A Novel Micro Pressure Sensor Fabrication without Problem of Stiction
Chien-Wei Liu
1, Chie Gau
2and Thin-Lin Horng
31
National Nano Device Laboratories
2
Institute of Aeronautics and Astronautics, National Cheng Kung University
3
Department of Mechanical Engineering, Kun Shan University of Technology NSC Project No.：NSC-91-2212-E-006-108
Abstract
The objective of this paper is to present a novel fabrication process for a or arrays of micro pressure sensors. The fabrication process is almost the reverse of the surface micromachining process used for the pressure sensor. This allows the use of SU-8 to form cavity that can be much deeper for pressure measurement. Therefore, the sensor made can provide a much wider range of pressure measurement. The fabrication process has completely absence of diaphragm sitction. In addition, arrays of pressure sensor can be readily made and integrated into a complicated micro system. More detailed design and fabrication techniques developed for this sensor will be presented.
Keywords: surface micromachining process, micro pressure sensor, SU-8
1. Introduction
Fabrication process of the micro pressure sensor has attracted much attention due to its wide industrial applications in various areas [1]. In general, fabrication of piezoresistive pressure sensor can be classified into the bulk [2,3] and the surface [4,5] micromachining process. However, the surface micormachining process has many advantages over the bulk micromachining process due to the fact that it can produce a much smaller size of cavity underneath the piezoresistive sensor by dry etch and thus, it can produce much more
number of sensors per wafer. In addition, large number of these sensors can be readily made and integrated with a microsystem. However, the diaphragm fabricated above the cavity is usually made of polysilicon layer that stiction can readily occur when the sacrificial layer deposited in the cavity is removed by the wet etch process. Special attention should be made to avoid the problem of stiction [6-9]. In addition, the height of the cavity cannot be made large enough which will not give enough space for diaphragm deformation and allow the pressure measurement in a wider range. This is attributed to the use of sacrificial material, such as silicon oxide, that cannot be deposited thick enough by the CVD process.
The current paper will present a novel fabrication process that can make large number of pressure sensors per wafer without the shortcomings described above.
This fabrication process is the reverse of the usual surface micromachining process for the pressure sensor.
Therefore, the diaphragm layer is deposited first on the silicon substrate, and then the sensor material. It is followed then by spin coat SU-8 layer and forming array of cavities for pressure measurement. This fabrication process has completely absence of stiction.
The height of the cavity made by SU-8 layer can be readily varied from a few microns to hundreds of microns. It gives a much wider range of pressure measurement. A large number of these sensors can be readily made and integrated into a complicated micro
system. Finally, the current pressure sensors made can provide a much better thermal insulation than the ones made by the previous surface micromachining process because of the use of the extremely low thermal conductivity materials such as the SU-8 and the Pyrex glass. This is highly important in the fabrication of a micro thermal system where thermal insulation should be seriously considered. More detail fabrication techniques and calibration of this sensor will also be provided and discussed in this paper.
2. Design and fabrication process
In general, the pressure sensor made by surface micromachining process is started fabricated by etching a cavity into the silicon substrate. Then, a polysilicon layer or a low stress nitride that acts as a diaphragm is deposited to enclose the cavity. Finally, piezoresistive material is deposited on top of the polysilicon layer to sense or measure the deformation of the layer or the pressure inside the cavity. However, before deposition of the polysilicon diaphragm a sacrificial layer such as silicon oxide or PSG is needed which is deposited to cover the cavity. Then, a planarization process, such as CMP, is required to fatten the sacrificial layer. However, the sacrificial layer can also be created by thermal oxidation of the silicon in the cavity, due to expansion of silicon oxide, until the oxide reaches the height on the top surface of the substrate. In this way, the planarization process can be omitted. However, the cavity depth can be made only on the order of 1µm, which significantly reduces the range for deflection of the diaphragm, and the range for pressure measurement.
Even using deposition of a thick sacrificial layer, usually with LP or PECVD, the maximum cavity depth that can be made is less than 10 µm. This makes the deflection of diaphragm or the range for the pressure measurement larger. However, in addition to the large residual stress there are a lot of problems to be resolved before successfully depositing a thick sacrificial layer.
