Capacitive micromachined ultrasonic transducer (CMUT) arrays for medical imaging

Transcription

1 Microelectronics Journal 37 (26) Capacitive micromachined ultrasonic transducer (CMUT) arrays for medical imaging Alessandro Caronti a, *, G. Caliano a, R. Carotenuto a,c, A. Savoia a, M. Pappalardo a, E. Cianci b, V. Foglietti b a Dip. di Ingegneria Elettronica, Università Roma Tre, Via della Vasca Navale 84, 146 Roma, Italy b Istituto di Fotonica e Nanotecnologie IFN-CNR, Via Cineto Romano 42, 156 Roma, Italy c Dip. I.M.E.T., Università degli Studi Mediterranea di Reggio Calabria, 8969 Reggio Calabria, Italy Received 6 August 25; received in revised form 18 October 25; accepted 24 October 25 Available online 13 December 25 Abstract Capacitive micromachined ultrasonic transducers (CMUTs) bring the fabrication technology of standard integrated circuits into the field of ultrasound medical imaging. This unique property, combined with the inherent advantages of CMUTs in terms of increased bandwidth and suitability for new imaging modalities and high frequency applications, have indicated these devices as new generation arrays for acoustic imaging. The advances in microfabrication have made possible to fabricate, in few years, silicon-based electrostatic transducers competing in performance with the piezoelectric transducers. This paper summarizes the fabrication, design, modeling, and characterization of 1D CMUT linear arrays for medical imaging, established in our laboratories during the past 3 years. Although the viability of our CMUT technology for applications in diagnostic echographic imaging is demonstrated, the whole process from silicon die to final probe is not fully mature yet for successful practical applications. q 25 Elsevier Ltd. All rights reserved. Keywords: Capacitive ultrasonic transducers (CMUTs); Medical imaging; Micromachining; MEMS 1. Introduction Since their first appearance in the mid 199s [1], capacitive micromachined ultrasonic transducers (CMUTs) have rapidly emerged as an alternative to conventional piezoelectric transducers, especially in the field of medical imaging [2 4]. The basic element of a CMUT is a capacitor cell with a fixed electrode (backplate) and a free electrode (membrane). The principle of operation is the well-known electrostatic transduction mechanism. If an alternating voltage is applied between the membrane and the backplate, the modulation of the electrostatic force results in membrane vibration with generation of ultrasounds. Conversely, when the membrane is subjected to an incident ultrasonic wave, the capacitance change can be detected as a current or voltage signal. A DC bias voltage must be used in reception for signal detection, and it is required in transmission for linear operation. In addition, * Corresponding author. Tel.: C ; fax: C address: caronti@uniroma3.it (A. Caronti) /$ - see front matter q 25 Elsevier Ltd. All rights reserved. doi:1.116/j.mejo both the transmit and receive sensitivities increase with increasing the bias voltage. Although the idea of generating acoustic waves by the electrostatic attraction force between the plates of a condenser is very old, recent advances in the microfabrication technology have made possible to fabricate electrostatic transducers consisting of a large number of membranes with precisely controlled geometrical and mechanical properties. For the operation in the megahertz range, as needed by the echographic applications, the lateral dimensions of the membranes are on the order of tens of microns, and the thickness is about 1 2 mm. Thanks to the surface micromachining, the electrode separation can be made very small, in the sub-micron range, which enables high electric fields inside the gap, required to achieve significant electrostatic force and transduction efficiency. The main advantages of CMUTs compared to piezoelectrics are the better acoustic matching to the propagation medium, resulting in wider immersion bandwidth and improved image resolution, the ease of fabrication, the ability to be integrated with electronic circuits on the same wafer, and the expected reduction of production costs. There is also a great potential for real-time 3D imaging, through the realization of 2D CMUT

