1203
1-4244-0842-3/07/$20.00©2007 IEEE
TRANSDUCERS & EUROSENSORS ’07
The 14th International Conference on Solid-State Sensors, Actuators and Microsystems, Lyon, France, June 10-14, 2007
DIAMAGNETICALLY LEVITATED MEMS ACCELEROMETERS
D. Garmire
1
, H. Choo
1
, R. Kant
2
, S. Govindjee
3
, C. H. Séquin
1
, R. S. Muller
1
, J. Demmel
1
1
Berkeley Sensor & Actuator Center, Berkeley, CA, USA
2
Stanford University, Palo Alto, CA, USA
3
ETH Zürich, Institut für Mechanische Systeme, Zürich, Switzerland
Abstract:
We introduce the theory and a proof-of-concept design for MEMS-based, diamagnetically-
levitated accelerometers. The theory includes an equation for determining the diamagnetic force above a
checkerboard configuration of magnets. We demonstrate both electronic probing and a rapid MEMS-
based interferometer technique for position sensing of the proof mass. Through a proof-of-concept
design, we show electrostatic-measurement sensitivity achieving 34 μg at a 0.1 V sense signal and
interferometer-measurement sensitivity achieving 6 μg for in-plane vibrations at 5 Hz. We conclude by
outlining batch-fabrication steps to produce levitated accelerometers.
Keywords:
accelerometer, diamagnetism, levitation, fast phase-shifting interferometer
1. INTRODUCTION
This research presents the theory and
verification for a MEMS-based, diamagnetically
levitated proof mass to improve the low-frequency
performance of accelerometers for applications
such as the study of earthquakes and structural
health monitoring. We have built a levitated proof-
mass device and obtained experimental results that
show electrostatic-measurement sensitivity of 34
μg at a 0.1 V sense signal and interferometer-
measurement
sensitivity
of
6
μg.
The
measurements are targeting 5 Hz vibrations. An
ADXL 203, a commercial MEMS accelerometer,
achieves 285 μg for the same vibrations. Previous
approaches to levitating a proof mass in a MEMS
process use electrostatic active control and suffer
from stiction [1]. In our accelerometer,
diamagnetism is employed to passively levitate a
proof mass containing pyrolytic graphite [2]. We
measure the position of the proof mass using off-
chip circuitry that differentially senses the change
in capacitance across a pair of comb drives. The
comb drives are fabricated in an SOI process [3].
2. EQUATIONS OF MOTION
We show that a lightly damped, non-stiff
suspension gives more precise results when
measuring low-frequency vibrations than does the
stiff suspensions that are typically used in MEMS
accelerometers. Diamagnetic levitation provides
the means to create these optimized suspensions.
The 1D mechanical theory for accelerator
response is formulated in Eq. 1. Calling
x
the
target displacement, the acceleration
)
(
t
x
is:
(
)
)
(
)
(
)
(
1
)
(
t
z
t
Kz
t
z
D
M
t
x
−
+
−
=
,
(1)
where
y
(
t
) is the proof-mass position,
z
(
t
) =
y
(
t
)-
x
(
t
) is the measured displacement, and
M
,
D
, and
K
are the mass, damping, and stiffness of the
system. We simulate the response of this system to
a 10 μg sinusoidal excitation over a frequency
range from 0.1 to 10 Hz (Fig. 1).
The assumed
noise of the position detector has a standard
deviation of 1 nm when sampling at 500 Hz.
10
-1
10
0
10
1
10
-8
10
-7
10
-6
10
-5
Frequency of Vibration
Standard Deviation of Error
Simulated
Accuracy
of
Suspensions
K/M=100000, D/M=100
K/M=100, D/M=100
K/M=100, D/M=10
K/M=0, D/M=0
Figure 1: The accuracy of accelerometers using
four different suspensions is measured by finding
the standard deviation of the suspension response
compared to a sinusoidal input at each frequency.
(Hz)
(g)
Isolated Monitoring
1/
¥
Hz tren
d
1/Hz tren
d
2
n
d
derivative error
Typical Suspension
Levitated Suspensions
2EM9.P
1204
1-4244-0842-3/07/$20.00©2007 IEEE
TRANSDUCERS & EUROSENSORS ’07
The 14th International Conference on Solid-State Sensors, Actuators and Microsystems, Lyon, France, June 10-14, 2007
3. DIAMAGNETIC FORCE
Using Fourier analysis and Taylor-series
expansions applied to the diamagnetic governing
equations [2] of a square plate of graphite above a
checkerboard configuration of magnets (Fig. 2),
we derive Eq. 2 which approximates the plate-
levitation height
z
when the proof-mass weight
balances the diamagnetic force (Fig. 3):
(
)
A
z
h
z
w
h
B
F
gA
h
graphite
graphite
load
total
̧
̧
¹
·
̈
̈
©
§
+
≈
+
2
0
5
16
235
.
