Smart Materials and Structures

PAPER Related content

- Finite element modeling of nonlinear
Design, fabrication, and testing of a low frequency piezoelectric energy harvesters with
magnetic interaction
MEMS piezoelectromagnetic energy harvester Deepesh Upadrashta and Yaowen Yang

- Design optimization of vibration energy

harvesters fabricated by lamination of
To cite this article: Egon Fernandes et al 2018 Smart Mater. Struct. 27 035017 thinned bulk-PZT on polymeric substrates
Andrés Vásquez Quintero, Nadine Besse,
Pattanaphong Janphuang et al.

- Modeling and design of Galfenol unimorph

energy harvesters
View the article online for updates and enhancements. Zhangxian Deng and Marcelo J Dapino

This content was downloaded from IP address 129.97.185.225 on 27/02/2018 at 19:33

 Smart Materials and Structures

Smart Mater. Struct. 27 (2018) 035017 (15pp) https://doi.org/10.1088/1361-665X/aaaba5

Design, fabrication, and testing of a low

frequency MEMS piezoelectromagnetic
energy harvester
Egon Fernandes1, Blake Martin1, Isabel Rua2, Sid Zarabi3, Hélène Debéda2,
David Nairn3, Lan Wei3 and Armaghan Salehian4,5
1
Energy Harvesting and Vibrations Lab, Mechanical and Mechatronics Engineering, University of
Waterloo, Waterloo, ON, N2L 3G1, Canada
2
Université de Bordeaux, Laboratoire IMS, UMR 5218, Bordeaux INP, 351 Cours de la Libération, BAT
A31, F-33405 Talence Cedex, France
3
Electrical and Computer Engineering, University of Waterloo, Waterloo, ON, N2L 3G1, Canada
4
Director of the Energy Harvesting and Vibrations Lab, Mechanical and Mechatronics Engineering,
University of Waterloo, Waterloo, ON, N2L 3G1, Canada

E-mail: [email protected], [email protected], [email protected],

[email protected], [email protected], [email protected], [email protected] and
[email protected]

Received 26 September 2017, revised 12 January 2018

Accepted for publication 30 January 2018
Published 20 February 2018

Abstract
This paper details a power solution for smart grid applications to replace batteries by harvesting
the electromagnetic energy from a current-carrying wire. A MEMS piezoelectromagnetic energy
harvester has been fabricated using PZT screen-printing technology with a centrally-supported
meandering geometry. The energy harvesting device employs a symmetric geometry to increase
its power output by reducing the effects of the torsional modes and the resultant overall strain
nodes in the system subsequently reduce the complexities for the electrode fabrication. The unit
is modelled using COMSOL to determine mode shapes and frequency response functions. A
12.7 mm by 14.7 mm unit is fabricated by screen-printing 75 μm-thick PZT on a stainless steel
substrate and then experimentally tested to validate the FEA results. Experimentally, the
harvester is shown to produce 9 μW from a wire carrying 7 A while operating at a distance of
6.5 mm from the wire. The design of the current work results in a greater normalized power
density than other MEMS based piezoelectromagnetic devices and shows great potential relative
to larger devices that use bulk or thin film piezoelectrics.

Keywords: smart materials, energy harvesting, piezoelectrics, smart grid

(Some figures may appear in colour only in the online journal)

1. Introduction wireless communications technology. Various sensors (cur-

rent, temperature, etc) are also integrated into the smart grid to
As the demand for electrical power increases, so does the allow for monitoring of critical components. A vast number
strain on the aging power grid infrastructure. Continued use of sensors, distributed across the entire grid, are required to
of electric components nearing failure has contributed to adequately monitor the system. Recent advancements in low-
several major blackouts over the past decades. Electric uti- power electronics have created opportunities for wireless
lities are modernizing their systems into a ‘smart grid’ that sensor nodes that combine sensors, power conditioning cir-
can detect critical changes through the use of sensors and cuitry, and radios all into one small package and allow for
remote sensing. These sensor nodes are typically powered by
5
Author to whom any correspondence should be addressed. batteries that only have a lifespan of a few years. The cost to

