capacitive micromachined ultrasonic transducers with ...membranes exhibits higher fill factor...

0885–3010/$25.00 © 2009 IEEE

136 IEEE TransacTIons on UlTrasonIcs, FErroElEcTrIcs, and FrEqUEncy conTrol, vol. 56, no. 1, JanUary 2009

Abstract—Capacitive micromachined ultrasonic transduc-ers (CMUTs) featuring piston-shaped membranes (piston CMUTs) were developed to improve device performance in terms of transmission efficiency, reception sensitivity, and frac-tional bandwidth (FBW). A piston CMUT has a relatively flat active moving surface whose membrane motion is closer to ideal piston-type motion compared with a CMUT with uni-formly thick membranes (classical CMUT). Piston CMUTs with a more uniform surface displacement profile can achieve high output pressure with a relatively small electrode separa-tion. The improved device capacitance and gap uniformity also enhance detection sensitivity. By adding a center mass to the membrane, a large ratio of second-order resonant frequency to first-order resonant frequency was achieved. This improved the FBW. Piston CMUTs featuring membranes of different geometric shapes were designed and fabricated using wafer bonding. Fabricating piston CMUTs is a more complex process than fabricating CMUTs with uniformly thick membranes. However, no yield loss was observed. These devices achieved ~100% improvement in transduction performance (transmis-sion and reception) over classical CMUTs. For CMUTs with square and rectangular membranes, the FBW increased from ~110% to ~150% and from ~140% to ~175%, respectively, compared with classical CMUTs. The new devices produced a maximum output pressure exceeding 1 MPa at the transducer surface. Performance optimization using geometric membrane shape configurations was the same in both piston CMUTs and classical CMUTs.

I. Introduction

capacitive micromachined ultrasonic transducers (cMUTs) have emerged as a promising technology

for applications in medical ultrasound. compared with

piezoelectric transducers in medical applications, cMUTs reported so far possess broader fractional bandwidth (FBW) but poorer transmission (TX) and reception (rX) efficiencies [1]. In its simplest form, a cMUT is made of a mass with a flat surface and a spring actuated by an elec-trostatic force, as shown in Fig. 1(a). Ideally the membrane displacement is uniform across the device area, maximiz-ing the volume displacement and hence also transduction (TX and rX) efficiencies for a specific vacuum cavity height. nearly all fabricated cMUTs reported in the lit-erature [2]–[6] featured membranes of uniform thickness. We reported on methods to improve cMUT performance by optimizing membrane geometry in the lateral mem-brane directions for more pistonlike membrane motion [7]. There is still a performance trade-off between transduc-tion efficiency and FBW when designing cMUTs with uniformly thick membranes (classical cMUTs). The room for improvement of such a cMUT is constrained because the mass and the spring constant are closely coupled. It is difficult to change one without substantially affecting the other, thus limiting design flexibility.

In this paper, we report on results for cMUTs featur-ing piston-shaped membranes (piston cMUT) as shown in Fig. 1(c). The piston cMUT is a cMUT with a relatively flat moving surface. consequently, the membrane motion is closer to ideal piston-type motion compared with a clas-sical cMUT with uniform membrane thickness. a center mass that could be made of silicon is added to each mem-brane, thus altering the membrane geometry in the verti-cal direction to have more ideal piston membrane motion. The motivation for this approach is to improve the device performance without substantially increasing the fabrica-tion process complexity or reducing device yield. The con-cept of cMUTs with piston-shaped membranes has been reported in a U.s. patent [8], and as a conference publica-tion in 2005 [9]. other researchers have also reported on work based on similar concepts for cMUTs [10]–[12] and other types of MEMs devices [13].

The piston cMUT offers advantages when compared with a classical cMUT. Most noticeably, because the cen-ter mass can be controlled independently of the membrane size or thickness, there is more design flexibility. specifi-cally, the spring constant, which is determined by both membrane and piston, and the membrane mass, which is mostly due to the center piston, are largely decoupled

Capacitive Micromachined Ultrasonic Transducers with Piston-Shaped

Membranes: Fabrication and Experimental Characterization

yongli Huang, Xuefeng Zhuang, Student Member, IEEE, Edward o. Hæggstrom, a. sanli Ergun, ching-Hsiang cheng, and Butrus T. Khuri-yakub, Fellow, IEEE

Manuscript received March 21, 2007; accepted June 18, 2008. The authors would like to acknowledge the financial support of the office of naval research and the national Institutes of Health. Xuefeng Zhuang was supported by a Weiland Family stanford Graduate Fellowship. dr. Hæggström acknowledges the Wihuri-Foundation and the academy of Finland for financial support.

y. Huang is with Kolo Technologies Inc., san Jose, ca.X. Zhuang and B. T. Khuri-yakub are with Edward l. Ginzton

laboratory, stanford University, stanford, ca (e-mail: xzhuang@ stanford.edu).

