Content uploaded by Vasiliki Giagka
Author content
All content in this area was uploaded by Vasiliki Giagka on Sep 07, 2023
Content may be subject to copyright.
A Comparative Study of Si3N4and Al2O3as
Dielectric Materials for Pre-Charged Collapse-Mode
CMUTs
Marta Saccher∗, Rob van Schaijk§, Shinnosuke Kawasaki‡, Johan H. Klootwijk§,
Amin Rashidi∗, Vasiliki Giagka∗†, Alessandro Stuart Savoia¶and Ronald Dekker∗§
∗Department of Microelectronics, Delft University of Technology, Delft, the Netherlands
†Department of System Integration and Interconnection Technologies, Fraunhofer IZM, Berlin, Germany
‡TNO/Holst Centre, Eindhoven, the Netherlands
§MEMS & Micro Devices, Philips, Eindhoven, the Netherlands
¶Department of Industrial, Electronic, and Mechanical Engineering, Roma Tre University, Rome, Italy
Abstract—Capacitive Micromachined Ultrasound Transducers
(CMUTs) have many advantages compared to other ultrasonic
transducer technologies, especially for implantable devices. How-
ever, they require a high bias voltage for efficient operation.
To eliminate the need for an external bias voltage, a charge
storage layer can be embedded in the dielectric. This study aims
to compare the performance of Si3N4and Al2O3when used
as a charge storage layer. By measuring the shift in the C-V
curve, Si3N4exhibits a larger shift than Al2O3, indicating a better
charge-trapping capability. When using the pre-charged CMUTs
as power receivers, the Si3N4version harvested up to 80 mW -
only a few mW more than the Al2O3- with an efficiency of about
50 %. Accelerated Lifetime Tests predict a lifetime of about 7.8
and 1.2 years for Si3N4and Al2O3respectively.
Index Terms—pre-charged CMUT, Capacitive Micromachined
Ultrasound Transducers, zero-bias transducers, ultrasonic power
transfer
I. INTRODUCTION
Capacitive Micromachined Ultrasonic Transducers
(CMUTs) have gained increasing attention as power
receivers for biomedical implants due to their manufacturing
scalability and biocompatibility. CMUTs can be operated
in two modes: non-collapse-mode, and collapse-mode. The
collapse-mode occurs when the top membrane makes partial
contact with the dielectric layer above the bottom electrode,
achieved by applying a high DC voltage that causes the top
membrane to snap down. The bias voltage typically ranges
from 30 V to 70 V, resulting in enhanced transmit and receive
sensitivity, and an output pressure up to three times greater
compared to conventional mode [1], [2]. This increase is due
to the higher electric field in the vacuum cavity caused by the
high bias voltage and smaller effective gap height. However,
such high DC biases are better avoided in the body.
To address the issue of external bias voltage, a charge
storage layer, such as Al2O3, can be incorporated into the
dielectric stack. By trapping a sufficient amount of charge in
this layer, a built-in bias is created, ensuring that the CMUT
This work was funded by the ECSEL Joint Undertaking project
Moore4Medical, grant number H2020-ECSEL-2019IA-876190.
(a) (b)
Bottom electrode (Al)
Substrate (Si)
Dielectric
(SiO2)
Top electrode (Al)
Membrane layer (Si3N4)
Charge layer
(Al2O3/Si3N4)
Fig. 1. (a) Schematic of collapse-mode CMUT with charge trapping layer in
the dielectric. (b) Microscope view of the fabricated CMUT array.
remains in permanent collapse-mode [3], [4]. Besides Al2O3,
other materials capable of retaining enough charge can also be
employed. Previous studies have utilized pre-charged CMUTs
with Si3N4or SiO2in the dielectric [2], [5]–[7]. Park et al.
[7] and Choi et al. [5] have demonstrated the feasibility of
pre-charged CMUTs with Si3N4as the dielectric, exhibiting
good charge retention over time, even at high temperatures.