0964-1726/18/035017+15$33.00 1 © 2018 IOP Publishing Ltd Printed in the UK

Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

replace these batteries can be immense as it involves both the MEMS-scale cantilever was fabricated using aluminum
battery and labor costs. Moreover, the installation of certain nitride as the piezoelectric material [14]. Frequently,
current sensors requires a power line to be disconnected and researchers increase the effective length of the system by
taken out of service causing disruptions to the given area. using two-dimensional geometries. For instance, one way in
Considering these issues, there is a need to develop a non- which the length can be increased is by using spiral geome-
invasive, low-cost, and self-powered solution for the sensor tries [15, 16]. However, previous research has shown that
nodes. spiral geometries are more susceptible to have a fundamental
Energy harvesting solutions that employ ambient or other vibration mode that is torsional and, therefore, not ideal for
sources of energy available to the system have proven as a vibrations energy harvesting [17]. Due to the resulting
viable option for powering electronics or wireless sensor orthogonality of the electric field to the material polarization
nodes. Extensive research has been done in low-frequency and placement of the electrodes, torsional strains are not
electromagnetic vibration energy harvesting [1]. Previous readily harvestable by the flat rectangular profile of planar 2D
work with piezoelectric cantilever beams has shown that both beam shapes and are, therefore, undesirable [18]. Previously,
current sensing [2, 3] and energy harvesting [4, 5] are pos- a quad-folded spiral cantilever beam electromagnetic har-
sible directly from AC conducting wires using non-invasive vester was modelled using a finite element method (FEM),
technology. The electromechanical vibration harvesting using optimized to fit a 10×10 mm2 area, and shown to theoreti-
bulk piezoelectric lead zirconate titanate (PZT) has been cally produce 2 μW from 1 A of current in a wire [19]. This
shown to produce better output than other comparable coil- design, however, is susceptible to torsion due to the spiral-like
based approaches where the harvester cannot encircle the design that as a result produces several strain nodes. The
conductor [6]. A Halbach array can also be used to focus the presence of strain nodes in the system require a more complex
magnetic field to one side of the magnets and increase the geometry for the placement of the electrodes as they need to
electromagnetic force between the magnet array and the wire be segmented to avoid a charge cancellation in order to
[7]. The piezoelectric solution can also be combined with increase the harvested power efficiency.
other electromagnetic scavenging techniques to develop Recently, a zigzag geometry was proposed for energy
hybrid energy harvesters [8]. These AC energy harvesting harvesters as a means of reducing the natural frequency and
solutions need to be scaled down to the micro-scale before maintaining a high power density when compared to a can-
becoming ubiquitous in the smart grid. Various micro-elec- tilever beam of similar dimensions [20, 21]. As with the spiral
tromechanical system (MEMS) solutions allow for mass geometry, this design also experiences some torsion at its
production that result in much smaller overall cost per unit. lower frequencies mainly due to the asymmetry of the design
Screen-printing, for example, is a low cost additive process that results in lower efficiency. Interestingly, a zigzag design
that allows for the creation of complex structures. Although a with a torsion-dominant fundamental mode was seen to pro-
significant amount of literature deals with the screen-printing duce more power than its bending-dominant second mode
of PZT to form simple cantilever beams [9–12], little research [22]. Symmetry has been applied to the zigzag design in
has been performed regarding the same fabrication process for [23, 24] in an effort to reduce the effect of torsion and sub-
two-dimensional cantilevers (zigzag for instance) to create sequently produce a higher power output due to bending-
low frequency energy harvesters. dominant fundamental modes. The ‘elephant’ design mirrors
Migration to MEMS energy harvesting solutions also has a three-beam zigzag geometry along the length of the clamped
some major hurdles. Piezoelectric cantilevers have been beam and forms a closed connection at its free end [23]. This
researched extensively for mechanical vibrations harvesting produced greater bending in the clamped beam where the
due to their simplicity [13]. For microsystems, however, highest amount of the strain is experienced. Their 25.4 mm by
cantilever beams tend to have a very high fundamental 25.4 mm harvester uses a piezoelectric layer on only the
resonant frequency, which is not ideal for ambient energy clamped beam to examine the reduced torsional effects and
harvesting where the available frequencies are low. This enhance pure bending for the fundamental mode. The ele-
becomes problematic for energy harvesting from power lines phant harvester was further able to combine the first two
as the harvester must be tuned to match the 60 Hz frequency modes through mass loading structural modification and
of the current flow in the North American power grid. shown to produce more power than any of the considered
Common methods to lower the fundamental frequency of mode shapes [25]. The ‘meandering’ design mirrors the zig-
a cantilever-based harvesting device include reducing the zag geometry along the length of the free end [24]. Similarly,
beam thickness, increasing the tip mass, or increasing the their 29 mm by 23 mm meandering design is shown to reduce
effective length of the beam. Most of these techniques torsion and experiences smaller shear strain at the anchor
become less practical when moving towards the micro-scale points. While the elephant and meandering designs show
because of certain limitations regarding either quality or size. improvement in low frequency energy harvesting through
For example, while MEMS fabrications methods allow for base excitation, they use bulk PZT and are in the meso scale,
piezoelectric thicknesses of about one micrometer, thin films and may be considered too large to be incorporated with
typically have much weaker electromechanical properties microelectronics on a single-chip and too costly to mass
than commercial bulk piezoelectrics and may not be suitable produce. Reducing the overall footprint even further will
for energy harvesting applications. Also, the tip mass has potentially increase the resonant frequency but is required to
limits due to size constraints of the packaging. Recently a allow for the realization of highly-scalable fabrication.

2
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Figure 1. Schematic of a piezoelectromagnetic energy harvester on an AC-carrying wire.

The goal of the current work is to design a piezo- 2. Piezoelectromagnetic energy harvesting working
electromagnetic energy harvester that is fabricated using a principle
low-cost screen-printing MEMS fabrication technique and
alleviates some of the issues stated in the previous work in the Figure 1 illustrates the design and working principle of the
literature. Screen-printed PZT thick films bridge the piezoelectromagnetic energy harvester. The harvester is
1–100 μm gap between thin-film and bulk piezoelectrics, and composed of a piezoelectric layer sandwiched between two
allow for the creation of complex geometries like zigzags. electrodes on a metallic substrate with a permanent magnet
PZT thick-films also make good candidates for MEMS energy attached to the tip. The harvester is mounted above an AC-
harvesting applications since their coupling is higher than carrying wire using a clamp where it is fixed at one end. The
magnet couples to the alternating magnetic field produced by
PZT thin-films. The proposed 12.7 mm by 14.7 mm centrally-
the wire. A sinusoidal electromagnetic force produced by the
supported meandering design has a low susceptibility to twist
AC current around the wire excites the cantilever and causes
and experiences fewer strain nodes that remove complexities
the piezoelectric material to experience cyclic elastic defor-
for the electrode fabrication. The harvester also employs PZT mation. Through the piezoelectric effect an AC voltage is
on every beam segment to increase the overall utilization of generated across the electrodes that produces an electric
the footprint for harvesting. The unit is designed, fabricated, current through the resistance load. The power produced can
and tested for harvesting the electromagnetic energy from an be calculated as the square of the root mean square (RMS)
AC current-carrying wire in North American transmission voltage across the load divided by its resistance. If the fre-
lines and, therefore, a resonant frequency of 60 Hz has been a quency of the electromagnetic force matches the resonant
major design criterion. A small footprint for the harvester was frequency of the harvester, the amplitude of the output power
also desired to help with further integration of the unit with will be maximized as it results in a resonant condition.
the sensor and wireless components in a single chip. Exper- The current-carrying wire emits a magnetic field
imental displacement and voltage frequency response func- according to the Biot–Savart law. The net force Fm on the
tions and measured power outputs as a function of current and magnet with flux density Bm can be described according to the
distance from a wire are also presented and compared to the following formula [26]
FEM results. Finally, the normalized power density of the d (Hx )
current work is compared to various energy harvesters in the Fm = - Bm ∮v dy
dV (1 )
literature to highlight the advantage of the proposed design.
where dV is the differential volume of the magnet and Hx is
the x-component of the wire’s magnetic field. Due to the
unit’s symmetric geometry and the symmetric positioning of
the unit and the magnet above the wire, it is reasonable to
assume that the motion will be primarily in bending and in the

3
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Table 1. 301 Stainless steel properties.

