Design of High Frequency Ultrasound Transducers using ...

Design of High Frequency Ultrasound Transducers using
Micromachining
Tung Manh, Anh Tuan. T. Nguyen, Lars Hoff and Tonni F.Johansen
Abstract— We present a new design of high frequency
ultrasound transducers using thick-film PZT on Silicon for
medical application. Air is used as backing material for the
PZT film, and the ultrasound pulses are transmitted through
the silicon substrate. The silicon substrate is micromachined to
build acoustic matching layers to optimize the acoustic coupling
between the active PZT layer and the tissue, and achieve a
large bandwidth. The number of matching layers in the design
could be either two or three, depending on the fabrication
methods to make the PZT thick-film, i.e. tape casting or screen
printing. A prototype of two matching layers for a 10MHz
transducer has been made. 1D analytical calculations and FEM
simulations show that the proposed design will have a 3 dB
bandwidth ranging from 7.5 MHz to 13 MHz, or 55%
bandwidth at center frequency of 10 MHz. Acoustic
measurement is ongoing.
I. INTRODUCTION
H IGH frequency ultrasound transducers are wanted in
medical imaging applications where high resolution is
needed [1]-[2]. Most of today’s medical ultrasound
transducers are based on an active ceramic layer of lead
zirconate titanate (PZT) vibrating in thickness mode. To
achieve frequencies in the range 10 to 100 MHz, the
thickness of the active ceramic layer must be between 20
and 200µm. Such a thin layer is difficult to lap down to a
correct thickness, whereas too thick to build up by thin film
techniques. In this paper, we propose using thick-film
technology and silicon micromachining as fabrication
technologies to bridge this gap.
Previous work in this field includes Zang et al. [1], who
fabricated a high frequency ultrasound transducer using PZT
thick-film on a Silicon (Si) substrate, which can be annealed
to a temperature of 750 0 C without reaction between the PZT
film and the Si substrate. This improved the film properties
compared to the previous work by Lukacs et al. [2], who
used Aluminum (Al) substrate. In both of these works, the
Manuscript received April 01, 2010. This work is funded by the
Research Council of Norway under Grant No. 176540.
Tung Manh, Anh Tuan T. Nguyen and Lars Hoff are with the Faculty of
Science and Engineering, Vestfold University College, P.O. Box 2243, N-
3103 Tønsberg, Norway (e-mail: tung.manh@ hive.no).
Tonni Franke Johansen is with the Department of Circulation and
Medical Imaging, University of Science and Technology, Trondheim,
Norway.
substrates served as temporary layers, and were etched off to
form epoxy backed transducers. This enhanced the
bandwidth of the transducers, but reduced the sensitivity.
We present a new design for an ultrasound transducer
where substrate is kept. Using Silicon micromachining
technology, the substrate is machined to form layers with
descending characteristic acoustic impedance from the PZT
film to the load material. These layers function as acoustic
matching between the thick-film and the load. The back side
of the thick-film is air, serving as the transducer backing.
This structure increases sensitivity and avoids problems with
reverberations in the backing material.
The paper contains a description of the design transducers
with details of the fabrication process to form matching
layers. Details of fabrication steps to build the entire
transducer can be found elsewhere [3]. The performance of
the designs is described analytically using the Mason stack
model [4]. Finite Element Method (FEM) simulations are
included to improve and verify the analytical 1D calculation.
II. DESCRIPTION OF THE TRANSDUCER
Fig. 1 (a) – (b) Cross sections of the transducers with two/ three matching
layers. Top view of (c) 1-3 composite and (d) 2-2 composite.
Cross sections of our transducer designs are shown in Fig.
1(a-b). The structure includes a PZT plate operating near its
half-wavelength resonance, covered by front and back
electrodes. The front is covered by two or three acoustic
matching layers, each having thickness λ/4; where λ is the
wavelength of the longitudinal bulk wave normal to the

layer, at the centre frequency of the transducer. For two
matching layers design [Fig. 1(a)], these two layers are: 1) a
layer consisting of a composite between Si and a polymer,
and 2) a layer of polymer. For three matching layers design
[Fig. 1(b)], the three layers are: 1) a substrate layer of Si, 2)
a composite layer between polymer and Silicon and 3) a
polymer layer. Since tape-casting method gives the
properties of PZT plate similar to bulk ceramic [5], the three
matching layers approach should be used. In contrast, the
screen printing method with sol-gel mixture provides a film
with acoustic impedance approximately equal to that of
Silicon, two matching layers must be used.
