Mass Sensitive Thin Film Bulk Acoustic Wave

Resonators
M. Loschonsky, D. Eisele, L. M. Reindl
University of Freiburg – IMTEK
Department of Microsystems Engineering, Laboratory for Electrical Instrumentation, Germany

Abstract—The mass sensitive effect of RF-Filter technology

based Film and Solidly-mounted Acoustic Bulk Wave single
resonators (FBARs and SBARs) is investigated for
metrological use. For a high resolution of a sensor system all
systematical errors like cross temperature sensitivity must be
eliminated. A high quality factor Q is needed for stable
oscillating for minimizing statistical errors. Experiments using
FBAR and SBAR single resonators have been carried out in a Figure 1. Layout of FBAR and SBAR-type thin film resonators.
vapor deposition facility and thermal chamber to analyze mass 1:substrate, 2, FBAR:carrier mebrane, 2,SBAR: Bragg-reflector, 3: piezo
layer with lower and upper electrodes (black).Red arrows show
and temperature sensitive effects. The usage of more than one
longitudinal polarization [2],[3].
resonant mode with distinct polarization might allow the
compensation of cross sensitivities.
Combining the new technological features of FBARS &
SBARS with the well established metrological knowledge of
I. INTRODUCTION QCM, a new and exciting perspective of integrated high
precision MEMS based mass sensitive device is achieved.
Thin Film Bulk Acoustic Wave Resonators (FBARs) and
Solidly-Mounted Acoustic Wave Resonators (SBARs) have Since the mass of a thin film bulk acoustic wave
found many new and exciting applications in the field of resonator is many orders of magnitude smaller due to the
communication especially in cell phones within the last few reduced size of the resonator than the mass of a quartz micro
years, due to their excellent RF specifications and the cost balance, the corresponding frequency shift due to the mass
effective MEMS processibility (see fig. 1). Mainly FBARs & loading effect is also many orders of magnitude larger than
SBARs are designed to fulfill the requirements for an with a quartz micro balance. The constraints given by the
application in communication technique. A high electro- specification of modern communication systems, however,
mechanic coupling constant k2 of the piezo layer is needed to are quite different from the requirements of micro balance
allow operation in a large bandwidth with a low insertion sensors. Thus, an optimal micro balance sensor exhibits
attenuation. For steep-skirted filters a high quality factor Q is probably entirely different features than an optimal RF filter.
needed. The temperature coefficient should be rather low to In our contribution we consider the suitability of FBARs and
guarantee the steep transition between the pass band and the SBARs particularly designed for mass sensitive applications,
stop band of filters for any usage in modern cell phones. The while comparing fundamental behavior of both types of
filter must show one single pass band and is decoupled to all resonators.
environmental influence by a proper mounting and housing
For metrological use of the quartz-microbalance like
technique.
mass sensitive effect of FBARs & SBARs, new aspects and
In filter technique FBARs & SBARS represent the new design targets are discussed which sensor resonators
technological enhancements of the well-known classical have to perform.
quartz resonator. Since the middle of the 1950’s quartz slabs
and few years later quartz resonators have been investigated II. THEORETICAL ASPECTS OF MASS RESOLUTION
for micro balancing. Quartz Crystal Microbalances (QCM)
have been used very successfully for more than forty years in Quartz Crystal Microbalances (QCM) differ from thin
process technology [1] and in chemical, biological and film bulk acoustic wave resonators in two main aspects: the
medical analysis systems as high resolution mass sensitive geometrical size of the oscillating mass and, therewith
device. Precious balancing demand primarily an accurate connected, the resonance frequency depends on the thickness
temperature compensated material together with a high of the piezoelectric layer.
quality factor of the resonator. A further aspect is an
additional interface to a sensitive layer for enabling
selectivity.

We acknowledge Agilent Technologies, Avago Technologies Inc. and

EPCOS AG for precious support.