Young’s modulus 193 GPa

Density 7880 kg m−3
Poisson’s ratio 0.24
Thickness 0.254 mm

Table 2. Neodymium magnet (NdFeB, N42) properties.

Young’s modulus 160 GPa

Density 7500 kg m−3
Poisson’s ratio 0.24
Dimensions 3.175 mm×3.175 mm×12.7 mm
Weight 0.96 g
Figure 2. Dimensioned (mm) schematic of energy harvester Magnetic remanence 1.32 T
substrate.

vertical direction, minimizing the possibility of a twist about Table 3. Screen-printed PZT parameters.
the z-axis, or rotation about the y axis. Additionally, the
Young’s modulus 43 GPa
electromagnetic force is purely translational and does not Density 5500 kg m−3
cause any rotation of the magnet about the x-axis. The magnet Piezoelectric coefficient d31 −40 pC N−1
is also physically constrained to the substrate such that there Thickness 0.075 mm
is no rotation about the y-axis. Simply put, assuming small
geometries, the force on the magnet is proportional to the
current in the wire and inversely proportional to the square of required during the fabrication process. Both the tip mass and
the distance between the magnet and wire. the PZT are attached on the same side of the substrate.
Finally, the geometry of the harvester also plays a significant
role in achieving a low resonant frequency. A preliminary
3. Design, modelling, and simulation implementation of this device was first presented by the
authors in [27].
The energy harvester has an overall footprint of
3.1. Design
14.7 mm×12.7 mm or 187 mm2, which excludes the extra
The objective of the proposed geometry is to maximize the space near the clamp that is used for fine-tuning the resonant
power density while maintaining a fundamental frequency in frequency. A 0.254 mm thick stainless steel substrate supports
the range of 60 Hz to match the fundamental frequency of the 75 μm thick strips of PZT on each beam. A fixed-end
North American power grid. A centrally-supported mean- boundary condition is imposed by clamping one end of the
dering (CSM) geometry is used in the current work and the central beam. The remainder of the structure is unrestricted
proposed design is shown in figure 2. It is shown that this and a neodymium magnetic proof mass is attached to the
geometry achieves less torsion and bending becomes the opposite end. The proposed geometry delivers nearly pure
dominant mode for the fundamental frequency. This results in bending to the clamped beam where the majority of the stress
higher energy harvesting efficiency. Furthermore, the geo- is located. The proposed geometry also minimizes the area of
metry also minimizes the area of strain nodes that will be strain nodes, which subsequently results in a greater har-
discussed in detail in the later section. vestable area.
A permanent magnet is attached as a tip mass in the
design to harvest energy from the electromagnetic field of the
3.2. Modelling and simulation
current-carrying wire and also to help tune the natural fre-
quency of the system to the desired frequency. A neodymium The energy harvester is modelled using the 3D FEA software
magnet has been selected due to its strong magnetic flux COMSOL Multiphysics and its associated MEMS module.
density and availability in a variety of sizes. With respect to The structural properties of the stainless steel and the neo-
the piezoelectric material, PZT is chosen as it is known to dymium magnet have been obtained from manufacturers
have high coupling coefficients and allows for various [28, 29] and are presented in tables 1 and 2, respectively.
available small-scale fabrication processes. For the proposed Finally, the material properties of the screen-printed PZT
design, a screen-printed PZT technology is used as this pro- were determined experimentally [30] and are presented in
cess allows for fabrication of films in a 1–100 μm thickness table 3.
range and also allows for fabrication of complicated geome- The COMSOL software contains several default levels of
tries. The substrate of the system is 301-grade stainless steel. meshing. Simulations were run with increasing levels of
This substrate was selected for its relatively low Young’s meshing until the incremental change in the resulting eigen-
modulus and compatibility with the high firing temperatures frequencies was below 1%. It was determined that the ‘Fine’

4
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Table 4. FEA meshing information. In order to obtain the frequency response functions
(FRFs) for the displacement and voltage results, two separate
Number of vertex elements 258
COMSOL simulations are performed in the current work.
Number of edge elements 7891
Number of boundary elements 62 501
First, the base excitation in the form of acceleration is pro-
Number of elements 146 059 vided to the clamped end of the harvester to obtain the dis-
placement FRFs for output displacement divided by the base
acceleration, similar to what is done in the experiments using
shaker excitations for validation. The second simulation
element mesh size was optimal as it provided the combination pertains to obtaining the voltage FRFs when the unit is
of solution convergence below the threshold and a reasonable operating in the real condition and mounted on the wire. For
simulation time. The meshing information for the final this purpose, an electromagnetic modelling is performed
simulations is presented in table 4. using a 2D model containing only cross-sections of the
An eigenfrequency analysis is performed using the Solid magnet and the wire to determine the electromagnetic force
Mechanics physics to obtain the natural frequency and mode on the tip magnet due to the magnetic flux around the wire.
shapes for the system. The eigenfrequency studies produce The Magnetic Fields physics in COMSOL is used for this
mode shapes that are independent of applied loads and are purpose. A 12.7 mm by 3.175 mm NdFeB magnet and an
used to design the harvester to have a fundamental resonant AWG10 wire with its conductor and insulation are modelled
frequency of 60 Hz. The corresponding mode shape is shown within a magnetically insulated space. A stationary study is
in figure 3. performed to determine the electromagnetic force. The
The displacement mode shape shown indicates no electromagnetic force from the 2D magnetic model is then
noticeable torsion for the frequency of interest. As a result, multiplied by the depth of the magnet to obtain the total
each leg experiences pure compression or tension along its volumetric force on the magnet due to the current carrying
entire length during cyclic motion as shown in the strain plot wire. The calculated force is then applied to the bottom face
in figure 5(a). This clearly demonstrates that one of the major of the magnet for the energy harvester model to obtain the
design goals for reducing torsion in the fundamental mode of voltage FRFs.
the system was achieved. The piezoelectric polarization plot
is shown in figure 4(b) and areas of warm colors (red) indicate
tension and positive charge whereas cold colors (blue) indi- 4. Geometry assessment
cate compression and a negative charge production.
Additionally, in figure 4(a), it can be seen that there are Before proceeding with fabrication, a study was performed to
no strain nodes. This is quite important as the presence of compare the power density of the centrally supported mean-
strain nodes will add complexities to the electrodes’ config- dering geometry of the current work with the quad-folded
urations, as they need to be segmented in order to avoid (QF) spiral geometry proposed in [19]. The QF design is
charge cancellation. Also, there is an alternating strain pattern chosen over the meandering or elephant designs for com-
seen across adjacent beams of the system. The PZT outputs a parisons to the proposed harvester due to the similarities for
voltage according to its strain profile and, therefore, a con- the intended application, i.e., harvesting from a current car-
tinuous electrode across the top and bottom of the entire rying wire. This allows for a fair comparison and serves to
structure cannot be used before polarization. Accordingly, the clearly highlight the advantages of the proposed design to
PZT has been placed along the length of each beam with better harvest from an AC current carrying wire. Specifically,
disconnects at each beam link to avoid the strain nodes. the purpose of the study was to confirm the energy harvesting

