Proceedings of PowerMEMS 2008+ microEMS2008, Sendai, Japan, November 9-12, (2008)
HARVESTER-ON-CHIP: DESIGN AND FABRICATION OF A PROOF OF
CONCEPT PROTOTYPE
Gonzalo Murillo1, Gabriel Abadal1, Francesc Torres1, Joan Lluis Lopez1, Joan Giner1,
Arantxa Uranga1, Nuria Barniol1
1
Dept. Enginyeria Electrònica, Universitat Autònoma de Barcelona, Bellaterra, Spain
Abstract: This paper presents a novel concept in energy scavenging generators, the energy Harvester-onChip (HoC). We define HoC as a chip only dedicated to harvest useful energy to supply to another chip. The
HoC proposed here is an in-plane gap-closing electrostatic transduction generator. Its design includes two
new ideas to increase the extracted energy, to use several microscavenging cells instead of a typical
millimetric mass and to take the whole chip as movable inertial mass, by means of anchoring these microcells.
This device, which is only a proof of concept prototype, has been fabricated in a commercial CMOS
technology. Preliminary physical characterizations have been performed in order to validate mechanical
behavior. Theoretically, for an environmental vibration acceleration of 0.2 m/s2 to 100Hz., a totally dedicated
chip could supply about 0.958 µW and a power density of 239.9µW/cm3 ,comparable with the state-of-art
values.
Key words: Energy Scavenging; Energy Harvesting; Micropower Generation; CMOS Integration; MEMS;
Bateryless; Harvester-on-Chip
1. INTRODUCTION
2. ENERGY SCAVENGING INNOVATIONS:
THE HARVESTER-ON-CHIP CONCEPT
From a decade ago, due to its energy density,
batteries are not able to follow the scaling down of the
other microelectronics parts, fulfilling the Moore’s
Law. The new tendency toward tiny, portable, wireless
and cheap systems is not compatible with a battery
with finite amount of energy, big size and difficult
recharge or replace. Therefore, the research about new
devices that extract useful energy from available
energy in the environment is increasing. This is the
basic idea of energy harvesting.
Among all types of energy sources, especially
attractive is the use of the environmental vibrations.
Basically, there are three types of transductions [1, 2]:
piezoelectric, electromagnetic and electrostatic. The
last one is the most suitable one to integrate with
standard electronic circuitry, which is actually based
on CMOS technology.
This paper is focused on the monolithic integration
of energy scavengers in a commercial CMOS
technology [3]. Those devices will play the role of
micropower generators that harvest energy from low
vibration environments, especially for very low power
devices and non-continuously working systems.
The frequencies are mainly and widely distributed
near 100 Hz. Note also that the acceleration
magnitudes range from about 0.1 to 10m/s2. Therefore,
we will consider a typical low-level vibration of 0.2
m/s2 to 100Hz, for this paper.
2.1 Vibrant Energy Harvester Study
In the Fig. 1, we can see the mechanical model of
an energy scavenging system, with a movable element
which is designed to get in resonance with an external
vibration, increasing the mechanical energy of the
system. This energy is converted in electrical energy
by means a transduction method, and has to be stored.
fmec
m
z(t)
y(t)
fdamp
ftrans
x(t)
Fig. 1: Model of a generic electrostatic scavenger.
The equation of movement of this system is:
mz&&(t ) + bz& (t ) + kz (t ) + ftrans = mAcc sin (ωt ) . (1)
225
Proceedings of PowerMEMS 2008+ microEMS2008, Sendai, Japan, November 9-12, (2008)
where ftrans is the force due to the transduction, w and
Acc are the frequency and the acceleration amplitude of
the environmental vibrations respectively, k is the
spring constant, b is the damping coefficient, m is the
inertial mass and z is the displacement.
It is important to have enough mass in order to
increase the work that the external acceleration will do
to increase the energy of the system, although the
transduction force, which will be really what does the
work that will finally be stored, must be coupled
correctly with the inertial force.
Basically, the electrostatic transduction is based on
a variable capacitance, whom internal energy increases
in each transition from maximum to minimum
capacitance (keeping constant the charge), by an
amount which is given by [1, 3, 4]:
∆U Q =cte
1
= Vin2 rC ∆C ,
2
(a)
Structures anchored to the vibrant environment
(b)
Fig. 2: Novel ideas of this work.