However, the reported pre-charged CMUTs are mostly focused
on non-collapse-mode operation, with the exception of Ho et
al. [6]. It is evident from these studies that trapping sufficient
charge to achieve a pre-charged collapse-mode CMUT is
challenging, and the choice of material for the charge-trapping
layer is crucial. Choi et al. [5] compared the performance of
Si3N4-SiO2and SiO2as charging layers, concluding that Si3N4
maintains its performance for a longer duration due to superior
charge retention. The objective of this study is to compare
the performance of Si3N4and Al2O3as charge storage layers
in the dielectric of collapse-mode CMUTs. Additionally, the
performance of CMUTs as power receivers is evaluated.
II. DEVICE PREPARATION AND CHARACTERIZATION
A. CMUT fabrication
The CMUTs utilized in this study are circular cells with
a diameter of 355 µm, fabricated using a standard sacrificial
release processing technique. The dielectric layer consists of a
stack of SiO2\charge layer\SiO2, a vacuum gap, and an addi-
tional layer of SiO2. Two versions of CMUTs were fabricated
and analyzed: the first version has an ALD-deposited Al2O3
charge layer, while the second has a PECVD-deposited Si3N4
charge layer. Both variants have a charge layer thickness of
200 nm, and the Equivalent Oxide Thickness (EOT) measures
320 nm and 350 nm for the Al2O3and Si3N4variant, re-
spectively. A 5×5 mm2single-element transducer, consisting
of 173 CMUT cells electrically connected in parallel, was
fabricated. Figure 1 depicts the schematic cross-section of the
CMUT, and a microscope photograph of the fabricated device.
B. CMUT pre-charging and characterization
In a previous study, we established that by applying an
electric field above 7−8 MV/cm for a few minutes, a stable
pre-charged CMUT is produced [4]. In this study, we investi-
gated the impact of charging the devices with varying electric
fields. We selected charging voltages corresponding to electric
fields ranging from 7.5 to 8.5 MV/cm, both with positive
and negative polarity with respect to the bottom electrode.
Each charging voltage was applied to a pristine device for 5
minutes. Subsequently, we measured the Capacitance-Voltage
(C-V) curves to assess the shift in collapse voltage resulting
from the charge stored in the dielectric. To minimize any
alteration of the stored charge in the device, we employed
a fast C-V protocol, with a voltage sweep rate of 80 V/ms.
For the C-V measurement, a small AC signal at 125 MHz was
superimposed on a 200 Hz triangular wave varying between
0 V and ±120 V. The change in the phase of the impedance
when the CMUT goes in and out of collapse is used to draw
the C-V curve. Two measurement protocols were used: a
DC bipolar sweep between positive and negative voltages for
uncharged devices, and a DC unipolar sweep for either only
positive or negative voltages, for devices charged with positive
and negative polarity, respectively.
III. RESULTS AND DISCUSSION
A. Comparison of charge trapping
In Fig. 2(a), we compare the C-V curves of the two CMUT
variants when charged with an electric field of 8 MV/cm with
positive or negative polarity, to that of corresponding pristine
devices. The C-V curve of both pristine devices exhibits
symmetry around 0 V, indicating that the CMUT is not in
a collapsed state and has no charge stored in the dielectric.
Conversely, the pre-charged devices show a shift in the point
of minimum capacitance of 60 to 80 V, indicating that they are
in a collapsed state even at 0 V bias, and the presence of charge
stored in the dielectric. Furthermore, we observe that the Si3N4
variant exhibits a larger shift in the C-V curve compared to
Al2O3for both polarities (Fig. 2(a)). This could be due to the
different deposition method (ALD vs PECVD) and not (only)
the different dielectric material of the two devices, and will
be investigated in the future. Additionally, charging the device
with a negative polarity results in a slightly larger shift in the
C-V curve compared to positive polarity.
Figures 2(b) and (c) depict the impact of different charg-
ing electric fields on the shift in the C-V curves for the
Si3N4and Al2O3variants, respectively. As expected, the
shift becomes more significant for higher charging electric
fields due to increased current tunneling through the CMUT
dielectric. Moreover, charging Al2O3devices with an electric
field exceeding 8 MV/cm was only feasible with negative
polarity, as devices charged with positive polarity experienced
dielectric breakdown. In contrast, the Si3N4variant tolerated
charging electric fields up to 8.5 MV/cm without encountering
any breakdown, regardless of polarity. Although the observed
difference in breakdown points between positive and negative
polarity was not evident for Si3N4, it is possible that the
difference exists but is smaller compared to Al2O3. This
can be attributed to the differing charging mechanisms and
characteristics of the two dielectric materials.