2 A. Caronti et al. / Microelectronics Journal 37 (26) Fig. 1. Basic steps of a CMUT fabrication process. arrays with a large number of elements [5], harmonic imaging applications, and high-frequency applications such as intravascular imaging (IVUS) [6]. The aim of this paper is to review the fabrication technology, modeling, design and characterization of CMUT arrays for medical imaging, as developed in our laboratories for some 3 years. 2. Fabrication process of CMUT arrays Capacitive micromachined ultrasonic transducers consist of an array of metallized micro-membranes suspended over a substrate. CMUTs are commonly fabricated by means of the surface micromachining technology, using standard integrated circuits techniques. Several processes have been reported in the literature to fabricate CMUTs, using different materials and thin-film deposition techniques [7,8]; integrations with CMOS electronics have been also presented [9,1]. Fig. 1 shows the basic fabrication steps of a process using PECVD silicon nitride as a membrane structural layer, evaporated chromium as a sacrificial layer, and sputtered aluminum for the metallization [11]. The device is fabricated onto a silicon wafer covered with thick thermal silicon dioxide on both sides. After aluminum sputtering and bottom electrode patterning, a thin-layer of silicon nitride is deposited by rf- PECVD (Fig. 1(a) and (b)). A chromium layer, acting as sacrificial layer, is evaporated and patterned into islands to define the cavities under the membranes (Fig. 1(c)). The excellent etching selectivity of chromium against silicon nitride allows a good control over the cell lateral dimensions and gap height. The first silicon nitride membrane layer is deposited at 35 8C using silane, ammonia, and nitrogen diluted in helium as reactant gases (Fig. 1(d)). The tensile stress of the film is controlled by varying the silane to ammonia flow ratio. An aluminium layer is then sputtered on top of the membranes and patterned to define the top electrodes and interconnections between adjacent cells (Fig. 1(e)). After a second silicon nitride deposition, the membranes are released by wet etching of the sacrificial layer through the etching holes defined around the perimeter of the membranes (Fig. 1(f) and (g)). Finally, the etchant holes are sealed by a third silicon Fig. 2. Optical microscope image of the CMUT membranes with top electrodes and sealed etching holes. Fig. 3. A portion of a 64-element 1D CMUT array (top-view).

3 772 A. Caronti et al. / Microelectronics Journal 37 (26) Fig. 4. Axisymmetric model of a CMUT cell with a membrane diameter of 4 mm. nitride deposition (Fig. 1(h)), in order to avoid water filling the cavities during immersed operation that degrades the device performance and eventually leads to a failure. 3. Design of CMUT arrays The overall performance of the CMUT array in immersion, in terms of center frequency, bandwidth and sensitivity, are strongly dependent on the design of the single cells, and their arrangement within each element as well. Our fabricated devices consist of circular membranes (Fig. 2). At the operating frequencies, the membrane dimension is typically lower than a wavelength in water. In order to design 1D CMUT linear arrays for medical imaging, the diameter of the membranes and their number in each element must satisfy the geometrical requirement imposed by the element-toelement distance (element pitch), so as to avoid grating lobes occurrence in the angular response of the array [12]. Conventional piezolectric arrays typically use a dice and fill technique, where the lateral dimension of the elements is defined by mechanical dicing and a polymer is infiltrated and cured into the kerfs. Using this approach, the fabrication of ultrasonic arrays operating above 2 MHz is very challenging with piezo-composites [13]. On the other hand, the CMUT technology is easily applicable to fine-scale arrays. A realized design for a 7-MHz CMUT linear array is based on 4-mm diameter membranes, where each element includes 4!288Z1152 membranes arranged with the configuration shown in Fig. 3. We employed finite element modeling (FEM) using the commercial software ANSYS (ANSYS Inc., Canonsburg, PA, USA) for the analysis, design and optimization of both CMUT single cells and array elements. Because the computation time of a FEM simulation can be very long, Table 1 Mechanical and electrical properties of the materials used in FEM simulations Parameters SiN x SiO Aluminum Young s modulus (GPa) Poisson s ratio Density (kg/m 3 ) Dielectric constant 7 4 an effort was made to build custom simplified models, which are nevertheless able to accurately simulate the various aspects of CMUT operation. 