You are on page 1of 4

Poster 119

EXPERIMENTAL AND THEORETICAL ANALYSIS OF SILICON-BASED


PIEZOELECTRIC TRANSDUCERS FOR ULTRASOUND IMAGING
S. Ballandras (l), S. Basrour (l), N.Griffaton (l), J.F. Gelly (2),F. Lanteri (2)

( I ) LPMO-CNRS, 32 Ave de I’Observatoire, 25044 BESANCON Cedex


(2)THOMSON MICROSONICS, 399 Route des Cvztes,
BP. 232, 06904 SOPHIA ANTIPOLIS Cedex

Abstract - The development of new Micromachined excite the fundamental flexural mode of the membrane.
Ultrasound Transducers (MUT) for medical imaging or Non This vibration results in a large normal displacement
Destructive Evaluation (NDE) has been allowed by the generating a cylindrical pressure field when radiating in
vulgarization of MicroElectroMechanical Systems (MEMS) fluids, just as the classical piston mode transducer does. Of
technologies. The possibility to massively produce very high course, the use of silicon substrate and its associated
quality transducers (high sensitivity, large S/R, good technologies allows the batch processing of many devices,
dynamics) directly integrated on silicon substrates with a high yielding the future fabrication of integrated imaging
throughput represents a significant advance in acoustic systems associating transducer arrays with front-end
imaging. The proposed work consists in the analysis of electronics for the treatment of excitation and detection
piezoelectric MUT to characterize their main properties signals (multiplexing, beam forming, electric matching,
(coupling coefficient, resonant frequencies, acoustic etc.). Furthermore, the possibility to manufacture 2D arrays
impedance) from both theoretical and experimental point of of very small sized transducers allows the development of
view. Computation using 3D finite element have been achieved 3D imaging probes with reasonable production costs.
and compared to experiments performed on large scale
PZTlSi transducers. Electrical response have been measured The present work consists in analyzing the physical
together the distribution of dynamical displacements using an properties of piezoelectric transducers composed of PZT
optical interferometer at resonant frequencies to check the ceramics reported on Si membranes. Theoretical
validity of the theoretical analysis. calculations have been performed to identify thee influence
of the different geometrical parameters in order to establish
design rules for PZT/Si flexural transducers. 3D finite
I. INTRODUCTION element calculations have been used in that matter,
allowing to evaluate the resonance frequency, the
Standard acoustic imaging probes used for biomedical electromechanical coupling factor or the dynamics of the
applications consist generally in an array of piezoelectric different tested configurations. In parallel, experimental
transducers working on their fundamental compression devices have been fabricated to validate the theoretical
mode to generate pressure fields propagating in the analysis. Elementary devices at large scale (working
patient’s body. Due to variation of the acoustic impedance frequency in the 10-100 kHz) have been fabricated and
from one organ to the other, this field is partially reflected tested without fluid loading. Electrical characteristics are
toward the probe which then used to detect any retum measured using standard network analyzer. Identification of
signal (the transducers are reciprocal). The use of the vibration shapes is performed using an optical
transducer arrays allows an electronic control of the interferometric probe giving access to the normal
pressure beam which is then spatially focussed, improving dynamical displacement field of the membrane for a given
the dynamics of the probe and allowing to define the excitation frequency. From this measurement, one can
equivalent of an acoustic pixel. The acoustic information is deduce the normal velocity and then compute the acoustic
then transformed in a gray scale image giving access to a impedance of the transducer. The electric admittance of
slice of the inspected body. This fundamentals of acoustic these elementary devices loaded by water has been also
imaging can be found in details in many reference books measured to determine their operational frequency
(see for instance [1,2]). bandwidth. As a conclusion, comparison between theory
and experiments is discussed.
Recent work has shown the interest of new architectures
of ultrasonic transducers for the development of high
density 2D probes well adapted to acoustic imaging [3,4,5]. 11. BASIC PRINCIPLE AND FABRICATION OF LARGE
These transducers are based on the dynamic excitation of SCALE TRANSDUCERS
flexural motions of silicon membranes. Such vibrations can
be produced either by electrostatic forces [3,4] or by The very basic principle of the piezoelectric MUT is
piezoelectric dilatation of the silicon membrane [4,5].In the reported in fig. 1 according to one of the very first devices
first case, the membrane consists in a silicon nitride film built by Bernstein & al [j]. It consists in a Silicon
fabricated on the silicon substrate using surface membrane covered by a PZT layer (thickness ranging from
micromachining techniques. In the second case, a 1 to 3 pm) actuated via two electrodes in such way that the
piezoelectric film has to be deposited on the surface of the fundamental flexural mode of the so-built bimorph can be
membrane, Both approaches are suitable to efficiently efficiently excited.