E. o. Hæggstrom is with the Electronics research Unit, University of Helsinki, Helsinki, Finland.

a. s. Ergun is with siemens corporate research, Mountain View, ca.

c.-H. cheng is with the research Institute of Innovative Products and Technologies, Hong Kong Polytechnic University, Hong Kong, china.

digital object Identifier 10.1109/TUFFc.2009.1013

from one another. compared with classical cMUTs, pis-ton cMUTs feature more uniform membrane displacement profiles so that they require smaller electrode separations to achieve the same or larger maximum output pressure. Piston cMUTs with a smaller electrode separation there-fore can achieve a higher intracavity electrostatic field for a given applied voltage, larger device capacitance, and in-creased capacitance change in reception. These improve-ments translate into better TX output pressure and rX sensitivity. another added benefit of the piston cMUT is that the higher-order resonant frequencies of the mem-brane can be designed to be far away from the funda-mental resonance frequency, hence eliminating the adverse impact of the second-order resonant frequency mode on the FBW.

In the following sections, we explain the piston cMUT concept and provide a brief description of the design pro-cess. after that, we introduce a fabrication process based on silicon-on-insulator (soI) wafer bonding technology to make piston cMUTs. We finally present characterization results and demonstrate the performance improvements achieved by a piston cMUT compared with a classical cMUT.

II. design and Fabrication

A. Design

changing the membrane thickness profile adds degrees of freedom to cMUT designs. Herein, a membrane with the center part thicker than the edge is called a piston membrane. The thick center part of the membrane is called a piston. a cMUT with piston membranes is called a piston cMUT.

Improving the maximum output pressure is a major goal in cMUT designs because, compared with PZT tech-nology, output pressure is arguably the only major weak-ness of the cMUT technology. For a cMUT design, elec-trode separation is a key parameter for determining the output pressure and sensitivity for a given applied voltage. The optimal design of electrode separation is a trade-off between the electrical field intensity, which determines the reception sensitivity, and the space needed for sufficient membrane displacement to generate high output pressure. a uniform membrane displacement profile is desired to improve both the output pressure and the reception sen-sitivity. The reason is that, for a cMUT with uniform displacement profile, the peak membrane displacement is

smaller than that of a classical cMUT for the same maxi-mum output pressure. Therefore, the electrode separation can be designed to be narrower. a cMUT with a smaller gap shows a larger capacitance change for a given elec-trode displacement due to impinging acoustic pressure, which improves the sensitivity.

The conclusion that piston cMUTs have higher trans-duction efficiency than classical cMUTs can also be reached from the cMUT equivalent circuit models [16], [17]. compared with a classical cMUT with the same cen-ter frequency, a piston cMUT design features a smaller electrode separation and a larger equivalent spring con-stant and mass to accommodate a stronger electrical field between the 2 electrodes. In the equivalent circuit, the cMUT membrane and the acoustic medium are connected in series. at the center frequency, the impedance of the medium is much larger than that of the cMUT membrane, which is determined by its equivalent spring constant and mass. Because the internal acoustic loss is small for a cMUT, the membrane impedance approaches zero at the center frequency in the equivalent circuit models regard-less of the values of its equivalent spring constant or mass. Therefore, the membrane velocity at the center frequency is largely independent of the membrane impedance. In transmission, the piston cMUT, with a higher equivalent spring constant, can sustain a higher intracavity electrical field than the classical cMUT. In the equivalent circuit analogy, this translates into a higher output pressure into the immersion medium. In reception, the membrane ve-locity generated by a certain impinging acoustic wave is similar for any cMUT with the same center frequency. With the piston cMUT, however, the electromechanical coupling factor is higher than that of a classical cMUT (due to the higher equivalent spring constant, which can withstand a higher intracavity electrical field), and more current is generated for improved receive sensitivity.