The composite layer is made as either 1-3 connectivity
[Fig. 1(c)] or 2-2 connectivity type [Fig. 1(d)] using
micromachining techniques [6]. The 1-3 type is preferred for
high frequencies (>20MHz) due to the fact of
manufacturing. The two electrodes function as the electric
port to connect the transducer to the electronics. They
should be made sufficiently thin so that their influence on
the acoustic performance of the transducer is small.
However, the electrodes must also be adequately thick and
dense to block the diffusion between Si substrate and the
thick-film [7]. An appropriate thickness is less than λ/10,
where λ is the wavelength of the longitudinal bulk wave in
the PZT plate.
The transducer is constructed as either a single circular
element or an annular array with ring elements for focusing.
Scanning for image formation will be done mechanically by
rotation and/or translation.
III. ANALYTICAL AND FEM SIMULATIONS
A. Analytical simulation
The performance of the designs was modelled analytically
using the Mason equivalent circuit model. In this model,
PZT ceramic of type pz27 is used as the active layer, the
substrate is -oriented Si, the electrodes are of Pt and
the polymer is epoxy resin. A water-like medium with
acoustic impedance 1.6 Mrayl is used as load. The material
parameters used in the modelling are listed in TABLE I.
As demonstration, only simulation corresponding to the
design with two matching layer transducer at the centre
frequency of 10MHz using 2-2 composite type as matching
layer is performed. Fabrication of such a layer is given in
IV. The second design can be found in [3].
In the 10MHz transducer design, the thickness of the PZT
plate is 177 µm, corresponding to λ/2 at frequency 11MHz,
covered by two Pt electrodes with thickness of 3.0 µm and
acoustic impedance of 84.7 Mrayls. The first matching layer
is composite of Si and epoxy with thickness of 152µm and
acoustic impedance of 8.9 Mrayls, whereas the last matching
layer is 63µm thick epoxy with acoustic impedance of
2.85Mrayl [Epofix (Struers, Solilull, UK)]. This epoxy has
acoustic impedance higher than the optimal value of
2.3Mrayl but was chosen due to its availability in our lab.
The effective elastic parameters of the composite
Si/Epoxy matching layer can be calculated using the
theoretical model developed by Smith et al. [8], with some
modifications. The resulting parameters are
2
⎡ p S
2(
c12 −c
) ⎤
S 13
p
c33 = v⎢c11 −( 1− v) ⎥+
( 1−v)
c11
(a)
S S p p
⎢ ( 1−
v)( c11 + c12 ) + v( c11+ c12)
⎥
⎣ ⎦
p S
2
⎡ ( c12 −c13
) ⎤
S p
c33 = v⎢c11−( 1− v) ⎥+
( 1−v
S p ) c11
(b)
⎢ ( 1−
vc ) 11 + vc⎥
11
⎣ ⎦
S p
ρ = vρ + ( 1−v)
ρ
Z = c ρ
c33
V =
ρ
where the overbars ( x ) refer to effective parameters of the
composite, superscript p refer to parameters of the polymer,
and superscript S to parameters of the silicon substrate.
cij are the elastic coefficients, ρ is the density, v is the
volume fraction of Si in the composite; Z , V are the
effective acoustic impedance and longitudinal wave speed in
the thickness direction of the composite, respectively. (a) is
the equation used for 1-3 composite whereas (b) is for 2-2
type. The others are for both.
The electro-acoustic transfer function, i.e. the transfer
function from the voltage over the electrodes to the pressure
at the transducer surface, is calculated for a 10MHz design
and plotted in Fig. 2. The upper graph shows the frequency
response, while the lower shows the impulse response, in the
time domain. These results give an effective 3-dB bandwidth
from 7.5 MHz to 13 MHz with centre frequency 10 MHz,
i.e. 55% relative bandwidth.
B. FEM simulation
Analytical model, based on the Mason equivalent circuit,
is well established in transducer modelling for describing
thickness mode vibrations. However, it is limited by being a
1D model. Hence, the composite layer must be described as
one material with a set of calculated effective parameters,
the anisotropic nature of Si is not handled, and lateral modes
as well as effects of a finite width are not included.
