ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 52, no. 12, december 2005
2259
Design, Fabrication and Characterization of a Capacitive Micromachined Ultrasonic Probe for Medical Imaging Giosu´e Caliano, Member, IEEE, Riccardo Carotenuto, Member, IEEE, Elena Cianci, Vittorio Foglietti, Member, IEEE, Alessandro Caronti, Member, IEEE, Antonio Iula, Member, IEEE, and Massimo Pappalardo, Member, IEEE Abstract—In this paper we report the design, fabrication process, and characterization of a 64-elements capacitive micromachined ultrasonic transducer (cMUT), 3 MHz center frequency, 100% fractional bandwidth. Using this transducer, we developed a linear probe for application in medical echographic imaging. The probe was fully characterized and tested with a commercial echographic scanner to obtain first images from phantoms and in vivo human body. The results, which quickly follow similar results obtained by other researchers, clearly show the great potentiality of this new emerging technology. The cMUT probe works better than the standard piezoelectric probe as far as the axial resolution is concerned, but it suffers from low sensitivity. At present this can be a limit, especially for in depth operation. But we are strongly confident that significant improvements can be obtained in the very near future to overcome this limitation, with a better transducer design, the use of an acoustic lens, and using well matched, front-end electronics between the transducer and the echographic system.
I. Introduction iezoelectric ultrasonic transducers are widely used in echographic systems both for nondestruction evaluation (NDE) and medical applications. The piezoelectric transducer technology, developed over the last 30 years, is nowadays a mature technology that gives bandwidth as high as 60–80% and a good sensitivity. Nevertheless, it is expensive and presents some restrictions in the transducer geometry design and cabling, especially for twodimensional (2-D) arrays. The electrostatic principle is an alternative to the piezoelectric effect for ultrasonic transducers. Electrostatic transducers made of a metallized, thin, dielectric membrane stretched over a rough conducting backplate were reported in the early 1950 [1] for air-coupled applications, and the first device for applications in liquids was devel-
P
Manuscript received February 3, 2004; accepted October 27, 2004. This work was granted financial support by the UMIC-EUREKA project (E!2145). G. Caliano, A. Caronti, A. Iula, and M. Pappalardo are with the Dipartimento di Ingegneria Elettronica, Universit` a Roma Tre, 00146 Roma, Italy (e-mail:
[email protected]). R. Carotenuto is with the Dipartimento I. M. E. T., Universit` a “Mediterranea” di Reggio Calabria, Loc. Feo de Vito, 89069 Reggio Calabria, Italy. E. Cianci and V. Foglietti are with the Istituto di Fotonica e Nanotecnologie-Consiglio Nazionale delle Richerche, 00156 Roma, Italy.
oped in 1979 [2]. These devices are based on the electrostatic attraction force between the backplate and the metallized membrane. In transmission, the membrane is forced into flexural vibration by a time-varying voltage; in reception, the membrane is put into vibration by the incident pressure wave. The capacitance modulation due to the membrane displacement is used to detect the signal. In recent years, Haller and Khuri-Yakub [3] and Eccardt et al. [4] have introduced a new generation of capacitive micromachined ultrasonic transducers (cMUTs). These devices consist of a 2-D array of miniaturized electrostatic cells electrically connected in parallel and driven in phase, fabricated using surface silicon micromachining. To perform operation in the megahertz range, the lateral dimensions of each cell are in the order of tens of microns, and to obtain sufficient sensitivity the number of cells is of the order of thousands [5]. This new generation of ultrasonic transducers seems to be competitive with piezoelectric transducers both for water- and air-coupled applications, thanks to the inherently low mechanical impedance of the thin vibrating membrane. In perspective, further advantages of this kind of ultrasonic transducers are the ability to integrate the front-end electronics on the silicon wafer used for the cMUT itself and to make complex 1D and 2-D array designs using simple photolithographic techniques. The first evidence that cMUT arrays are able to obtain echographic images of a test object was given by Oralkan et al. [6] by means of an off-line postprocessing reconstruction system. Very recently, Panda et al. [7] showed echographic images of in vivo human body obtained with a cMUT array probe developed by Sensant Corp. (San Leandro, CA) and coupled to a commercial Esaote system. Mills and Smith [8] also showed echographic images obtained with the same cMUT array probe developed by Sensant and engineered by GE Healthcare (Chalfont St. Giles, UK) coupled to a commercial GE echographic system. In previous works [9]–[11] we described the cMUT micromachining technology developed in our laboratories, and we discussed the design criteria and the analytical and numerical [finite element modeling (FEM) approach] model used to foresee the performances of this kind of transducers [12], [13]. Some other relevant modeling works can be found in [14]–[17]. In this work we demonstrate, with experimental evidence, the practical use of our tech-
c 2005 IEEE 0885–3010/$20.00
2260
ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 52, no. 12, december 2005 TABLE I Geometrical Parameters of the cMUT Array. Nominal frequency [MHz] Number of elements Element pitch [mm] Element length (elevation) [mm]
3.0 64 .25 12
nology to realize multielement probes able to be used in commercial echographic systems for medical diagnostics. We fabricated, by surface micromachining, on a thin silicon wafer, an array of 64 cMUT elements, each obtained connecting in parallel 1152 capacitive micro-cells. The silicon array was backed to a substrate, wire-bonded to a pair of printed circuit boards (PCB), which provide the electrical polarization, and wired to a standard plug. The engineering of the probe was completed by encapsulating the silicon array and the PCBs in a special box, and protecting the active surface of the silicon array by means of a special polymer layer. The probe was characterized with standard measurements and simply coupled to a commercial echographic system (Technos by Esaote Spa, Genova, Italy), without any ad hoc matching circuit box. Echographic images of a test object and of internal organs of a human body were obtained, demonstrating the practical possibility to use the technology developed in our laboratories to make cMUT echographic probes.