The smaller electrode separation can also increase the total capacitance of the cMUT design. resulting in an im-pedance comparable to that of piezoelectric transducers. consequently, the interface of the new cMUT transducer to electronics is easier than that of the existing cMUT transducers. The design also increases the fill factor and reduces the parasitic capacitance, compared with corre-sponding cMUTs with uniform membranes. The fill factor increase is due to the fact that the piston cMUT device can carry larger membranes than the corresponding clas-sical device. The thicknesses of the intercavity wall, which is largely determined by fabrication, are the same for clas-sical and piston cMUTs. Therefore, a cMUT with larger

137HUanG ET al.: cmuts with piston-shaped membranes: fabrication and experimental characterization

Fig. 1. (a) an ideal cMUT, (b) a classical cMUT, and (c) a piston cMUT.

membranes exhibits higher fill factor compared with a de-vice with smaller membranes. The parasitic capacitance of a piston cMUT is lower than that of a corresponding classical cMUT also because of the relative smaller sup-porting wall area.

Fig 1 HErE. The piston cMUT design also aims to im-prove the device bandwidth. The membrane of any cMUT supports a fundamental and numerous higher-order reso-nant modes. Usually at a frequency close to the second resonance mode of the membrane, the average membrane velocity is close to zero, resulting in a minimal average membrane displacement. small average membrane dis-placement means that the cMUT is inefficient at this fre-quency, i.e., a frequency cut off. Therefore, another design criterion for a piston cMUT is to maximize the separation between the fundamental (f1) and the second-order reso-nant1 (f2) frequencies. The separation is evaluated using the ratio of the second-order resonant frequency to the fundamental-order resonant frequency of the membrane (f2/f1). For comparison, the f2/ f1 ratio of various mem-brane configurations is investigated.

ansys (ansys 8.0, ansys Inc., canonsburg, Pa) simulations similar to [14] were used to investigate how the f2/ f1 ratio could be improved through membrane de-sign. First, we studied how the membrane thickness af-fects the f2/ f1 ratio. Fig. 2(a) and 2(b) show simulation results of circular/square membranes of different thick-nesses with a fixed resonant frequency and membrane size, respectively. In the first case, the membrane size was varied with different membrane thicknesses to maintain the resonant frequency of the membrane fixed at 3 MHz.

In the second case, the membrane size was fixed and the membrane thickness was varied. In both cases, the f2/ f1 ratios slightly decreased with increasing membrane thick-ness. However, the decrease was so small that changing the membrane thickness seemed not to provide an efficient way to improve the f2/ f1 ratio.

Instead of changing the membrane thickness, the f2/ f1 ratios of the 7 different membrane shapes were investigat-ed by means of ansys simulations (Fig. 3). These mem-brane shapes were selected because they cover efficiently the transducer surface, leaving little unutilized space. The thickness of each membrane was 1 μm and the size of each membrane shape was chosen so that its resonant fre-quency was fixed at 3 MHz. The rectangular membrane anchored along its edges and the triangle membrane an-chored at its corners showed great improvement over the other membrane shapes.

The f2/ f1 ratio was simulated as a function of pis-ton thickness for the piston cMUT (Fig. 4). Fig. 5 shows considerable improvement in the f2/ f1 ratio with increas-ing piston thickness. Moreover, the piston part adds mass to the membrane, which allowed increasing the equiva-lent spring constant for a fixed resonance frequency. The membrane flatness also improved with the piston design as compared with the conventional design. The flatness was defined as average membrane deflection over maxi-mum membrane deflection when the membrane was under uniform pressure loading. The way by which the flatness improves the device performance is discussed in [7]. The above mentioned factors should improve the cMUT trans-duction and the FBW simultaneously.

as with classical cMUTs, the membrane geometries in a piston cMUT can be configured in several fashions, such as square, circular, rectangular, and tent. a tent membrane is a membrane supported by isolated posts to improve the transducer fill factor. The piston geometries can vary in a similar manner. In this study, all 4 variations of membrane geometries were included in the mask layout to allow comparison between them. To ensure a fair com-


Fig. 2. simulated, in vacuum, f2/f1 ratios of a classical circular and a square membrane as a function of membrane thickness. case a: fixed 3 MHz resonance frequency. case b: fixed size 45 μm radius for circular membrane and 50 μm side length for square membrane.

Fig. 3. simulated, in vacuum, f2/f1 ratios of seven different membrane shapes with uniform membrane thickness of 1 μm.

