Characterization of A Polymer-Based MEMS Packaging ... - IEEE Xplore
Characterization of A Polymer-Based MEMS Packaging Technique Pejman Monajemi', Paul J. Joseph2, Paul A. Kohl2, Farrokh Ayazi' 'School of Electrical and Computer Engineering 2School of Chemical and Biomolecular Engineering
Georgia Institute of Technology, Atlanta, GA 30332 Phone: 404-894-9496; e-mail: ayazigece.gatech.edu
Abstract
This paper presents characterization of a low-cost polymer-based technique for wafer-level packaging of microelectromechanical systems (MEMS). The packaging process does not impose any size limitation to the device and can be applied to a wide variety of MEMS devices regardless of substrate type. Our technique utilizes thermal decomposition of a sacrificial polymer through a polymer overcoat cap, followed by metal deposition to create hermiticity. The method has been applied to surface and bulk micromachined silicon structures including resonators, tunable capacitors and accelerometers. Mechanical and electrical characterization of the packaged devices is reported to be very close to the corresponding values before packaging. Keywords: MEMS, HARPSS, thermal decomposition, sacrificial polymer, polymer overcoat. Introduction MEMS packaging adds complexity and significant cost to the microsystem manufacturing. In many applications, hermiticity is required to isolate the microelectromechanical systems (MEMS) from the environment [1]. Several techniques have been reported for wafer-level packaging of MEMS devices, including a variety of wafer bonding [2-6] and sacrificial-film-based [7, 8] methods. Most of the reported techniques are either costly, have size limits, require high-temperature, or are device-specific. This work reports on characterization of a low-cost waferlevel packaging process that is applicable to both surface and bulk-micromachined structures, after their fabrication is completed. Our technique is based on thermal decomposition of a polymeric sacrificial material, Unity 2000 (Promerus, LLC) [9, 10], through a spin-coated solid polymer overcoat, Avatrel (Promerus, LLC). The sacrificial material undergoes thermolytic degradation in an oven to create an air cavity on top of the active parts of the device. Thermal decomposition of sacrificial polymers happens at low temperatures below 250°C. This packaging scheme has certain advantages over current packaging methods. First, thermal heating is stictionfree and does not require etching holes in the overcoat. This simplifies the process. Second, the cap formation is done by regular photolithography. This eliminates the errors associated with aligning a cap with cavity to the MEMS wafer. Also, the package is resizable. In other words, thickness of the sacrificial polymer and the overcoat can be arbitrarily tailored based on the device geometry, size and application. Cavities as thin as 1 im can be created by
lithography. The sacrificial polymer can be dispensed to create cavities as thick as few millimeters (which is impossible by wafer-to-wafer bonding or by deposition of sacrificial inorganic films). The polymer overcoat can be spin coated to create capsules as thin as few micrometers or as thick as few 100 micrometers. Moreover, thermal release of MEMS and package can be done simultaneously by using the sacrificial polymer in the device fabrication and packaging. Thermally depolymerizable polycarbonates have been used to create microfluidic channels on silicon [11]. Patterning these polymers requires electron beam lithography and the depolymerization starts above 300°C. Also creation of air channels has been reported in [12], where fluorocarbons deposited by chemical vapor deposition (CVD) are thermally depolymerized at about 300°C. In our method, the sacrificial polymer is created by spin coating or dispensing and can be photodefined using regular optical exposure tools. The polymer encapsulation is cost-effective and requires common processing tools including spinners, mask aligners, and ovens. Hermiticity is obtained by sputtering, evaporation or electroplating of a metal having a thermal coefficient of expansion (CTE) close to that of substrate. Metal-Organic Package Description Figure l.a shows the schematics of a metal-organic packaged beam resonator in silicon-on-insulator (SOI) substrate. Figure l.b is the electron micrograph of a packaged beam resonator. The cavity is only 80imx 120imx 15pim.
(a)
(b
........
(b) . . .
Fig. 1. a) Schematics of a packaged SOI beam resonator (15,um thick), b) Electron microscope view of a packaged resonator [10]
1-4244-0261-1/06/$20.oo A©2006 IEEE.
