Fabrication and performance of d 33 -mode lead-zirconate-titanate (PZT) MEMS accelerometers

Fabrication and performance of d 33 -mode lead-zirconate-titanate (PZT) MEMS accelerometers H. G. Yu, R. Wolf*,K. Deng +,L.Zou +, S. Tadigadapa and S. Trolier-McKinstry* Department of Electrical Engineering, Penn State University, University Park, PA16802. *Materials Science and Engineering Department, Penn State University + Wilcoxon Research Laboratory, Gaithersburg, Maryland ABSTRACT Piezoelectric accelerometers fabricated from Lead-Zirconate-Titanate (PZT) thin films are expected to achieve higher sensitivities and better signal-to-noise ratios (SNR) in comparison with capacitive and piezoresistive accelerometers. This paper will present, for the first time, the fabrication and performance of piezoelectric, bulkmicromachined accelerometers using PZT thin films operating in the d 33 -mode. Using sol-gel techniques, 0.6µm thick PZT films with high piezoelectric coefficients were deposited. Measurements on these PZT films show a remnant polarization P r >19µC/cm 2, dielectric constants ε r > 800, and d 33 coefficient of 120 pc/n. The PZT accelerometers operating in the d 33 mode were successfully fabricated. Interdigitated capacitors were used to achieve the d 33 mode of operation and deep reactive ion etching was used to define the proof-mass of the accelerometers. Measurements on these accelerometers show sensitivitiesranging from 0.85-1.67 mv/g with resonancefrequenciesranging from 22.4-15.4 khz respectively. In addition to the improved sensitivity, advantages of d 33 -mode accelerometers include use of thinner PZT films, and the ability to optimize the impedance of the device to achieve a higher SNR. The performance of MEMS d 33 -mode accelerometers will also be compare with the previously reported d 31 -mode accelerometers using PZT thin films. I. INTRODUCTION The popularity of microelectromechanical systems (MEMS) accelerometers in recent years has been due primarily to their wide range of automotive uses in safety, vehicle stability and electronic suspension systems. However, with the ability to manufacture these miniature, high-performance sensors at low cost, the market is continuously expanding with applications in consumer products, the biomedical field, industrial monitoring and the military [1]. Of the various types of accelerometers such as piezoresistive [2] and capacitive [3], piezoelectric accelerometers offer a high Q (>80), high output impedance, and low damping [4]. Traditionally, ZnO, AlN and lead zirconate titanate (PZT) films have been employed in piezoelectric microaccelerometers. Of these, PZT films exhibit the highest piezoelectric coefficients and therefore offer the opportunity for highest sensitivity. Thin film piezoelectric coefficients of PZT are an order of magnitude larger than those of ZnO and AlN [5]. Additionally, the piezoelectric response is anisotropic in all piezoelectric materials. In PZT, the magnitude of the piezoelectric charge coefficient (d) varies with the angle between the applied stress and poling directions. For most perovskite ferroelectrics, including most PZT s, in the case where the applied stress and poling directions are perpendicular (d 31 -mode response) the coefficient is ~ 2 times smaller than that in which these directions coincide (d 33 -mode response). To this point, MEMS accelerometers have almost exclusively employed d 31 -mode sensing elements. In this work, bulk micromachined accelerometers were designed and fabricated using standard silicon processing methods. Devices incorporated interdigitated (IDT) electrodes [6] to exploit the d 33 -mode piezoelectric response of a PZT thin film. The relatively simple fabrication process involved only 3 photomask levels and included 2 deep-trench reactive ion-etching (DRIE) steps. The accelerometer was designed to have a sensitivity of 1 mv/g with a resonance frequency greater than 10 khz for application in industrial vibration monitoring. 130 MEMS Components and Applications for Industry, Automobiles, Aerospace, and Communication, Henry Helvajian, Siegfried W. Janson, Franz Lärmer, Editors, Proceedings of SPIE Vol. 4559 (2001) 2001 SPIE 0277-786X/01/$15.00