1 The second resonant frequency means the second resonant frequency with symmetry of actuation. This is the first minimum of the average velocity in a harmonic simulation. Usually it is not the second Eigen mode of the membrane.

parison, the membrane and piston parameters were chosen in such a way that the fundamental resonances were the same for all designs.

To compare the performances side-by-side, classical cMUTs comprising square-, circular-, rectangular-, and tent-shaped membranes were designed and fabricated. In designing these cMUTs, the targeted first-resonant fre-quencies were the same as those of the piston cMUTs. To enable fair performance comparison, these devices fea-tured the same element sizes (5500 μm by 500 μm) as the piston cMUTs. Tables I and II summarize the key param-eters for the cMUTs fabricated in this study.

B. Fabrication

single crystal silicon was chosen as the material for the membrane and the piston. The advantages of single crystal silicon membranes are discussed in [6]. From a pro-cessing compatibility standpoint, a cMUT can easily be fabricated to have a silicon membrane and a piston.

The piston cMUT is fabricated using a silicon fusion bonding technique similar to that reported in [6] but with modifications. Two soI wafers, with device layer thick-ness of 4.5 μm and 1.0 μm, respectively, as well as one prime silicon wafer are needed to build the piston cMUT. The first soI wafer, with a 4.5-μm device layer, is used to form the piston layer. The second soI wafer, with the 1.0-μm device layer provides the membrane layer.

Fig. 6 illustrates the process flow. on the device layer (4.5 μm) of the first soI wafer, a layer of thermal oxide is grown at 1000°c (a). This oxide serves as an etch stop for later processing stops. next, this soI wafer is fusion bonded to the second (1.0 μm) soI wafer, with the device layers facing each other (b). The bonding is followed by an oxygen annealing step at 1100°c for one hour. The handle


Fig. 4. drawings of (a) a 70-μm radius circular piston membrane, and (b) a 120-μm square piston membrane of a piston cMUT with non-uniform membrane thickness.

Fig. 5. simulated, in vacuum, f2/f1 ratios for membrane configurations shown in Fig. 4, as a function of piston thickness tp.

TaBlE I. Piston cMUT Parameters.

design 1 design 2 design 3 design 4

shape dim. (μm) shape dim. (μm) shape dim. (μm) shape dim. (μm)

Membrane square 120 × 120 × 1 rect. 540 × 100 × 1 square 130 × 130 × 1 Tent 95 × 95 × 1Piston square 70 × 70 × 4.5 rect. 500 × 60 × 4.5 circ. r = 45, t = 4.5 rect. 25 × 95 × 4.5Gap (μm) 0.44 0.44 0.44 0.44

TaBlE II. classical cMUT Parameters.

design 5 design 6 design 7 design 8

shape dim. (μm) shape dim. (μm) shape dim. (μm) shape dim. (μm)

Membrane square 88 × 88 × 1 rect. 540 × 73 × 1 square 102 × 102 × 1 Tent 70 × 70 × 1Gap (μm) 0.85 0.85 0.85 0.85

wafer of the piston soI wafer is removed by mechani-cal grinding and heated tetramethylammonium hydroxide (TMaH) etching. The buried oxide layer is removed in 6:1 buffered oxide etchant (BoE) (c). at this point, the necessary membrane and piston layers for the final piston cMUT structure have been formed.

next, a thermal oxide layer is grown on the device layer (d). The thickness of this oxide layer defines the vacuum cavity height between the piston and the rigid bottom electrode. The oxide is patterned and etched to form the cavity area (e). The subsequent step defines the piston and the cavity patterns using the same photo resist hard mask. The silicon is removed using plasma etching and the oxide is removed using 6:1 BoE etching (f). The self-aligning nature of these 2 steps eliminates any alignment errors between the relative positions of the piston and the membrane. such alignment errors can cause unpredictable higher-order resonant modes that degrade device perfor-mance.

The patterned wafer is then fusion bonded to a prime quality silicon wafer with a pre-grown thermal oxide layer serving as electrical insulation between the 2 electrodes of the piston cMUT (g). again, the wafer is grinded and etched back to the silicon device layer (h).

Electrical contact to the silicon pistons provides en-hanced electrostatic force intensity in the cavity. These contacts are coupled to the pistons by etching vias through the membrane and oxide layers between the membrane and the piston. Vias are also etched to access the bot-tom electrode (i). next, a 1.0-μm-thick aluminum layer is sputtered and patterned to form the top and bottom elec-trodes (j). The last step is to make an electrical isolation between the cMUT elements by means of silicon plasma etching (k).