1-4244-0074-0/06/$20.00 © 2006 IEEE. 111

Authorized licensed use limited to: Central Electronics Engineering Research Institute. Downloaded on September 01,2023 at 09:44:33 UTC from IEEE Xplore. Restrictions apply.
 A. Mass Sensitivity of Acoustic Resonators resonator and the electrical circuitry, and aging. A smaller,
The Sauerbrey equation (1) describes the relation but also detectable effect is caused by a varying pressure.
between a fractional increase in mass and the frequency shift
due to the mass loading. B. Theoretical Frequency Resolution
The accuracy of detecting a frequency shift due to mass
∆m ∆f
− = ⇔ ∆f ~ ∆m ⋅ f 2 loading on the top of a resonator surface is primarily based
m f (1) on the technical measurement equipment and the methods of
signal processing. Modern definitions of signal processing
If the oscillator is operated on the nth harmonic, n·f0, the include statistical estimation theories [4] which allow best
frequency shift ∆fn is n times the frequency shift of the estimated parameter values with highest certainty.
fundamental ∆f0:
For parameter estimation of a sinusoidal signal, as described
∆f n = n ⋅ ∆f 0 (2) in [6], the theory of the Cramer-Rao-Lower-Bound can be
applied to estimate amplitude, phase and frequency of a
In eq. 1, both the mass of the resonator and the added mass sinusoidal oscillating signal.
are usually treated as areal mass densities, that is, mass per Considering a potential mass sensitive metrological system
unit area. If the film coated onto the crystal is laterally based on an oscillator controlling thin film bulk acoustic
homogeneous, the difference is unessential. Since a large wave resonator with following system parameters: 7ng
crystal’s own mass is large it is less sensitive to the absolute oscillating resonator mass; 1.88 GHz of resonance
mass (in grams). However, the mass of a laterally frequency; 20 dB Signal-to-Noise-Ratio (SNR); 2500
homogenous film (in grams) coated on top of this large number of points (N) at a sampling time of 1s (tA).
crystal increases in proportion to the area of the crystal as
well. Therefore the sensitivity of the QCM width scales with The frequency estimation (3) of a sinusoidal signal
the square of the frequency, irrespective of its lateral size. embedded in White Gaussian Noise is hence
The situation is different if the target is not sensing of a film 6
thickness, but sensing of a small mass in an absolute sense. var( fˆ0 ) ≥ (3)
4π 2 ⋅ SNR ⋅ N ⋅ ( N 2 − 1)
Such a mass could, for instance, be the mass of a single
nanoparticle. In this case, it is advantageous to make the 6
resonating structure as small as possible in both the lateral var( fˆ0 ) ≥ (4)
and the vertical dimension. Using a small crystal decreases 4π 2 ⋅ 100 ⋅ 2500 ⋅ ( 2500 2 − 1)
the mass and increases the sensitivity accordingly. Thickness
shear resonators need to have an aspect ratio (width to It follows from the above for (1):
height) of about 30. Otherwise, the distortions of the shear
wave induced by the finite lateral extensions become too ∆f
strong. = - 0,3056 ⋅ 10 −12
f (5)
Under these circumstances one is rewarded twice for
choosing a very thin resonator. Firstly, thin resonators have a The detectable minimum loaded mass is therefore expressed
high frequency which enters into the Sauerbrey equation. by the Cramer-Rao-Lower-Bound.
Secondly, a thin resonator can also have a small width. The
minimum lateral extension scales as the thickness, which
∆mCRLB = 2,1392 ⋅ 10−21 g (6)
means that the minimum area scales as the square of the
thickness. Taking all these factors into account, one finds
that the sensitivity of the QCM with regard to the According to [Kay], the Cramer-Rao-Lower-Bound for the
determination of absolute masses (not areal mass densities) frequency decreases as the SNR increases and as the bound
scales as the fourth power of f. decreases to 1/N3, making it quite sensitive to data record
length.
The actual resolution will be due to the minimum resolvable
frequency shift. Mainly, there are two limiting factors for C. Temperature Cross Sensitivity
this resolution: The first is due to limited quality factor of the
The temperature sensitivity is the dominant source of
resonator which results in a phase noise in an implemented systematic error. This cross sensitivity can be minimized by
oscillator circuitry. This determines the signal-to-noise ratio
inherent material properties, as in Quartz, by a composite
of the sensor signal and results as a statistical error in a arrangement of materials with positive and negative
limited frequency resolution of the frequency estimator.
temperature coefficients [3], and numerically. For the
The second limiting factor is the systematic error of the set- numerical compensation a resonator with at least two
up caused by remaining cross sensitivities of environmental dominant modes is needed. The applied oscillator circuitry
influences, mainly the temperature coefficient of the has to resonate on both frequencies simultaneously. If the
ratio of the mass sensitivity of both modes is different to the

112

Authorized licensed use limited to: Central Electronics Engineering Research Institute. Downloaded on September 01,2023 at 09:44:33 UTC from IEEE Xplore. Restrictions apply.
 ratio of the temperature sensitivities, a numerical or
electronic compensation of the temperature cross sensitivity
can be achieved [6].