II. ACCELEROMETER DESIGN An annular design has been employed here. Figure 1 shows front and backside schematic diagrams of the accelerometer. The accelerometer consists of a cylindrical proof mass attached to the silicon frame by an annular diaphragm. Since the accelerometer design is totally symmetrical, the piezoelectric effect arising from motion parallel to the diaphragm is canceled out. Two accelerometer designs with different diaphragm radii have been fabricated. A thin film of PZT is deposited on top of the diaphragm area and is used as the sensing material. The fabricated accelerometers have interdigitated electrode capacitors for d 33 mode of operation. Additionally, three different electrode widths were used while keeping the spacing between the electrodes fixed. Since the capacitance of the devices depends primarily on the electrode spacing and not on the electrode width, the nominal capacitances of all the electrode geometries are the same. However, the different electrode width influences the poling efficiency of the piezoelectric film since the field Electrode for the inner capacitor Electrode for the outer capacitor Clamped Area d r 1 w Proof Mass r 2 (a) Electrode for the common (b) Figure 1. Schematic diagram of a d 33 annular MEMS accelerometer. (a) Top view of the accelerometer showing interdigitated electrodes w and d are width of the electrode and distance between electrodes, respectively, (b) Backside view; r 1 is the radius of the proof mass and r 2 is the radius of the annular membrane under the electrodes will not be parallel to the plane of the piezoelectric film and consequently these areas will not be poled in the d 33 (parallel to the film) directions. So devices with different electrode width were fabricated to investigate the associated degradation in the sensitivity of the accelerometers. However, small width electrodes reduced the yield of the fabricated devices. On each diaphragm, two separate interdigitated capacitors were located; one on the inner diameter of the diaphragm and the other on the outer diameter of the diaphragm. These two areas also represent the regions of maximum stress when the proof mass moves relative to the frame under the influence of applied normal acceleration. Devices with annular diaphragms of inner radius 1mm and outer radius 2mm were fabricated. The interdigitated capacitor located at the inner radius had a nominal value of 100pF while the capacitor located at the outer radius had a nominal value of 150pF. The cylindrical proof mass of the fabricated devices was geometrically defined by lithography process and had a diameter of 2mm while its thickness was defined by the thickness of the wafer. Thus the fabrication process used accurately defined the dimensions of the proof mass. However, as will be described later, the same was not true for the thickness of the diaphragm. Thus the performance of the devices is expected to show variability on account of the corresponding variation in the thickness of diaphragms. Proc. SPIE Vol. 4559 131

III. FABRICATION PROCESS The PZT MEMS accelerometers using the d 33 mode are fabricated with three masks. The process requires front-to-back side alignment and two DRIE steps. The three mask levels consist of top electrode definition, backside individual accelerometer area definition, and backside proof mass area definition. The fabrication process flow is shown in Figure 2. Prepared wafers are 4-inch diameter, n-type <100> (1-10 Ω -cm), double-side-polished silicon wafers Cr-Au layer 0.14 µm PZT layer 0.6 µm ZnO layer 0.5 µm SiO 2 layer 0.7 µm Si Substrate 450 µm PRlayer 10µm 1. Deposit Cr-Au/ PZT/ZnO/SiO 2 layer on the Si substrate 2. Apply photoresist (PR), pattern interdigitated electrode area and etch 3. Define backside etch area for controlling thickness of the membrane 4. Define proof mass and perform first DRIE step for determination of membrane thickness Figure 2. Fabrication Flow Diagram 5. Complete second DRIE step to etch through the substrate after etching defined proof mass area SiO 2 with BOE 132 Proc. SPIE Vol. 4559

having SiO 2 layers on them. A 0.5 µm thickzro 2 layer followed by a 0.6 µm thick PZT layer was deposited by a sol-gel process. The top metal layers of Cr (20nm) and Au (120nm) were deposited via thermal and electron-gun evaporation, respectively on top of the PZT layer. All of the processes until this step are performed as blanket depositions without the use of any lithographic step. The first photolithography step is a front-to-backside alignment. After development, the top Au and Cr layers were etched using a RIE etch process to form the interdigitated capacitors on top of the PZT. Reactive ion etching was used to get a better pattern definition, which is especially required for defining