4. Static operation: FEM modeling of the single cell The electrostatic analysis allows the computation of the membrane deformation caused by an applied bias voltage. An axisymmetric model of the CMUT cell is shown in Fig. 4; this includes a silicon nitride membrane with a buried aluminum electrode, a thin silicon nitride insulation layer over the bottom electrode, and a vacuum gap between the membrane and the insulation layer. The bottom electrode is assumed to be infinitesimally thick, because it does not affect the membrane displacement and the electrostatic solution. The metal interconnects linking adjacent cells are neglected by the axisymmetric model. The PLANE121 electrostatic elements, and the PLANE82 structural elements, were used to mesh the structure. The coupled electrostatic-structural analysis was performed using the ANSYS macro ESSOLV, which automatically iterates between an electrostatic field solution and a structural solution until the field and the structure are in equilibrium. The material parameters used in the simulations are listed in Table 1. If the CMUT is operated in conventional regime, the best performance are obtained when the bias voltage is close to the collapse voltage, which is the voltage at which the electrostatic U [nm] center average Static membrane displacement Bias voltage [V] Fig. 5. Center and average membrane displacement as a function of the bias voltage.

4 A. Caronti et al. / Microelectronics Journal 37 (26) C [ ff ] Static cell capacitance Bias voltage [V] Fig. 6. Cell capacitance versus the bias voltage. force overcomes the elastic restoring force of the membrane, and the membrane collapses onto the substrate [14]. The collapse voltage was calculated as 232 V for the cell of Fig. 4, with an uncertainty of 1V based on a non-convergence criterion. The effects of membrane metallization upon the static displacement of the diaphragm, its mechanical resonance frequency, the cell capacitance, and the collapse voltage, for different sizes of the top electrode, are analyzed in [15]. A plot of the membrane displacement is shown in Fig. 5 as a function of the applied bias voltage. The cell capacitance C was computed using the ANSYS macro CMATRIX, relating the charges on the electrodes with the voltage drop. A plot of the capacitance versus the bias voltage, before the collapse to take place, is shown in Fig Dynamic operation: FEM modeling of the CMUT array element The most simple model to analyze the dynamic operation of an immersion CMUT is the unbounded transducer model, Average pressure [db re Pin] pitch = 11% D mem pitch = 12% D mem pitch = 14% D mem pitch = 16% D mem BW 3dB = 135%@ 8. MHz Fig. 8. FEM simulated transmit pressure on the surface of the unbounded CMUT with increasing distance between the membranes. representing infinity of identical membranes, all driven in phase. The unbounded FEM model for the CMUT element with the membrane arrangement of Fig. 3 includes two quarters of membranes in contact with a fluid column with rigid walls, as shown in Fig. 7. The validity of this approach is stated by the principle of image sources [18]. The unbounded model can be run fast to provide a first order approximation of the actual CMUT array element response. Average displacement [nm] piston unbounded CMUT CMUT array element Average pressure [db re Pin] piston unbounded CMUT CMUT array element Fig. 7. FEM model of an unbounded transducer with the membrane arrangement shown in Fig. 3. Absorbing acoustic elements are placed on top to avoid reflections of outgoing pressure waves Fig. 9. Simulated average displacement (top) and surface pressure (bottom) of a CMUT array element, and its equivalent piston.

