- Email: [email protected]
A silicon carbide capacitive pressure sensor for in-cylinder pressure measurement
Available online at www.sciencedirect.com
Sensors and Actuators A 145–146 (2008) 2–8
A silicon carbide capacitive pressure sensor for in-cylinder pressure measurement Li Chen a,∗ , Mehran Mehregany b a
b
Department of Material Science and Engineering, Case Western Reserve University Cleveland, OH 44106, USA Department of Electrical Engineering and Computer Science, Case Western Reserve University Cleveland, OH 44106, USA Received 9 July 2007; received in revised form 13 September 2007; accepted 18 September 2007 Available online 2 October 2007
Abstract This paper reports a research prototype of a low-cost, miniature, mass-producible sensor for measurement of high-pressure at operating temperatures of 300–600 ◦ C, e.g., in-cylinder engine pressure monitoring applications. This all-silicon carbide (SiC) capacitive sensor, i.e., a SiC diaphragm on a SiC substrate, takes advantage of the excellent harsh-environment material properties of SiC and is fabricated by surface micromachining. The sensor is packaged in a high-temperature ceramic package and characterized under static pressures of up to ∼5 MPa (700 psi) and temperatures of up to 574 ◦ C in a custom chamber. An instrumentation amplifier integrated circuit is used to convert capacitance into voltage for measurements up to 300 ◦ C; beyond 300 ◦ C, the capacitance is measured directly from an array of identical sensor elements using a LCZ meter. After high-temperature soaking and several tens of temperature/pressure cycles, packaged sensors continue to show stable operation. For monitoring the dynamic cylinder pressure in the combustion chamber, the sensor is packaged in a custom probe and inserted into the cylinder head of a research internal combustion engine. The sensor efficacy is verified against the reference probe used for monitoring pressure in the research engine. © 2007 Elsevier B.V. All rights reserved. Keywords: Silicon carbide; Poly-SiC; Capacitive; High-temperature; Harsh-environment; Pressure sensor; Engine pressure; In-cylinder pressure
1. Introduction To enhance fuel efficiency, reduce emissions and improve reliability for future vehicles, it is necessary to optimize the combustion process by using a pressure-based engine management solution. For this type of high-temperature, harsh-environment application, a robust pressure sensor using high-temperature material has to be developed. At the same time, the cost per sensor must be low enough in automotive quantities to allow such technology insertion. Prototypes of SiC piezoresistive and capacitive pressure sensors have been reported in recent years but only the piezoresistive prototypes are all-SiC [1]. Others have utilized a SiC pressuresensing diaphragm on a Si substrate [2–4], but the devices suffer from thermal expansion mismatch between the diaphragm and the substrate. Concerns related to the piezoresistive approach
∗
Corresponding author. Tel.: +1 216 368 4593; fax: +1 216 368 3209. E-mail address: [email protected] (L. Chen).
0924-4247/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2007.09.015
include: (i) the relatively small gage factor of SiC at high temperatures; (ii) temperature sensitivity of the piezoresistive readout; (iii) junction leakage current at high temperatures; (iv) the challenge of bulk-etching a SiC substrate to realize the sensing diaphragm. The capacitive approach overcomes these concerns, albeit at the cost of a high-temperature interface electronics requirement, a separate development in our group to realize SiC interface integrated circuits capable of 300–600 ◦ C operation [5]. An all-SiC capacitive implementation overcomes both the piezoresistive concerns, as well as the thermal mismatch issue. In this paper, a surface-micromachined capacitive pressure sensor is presented, which is comprised of a polycrystalline SiC (poly-SiC) diaphragm on a poly-SiC substrate. High-pressure operation of the sensor die in a high-temperature ceramic package is demonstrated up to 574 ◦ C (limited by the testing setup) in a custom chamber. To demonstrate its operation in dynamic pressure environment in an internal combustion engine, the sensor chip is packaged into a custom probe and inserted into a single-cylinder internal combustion test engine.
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
3
2. Sensor design A series of edge-clamped circular sensing diaphragms of different diameter, operating in non-touch (small deflection) and touch modes, were designed for a range of pressure application specifications. The sensor’s operational modes were modeled by finite element analysis (FEA) to predict performance. A three-dimensional analysis was performed for enhanced visualization, even though more computationally expensive than a simpler two-dimensional axisymmetric model. The Young’s modulus of our heavily doped poly-SiC is ∼330 GPa, previously determined by bulk-micromachined membranes and surface-micromachined resonators fabricated from our [6]. Fig. 1 shows an example three-dimensional model and the predicted performance for a diaphragm operated in small deflection mode up to ∼1.8 MPa (250 psi). The sensor diaphragm has a diameter of 136 m and a thickness of 3 m, while the zero-pressure capacitive gap is 1.5 m. Fig. 2 shows the contact-mode operation of a 194 m-diameter diaphragm sensor
Fig. 2. Sensor modeling results: (a) three-dimensional finite element simulation for contact-mode operation in the pressure range of ∼1.8 MPa (250 psi) to ∼3.5 MPa (500 psi) and (b) linear fit to the model prediction showing a non-linearity of 0.5% (deflection not to scale in (a)).
Fig. 1. Sensor modeling results: (a) three-dimensional finite element simulation for small deflection (i.e., non-contact mode) operation up to a pressure load of ∼1.8 MPa (250 psi) and (b) linear fit to the model prediction showing a nonlinearity of 0.8% (deflection not to scale in (a)).
