Simulation and Optimization of MEMS Piezoelectric

Energy Harvester with a Non-traditional Geometry

S. Sunithamani1, P. Lakshmi1, E. Eba Flora1
1
Department of EEE, College of Engineering, Anna University, Chennai, India

Abstract: Piezoelectric energy harvester with COMSOL for the conversion of mechanical
converts mechanical vibration into electrical into electrical energy.
energy via piezoelectric effect. In this paper a
MEMS based energy harvester with a non- 2. Modelling in COMSOL Multiphysics:
traditional geometry is designed. The design of
the energy harvester consists of a T-shaped Piezoelectric application mode is used to
cantilever structure with triangular shape at the model and simulate the non-traditional geometry
tip. Simulation results demonstrated that the new and moving mesh application mode is used to
cantilever structure can improve the strain optimize the thickness of piezoelectric layer that
distribution and generate more voltage than the maximize the output electrical energy.
triangular and rectangular structures at the
resonance frequency. The proposed structure is 2.1 Geometry
simulated using the software COMSOL
Multiphysics and the results are compared with A 3 dimensional geometry is considered for
other geometries such as rectangular and the simulation. The MEMS piezoelectric energy
triangular. harvester has a unimorph T-shaped cantilever
with triangular shape at the tip, as shown in
figure1.The device is made by a stainless steel
Keywords: Energy harvesting, unimorph substrate with a piezoelectric layer on the top,
cantilever, non-traditional geometry, poled along the thickness direction. The total
piezoelectric energy harvester. length of the substrate is 27000µm and total
width is 3000µm. The thickness of the substrate
1. Introduction T0 = 200µm and that of piezoelectric layer T1 is
set initially as 60µm and then varied from 10µm
The energy harvesting devices converting to 400µm with the moving mesh application.
ambient energy into electrical energy have
attracted much interest within academic
community and industry. Ambient energy is
available in the form of vibration, light,
temperature gradient etc. Among these energy,
mechanical vibration is the most widespread and
wasted energy in the environment. This
mechanical vibration can be converted into
electrical energy by three kinds of generators
such as electromagnetic, electrostatic and
piezoelectric. Piezoelectric generators are mostly
used because piezoelectric materials have the
advantage of large power and ease of
application.
The most studied energy harvesters are based
on the piezoelectric effect and are made with
MEMS technology. The geometry of
Figure 1. Structure of piezoelectric energy harvester
piezoelectric cantilever beam greatly affects its with non-traditional geometry. L=27000µm,
vibration energy harvesting ability [1]. In this L0=2000µm, L1=18000µm, L2=7000µm, W=3000µm,
paper MEMS based energy harvester with a non- W0=W1=1000µm, T0=200µm, T1=210 µm.
traditional geometry is designed and simulated

Excerpt from the Proceedings of the 2012 COMSOL Conference in Bangalore

 piezoelectric layer. The substrate is made of
stainless steel using the decoupled isotropic
material available in the COMSOL material
library. The material parameters of the substrate
are as follows: its density ρ=7850 kg/m3,
young’s modulus E=200 x 109 Pa, its Poisson’s
ratio μ=0.33.
The active layer of unimorph is made up of
ultra low-temperature piezoelectric material (0-3
composite) [2] and it is modelled using the
following set of properties

-Elastic compliance tensor

 50 20 20 0 0 0 
Figure 2. Piezoelectric Energy Harvester mesh. SE =  20 50 20 0 0 0  x 10-12 Pa-1
 
 20 20 50 0 0 0
2.2. Meshing  
 0 0 0 70 0 0 
 0 0 0 0 70 0 
Figure 2 shows the meshing of piezoelectric  
 0 0 0 0 0 70 
energy harvester.The mesh consists of 238 quad -Piezoelectric tensor
elements for a total number of degrees of
d =  0 0 0 0 11 0  x 10-12 CN-1
freedom 10639. The swept mesh tool is used to
 0 0 0 11 0 0 
create mesh using opposite vertical surfaces of  2.5 2.5 5 0 0 0
each layer as a source face and target face.
-Relative permittivity matrix
2.3. Piezoelectric Equations εT = 50 0 0  x ε0
0 50 0 