Charging |Efield| [MV/cm]
uncharged
7.5
8
8.5
-150-100 -50
0
50 100 150 200-200
0.8
0.9
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Si3N4Al2O3
Al2O3
Si3N4
+ charging
uncharged
- charging
+ charging
- charging
Bias voltage [V] Bias voltage [V] Bias voltage [V]
-150-100 -50
0
50 100 150 200-200
Capacitance [a.u.]
Capacitance [a.u.]
Capacitance [a.u.]
0.8
0.9
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
C-V shift
-150-100 -50
0
50 100 150 200-200
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Charging |Efield| [MV/cm]
uncharged
7.5
8
8.25
+ charging
- charging
(a) (b) (c)
Fig. 2. (a) Comparison of C-V curve shift between Si3N4and Al2O3samples when charged with ±8 MV/cm electric field. (b) Comparison of C-V curve
shift for Si3N4samples when charged with different electric fields and polarity. (c) Same as (b) but for Al2O3.
0 2 4 6 8 10 12 14
Frequency [MHz]
70
60
50
40
30
20
10
0
Re [Ω]
Im [Ω]
0
-200
-400
-600
-800
-1000
-1200
-1400
-1600
R
Si3N4
Al2O3
X
3.6 Ω
395 Ω
497 Ω
250 pF
260 pF
730 pF
730 pF
23 µH
28 µH
3.6
Ω
Fig. 3. Electrical impedance of the pre-charged collapse-mode CMUTs
measured in water and corresponding BVD model values for the two device
variants.
B. Pre-charged CMUT electro-mechanical characteristics
To evaluate the pre-charged collapse-mode CMUTs as wire-
less power receivers, we chose to compare devices charged
with an electric field of −8 MV/cm as this was the highest
electric field with which both variants could be charged.
The CMUTs were wire-bonded to a PCB and coated with
a thin layer of PBR (polybutadiene rubber). After charging
the CMUTs, the shift in the C-V curve was 77 V and 68 V
for the Si3N4and Al2O3variants, respectively, which can be
considered as a built-in DC bias.
After charging, we measured their impedance in water with
an impedance analyzer (Keithley E4990A, Keysight Tech-
nologies), and extracted the values of the Butterworth-Van
Dyke (BVD) model components (Fig. 3). Both devices have
a center frequency of around 2 MHz in water, and a typical
bandwidth between 105 % and 120 %. Additionally, the BVD
model incorporates a 3.6 Ω parasitic resistance connected in
series with the CMUT model. This resistance arises from the
electrical interconnects between the CMUT array and the wire-
bonding pads.
C. Maximum power transfer
To evaluate the pre-charged collapse-mode CMUTs as ul-
trasound power receivers, we conducted a power transfer
experiment. A lower power carrier frequency is favorable in
ultrasonic power delivery since it results in lower propagation
losses for ultrasonic waves traveling through the tissue. There-
fore, for this study, we chose a carrier frequency of 1 MHz.
We used a circular single-element PZT piston transducer with
a center frequency of 1 MHz and a diameter of 39 mm as the
transmitting (TX) transducer. The receiving (RX) transducers
were the two pre-charged CMUT variants. Both the TX and
RX transducers were immersed in a water tank and aligned
at the center, with a distance of 22 cm between them, which
corresponds to the TX’s natural focus distance. The pressure
field generated by the TX transducer at this distance was
previously measured by scanning a plane using a needle
hydrophone parallel to the surface of the TX (Fig. 4(a)).
