Three-dimensional micro electromechanical system piezoelectricultrasound transducerArman Hajati, Dimitre Latev, Deane Gardner, Azadeh Hajati, Darren Imai et al. Citation: Appl. Phys. Lett. 101, 253101 (2012); doi: 10.1063/1.4772469 View online: http://dx.doi.org/10.1063/1.4772469 View Table of Contents: http://apl.aip.org/resource/1/APPLAB/v101/i25 Published by the American Institute of Physics. Related ArticlesEfficient counter-propagating wave acoustic micro-particle manipulation Appl. Phys. Lett. 101, 233501 (2012) Piezoelectric and electrostrictive effects in ferroelectret ultrasonic transducers J. Appl. Phys. 112, 084505 (2012) Piezoelectric resonator arrays for tunable acoustic waveguides and metamaterials J. Appl. Phys. 112, 064902 (2012) Focused high frequency needle transducer for ultrasonic imaging and trapping Appl. Phys. Lett. 101, 024105 (2012) A rigid, monolithic but still scannable cavity ring-down spectroscopy cell Rev. Sci. Instrum. 83, 043115 (2012) Additional information on Appl. Phys. Lett.Journal Homepage: http://apl.aip.org/ Journal Information: http://apl.aip.org/about/about_the_journal Top downloads: http://apl.aip.org/features/most_downloaded Information for Authors: http://apl.aip.org/authors

Three-dimensional micro electromechanical system piezoelectric ultrasoundtransducer

Arman Hajati,1 Dimitre Latev,1 Deane Gardner,1 Azadeh Hajati,2 Darren Imai,1

Marc Torrey,1 and Martin Schoeppler1

1FUJIFILM Dimatix, Inc., Santa Clara, California 95050, USA2TABA Medical Imaging Center, Mirzay-Shirazi Blvd., Shiraz 71888-34777, Iran

(Received 26 October 2012; accepted 30 November 2012; published online 17 December 2012)

Here we present the design and experimental acoustic test data for an ultrasound transducer

technology based on a combination of micromachined dome-shaped piezoelectric resonators

arranged in a flexible architecture. Our high performance niobium-doped lead zirconate titanate

film is implemented in three-dimensional dome-shaped structures, which form the basic resonating

cells. Adjustable frequency response is realized by mixing these basic cells and modifying their

dimensions by lithography. Improved characteristics such as high sensitivity, adjustable wide-

bandwidth frequency response, low transmit voltage compatible with ordinary integrated circuitry,

low electrical impedance well matched to coaxial cabling, and intrinsic acoustic impedance match

to water are demonstrated. VC 2012 American Institute of Physics.

[http://dx.doi.org/10.1063/1.4772469]