III. EXPERIMENTS
Both types of resonators, FBAR and SBAR, where used
in the experimental investigations in the frequency range
from 500MHz to 8.5 GHz. The fundamental resonance
frequency of the FBAR was 886 MHz. Additionally, the
FBAR showed three distinct parasitic resonances at 2.3, 4,
and 5.1 GHz. These frequencies are near the 3rd, 5th and 7th
harmonic of the fundamental, but clearly too different from
them to depend on dispersion of material properties alone.
Thus, the polarization of the parasitic resonances should be Figure 3. left-side: Front view of a vapor deposition facility type Leybold
considerable different from the fundamental which results in Univex 500, right-side: Measurement Set-up with Agilent E5071B.
a different mass sensitivity. The fundamental resonance
frequency of the SBAR was 1.9 GHz. In the considered In Fig. 5 the fundamental and the 3rd harmonic can be seen as
frequency range, one distinct higher parasitic mode was sharp resonances, next to some other smaller anharmonics.
detectable, at the 3rd harmonic frequency 5.7 GHz. The Q of these resonances is lower when compared to the
FBAR. This is due to the higher fundamental frequency of
A. Experimental Set-Up the SBAR which also results in a lower thickness of the
electrode metallization and thus in higher ohmic losses.
The measurement set-up was built up using an Agilent
E5071B network-analyzer with 20 001 measurement points Both figures show, additional to the acoustic resonances,
in the frequency range from 500 MHz to 8.5 GHz. The some high-Q electro-magnetic resonant modes of the
resonators were mounted on a metallic TO-39 chip carrier, vacuum recipient. At frequencies over 5 GHz the amplitude
wired to the network analyzer (see fig. 2), and placed into a of S11 is larger than one, which means that the corresponding
vapor deposition facility as shown in fig. 3. The temperature circuitry is considered as active. This is due to problems in
measurements were carried out using a WTC Binder MK 53 the calibration procedure. The calibration was done by using
thermal chamber. a Calibration Kit with SMA connectors which, of course,
avoids any coupling to the resonant electro-magnetic modes
Fig. 4 and fig. 5 show typical amplitudes of the reflectivities
of the vacuum recipient. In the measurements the bond wire
|S11| of a FBAR and a SBAR, respectively, in the frequency couples increasingly to these electro-magnetic modes with
range of 500 MHz to 8.5 GHz measured with 20 001 points. higher frequencies. These electro-magnetic modes
In Fig. 4 the fundamental can be seen on the left side as a deteriorate also the frequency accuracy of the investigated
strong and sharp resonance. The parasitic modes appear at acoustic modes.
higher frequencies near the 3rd, 5th and 7th multiple of the
fundamental. At 4 GHz a low-Q resonant mode is
superimposed to the acoustic mode. 10

0
|S11| in dB

-5

-10

-15

-20

-25
2 4 6 8

Figure 2. Bottom view of single resonator mount in a vapor deposition frequency in GHz
facility type Leybold Univex 500.

Figure 4. Frequency response |S11 | of a FBAR.

113

Authorized licensed use limited to: Central Electronics Engineering Research Institute. Downloaded on September 01,2023 at 09:44:33 UTC from IEEE Xplore. Restrictions apply.
 10 867,6

sensitivity: -40 ppm/Å

5 867,4
measured data
linear fit

frequency in MHz
0 867,2
|S11| in dB

-5 867,0

-10 866,8

-15 866,6

-20 866,4

-25 866,2
2 4 6 8 0 5 10 15 20 25 30

frequency in GHz thickness in Å

Figure 5. Frequency response |S11 | of a SBAR. Figure 6. Mass Sensitivity of a FBAR for the fundamental mode.

B. Mass Sensitive Effect 2,300

For the evaluation of the mass sensitivity, Chrome was sensitivity: -62 ppm/Å
evaporated to total a thickness of 3 nm in 0.3 nm steps on 2,299
measured data
both types of resonators. A calibrated thermocouple was linear fit

frequency in GHz
2,298
used to control the temperature cross sensitivity and after
each evaporation step the frequency response was recorded 2,297

at a pressure of 6·10-6 mBar and at a operation temperature of

24°C. The investigated resonance frequencies where 2,296

extracted from the measured data by accurate determination 2,295

of the minima in the amplitude of the frequency response.
The frequency shift due to the variation of mass per area 2,294

