Thin-Film Piezoeelectric-on-Silicon Particle Mass
Sensors
Brandon P. Harrington and Reza Abddolvand
Arash Hajjam, James C. Wilson, and Siavash
Pourkam
mali
School of Electrical and Computer Enginneering
Oklahoma State University
Tulsa, OK, USA
[email protected]
University off Denver
Denver, CO
O, USA
Abstract — In this paper, high quality factor (Qair>18000), high
frequency (~27MHz and ~54MHz), lateral--extensional mode
thin-film piezoelectric-on-silicon resonators arre used as aerosol
particle mass sensors. Using an aerosol particcle generator, mass
sensitivities of -4.2Hz/pg and -42Hz/pgg are measured
respectively which confirms the benefits of employing higher
frequency resonators given the quality factor iis not deteriorated.
These results are in good agreement with th
he theoretical and
simulated values. Our work suggests the TPoS
S resonators as an
easily integrated, low-loss platform for particle sensing
applications.
I.
INTRODUCTION
make atmospheric
Currently, the equipment required to m
particle measurements are bulky and expennsive and do not
provide desired levels of performance [1] [[2]. Atmospheric
aerosol particles ranging in size from a few nm to tens of um
have a direct impact on air quality and thhereby have a far
reaching effect on health [3] and global clim
mate [4]. MEMS
resonators functioning as mass sensors can prrovide a low-cost,
high-sensitivity, and easy to integrate platfforms for aerosol
measurement instruments. MEMS resonatoors with their high
frequencies and very small masses can achieve greater mass
sensitivities. However, without addressingg the operational
noise level, gain in sensitivity would not neccessarily yield in a
better sensor. The resonator’s high mechaniical quality factor
(Q) can work to improve the signal-to-noisee ratio (SNR) and
therefore lower minimum detectable mass w
while allowing low
power operation. To date, many approaches to MEMS mass
sensors have been investigated.
Capacitive-transduced, very high-Q ((>100,000) mass
sensors have been demonstrated in the paast. However, for
capacitive devices to operate at high frequeency, very narrow
transduction gaps are required [5]. For m
mass sensing this
small gap is undesirable. First the significcant gas damping
forbids this type of device from operation in air with a high Q.
Lastly, the small transduction gap providdes an additional
failure mechanism by allowing particles trappped in the gap to
obstruct the resonator’s motion.
978-1-4244-6401-2/10/$26.00 ©2010 IEEE
Thermally actuated resonant masss sensors are a relatively
new idea. They offer operation in atmospheric
a
environments
and exhibit great robustness against deposition of
contaminants and particulates on theeir structures [6]. On the
other hand, their high power consum
mption (tens of mW) could
limit remote or portable use.
Recently, nanomechanical resonaators have been pursued to
push the minimum detectable mass liimits closer to single atom
level. Such tiny resonators benefit from
f
their extremely small
masses to achieve this performance.. However, these devices
typically suffer from relatively lo
ow quality factors, low
resonant frequencies, and nanoscalee transduction challenges
[7].
A thin-film piezoelectric-on-siliicon (TPoS) resonator as
seen in Fig. 1, offers a unique solution in comparison to the
previously mentioned approaches. In combination with the
low loss single-crystal silicon reesonant body, the high
piezoelectric transduction coefficientt results in a low motional
impedance device. The TPoS resonaator design is able to scale
its frequency into the GHz regime while
w
maintaining a high Q
in air with little detriment [8]. Osciillators operating at 1GHz
with low phase noise based on high
h-Q TPoS resonators have
already been realized in [9] so an exttension to high frequency,
high-Q, low-noise mass sensing witth these devices is readily
obtainable.
Silicon
Metal Electrode
Piezoelectric Layer
Ground Plane Via
Figure 1: A 3rd order lateral extensional thin-ffilm piezoelectric-on-silicon
resonator (TPoS).
238
THEORY
In a resonant mass sensor, the incidennt mass particles
perturb the measuring resonator through aadditional inertial
mass causing the resonant peak to shift. If the resonator
resides within an oscillator circuit, the reesulting oscillator
frequency becomes dependent on the absorbeed mass.
0.994) mass sensitivity of
in Fig. 2 has a nearly linear (R2=0
-43.9Hz/pg.
55.274
Frequency (MHz)
II.