Figure 3. Fundamental mode shape showing total displacement (mm) (a) Top view. (b) Isometric view.

5
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Figure 4. Fundamental mode shape (a) Strain. (b) Piezoelectric polarization.

advantages of the proposed geometry while minimizing the

impact of other contributing factors such as piezoelectric
material, footprint, tip mass, and magnetic properties.
As shown in figure 5, the QF cantilever design studied in
[19] utilizes four beams that are folded around in a spiral
shape, and attach to a central mass at the end. In this design,
the entire length of each beam contains a piezoelectric layer.
The geometry contains four fixed constraints, one in each end
corner, and the central tip mass acts like a guided end due to a
symmetric geometry. A COMSOL stress analysis for this
structure shows that each beam segment experiences both
tension (red) and compression (blue), and indicates that large
areas exist where very little strain (grey) is experienced. These
areas pertain to the strain node sections where power har-
vesting has minimal efficiency. The combination of both
tension and compression in a single beam segment requires
complicated electrode patterning for energy harvesting to
ensure minimal voltage cancellation. These large strain nodes
Figure 5. Quad-folded (QF) cantilever strain plot.
are left without electrodes. Moreoever, due to the spiral shape,
the structure experiences torsional stresses, which are not
cancellation. This is a major advantage for the screen-printing
readily harvestable.
fabrication.
The CSM geometry proposed in this study is a fixed-free
The QF geometry is compared with variations of the
folded design similar to a cantilever, shown in figure 6. The
CSM geometry shown in figure 6 to determine the best
central beam is clamped at one end and extends into a zigzag
geometry for the piezoelectromagnetic energy harvester.
pattern that wraps around to form a closed loop which sup- These designs are modelled using COMSOL for their power
ports a tip mass. Similar to a cantilever beam, this structure output comparison. PZT-5A and a silicon substrate are chosen
experiences the greatest amount of strain near its clamped for this analysis for all designs. The same dimensions and
end. The symmetry of this design reduces torsion in the magnetic properties are used for the tip mass in all cases.
beams, which results in greater bending effects and a more Also, the designs shown in figure 7 have similar footprints to
efficient energy harvester. Each beam supports a layer of that of the QF design to ease the comparisons. Eigen-
piezoelectric material, and contains a small disconnect frequency and frequency response analyses are performed to
between each beam segment to allow for individual polar- determine the resonant frequency and power output per
ization or electrodes. Important to note is that each beam footprint using the optimal load for each case. A wire current
segment experiences either tension or compression. This of 1 A and a distance of 4 mm are used and the results of the
permits a single continuous electrode for each beam, allowing simulations are presented in table 5.
for full utilization of the piezoelectric material and removes The results from the study show that both CSM designs
the need for segmented electrodes to avoid a charge produced larger power and had at least three times more

6
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Figure 6. Centrally supported meandering geometry with 5 beams (left) and 9 beams (right). The unit is fixed at the left side of the
center beam.

Figure 7. Piezoelectromagnetic energy harvester fabricated using screen-printing technology (a) Top view. (b) Cross-sectional schematic of
one beam.

Table 5. Comparison of quad-folded and two centrally-supported meandering designs.

Quad-folded 9 Beams 5 Beams

Eigenfrequency (Hz) 56.53 56.39 56.61
Optimal load, Rl (Ω) 10 000 50 000 10 000
Frequency @ Rl (Hz) 56.53 56.58 57.82
Power @ Rl (μW) 3.1 11.2 11.8
Footprint (mm2) 10×10=100 11.88×10=119 10×10.5=105
Power/footprint (uW mm−2) 0.031 0.094 0.112
Force on magnet (N) 0.000 26 0.000 26 0.000 26
Number of fixed constraints 4 1 1
PZT thickness (μm) 1 181 62
Area of PZT, AP (mm2) 39 72 64
Area of electrodes, AE (mm2) 21 72 64
AE/AP 54% 100% 100%
Power/active area 0.15 0.16 0.18

power per footprint area. The optimal loads of the designs thickness of the PZT is found to be larger for these designs to
were found to be in the 10 kΩ range. For fair comparison, all tune to the same fundamental frequency. This subsequently
units have been designed to have the same fundamental fre- allows for a higher energy harvesting efficiency. Furthermore,
quencies. Since the CSM designs have fewer fixed constraints the CSM designs also offer a greater harvestable area (area of
and therefore are more flexible than the QF cantilever, the the electrodes, AE, divided by the area of PZT, AP) with less

7
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

strain node areas, a greater active area for PZT, and even a
greater power output per active area. For the reasons men-
tioned, the CSM was further optimized as described in the
previous section and chosen for this application. Finally, the
proposed unit and its symmetricity around the fixed ends
allows for a design that is less susceptible to the torsional
modes.