To maximize the power transfer efficiency, a matching
network was connected to the CMUTs. In this study, we
opted for a parallel matching approach, where the value of
the optimal load is the conjugate of the measured impedance
at the chosen frequency. At 1 MHz, the BVD model of the two
CMUTs devices can be simplified with a resistor in parallel
with a capacitor. According to the maximum power transfer
theorem, the optimal load connected in parallel to the CMUTs,
consists of an inductor (Lload) and a resistor (Rload ). The
optimal inductor compensates for the equivalent capacitance of
the BVD model, while the resistor should have the same value
as the equivalent resistor of the BVD model. We determined
the value of the matching inductor by testing components
with values close to the conjugate of the equivalent electrical
capacitance of the CMUT variants. A potentiometer was then
used to find the optimal resistance value for maximizing the
received power (Fig. 4(b)).
We applied a sine burst signal with an amplitude of 70 V
and 10 cycles at a pulse repetition frequency of 1 kHz to
the TX transducer, resulting in an instant acoustic intensity
of 720 mW/cm2averaged across the CMUT surface. This
intensity adheres to the FDA safety limit for the use of ultra-
sound in the body. The power at Rload was calculated using
P = V2
pp/(8 ·Rload ), where Vpp is the peak-to-peak voltage at
the load. The maximum harvested power was 80 mW with the
Si3N4CMUT and 77 mW with the Al2O3variant. The inset
in Fig. 4(b) shows the corresponding acoustic-to-electrical
power conversion efficiency of the CMUTs, calculated as
the ratio between the acoustic power at the CMUT surface
(instant acoustic intensity ×CMUT area) and the maximum
harvested power, which reaches approximately 50 %. The
marginal difference in the maximum harvested power between
the two devices can be attributed to an imperfect load matching
for the Al2O3version, primarily due to the limitations in
available inductor values. However, it is important to note that
this discrepancy is not significant enough to give preference
to one material over the other.
Considering the integration of the CMUT devices into
an ASIC, the requirement for a matching inductor poses
a significant challenge, especially when inductors of tens
of µHare needed, as in this case. To address this issue,
we evaluated the power conversion efficiency with a purely
resistive load by repeating the previous experiment. In this
case, the load value needs to be matched to the modulus
of the impedance of the simplified BVD model. Under these
conditions, the maximum harvested power was 35 mW with
the Si3N4variant and 30 mW with the Al2O3variant (Fig.
4(c)). The corresponding power conversion efficiency was
approximately 20 %. Table I summarizes the data obtained
from the power transfer experiment.
D. Lifetime
To estimate the lifetime of the devices, Accelerated Lifetime
Tests (ALT) have been performed with temperature as an
accelerating factor. We placed the pre-charged CMUTs on a
hotplate set at temperatures ranging from 140 ◦Cto 160 ◦C,
and the impedance was measured at regular time intervals. The
time to failure at different test temperatures, defined as when
more than half of the CMUT membranes are out of collapse-
(a) (b) (c)
200
220
240
260
280
300
Peak-to-peak pressure [kPa]
-4
-3
-2
-1
0
1
2
3
4
-4 -3 -2 -1 0 1 2 3 4
RXx [mm]
RXy [mm]
30
40
50
60
70
80
90
Received power [mW]
Received power [mW]
5
10
15
20
25
30
35
40
Vpp [V]
Vpp [V]
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 10.5 1.5
Rload [kΩ]Ronly_load [kΩ]
10
0 1
R [kΩ]
η [%]
2 3 4 5
20
30
40
50
5
10
15
20
25
30
35
40
2
3
4
5
6
7
8
9
10
11
12
0 0.5 1 1.5
R [kΩ]
50
40
30
20
10
η [%]
Al2O3
Si3N4
Al2O3, Lload=32 µH
Si3N4, Lload=32 µH
Si3N4
Al2O3
Si3N4
Al2O3
Fig. 4. (a) Pressure field at the location of the RX CMUT. The dashed line indicates the position of the CMUT. (b) and (c) Measurements with and without
the matching inductor, respectively; power at Rload, and corresponding peak-to-peak voltage at load, for the two CMUT variants. Figure insets indicate the
acoustic to electrical power conversion efficiency.