(i.e., otherwise the same diaphragm thickness and zero-pressure gap) for the pressure range of ∼1.8–3.5 MPa (250–500 psi). The capacitance characteristics presented in Figs. 1b and 2b were calculated corresponding to the diaphragm deflection, which accounts for non-linear deflection behavior, as well as membrane middle-plane stretching and film residual stress. These modeling results demonstrate that acceptable sensitivity and linearity can be achieved for a selected pressure range even when the diaphragm thickness and zero-pressure capacitive gap is fixed by the process. The sensor’s model is used to verify the measured performance data by incorporating measured process parameters in place of nominal design values. The sensor architecture implements a suspended poly-SiC diaphragm with multiple surrounding anchor points and release access channels to facilitate the diaphragm cavity’s subsequent wafer-scale release and sealing. Although the diaphragm clamped edge boundary condition used in FEA modeling may seem to be interrupted by release access channels, the ensuing patterned LTO seal ring around the periphery provides
4
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
robust edge clamping as long as there are enough alternating anchors/release channels. FEA showed that 12 or more anchor segments around the diaphragm periphery (each anchor or release channel segment corresponding to 15◦ arc in our design) essentially simulates fully clamped edge. When this number drops to 8 or less, the deviation is appreciable. In this context, the fact that our diaphragm thickness is a factor two larger than the boundary step height and that the diaphragm radius is large compared to its thickness is also helpful. Two types of die layouts, each 4.5 mm by 4.5 mm, are designed for the two temperature regimes of data retrieval. Type A dies have a single sensing element (variable capacitor) and a reference element (fixed capacitor); the two elements are used in a differential bridge circuit in conjunction with an instrumentation amplifier integrated circuit (IC) capable of operation up to 300 ◦ C. For temperatures in excess of 300 ◦ C, Type B dies having an array of sensor elements to increase capacitance (change) are used, allowing capacitance readout directly by a LCZ meter since the capacitance changes are relatively large. The relatively large chip size was selected to simplify
Fig. 3. Cross-sectional description of the sensor fabrication process: (a) after lower electrode patterning; (b) after electrical isolation of the lower electrode; (c) after LTO sacrificial layer deposition and subsequent patterning of the anchor holes; (d) after diaphragm patterning; (e) after wafer-scale release; (f) after diaphragm cavity sealing; (g) after metallization patterning.
Table 1 Summary of sensor performance data
Range (psi) Sensitivity Nonlinearity (%) Hysteresis (%) Resolution (psi) Temperature Coefficient (%)
RT
300 ◦ C
574 ◦ C
750 272 V/psi 0.5 0.7 2 N/A
750 251 V/psi 1.7 2.4 2 0.04
100 7.2 fF/psi 2.4 N/A 7 0.05
handling during backend processes; it also allowed uniform die size with sensor diaphragm array implementations (e.g., 130 elements for the case of 194 m-diameter diaphragm sensor chip). 3. Device fabrication Fig. 3 depicts the sensor fabrication process. A CVD-grown, 100 mm diameter, high-resistivity poly-SiC wafer is used for the substrate. A 300-nm thick, low-stress Si3 N4 film is deposited by LPCVD, followed by deposition of an in situ ammonia-doped, 200-nm thick, low-resistivity LPCVD poly-SiC film [6]. This poly-SiC film is patterned by fluorine-based plasma chemistries directly using photoresist as masking layer to form the bottom electrodes and routing interconnects, and has a resistivity of 0.02 cm. Next, a 400-nm thick, low-stress LPCVD Si3 N4 film is deposited and patterned to open electrical contact windows to the bottom electrodes, as well as to electrically isolate the lower electrodes during contact-mode operation. A 1.4-m thick LTO film is then deposited as the sacrificial spacer, and multiple anchor holes are patterned. A 2.8-m thick, in situ ammonia-doped, low-resistivity polySiC film is deposited by LPCVD and patterned to form the sensing diaphragm. This poly-SiC film has a resistivity of 0.005 cm and a residual tensile stress of 190 MPa. Due to the relative large thickness of this SiC layer, a 200-nm thick sputtered nickel thin film was used as hard mask patterned in a wet etch. The SiC film was etched in a STS® Inductively Coupled Plasma (ICP) etcher using SF6 plasma. The LTO sacrificial spacer is next removed in HF through the multiple release access channels surrounding the diaphragm. A 2.5-m thick LTO layer is then deposited to seal the released diaphragm cavity. This film is densified at 1000 ◦ C for 30 min and then patterned into a ring over the periphery of the diaphragm. Finally, a multi-layer stack of Ni (200 nm)/Cr (20 nm)/Au (500 nm) is deposited for high-temperature metallization, followed by rapid thermal annealing at 750 ◦ C for 10 s. Ni is chosen to form a good ohmic contact with SiC; Au is used as an over-layer, with Cr for adhesion in order to provide for reliable Au wire-bonding in subsequent packaging. Fig. 4 shows SEM diagnostic views after completion of sensor fabrication. Regarding long-term reliability of the employed metallization at high temperatures, the findings from the sensor testing at high temperatures (described in the following section) over the past 7 months are promising in performance consistency, though more work is needed to quantify our observations.
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
5
Fig. 4. SEM inspection of fabricated poly-SiC sensors: (a) tilted view of a vacuum-sealed, 194 m-diameter diaphragm with the insets showing magnifications of the diaphragm surface and periphery and (b) cross-sectional views of a sensor similar to that in (a) with the insets showing magnifications of the diaphragm/gap and the sealed edge.