5 774 A. Caronti et al. / Microelectronics Journal 37 (26) Fig. 8 shows the average transmit pressure of a device with the cell configuration of Fig. 4, for increasing values of the centerto-center distance between the membranes (membrane pitch). As can be seen, the acoustic coupling determines the immersion bandwidth: the smaller the pitch, the larger the acoustic coupling and the transmit bandwidth. The dip around 22 MHz is in between the first and second symmetrical modes of the membranes, where the average displacement is minimum, and can be called a mechanical anti-resonance. Unlike piezoelectric transducers, fractional bandwidths as high as 13% can easily be obtained with CMUTs, thus improving the axial resolution of the reconstructed image; this also enables other modern techniques like tissue harmonic imaging [19]. Acoustic coupling has other important effects, which cannot be investigated by means of the simple unbounded model. Because of the negligible mass, acoustic interactions occur between the membranes of a CMUT with finite dimensions, such as the long and narrow 1D array elements used in medical imaging applications (Fig. 3). We have deeply investigated the mechanism of acoustic interactions between the CMUT membranes in immersion, through both finite element modeling and experimental measurements [16 18]. We developed a reduced 3D FEM model of one element by using appropriate periodicity conditions, so as to represent an infinitely long CMUT array element. All the membranes in the Hydrophone signal [mv] acoustic interactions V DC = 14V substrate ringing Time [ms] Normalized transmit pressure Measured FEM FFT amplitude [db] Signal amplitude [V] Time [ms] Fig. 11. Pulse-echo impulse response, and its Fourier transform, as measured in oil with a voltage amplifier. element are assumed to be driven in phase. As a result of the pressure waves propagation, the membranes are subjected to different acoustic loading, depending on their position in the element, and dips occur in the average displacement and transmit pressure, at frequencies where interaction effects are stronger [18]. The average displacement and pressure responses of a typical CMUT array element are shown in Fig. 9, where a comparison is reported with the results of the unbounded model and the behavior of a piston having the same mechanical impedance as the CMUT element. The remarkable bandwidth of the piston is due to the absence of both the higher order modes and the fluid reactive effects, originating from the membrane flexural motion. 6. Experimental results and images We validated our models through optical displacement measurements, hydrophone measurements, and pulse-echo measurements. Fig. 1 displays the pressure impulse response, and its Fourier transform, as measured by a hydrophone on the acoustic axis of a CMUT array element in water (top); the element was biased at 14 V and driven with a short pulse through FFT amplitude [db] acoustic interactions surface average far-field on axis FFT amplitude [db] Fig. 1. Transmit pressure of a CMUT array element in water: hydrophone measurement (top), and FEM simulation (bottom). Fig. 12. Sixty-four-element CMUT linear array mounted on a PCB, with an epoxy resin protecting the wire-bond connections.

6 A. Caronti et al. / Microelectronics Journal 37 (26) Fig. 13. Final assembly of the CMUT array probe with an echographic cable. the pulser/receiver GE-Panametrics 58. A comparison with a FEM simulation of the transmit pressure is shown on the bottom of the figure The dashed line represents the average surface pressure, which is then propagated and compensated for diffraction losses (solid line). The effect of water attenuation is not included. The acoustic interaction effects are predicted by FEM and can be observed experimentally, resulting in a moderate degradation of the output pressure. The mechanical antiresonance of the membranes, limiting the immersion bandwidth of the device below 2 MHz, is also found in a very good agreement between FEM simulations and experiments. In addition, silicon substrate ringing modes are observed at 7.5 MHz and harmonics. These modes have been previously reported by other researchers [2], and can be eliminated through proper high impedance backing of the CMUT; substrate ringing is not predicted by FEM because the substrate is not included in the model. The most critical components of an ultrasound imaging system are the transducer and its front-end electronics. In order to exploit all the advantages of the CMUT technology, the design of a low-noise wideband electronic system is an important issue. Custom-designed electronic systems, based on voltage (non-inverting) and trans-impedance (inverting) Fig. 14. Echographic images obtained with the 64-element CMUT probe: (top) cyst phantom; (bottom) human carotid.