Fig. 7 shows photographs of finished piston cMUTs with different membrane and piston patterns. Perspective views of the piston membranes are also shown in the fig-ures. The surface profiles obtained by an optical surface


Fig. 6. Piston cMUT fabrication steps.

profiler (WyKo, nT3000, Veeco Instruments, Inc., Wood-bury, ny) of both piston cMUTs and classical cMUTs are shown in Fig. 8 for comparison. These optical pictures show that the piston cMUT membrane is flatter.

III. results

In this section, the detailed characterization of a piston cMUT (design 1) with a square membrane (120 μm by 120 μm by 1 μm) and a square piston (70 μm by 70 μm by 4.5 μm) in the membrane center shown in Fig. 7(a) is presented. These results are compared with those of a classical cMUT (design 5) carrying square membranes (88 μm by 88 μm by 1 μm). characterization results for piston cMUTs with various other membranes and piston configurations follow. all tested cMUT elements were of the same size (5500 μm by 500 μm). The cMUTs were all characterized in conventional (noncollapse) mode opera-tion.

A. Reception Sensitivity

The reception sensitivity of the cMUTs immersed in oil was measured. a Panametrics V109 flat-focus trans-

ducer (Panametrics, Inc., Waltham, Ma) with a 12.5-mm circular aperture was used as the transmitter and the cMUTs as receiver. The transmitter and the receiver were placed 86.1 mm apart in an oil tank. a 2.5-MHz, 10 Vp-p, 30-cycle sinusoidal wave train was applied to the trans-mitter, and the signals from the cMUT were recorded by the oscilloscope with a 50-Ω input impedance. Fig. 9 shows the sensitivity of the piston cMUT and of the clas-sical cMUT as a function of applied bias voltage. In both cases, the receive sensitivity increased as the bias voltage increased because of the improved transduction efficiency [15]. The sensitivity of the piston cMUT and of the clas-sical cMUT was 8.25 mV/kPa at 55 Vdc (~85% of the collapse voltage) and 4.23 mV/kPa at 40 Vdc (~85% of the collapse voltage), respectively. These values are cor-rected for diffraction and attenuation losses.

B. Transmission Efficiency

In the oil transmission test, cMUTs were used as the acoustic source, and a hydrophone (PZT-Z44–0400, onda corporation, sunnyvale, ca) was used to detect, on the acoustic axis, the ultrasonic pressure generated by the cMUT at a distance of 13.7 mm from the cMUT surface. a 2.5-MHz, 10 Vp-p, 30-cycle sinusoidal wave train was


Fig. 7. Photographs of fabricated piston cMUTs and schematic pictures of their cell geometries.

applied to the cMUT. The signal from the hydrophone was recorded by the oscilloscope with 50-Ω input imped-ance. Fig. 10 shows the transmission efficiency of the pis-ton cMUT and of the classical cMUT as a function of bias voltage. In both cases, the output pressure improved as the bias voltage increased. The output pressure of the piston cMUT and that of the classical cMUT were 5.75 kPa/Vac at 55 Vdc (~85% of the collapse voltage) and 3.15 kPa/Vac at 40 (~85% of the collapse voltage), re-spectively. These values are corrected for diffraction and attenuation losses.

C. Pulse-Echo Measurements

The pulse-echo tests were performed against a 5-cm-thick aluminum reflector. The separation between the

cMUT and reflector was 1.0 cm. a 3-V, 10-MHz positive square pulse was applied to the cMUT. Fig. 11 shows the received pulses recorded by the oscilloscope with the pis-ton cMUT and with the classical cMUT. Fig. 12 shows the spectral comparison between the 2 devices. The low frequency cut-offs of the piston cMUT and of the clas-sical cMUT are nearly identical (~1 MHz), but the pis-ton cMUT has a much higher high-frequency cut-off (~6 MHz) than the classical cMUT (3.2 MHz). This result confirms the theoretical prediction presented in the previ-ous section of the paper. The FBW of the piston cMUT and the classical cMUT are ~150% and ~110%, respec-tively. These values are not corrected for the immersion medium diffraction or attenuation but are corrected for the spectrum of the input signal.


Fig. 8. surface profile comparison between classical cMUTs and piston cMUTs carrying differently shaped membranes. In the online version of each picture, maximum displacement is indicated by blue and zero displacement is marked red. The profiles are obtained in ambient air pressure.