4. Sensor characterization 4.1. Static measurements Table 1 summarizes the sensor performance data achieved in this work based on bench-top testing described in Section 4.1. 4.1.1. Voltage readout with Type A dies A 300 ◦ C bulk CMOS instrumentation amplifier IC recently developed by Yu and Garverick of the Case Western Reserve University’s Mixed Signal IC Lab was used to convert capaci-
tance into voltage [7]. The IC is highly integrated, requiring just two off-chip capacitors, which minimizes the possibility of components and interconnect failure at high temperature. The usual bulk CMOS degradations at high temperature due to increased junction leakage, unstable threshold voltage, and reduced mobility are compensated using a number of analog design techniques, including constant-gm baising, switched-capacitor (SC) signal processing, fully differential topology and correlated double sampling (CDS). Critical analog blocks, such as the opamps, have been carefully designed to improve high-temperature performance. The IC includes a CDS programmable-gain pre-
6
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
Fig. 5. Sensor package: (left) top view of the high-temperature ceramic DIP package with SiC pressure sensor dies and interface IC and (right) backside view of the DIP package, showing the 4-wire I/O configuration.
amplifier that can be configured as a charge amplifier to interface with capacitive sensors. A delta modulator is used to amplify and convert discrete-time CDS amplifier output to continuous-time analog output. The circuits were fabricated using a low-cost 1.5 m bulk CMOS process through MOSIS. For testing Type A dies, a hybrid high-temperature ceramic dual-in-line package (DIP) was designed with simple 4-wire I/O (see Fig. 5), both sensor and IC chips were mounted on the ceramic package substrate using AREMCO 569-Ceramabond® high-temperature adhesives. The sensor performance was characterized in a custom high-pressure chamber with temperature control. Pressure measurements were performed at temperatures up to 300 ± 5 ◦ C and pressures up to ∼5 MPa (700 psi). Fig. 6 shows measurement data for a typical sensor over the studied pressure and temperature ranges, as well as the non-touch and touch modes of operation predicted by finite element modeling using actual sensor dimensions from the SEM diagnostic views. The data and model are in good agree-
Fig. 6. Data and model for a sensor similar to the one in Fig. 4. Non-contact and contact modes of linear operation are boxed as Zones I & II, respectively.
Fig. 7. Type B sensor performance example: (a) data and model for a typical device with 86 m-diameter sensor elements operating in small deflection (i.e., non-contact) mode and (b) data and model for a typical device with 280 mdiameter sensor elements operating in small deflection and contact modes. In (b), non-contact and contact modes of linear operation are boxed as Zones I & II, respectively.
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
ment. This 194 m-diameter sensor example shows a room temperature sensitivity of 272 V/psi in the linear range of ∼1.4–3.2 MPa (200–450 psi), with nonlinearity of 0.5% and hysteresis of 0.7%. At 300 ◦ C, the sensitivity is 251 V/psi, and the nonlinearity and hysteresis are 1.7% and 2.4%, respectively. The temperature coefficient of capacitance change is less than 0.04%. 4.1.2. Capacitance readout with Type B dies A typical Type B die uses an array of sensor elements to provide large capacitance changes. Packaging of the Type B die in the DIP used Au wire bonding. A four-wire measurement method using a LCZ meter was utilized to reach characterization temperatures higher than 300 ◦ C. The data and model for a typical Type B die example (172 m-diameter diaphragm array) operating in small deflection (non-contact) mode at temperatures up to 574 ◦ C and pressures up to 100 psi are presented in Fig. 7(a). Although the stainless-steel custom chamber was designed to withstand 1500 psi limit, lacking of chamber wall cooling makes the testing at higher pressure in the temperature range of 400–600 ◦ C a safety concern, hence limiting us to 100 psi testing for the high-temperature range at this time. At 574 ◦ C, the achieved sensitivity and nonlinearity are 7.2 fF/psi and 2.4%, respectively; the temperature coefficient of capacitance change is 0.05%. The readout error of 0.05 pF – due to the precision limit of the LCZ meter’s display – is indicated by error bars in Fig. 7(a). The repeatability data is shown in Fig. 7(b) for a sample die with 282 m-diameter diaphragm testing at 250 ◦ C up to 500 psi on three different dates. Stability studies were carried out after several tens of thermal (25–500 ◦ C) and pressure (full range) cycles. Sensor readout variation at different temperatures was recorded for 48 h and was below 0.1 pF.
7
4.2. Dynamic measurements For monitoring the dynamic cylinder pressure in combustion chamber with the characteristic of cyclic 4-stroke dynamic pressure loading (i.e., suction, compression, working and exhaust), a stainless steel probe package is designed. SiC sensor die was mounted onto the ceramic carrier using the same hightemperature ceramic adhesive mentioned in Section 4.1 above. The gold-coated metal pins were plugged into connectors on feed-through wires. The tube-shape probe package was then threaded into the cylinder head of a transparent single-cylinder research internal combustion engine MEGATECH® MARK III in an engine lab in the Mechanical Engineering Department of Case Western Reserve University (see Fig. 8). The transparent cylinder allows visual observation of the combustion flame after engine was started. A commercially available semiconductor piezoresistive pressure transducer from Omega is as a reference. Sensor diaphragms in this work were designed to have over 20 kHz frequency bandwidth to cover application requirements from engine environment. Thus, the dynamic resolution of the signal-monitoring instrument should be sufficient to follow sensor response. An automated data acquisition system was setup. An LCR meter having milliseconds resolution and a frequency range of 20–2 MHz was used to record the dynamic data. A Labview code was written to directly transfer raw data into a computer. Alcohol was used as fuel source in this test engine. The air/fuel ratio could only be adjusted manually. Due to mechanical deficiencies, this test engine could only run smoothly below 1000 rpm; at higher rpm, stall occurred frequently. The engine was run at 750 rpm under partial loads, which corresponded to a temperature of about 150 ◦ C on the sensor
Fig. 8. (Left) Transparent single-cylinder combustion engine with custom-designed sensor package placed onto the cylinder head; (right) the close-up view of the combustion process while engine running.