 0 0 50 
Piezoelectric materials produce electrical
charge when it is mechanically deformed. The -Density ρ = 3000 kg m-3
piezoelectric energy harvester is governed by the
following equations written in the strain-charge The vertical acceleration is applied using
format body load FZ equal to aρ in each subdomain,
S = sE T + d E where a = 0.1g represents the acceleration
(1) magnitude and ρ is the density of the material
D = εT E + d T [2].

Where S is the mechanical strain vector, sE 3. Boundary conditions

elastic compliance tensor (Pa-1), T mechanical
stress vector (Nm-2), D the elastic displacement 3.1 Piezo solid model
vector (Cm-2), εT the dielectric permittivity tensor
(Fm-1), E the electric field vector (Vm-1) and d One end of the unimorph cantilever is fixed
the transverse piezoelectric coefficient tensor while other is free for vibration. Therefore the
(CN-1). For the substrate layer only mechanical fixed constraint condition is applied for the
behaviour is considered using stress-strain vertical faces of both the layers, while all other
relationship faces are free of displacement. In order to pole
S = sT (2) piezoelectric layer along thickness direction, the
Where s is the compliance of stainless steel electrical behaviour of PZT must be considered
substrate. and it is modelled with the electrostatic boundary
conditions. The upper and lower face of PZT
2.4 Subdomain settings layer are selected as floating and ground
potentials respectively, while all other faces of
The geometry consists of two subdomains, piezoelectric layer are kept as zero charge [3].
one for the substrate layer and another for the

Excerpt from the Proceedings of the 2012 COMSOL Conference in Bangalore

3.2 Moving mesh ALE

Moving mesh application mode is used to

optimize the thickness of the PZT layer. The
bottom face of the piezoelectric subdomain is
constrained as clamped while the vertical faces
were clamped along thickness and left
unconstraint along tangential direction allowing
them to stretch freely. The upper face of the
piezoelectric layer was tangentially constraint
and displaced in the normal direction to the
surface by a given displacement, using a
parameter deltaThickness. The parameter
deltaThickness is changed from 10μm to 400μm
obtaining parameterized moving mesh. Figure 4 Tip displacement (nm) Vs Thickness(μm).

4. Modelling and Optimization 4.2. Thickness optimization of piezoelectric

layer for three different geometries
4.1. FEM modelling of piezoelectric energy
harvester with three different geometries Moving mesh Arbitrary Lagrangian
Eulerian (ALE) method is used to compute the
A finite element model is the basis of new piezoelectric layer dimensions. The
simulation analysis, so at first piezoelectric displacement at the free tip of the piezoelectric
energy harvester in figure1 is modelled by using energy harvester, the open circuit voltage and the
the software COMSOL Multiphysics and then generated charge collected on the electrodes
the model is used to study the deformation, strain where computed for piezoelectric layer thickness
and generated voltage distribution of a varying from 10μm to 400μm. The obtained tip
piezoelectric energy harvester. displacement is shown in figure 4 in log-log
Material properties are applied as described scale. The tip displacement decreases with the
in 2.3 and 2.4 and boundary conditions are increase in PZT layer thickness.
applied as explained in section 3. The FEM Figure 5 shows open circuit voltage and the
modelling of rectangular, triangular and non- charge generated on the piezoelectric energy
traditional geometries are obtained as in figure 3. harvester electrodes when the piezoelectric layer
thickness is varied. The charge reaches
maximum for the low piezoelectric layer
thickness, decreasing then when the rigidity of
the piezoelectric layer becomes significant and it
influences the whole converter rigidity. On the
other hand, the converter capacitance C
monotonically decreases with thickness with an
inverse proportionality. For this reason, being the
output voltage V the ratio between charge Q and
capacitance C, the voltage increases for a
constant charge level Q, and it remains instead
constant when both Q and C decrease.
The electrical energy converted from
mechanical vibrations are stored in the
piezoelectric material was computed with the
following equation and it is plotted in figure 6 as
a function of the piezoelectric layer thickness
Figure 3 FEM modelling of rectangular, triangular 1
and non-traditional geometries. E  QV (3)
2