Bulk piezoelectric transducers, in which a plate of lead

zirconate titanate (PZT) piezoelectric ceramic is mechani-

cally diced and backfilled with a polymer filler to form a

one- or two-dimensional array of elements, have dominated

ultrasonic transducer technology for a long time.1 Micro

electromechanical system (MEMS) based ultrasound trans-

ducers have recently emerged as an alternative aiming to

offer advantages such as increased bandwidth, flexible geo-

metries, natural acoustic match with water, reduced voltage

requirements, mixing of different resonant frequencies, and

potential for integration with supporting electronic circuits

especially for miniaturized high frequency applications.2,3

Capacitive micro-machined ultrasound transducer

(CMUT) technology introduced in the 1990s has offered

promising results such as better acoustic matching, a broader

bandwidth, improved flexibility in the design of complex

arrays, size reduction, and the potential for integration of

front-end electronics on the same wafer. Nevertheless,

CMUTs suffer from serious issues, including low transmit sen-

sitivity (�10 kPa/V), very large drive and bias voltage require-

ments (up to 200 V), difficult fabrication, long-term device

reliability, electrical safety issues, acoustic cross-talk, and

high electrical impedance as a result of limited capacitance.4

Partially collapsed CMUT transducers employing high-K

dielectrics may address some but not all of these drawbacks.5

Thin film piezoelectric materials offer a number of

advantages including the large motions that can be gener-

ated, often with low hysteresis, the high available energy

densities, as well as high sensitivity and low power require-

ments and have been used in various diaphragm based6,7 and

cantilever based8–10 MEMS transducers. Piezoelectric

micro-machined ultrasound transducers (PMUT) have been

sought as a solution that can be actuated at much lower volt-

age levels and provide much lower electrical impedance per

area, even in comparison with conventional bulk piezoelec-

tric transducers. Muralt et al. fabricated and characterized a

PMUT based on 2 lm thick PZT film deposited by spin-

coating method.11,12 Wang et al. reported a low frequency

(76 kHz) membrane-based PMUT using 3.5 lm thick PZT

film.13 A low-cost piezoelectric thin film platform for ultra-

sound imaging and sensing applications in the frequency

range of >50 kHz up to 10 MHz was presented by Klee

et al.14–16 Zhou et al. fabricated very high frequency

(>100 MHz) kerfless ultrasound linear arrays using thick

piezoelectric films.17,18 Flexure-mode 2D PMUT arrays con-

taining 81 elements were reported by Dausch et al.19,20

However, difficulties in fabricating thin-film piezoelec-

tric MEMS structures combined with technical issues such

as very low electromechanical coupling and small bandwidth

have prevented PMUT devices from becoming a viable and

practical solution.11 Traditional PMUT devices are based on

the flexural motion of a thick membrane, which is actuated

by an applied piezoelectric film in the bending mode. The

stiffness of the structure is predominantly determined by the

membrane layer, typically silicon, which causes very poor

electromechanical coupling (less than 5%) and small acous-

tic bandwidth.11 In addition, these designs exhibited rela-

tively high Q-factors and reduced bandwidth compared to

ordinary bulk PZT transducers. PMUT performance may

also be constrained by the limited piezoelectric performance

of sol-gel and aerosol PZT films compared to bulk PZT.18,21

In this paper, we are presenting a micro-machined ultra-

sound technology called ClarinetTM aiming to address the

aforementioned issues. Analogous to IC technology, Clarinet

can be seen as a monolithic integrated ultrasonic circuit man-

ufactured by lithography with four main advantages over

bulk piezoelectric transducers: cost, performance, customiza-

tion, and miniaturization capability. In this paper, design and

test results of a wideband 5 MHz linear array are presented

and discussed as an example; however, transducers in a wide

range of frequencies (1-40 MHz) have been fabricated in the

same wafer by changing the element dimensions.

As shown in Fig. 1, the basic resonating cell consists of

active piezoelectric film in the form of a three dimensional

(3D) dome with thin metallic top and bottom electrodes. The

pre-shaped structure of the dome eliminates the stiff silicon

0003-6951/2012/101(25)/253101/5/$30.00 VC 2012 American Institute of Physics101, 253101-1

APPLIED PHYSICS LETTERS 101, 253101 (2012)

membrane layer otherwise required for the traditional bend-

ing mode. Electrical energy is efficiently converted via piezo-

electric effect to elastic energy through the “stretching” mode

in addition to the typical “bending” mode.22 Subsequently,

the elastic energy is converted into the desired acoustic

energy through the interaction of the dome and the medium.

Consequently, a significant electromechanical coupling, as

high as 45%, and strong acoustic sensitivity are achieved.

One of the biggest challenges in the development of

high performance and reliable PMUTs has been the lack of a

reliable process to deposit high-quality piezoelectric films.6,7

As reported in Ref. 23, we have developed a reliable and

repeatable process to sputter a dense, high performance

niobium-doped lead zirconate titanate (PNZT) film that has

an unusually high level of Nb dopant (13%). The additional

Nb dopant results in a �70% higher piezoelectric coefficient

than sputtered PZT films previously reported. As shown in

Fig. 2, the x-ray diffraction patterns of the PNZT film dem-

onstrate that film is in a perovskite phase with predominantly

(100) orientation, which partly accounts for its high piezo-

electric performance (e31,f¼�23 C/m2). One of the unique

properties of the PNZT film is that the hysteresis loop is

shifted toward the positive electric field direction. Conse-

quently, the polarization axes have been aligned in a certain

direction beforehand, making a post-deposition polarization

process unnecessary. As a result, no poling or aging proce-

dure was required in the fabrication of the transducers.23

Semi-spherical membranes cells vibrate at various

modes, which are the solutions of Bessel functions. Consid-

ering that the piezoelectric excitation of the semi-spherical

dome is almost independent of the angle h, the preferred

mode shapes are (0, 1) mode, (0, 2) mode, (0, 3) mode in

which the number of nodal diameter is 0. Fig. 3 shows the

simulated acoustic intensity sensitivity generated by a sin-

gle dome with the cavity diameter of 75 lm as a function of

frequency up to 20 MHz. The mode shapes of the single

dome are measured by laser Doppler vibrometer (MSA-500

Micro System Analyzer by Polytec). (0, 1) mode and (0, 2)