8
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
References
Fig. 9. Typical cylinder pressure measured with a reference pressure sensor and the SiC capacitive pressure sensor at 750 rpm.
chip as monitored by a thermocouple. The relatively low sensor surface temperature is also due to the gap between the sensor chip location and cylinder head hole because of the modified fitting design. This condition simulates the car running in idle state; the engine actual working status is presented in Fig. 8 with combustion flame on. Fig. 9 shows the output signal (pressure) measured with both sensors after calibrations in terms of time (or crank angle). The pressure surge time is about 34 ms, and the period of each cycle is about 133 ms. The SiC sensor not only matches the reference sensor measurements but also shows better performance at the low-pressure regime of the cycle. 5. Conclusion Research prototypes of an all-SiC, surface-micromachined capacitive pressure sensors suitable for in-cylinder pressure monitoring were demonstrated. The sensor elements are comprised of poly-SiC diaphragms fabricated on a poly-SiC substrate, overcoming the diaphragm/substrate thermal mismatch, which is so critical to high-temperature applications. Operation of packaged devices is demonstrated up to 574 ◦ C. The stability and performance reproducibility after several tens of temperature/pressure cycles are promising. Finally, an on-engine test package was developed to demonstrate sensor operation in a simulated engine environment. Acknowledgments The authors thank Prof. Steven Garverick and his group for the high-temperature IC used in characterizing Type A dies. Thanks also go to Prof. Chih-Jen Sung and Mr. David Conger for letting us use their research test engine for dynamic measurement in Mechanical Engineering department. This work was supported in part by Grant # NBCH1050002 from DARPA. Mehregany as a scientific advisor/partner of and Case Western Reserve University as a licensor of related technology to FLX Micro, Inc. (Cleveland, OH) have a financial interest in the products or methods being investigated in this study.
[1] R. Okojie, A. Ned, A. Kurtz, Operation of a 6H–SiC pressure sensor at 500 ◦ C, in: Proceedings of the International Conference on Solid-State Sensors and Actuators, Chicago, IL, 1997, pp. 1407–1409. [2] R. Ziermann, J. Berg, W. Reichert, E. Obermeier, M. Eickhoff, G. Krotz, A high temperature pressure sensor with SiC piezoresistors on SOI substrates, in: Proceedings of the International Conference on Solid-State Sensors and Actuators, Chicago, IL, 1997, pp. 1411–1414. [3] J. Von Berg, R. Ziermann, W. Reichert, E. Obermeier, M. Eickhoff, G. Krotz, U. Thoma, C. Cavalloni, J.P. Nendza, Measurement of the cylinder pressure in combustion engines with a piezoresistive -SiC-on-SOI pressure sensor, in: Proceedings of the fourth International High Temperature Electronics Conference, Albuquerque, NM, 1998, pp. 245–249. [4] D. Young, J. Du, C. Zorman, W. Ko, High-temperature single-crystal 3C-SiC capacitive pressure sensor, IEEE Sens. J. 4 (4) (2004) 464–470. [5] A. Patil, X. Fu, C. Anupongongarch, M. Mehregany, S. Garverick, Characterization of silicon carbide differential amplifiers at high temperature, in: Proceedings of the IEEE Compound Semiconductor IC Symposium, Portland, OR, October, 2007, pp. 139–142. [6] J. Trevino, X. Fu, C. Zorman, M. Mehregany, Low-stress heavily doped polycrystalline silicon carbide for MEMS applications, in: Proceedings of IEEE MEMS Conference, Orlando, FL, 2005, pp. 451–454. [7] X. Yu, High-temperature bulk CMOS integrated circuits for data acquisition, Ph.D. Dissertation, Department of Electrical Engineering and Computer Science, Case Western Reserve University, 2006.
Biographies Li Chen received his BS in materials science and engineering from Shanghai University of Science and Technology in 1996, and his MS in materials science and engineering from Case Western Reserve University in 2003. From 1996 to 1999, he was an engineer in the Automotive Steel Research Center at Shanghai BAOSTEEL Corporation, China. In 2000, he joined OLYMPUS Microscope (China) as a sales representative. He is currently a PhD candidate on a Case Prime Fellowship at Case Western Reserve University, Cleveland, OH. His research interests are in the areas of SiC MEMS materials characterization and fabrication technologies, sensor device development for harsh-environment and demanding applications, and MEMS packaging. Mr. Chen is a student member of the IEEE and Material Research Society (MRS). Mehran Mehregany received his BS in electrical engineering from the University of Missouri in 1984, and his MS and PhD in electrical engineering from Massachusetts Institute of Technology in 1986 and 1990, respectively. From 1986 to 1990, he was a consultant to the Robotic Systems Research Department at AT&T Bell Laboratories, where he was a key contributor to ground-breaking research in microelectromechanical systems (MEMS). In 1990, he joined the Department of Electrical Engineering and Applied Physics at Case Western Reserve University as an assistant professor. He was awarded the Nord Assistant Professorship in 1991, was promoted to associate professor with tenure in July 1994 and was promoted to full professor in July 1997. He held the George S. Dively Professor of Engineering endowed chair from January 1998 until July 2000, when he was appointed to the Goodrich Professor of Engineering Innovation endowed chair. He served as the director of the MEMS Research Center at Case from July 1995 until December 2002. Since January 2003, he has been Chairman of the Electrical Engineering and Computer Science Department at Case. Prof. Mehregany is well known for his research in the area of MEMS; he has over 270 publications describing his work, holds 15 U.S. patents, and is the recipient of a number of awards/honors. He served as the editor-in-chief of the Journal of Micromechanics and Microengineering (January 1996 to December 1997), assistant-to-the-president of the Transducers Research Foundation (1994–2004) and is currently an editor for the Journal of Microelectromechanical Systems. His research pursues innovation in high-performance sensors and actuators for demanding and harsh-environment applications by combining leading-edge capabilities in devices and circuits, design and modeling, materials and fabrication, and testing and packaging.