II. cMUT Array: Geometrical Design and Technology The criteria used to design the geometrical dimensions of the cMUT array are the same as those generally used in standard, commercially available, piezoelectric probes [18]. These criteria, in our case, assure an acceptable trade-off between beam width and grating lobe levels up to 6 MHz for linear scanning and up to 3 MHz for sector scanning. The relevant geometrical parameters are summarized in Table I. In order to limit the complexity of the prototype design, the number of elements was chosen to be only 64, and current echographic piezoelectric probes usually have a much larger number of elements, typically 128 for sector scanning or 128 up to 192 for linear scanning. A larger number of active elements improves the resolution in the lateral direction and enlarges the width of the field of view. It was assumed that 64 elements were sufficient to demonstrate the possibility to use cMUT probes in practical applications. The technology developed to fabricate our silicon cMUT array was explained in previous works [9]–[11], [13], [19]– [21], to which we refer for details; therefore, in this section, only the basic steps of the fabrication process of the cMUT array are briefly described. The fabrication process is based on the surface micromachining technology. To realize the basic structure of the cMUT (i.e., an array of micro-cells, each equipped with its own metallized membrane suspended on a fixed bottom
Fig. 1. Bottom chromium electrodes of the array.
electrode), seven thin film depositions and seven lithographic steps are used. The device is fabricated on top of a silicon oxidized substrate. The oxide layer is used to electrically insulate the bottom electrodes of the elements; the oxide thickness has been chosen high enough to reduce the parasitic capacitance and to increase the metal-oxide-semiconductor (MOS) diode voltage breakdown which, however, exists between the electrodes and the silicon substrate [22]. The electrodes are patterned on a metal layer deposited over the oxide layer (see Fig. 1) and subsequently protected by a thin film of silicon nitride deposited with a radio frequency-plasma enchanced chemical vapor deposition (RF-PECVD) technique. We first used aluminum as metal for the bottom electrodes, but we observed an increased roughness of its surface after deposition of the PECVD silicon nitride layer. The surface roughness increased the possibility of a disruptive electrical discharge between the bottom-fixed electrode and the metallized membrane. Unlike aluminum, chromium does not undergo modification during the passivation step. In order to create the micro-cells structure, a chromium sacrificial layer is evaporated onto the silicon nitride layer. By means of a photolithographic step, this layer then is patterned into small circular islands that will define the cavities existing under the membrane of the micro-cells. Fig. 2 shows a portion of two bottom electrodes covered with the patterned chromium sacrificial islands. A PECVD silicon nitride film is deposited all over the silicon substrate covering the chromium sacrificial islands with a layer thickness of 4000 ˚ A. The deposition is performed at 350◦ C using as reactant gases silane, ammonia, and nitrogen diluted in helium. The film is conformal, and the step coverage is 100%. The micro membranes are released by wet etching of the sacrificial islands through holes performed by dry-etching through the silicon nitride layer covering the islands. Fig. 3 shows the silicon nitride borders of the micro-cell.
caliano et al.: probe for cmut
2261
(a)
Fig. 2. Sacrificial chromium islands defining the micro-cells.
(b) Fig. 4. (a) Optical microscopy image of silicon nitride free-standing membranes. The chromium electrode underneath is visible through the optically transparent membrane layer. (b) AFM section of a released membrane.