7 776 A. Caronti et al. / Microelectronics Journal 37 (26) amplifier configurations, were realized to evaluate the pulseecho performance of CMUT arrays [21]. A pulse-echo impulse response of a CMUT array element immersed in oil, as measured with a voltage amplifier, is shown in Fig. 11. The transducer is provided with a backing on the rear side to eliminate substrate ringing. The ringing visible in the echo signal is mainly caused by the acoustic interactions between the membranes, as also observed in the FFT spectrum around 3.5 MHz. Our research is mainly focused on the development of CMUT array probes for medical imaging. For this purpose, the CMUT array was wire-bonded to a PCB (Fig. 12), the active membrane area was covered with a thin-layer of silicone rubber for continuative immersed operation, and the transducer was housed in a probe containing biasing and decoupling circuits. A picture of the final probe with the cable connecting to a commercial echographic system (Esaote Technos) is shown in Fig. 13. Fig. 14 displays the image of an echographic cyst phantom immersed in a uniform parenchyma mimicking the human body (top), and the in vivo image of a human carotid (bottom), as obtained by the CMUT probe. The image quality is comparable with that achieved with a commercial piezoelectric probe having similar features, although the current sensitivity of our CMUT probes is about 1 db below [22,23]. 7. Conclusion In this paper, we presented a summary of the fabrication technology, design, modeling and characterization of capacitive micromachined ultrasonic transducer (CMUT) arrays for medical imaging, as developed in our laboratories in the past 3 years. We demonstrated, by experimental evidence, the viability of our technology to realize multi-element probes, which can be connected to standard commercial echographic systems to generate images. At present, the images obtained with state-of-the-art CMUT probes can be comparable or even superior in quality to those of commercial PZT probes. Thanks to the wider pulse-echo bandwidth, typically greater than 1%, CMUT images feature improved axial resolution, which allows for small targets to be resolved. However, further improvements in both sensitivity and resolution are needed to fully compete with piezoelectric arrays, especially in the applications where high depth of penetration is required. This paper also reports on finite element modeling of CMUT arrays, that we are using extensively to design and optimize the performance of these devices. In particular, we have recently reported the first comprehensive analysis of the acoustic coupling in CMUTs, through both finite element modeling and experiments. The results show that interaction effects between the membranes can be notable in the operational bandwidth of the transducer, giving rise to a degradation of the transducer output pressure and efficiency. The CMUT technology developed in our laboratories is not fully consolidated yet. Our probes suffer from low sensitivity and reliability issues. However, we are strongly confident that significant improvements can be obtained in the near future, by conducting research in many directions: the materials quality, fabrication process stability, CMUT design, front-end electronics, and probe engineering are the fields where we have been working on to translate CMUT array demonstrators into successful practical applications. References [1] M.I. Haller, B.T. Khuri-Yakub, A surface micromachined electrostatic ultrasonic air transducer, Proc. IEEE Ultrason. Symp. (1994) [2] O. Oralkan, A.S. Ergun, J.A. Johnson, M. Karaman, U. Demirci, K. Kaviani, T.H. Lee, Capacitive micromachined ultrasonic transducers: next-generation arrays for acoustic imaging?, IEEE Trans. Ultrason Ferroelectr. Freq. Control 49 (11) (22) [3] D.M. Mills, L.S. Smith, Real-time in-vivo imaging with capacitive micromachined ultrasound transducer (cmut) linear arrays, in: Proceedings IEEE Ultrasonics Symposium, 23, pp [4] S. Panda, C. Daft, P. Wagner, I. Ladabaum, P. Pellegretti, F. Bertora, Capacitive microfabricated ultrasonic transducer (CMUT) probes: imaging advantages over PZT probes, Presented at 3rd International Workshop on Micromachined Ultrasonic Transducers, Lausanne, Switzerland, June 26 27, 23. [5] O. Oralkan, A.S. Ergun, C.H. Cheng, J.A. Johnson, M. Karaman, T.H. Lee, B.T. Khuri-Yakub, Volumetric ultrasound imaging using 2-D CMUT arrays, IEEE Trans. Ultrason. Ferroelectr Freq. Control 5 (11) (23) [6] J.G. Knight, F.L. Degertekin, Fabrication and characterization of CMUTs for forward looking intravascular ultrasound imaging, in: Proceedings IEEE Ultrasonics Symposium, 23, pp [7] X.C. Jin, I. Ladabaum, B.T. Khuri-Yakub, The microfabrication of capacitive ultrasonic transducers, IEEE/ASME J. Microelectromech. Syst. 7 (3) (1998) [8] E. Cianci, V. Foglietti, D. Memmi, G. Caliano, A. Caronti, M. Pappalardo, Fabrication of capacitive ultrasonic transducers by a low temperature and fully surface-micromachined process, Precision Eng. 26 (22) [9] P.C. Eccardt, K. Niederer, T. Scheiter, C. Hierold, Surface micromachined ultrasound transducers in CMOS technology, in: Proceedings IEEE Ultrasonics Symposium, 1996 pp [1] R.A. Noble, A.R.D. Jones, T.J. Robertson, D.A. Hutchins, D.R. Billson, Novel, wide bandwidth micromachined ultrasonic transducers, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 48 (6) (21) [11] E. Cianci, A. Minotti, V. Foglietti, G. Caliano, M. Pappalardo, Onedimensional capacitive micromachined ultrasonic transducer arrays for echographic probes, Microelectron. Eng (24) [12] G. Kino, Acoustic Waves, Devices, Imaging, and Analog Signal Processing, Prentice-Hall, Englewood Cliffs, NJ, [13] T.A. Ritter, T.R. Shrout, R. Tutwiler, K.K. Shung, A 3-MHz piezocomposite ultrasound array for medical imaging applications, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 49 (2) (22) [14] F.V. Hunt, Electroacoustics. The Analysis of Transduction, and Its Historical Background, American Institute of Physics, [15] A. Caronti, R. Carotenuto, G. Caliano, M. Pappalardo, The effects of membrane metallization in capacitive microfabricated ultrasonic transducers, J. Acoust. Soc. Am. 115 (2) (24) [16] A. Caronti, D. Fiasca, G. Caliano. M. Pappalardo, An experimental study of cross-coupling and acoustic interactions in 1-D capacitive micromachined ultrasonic transducer arrays, Presented at US Navy Workshop on Acoustic Transduction Materials and Devices, State College, Pennsylvania, May, 6 8, 23. [17] A. Caronti, D. Fiasca, G. Caliano, E. Cianci, M. Pappalardo, Experimental study of acoustic coupling in CMUT arrays by optical interferometry, in: Proceedings IEEE Ultrasonics Symposium, 23, pp [18] A. Caronti, A. Savoia, G. Caliano, M. Pappalardo, Acoustic coupling in capacitive microfabricated ultrasonic transducers: modeling and experiments, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 52 (12) (25), in print.

8 A. Caronti et al. / Microelectronics Journal 37 (26) [19] M.A. Averkiou, Tissue harmonic imaging, in: Proceedings IEEE Ultrasonics Symposium, 2, pp [2] I. Ladabaum, P. Wagner, C. Zanelli, J. Mould, P. Reynolds, G. Wojcik, Silicon substrate ringing in microfabricated ultrasonic transducers, in: Proceedings IEEE Ultrasonics Symposium, 2, pp [21] A. Caronti, G. Caliano, A. Savoia, M. Pappalardo, R. Carotenuto, A lownoise, wideband electronic system for pulse-echo ultrasound imaging with CMUT arrays, in: Proceedings IEEE Ultrasonics Symposium, 24, pp [22] G. Caliano, R. Carotenuto, E. Cianci, V. Foglietti, A. Caronti, M. Pappalardo, A CMUT linear array used as echographic probe: fabrication, characterization, and images, in: Proceedings IEEE Ultrasonics Symposium, 24, pp [23] G. Caliano, R. Carotenuto, E. Cianci, V. Foglietti, A. Caronti, A. Iula, M. Pappalardo, Design, fabrication, and characterization of a capacitive micromachined ultrasonic probe for medical imaging, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, 52 (12) (25), in print.