modes with corresponding acoustic intensity of 0.15 mW

(RMS) and 0.13 mW (RMS) at 1 V can be identified, which

translate into impressive acoustic power density of 3 and

2.6 W/cm2 at 1 V excitation level. However, the corre-

sponding bandwidth of 15% and 9.5% is too narrow for

ultrasound imaging applications.

FIG. 1. Cross-section of three-dimensional

dome-shaped piezoelectric membrane, which is

the fundamental element of the transducer.

Schematic (left) and scanning electron micro-

scope image (right).

FIG. 2. FUJIFILM Dimatix proprietary RF-sputtered thin film d31XL PNZT

film. (a) X-ray diffraction shows a perovskite phase with (100) orientation.

(b) P-E hysteresis across the wafer. The hysteresis loop is shifted towards

the positive electric field, suggesting that the polarization axes have been

aligned in a certain direction beforehand, making a post-deposition polariza-

tion process unnecessary.17

FIG. 3. Simulated acoustic intensity spectrum (bottom) of a single dome at

1 V excitation level shows two modes of vibration. The simulated acoustic

pressure sensitivity (in kPa/V) distribution (top) and their mode shapes

measured by laser Doppler vibrometer (middle).

253101-2 Hajati et al. Appl. Phys. Lett. 101, 253101 (2012)

Unlike a bulk piezoelectric transducer in which the reso-

nant frequency is fixed by the thickness of the ceramic, the

resonant frequency of a micro-machined dome-shaped ele-

ment is mainly determined by its dimensions as defined by li-

thography. Therefore, an array of paralleled domes may be

constructed as a network of resonators similar to a multistage

analog LCR electronic filter, in which the component values

of the electrical equivalent circuit can be adjusted by modi-

fying the dimensions of the domes by lithography. In this

way, all the theoretical tools of electrical analysis and filter

design can be implemented in the acoustic design and any

filter realizable in electrical theory such as Butterworth and

Chebyshev filters can, in principle, also be synthesized as an

acoustic transducer. Basically, the process is very similar to

the design of RF and microwave band-pass filters using

microstrips.24

To design and optimize a transducer with a desired wide

bandwidth spectrum, a lumped-parameter electro-mechanical/

acoustic model is developed. Each element’s motion Xi indu-

ces an acoustic pressure in the medium and on the surface of

the transducer itself in the form of Piðr;xÞ ¼ Riðr;xÞþ jIiðr;xÞ, which can be simulated using 2D axisymmetric

FEM model of single domes. Accordingly, the complex

acoustic coupling between the elements of the array through

the medium can be estimated as ZjiðxÞ ¼Ð

AjPiðj~rj�~ri j;xÞdA

xX0iAj

by

numerical integration. The real part, RjiðxÞ, and the imagi-

nary part, IjiðxÞ, of the acoustic coupling can be seen as the

damping force and inertial loading applied to dome j from the

ith dome, respectively. It is notable that the lumped displace-

ment of every element is a weighted average of the dome dis-

placement over its surface area Xi ¼Ð Di=2

0xiðrÞdr

Ð Di=2

0dr

or

Xi ¼Ð Di=2

0xiðrÞrdr

Ð Di=2

0rdr

and the same normalization should be applied

to calculate every lumped parameter. The effective stiffness,

mass, and piezoelectric actuation force giv can also be esti-

mated from the simple 2D model of each dome. The dynamic

electromechanical model of the array can be constructed in

the matrix form of �½M�Xx2 þ j½ZðxÞ�Xxþ ½K�X ¼ ½G�v.

Consequently, the frequency response of arrays with

various dome mixtures and configurations can be analyzed:

X ¼ ½�Mx2 þ jxZðxÞ þ K��1½G�v.