Fig. 9. Ultrasonic reception, in oil, by a cMUT as a function of bias voltage. a PZT transducer with a constant output was used as the acoustical source. (a) Piston cMUT (design 1), (b) classical cMUT (design 5).

D. Large-AC Actuation

The large ac experiments were carried out to establish the maximum output pressure that the fabricated piston cMUTs can deliver. The setup for this experiment was the same as the one in the TX experiments. Generally the output pressure is proportional to the membrane displace-ment. For single-pulse actuation, under 0 Vdc, the cMUT has the largest initial separation between 2 electrodes, thus the membrane has the maximum displacement distance available, i.e., the largest membrane displacement can be achieved. Therefore, under this condition, the highest out-put pressure of single pulse actuation can be realized. This argument is confirmed by an analytical model based on an analysis for the dynamic behavior of cMUTs [16]. Fig. 13 shows the transmitted pulse and its spectrum measured in a large-ac transmission test. The maximum output pressure generated on the device surface was 1.28 MPa (peak-to-peak) at the center of the output spectrum. The

input was a 160 V negative spike applied from a Panamet-rics pulser (Model 5058Pr, Panametrics Inc., Waltham, Ma) to the unbiased transducer. The output value is not corrected for the medium diffraction or attenuation, but is corrected for the spectrum of the input signal. The maxi-mum output pressure of a classical cMUT can also be obtained from its large ac response [16].

E. Comparison of Different Piston CMUT Designs

Pulse-echo tests were performed with piston cMUTs carrying square- (design 1), rectangular- (design 2), and tent-shaped (design 4) membranes. Their spectra are shown in Fig. 14. These spectra are corrected for the spec-trum of the input signal only. compared with classical cMUTs with different membrane shapes [7], the results are similar. The rectangular cMUT exhibits the widest bandwidth while the tent cMUT has a slower high fre-quency roll-off than either of the other 2 designs. This


Fig. 10. Ultrasonic transmission, in oil, by cMUTs received by a broadband hydrophone as a function of bias voltage. (a) Piston cMUT (design 1), (b) classical cMUT (design 5).

Fig. 11. received, in oil, pulse-echo signals against a flat steel reflector recorded by an oscilloscope: (a) piston cMUT (design 1), (b) classical cMUT (design 5). Bias voltages represent the same fraction of the collapse voltage for the respective designs.

indicates that, because the observed phenomena are iden-tical for both the piston cMUT and the classical cMUT, the device performance optimization by means of mem-brane shaping [7] is the same for a piston cMUT and for a classical cMUT.

IV. conclusions

cMUTs featuring a piston-shaped membrane have been demonstrated. although the fabrication process is more complex than that of a classical cMUT, the device yield is similar when a wafer bonding technique is used to fabricate the cMUTs. characterization results show com-pelling evidence that a piston cMUT exhibits superior performance when compared with a classical cMUT, with regard to output power, reception sensitivity, and FBW.

We achieved an output pressure of 1.28 MPa at the trans-ducer surface of a piston cMUT by applying a large ac actuation voltage only. The piston cMUT performance was improved by optimizing the membrane geometry in a similar fashion to that of a classical cMUT. The piston cMUT is an important step in narrowing the performance gap between cMUTs and piezoelectric transducers with regard to transmission and reception sensitivities, while it further advances the advantage in FBW.

references

[1] d. M. Mills, “Medical imaging with capacitive micromachined ultra-sound transducer (cMUT) arrays,” in Proc. IEEE Ultrasonics Sym-posium, 2004, pp. 384–390.

[2] a. G. Bashford, d. W. schindel, and d. a. Hutchins, “Microma-chined ultrasonic capacitance transducers for immersion applica-tions,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 45, no. 2, pp. 367–375, Mar. 1998.

[3] E. cianci, V. Foglietti, G. caliano, and M. Pappalardo, “Microma-chined capacitive ultrasonic transducers fabricated using silicon on insulator wafers,” Microelectron. Eng., vol. 61–62, pp. 1025–1029, 2002.

[4] P. c. Eccardt, K. niederer, T. scheiter, and c. Hierold, “surface mi-cromachined ultrasound transducers in cMos technology,” in Proc. IEEE Ultrasonics Symp., 1996, pp. 959–962.

[5] X. Jin, I. ladabaum, and B. T. Khuri-yakub, “The microfabrication of capacitive ultrasonic transducers,” J. Microelectromech. Syst., vol. 7, pp. 295–302, 1998.