5. Fabrication

The materials used in the fabrication process consist of a Figure 8. Fabricated energy harvester with magnet.
stainless steel substrate, thin gold electrodes, and PZT. The
stainless steel substrate was laser cut (KJ Laser Micro- 6. Experimental results
machining) according to the designed geometry. Stainless
steel was selected as a substrate for its compatibility with the Experimental results are obtained to validate the COMSOL
high firing temperature required by the screen-printing pro- model. It should be noted that the COMSOL model assumes
cess and for its flexibility when compared with the alumina an excitation force with constant magnitude whereas experi-
substrates normally used with thick films. Commercially mentally the force on the magnet varies as the harvester
available gold ink (ESL8836 from ElectroScience Labora- undergoes vibrations. Therefore, experiments are used to
assess the validity of this assumption for the range of the
tories) is used for the electrode layers, whereas the PZT is
current values used for this work.
prepared at the IMS laboratory [31, 32].
A DEK HORIZON 03iX machine is used for the printing
6.1. Shaker test—no tip mass
process, and a specially built structure holds the substrate
during all levels of printing. The bottom gold electrode, PZT To validate the COMSOL model the fabricated unit was
layer, and top gold electrode are successively printed and experimentally tested. Initially, a shaker was used as the
dried at 120 °C using 325 μm, 200 μm, and 325 μm mesh excitation source for the energy harvester to validate the
screens, respectively. The dried samples then undergo hot dynamic behavior of the system. As a preliminary step, the
isostatic pressing to reduce the porosity of the PZT, thus energy harvester is tested prior to affixing the magnet. This
improving the material’s electromechanical properties. minimizes the introduction of errors from the process of
Afterwards, the samples are co-fired for two hours at 900 °C epoxying the tip magnet and provides a simpler case for
in air atmosphere and an optical profiler found the PZT comparison to the COMSOL results.
thickness to be approximately 75 μm. Figure 7 shows the Figure 9 shows the test setup used to obtain the dis-
fabricated unit after firing and a schematic of the unit cross- placement FRFs of the energy harvester. A Modal Shop
section. 2075E dual-purpose electrodynamic shaker is used to excite
The next step of the fabrication is the polarization of the the energy harvester using a base excitation. An LMS SCA-
DAS mobile data acquisition system is used to control the
PZT layers to exhibit piezoelectric properties. A strain-mat-
shaker, and to perform a frequency sweep analysis using the
ched polarization scheme is used to mitigate voltage cancel-
Sine Control module. The sinusoidal input signal is amplified
lation across the harvester [24]. The PZT of each beam is
through a Modal Shop 2050E09 power amplifier. An accel-
poled such that each adjacent member has the opposite poling erometer is mounted on the shaker to provide controlled
direction to match the strain plots in figure 4(a). Beams of the feedback during the frequency sweep and maintain the
same strain were first micro-wired together using a ball- desired acceleration. During excitation, a Polytec OFV-505
bonding technique. Then both positive and negative groups laser vibrometer and OFV-5000 controller are used to mea-
were polarized by applying 0.250 kV for 10 min at 280 °C sure the tip displacement. An additional distance of 3 mm
under nitrogen according to the sign of figure 4(a). By from the clamp is used in the experimental tests to reduce the
employing this approach, all of the electrodes have the same stress experienced in the PZT near the clamp and minimize
voltage polarity under strain. After poling, micro-wire con- chances for damage to the harvester.
nections are made between the top electrodes of all the beams. A frequency sweep from 30 Hz to 500 Hz with an input
This allows for a single continuous electrode to be used acceleration of 0.2 g was performed to capture the first three
across the bottom layer of the energy harvester. Finally, the modes, and the displacement FRFs of the experimental and
magnet is glued to the tip of the substrate using a thin layer of simulation results are shown in figure 10. A sweep rate of
epoxy and left to cure for two days. The completed energy 0.1 Hz sec−1 was used to obtain the FRFs and was slowed
harvester is shown in figure 8. down near resonance peaks. The slow sweep rate and the low
acceleration level used for this validation reduce the presence
of nonlinearities in the response, and allow for a fair

8
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Figure 9. (a) Experimental test setup equipment. (b) Close up of EH.

Figure 11. Experimental and model displacement FRFs of the energy

harvester with a tip mass.
Figure 10. Experimental and model displacement FRFs of the energy
harvester without a tip mass.

comparison with the discrete steady-state responses from the

COMSOL model. The displacement measurement is per-
formed at the center of the tip of the harvester as also shown appear in the displacement FRF of figure 10 because the
as a laser target in figure 9(b). sensing point is chosen exactly in the center of the tip mass
Experimentally, the first two bending modes occur at area. Due to symmetry the torsional mode is not seen at this
164.75 Hz and 443 Hz, and the first torsional mode occurs at centerline but the frequency is taken from the eigenfrequency
403.25 Hz. COMSOL predicts the first two bending modes at analysis. Also in figure 10 it is seen that there is a good
163.61 Hz and 439.38 Hz, errors of less than 1%. COMSOL agreement in the amplitudes for the displacement FRFs when
predicts the first torsional mode at 401.24 Hz, an error of comparing experimental and simulation results that further
0.5%. The torsional mode for the COMSOL model does not validates the model as a viable tool for future optimization.

9
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Figure 12. Experimental test setup of the piezoelectromagnetic

energy harvester on a current-conducting wire.

Figure 15. Power output with respect to load resistance (Wire

current=1 A at a distance of 12.5 mm).

Figure 13. Tip displacement and voltage FRFs showing a resonant

frequency at 60.3 Hz (wire current=1 A at a distance of 12.5 mm).