Sensors and Actuators A 145–146 (2008) 2–8
A silicon carbide capacitive pressure sensor for in-cylinder pressure measurement Li Chen a,∗ , Mehran Mehregany b a
b
Department of Material Science and Engineering, Case Western Reserve University Cleveland, OH 44106, USA Department of Electrical Engineering and Computer Science, Case Western Reserve University Cleveland, OH 44106, USA Received 9 July 2007; received in revised form 13 September 2007; accepted 18 September 2007 Available online 2 October 2007
Abstract This paper reports a research prototype of a low-cost, miniature, mass-producible sensor for measurement of high-pressure at operating temperatures of 300–600 ◦ C, e.g., in-cylinder engine pressure monitoring applications. This all-silicon carbide (SiC) capacitive sensor, i.e., a SiC diaphragm on a SiC substrate, takes advantage of the excellent harsh-environment material properties of SiC and is fabricated by surface micromachining. The sensor is packaged in a high-temperature ceramic package and characterized under static pressures of up to ∼5 MPa (700 psi) and temperatures of up to 574 ◦ C in a custom chamber. An instrumentation amplifier integrated circuit is used to convert capacitance into voltage for measurements up to 300 ◦ C; beyond 300 ◦ C, the capacitance is measured directly from an array of identical sensor elements using a LCZ meter. After high-temperature soaking and several tens of temperature/pressure cycles, packaged sensors continue to show stable operation. For monitoring the dynamic cylinder pressure in the combustion chamber, the sensor is packaged in a custom probe and inserted into the cylinder head of a research internal combustion engine. The sensor efficacy is verified against the reference probe used for monitoring pressure in the research engine. © 2007 Elsevier B.V. All rights reserved. Keywords: Silicon carbide; Poly-SiC; Capacitive; High-temperature; Harsh-environment; Pressure sensor; Engine pressure; In-cylinder pressure
1. Introduction To enhance fuel efficiency, reduce emissions and improve reliability for future vehicles, it is necessary to optimize the combustion process by using a pressure-based engine management solution. For this type of high-temperature, harsh-environment application, a robust pressure sensor using high-temperature material has to be developed. At the same time, the cost per sensor must be low enough in automotive quantities to allow such technology insertion. Prototypes of SiC piezoresistive and capacitive pressure sensors have been reported in recent years but only the piezoresistive prototypes are all-SiC [1]. Others have utilized a SiC pressuresensing diaphragm on a Si substrate [2–4], but the devices suffer from thermal expansion mismatch between the diaphragm and the substrate. Concerns related to the piezoresistive approach
∗
Corresponding author. Tel.: +1 216 368 4593; fax: +1 216 368 3209. E-mail address: [email protected] (L. Chen).
0924-4247/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2007.09.015
include: (i) the relatively small gage factor of SiC at high temperatures; (ii) temperature sensitivity of the piezoresistive readout; (iii) junction leakage current at high temperatures; (iv) the challenge of bulk-etching a SiC substrate to realize the sensing diaphragm. The capacitive approach overcomes these concerns, albeit at the cost of a high-temperature interface electronics requirement, a separate development in our group to realize SiC interface integrated circuits capable of 300–600 ◦ C operation [5]. An all-SiC capacitive implementation overcomes both the piezoresistive concerns, as well as the thermal mismatch issue. In this paper, a surface-micromachined capacitive pressure sensor is presented, which is comprised of a polycrystalline SiC (poly-SiC) diaphragm on a poly-SiC substrate. High-pressure operation of the sensor die in a high-temperature ceramic package is demonstrated up to 574 ◦ C (limited by the testing setup) in a custom chamber. To demonstrate its operation in dynamic pressure environment in an internal combustion engine, the sensor chip is packaged into a custom probe and inserted into a single-cylinder internal combustion test engine.
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
3
2. Sensor design A series of edge-clamped circular sensing diaphragms of different diameter, operating in non-touch (small deflection) and touch modes, were designed for a range of pressure application specifications. The sensor’s operational modes were modeled by finite element analysis (FEA) to predict performance. A three-dimensional analysis was performed for enhanced visualization, even though more computationally expensive than a simpler two-dimensional axisymmetric model. The Young’s modulus of our heavily doped poly-SiC is ∼330 GPa, previously determined by bulk-micromachined membranes and surface-micromachined resonators fabricated from our [6]. Fig. 1 shows an example three-dimensional model and the predicted performance for a diaphragm operated in small deflection mode up to ∼1.8 MPa (250 psi). The sensor diaphragm has a diameter of 136 m and a thickness of 3 m, while the zero-pressure capacitive gap is 1.5 m. Fig. 2 shows the contact-mode operation of a 194 m-diameter diaphragm sensor
Fig. 2. Sensor modeling results: (a) three-dimensional finite element simulation for contact-mode operation in the pressure range of ∼1.8 MPa (250 psi) to ∼3.5 MPa (500 psi) and (b) linear fit to the model prediction showing a non-linearity of 0.5% (deflection not to scale in (a)).
Fig. 1. Sensor modeling results: (a) three-dimensional finite element simulation for small deflection (i.e., non-contact mode) operation up to a pressure load of ∼1.8 MPa (250 psi) and (b) linear fit to the model prediction showing a nonlinearity of 0.8% (deflection not to scale in (a)).