To accurately analyze the mass sensinng limit for any
resonator one must look at the different addittions of noise into
the system. Fortunately much work has beenn done in this area
showing the dominant source of noise to be thermomechanical fluctuations. When limited by thhis phenomenon, it
has been shown that the minimum detectablee change in mass,
(δM) in a PLL-based is [10]:
∆
55.271
55.269
,
(1)
0
where the Meff is the effective mass of the reesonator, Eth is the
thermal energy, Ec is the maximum drive energy at critical
vibration amplitude, Δf is the measurementt bandwidth, Q is
the resonator quality factor, and ω0 is the aangular resonance
frequency. The first ratio represents the inveerse of the signalto-noise ratio and shows that the more poweer is driven to the
resonator while it remains linear, the smaller the detectable
change in mass will be. In addition, if the measurement
bandwidth is increased, the thermo-mechannical noise grows,
lessening the sensitivity.
The two most common approaches to reeducing δM are to
make the device less massive (lower Meff) oor to increase the
frequency-Q product (f.Q). Most approachees to mass sensing
have been focused on the former, Meff reducttion. In doing so,
these devices have been hampered by reduceed Q. The goal of
the devices in this paper is to increase the ff.Q in an effort to
improve the minimum detectable mass changge (δM) and mass
sensitivity (δf/δM).
Even so, in the effort to increase f.Q, tthe overall sensor
performance can still be limited. A pushh to ever higher
frequencies can lead to a diminished Q yet sstill a greater f.Q.
Q has been shown to be inversely proportiional to the Allan
deviation, the standard measure of short-term
m instabilities [11].
Therefore, to truly improve the performancee for an oscillatorbased mass sensor, Q should remain constaant or improve as
the resonant frequency increases.
Sincce current TPoS
resonator f.Q of 6.6x1012 [9] have not approached the
maximum theoretical f.Q for the single-crystaal silicon resonant
structure, 1014 as predicted by [122], simultaneous
improvement in frequency and quality factor is still
obtainable.
III.
55.272
55.270
IV.
40
0
Number of Particles
P
60
80
RESONATOR DESIGN AND
A
FABRICATION
To study the effectiveness of the
t resonators as a mass
sensor, 27MHz and 54MHz reson
nators were designed and
fabricated. The resonators were fabricated
f
on 30μm thick
device layer SOI wafers with a piezo
oelectric aluminum nitride
layer, sandwiched between two moly
ybdenum layers that form
the electrodes and electrical interco
onnects. The five mask
procedure is a low temperature (< 250C), post-CMOS
compatible process starting with thee layered stack on top of
the SOI wafer. After a series of
o dry etches define the
electrode shape and resonator body, the handle silicon layer is
anisotropically wet etched from the backside stopping at the
SOI buried oxide layer. The reson
nators are released with a
wet etch to remove the exposed ox
xide layer. The complete
fabrication process has been descrribed in detail previously
[13]. SEM images of fabricated devices are shown in Fig. 3
d
with 27MHz and
and 4. Prior to testing, the two devices
54MHz resonant frequencies weree tested in air using a
network analyzer; their frequency responses are seen in Fig. 3
and 4 respectively.
0
Air Operation
f0 = 27.37MHz
Qunloaded = 23000
Rm = 262Ω
mass = 3.09μg
-20
SIMULATION
To evaluate the TPoS mass sensorr’s sensitivity, a
COMSOL Multiphysics structural mechannics simulation is
utilized. The eigen-frequencies of a single-ccrystal silicon slab
of dimensions 114.8μm by 76.5μm by 30μm
m with no support
tethers or loss mechanisms are examined. O
One micron cubic
nylon particles are randomly distributed acrross the sensor to
imitate deposited aerosol particles. The frequency of the
targeted mode shifts as a response to tthe added mass.
Through 75 particles, a simulated 54MHz maass sensor as seen
20
Figure 2: Simulated frequency shift from added
a
particles on a 54MHz
mass sensor. A meshed single-crystal silicon slab with eight particles
wn in the top right.
randomly distributed on the surface is show
|S12| (dB)
2
55.273
-40
-60
-80
27
28
29
Frequency (MHz)
(
Figure 3: SEM image and frequency responsse for a fabricated 27MHz TPoS
resonator.
239
25
26
VI.
-20
Aiir Operation
f0 = 53.99MHz
Quunloaded = 19000
Rm = 3.8kΩ
m
mass = 0.63μg
As expected, the frequency decrreases linearly due to the
additional mass loading as the partiicles are deposited on the
device, (Fig 6). The frequency respo
onses show an increase in
loss and therefore motional impeedance as more mass is
deposited (Fig. 7).
-60
Frequency(MHz)
-80
-100
50
52
54
56
58
Frequency (MHz)
Figure 4: SEM image and frequency response for a ffabricated 54MHz
TPoS resonator.