In addition, during removal of the sacrificial oxide layer, the top diaphragm will be driven downward by the
surface tension of drying water to touch the bottom wall of the cavity. Once the diaphragm touches the bottom, it can never be separated and recovered from the wall, and eventually result in fatal and permanent damage. Many stiction relief methods that have been proposed and tested have been partially successful [6-10].
In order to resolve the narrow depth of the cavity and stiction of the diaphragm, one needs to use other sacrificial material that can be deposited a lot thicker.
The most promising one is the use of the photoresists, i.e. SU-8. The SU-8 can be readily spun coat on the substrate from a few microns to a hundred microns thick.
This can give us a much deeper cavity depth if diaphragm layer can be properly deposited on the top to enclose the cavity. However, the high deposition temperature of the LPCVD polysilicon or nitride process for diaphragm formation has precluded the use of SU-8 as sacrificial layer. In fact, this difficulty can be readily overcome by reversing the fabrication process of the pressure sensor. That is, one can first deposit the diaphragm, and then the piezoresistive layer and the metal lines on the silicon substrate, as shown in Fig. 1a.
The piezoresistive layer is actually the polysilicon layer implanted with certain concentration of boron or phosphorous. In the current process, one would like to select the concentration of boron such that the resistivity variation with the temperature in the polysilicon layer can be minimized. From the experimental data plotted in Shen and Gau [11], for boron concentration greater than 1020atms/cm3 the resistivity variation with temperature can be negligible small. Therefore, this concentration of boron is adopted in the implantation process for the polysilicon layer. After implantation, the polysilicon is annealed for 30mins at 1100oC. Further explanation to release the stress or deflection will be given in section 3.2. It is then followed by spin coat the SU-8 layer at desired thickness. Thus, the formation of SU-8 layer will not go through a high temperature process. Another advantage is that the SU-8 layer can be readily patterned to form cavities by lithography. Finally,
a PMMA plate can be bonded, using epoxy resin, with the patterned SU-8 layer on the top to enclose the cavities, as shown in Fig. 1b. Once the silicon substrate is completely removed by wet etch, a successful pressure sensors can be readily achieved, as shown in Fig. 1c. The completed result of the micro pressure sensor is shown in Fig. 2. In fact, arrays of pressure sensors can also be made by the same process, except that a number of cavities should be formed in the SU-8 layer by lithography.
3. Discussion on fabrication techniques 3.1 SU-8 Lithography Process
Among the thin film materials available, the SU-8 appears to be the most suitable one since it can be readily spun on the substrate at desired thickness, and patterned into required shape of channel by photolithography. Therefore, different sizes and shapes of cavity or arrays of cavities combined with a channel system can be readily made. To spin coat a thick layer of SU-8, the substrate has to be cleaned first by acetone and then rinsed by IPA solution more than once. HMDS is also needed to make the surface to be hydrophobic. If the surface of the substrate is not clean enough or hydrophobic to increase the adhesion force of the SU-8 layer, the SU-8 layer can readily shrink after soft bake due to the serious thermal expansion of the SU-8 layer.
In fact, uniformity over the SU-8 layer cannot be readily achieved after substrate cutting. This is attributed to the formation of the protruded bead on the wafer edge, which is induced by the large surface tension of SU-8 resist during the spin coat process. The protruded bead on the edge of the 4”or 6”substrate can be eliminated by the reflow of the SU-8 layer during soft bake, or it can be removed by an edge beads removing process on an automatic photoresist treatment system. However, for protruded bead on the edge of a much smaller epoxy-glass chip, it can be neither eliminated by the reflow of the SU-8 layer during soft bake due to poor thermal conductivity of a epoxy-glass chip, nor by the edge beads removing process. Therefore, a compromise
is made. First, a portion of edge bead on the substrate is removed by using a silicone scraper after spin coat of SU-8. Then, the SU-8 layer can be setting on the substrate for 20-30 minutes for reflow before soft bake, since the viscosity of the SU-8 layer is still very low and the SU-8 layer can be easily reflowed by gravity.
In order to avoid further shrinking and release the residual stress of a thick SU-8 layer after soft bake, the mode of a ramping process in the soft bake also plays an important role. A two-step ramping mode was used in the present soft bake process. First, the temperature was ramped to 65oC from the room temperature with a 2oC/min ramp rate, and then held there for 15minutes.
Next, the temperature was ramped to 95oC with the same ramp rate and then held there for 30 minutes.
