1Modes and damping in cMUT transducers for acoustic emission D.W. Greve and W. Wu Department of Electrical and Computer Engineering Carnegie Mellon University, Pittsburgh, PA, USA I.J. Oppenheim Department of Civil and Environmental Engineering, Carnegie Mellon University, Pittsburgh, PA, USA Abstract—Acoustic emissions caused by the initiation and growth of cracks can be detected by cMUT transducers. A particular advantage of this type of transducer is the possibility of fabricating transducers with multiple resonant frequencies. However previously reported designs had several modes within the frequency range of interest. In this paper we consider the design of an improved transducer which exhibits large spacing between resonant modes. In addition we discuss the prediction and measurement of the quality factor. Keywords-acoustic emission, transducer, MEMS, cMUT I. INTRODUCTION Acoustic emissions are ultrasonic pulses produced in sol-ids when irreversible damage occurs under mechanical load-ing. Most commonly piezoelectric transducers with con-trolled damping are used to detect acoustic emissions [1] although alternative sensors have been explored [2,3,4,5]. We have previously reported the application of cMUT transducer arrays for the detection of acoustic emis-sion events [6]. The cMUT array makes it possible to have multiple transducers that sample different portions of the broadband acoustic emission signal. Using the poly-MUMPS process, we designed and fabricated a cMUT transducer ar-ray consisting of polysilicon plates suspended by springs located inside the plate [6,7]. This design exhibited two elec-trically active modes within the frequency range of interest, and quality factors that were lower than desired. Here, we combine finite element simulation and experimental studies to understand the nature of the modes exhibited by these transducers, to quantify damping, and to compare the meas-ured damping with an improved analytic model. These re-sults lead to an improved transducer design for which we report the results of initial device characterization. II. OLD DESIGN SIMULATIONS We first consider the performance of a previous, non-optimized design (referred to as design #1 below). That de-sign used the POLY0 and POLY1 layers of the poly-MUMPs process to form fixed and suspended electrodes, respectively. The operating principle of this transducer is similar to the cMUT in that a DC bias is applied and vibra-tions of the electrodes with respect to each other cause an electric current in the external circuit. However in contrast to the cMUT acoustic energy is coupled to the substrate and the upper electrode is free to move with respect to the lower electrode. In order to obtain reasonable signal levels many small units are joined together to form a single large diaphragm. Figure 1 shows one unit, its connections to neighboring units, and a cross-section of the structure. Each unit consists of a POLY1 plate suspended from four supports. Transducers with different resonant frequencies are designed by varying the length of the springs and the total size of each unit. Fig-ure 1 also shows etch release holes that are provided to en-able removal of the sacrificial layer but which also have the effect of damping the diaphragm resonance. Figure 1. One unit of the diaphragm and connections to neighboring units. Experiments with design #1 showed two important short-comings. First, we observed that there were at least two closely spaced vibrational modes. And secondly, the damp-ing in air was high resulting in low Q. Low Q leads to a wide bandwidth and in addition low sensitivity [6,7]. The purpose of this work is to understand the deficiencies of this design and to develop an improved transducer design. In order to understand the vibrational modes, the eigen-modes were calculated using the structural mechanics module of FEMLAB. Figure 2 shows the results of simula-tions of one of the six transducers (unit size 310 × 310 μm, spring dimensions Ls1 = 8 μm, Ls2 = 27 μm). One quarter of a2unit is shown and the calculation assumed symmetry about the two rearmost edges. The lowest predicted eigenmode (Fig. 1, top) is at 225 kHz which is in fair agreement with the measured lower resonant frequency of 187 kHz. Note that there is significant bending of the plate, that is, the motion is not simply the vi-bration of a rigid plate supported on springs. The second eigenmode at 261 kHz corresponds to a tilting motion of the entire unit and consequently this mode is electrically inac-tive. The next mode at 316 kHz is shown in the lower figure. Again the high degree of plate flexibility is evident. These simulations are consistent with our experimental observation of two closely spaced resonances [6,7]. These modes are closely spaced because the vibrational modes of the plate mass- spring system are close to the vibrational modes of a freely suspended plate. Consequently we have looked for other designs which have plates which are more rigid and/ or springs that are more flexible. Figure 2. Design #1 fundamental mode (top) and third mode (bottom). Each figure shows the z displacement of the simulated eigenmode (color) and the undistorted structure (outline). III. IMPROVED TRANSDUCER DESIGNS Objectives for an improved design include (1) wider separation of the two lowest electrically active resonances and (2) decreased damping. To this end we considered two designs. Design #2 used only the POLY1 layer but the springs were moved to the edge and the size of each unit was reduced. Design #3 used both POLY1 and POLY2 plates to form the diaphragm. Simulation results for these two designs along with simulation results for the previous design are pre-sented in Table I. TABLE I. SIMULATION RESULTS FOR VARIOUS DESIGNS design 1(expt) design 1 (simulated) design 2 POLY1 (simulated) design 3 POLY1+POLY2(simulated) f1 [kHz] 187 225 179 179 f2 [kHz] 275 316 1024 1518 An improved separation of the lowest modes was ob-tained with both new designs. Figure 3 shows simulation results for design #2 which is the one chosen for fabrication. Again only one quarter of the structure is shown although in this case the symmetry planes were along the two edges at the right of the figure. Figure 3. Design #2 fundamental mode (top) and next highest mode (bottom). Transducers with a range of resonant frequencies have been designed with similar geometry using the single-poly diaphragm. Table II