This model is a quick and effective way of estimating

the array’s first mode’s frequency spectrum. Using this

model, the architecture of a 64 channel 5 MHz linear array

is analyzed and optimized to achieve a wide fractional

bandwidth (>50%) desired for a good imaging axial resolu-

tion. Nevertheless, analysis of higher modes requires a full

3D finite element model analysis of the array, which can be

computationally intensive. Fig. 4 shows the optical image

and also the scanning electron microscope image of the fab-

ricated device exploiting 5 different dome sizes with cav-

ities ranging from 74 to 90 lm diameter. The pitch, active

elevation, number of dome elements, capacitance and elec-

trical impedance of each channel are 400 lm, 2.3 mm,

57 domes, 1 nF, and 45X/� 56�, respectively. The bandwidth

broadening and sensitivity improvement have been achieved

through a number of complex interaction mechanisms. Similar

to a coupled-resonator optical waveguide25 or a microstrip loop

resonator,26,27 the strong coupling between elements caused by

the acoustic loading induces degenerate mode splitting for simi-

lar dome sizes. In addition, using multiple domes in a row

increases the effective width of the channel to approximately

one wavelength. It causes a significant improvement in the

effective acoustic impedance and the real acoustic power,

which results in further increase of sensitivity and bandwidth.

Finally, exploiting different dome sizes in an optimized order

and architecture creates a wide composite bandwidth by over-

lapping the frequency spectrum of multiple resonators similar

to a 5th order linear filter.

The frequency spectrum of the transducer has been meas-

ured by a swept-sine network analyzer (Agilent E5061B-3L5)

to achieve a better accuracy especially at low-amplitude

regions of the signal. The low-frequency (LF) output of the

gain-phase test port was amplified by a broadband solid-state

FIG. 4. Optical and SEM images of a 64 channel 5 MHz linear array exploit-

ing 5 different dome sizes with cavities ranging from 74 to 90 lm diameter.

FIG. 5. (Top) S11 measurements show 5 distinct high-Q peaks per mode in

air which become merged and well-damped in water indicating a wide

achieved bandwidth and strong acoustic coupling. (Bottom) Transmit pres-

sure sensitivity vs. frequency measured at 30 mm by hydrophone.

253101-3 Hajati et al. Appl. Phys. Lett. 101, 253101 (2012)

class A power amplifier (325LA by E&I) to excite the trans-

ducer. The realized underwater acoustic pressure was meas-

ured by a hydrophone (HGL-1000 by ONDA Corp.) followed

by a 20 dB pre-amplifier (AH-2010 by ONDA Corp.) and was

fed back to the network analyzer. As shown in Fig. 5, the S11

parameter and the pressure frequency response of the 5 MHz

transducer (which contains 5 different dome sizes) were

measured in air and in water at 30 mm depth. The integration

of 5 different dome sizes in the transducer design results in 5

distinct pairs (one per mode) of high-Q peaks of the S11 pa-

rameter measured in air. These peaks merge together when

measured in water as a result of the strong acoustic damping

caused by water also the resulting acoustic coupling between

the domes and generates a response similar to a wide-band

resonator. Exploiting the 5th order filter design, the transducer

generates two wide bands: 1st and 2nd modes centered at

5 MHz (�3 dB bandwidth of 55%) and 10.6 MHz (�3 dB

bandwidth of 34%) and a peak sensitivity of 85 kPa/V and

115 kPa/V, respectively.

Tissue harmonic imaging (THI) is an imaging technique

which addresses ultrasound limitations like penetration and

resolution.1 Traditionally, the bandwidth of the transducers

has been a limiting factor for THI since the transducer must

be both an efficient transmitter at one frequency and an effi-

cient receiver at the second harmonic. It is clear that the

bandwidth of the conventional transducers (approximately

70%–80% �6 dB fractional bandwidth) is sub-optimal for

harmonic imaging. Unlike the conventional transducers, the

3D-MEMS piezoelectric transducers enable several wide

bands in the frequency response as a result of each mode of

vibration. An efficient THI scheme can be implemented by

exploiting the first band as the transmitter and the second

band which lies at about twice frequency of the first band as

the receiver. Furthermore, each frequency band can be used

as an independent imaging channel, which can be excited by

an excitation waveform at its associated frequency such as a

pulse with the duration of T � 12fi= or chirp waveforms. The

received signal can be analyzed into various channels using

corresponding band-pass filters or by various demodulation

techniques. Consequently, the frame rate can be improved

proportionally to the number of channels.