Finally, the temperature was ramped down to the room temperature slowly. A exposure dose with 3 mJ/(cm2• µm-1) is selected for the exposure process. After exposure of the SU-8 layer, a post exposure bake is required to make the edge between the exposed and unexposed region more sharp and clear. Since the SU-8 material has a very high thermal expansion coefficient (52 ppm/K), to bake it at a high temperature may significantly distort and deform the shape of the cavity after development. To demonstrate the seriousness of this problem, an array of cavities made for an array of pressure sensors, each has an area of area is 200 µm×200 µm, that connects to a long channel with very narrow size tunnels, i.e. 1 µm in width, was made as shown in Fig. 3. Fig. 3(a) shows that to bake the SU-8 cavities and channel system immediately at 95oC, not only the shape has deformed, but also the size of the cavity or channel has been significantly reduced. The original width of the channel is 500µm. But after the post bake, the width of the channel has shrunk to approximately 400 µm. The very dense rings of stress appear along the surrounding of the small cavities. The small tunnel between the cavity and the channel at this baking condition has been completely blocked. All the serious deformation and the very dense rings of stress
are caused by the high thermal expansion coefficient of the SU-8 layer. However, the number of stress rings is significantly reduced, as shown in Fig. 3(b), as the post bake temperature is reduced to 65oC. However, the shape of the channel and the cavities is still significantly deformed due to uneven expansion in different parts of the structure. In addition, the small tunnel is still blocked by the expansion. Further reduction in the baking temperature may take a very long time to evaporate the solvent. Another way of reducing the uneven expansion is to slowly increase the bake temperature until it reaches the desired level. With a 2oC/min ramping rate in the oven until it reaches 65oC and keeping there for half an hour, the post bake result of the channel and the cavities after development is shown in Fig. 3(c). The stress rings around the cavities are all gone. No any deformation is found. Even the shape and the size of the small tunnel between the cavity and the channel maintain the original structure and do not change at all. After developing and rinse, the SU-8 layer finally undergoes a high dose exposure hard bake. The exposure dose used was 4000mJ/cm2. From a cross section view of a cavity made by the present SU-8 lithography process, the cavity sidewall is very near vertical as designed, as indicated in Liu [12].
3.2 Release of residual stress in the polysilicon diaphragm
In order to avoid cracks or wrinkles of the deposited polysilicon diaphragm that may completely damage the pressure sensor due to the release of large residual stress in the polysilicon film after complete removal of the silicon substrate by TMAH wet etch process, the residual stress in the polysilicon layer has to be tested and measured before the pressure sensor system is constructed. First, a 0.3 µm thick LPCVD TEOS-based oxide layer is deposited on the both sides of the wafer as an etch stop in the following TMAH wet etch process. The bending of the wafer due to the deposited TEOS layer on both sides is measured with a film stress analyzer (TENCOR FLX-2320, USA) in
NDL. Then, a 1.8 µm thick LPCVD polysilicon layer used as a tested pressure diaphragm is deposited on the both sides of the wafer, with a SiH4 flow rate of 105 sccm, a pressure of 160 mtorr and at 620oC. To measure the residual stress of a single polysilicon layer, the polysilicon layer on the backside has to be removed by the TMAH wet etch process. After standard furnace annealing with a N2 flow rate of 7000 sccm at 950-1100oC for 30-120 minutes to release the residual stress of the polysilicon layer, the bending of the wafer will be measured with the film stress analyzer.
Therefore, the residual stress in the polysilicon layer can be readily calculated from the bending of the substrate.
The initial residual stress in the 1.8µm thick polysilicon layer without annealing is compressive and is 360 MPa.
Comparison of the residual stress with the published results [12] in the polysilicon layer for annealing at various temperatures and time periods is shown in Fig. 4.
However, in Zhang’ data [13], a 0.5 µm thick LPCVD polysilicon layer is deposited at 300 mtorr and 620oC with a SiH4 flow rate of 70 sccm. It is found that the residual stress in the present experiments is significantly lower than the published data due to the different deposition conditions of the LPCVD polysilicon layer, especially for the SiH4 flow rate and the deposition pressure. Therefore, such a low compressive, residual stress of 25 MPa in the polysilicon diaphragm, due to annealing at 1100oC for greater than 30 minutes, will not crack or wrinkle the substrate surface after one side of the polysilicon layers is removed.