0
μ
π
χ
ρ
, (2)
where
h
total
is the total plate thickness,
ȡ
is the
plate density (2300 kg/m
3
),
g
is 9.8 m/s
2
,
A
is the
plate area (8mm × 8mm),
F
load
is the vertical load
applied to the plate
,
Ȥ
is graphite’s magnetic
susceptibility (450×10
-6
), μ
0
is 4
ʌ
×10
í
7
N·A
í
2
,
B
is the magnetic flux density (1 Tesla for NdFeB),
w
×
w
is the magnet size (100 μm × 100 μm), and
h
graphite
is the graphite-layer thickness.
We verify Eq. 2 by placing weights onto a
levitated plate of pyrolytic graphite (7.7×10
-5
N)
while monitoring the vertical displacement
through a side-mounted microscope (Fig. 3a). We
reduce tilting of the proof mass by positioning the
weight while monitoring the angular change of a
reflected laser beam. We also compare adding
additional pieces of pyrolytic graphite (Fig. 3b).
Figure 2: Two equipotential surfaces of the
diamagnetic field above alternating north- and
south-pole magnets (potential increases closer to
the magnets).
0.7
0.8
0.9
1
1.1
1.2
1.3
x 10
-3
0
2
4
6
8
x 10
-4
Force (N)
Levitation Height (m)
Modeled Data
Measured Data
4
6
8
10
12
14
x 10
-4
0
2
4
6
8
x 10
-4
T
h
i
c
k
n
e
s
s
(
m
)
Levitation Height (m)
Modeled Data
Measured Data
Figure 3: Theoretical (Eq. 2) and experimental
measurements of the levitation height of the proof
mass (a) under varying load and (b) with varying
thicknesses of graphite.
4. PROOF-OF-CONCEPT RESULTS
To test the MEMS levitated accelerometer, we
mount a fabricated SOI proof mass to several
layers of pyrolytic graphite. We levitate it above
NdFeB magnets and align the electrostatic-sensing
combs to it using a 3-axis stage (see Fig. 4a). We
measure and compare the responses of our
levitated accelerometer to those of the ADXL 203
accelerometer when both are subjected to tapping
impulses applied to the support table (see Fig. 4).
We reflect a laser beam off of the proof mass and
record its position on a CCD image sensor to
measure out-of-plane tilting motions of the plate
(Fig. 6). By processing the results using off-chip
circuitry, we show that the noise level of the
levitated accelerometer is at least a factor of 8.4
smaller than the noise level of the ADXL 203
using only 0.1V for the differential sense (specific
results are at 5 Hz). Upon excitation of the air
table, a higher-frequency (40 Hz) tilting-mode
coupling is measured. We can reduce the
amplitude of this mode by packing more magnets
under the levitated proof mass.
(a)
(b)
N
N
S
Graphite Square
2EM9.P
1205
1-4244-0842-3/07/$20.00©2007 IEEE
TRANSDUCERS & EUROSENSORS ’07
The 14th International Conference on Solid-State Sensors, Actuators and Microsystems, Lyon, France, June 10-14, 2007
At rest, the standard deviations of the ADXL
and our levitated accelerometer
measure,
respectively, 285 μg and 34 μg when we use a 0.1
V sense signal and average signals over 0.1
seconds. A larger sense voltage increases the
sensitivity of the detector (0.34 μg possible at a
10V applied signal). Note that added aluminum,
produces the desired damping of in-plane motion.
Figure 4: (a) Schematic of experimental setup;
(b) microscope photo of levitated proof mass;
(c) the magnets under the levitated proof mass.
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
-0.01
0
0.01
time (seconds)
acceleration (g)
levitated accelerometer
ADXL
0.2
0.4
0.6
0.8
1
-1
-0.5
0
0.5
1
x 10
-3
time (seconds)
acceleration (g)
levitated accelerometer
ADXL
Figure 5: Comparison of an ADXL 203 and the
levitated accelerometer excited (a) and at rest (b).
0
0.2
0.4
0.6
0.8
1
0.05
0.052
0.054
0.056
0.058
time (seconds)
angle (radians)
Figure 6: Tiling of the levitated proof mass occurs
in the first 0.4 seconds and then stabilizes.
Next, we examine the motion of a diamagnetic
proof-mass (D/M = 73 Hz, K/M = 4 Hz
2
) with a
fast
phase-shifting
interferometer.