1st Annual International IEEE-EMBS Special Topic Conference on Microtechnologies in Medicine & Biology 97
October l2-14,2000, Lyon, France - 0-7803-6603-4/00/$10.00~2000 IEEE
Poster 119
111. COMPARISON BETWEEN THEORETICAL AND
MEASURED ADMITTANCES

Different meshes have been performed to treat the


practically considered structures. The PZT thickness was
fixed to 127 pm or 267 gm according to the plate used
experimentally, Only simple rectangular and octagonal
shapes of membranes and PZT have been considered to
answer the requirements of simplicity and low fabrication
cost. From a theoretical point of view, the PZT has been
I I / assumed in perfect contact with the Silicon. No mechanical

~ J, $ : s e
clamping was assumed in order to avoid biasing the results
by non proper assumptions. Following this approach, the
PZT was excited by an harmonic homogeneous potential of
PZT layer magnitude 1V (dependence versus time assumed equal to
Pressure radiation do').on the top electrode, the bottom electrode (in contact
with the silicon) being grounded. This allows to directly
obtain the admittance by simply computing the electrode
Fig. 1Scheme ofthe basic principle of a PZT/Si transducer current on one or the other electrode (according to the
chosen convention). Figure 2 shows a comparison between
In reference [SI, the PZT layer is deposited using theory and experiments performed on a very large
sputtering techniques. However, the PZT must be cracked spectrum. This allows a comprehensive analysis of the
to acquire its piezoelectric characteristics. This can be validity of the finite element calculation by checking
efficiently done using a Rapid Thermal Annealing (RTA) whether the adopted theoretical approach is able or not to
oven. However, this technological step requires the PZT to faithfully predict the actual modes of the structure.
be deposited on a TiPt layer well suited to avoid any
diffusion of the heavy components of the PZT (mainly Pb) Different informations can be obtained from this
in the Silicon. This protection layer can be opportunely comparison. First, it obviously appears that not all the
used as counter electrode when exciting the device. The top modes are predicted by the finite element computations
electrode can be pattemed using conventional techniques (acoustic losses have been neglected here). Second, a
(lithography, lift off, wet etching, etc.) and can be significant frequency shift separates calculation and
advantageously composed of Aluminum. experimental results. Finally, numerous modes arise from
the fundamental flexural vibration (a few tenth of kHz) to
In our experiments, rather large silicon membranes have the longitudinal compression mode of the bimorph (around
been built allowing an easy measurements of vibration 10 MHz).
distribution using optical means (see section IV). The top
side of the silicon wafer has been covered by an
homogeneous Aluminum film (a few thousand Angstroms)
and the shape of the counter electrodes have been patterned
by photolithography and wet etching. The PZT electrodes
have been also performed, starting from 3 inches square
wafers which have been then cut using ultrasound
micromachining to fit the membrane dimensions precisely.
The PZT plates have been then glued on silicon by using a
very thin PMMA layer. The adhesion of both material has
been mechanically checked and was found homogeneous
on the whole membrane area. IOOWO A-
10 100 io00 IOW0 IOW
Frequency (in W )
Due to the rather large dimension of the membrane and --
the small thickness of the above-mentioned metal layer, one
can neglect their acoustic contribution when analyzing the
electromechanical behavior of the structure of fig.1 without
any loss of generality (at very low frequencies, only the
central frequency may be affected by the presence of the
metal).
A
As a consequence, 3D finite element calculations have -
been implemented to predict the electromechanical working 0W 2 5
_ _ a5.
I

of such transducers (for a comprehensive insight of such 0 WO2 -- .


65 70 75 eo L5-3 95
calculations, see for instance [ 6 ] ) . Frequency (in kM)
Fig. 2example of experimental and theoretical admittances of a large
scale transducer (membrane lateral dimension I .2 cm, Si membrane
thickness 150 pm. PZT thickness 256 pm)