For a further test, measurements of the deflection for the suspended polysilicon diaphragm after release is made and the results are shown in Fig. 5. It is found that the deflection of the suspended polysilicon diaphragm after release is significantly reduced when the thickness of the polysilicon layer is greater than 2.4 µm. To design a pressure sensor with diaphragm thickness of 2.4 µm and area of 200 µm×200µm for the present sensor, the deflection of the diaphragm will be very small and is estimated approximately 200 nm. Further explanation
for this size of the pressure sensor is given in the next section
4. Design for polysilicon diaphragm and circuitry
The size of the polysilicon diaphragm used in the pressure sensor is determined by the pressure range to be measured, the maximum strain that the diaphragm can sustain and the sensitivity required. The deformation for a square shell or plate with four edge clamped under a uniform normal pressure force can be calculated by a computer software ANSYS. For a pressure range from 1 to 5 atms, the maximum deflection occurred at the center of the diaphragm versus the pressure force is presented in Fig. 6 for different thickness of diaphragm. The maximum strain the polysilicon layer can sustain is found 0.71% [4].
Therefore, the rectangle presented in Fig. 6 represents the safety region for the size of the diaphragm. For a better sensitivity, i.e. more deflection, the size of 200µm
× 200µm wide at a thickness of approximately 2.4 µm is selected.
There are a total of 4 sensors designed to locate on the edge of the diaphragm to give the maximum sense of the stress variation due to the deformation of the diaphragm. The piezoresistive sensor can be arranged with a Wheatstone bridge circuitry to amplify the output signals and obtain a high sensitivity. Usually, the output signals of the circuitry are very small (from a few to few hundred microvoltage) and possibly with some noise.
Therefore, a data acquisition combined with signal processing system has to be used to obtain accurate electric signals from the pressure sensor. Although the Wheatstone bridge circuitry system has a higher sensitivity, the data acquisition process is more complicated and time consuming due to the complicated signal processing process. Therefore, a series connection with all the four sensors is designed and used to give the maximum sensitivity, but reduce the processing time of signals and give more rapid signal
response for the pressure variation in the cavity.
5. Calibration
The completed pressure sensor for a series connection circuitry is wire bonded to acquire electric signals from the sensor for calibration. A PMMA plate with a large cavity is glued on the chip sensor. The cavity is connected to a valve that is used to connect the gas flow with the pressure at desired levels from a high-pressure gas system. The variation of the total resistance of the four sensors is measured at selected pressure levels and plotted as shown in Fig. 7. The very linear relationship between the pressure applied and the total resistance measured indicate the successfulness of the sensors. The large variation of the resistance with the pressure indicates the very high resolution and sensitivity of this sensor system.
6. Conclusions
A novel fabrication process for a or arrays of pressure sensors has been presented. The current fabrication process is almost the reverse of the usual surface micromachining process for the pressure sensor.
This allows the use of the SU-8 layer to make a much deeper size of cavity and allow much wider range of pressure measurement. The fabrication process has completely absence of stiction problem. In addition, this kind of sensors can be readily integrated with a complicated micro system. However, care should be taken for the cavity formation in the SU-8 layer and the release of the residual stress in the diaphragm.
Calibration for this sensor has also been presented.
Acknowledgment
This research was sponsored by both the National Science Council of Taiwan under contract no. NSC 91-2212-E-006-108 and the National Nano Device Laboratories (NDL) under contract no. NDL 91S-C-062.
References
1. S. M. Sze, Semiconductor Sensors, John Wiley and
Sons, Inc., New York, 1994.
2. O. N. Tufte, P. W. Chapman and D. Long, Journal of Applied Physics, Vol. 33, No. 11, pp.3322-3327, 1962.
3. S. K. Clark and K. D. Wise, “Pressure sensitivity in anisotropically etched thin-diaphragm pressure sensors,” IEEE Trans. Electron Devices, Vol.
ED-26, pp.1887-1896, 1979.
4. H. Guckel, “Surface micromachined pressure transducer,” Sens. Actuators, Vol. A28, pp.133-146, 1991.
5. S. Sugiyama, K. Shimaoka, and O. Tabata,
“Surface micromachined micro-diaphragm pressure sensors,” in Proc. 6th Int. Conf. Solid-State Sensors and Actuators (Transducer’91), 1991, pp.188-191.