[6] y. Huang, a. s. Ergun, E. Haeggstrom, M. H. Badi, and B. T. Khuri-yakub, “Fabricating capacitive micromachined ultrasonic transducers with wafer-bonding technology,” J. Microelectromech. Syst., vol. 12, pp. 128–137, 2003.

[7] y. Huang, E. o. Haeggstrom, X. Zhuang, a. s. Ergun, and B. T. Khuri-yakub, “optimized membrane configuration improves cMUT performance,” in Proc. IEEE Ultrasonics Symp., 2004, pp. 505–508.

[8] B. Khuri-yakub, y. Huang, and a. s. Ergun, “Micromachined ultra-sonic transducers and method of fabrication,” U.s. Patent 6,958,255, 2005.

[9] y. Huang, E. o. Haeggstrom, X. Zhuang, a. s. Ergun, and B. T. Khuri-yakub, “capacitive micromachined ultrasonic transducers


Fig. 12. spectrum comparison of pulse-echo signals shown in Fig. 11 be-tween a piston cMUT (design 1) and a classical cMUT (design 5).

Fig. 13. received signal by the hydrophone, shown in frequency and time domain, during the transmission test in oil. The transmitter was a piston cMUT (design 1).

Fig. 14. Pulse-echo spectra, in oil against a flat steel reflector, of piston cMUTs with square (design 1), rectangular (design 2), and tent (design 4) membranes. The input signal is a single 3 V unipolar square pulse and the piston cMUT is biased at 50 V.

(cMUTs) with piston-shaped membranes,” in Proc. IEEE Ultrason-ics Symp., 2005, pp. 589–592.

[10] s. Zhou, P. reynolds, and J. a. Hossack, “Improving the perfor-mance of capacitive micromachined ultrasound transducers using modified membrane and support structures,” in Proc. IEEE Ultra-sonics Symp., 2005, pp. 1925–1928.

[11] M. n. senlik, s. olcum, and a. atalar, “Improved performance of cMUT with nonuniform membranes,” in Proc. IEEE Ultrasonics Symposium, 2005, pp. 597–600.

[12] J. G. Knight and F. l. degertekin, “capacitive micromachined ultra-sonic transducers for forward looking intravascular imaging arrays,” in Proc. IEEE Ultrasonics Symp., 2002, vol. 2, pp. 1079–1082.

[13] B. W. chui, y. Hishinuma, r. Budakian, H. J. Mamin, T. W. Ken-ny, and d. rugar, “Mass-loaded cantilevers with suppressed high-er-order modes for magnetic resonance force microscopy,” in Proc. Transducers 2003, pp. 1120–1123.

[14] G. G. yaralioglu, s. a. Ergun, and B. T. Khuri-yakub, “Finite-ele-ment analysis of capacitive micromachined ultrasonic transducers,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 52, no. 12, pp. 2185–2198, dec. 2005.

[15] a. s. Ergun, G. G. yaralioglu, and B. T. Khuri-yakub, “capacitive micromachined ultrasonic transducers: Theory and technology,” J. Aerosp. Eng., vol. 16, pp. 76–84, 2003.

[16] I. ladabaum, X. Jin, H. T. soh, a. atalar, and B. T. Khuri-yakub, “surface micromachined capacitive ultrasonic transducers,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 45, no. 3, pp. 678–690, 1998.

[17] a. lohfink and P. c. Eccardt, “linear and nonlinear equivalent cir-cuit modeling of cMUTs,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 52, no. 12, pp. 2163–2172, 2005.

Yongli Huang received the B.s. and M.s. de-grees in physics from Fudan University in 1987 and 1990, respectively, and the M.s. degree in electrical engineering from University of Hawaii at Manoa in 1996 and the Ph.d. degree in electrical engineering from stanford University in 2005.

He worked as a research associate in the de-partment of Material science and Engineering at University of Electronic science and Technology of china from 1990 to 1992. He joined siTek Inc., a subsidiary of BEI Electronics Inc., campbell,

ca, in 1997 as a member of the technical staff and was promoted to a principal engineer in 1999. He is currently with Kolo Technologies Inc., san Jose, ca, to develop micromachined ultrasonic transducers. His research interests include MEMs technology, sensors and actuators, micromachined ultrasonic devices, inertial sensors, and optical devices.