Generally, we can define a HoC as a chip only
dedicated to scavenge energy, which is supplied to
another chip or system. This outer system can be a
substrate, as shown in Fig. 3, which can be anchored
electrically and mechanically to the HoC by means of
a similar flip-chip process. Moreover, since we
fabricate the chip in a commercial CMOS technology,
we can reuse the same chip, already used to contain all
low power circuits, sensors and receivers, to make
MEMS elements which are able to extract useful
energy from the environmental vibrations.
(2)
where rC is the ratio between maximum and minimum
capacitance and DC is the difference between the
maximum and minimum capacitance as well.
Finally the power that generates the scavenger is
the product of the extracted energy per capacitance
cycle by the cycle numbers in a period, and by the
frequency, f:
Power = ncycles ∆U cycle f
Multi scavenging
cells device
Typical millimetric
electrostatic
scavenging device
(3)
2.3 Novel ideas: Harvester-on-chip
In the state of the art [2], we can see several
scavengers with electrostatic transduction. These
devices used to be millimetric inertial mass with comb
drivers in both sides. In order to maximize the
transduction surface, several multi micro scavenging
cells are proposed (Fig. 2.a). Another new concept is
the use of the whole chip as inertial mass, i.e. the chip
is anchored to the vibrant environment by the cells,
instead of by the substrate (Fig. 2.b). Since the
capacitance and the mass are increased with these
ideas, then the extracted energy will increase too.
Fig. 3: Assembling of a HoC with a support substrate,
by means of flip-chip bonding.
The joint of these ideas, where we have a lot of
scavenging cells anchored to the environment and
coupled to the chip by means of individual suspensions,
converges towards the new concept of energy
Harvester-on-Chip (HoC). Therefore, the chip mass,
the equivalent whole spring constant of all the
suspensions, and the total variable capacitance are the
global main design parameters.
3. FABRICATION PROCESS
The chip has been fabricated using a 0.35µm
CMOS technology, which has four metal layers and
two polysilicon layers.
In order to exploit the back-end layers to make
MEMS structures, the metal layers are used as like
structural layers, and the oxide layers as sacrificial
226
Proceedings of PowerMEMS 2008+ microEMS2008, Sendai, Japan, November 9-12, (2008)
to the gravity and overall the capillary forces which
appear between when there is a gas-liquid interface in
the typical drier step.
layers. Therefore, when the structures were fabricated
in the commercial foundry, a wet-etching post-process,
with an HF-based solution, was performed in order to
release the movable structures (Fig. 4). The vias
between metals are used to increase the thickness of
the metal layers [3].
PROT2
1000nm
1000nm
PROT2
PROT1
900nm
PROT1
Metal4
925nm
Metal4
925nm
IMD3
1000nm
IMD3
1000nm
Metal3
640nm
Metal3
640nm
IMD2
620nm
IMD2
620nm
vias
Metal2
BHF
1000nm
1000nm
900nm
640nm
Metal2
640nm
IMD1
645nm
IMD1
645nm
Metal1
665nm
Metal1
665nm
645nm
282nm
290nm
ILDFOX
Poly1
FOX
645nm
282nm
290nm
Poly2
ILDFOX
Poly1
FOX
200nm
Bulk Si
COMB DRIVERS
PAD TO BE
ANCHORED
SUSPENSIONS
CROSS -CONNECTED PAIR
OF SCAVENGING CELLS
Bulk Si
In-foundry standard CMOS fabrication
process
Out-foundry maskless BHF etching
Fig. 4: Fabrication process steps.
Fig. 6: Layout of whole chip fabricated.
5. PRELIMIARY CHARACTERIZATION
4. DESIGN OF THE PROTOTYPE
Several chips have been processed with different
etching times, in order to find the optimum one which
achieves to release the structure, but without important
damages in the structural metal layers.
The first characterizations performed were optical
images, which are shown in the Fig. 7. We can see in
the comb driver fingers detail that these fingers are out
of focus respect to the others. It was the first indication
that the suspended masses are collapsed with the
substrate.