To measure the LF transmit acoustic pulse in the time-

domain, the transducer was excited by a train of 100 ns

pulses generated by an arbitrary function generator (Tektro-

nix AFG3102) amplified by the 325LA amplifier. By trans-

mitting a pulse width of 100 ns, the 1st mode is exclusively

excited while higher modes, including the 2nd mode, are

suppressed. Fig. 6 depicts the resulting acoustic pulse and its

normalized FFT spectrum. Activating only 4 channels with-

out any mechanical or electrical focusing generates a nega-

tive peak pressure of 1.5 MPa and short pulse duration

(0.57ls) enabling 0.4 mm axial resolution capability. Higher

acoustic pressure up to the maximum FDA-approved me-

chanical index (MI of 1.9) can be easily achieved by focus-

ing 16 to 20 channels per the typical linear imaging practice.

This 5 MHz linear array, as an example of a 3D-MEMS

piezoelectric ultrasound transducer, demonstrates unique

features such as high sensitivity (more than 100 kPa/V), ad-

justable wide-bandwidth frequency response (greater than

55%), CMOS-compatible low transmit voltage (2-20 V), low

electrical impedance (less than 50 X), efficient electrome-

chanical coupling (greater than 45%), and reliable mono-

lithic fabrication. We envision that Clarinet TechnologyTM

can be implemented as a desirable solution in various

medical imaging applications. Exploiting its small form-

factor, high sensitivity, low voltage level, and low imped-

ance well matched to micro-coaxial cables, the transducer

can be incorporated into high performance endoscopy ultra-

sound (EUS) catheters to achieve high quality GI imaging

especially in depicting the gut wall as a series of layers

correlating with histologic features.28 Similarly, it enables

low-voltage and miniaturized transesophageal echocardio-

gram (TEE) and intracardiac echocardiogram (ICE) probes

that can provide superior image quality.29,30 In addition, this

technology is an obvious solution for high-frequency, high

resolution, and miniaturized intravascular ultrasound (IVUS)

to provide accurate quantitative and qualitative information

regarding the lumen and outer vessel wall.31 Besides in-vivoimaging, it can enable high performance, low-power, low-

voltage, and portable 3D/4D sonography32 and an affordable

3D ultrasound stethoscope. Finally, the integration of low-

voltage CMOS chips with the MEMS structure can enable

unforeseen applications.

1R. Cobbold, Foundations of Biomedical Ultrasound (Oxford Univ. Press,

2007).2B. T. Khuri-Yakub, O. Oralkan, and M. Kupnik, IEEE Spectrum 46(5),

44–54 (2009).3B. T. Khuri-Yakub and O. Oralkan, J. Micromech. Microeng. 21(5),

054004 (2011).4P. Zhang, J. Microelectromech. Syst. 21(3), 668–680 (2012).5A. Roest, K. Reimann, M. Klee, T. van Beek, and J. Fraser, U.S. patent

US8203912 (19 June 2012).6M. Lebedev, J. Akedo, and Y. Akiyama, Jpn. J. Appl. Phys. 39, 5600–

5603 (2000).7T. Iijima and K. Kunii, Jpn. J. Appl. Phys. 40, 5740–5742 (2001).8H. J. Nam, Y. S. Kim, S. M. Cho, C. S. Lee, J. U. Bu, J. W. Hong, and Z.

G. Khim, J. Appl. Phys. 41, 7153–7157 (2002).9T. L. Ren, L. T. Zhang, J. S. Liu, L. T. Liu, and Z. J. Li, Microelectron.

Eng. 66, 683–687 (2003).10S. Trolier-McKinstry and P. Muralt, J. Electroceram. 12(1–2), 7–17

(2004).11P. Muralt, IEEE Ultrason. Ferroelectr. Freq. Control 52(12), 2276–2288

(2005).12J. Baborowski, N. Ledermann, and P. Muralt, “Piezoelectric microma-

chined transducers (PMUT’s) based on PZT thin films,” Proceedings ofthe Ultrasonics Symposium (2002), Vol. 2, pp. 1051–1054.