Figure 16. Power output as a function of electric current and distance

from wire for the optimal load.

frequency provides the greatest power output from the energy

harvester, and considering the good match in the results
without a tip mass, the frequency sweep is performed for the
25 Hz–80 Hz that embraces the first peak. An input accel-
eration of 0.2 g was used. The experimental and simulation
FRFs are presented in figure 11.
In figure 11 the experimental fundamental frequency is
measured to be about 47.6 Hz and the COMSOL result is
48.1 Hz, an error of 1%. The experimental fundamental fre-
Figure 14. Power curves across 10 kΩ to 1 MΩ resistive loads (wire quency of 47.6 Hz being lower than the design requirement
current=1 A at a distance of 12.5 mm). 60 Hz is due to the additional 3 mm distance used from the
clamp to minimize potential damage to the unit. This was
fixed later when the unit was tested on the wire to attain the
6.2. Shaker test—tip mass included
desired frequency of 60 Hz. Furthermore, figure 11 indicates
Next, the magnetic tip mass is epoxied to the harvester and that the amplitudes of the displacement FRFs match very
the displacement FRF is obtained. Since the fundamental well. This indicates that the COMSOL model accurately

10
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

Figure 17. Voltage frequency responses for increasing wire currents demonstrating nonlinear effects (d=6.5 mm) (a) Voltage amplitude. (b)
Voltage FRF.

predicts the dynamic behavior of the proposed energy har- minimize the presence of nonlinear effects. The harvester was
vester design. mounted above the wire at a distance of 12.5 mm between the
bottom of the magnet and the top surface of the wire carrying
a 1 A current using a sine sweep signal over the range of
6.3. Wire test
frequencies shown in figure 13. The experimental and simu-
The final experimental tests were performed on a unit using a lated tip displacement and open-circuit voltage FRFs for a
wire excitation using the setup shown in figure 12. The unit range of 55 Hz to 65 Hz are shown in figure 13.
was mounted above a 10 AWG current-carrying wire using an The experimental and simulated results show a strong
acrylic base and 1.5 mm thick spacers to control the distance agreement in both the amplitude and resonant frequency. This
between the magnet and the wire. The vertical distance indicates that the COMSOL model accurately predicts the
between the magnet and wire was measured using a plastic dynamic behavior of the proposed energy harvester design.
caliper. An LMS SCADAS mobile data acquisition system is The experimental resonant frequency was found to be 60.3 Hz
used to control the current amplitude and frequency in the that is suitable for harvesting from the North American power
wire and to perform a frequency sweep analysis using the grid. It should be noted that this frequency was tuned by
Sine Control module. A Fluke i400s AC clamp is used to adjusting the distance from the clamp. The COMSOL model
provide the feedback control for the wire’s current amplitude. predicted a resonant frequency of 60.2 Hz, which shows a
It should be noted that for model validation purposes the negligible error. The damping ratio is calculated by using the
setup is used to measure the voltage and displacement FRFs quadrature peak picking method [33]. The damping ratio zeta
for the unit for an input electric current through the wire of a was found to be approximately 0.0027 from the voltage FRF
constant magnitude that is swept across a range of frequencies test results and applied to the COMSOL model.
and controlled using the sine sweep module. Therefore, the
sine sweep electric current signal does not have a 60 Hz
6.4. Load sweep
dominant frequency, and the peaks shown further in the FRF
plots indicate the natural frequency of the unit after the pre- Next, the energy harvester is tested with different resistive
viously used 3 mm gap to avoid damages during the vibration loads to determine the optimal load resulting in the maximum
testing was removed. A similar arrangement is generally power transfer. An Elenco resistor substitution box is used to
adopted for modal analysis and testing where the system is conveniently provide the numerous resistances. The power is
provided with a constant mechanical input in the form of calculated by using the square of the measured RMS voltage
displacement or acceleration across a range of frequencies across a resistor and dividing by its resistance value
instead of an excitation of a given frequency. The output (Pave = VRMS 2 Rload). The power outputs across 10 kΩ to 1
voltage produced by the system is measured by the LMS MΩ are shown in figure 14.
system with an input impedance of 1 MΩ using a 10:1 probe. No noticeable shift in resonant frequency is seen across
Any resistive load is adjusted to account for the 10 MΩ input the load sweeps. The difference in power output between the
resistance. The frequency resolution of the data acquisition is 500 kΩ and the 1 MΩ loads appears small, with the 1 MΩ
0.01 Hz. resistor resulting in the highest power output. This shows that
The harvester was initially tested with a relatively low the resistor values resulting in the greatest power transfer are
input force (both low current and high distance from wire) to in this proximity. The simulation and test results are shown to

11
 Smart Mater. Struct. 27 (2018) 035017
Table 6. Normalized power density for previous piezoelectromagnetic harvesters and the present work.

Deposition Magnet dis- Magnetic rema- Wire cur- Effective volume Normalized power density
Reference method tance (mm) nence (T) rent (A) Power output (μW) Frequency (Hz) (mm3) μW/(Ta Aa mm)
The present Screen-prin- 6.5 1.32 7 9 60 169.7 0.243
design ted PZT
Paprotny [19] Sputtered AlN 2.5a 1.32 1 2 (simulated) 57a 111.81 0.085
Olszewski [14] Sputtered AlN 2a 1.3 2 1.5 42.3 20.69 0.112
12

Paprotny [6] Bulk PZT 2.5 1.48 20 2700 60 1731.06 0.329

Chen [8] Bulk PZT 2 1.4a 2.5 295.3 50 1282.74a 0.263
He [7] Bulk PMgN-51 4 1.45 10 566 50 265.92 2.349
a
Values interpolated from data in the referenced literature.