the smallest width electrodes. In the fabrication of these devices the PZT and ZrO 2 layers were not patterned. Instead, backside DRIE etch was used to delineate the entire device. This was followed by separating the individual devices by breaking them at the thin membrane areas where the etching stopped on the silicon dioxide layer. By using this technique we were able to reduce the number of mask steps (from for a d 31 mode device [7]) and simplify the fabrication process. For the DRIE (Deep Reactive Ion Etch) preparation to define the die frames, diaphragms, and proof masses of the annular structures, thick photoresist AZ4620 was used and spun up to a thickness of ~13µm. The first DRIE step performed determines the thickness of the diaphragms. The etch was performed for 25 minutes and resulted in an etched depth of 50 +1.5µm. This is followed by a buffered oxide etch (BOE) to remove the patterned oxide on the diaphragm areas, after which a second DRIE step is used to etch through the wafers (Figure 2). Visual end point detection was used by examining the areas between adjacent accelerometer dies for the complete removal of substrate silicon. A deep reactive ion etch over a wafer is not uniform and tends to etch the edges of the wafer much faster than the center of the wafer in a radially symmetric pattern. This etch rate nonuniformity has been found to vary by as much as by 5-7% [7]. Therefore, for a through wafer-etch, the thickness of the diaphragms produced can be expected to vary by as much as 25-35µm acrossthewafer. IV. EXPERIMENTAL RESULTS Two kinds of tests were performed for finding the sensitivities and resonance frequencies of the accelerometers. One is a frequency response measurement and the other is an impedance measurement. In the frequency response measurement, the output of the PZT accelerometer under test is compared to a calibrated reference sensor in the frequency domain on mechanical excitation. Both the sensors are mounted on the same shaker table and the relative response in db was measured. The relative response was converted into an absolute sensitivity figure for the PZT accelerometer using the sensitivity specification of the reference (calibrated) sensor. The PZT accelerometer is mounted on top of a reference sensor to get same amount of acceleration. Figure 3 shows a schematic diagram of the measurement set-up. A charge amplifier circuit was used for the measurement of the output of the sensors. The circuit was designed to have a charge sensitivity of 10mV/pF in the frequency range of interest. MEMS Accelerometer SR 785 Dynamic Signal Analyzer Reference Accelerometer Shaker Power Amplifier Figure 3. Schematic of the equipment setup for the frequency response measurement The response characteristics of the accelerometers show the classic spring-mass-dashpot response curve. The low frequency flat response is followed by an increase in the response as the first resonance frequency is approached and Proc. SPIE Vol. 4559 133

is followed by a rapid decay in the response. The Q-factor determines the sharpness and the peak height of the response at the resonance frequency. Three different devices, labeled as devices 1-3, were tested for their performance. For devices 1and 2 the inner capacitor was used for the measurements reported here whereas for device 3, the outer capacitor was used. Figure 4 shows a typical frequency response for the accelerometers. The flat area before the resonance peak was used for sensitivity calculations. All accelerometers showed a decrease in their response at frequencies below 2500Hz. Since the PZT films have been shown to have a flat response down to 250Hz [7], we believe that this reduction in response at the low frequencies is due to the charge amplifier circuit that was used in these measurements. As mentioned earlier, the nonuniformity of the deep-etch resulted in devices with different sensitivity and resonance characteristics. Devices located at radially symmetric positions on the silicon wafer had approximately the same diaphragm thickness and were observed to have similar sensitivity (0.13-0.17pC/g) and resonance frequencies (~15kHz). However, considerable differences in the sensitivity and resonance frequency were obtained when the devices were located at different radial positions on the wafer. For example a similar device from the center of the wafer showed 22.4kHz and a 0.077pC/g as resonance frequency and charge sensitivity respectively. Figure 4. Frequency response