 2012;100(22):223905.
vibration from our environment into electric energy. A few bump- [18] Foisal ARMd, Hong C, Chunga GS. Multi-frequency electromagnetic energy har-
stoppers are designed to prevent the electrostatic collapse. This device vester using a magnetic spring cantilever. Sens Actuat, A 2012;182(May):106–13.
[19] Yang B, Lee C. Non-resonant electromagnetic wideband energy harvesting me-
is fabricated with a compact sized of 1.3 × 1.8 × 0.08 cm3. The air chanism for low frequency vibrations. Microsyst Technol 2010;16(March):961–6.
damping effect for the vibration energy harvesting has been in- [20] Sardini E, Serpelloni M. An efficient electromagnetic power harvesting device for
vestigated at different package pressure. At a low air pressure of 3 Pa, low-frequency applications. Sens Actuat, A 2011;172(September):475–82.
[21] Galchev T, McCullagh J, Peterson RL, Najafi K. A vibration harvesting system for
our device can harvest a power output of 4.95 μW when a low vibration bridge health monitoring applications. In: 10th International workshop on micro
acceleration amplitude of 0.09 g is applied, which gives a high NPD of 3 and nanotechnology for power generation and energy conversion applications
mW/cm3/g2. Furthermore, the bandwidth of the device has also been (PowerMEMS 2010), November 2010. p. 179–82.
[22] Wang P, Tanaka K, Sugiyama S, Dai X, Zhao X, Liu J. A micro electromagnetic low
broadened, thanks to the amplitude limitation scheme from the air level vibration energy harvester based on MEMS technology. Microsyst Technol
damping and the mechanical collision. The broad bandwidth of the April 2009;15:941–51.
device can solve the two major challenges for the practical application [23] Tang Q, Li X. Two-stage wideband energy harvester driven by multimode coupled
vibration. IEEE/ASME Trans Mechatron 2015;20(February):115–21.
of vibration energy harvester. Firstly, it makes it possible to harvest [24] Liu H, Zhang S, Kobayashi T, Chen T, Lee C. Flow sensing and energy harvesting
energy from random vibrations in our daily life. Secondly, the broad characteristics of a wind-driven piezoelectric Pb(Zr0.52, Ti0.48)O3 microcantilever.
bandwidth makes it feasible to pile up multi-devices for high power Micro Nano Lett 2014;9(4):286–9.
[25] Liu H, Zhang S, Kathiresan R, Kobayashi T, Lee C. Development of piezoelectric
output. With the excellent all-round performance and compatible fab-
microcantilever flow sensor with wind-driven energy harvesting capability. Appl
rication process, the devices have been used for a self-powered wireless Phys Lett 2014;104(5):053901.
temperature sensor node, which demonstrates promising application [26] Gu L, Livermore C. Passive self-tuning energy harvester for extracting energy from
for WSN technology in the future. rotational motion. Appl Phys Lett 2010;97(8):081904.
[27] Xiong H, Wang L. Piezoelectric energy harvester for public roadway: on-site in-
stallation and evaluation. Appl Energy 2016;174:101–7.
Acknowledgement [28] Jung I, Shin Y, Kim S, Choi J, Kang C. Flexible piezoelectric polymer-based energy
harvesting system for roadway applications. Appl Energy 2017;197:222–9.
[29] Li S, Peng Z, Zhang A, Wang F. Dual resonant structure for energy harvesting from
This work was supported by National Natural Science Foundation of random vibration sources at low frequency. AIP Adv 2016;6(January):015019.
China (Project No.: 51505209) and Shenzhen Science and Technology [30] Aktakka EE, Peterson RL, Najafi K. A CMOS-compatible piezoelectric vibration
energy scavenger based on the integration of bulk PZT films on silicon. In: 2010
Innovation Committee (Projects Nos.: JCYJ20170412154426330 and IEEE international electron devices meeting (IEDM), December 2010.
KQTD2015071710313656). Fei Wang is also supported by Guangdong [31] Elfrink R, Kamel TM, Goedbloed M, Matova S, Hohlfeld D, van Schaijk R, et al.
Natural Science Funds (Project Nos.: 2015A030313812 and Vibration energy harvesting with aluminum nitride-based piezoelectric devices. J

370
Y. Zhang et al. Applied Energy 212 (2018) 362–371

Micromech Microeng 2009;19(August):094005. [52] Renaud M, Altena G, Goedbloed M, de Nooijer C, Matova S, Naito Y, et al. A high
[32] Lei A, Xu R, Thyssen A, Stoot AC, Christiansen TL, Hansen K, et al. MEMS-based performance electrostatic MEMS vibration energy harvester with corrugated in-
thick film PZT vibrational energy harvester. In: 24th International conference on organic SiO2-Si3N4 electret. In: 17th International conference on solid-state sensors
micro electro mechanical systems, March 2011. p. 125–8. actuators and microsystems, Barcelona, June 2013. p. 693–6.