This
interferometer uses a vibrating MEMS plate to
achieve rapid phase-shifting and therefore a high
sampling rate (50 Hz) [5] (Fig. 7). We use a Fisher
Scientific U56001 vibration generator to supply a
1.6 N impulse to the 300 Kg Newport air table.
From the response measured by the levitated
accelerometer (Fig. 7), we are able to determine
the table’s suspension values, K/M
§
11.1 Hz
2
and
D/M
§
2.3 Hz. The noise in the system has a
standard deviation of 6.0μg, not the limit of the
detector but the limit of isolating the table from
surrounding vibrations.
0
2
4
6
8
-1
-0.5
0
0.5
1
x 10
-4
Figure 7: The fast phase-shifted interferometer (a)
measures the acceleration (b) of the 300 kg air
table after a 1.6 N impulse is applied to it.
(a)
(b)
(c)
(a)
(b)
Measured Acceleration (g)
Acceleration (g)
Acceleration (g)
Angle (radians)
Time (seconds)
Time (seconds)
Time (seconds)
N
oise has 6.0
μg
standard deviation
Time (seconds)
(b)
A
pp
lied im
p
ulse to the table
A
pp
lied im
p
ulse to the table
(a)
2EM9.P
1206
1-4244-0842-3/07/$20.00©2007 IEEE
TRANSDUCERS & EUROSENSORS ’07
The 14th International Conference on Solid-State Sensors, Actuators and Microsystems, Lyon, France, June 10-14, 2007
5. IDEAL FABRICATION AND DESIGN
We show the feasibility of a MEMS process to
fabricate a levitated accelerometer with the
necessary
levitation
height
and
dynamic
properties. Evaluating Eq. 2, we calculate a
levitation height of 26 μm for a 2 μm-thick
pyrolytic graphite layer [4] on top of a 20 μm-
thick silicon plate with square magnets (100 μm
on a side) fabricated on the substrate. The magnet
arrays can be made separately and bonded to the
device to reduce complexity. Thus, the plate is
separated by a 6 μm-gap from the substrate and
magnets. Depositing a 0.122 μm layer of
aluminum on top of the plate results in eddy
current damping of D/M =
c
B
2
/
ȡ
h
aluminum
/
h
total
=
100 Hz, where
c
is aluminum’s conductivity
(37.7×10
6
/ohm m).
Fig. 8 shows a conceptual view of a processed
device. The magnets can be made separately and
placed in a backside trench or opening. Additional
magnets should be placed above to keep the proof
mass in place during the release of the proof mass
from its surrounding, and to lessen out-of-plane
tilting effects during operation.
Figure 8: The levitated proof mass construction.
Differential
comb-drives
are
used
in
electrostatic sensing for accurately determining
position. Differential vertical-offset combs can be
fabricated directly in SOI to electrostatically sense
out-of-plane motion [3] (Fig. 9).
Figure 9: Vertical sensing using offset combs.
The levitated accelerometer can be designed to
use either electrostatic sensing or interferometric
sensing. As demonstrated, using fast phase-
shifting interferometer techniques, we can
inexpensively and rapidly monitor lateral motions
at frequencies above 100 Hz with a precision
better than 2 nm per frame.
6. CONCLUSIONS
We have shown that diamagnetically levitated
accelerometers perform well at measuring low-
frequency vibrations below 5 Hz. Using
electrostatic measurements, we achieve 34 μg
sensitivity at a 0.1 V sense signal. Using
interferometer measurements, we achieve 6 μg
sensitivity. Furthermore, we have shown that such
accelerometers are possible in MEMS.
REFERENCES
[1]
B. Damrongsak and M. Kraft.
Electrostatic
suspension control for micromachined inertial
sensors employing a levitated-disk proof mass
.
Proc. MME 2005 Conference, Sept. 2005, pp.
240-243.
[2]
M. V. Berry et al.
Of flying frogs and
levitrons
. Eur. J. Phys. Vol. 18, 1997, pp. 307-313.
[3]
H. Choo et al.
A simple process to
fabricate self - aligned, high - performance
torsional microscanners: demonstrated use in a
two-dimensional scanner
. 2005 IEEE/LEOS
Optical MEMS, August 1-4, 2005, pp. 21-22.
[4]
Kostecki et al.
Surface studies of carbon
films from pyrolized photoresist
. Thin Solid Films,
396, (2001).
[5]
H. Choo et al.
Fast, MEMS-Based, Phase-
Shifting Interferometer
. Solid-State Sensor and
Actuator Workshop, June 4-8, 2006, pp.94-95,
Hilton Head, SC USA.
(attached on backside)
2EM9.P