E Fernandes et al
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

be in excellent agreement. To obtain the final optimal value the energy harvester to a specified frequency. Finally, it is
for the load resistance, the power plot is produced for the seen in figure 17 that a larger harvesting frequency bandwidth
optimal range found in figure 14 and is shown in figure 15. is obtained with increasing excitation force.
An area of the plot between approximately 1 MΩ to 2 MΩ,
the optimal range, shows small variation (less than 5%) in 6.7. Comparisons to other works
power production and can serve as the optimal load
resistance. In this section, the results of the proposed design are com-
pared to the previous work in the literature and presented in
table 6. As previously mentioned, the overall power output is
6.5. Power output versus the current and distance to the wire
a function of the current in the wire, the distance between the
With the electromechanical model experimentally validated magnet and the wire, and the remanence of the magnets.
and the optimal load resistance range determined, the power Therefore, all of these must be considered for a fair com-
outputs were measured using various distances to the wire and parison. The effective volume considers the sum of the
various electric current values. Figure 16 shows the exper- volume of all beams and magnets. The normalized power
imental power outputs measured across a 1 MΩ resistive load density of the piezoelectromagnetic energy harvesters is cal-
as a function of current and distance to the wire for the culated by dividing the power output by both the effective
harvester. volume and the magnetic force.
It is shown that at a distance of approximately 6.5 mm to By comparing the normalized power densities shown in
a wire carrying a 7 A current, approximately 9 μW is dis- the table, one can see that the present work shows significant
sipated across the 1 MΩ load. Since the force on the magnet is improvement over previous MEMS devices, [19] and [14],
inversely proportional to the square of the distance and the with a 2.9 and 2.2 times increase in the normalized power
power output is proportional to the square of the force, the density, respectively. This improvement validates the advan-
overall power output grows significantly as the distance tages of the centrally-supported meandering geometry and the
between the magnet and wire are minimized. Under large screen-printing fabrication process. The proposed geometry
forces, however, nonlinear effects become more evident and minimizes the number of overall strain nodes that plague
result in a lower experimental power than expected. Further other geometries. The results for the present work also show
tests at smaller distances and larger current values were voi- good potential when compared to the larger scale units in
ded to prevent damage to the unit during these vibrations. [6–8]. Additionally, in these larger scale devices, bulk PZT is
used and is known to have higher electromechanical coupling
compared to the MEMS fabricated technology, and, therefore,
6.6. Nonlinear effects
is expected to have higher normalized power densities. As
As the current in the wire increases, equation (1) predicts a such, the benefits of the MEMS screen-printing fabrication
larger total force on the magnet. This leads to larger dis- technology remains one of the main advantages of the present
placements in the system and, consequently, the open-circuit work. Also, it is seen that the result of the proposed design is
voltage output will increase. This behavior is observed in quite comparable to that of the bulk PZT units for [6], [8].
figure 17(a), which presents the experimental open-circuit
voltage output of the energy harvester for various wire cur-
rents, increasing in 1 A increments from 1 to 8 A. In addition 7. Conclusion
to the increase in voltage output, it is observed in figure 17(a)
that the resonant frequency decreases as the wire current A MEMS piezoelectromagnetic energy harvester has been
increases. It is also observed that the peak frequencies are fabricated through screen-printing of PZT on a stainless steel
shifted to the left for the higher current values. This indicates substrate with a centrally-supported meandering geometry.
of some nonlinear effects that should be investigated further The unit is used to harvest electrical energy in a non-invasive
through careful observation of the output-input ratios-FRFs. manner from the transmission lines to help power the sensors
The experimental voltage FRFs presented in figure 17(b) for a smart grid. The device is modelled using COMSOL, and
also demonstrate a decrease in both the voltage FRF peak validated against its experimental results for the displacement
value and the resonant frequency as the wire current increa- and power frequency response functions. The natural fre-
ses. These effects are attributable with varying degrees to a quency design requirement is chosen as 60 Hz to match the
decrease in the quality factor and the stiffness of the unit fundamental frequency of the current in transmission lines. It
(some softening effects) occurring as the wire current is shown that the proposed geometry experiences dominant
increases and the unit experiences overall larger displace- bending when compared to the previous MEMS-based work
ments during its vibrations [34]. Additionally, variations of on this technology, helps increase the power efficiency, and
the magnetic forces experienced by the unit as its distance to simplifies the electrode geometry. As a result, the present
the wire changes during these oscillations can be partially a work results in an improved normalized power density
cause for this. Although further testing is required to truly compared to the previous MEMS-based piezoelectromagnetic
quantify the nonlinear effects and their potential sources, harvesting technologies for the proposed application.
these results serve to highlight the importance that, for a given Currently, work on integrating the energy harvester with
application, the input force must be considered when tuning power conditioning and wireless communications circuitry is