Fig. 2 Electro-acoustic transfer function and output waveform of 10MHz
transducer
33

Fig. 3 Electrical impedance of a 10MHz transducer
To obtain more realistic simulations, and to get an
estimate of the performance and limitations of the 1D
analytical model, FEM models were created using
COMSOL Multiphysics software version 3.4 (Comsol AB,
Stockholm, Sweden). In FEM model, the anisotropic nature
of Si is included, and the composite is made between Si and
Epoxy posts. One half of a composite unit cell was modeled
for 2-2 composite, whereas a quarter of cell was modeled for
1-3 type. Symmetry boundary conditions were applied, as
described in [9]. The sizes of the Si and polymer bars in the
composite layer of the 10MHz transducer were designed to
be 9µm and 23µm, respectively, to avoid interference from
the lateral resonances into the thickness mode [10]. There is
a good agreement between FEM simulation and analytical
calculation (Fig.3). For 3 matching layer design, similar
result is also achieved.
IV. FABRICATION OF THE MATCHING LAYERS
The novel and most important part in the transducer is the
micromachined matching layer. This layer has been
fabricated by using anisotropic wet etching method to form
deep vertical trenches (2-2 type) into Silicon wafers,
followed by vacuum impregnated these grooves by epoxy
resin.
The process starts with a 550 µm thick oriented Si
wafer. A 1µm thin SiO2 was grown on the surfaces of the
wafer. The main flat of the wafer is supposed to be
parallel to the planes. However, the accuracy of this
flat is approximately 1 0 , which is not efficient to have good
alignment for our purposes. A two steps etching process
therefore has been made. First, a mask with fan structures of
0.1 0 increasing step was transfer to the SiO2 layer by
lithography. A 1.5 hours etch in KOH 25% at 80 0 C was then
performed. All lines which are imprecisely aligned along
planes were undercut and destroyed due to the nature
of anisotropic wet etching. Remain lines were used to be the
alignment marks for the next lithography. Fig. 4(a) shows
the fan structure patterns on the plastic mask whereas the
etched structures on Silicon after the first wet etching step
are showed in Fig. 4(b). Due to anisotropic etching of the
Silicon crystal, the circle on the mask turned to be hexagonal
shape and only lines which are precisely aligned along
planes have been survived completely.
The grating pattern, i.e. 2-2 composite, was then
transferred to the protected SiO2 layer, followed by
anisotropic wet etching of the wafer for 1.5 hours to form
deep trenches. A closed look of the etched structure before
polymer filled is shown in Fig. 5. Obviously, straight walls
have been achieved but there are steeps at the bottom of the
trenches due to anisotropic behaviour of Silicon crystal.
Then, the wafer was vacuum impregnated by polymer. It
was done by placing the wafer into a vacuum chamber and
pouring epoxy resin onto wafer’s surface. By applying a
vacuum, small air bubbles in polymer were removed. The
polymer was cured at room temperature for 24 hours. The
sample after vacuum impregnated and cured is shown in Fig.
6, where bright is Si and grey is polymer. It is clearly shown
that air bubbles are removed in polymer.
The sample was then lapped down to the desired
thickness using a precise lapping machine (MultiPrep TM
System, Allied High Tech Products Inc, USA). A PZT active
layer was then glued on the sample for characterization.
Characterization of these matching layers is ongoing.
Fig. 4 (a) Fan structures on the glass mask (b) Patterns on Silicon wafer
after the first step in Silicon wet etching process
Fig. 5 Closed look of the etched structure before filled in with polymer

Fig. 6 Sample after polymer filled, bright is Silicon and grey is Polymer
V. CONCLUSION
The novel design of ultrasound transducers based on
micromachining has been presented. A test prototype of the
composite layers using wet etching has been made.
Theoretical and FEM simulation show that transducers using
this structure to be matching layers provide 55% relative
bandwidth. This is a little smaller than the optimal
bandwidth that can be achieved with two matching layers
but it is still reasonable. This is because we didn’t use epoxy
with optimal acoustic impedance. Acoustic measurement of
this structure is ongoing.
Active
layer
Passive
layers
Pz27
[110]oriented
Silicon
Platinum
Epoxy
TABLE I
MATERIAL PROPERTIES USED FOR SIMULATIONS
REFERENCES
[1] Q.Q. Zhang, F.T. Djuth, Q.F. Zou, C,H. Hu, J.H. Cha, K.K. Shung,
“High frequency broadband PZT thick film ultrasonic transducers for
medical imaging applications”.Ultrasonics, vol.44, p711-715, 2006.
[2] M. Lukacs, M. Sayer, F.S. Foster, ”Single element high frequency
(