V.
27.2908
53.8925
27.2906
53.8920
53.8915
27.2904
53.8910
27.2902
53.8905
27.2900
53.8900
27.2898
53.8895
27.2896
53.8890
Frequency (MHz)
|S12| (dB)
-40
RESUL
LTS
0
20
40
Time(min)
Figure 6: Change in frequency as aerosol-d
dispersed particles are
deposited on the 27MHz and 54MHz reson
nators
CHARACTERIZATION
N
An aerosol particle generator, shown in F
Fig. 5, was used to
deposit particles with known size (and thereffore known mass)
onto the fabricated resonators. To form uniiform particles, in
the particle generator atomized droplets of m
methylene blue in
ethanol are pumped through a Kr-85 biipolar charger to
establish a near Boltzmann charge distributioon centered at zero
charge for the droplets. As the solvent driies, the remaining
solid spherical aerosol particles are injecteed into a column
with a large electrostatic field which defl
flects the moving
charged particles based on their mass andd charge. In this
manner particles are separated based on thheir mass and the
specific desired diameter, nominally 1μm
m for the tests
discussed in this paper, is selected.
-59
-61
dB
-63
10m
0m
20m
30m
-65
-67
-69
-71
-73
53.885 53.887 53.889 53.891 53.893
5
53.895 53.897
Frequency (M
MHz)
Figure 7: Frequency responses through tim
me as more particles are
collected on the resonator.
The high loss (relative to the meaasured frequency response
of Fig. 4) seen in the captured data of Fig. 7 is attributed to the
d to make the particle
vacuum interconnections required
measurements. SEM images of the 27MHz device after
testing can be seen in Fig. 8. Using this image, the number of
p
and therefore
particles are counted (~330 particles)
approximation for the mass depositeed on the sensor (Mp) can
be calculated using
(2)
Figure 5: Aerosol particle generator with the capabiliity to select
outgoing particle size
Particles are propelled and deposited onn the device under
test in a low vacuum environment (~50-1100Torr) inside a
glass bell jar. The device which has been wire-bonded to a
PCB is mounted in the tool and attached to a network analyzer
through vacuum feedthroughs. Prior to deposition, the particle
nozzle is aligned over the target device using a microscope
and a micropositioning stage. Frequenccy spectrums are
captured at regular intervals through the depoosition.
where N is the number of particlees, D is the approximate
diameter, and ρ is the density of thee aerosol particle resulting
in a calculated mass of 213pg. Likeewise, the same procedure
was performed on the 54MHz device (Fig. 9) resulting in 110
particles or a calculated mass of 70.8pg. The overall
frequency shifts seen in Fig. 6 are -904.5Hz and
-296.3Hz
for the 27MHz and 54MHz resonators respectively.
nsitivity for the 27MHz is
Therefore, the experimental mass sen
found to be -4.2Hz/pg while that of
o the 54MHz design is 41.9Hz/pg, a tenfold improvementt. The increase can be
attributed to the double in operating frequency and a five
times reduction in structural mass.
240
VII. CONCLUSSIONS
The thin-film piezoelectric-on-siilicon (TPoS) based mass
sensors shown in this paper offer low motional impedance
(<3kΩ), high Q (>18000), and high frequency (27MHz) with
mass resolution up to -4.2Hz/pg show
wn in this work. Through
doubling the frequency to 54MHz wh
hile maintaining a high Q,
the TPoS mass sensor shows roughlly a 75% reduction in the
minimum detectable mass and a teen times improvement (41.8Hz/pg) in experimental mass sensitivity
s
while agreeing
well with theoretical and simulated
d results. The completed
device has been shown to operate in air making the sensor
function with less power and in moree lenient packaging.
Figure 8: SEM image of the 27MHz mass sensor afteer exposure to
aerosol particles. This device has ~330 particles on it.
VIII. REFEREN
NCES
[1]
[2]
[3]
Figure 9: SEM image of 54 MHz mass sensor after exxposure to
aerosol particles. This device has ~110 particles on it.
Assuming the resonator’s compliance remains constant
during the deposition, the deposited mass ccan be calculated
using
[4]
[5]
(3)
where ∆f0 is the change in frequency, f0 is the resonant
frequency, ∆Meff is the change in mass, and Meff is the
effective mass of the resonator [11]. Since thhe deposited mass
is small and not affecting the mode shape, tthe mass ratio can
be written as the change in mass divided by total effective
mass. With (3), the expected mass depositeed on the 27MHz
resonator calculated from the -904.5Hz shiift is found to be
205pg, with a 3.6% error. For the 54MHz reesonator, -2963Hz
frequency shift translates into the expectedd mass of 69.2pg,
resulting in a theoretical error of ~2.3% and a simulation error
of ~5%.