13
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

in progress [32]. The ultimate goal is to combine the har- 2017: 17th Int. Conf. on Micro and Nanotechnology for
vester, circuitry, and a current sensor into a single chip to Power Generation and Energy Conversion Applications
obtain a self-contained sensing solution for transmission [11] Xu R, Lei A, Dahl-Petersen C, Hansen K, Guizzetti M,
Birkelund K, Thomsen E V and Hansen O 2012 Screen
power lines. Future work will also involve improving the printed PZT/PZT thick film bimorph MEMS cantilever
fabrication process to provide a refined poling process to device for vibration energy harvesting Sens. Actuators A
remove complicated wiring and may reduce the overall 188 383–8
damping. [12] Costache F, Pawlik B and Rieck A 2017 Development of a
compact, low-frequency vibration, piezoelectric MEMS
energy harvester Multidisciplinary Digital Publishing
Institute Proceedings 588
Acknowledgments [13] Roundy S and Wright P K 2004 A piezoelectric vibration
based generator for wireless electronics Smart Mater. Struct.
The authors would like to thank Xiaodong Zhang for his help 13 1131
with the electromagnetic modelling in COMSOL and Isabelle [14] Olszewski O Z, Houlihan R, Mathewson A and Jackson N
2016 A low frequency MEMS energy harvester scavenging
Favre for her help with the micro-wiring of the units. This energy from magnetic field surrounding an AC current-
work was supported by grants from the Waterloo Institute of carrying wire J. Phys.: Conf. Ser. 757 012039
Sustainable Energy, Cisco Systems, UWUB research funding [15] Ibrahim M and Salehian A 2015 Modeling, fabrication, and
and International Research Partnership Grant. experimental validation of hybrid piezo-magnetostrictive
and piezomagnetic energy harvesting units J. Intell. Mater.
Syst. Struct. 26 1259–71
[16] Monin T, Tessier-Poirier A, Léveillé E, Juneau-Fecteau A,
ORCID iDs Skotnicki T, Formosa F, Monfray S and Fréchette L 2016
First experimental demonstration of a self-oscillating fluidic
Armaghan Salehian https://orcid.org/0000-0001- heat engine (SOFHE) with piezoelectric power generation
5032-4124 J. Phys.: Conf. Ser. 012039
[17] Karami M A, Yardimoglu B and Inman D J 2010 Coupled out
of plane vibrations of spiral beams for micro-scale
applications J. Sound Vib. 329 5584–99
References [18] Karami M A and Inman D J 2012 Parametric study of zigzag
microstructure for vibrational energy harvesting
J. Microelectromech. Syst. 21 145–60
[1] Naifar S, Bradai S, Viehweger C and Kanoun O 2017 Survey [19] Paprotny I, White R and Wright P 2010 Modeling, design and
of electromagnetic and magnetoelectric vibration energy fabrication of (10×10×4 mm3) MEMS AC energy
harvesters for low frequency excitation Measurement 106 scavenger for smart grid applications Proc. 10th Int.
251–63 Workshop Micro Nanotechnol. Power Generat. Energy
[2] Leland E, Wright P and White R 2007 Design of a MEMS Convert. Appl. pp 107–10
passive, proximity-based AC electric current sensor for [20] Karami M A and Inman D J 2011 Analytical modeling and
residential and commercial loads PowerMEMS 77–80 experimental verification of the vibrations of the zigzag
[3] Lao S B, Chauhan S S, Pollock T E, Schröder T, Cho I S and microstructure for energy harvesting Journal Vib. Acoust.
Salehian A 2014 Design, fabrication and temperature 133 011002
sensitivity testing of a miniature piezoelectric-based sensor [21] Sharpes N, Abdelkefi and Priya S 2014 Comparative analysis
for current measurements Actuators 162–81 of one-dimensional and two-dimensional cantilever
[4] Leland E S, White R M and Wright P K 2006 Energy piezoelectric energy harvesters Energy Harvesting and
scavenging power sources for household electrical Systems 1 209–16
monitoring PowerMEMS Proc. 6th Int. Workshop Micro [22] Abdelmoula H, Sharpes N, Abdelkefi A, Lee H and Priya S
Nanotechnol. Power Generat. Energy Convert. Appl. 2017 Low-frequency Zigzag energy harvesters operating in
pp 165–8 torsion-dominant mode Appl. Energy 204 413–9
[5] Pollock T 2014 Design, modelling, fabrication & testing of a [23] Sharpes N, Abdelkefi A and Priya S 2015 Two-dimensional
miniature piezoelectric-based EMF energy harvester concentrated-stress low-frequency piezoelectric vibration
Master’s Thesis University of Waterloo energy harvesters Appl. Phys. Lett. 107 093901
[6] Paprotny I, Xu Q, Chan W W, White R M and Wright P K [24] Berdy D F, Srisungsitthisunti P, Jung B, Xu X, Rhoads J F and
2013 Electromechanical energy scavenging from current- Peroulis D 2012 Low-frequency meandering piezoelectric
carrying conductors IEEE. S. J. 13 190–201 vibration energy harvester IEEE Trans. Ultrason.
[7] He W, Li P, Wen Y, Zhang J, Lu C and Yang A 2013 Energy Ferroelectr. Freq. Control 59 846–58
harvesting from electric power lines employing the Halbach [25] Sharpes N, Abdelkefi A and Priya S 2016 Mode shape
arrays Rev. Sci. Instrum. 84 105004 combination in a two-dimensional vibration energy harvester
[8] Chen W, Cao Y and Xie J 2015 Piezoelectric and through mass loading structural modification Appl. Phys.
electromagnetic hybrid energy harvester for powering Lett. 109 033901
wireless sensor nodes in smart grid J. Mech. Sci. Technol. 29 [26] Wagner B, Kreutzer M and Benecke W 1993 Permanent
4313–8 magnet micromotors on silicon substrates
[9] Clement P, Perez E D C, Gonzalez O, Calavia R, Lucat C, J. Microelectromech. Syst. 2 23–9
Llobet E and Debéda H 2016 Gas discrimination using [27] Fernandes E, Zarabi S, Debéda H, Lucat C, Nairn D,
screen-printed piezoelectric cantilevers coated with carbon Wei L and Salehian A 2016 Modelling and fabrication of a
nanotubes Sens. Actuators B 237 1056–65 compliant centrally supported meandering piezoelectric
[10] Katsumura H 2017 Development of piezoelectric vibration energy harvester using screenprinting technology J. Phys.:
energy harvesters for battery-less smart shoes PowerMEMS Conf. Ser. 773 012109
[28] AK Steel www.aksteel.com

14
Smart Mater. Struct. 27 (2018) 035017 E Fernandes et al

[29] K&J Magnetics www.kjmagnetics.com contained piezoelectric-based monitoring device for smart
[30] Castille C 2010 Etude de MEMS piézoélectriques libérés et grids PowerMEMS 2017: 17th Int. Conf. on Micro and
microstructurés par sérigraphie.Application à la détection en Nanotechnology for Power Generation and Energy
milieu gazeux et en milieu liquide PhD Thesis Université de Conversion Applications
Bordeaux [33] Inman D J 2014 Engineering Vibration (Upper Saddle River,
[31] Debéda H, Clément P, Llobet E and Lucat C 2015 One-step NJ: Pearson)
firing for electroded PZT thick films applied to MEMS [34] Shen D, Park J, Ajitsaria J, Choe S, Wikle H C III and Kim D
Smart Mater. Struct. 24 025020 2008 The design, fabrication and evaluation of a MEMS
[32] Debéda H, Rua-Taborda I, Fernandes E, Zarabi S, Nairn D, PZT cantilever with an integrated Si proof mass for vibration
Wei L and Salehian A 2017 Printed MEMS-based self– energy harvesting J. Micromech. Microeng. 18 055017

15