curve for device 3, outer capacitor with 5µm electrode width The resonance frequency of the accelerometers was also measured using impedance measurement of the PZT interdigitated capacitor as a function of frequency. At the resonance frequency, the impedance response shows rapid change. This measurement is shown in Figure 5 and the resonance frequency of 15.2kHz agrees quite well with the frequency response measurement. V. DISCUSSION The sensitivity of the PZT accelerometers was also estimated for the value of the normal force applied in the experiments. The applied normal force was calculated by multiplying the proof mass of the PZT accelerometer with the applied normal acceleration (output of the reference sensor). Analytical expressions were used for estimating the stress generated in the electrode regions of the diaphragm under the influence of the applied normal force [9]. Using this stress value and the number of interdigitated capacitors, the output charge was estimated as Q out = Σd 33 σ 33i A i =d 33 Σσ 33i A i i = 1,2,,n (1) 134 Proc. SPIE Vol. 4559

90000 85000 Impedance Phase -86-86.5 Impedance 80000 75000 70000 65000-87 -87.5-88 -88.5 Phase 60000-89 12000 13000 14000 15000 16000 17000 18000 Frequency Figure 5. Impedance response measurement for device 3. where,q out is the output charge, n is the number of interdigitated capacitors connected in parallel, d 33 is the piezoelectric coefficient of the PZT film, σ 33i is the stress generated at the i th interdigitated capacitor under the influence of the applied normal force, and A i is the cross sectional area of the i th interdigitated capacitor. A d 33 value of 120pC/N was used from the experimentally measured values of similar PZT films. To predict the sensitivity value of the accelerometers, using eq. (1), the thickness of the membrane is also required. However, the exact thickness of the membranes could not be experimentally measured since the etched depth (membrane area) was too large to use a surface profilometer. Therefore, the measured resonance frequencies were used to estimate the annular membrane thickness [8]. Table 1 summarizes the calculated membrane thickness from the resonance frequency for the three devices. A large thickness difference ~16µm is observed between the thickness of the devices reported here. Membrane thickness is one of the primary factors that affect the accelerometer characteristics. Therefore, uniform membrane thickness over the wafer is necessary for the control of the accelerometer characteristics. This, however, could not be achieved using the DRIE machine that was used for the fabrication of the devices reported here. Devices 2 and 3 are from radially equivalent locations on the silicon wafer and have similar membrane thickness whereas the device 1 is from the center of the wafer and has a membrane ~16µm thicker than those of devices 2 and 3. Measured resonance frequency (khz) Calculated membrane Thickness (µm) Device 1 22.4 71.1 Device 2 15.4 55.4 Device 3 15.3 55.1 Table 1. Resonance frequency and membrane thickness Using these membrane thickness values, the sensitivities of the fabricated accelerometers were calculated and are listed in Table 2. Given the errors in the estimated thickness of the diaphragms, and the errors in the d 33 coefficients used, the predicted values and the experimental results seem to agree quite well for capacitors located at the inner diameter of the diaphragm (device 1and 2) but seems to breakdown at the outer diameter of the diaphragm (device 3). These inadequacies of the predicted values will be addressed in future work which will include experimental determination of the thickness of the diaphragms using a vertical interference microscope and using finite element Proc. SPIE Vol. 4559 135

models to predict the stress generated in the diaphragms. A voltage sensitivity of 0.081-0.247mV/g for the d 31 mode PZT accelerometers was calculated by dividing the reported charge sensitivity of 0.5-7.6pC/g by the capacitance of the devices [7]. Comparing these to the values of d 33 accelerometer reported here, the voltage sensitivity of the d 33 accelerometers seems ~10 time higher than that of d 31 accelerometers. This is to be expected since the relative reduction in the cross sectional area (i.e. capacitance) of the devices is larger than the reduction in the charge collected through the same area (i.e. output) [10]. However, for the very same reason, the d 33 accelerometers have lower charge sensitivity as compared