[33] Shen D, Park JH, Ajitsaria J, Choe SY, Wikle HC, Kim DJ. The design, fabrication [53] Asanuma H, Oguchi H, Hara M, Yoshida R, Kuwano H. Ferroelectric dipole electrets
and evaluation of a MEMS PZT cantilever with an integrated Si proof mass for for output power enhancement in electrostatic vibration energy harvesters. Appl
vibration energy harvesting. J Micromech Microeng April 2008;18:055017. Phys Lett October 2013;103:162901.
[34] Tang QC, Yang YL, Li XX. Bi-stable frequency up-conversion piezoelectric energy [54] Suzuki Y. Recent progress in MEMS electret generator for energy harvesting. IEEJ
harvester driven by non-contact magnetic repulsion. Smart Mater Struct November Trans Electr Electron Eng 2011;6:101–11.
2011;20:125011. [55] Orrego S, et al. Harvesting ambient wind energy with an inverted piezoelectric flag.
[35] Liu H, Tay CJ, Quan C, Kobayashi T, Lee C. Piezoelectric MEMS energy harvester for Appl Energy 2017;194:212–22.
low-frequency vibrations with wideband operation range and steadily increased [56] Yurchenko D, Lai ZH, Thomson G, Val DV, Bobryk RV. Parametric study of a novel
output power. J Microelectromech Syst 2011;20(October):1131–42. vibro-impact energy harvesting system with dielectric elastomer. Appl Energy
[36] Aktakka EE, Najafi K. Three-axis piezoelectric vibration energy harvester. In: 28th 2017;208:456–70.
IEEE international conference on micro electro mechanical systems (MEMS 2015); [57] Beeby SP, Tudor MJ, White NM. Energy harvesting vibration sources for micro-
2015. p. 1141–4. systems applications. Meas Sci Technol 2006;17:175–95.
[37] Aktakka EE, Najafi K. A micro inertial energy harvesting platform with self-supplied [58] Külah H, Najafi K. Energy scavenging from low-frequency vibrations by using fre-
power management circuit for autonomous wireless sensor nodes. IEEE J Solid- quency up-conversion for wireless sensor applications. IEEE Sens J 2008;8:261–8.
State Circuits 2014;49(September):2017–29. [59] Liu H, Lee C, Kobayashi T, Tay C, Quan C. Piezoelectric MEMS-based wideband
[38] Lo H, Tai YC. Parylene-based electret power generators. J Micromech Microeng energy harvesting systems using a frequency-up-conversion cantilever stopper.
2008;18:104006. Sensors Actuat A 2012;186:242–8.
[39] Boisseau S, Despesse G, Ricart T, Defay E, Sylvestre A. Cantilever-based electrets [60] Liu H, Lee C, Kobayashi T, Tay C, Quan C. Investigation of a MEMS piezoelectric
energy harvesters. Smart Mater Struct 2011;20:105013. energy harvester system with a frequency-widened-bandwidth mechanism in-
[40] Takahashi T, Suzuki M, Nishida T, Yoshikawa Y, Aoyagi S. Vertical capacitive en- troduced by mechanical stoppers. Smart Mater Struct 2012;21:035005.
ergy harvester positively using contact between proof mass and electret plate - [61] Liu H, Lee C, Kobayashi T, Tay C, Quan C. A new S-shaped MEMS PZT cantilever for
stiffness matching by spring support of plate and stiction prevention by stopper energy harvesting from low frequency vibrations below 30 Hz. Microsyst Technol
mechanism. In 28th IEEE international conference on micro electro mechanical 2012;18:497–506.
systems (MEMS 2015); 2015. p. 1145–8. [62] Lin SC, Lee BS, Wu WJ, Lee CK. Multi-cantilever piezoelectric MEMS generator in
[41] Crovetto A, Wang F, Hansen O. An electret-based energy harvesting device with a energy harvesting. In: 2009 IEEE international ultrasonics symposium (IUS). IEEE;
wafer-level fabrication process. J Micromech Microeng 2013;23(October):114010. 2009. p. 755–8.