(i.e., otherwise the same diaphragm thickness and zero-pressure gap) for the pressure range of ∼1.8–3.5 MPa (250–500 psi). The capacitance characteristics presented in Figs. 1b and 2b were calculated corresponding to the diaphragm deflection, which accounts for non-linear deflection behavior, as well as membrane middle-plane stretching and film residual stress. These modeling results demonstrate that acceptable sensitivity and linearity can be achieved for a selected pressure range even when the diaphragm thickness and zero-pressure capacitive gap is fixed by the process. The sensor’s model is used to verify the measured performance data by incorporating measured process parameters in place of nominal design values. The sensor architecture implements a suspended poly-SiC diaphragm with multiple surrounding anchor points and release access channels to facilitate the diaphragm cavity’s subsequent wafer-scale release and sealing. Although the diaphragm clamped edge boundary condition used in FEA modeling may seem to be interrupted by release access channels, the ensuing patterned LTO seal ring around the periphery provides
4
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
robust edge clamping as long as there are enough alternating anchors/release channels. FEA showed that 12 or more anchor segments around the diaphragm periphery (each anchor or release channel segment corresponding to 15◦ arc in our design) essentially simulates fully clamped edge. When this number drops to 8 or less, the deviation is appreciable. In this context, the fact that our diaphragm thickness is a factor two larger than the boundary step height and that the diaphragm radius is large compared to its thickness is also helpful. Two types of die layouts, each 4.5 mm by 4.5 mm, are designed for the two temperature regimes of data retrieval. Type A dies have a single sensing element (variable capacitor) and a reference element (fixed capacitor); the two elements are used in a differential bridge circuit in conjunction with an instrumentation amplifier integrated circuit (IC) capable of operation up to 300 ◦ C. For temperatures in excess of 300 ◦ C, Type B dies having an array of sensor elements to increase capacitance (change) are used, allowing capacitance readout directly by a LCZ meter since the capacitance changes are relatively large. The relatively large chip size was selected to simplify
Fig. 3. Cross-sectional description of the sensor fabrication process: (a) after lower electrode patterning; (b) after electrical isolation of the lower electrode; (c) after LTO sacrificial layer deposition and subsequent patterning of the anchor holes; (d) after diaphragm patterning; (e) after wafer-scale release; (f) after diaphragm cavity sealing; (g) after metallization patterning.
Table 1 Summary of sensor performance data
Range (psi) Sensitivity Nonlinearity (%) Hysteresis (%) Resolution (psi) Temperature Coefficient (%)
RT
300 ◦ C
574 ◦ C
750 272 V/psi 0.5 0.7 2 N/A
750 251 V/psi 1.7 2.4 2 0.04
100 7.2 fF/psi 2.4 N/A 7 0.05
handling during backend processes; it also allowed uniform die size with sensor diaphragm array implementations (e.g., 130 elements for the case of 194 m-diameter diaphragm sensor chip). 3. Device fabrication Fig. 3 depicts the sensor fabrication process. A CVD-grown, 100 mm diameter, high-resistivity poly-SiC wafer is used for the substrate. A 300-nm thick, low-stress Si3 N4 film is deposited by LPCVD, followed by deposition of an in situ ammonia-doped, 200-nm thick, low-resistivity LPCVD poly-SiC film [6]. This poly-SiC film is patterned by fluorine-based plasma chemistries directly using photoresist as masking layer to form the bottom electrodes and routing interconnects, and has a resistivity of 0.02 cm. Next, a 400-nm thick, low-stress LPCVD Si3 N4 film is deposited and patterned to open electrical contact windows to the bottom electrodes, as well as to electrically isolate the lower electrodes during contact-mode operation. A 1.4-m thick LTO film is then deposited as the sacrificial spacer, and multiple anchor holes are patterned. A 2.8-m thick, in situ ammonia-doped, low-resistivity polySiC film is deposited by LPCVD and patterned to form the sensing diaphragm. This poly-SiC film has a resistivity of 0.005 cm and a residual tensile stress of 190 MPa. Due to the relative large thickness of this SiC layer, a 200-nm thick sputtered nickel thin film was used as hard mask patterned in a wet etch. The SiC film was etched in a STS® Inductively Coupled Plasma (ICP) etcher using SF6 plasma. The LTO sacrificial spacer is next removed in HF through the multiple release access channels surrounding the diaphragm. A 2.5-m thick LTO layer is then deposited to seal the released diaphragm cavity. This film is densified at 1000 ◦ C for 30 min and then patterned into a ring over the periphery of the diaphragm. Finally, a multi-layer stack of Ni (200 nm)/Cr (20 nm)/Au (500 nm) is deposited for high-temperature metallization, followed by rapid thermal annealing at 750 ◦ C for 10 s. Ni is chosen to form a good ohmic contact with SiC; Au is used as an over-layer, with Cr for adhesion in order to provide for reliable Au wire-bonding in subsequent packaging. Fig. 4 shows SEM diagnostic views after completion of sensor fabrication. Regarding long-term reliability of the employed metallization at high temperatures, the findings from the sensor testing at high temperatures (described in the following section) over the past 7 months are promising in performance consistency, though more work is needed to quantify our observations.
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
5
Fig. 4. SEM inspection of fabricated poly-SiC sensors: (a) tilted view of a vacuum-sealed, 194 m-diameter diaphragm with the insets showing magnifications of the diaphragm surface and periphery and (b) cross-sectional views of a sensor similar to that in (a) with the insets showing magnifications of the diaphragm/gap and the sealed edge.