To estimate the ratio of minimum detectable mass
(δM54MHz/ δM27MHz) using (1), the measureement bandwidth,
Δf, and the thermal energy, Eth, can be assum
med to be constant
between the two mass sensors. The result,
(4)
where V is the volume of the resonator and <
<x2> is the critical
displacement amplitude, shows that increasinng f.Q will reduce
the minimum detectable mass. Howeverr, measuring the
critical displacement amplitude is tricky forr TPoS resonators
so for the sake of simplicity it will be aassumed to scale
inversely and linearly with the frequenncy which is a
reasonable assumption in our case where thee device thickness
is the same. Using the experimental data collected and (4),
MHz design will
the minimum detectable mass for the 54M
roughly be one quarter of the 27MHz design..
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
241
W. D. Bowers and R. L. Chuan, "Surfface acoustic-wave piezoelectric
crystal aerosol mass microbalance." Reeview of Scientific Instruments,
Vol. 60, pp. 1297-1302, 1989.
P. A. Baron and K. Willeke, Aerrosol Measurement Principles,
Techniques, and Applications, 2nd Edition. New York : WileyInterscience, 2005.
nswers about the ozone layer, in
D. W. Fahey, “Twenty questions and an
scientific assessment of ozone depletion.” Global Ozone Research and
Monitoring Project Report No. 47. Geneva
G
: World Meteorological
Organization, 2003.
J.B. Kerr and G. Seckmeyer, “Surfacee ultraviolet radiation: past and
future, in scientific assessment of ozzone depletion.” Global Ozone
Research and Monitoring Project Reeport No. 47. Geneva : World
Meteorological Organization, 2003.
Z. Hao, R. Abdolvand, and F. Ayazi., "A high-Q length-extensional
bulk-mode mass sensor with annexed sensing platforms." Proc. 19th
IEEE International Conference on Miccro Electro Mechanical Systems
(MEMS2006). pp. 598-601, 2006.
A. Hajjam, J. C. Wilson, A. Rahafroo
oz, S. Pourkamali., "Fabrication
and characterization of resonant aerossol particle mass sensor." Proc.
23rd IEEE International Conference on Micro Electro Mechanical
010.
Systems (MEMS2010). pp. 863-866, 20
Z. J. Davis, G. Abadal, O. Kuhn, O. Hansen,
H
F. Grey, and A. Boisen,
“Fabrication and characterization of naanoresonating devices for mass
detection.” J. Vacuum Science Techno
ology B, vol. 18, no. 2, pp. 612616, 2000.
B. P. Harrington, M. Shahmohammadii, and R. Abdolvand., "Toward
ultimate performance in GHz MEMS resonators: low impedance and
high Q." Proc. 23rd IEEE Internationaal Conference on Micro Electro
Mechanical Systems (MEMS 2010). pp. 707-710, 2010.
on, R. Abdolvand, and F. Ayazi,
H. Mirilavasani, W. Pan, B. P. Harringto
“A 76dBΩ 1.7GHz 0.18μm CMOS Tun
nable Transimpedance Amplifier
Using Broadband Current Pre-Amplifiier for High Frequency Lateral
Micromechanical Oscillators.” Solid-Sttate Circuits Conference Digest
of Technical Papers (ISSCC 2010), p. 318-319, 2010.
L Roukes., "Ultimate limits to
K. L. Ekinci, Y. T. Yang, and M. L.
inertial mass sensing based upon nano
oelectromechanical systems." J.
Applied Physics, Issue 5, Vol. 95, pp. 26
682-2689, 2004.
J. Vig, and F. Walls, "A review of sensor sensitivity and stability,"
Proc. 2000 IEEE/EIA International Freq
quency Control Symposium and
Exhibition, p. 30–33, 2000.
R. Tabrizian, M. Rais-Zadeh, and F. Ayazi, "Effect of phonon
interactions on limiting the f.Q product of micromechanical
resonators." Proc. 15th International Con
nference on Solid-State Sensors,
Actuators, and Microsystems (TRANSD
DUCERS 2009), pp. 2131-2134,
2009.
B. P. Harrington and R. Abdolvaand, "Q-enhancement through
minimization of acoustic energy radiaation in micromachined lateralmode resonators." Proc. 15th Internatio
onal Conference on Solid-State
Sensors, Actuators, and Microsystemss (TRANSDUCERS 2009). pp.
700-703, 2009.