Excerpt from the Proceedings of the 2012 COMSOL Conference in Bangalore

Figure 5 voltage (mV) Vs thickness (μm) and charge
Figure 7 Voltage (V) Vs Frequency (Hz) for three
(e-13) Vs thickness (μm).
different geometries.

5.2 Strain analysis

The strain is the most important factor of

affecting electric energy generation. According
to piezoelectric effect equation, the generated
voltage output of the piezoelectric vibrator is
proportional to strain. In other words, the more
strain of each location is the more generated
voltage output. Finite element simulation used to
obtain strain curves of a rectangular, triangular
and non-traditional piezoelectric energy
harvester in the direction of length and is shown
in figure 8.
Figure 6 Stored electrical energy (fJ) Vs thickness It can be seen from figure 8, the strain of the
(μm). piezoelectric energy harvester with non-
traditional geometry is more than those of
The energy shows a maximum for a rectangular and triangular piezoelectric energy
thickness value of 210 μm and it is considered as harvester. Simulation result shows that the
the optimal piezoelectric layer thickness. In the distribution of strain can be improved effectively
optimal condition the thickness ratio tPZT/tsubstrate by using piezoelectric energy harvester with non-
has a value of 1.05. traditional geometry, which will be useful to
harvest more vibration energy.
5. Results

5.1 Frequency analysis

In order to obtain the resonant frequency,

frequency analysis is carried out at the optimal
thickness value of PZT layer which is 210μm.
The Eigen frequency for non-traditional,
rectangular and triangular geometry are obtained
and plotted in figure 7. The graph shows that at
resonant frequency, the generated voltage of
non-traditional geometry is greater than that of
other two geometries.
Figure 8 Strain curves of three different geometries
along X direction.

Excerpt from the Proceedings of the 2012 COMSOL Conference in Bangalore

 Harvesting Device” Proceeding of the 2011
IEEE Students' Technology Symposium, 265 -
269 (2011), IIT Kharagpur.

8. Acknowledgements

The authors would like to thank Mr. Suyog N

Jagtap for helpful suggestion during simulation.

Figure 9. Performance comparison.

Figure 9 shows the comparison of

deformation and strain of three different
piezoelectric energy harvester geometries.

Conclusions
A piezoelectric energy harvester with non-
traditional geometry was designed and simulated
in COMSOL Multiphysics. The thickness of
PZT layer was optimized to give maximum
stored electrical energy. Frequency analysis and
strain analysis were carried out for the optimized
thickness of 210μm. Simulation results
demonstrate that the piezoelectric energy
harvester with non-traditional geometry
improves strain and generate more voltage at
resonant frequency than the rectangular and
triangular piezoelectric energy harvester.

7. References

[1] Z.S.Chen, Y.M.Yang and G.Q.Deng,

“Analytical and Experimental Study on
Vibration Energy Harvesting Behaviors of
Piezoelectric Cantilevers with Different
Geometries” International conference on
sustainable power generation and supply, 1 - 6 (
2009).
[2] M.Guizzetti, V.Ferrari, D.Marioli and
T.Zawada, “Thickness optimization of a
piezoelectric converter for energy harvesting,”
Proceedings of the COMSOL Conference, 2009.
[3] Suyog N Jagtap and Roy Paily, “Geometry
Optimization of a MEMS-based Energy

Excerpt from the Proceedings of the 2012 COMSOL Conference in Bangalore