4. Sensor characterization 4.1. Static measurements Table 1 summarizes the sensor performance data achieved in this work based on bench-top testing described in Section 4.1. 4.1.1. Voltage readout with Type A dies A 300 ◦ C bulk CMOS instrumentation amplifier IC recently developed by Yu and Garverick of the Case Western Reserve University’s Mixed Signal IC Lab was used to convert capaci-
tance into voltage [7]. The IC is highly integrated, requiring just two off-chip capacitors, which minimizes the possibility of components and interconnect failure at high temperature. The usual bulk CMOS degradations at high temperature due to increased junction leakage, unstable threshold voltage, and reduced mobility are compensated using a number of analog design techniques, including constant-gm baising, switched-capacitor (SC) signal processing, fully differential topology and correlated double sampling (CDS). Critical analog blocks, such as the opamps, have been carefully designed to improve high-temperature performance. The IC includes a CDS programmable-gain pre-
6
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
Fig. 5. Sensor package: (left) top view of the high-temperature ceramic DIP package with SiC pressure sensor dies and interface IC and (right) backside view of the DIP package, showing the 4-wire I/O configuration.
amplifier that can be configured as a charge amplifier to interface with capacitive sensors. A delta modulator is used to amplify and convert discrete-time CDS amplifier output to continuous-time analog output. The circuits were fabricated using a low-cost 1.5 m bulk CMOS process through MOSIS. For testing Type A dies, a hybrid high-temperature ceramic dual-in-line package (DIP) was designed with simple 4-wire I/O (see Fig. 5), both sensor and IC chips were mounted on the ceramic package substrate using AREMCO 569-Ceramabond® high-temperature adhesives. The sensor performance was characterized in a custom high-pressure chamber with temperature control. Pressure measurements were performed at temperatures up to 300 ± 5 ◦ C and pressures up to ∼5 MPa (700 psi). Fig. 6 shows measurement data for a typical sensor over the studied pressure and temperature ranges, as well as the non-touch and touch modes of operation predicted by finite element modeling using actual sensor dimensions from the SEM diagnostic views. The data and model are in good agree-
Fig. 6. Data and model for a sensor similar to the one in Fig. 4. Non-contact and contact modes of linear operation are boxed as Zones I & II, respectively.
Fig. 7. Type B sensor performance example: (a) data and model for a typical device with 86 m-diameter sensor elements operating in small deflection (i.e., non-contact) mode and (b) data and model for a typical device with 280 mdiameter sensor elements operating in small deflection and contact modes. In (b), non-contact and contact modes of linear operation are boxed as Zones I & II, respectively.
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
ment. This 194 m-diameter sensor example shows a room temperature sensitivity of 272 V/psi in the linear range of ∼1.4–3.2 MPa (200–450 psi), with nonlinearity of 0.5% and hysteresis of 0.7%. At 300 ◦ C, the sensitivity is 251 V/psi, and the nonlinearity and hysteresis are 1.7% and 2.4%, respectively. The temperature coefficient of capacitance change is less than 0.04%. 4.1.2. Capacitance readout with Type B dies A typical Type B die uses an array of sensor elements to provide large capacitance changes. Packaging of the Type B die in the DIP used Au wire bonding. A four-wire measurement method using a LCZ meter was utilized to reach characterization temperatures higher than 300 ◦ C. The data and model for a typical Type B die example (172 m-diameter diaphragm array) operating in small deflection (non-contact) mode at temperatures up to 574 ◦ C and pressures up to 100 psi are presented in Fig. 7(a). Although the stainless-steel custom chamber was designed to withstand 1500 psi limit, lacking of chamber wall cooling makes the testing at higher pressure in the temperature range of 400–600 ◦ C a safety concern, hence limiting us to 100 psi testing for the high-temperature range at this time. At 574 ◦ C, the achieved sensitivity and nonlinearity are 7.2 fF/psi and 2.4%, respectively; the temperature coefficient of capacitance change is 0.05%. The readout error of 0.05 pF – due to the precision limit of the LCZ meter’s display – is indicated by error bars in Fig. 7(a). The repeatability data is shown in Fig. 7(b) for a sample die with 282 m-diameter diaphragm testing at 250 ◦ C up to 500 psi on three different dates. Stability studies were carried out after several tens of thermal (25–500 ◦ C) and pressure (full range) cycles. Sensor readout variation at different temperatures was recorded for 48 h and was below 0.1 pF.
7
4.2. Dynamic measurements For monitoring the dynamic cylinder pressure in combustion chamber with the characteristic of cyclic 4-stroke dynamic pressure loading (i.e., suction, compression, working and exhaust), a stainless steel probe package is designed. SiC sensor die was mounted onto the ceramic carrier using the same hightemperature ceramic adhesive mentioned in Section 4.1 above. The gold-coated metal pins were plugged into connectors on feed-through wires. The tube-shape probe package was then threaded into the cylinder head of a transparent single-cylinder research internal combustion engine MEGATECH® MARK III in an engine lab in the Mechanical Engineering Department of Case Western Reserve University (see Fig. 8). The transparent cylinder allows visual observation of the combustion flame after engine was started. A commercially available semiconductor piezoresistive pressure transducer from Omega is as a reference. Sensor diaphragms in this work were designed to have over 20 kHz frequency bandwidth to cover application requirements from engine environment. Thus, the dynamic resolution of the signal-monitoring instrument should be sufficient to follow sensor response. An automated data acquisition system was setup. An LCR meter having milliseconds resolution and a frequency range of 20–2 MHz was used to record the dynamic data. A Labview code was written to directly transfer raw data into a computer. Alcohol was used as fuel source in this test engine. The air/fuel ratio could only be adjusted manually. Due to mechanical deficiencies, this test engine could only run smoothly below 1000 rpm; at higher rpm, stall occurred frequently. The engine was run at 750 rpm under partial loads, which corresponded to a temperature of about 150 ◦ C on the sensor
Fig. 8. (Left) Transparent single-cylinder combustion engine with custom-designed sensor package placed onto the cylinder head; (right) the close-up view of the combustion process while engine running.