[42] Wang F, Hansen O. Electrostatic energy harvesting device with out-of-the-plane gap [63] Harne RL, Wang KW. A review of the recent research on vibration energy harvesting
closing scheme. Sens Actuat, A 2014;211(March):131–7. via bistable systems. Smart Mater Struct 2013;22(2):023001.
[43] Chiu Y, Lee YC. Flat and robust out-of-plane vibrational electret energy harvester. J [64] Vocca H, Neri I, Travasso F, Gammaitoni L. Kinetic energy harvesting with bistable
Micromech Microeng 2013;24(December):015012. oscillators. Appl Energy 2012;97:771–6.
[44] Guilllemet R, Basset P, Galayko D, Cottone F, Marty F, Bourouina T. Wideband [65] Younesian D, Alam M-R. Multi-stable mechanisms for high-efficiency and broad-
MEMS electrostatic vibration energy harvesters based on gap-closing interdigited band ocean wave energy harvesting. Appl Energy 2017;197:292–302.
combs with a trapezoidal cross section. In: 26th International conference on micro [66] Erturk A, Inman D. Broadband piezoelectric power generation on high-energy orbits
electro mechanical systems (MEMS), March 2013. p. 817–20. of the bistable Duffing oscillator with electromechanical coupling. J Sound Vib
[45] Cottone F, Basset P, Guilllemet R, Galayko D, Marty F, Bourouina T. Non-linear 2011;330(May):2339–53.
MEMS electrostatic kinetic energy harvester with a tunable multistable potential for [67] Zhou S, Cao J, Inman DJ, Lin J, Liu S, Wang Z. Broadband tristable energy har-
stochastic vibrations. In: 17th International conference on solid-state sensors, ac- vester: Modeling and experiment verification. Appl Energy 2014;133:33–9.
tuators and microsystems (transducers & eurosensors XXVII), October 2013. p. [68] Miki D, Honzumi M, Suzuki Y, Kasagi N. Large-amplitude MEMS electret generator
1336–9. with nonlinear spring. IEEE 23rd international conference on micro electro me-
[46] Suzuki Y, Miki D, Edamoto M, Honzumi M. A MEMS electret generator with elec- chanical systems, Hong Kong, January 2010. p. 176–9.
trostatic levitation for vibration-driven energy-harvesting applications. J [69] Nguyen SD, Halvorsen E. Nonlinear springs for bandwidth-tolerant vibration energy
Micromech Microeng 2010;20(10):104002. harvesting. J Microelectromech Syst 2011;20(6):1225–7.
[47] Minakawa Y, Chen R, Suzuki Y. X-shaped-spring enhanced MEMS electret generator [70] Bao MH. Analysis and design principles of MEMS devices. Amsterdam: Elsevier;
for energy harvesting. In: 17th International conference on solid-state sensors, ac- 2005 [Chapter 3].
tuators and microsystems (transducers & eurosensors XXVII), October 2013. p. [71] Li S, et al. Bi-resonant structure with piezoelectric PVDF films for energy harvesting
2241–4. from random vibration sources at low frequency. Sens Actuat, A 2016;247:547–54.
[48] Tao K, Lye SW, Miao J, Tang L, Hu X. Out-of-plane electret-based MEMS energy [72] Crovetto A, Wang F, Hansen O. Modeling and optimization of an electrostatic en-
harvester with the combined nonlinear effect from electrostatic force and a me- ergy harvesting device. IEEE/ASME J Microelectromech Syst
chanical elastic stopper. J Micromech Microeng 2015;25(September):104014. 2014;23(October):1141–55.
[49] Asanuma H, Hara M, Oguchi H, Kuwano H. Nonlinear restoring force of spring with [73] Xu Y, et al. Spray coating of polymer electret with nano particles for electrostatic
stopper for ferroelectric dipole electret-based electrostatic vibration energy har- energy harvesting. Micro Nano Lett 2016;11:640–4.
vesters. AIP Adv 2016;6(July):075206. [74] Zhang Y, et al. Electrostatic energy harvesting device with dual resonant structure
[50] Paracha M, Basset P, Galayko D, Marty F, Bourouina T. A silicon MEMS DC/DC for wideband random vibration sources at low frequency. Rev Sci Instrum
converter for autonomous vibration-to-electrical-energy scavenger. IEEE Electron 2016;87:125001.
Device Lett 2009;30(5):481–3. [75] Stephen NG. On energy harvesting from ambient vibration. J Sound Vib
[51] Wang F, Hansen O. Invisible surface charge pattern on inorganic electrets. IEEE 2006;293:409–25.
Electron Device Lett 2013;34(July):1047–9.

371