m/A was fitted by a linear regression line to the

measurement points. The mass sensitivity is giving by the 2,293
0 5 10 15 20 25 30
slope of the linear regression line. thickness in Å
1) Mass Sensitivity for FBARs
Fig. 6-9 show the frequency shift of a FBAR for the Figure 7. Mass Sensitivity of a FBAR for the 1st parasitic mode near the
fundamental and the three parasitic higher modes. The mass 3rd harmonic frequency.
sensitivity for fundamental and the parasitic resonances in
the higher frequency range -40, -62, -85, and -170 ppm/Ǻ,
respectively. The sensitivities do not follow the eq. 2 which 4,010

might be explained by the different polarization of the 4,008

sensitivity: -85 ppm/Å

modes. The deviation of the measured data to the linear measured data
linear fit
regression line is due to the interference of the resonance
frequency in GHz

4,006

peaks with some high-Q electro-magnetic resonant modes of

the vacuum recipient. 4,004

2) Solidly-mounted Film Bulk Acoustic Wave Resonator 4,002

Fig. 10-11 show the frequency shift of a SBAR for the 4,000
fundamental and the 3rd harmonic. The mass sensitivities are
-127 and -375 ppm/Ǻ in well agreement with eq. 2. 3,998

3,996
0 5 10 15 20 25 30

thickness in Å

Figure 8. Mass Sensitivity of a FBAR for the 2 nd parasitic mode near the
5th harmonic frequency.

114

Authorized licensed use limited to: Central Electronics Engineering Research Institute. Downloaded on September 01,2023 at 09:44:33 UTC from IEEE Xplore. Restrictions apply.
 5,175G
Fig. 16 and 17 show the temperature sensitivity of the
sensitivity: -170 ppm/Å
5,170G investigated SBAR device is for the fundamental and the 3rd
measured data
linear fit harmonic which are -26 and -80 ppm/°C, respectively. These
two values fulfill an analogue equation to eq. 2 for
frequency in GHz

5,165G

temperature sensitivity. The ratio of the sensitivities for the

5,160G
fundamental and the 3rd harmonic on a mass load and on a
5,155G temperature load is relatively close to each other, but still
allows a numerical temperature compensation (see Tab. I).
5,150G 844,5

sensitivity: -28 ppm/°C

5,145G 844,0
measured data
linear fit

frequency in MHz
5,140G 843,5
0 5 10 15 20 25 30

thickness in Å 843,0

Figure 9. Mass Sensitivity of a FBAR for the 3rd parasitic mode near the
842,5
7th harmonic frequency.
842,0
1,914

sensitivity: -127 ppm/Å 841,5

1,912
measured data
linear fit 841,0
frequency in GHz

1,910
0 20 40 60 80 100

1,908
temperature in °C

1,906 Figure 12. Temperature Sensitivity of of a FBAR for the 1st

harmonic.
1,904 3,914

sensitivity: -33 ppm/°C

1,902 3,912
measured data
linear fit
frequency in GHz

1,900 3,910
0 5 10 15 20 25 30

thickness in Å 3,908

st
Figure 10. Mass Sensitivity of a SBAR.of the 1 harmonic. 3,906

5,820
3,904
sensitivity: -375 ppm/Å
5,810
measured data 3,902
5,800 linear fit
frequency in GHz

3,900
5,790 0 20 40 60 80 100

temperature in °C
5,780

5,770
Figure 13. Temperature Sensitivity of a FBAR for the 1st parasitic mode
5,760
near the 3rd harmonic frequency.
2,228
5,750
sensitivity: -38 ppm/°C
5,740 2,226
measured data
0 5 10 15 20 25 30
linear fit
frequency in GHz

thickness in Å 2,224

Figure 11. Mass Sensitivity of a SBAR of the 3rd harmonic. 2,222

2,220
C. Temperature Sensitve Effect
The measured temperature sensitivity of the investigated 2,218

FBAR device is for the fundamental and the three higher 2,216
anharmonics near the 3rd, 5th and 7th harmonic frequencies
more or less constant in the range of 28-38 ppm/°C (see 2,214

fig.12 to 15). This amazing fact might also be explained by 0 20 40 60 80 100

the different polarizations of the parasitic modes at higher temperature in °C

frequencies. This fact shows again that these higher parasitic Figure 14. Temperature Sensitivity of a FBAR for the 2nd parasitic mode
modes are definitely different from the higher harmonics. near the 5th harmonic frequency.