8
L. Chen, M. Mehregany / Sensors and Actuators A 145–146 (2008) 2–8
References
Fig. 9. Typical cylinder pressure measured with a reference pressure sensor and the SiC capacitive pressure sensor at 750 rpm.
chip as monitored by a thermocouple. The relatively low sensor surface temperature is also due to the gap between the sensor chip location and cylinder head hole because of the modified fitting design. This condition simulates the car running in idle state; the engine actual working status is presented in Fig. 8 with combustion flame on. Fig. 9 shows the output signal (pressure) measured with both sensors after calibrations in terms of time (or crank angle). The pressure surge time is about 34 ms, and the period of each cycle is about 133 ms. The SiC sensor not only matches the reference sensor measurements but also shows better performance at the low-pressure regime of the cycle. 5. Conclusion Research prototypes of an all-SiC, surface-micromachined capacitive pressure sensors suitable for in-cylinder pressure monitoring were demonstrated. The sensor elements are comprised of poly-SiC diaphragms fabricated on a poly-SiC substrate, overcoming the diaphragm/substrate thermal mismatch, which is so critical to high-temperature applications. Operation of packaged devices is demonstrated up to 574 ◦ C. The stability and performance reproducibility after several tens of temperature/pressure cycles are promising. Finally, an on-engine test package was developed to demonstrate sensor operation in a simulated engine environment. Acknowledgments The authors thank Prof. Steven Garverick and his group for the high-temperature IC used in characterizing Type A dies. Thanks also go to Prof. Chih-Jen Sung and Mr. David Conger for letting us use their research test engine for dynamic measurement in Mechanical Engineering department. This work was supported in part by Grant # NBCH1050002 from DARPA. Mehregany as a scientific advisor/partner of and Case Western Reserve University as a licensor of related technology to FLX Micro, Inc. (Cleveland, OH) have a financial interest in the products or methods being investigated in this study.
[1] R. Okojie, A. Ned, A. Kurtz, Operation of a 6H–SiC pressure sensor at 500 ◦ C, in: Proceedings of the International Conference on Solid-State Sensors and Actuators, Chicago, IL, 1997, pp. 1407–1409. [2] R. Ziermann, J. Berg, W. Reichert, E. Obermeier, M. Eickhoff, G. Krotz, A high temperature pressure sensor with SiC piezoresistors on SOI substrates, in: Proceedings of the International Conference on Solid-State Sensors and Actuators, Chicago, IL, 1997, pp. 1411–1414. [3] J. Von Berg, R. Ziermann, W. Reichert, E. Obermeier, M. Eickhoff, G. Krotz, U. Thoma, C. Cavalloni, J.P. Nendza, Measurement of the cylinder pressure in combustion engines with a piezoresistive -SiC-on-SOI pressure sensor, in: Proceedings of the fourth International High Temperature Electronics Conference, Albuquerque, NM, 1998, pp. 245–249. [4] D. Young, J. Du, C. Zorman, W. Ko, High-temperature single-crystal 3C-SiC capacitive pressure sensor, IEEE Sens. J. 4 (4) (2004) 464–470. [5] A. Patil, X. Fu, C. Anupongongarch, M. Mehregany, S. Garverick, Characterization of silicon carbide differential amplifiers at high temperature, in: Proceedings of the IEEE Compound Semiconductor IC Symposium, Portland, OR, October, 2007, pp. 139–142. [6] J. Trevino, X. Fu, C. Zorman, M. Mehregany, Low-stress heavily doped polycrystalline silicon carbide for MEMS applications, in: Proceedings of IEEE MEMS Conference, Orlando, FL, 2005, pp. 451–454. [7] X. Yu, High-temperature bulk CMOS integrated circuits for data acquisition, Ph.D. Dissertation, Department of Electrical Engineering and Computer Science, Case Western Reserve University, 2006.
Biographies Li Chen received his BS in materials science and engineering from Shanghai University of Science and Technology in 1996, and his MS in materials science and engineering from Case Western Reserve University in 2003. From 1996 to 1999, he was an engineer in the Automotive Steel Research Center at Shanghai BAOSTEEL Corporation, China. In 2000, he joined OLYMPUS Microscope (China) as a sales representative. He is currently a PhD candidate on a Case Prime Fellowship at Case Western Reserve University, Cleveland, OH. His research interests are in the areas of SiC MEMS materials characterization and fabrication technologies, sensor device development for harsh-environment and demanding applications, and MEMS packaging. Mr. Chen is a student member of the IEEE and Material Research Society (MRS). Mehran Mehregany received his BS in electrical engineering from the University of Missouri in 1984, and his MS and PhD in electrical engineering from Massachusetts Institute of Technology in 1986 and 1990, respectively. From 1986 to 1990, he was a consultant to the Robotic Systems Research Department at AT&T Bell Laboratories, where he was a key contributor to ground-breaking research in microelectromechanical systems (MEMS). In 1990, he joined the Department of Electrical Engineering and Applied Physics at Case Western Reserve University as an assistant professor. He was awarded the Nord Assistant Professorship in 1991, was promoted to associate professor with tenure in July 1994 and was promoted to full professor in July 1997. He held the George S. Dively Professor of Engineering endowed chair from January 1998 until July 2000, when he was appointed to the Goodrich Professor of Engineering Innovation endowed chair. He served as the director of the MEMS Research Center at Case from July 1995 until December 2002. Since January 2003, he has been Chairman of the Electrical Engineering and Computer Science Department at Case. Prof. Mehregany is well known for his research in the area of MEMS; he has over 270 publications describing his work, holds 15 U.S. patents, and is the recipient of a number of awards/honors. He served as the editor-in-chief of the Journal of Micromechanics and Microengineering (January 1996 to December 1997), assistant-to-the-president of the Transducers Research Foundation (1994–2004) and is currently an editor for the Journal of Microelectromechanical Systems. His research pursues innovation in high-performance sensors and actuators for demanding and harsh-environment applications by combining leading-edge capabilities in devices and circuits, design and modeling, materials and fabrication, and testing and packaging.