TABLE I
SUM MARY O F POW ER TR AN SFE R EX PER IM ENT
Req[Ω] Ceq [pF]Lload [µH] Rload [Ω]Ronly load [Ω]
Si3N4
900 920 32 1000 200
Power 80 35
η[%] 50 22
Al2O3
800 880 32 860 150
Power 77 31
η[%] 48 19
mode, was fitted using a previously defined method [4] to
estimate the lifetime of the two CMUT variants at 37 ◦C(Fig.
5). The predicted lifetime of the Si3N4variant is about 7.8
years, and 1.2 years for the Al2O3variant. It should be noted
that, although these ALT conditions may have accelerated the
failure mechanisms, further tests are necessary to evaluate
their lifetime under actual operational conditions. Nonetheless,
these results demonstrate the potential of these devices.
IV. CONCLUSION
In this study, the performance of CMUTs with Si3N4and
Al2O3charge storage layers was compared in terms of charge
2.25 2.3 2.35 2.4 2.45 2.5
10⁴
10⁵
Time to failure [s]
Temperature [1000/K]
Si3N4
Al2O3
Fig. 5. Fitting of the ALT results to predict the CMUT lifetime.
trapping and power conversion efficiency. It was observed that,
under the same charging electric field, a higher amount of
charge could be stored in the Si3N4variant compared to the
Al2O3variant. Furthermore, the Si3N4variant demonstrated
slightly higher power conversion efficiency than Al2O3, with
both variants achieving approximately 50 % efficiency for a
matched load and around 20 % efficiency for a purely resis-
tive load, indicating promising potentials for their integration
into an ASIC to produce a single chip implantable medical
device. Moreover, the estimated lifetime at body temperature
yielded promising results for both variants. In conclusion, both
materials showed good performance, and when selecting the
dielectric material for pre-charged collapse-mode CMUTs, a
trade-off between other factors such as ease of deposition,
charging time, and charge retention should be carefully con-
sidered.
ACKNOWLEDGMENT
The authors would like to thank Eugene Timmering for
fabricating the CMUT devices, Paul Dijkstra and Jacco Scheer
for their assembly, Bart Mos for the acoustic measurements.
REFERENCES
[1] K. K. Park, O. Oralkan, and B. T. Khuri-Yakub, “A comparison between conventional
and collapse-mode capacitive micromachined ultrasonic transducers in 10-MHz 1-D
arrays,” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control,
vol. 60, no. 6, pp. 1245–1255, 2013.
[2] A. Kshirsagar, A. Sampaleanu, R. Chee, W. Moussa, and R. J. Zemp, “Pre-charged
CMUTs with efficient low-bias voltage operation for medical applications,” in 2013
IEEE International Ultrasonics Symposium.
[3] M. Saccher, S. Kawasaki, and R. Dekker, “The long-term reliability of pre-charged
CMUTs for the powering of deep implanted devices,” in 2021 IEEE International
Ultrasonics Symposium, pp. 1–4.
[4] M. Saccher, S. Kawasaki, J. H. Klootwijk, R. Van Schaijk, and R. Dekker, “Modeling
and Characterization of Pre-Charged Collapse-Mode CMUTs,” IEEE Open Journal
of Ultrasonics, Ferroelectrics, and Frequency Control, vol. 3, pp. 14–28, 2023.
[5] W. Y. Choi, C. H. Lee, Y. H. Kim, and K. K. Park, “Comparison of Si3N4–SiO2and
SiO2Insulation Layer for Zero-Bias CMUT Operation Using Dielectric Charging
Effects,” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control,
vol. 67, no. 4, pp. 879–882, 2020.
[6] M.-C. Ho, M. Kupnik, K. K. Park, and B. T. Khuri-Yakub, “Long-term measurement
results of pre-charged CMUTs with zero external bias operation,” in 2012 IEEE
International Ultrasonics Symposium, pp. 89–92.
[7] K. K. Park, M. Kupnik, H. J. Lee, O. Oralkan, B. T. Khuri-Yakub, and Ieee,
Zero-Bias Resonant Sensor with an Oxide-Nitride Layer as Charge Trap, ser. IEEE
Sensors. IEEE, 2010, pp. 1024–1028.