6. T. A. Core, W. K. Tsang and S. J. Sherman,
“Fabrication Technology for an Integrated Surface Micromachined Sensor,” Solid State Technology, October, pp.39-47, 1993.
7. C. J. Kim, J. Y. Kim and B. Sridharan,
“Comparative evaluation of drying techniques for surface micromachining,” Sens. Actuators, Vol.
64A, pp.17-26, 1998.
8. K. Komvopoulos, “Surface engineering and microtribology for microelectromechanical systems,” Wear, Vol. 200, Issue 1-2, Dec., pp.305-327, 1996.
9. R. Legtenberg, J. Elders, M. Elwenspoek, “Stiction of surface microstructure after rinsing and drying:
model and investigation of adhesion mechanisms,”
Proc. 7th Int. Conf. Solid-State Sensors and Actuators (Transducer’93), Yogohama, Japan, 1993, pp.198-201.
10. Marc J. Madou, Fundamentals of Microfabrication, 2nd ed., CRC Press, 2002, pp.276-277.
11. C. H. Shen and C. Gau, “Thermal Chip Fabrication with Arrays of Sensors and Heaters for Micro-Scale Impingement Cooling Heat Transfer Analysis and Measurements,” Biosensors and Bioelectronics, 2004. (in press)
12. C. W. Liu, Fabrication Development for Micro Channel System by MEMS Technology with Measurements of the Inside Thermal Transport Process, PhD thesis, National Cheng Kung University, Tainan, Taiwan, ROC, 2004.
13. X. Zhang, T. Y. Zhang, M. Wong and Y. Zohar,
“Residual-stress relaxation in polysilicon thin film by high-temperature rapid thermal annealing,”
Sens. Actuators, Vol. A64, pp.109-115, 1998.
無粘黏問題的新式微壓力感測器製程
劉建惟
1高騏
2洪興林
31
國家奈米元件實驗室
2
國立成功大學航空太空工程學系(所)
3
崑山科技大學機械工程系
國科會計畫編號：NSC-91-2212-E-006-108
摘要
本研究主要目的乃提出一套幾乎與標準微型面 加工製作步驟相反的新式微壓力感測器製程技術。
感壓凹穴以SU-8 厚膜結構取代，因此可增大壓力感
測器的感壓範圍而不會有輸出訊號達飽和的問題。
此外，壓力感測器之感壓薄膜成型過程不須經過去離 子水清洗，感壓薄膜不會有粘黏現象發生，所以微壓 力感測器陣列將可容易地被整合於微型系統內，其他 詳細的製程結果將會在本文中作更深入的討論。
關鍵字：微型面加工，微壓力感測器，SU-8
Figure 1: Fabrication process of the pressure sensor: (1) silicon substrate, (2) polysilicon diaphragm, (3) piezoresistive sensor, (4) metal lead, (5) Pyrex glass, (6) active cavity and 7) SU-8 layer.
Figure 2: Photograph of the pressure sensor.
Figure 3: Photographs and the profile meter results of the SU-8 channel structure with different PEB steps at (a) 95oC and (b) 65oC and (c) with a 2oC/min ramping rate in the oven until it reaches 65oC.
Figure 4: Comparison of the residual stress with the published results for the polysilicon layer at various annealing temperatures and time periods
Figure 5: Deflection of a suspended polysilicon diaphragm after annealing at 1100oC for 1 hour. (a) The cross section view of a suspended polysilicon diaphragm, and (b) deflection of the pressure diaphragm at different length and thickness
(a)
(b)
(c)
(a)
(b)
(c)
(a)
(b)
(c)
150umx150um
Pressure(atm)
0 1 2 3 4 5 6
Strain,max(%)
0.0 0.5 1.0 1.5 2.0 2.5
1um 2um 3um
200umx200um
Pressure(atm)
0 1 2 3 4 5 6
Strain,max(%)
0.0 1.0 2.0 3.0 4.0 5.0
1um 2um 3um
250umx250um
Pressure(atm)
0 1 2 3 4 5 6
Strain,max(%)
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0
1um 2um 3um
Figure 6: The maximum strain variation with pressure for different sizes and thicknesses of polysilicon pressure diaphragm.
Figure 7: Characteristic curve of the present micro pressure sensor.