98 1' Annual International IEEE-EMBS Special Topic Conference on Microtechnologies in Medicine & Biology
October 12-14,2000, Lyon, France - 0-7803-6603-4/00/$10.0002000 IEEE
Poster 119
The frequency shift is rather simple to explain, taking The device tested in fig. 2 has been measured operating at
into account the fact that not all the material constants the central frequency of the fundamental flexural mode
required by the theoretical analysis was provided by the response. The size of the image corresponds to the silicon
PZT manufacturer, and more specially those conceming membrane surface, the pixel is equal to 100 pm. The
shear effects which are generally considered as undesired displacement distribution on the Silicon side of the bimorph
parasitical effects. The material data set has been then is reported in fig.4 under the form of magnitude and phase.
completed using commonly used values (according to what The theoretically predicted distributions are found very
published in the literature). It is more difficult to explain the close to the one measured, showing the capability of the
absence of particular modes on the theoretical curve in finite element analysis to well describe such vibration.
regard with experiments. However, since square
membranes have been used, a very particular care must be
devoted to boundary conditions. Actually, the square
membrane exhibit degenerated modes capable to convert in
differentiated vibrations if any symmetry defect arises. This
point will be discussed again in the next section. The
measurement of well defined longitudinal modes can be
considered as a positive issue since it indicates that the
piezolectrically excited modes are well transmitted to the
Silicon membrane. Again, the slighter discrepancy between
theory and experiments is mainly due to the incompleteness
of the data set. Finally, it is shown that the proposed (a) Magnitude
analysis requires a very good knowledge of the material
constants and of the actual boundary conditions to provide a
reliable prediction of the structure modes. Since the
fundamental flexural mode has been theoretically predicted,
it can be remarked that standard 3D finite element analyses
can be implemented to optimize its characteristics. For
.
k
instance, choosing the appropriate PZT thickness together i
with the membrane dimensions allows to theoretically
found coupling coefficients close to 20%. However, this
figure has to be checked experimentally. The measured
electromechanical coupling coefficient of the flexural mode
of the structure tested in fig2 was compared to
theoretically predicted.

IV. MECHANICAL DISPLACEMENT


CHARACTERIZATION

A complementary assessment of the theoretical


calculation versus experiments can be achieved by
comparing the displacement distributions provided by the
finite element calculation and those obtained using the
experimental setup described below. It is based on the
measurement of the normal displacement of the sample to
be tested by using an optical interferometer probe. A
principle scheme of the setup is shown in fig.4.

- plisne deDove (b) Magnitude

Fig3 Principle scheme of the optical interferometer probe used to


measure normal displacements of our transducers. (b) Phase

1" Annual Intemational IEEE-EMBS Special Topic Conference on Microtechnologies in Medicine & Biology 99
October 12-14,2000, Lyon, France - 0-7803-6603-4/00/$10.0002000 IEEE
Poster 119
Fig. 4 Experimentally measured displacement distribution throughout demonstrated, allowing to develop reliable low cost NDE
the silicon membrane for different mode orders :(a) 3 1 flexural mode (b) acoustic probes using very accessible technologies.
31-13 flexural mode

V. CONCLUSION
REFERENCES

[I] G.S. Kino, "Acoustic waves, devices, imaging and analog signal
First Of a study devoted to piezoelech.ic MUTs
processing", Prentice-Hall, Inc., Englewoods Cliffs, New Jersey,
have been reported in this paper. They consist in 1987
experimental measurements and theoretical predictions of [21 W. p, acPhysicalAcoustics~~, pa Academic press,
bimorph structures composed of a silicon membrane New York, 1964
supporting a piezoelectric plate capable to excite its
[5] I. Ladabaum, B.T. Khuri-Yakub, D. Spoliansky, M.I. Naller
fundamental flexural vibration mode. Both electrical and "Micromachined Ultrasonic Transducers (MUTs)". IEEE Ultrasonics
mechanical characteristics of the mode can be predicted and
Symposium, ,99j, pp jo1-jo4
measured' The possibility to use standard 3D finite element 141 p,C. K, Niederer, B,Fischer, "Micromachined transducers
harmonic analyses of such devices for optimization purpose for ultrasound applications", Proc. of the IEEE Ultr. Symp., pp 1609-
depend on the capability of the designer to have access to a
1618, 1997
data set Of Inaterial 'Onstant together with a (51 J.J. Bemstein, S.L. Finberg, K. Houston, L.C. Niles, H.D. Chen, L.E.
fabrication process and a technok%ically robust structure.
cross, K,K, Li, K, Udayakumar ,gMicromachinedhigh Frequency
The 'Onsidered 'quare shape Of the membrane is not the ferroelectric sonar transducers" IEEE Trans, Ultrasonics,
best configuration to fit this requira". Using classical Ferroelectricsand Frequency Control, @, sepr, 1997, pp
KOH etching of the Silicon, it i s possible to perform
960-969
membranes better suited in that approach' [6] R. Lerch, "Simulation of piezoelectric devices by two and three
However, the possibili~to fabricate low pzT/si dimensional finite elements", 1EEE Trans, on UFFC, Vol. 37, ne 2,
transducers with simple gluing techniques has been nicely pp. 233-247,1990

100 1 Annual International IEEE-EMBS Special Topic Conference on Microtechnologies in Medicine &. Biology
October 12-14,2000, Lyon, France - 0-7803-6603-4/00/$10.0002000 IEEE

You might also like