to d 31 accelerometers. Type Charge Sensitivity (pc/g) Voltage Sensitivity (mv/g) Value Measured Calculated Measured Calculated Device 1 0.077 0.045 0.85 0.5 Device 2 0.13 0.08 1.67 1.03 Device 3 0.17 0.84 1.07 5.29 Table 2. Comparison of measurement and calculation of charge and voltage sensitivity for d 33 accelerometers Using PZT accelerometer in the d 33 mode of operation has resulted in several advantages and improvements. First, the fabrication process has been considerably simplified and requires only three mask layers. This has reduced the processing time significantly and eliminated the need for defining PZT and bottom electrode areas, which are required in the fabrication of d 31 mode devices. The second advantage of d 33 mode of operation is that very thin films of PZT can now be used to realize devices with any desired capacitance by simply varying the electrode spacing. The third and most important advantage of d 33 accelerometers is the improved voltage sensitivity of these devices. The voltage sensitivity improvement of d 33 accelerometers was achieved because of the high d 33 coefficient value of PZT thin films and a reduction in capacitance obtained by using the interdigitated capacitor configuration. The associated electronics can now be reduced to a simple voltage amplifier. The advantages of a voltage amplifier against a charge amplifier include a simpler circuit and a larger bandwidth. VI. CONCLUSIONS PZT MEMS accelerometers operating in the d 33 -mode have been designed, fabricated, and tested. PZT accelerometers operating in the d 33 mode were fabricated. Interdigitated capacitors were used to achieve the d 33 mode of operation and deep reactive ion etching to define the proof-mass of the accelerometers. Measurements on these accelerometers show sensitivities ranging from 0.85-1.67 mv/g with resonance frequencies ranging from 22.4-15.4 khz respectively. In addition to the improved sensitivity, advantages of d 33 -mode accelerometers include use of thinner PZT films, and the ability to optimize the impedance of the device to achieve a higher signal-to noise-ratio. Preliminary estimates of the accelerometer performance seem to agree reasonably with the measured values. d 33 -accelerometers showed a voltage sensitivity improvement by a factor of ~10 as compared to similar d 31 accelerometers reported earlier. Future work will include, finite element modeling of the accelerometers, optimization of the interdigitated electrode design to obtain highest voltage sensitivity, and temperature compensation. ACKNOWLEDGEMENTS The authors thank the National Institute of Standards and Technology for supporting this research under WR- ATP-0001, and the Penn State and Cornell Nanofabrication Facilities. REFERENCES [1] N. Yazdi, F. Ayazi, and K. Najafi, Micromachined Inertial Sensors, Proceedings of the IEEE, 86 [8] 1640-59 (1998). [2] A.Partridge,J.K.Reynolds,B.W.Chui,E.M.Chow,A.M.Fitzgerald,L.Zhang,N.I.Maluf,andT.W.Kenny, A High-Performance Planar Piezoresistive Accelerometer, Journal of Microelectromechanical Systems, 9 [1] 58-66 (2000). [3] N. Yazdi and K. Najafi, An All-Silicon Single-Wafer Micro-g Accelerometer with a Combined Surface and Bulk Micromachining Process, Journal of Microelectromechanical Systems, 9 [4] 544-50 (2000). 136 Proc. SPIE Vol. 4559

[4] A. Iula, N. Lamberti, and M. Pappalardo, Analysis and Experimental Evaluation of a New Planar Piezoelectric Accelerometer, IEEE/ASME Transactions on Mechatronics, 4 [2] 207-12 (1999). [5] P. Muralt, PZT Thin Films for Microsensors and Actuators: Where Do We Stand?, IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 47 [4] 903-15 (2000). [6] B. Xu, Y. Ye, J. J. Bernstein, R. Miller, and L. E. Cross, Dielectric Hysteresis from Transverse Electric Fields in Lead Zirconate Titanate Thin Films, Applied Physics Letters, 74 [23] 3549-51 (1999). [7] Li-Peng Wang, Microelectromechanical Systems (MEMS) Sensors based on Lead Zirconate Titanate (PZT) Films Ph.D. thesis, The Pennsylvania State University, May 2001. [8] Robert D. Blevins, Formulas for Natural Frequency and Mode Shape, Krieger Publishing Company (1979) p244 [9] Warren C. Young, Roark s Formulas for Stress & Strain 6 th edition, McGraw-Hill, Inc (1989) pp 398,402 [10] B. Xu, R. G. Polcawich, S. Trolier-McKinstry, Y. Ye, L. E. Cross, J. J. Bernstein, and R. Miller, Sensing Characteristics of In-Plane Polarized Lead Zirconate Titanate Thin Films, Appl. Phys. Lett. 75(26) 4180-4182 (1999). Proc. SPIE Vol. 4559 137