13Z. H. Wang, W. Zhu, O. Kiang Tan, Ch. Chao, H. Zhu, and J. Miao, Appl.

Phys. Lett. 86, 033508 (2005).

FIG. 6. Measured time-domain acoustic pulse and its corresponding FFT

spectrum measured at 9 mm depth by hydrophone. 4 channels are excited by

100 ns pulse trains.

253101-4 Hajati et al. Appl. Phys. Lett. 101, 253101 (2012)

14M. Klee, H. W. Brand, E. Krafczyk, H. P. Loebl, and P. K. Bachmann,

U.S. patent 6,548,937 B1 (15 April 2003).15M. Klee, H. Boots, B. Kumar, W. Keur, M. D. Wild, P. Dirksen, K.

Reimann, O. Wunnicke, C. Renders, H. Van Esch et al., J. Acoust. Soc.

123(5), 3376 (2008).16M. Klee, R. Mauczok, C. van Heesch, H. Boots, M. de Wild, B. op het

Veld, W. Soer, G. Schmitz, and M. Mleczko, IEEE Ultrason. Symp.,Orlando, FL, 18–21 Oct. (2011), pp. 196–199.

17Q. Zhou, J. M. Cannata, R. Meyer, D. J. Van Tol, S. Tadigadapa, W. J.

Hughes, K. K. Shung, and S. Trolier-McKinstry, IEEE Ultrason. Ferroe-

lectr. Freq. Control 52(3), 350–357 (2005).18Q. Zhou, S. Lau, D. Wu, and K. Shung, Prog. Mater. Sci. 56(2), 139–174

(2011).19D. E. Dausch, K. H. Gilchrist, J. B. Castellucci, and D. R. Chou, IEEE

Ultrason. Symp., New York, NY, 2007, pp. 1053–1056 .20D. E. Dausch, K. H. Gilchrist, J. R. Carlson, J. B. Castellucci, D. R. Chou,

and O. T. von Ramm, Ultrasonics Symposium (IUS), 2010 IEEE, SanDiego, CA, 2010, pp. 451–454.

21B. H. Chen, L. Wu, M. C. Chure, Y. C. Chen, in Symposium onPiezoelectricity, Acoustic Waves and Device Applications (2010),

pp. 310–314.

22A. Hajati and S. G. Kim, Appl. Phys. Lett. 99, 083105 (2011).23Y. Hishinuma, Y. Li, J. Birkmeyer, T. Fujii, T. Naono, and T. Arakawa,

Nanotechnology 2, 137–140 (2012).24J. S. G. Hong and M. J. Lancaster, Microstrip Filters for RF/Microwave

Applications (Wiley, 2001).25H. C. Liu and A. Yariv, Opt. Express 19(18), 17653–17668 (2011).26J. S. Hong and M. J. Lancaster, IEEE Microw. Guid. Wave Lett. 5(11),

371–372 (1995).27I. Wolff, Electron. Lett. 8(12), 302–303 (1972).28J. P. Mcgahan and B. B. Goldberg, Diagnostic Ultrasound, 2nd ed.

(Informa Healthcare, New York. 2008).29A. Chenzbraun, F. J. Pinto, and I. Schnittger, Clin. Cardiol. 17, 438–444

(1994).30G. Rigatelli and M. Hijazi, J. Invasive Cadiol. 18(5), 225–233 (2006).31C. V. Bourantas, K. K. Naka, S. Garg, S. Thackray, D. Papadopoulos, F.

M. Alamgir, A. Hoye, and L. K. Michalis, Echocardiography 27(10),

1282–1290 (2010).32R. Lazebnik and T. Desser, “Clinical 3D ultrasound imaging: beyond

obstetrical applications,” Diagnostic Imaging (2007), pp. 1–5, Reference

No. A07019012, http://imaging.ubmmedica.com/dimag/cme/articles/ultra-

sound-clinical-3d-imaging-beyond-obstetrical-applications.pdf.

253101-5 Hajati et al. Appl. Phys. Lett. 101, 253101 (2012)