115

Authorized licensed use limited to: Central Electronics Engineering Research Institute. Downloaded on September 01,2023 at 09:44:33 UTC from IEEE Xplore. Restrictions apply.
 5,163
Since these two sensitivities are quite different to each other
sensitivity: -33 ppm/°C
5,160 such a device is well suited for any temperature
measured data
linear fit compensation.
frequency in GHz

5,157
Tab. II shows the corresponding sensitivities of the
5,154 fundamental and the 3rd harmonic of a SBAR on mass and
temperature loading. Since the ratio of the mass sensitivities
5,151
to the ratio of the temperature sensitivities the corresponding
5,148
numerical equation is ill-conditioned.
FBAR Resonance frequencies
5,145
Frequency
Mass
5,142 -40 ppm/Ǻ -62 ppm/Ǻ - 85ppm/Ǻ -170ppm/Ǻ
0 20 40 60 80 100 sensitivity
temperature in °C Temperature
-28 ppm/°C -38 ppm/°C -33 ppm/°C -33 ppm/°C
sensitivity
Ratio of
Figure 15. Temperature Sensitivity of a FBAR for the 3rd parasitic mode both 1,429 1,632 2,576 5,152
near the 7th harmonic frequency. sensitivities

1,895
TABLE I. MASS AND TEMPERATURE SENITIVITY OF THE
sensitivity: -26 ppm/°C INVETSIGATED FBAR.
1,894
measured data
linear fit SBAR Resonance frequencies
frequency in GHz

1,893 Frequency 1,91-1,912 GHz 5,74-5,81 GHz

Mass sensitivity -127 ppm/Ǻ -375 ppm/Ǻ
1,892
Temperature sensitivity -26 ppm/°C -80 ppm/°C
Ratio of both sensitivities 4,885 4,688
1,891

1,890 TABLE II. MASS AND TEMPERATURE SENITIVITY OF THE

INVETSIGATED SBAR.
1,889
In the next step FBAR and SBAR devices have to be
1,888
designed and processed especially suited for mass sensitive
0 20 40 60 80 100 applications. This implies at least two high-Q resonant
temperature in °C modes with distinct different mass and temperature
sensitivities. These new types of resonators have to be
Figure 16. Temperature Sensitivity of the 1st harmonic of a SBAR. integrated into a double-mode oscillator circuitry.
5,770

sensitivity: -80 ppm/°C ACKNOWLEDGMENT

5,760
measured data
linear fit
R. B. G. thanks Dr. Armin Dadgar and Prof. Dr. Alois
Krost from Otto-von-Guericke-University in Magdeburg,
frequency in GHz

5,750

Germany, Prof. Dr. Franz Luwig Dickert from University of

5,740
Vienna, Prof. Dr. Diethelm Johannsmann from TU Clausthal
5,730
and Dr. Claas Müller, University of Freiburg, Germany for
scientific discussions.
5,720

5,710
REFERENCES
[1] G. Sauerbrey, Verwendung von Schwingquarzen zur Wägung dünner
5,700 Schichten und zur Mikrowägung, Zeitschrift für Physik 155, pp206-
0 20 40 60 80 100
222, 1959
temperature in °C [2] K. M. Lakin, J. Wang, Acoustic Bulk Wave Composite Resonators,
Figure 17. Temperature Sensitivity of the 3rd harmonic of a SBAR. Appl. Phys. Let., 1981, 39, pp. 125-127.Y.
[3] K. M. Lakin, Thin Film Resonator Technology, IEEE Trans Ultrason
Ferroelectr Freq Control. 2005, 52, pp 707-716
IV. CONCLUSION AND OUTLOOK [4] S. Kullback, Information Theory and Statistics, Dover Publications,
Thin film bulk acoustic wave resonators of FBAR and 1959
SBAR type have been investigated for a metrological mass [5] S. Kay, Fundamentals of Statistical Signal Processing Vol I & II,
Prentice Hall, 1993
sensitive application.
[6] S. Schodowski, Resonator Self-Temperature-Sensing Using a Dual-
Tab. I shows the sensitivities of the fundamental and the Harmonic-Mode Crystal Oscillator, Proc. 43rd Annual Symposium
three higher anharmonics near the 3rd, 5th and 7th harmonic on Frequency Control, pp. 2-7, 1989
frequencies of a FBAR on mass and temperature loading.
116

Authorized licensed use limited to: Central Electronics Engineering Research Institute. Downloaded on September 01,2023 at 09:44:33 UTC from IEEE Xplore. Restrictions apply.