Xuefeng (Steve) Zhuang (s’02) received the B.s. degree from louisiana state University, Ba-ton rouge, la, in 2002, and the M.s. degree from stanford University, Palo alto, ca, in 2004, both in electrical engineering. He is currently pursuing a Ph.d. degree in electrical engineering at stan-ford University. His research interests include mi-cromachined sensors and actuators. His current research focuses on the new fabrication and pack-aging techniques for capacitive micromachined ultrasonic transducers. He won the lsU Univer-

sity Medal and the Mclaughlin dean’s Medal (2002). He is a stanford Graduate Fellow (2002). He is a student member of the IEEE and sPIE.

Edward Hæggström received the d.sc. degree in applied physics from the University of Helsinki, Finland, in 1998, and the MBa degree in innova-tion management from the Helsinki University of Technology in 2001. He was a visiting scholar at E. l. Ginzton laboratory, stanford University, Palo alto, ca, in 2002 and 2003, and a project leader at cErn of switzerland in 2004 and 2005. He is currently a professor of electronics in the

department of Physics, University of Helsinki. His principal research interests are in ultrasonics.

Arif Sanli Ergun was born in ankara, Turkey, in 1969. He received his B.sc., M.sc., and Ph.d. degrees in 1991, 1994, and 1999, respectively, all in electrical and electronics engineering, from Bilkent University, ankara, Turkey. He was a re-search assistant in Bilkent University between 1991 and 1999, and an engineering research asso-ciate at E. l. Ginzton laboratory, stanford Uni-versity, Palo alto, ca, between 2000 and 2006. Between 2006 and 2008 he worked at siemens corporate research as a research scientist, where

he is still working as a consultant. He is a member of the IEEE and the Electron devices society.

Ching-Hsiang Cheng is a research engineer and also assistant professor in the research Institute of Innovative Products & Technologies, Hong Kong Polytechnic University, Hung Hom, Kow-loon, Hong Kong. ching-Hsiang cheng was born in Taipei, Taiwan. He received the B.s. degree in mechanical engineering from national Taiwan University in 1993. after completing military ser-vice in 1995, he was admitted to cornell Univer-sity in Ithaca, ny, and received master’s degrees in both mechanical and electrical engineering from

cornell University in 1998. Following his master’s degree, he joined Pro-fessor Khuri-yakub’s ultrasonics group in electrical engineering at stan-ford University, Palo alto, ca, and received his Ph.d. degree in 2005. after receiving his doctoral degree, he was invited to join Professor Ken-neth Goodson’s microscale heat transfer group in mechanical engineer-ing, also at stanford University, as a postdoctoral scholar. after complet-ing his postdoctoral research, he joined Industrial Technology research Institute (ITrI) in Taiwan as a researcher and became a project leader after three months. In 2006, he was appointed as a research engineer and also assistant professor at the research Institute of Innovative Products and Technologies (rIIPT) of the Hong Kong Polytechnic University. His research interests include polymeric strain and force sensors, microfluidic systems with integrated sensors and actuators, capacitive micromachined ultrasonic transducers (cMUT), electrical through-wafer interconnects, nanoimprint technology, and microelectromechanical systems (MEMs) for health-care applications.

Butrus (Pierre) T. Khuri-Yakub (s’70–s’73–M’76–sM’87–F’95) is a professor of electrical engi-neering at stanford University. He received the B.s. degree in 1970 from the american University of Beirut, the M.s. degree in 1972 from dart-mouth college, and the Ph.d. degree in 1975 from stanford University, all in electrical engineering. He was a research associate (1965–1978) then se-nior research associate (1978–1982) at the E. l. Ginzton laboratory of stanford University and was promoted to the rank of professor of electrical

engineering in 1982. His current research interests include medical ultra-sound imaging and therapy, micromachined ultrasonic transducers, smart biofluidic channels, microphones, ultrasonic fluid ejectors, and ul-trasonic nondestructive evaluation, imaging, and microscopy. He has au-thored more than 400 publications and has been principal inventor or co-inventor of 76 U.s. and internationally issued patents. He was award-ed the Medal of the city of Bordeaux in 1983 for his contributions to nondestructive evaluation, the distinguished advisor award of the school of Engineering at stanford University in 1987, the distinguished lecturer award of the IEEE UFFc society in 1999, a stanford Univer-sity outstanding Inventor award in 2004, and a distinguished alumnus award of the school of Engineering of the american University of Beirut in 2005.


capacitive micromachined ultrasonic transducers with ...membranes exhibits higher fill factor...

Documents