Fig. 3. Scanning electron microscope (SEM) image of a portion of a released membrane over a sealed cavity. (a)
The tensile stress of the micromembrane can be tuned varying the silane to ammonia flow ratio. Fig. 4(a) shows the optical microscopy image of silicon nitride freestanding membranes. Fig. 4(b) shows the profile of a released membrane obtained with the atomic force microscope (AFM); the small deflection at the center is due to intrinsic tensile stress of the film. The next critical step in the process is the sealing of the etchant holes, which is necessary to avoid water filling of the cavities in immersion operation. A PECVD silicon nitride deposition is performed, with the required step coverage and stress properties, able to fill the holes without significantly penetrating under the active area of the membrane. A film about 20% thicker than the cavity height is needed to seal the holes; it is deposited over the entire device, so it must be removed from the membrane areas, masking small regions around the holes and etching all around (see Fig. 5). An aluminum layer then is sputtered on top of the membranes and patterned to define the upper circular electrodes and interconnections between adjacent cells. The wafer then is protected by a PECVD silicon nitride film characterized by a low value of residual stress that is tai-
(b)
Fig. 5. (a) Optical microscope image of the circular membranes. (b) SEM image of a sealed hole.
lored in order to minimize the membrane bending, which would limit its deflection during operation. The final vibrating structure of the single micro-cell consists of a silicon nitride membrane, a metal electrode, and a silicon nitride protection layer; the AFM section profile is shown in Fig. 6.
III. Probe Engineering We did not find experimental evidence of reflected echoes from the bottom of the silicon die of the cMUT array in air-coupled operation. However, substrate ringing in water-coupled operation has been observed experimentally by other researchers [23]. We excited one element of an air-coupled, not-backed array with a Panametrics 5800 (energy 12.5 µJ, damping 500 Ω, polarization 100 VDC). Fig. 7(a) shows the displacement versus time of the center of one membrane of the fired element and that of the center of one nearest rail [see Fig. 7(b)]. This measurement was
2262
ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 52, no. 12, december 2005
Fig. 6. AFM section profile of a finished cMUT cell.
made with a Polytec MSV300 laser microvibrometer system (Polytec GmbH, Waldbronn, Germany). As you can see, the membrane displacement becomes visible with a delay of about 160 ns, due to its large rise time (the array is air coupled and the single membrane is free to oscillate), and the rail also starts to oscillate practically at the same time. The ratio between these two amplitudes is about 500. After a time delay of 130 ns from the start of the oscillation of the membrane, which corresponds to the time of flight of a longitudinal wave in the thickness of the silicon substrate (vl = 8340 m/s), no significant amplitude variation is observed in the rail displacement waveform. Even if this experiment suggests that very little energy comes back from the bottom of the silicon, the cMUT silicon array die was glued onto a plate of 3 mm in thickness made of Alumina powders and Stycast 1264 epoxy encapsulant (Emerson & Cuming Inc., Billerica, MA), in order to minimize spurious plate waves propagation in the substrate and to realize an adequate mechanical support. The 64-elements cMUT die was designed with alternated odd and even elements to allow a simple electrical connection to the biasing circuits via a special PCB, shown in Fig. 8 (the elements on the right side of the die are only for test purposes). The array was wire bonded to the PCB using a 50 µm aluminum wire [Fig. 9(a)]. This wiring was protected against thermal and mechanical shock using a semiconductor encapsulant for chip-on-board application. We use a Tra-Bond 933-1.5 epoxy encapsulant (Tra-Con Inc., Bedford, MA) formulated to provide a smooth crown over the wire bonds of the device [Fig. 9(b)]. The active surface of the array then was covered with a thin layer (thickness 1.8 mm, velocity of sound ≈ 1000 m/s, and attenuation ≈ 2 dB/mm at 3 MHz) of silicone elastomer (MED-6033, Nusil Technology LLC, Carpinteria, CA) in order to protect the active area of the transducer. In this first prototype, we do not use an acoustic lens, but the silicone elastomer easily can be shaped to perform this important feature. The entire transducer with backing, connection-PCB, and front-protection is housed in a two-part aluminum enclosure with two printed circuit boards (perpendicularly mounted) containing biasing and decoupling circuits and the terminal connectors of the cable that connect the probe
(a)
(b) Fig. 7. Silicon substrate ringing. (a) Displacement versus time of the center of the membrane and of the rail. (b) Membrane and rail measurement points.
to the echographic system. A picture of the completed probe is shown in Fig. 10.
IV. Probe Testing The cMUT silicon array die was first tested in air and showed an evident resonant behavior [12]. Figs. 11(a) and (b) show the superimposed real and imaginary parts of the electrical impedance of the elements, Fig. 12 shows the resonant frequency spreading of the elements (