Several FEM simulations have been performed
using COVENTOR® software, e.g. whole chip modal
simulations, suspensions behavior, structures harmonic
responses, etc. (Fig. 5) Also, a program using
MATLAB code has been developed in order to solve
the differential equations system. According to the
results, the expected resonant maximum displacement
is around 12.5 µm. With the theoretical expressions
and using these simulation tools, a proof of concept
prototype was designed. This chip have four pairs of
electrically cross-connected scavenging cells pair,
distributed near the four corners.
Position & Velocity
20
0.01
position
velocity
0.005
0
0
-10
-20
0
Velocity(m/s)
Position(µm)
10
-0.005
0.05
0.1
0.15
Time (sec)
0.2
-0.01
0.25
Fig. 5: Displacement and velocity graphics (left) and
modal FEM simulation of the whole chip (right).
Fig. 7: Optical views of the chip fabricated.
In the Fig. 6 it is shown the whole chip layout and
the detail of a cell layout. It is basically a suspended
mass, with the typical dimensions of a pad, with
fingers in both sides, which form a comb driver. Four
U-type suspensions were designed to achieve a flexible
spring in the surface chip plane, but the hardest spring
possible in the perpendicular direction. If the
suspension is not enough hard in this direction, the
suspended mass could collapse with the substrate due
From the SEM images shown in the Fig. 8 , we can
see that all the sacrificial oxide is removed, and
therefore, the structure is released. We can also see in
the details that the aluminum of the metallic layer has
been etched, because the etching time was too long.
Finally, a 3D optical profiler, based on a confocal
microscope, was used for obtaining 3D images and 2D
profiles of the chips.
227
Proceedings of PowerMEMS 2008+ microEMS2008, Sendai, Japan, November 9-12, (2008)
Afterwards, several chips were processed with a
critical point drier, in order to avoid these forces. The
3D profiles of the scavenging cells in these chips show
that the cells are not collapsed. A good releasing and
drying or a partial releasing could explain these results.
More detailed analysis must be performed in order to
obtain a mechanically and electrically functional chip.
7. CONCLUSION
This paper presents several novel ideas in energy
scavenging systems, which converges in the new
concept of Harvester-on-chip. A complete theoretic
study of this type of devices has been developed, and
the electromechanical behavior has been simulated by
means of FEM tools and an ad hoc code which solve
the complete differential equations system. Using these
analyses, a proof of concept prototype has been
designed and fabricated with a standard CMOS
technology. Preliminary physical analyses have been
performed and good patterning and releasing has been
validated, but sticking problems must be solved yet.
For this prototype, a maximum power about 76.67
nW is expected for a low-level environmental
vibrations, with an acceleration of 0.2 m/s2 and a peakfrequency of 100Hz. Therefore, for a whole energy
harvesting chip, a theoretical power density about
239.9µW/cm3, which is comparable with the state-ofart values, has been calculated.
Fig. 8: SEM images of a scavenging cell, with detail of
suspensions and layers stack.
In Fig. 9 , we can see that the suspended masses
are collapsed to the substrate, because the difference
between the fingers height is the same than the
distance between the bottom cell and the substrate, i.e.
around 1.2 µm.
7. ACKNOWLEDGMENTS
This work has been supported by project
MEMSPORT (TEC2006-03698/MIC). The authors
want to thank Marta Duch, Marta Gerboles and
Jaume Esteve from IMB-CNM for their support
on the post-CMOS process.
(a)
8. REFERENCES
(b)
1.2 µm
[1] Roundy S, Wright P K,Rabaey J M 2004 Energy
Scavenging for Wireless Sensor Networks: With
Special Focus on Vibrations
[2] Beeby S P, Tudor M J,White N M 2006 Energy
Harvesting Vibration Sources for Microsystems
Applications Meas. Sci. Technol 17 175–195
[3] Murillo G, Abadal G, Torres F, Lopez J L, Giner
J, Uranga A,Barniol N 2008 On the Monolithic
Integration of Cmos-Mems Energy Scavengers 23rd
Conference on Design of Circuits and Integrated
Systems (DCIS) (Grenoble, France, 12-14 November
2008)
[4] Miranda J O M 2003 Electrostatic Vibration-toElectric Energy Conversion
Fig. 9: 3D profile of a pair of scavenging cells with
cross section detail in fingers: collapsed (a), and noncollapsed (b) structures.
6. DISCUSSION
These first physical characterizations showed that
the suspended scavenging cells are stuck to the
substrate, due to the capillary forces during the drying.
228