139
In Figure l.a, the metal-organic square overcoat has a width, L, polymer thickness, tpolymer, and metal thickness, tMetal. The air cavity is created by thickness of the sacrificial polymer, tCaviy. The maximum cap bending due to the pressure difference between inside and outside the cavity, Ap, should be less than tCaviy.. The package design should predict the proper thickness of the sacrificial polymer to completely cover the MEMS holes. Also, the minimum overcoat thickness, required to tolerate the pressure difference, Ap should be calculated. For the rectangular cap of Figure l.a., the package bending, Z, is formulated in (1) [13]. Here tc, E and, v are the cap effective thickness, Young's modulus, and Poisson ratio, respectively: Ap-
2 ° t Z+
2L2
2
4E
t3CZ +
4C 3(1- v )L
(
2
+
1674E t3Z+
3(l1_V2 )r4
(7-v)tCz3 3(l1_V)r4
(2)
The three terms in equations (1, 2) represent thin film residual stress, plate bending due to flexural rigidity, and inplane stretching, respectively. For thin-films, the second term is usually the largest term. For the square cap with a cavity height, tCavily, tc should be designed so that the maximum plate bending (Zmax) is less than tcavity:
3(1
V2 )L4
(a) Insulator
2)
128(l1-v)L4CtLZ3
For a circular cap with radius, r, equation (1) can be rewritten as (2):
Ap4o-t r
and expedite the evacuation. The metal film is deposited as in Figure 2.d. Finally to get access to the pads, the metal film and consequently the isolation film are patterned and etched (Figure 2.e).
(b) Polymer overcoat `
(c) Metal overcoat >
Ap3
Zmax - 2 Et (3) For the metal-organic package, the Young's modulus of the metal film is orders of magnitude larger than that of the polymer. For gold, E=79GPa and vo0.44. Using a 1I m thick gold overcoat in equation (3) and the mentioned values for E, v, and Ap=latm yields Zmax =1.57[tm. Therefore a sacrificial polymer with minimum thickness of 2pm is required to ensure that the MEMS package will survive the pressure difference without collapsing. The polymer overcoat can be selected among variety of photo-definable polymers with high glass transition temperature.
(d)
(e) Fig.2. Polymer-based packaging process.
Table 1 lists the processing steps in detail for Unity and Avatrel. For a Unity formulated with photo-acid generator MEMS Packaging Process Figure 2 describes the packaging steps. In Figure 2.a., the (PAG), the development involves PAG-assisted sacrificial material is first spin-coated or dispensed to cover decomposition of the exposed area at around 110°C on a the active MEMS components (such as beam in a beam hotplate followed by a short dip in isopropanol alcohol. TABLE I resonator). In Figure 2.b, a layer of insulator such as silicon PROCESSING PARAMETERS POLYMER dioxide is deposited prior to polymer overcoat formation. This Parameter Unity (9pm) Avatrel (10im) layer increases the rigidity of the package [14] and creates Spin coating speed 3800rpm 4200rpm isolation between MEMS bond pads and metal package. The Soft bake l0min/110°C l0min/1000C polymer overcoat is spin coated and patterned to cover the MEMS except the bond pads. The sacrificial material is then IJ/cm2 (240nm) 0.2J/cm2 (365nm) Exposure Post-exposure bake l0min/1000C decomposed at 200-250°C. This is the highest temperature step in this process. The by-products of thermal Development Spray 1min -1 0min/1 10°C decomposition (oxygen, hydrogen, carbon monoxide, etc.) Decomposition 1°C/min to 1600C, hold for 1 hour permeate through the polymer overcoat (Figure 2.c). The 1°C/min to 2500C, hold for 2 hours organic package is left in a metal deposition chamber such as an electron beam evaporator or a DC sputterer. The package Dispensing of sacrificial polymer can be done using a syringe should be left for enough time prior to deposition to create to create a dome-shaped capsule. This is suitable for bulk vacuum inside the cavity. Heating can increase gas diffusivity 140 1-4244-0261-1/06/$20.00 A©2006 IEEE. -
micromachined structures with wide and deep cavities. This method is especially suitable for MEMS devices with delicate components that may break during sacrificial material spin coating. The examples include MEMS accelerometers and gyroscopes.
(a)
Characterization of the Polymer Package First, the electrical, mechanical, thermal, and optical parameters of Avatrel were measured, as listed in table 2. The small dielectric constant is suitable for low-loss RF MEMS packaging. The glass transition temperature of Avatrel (-3000C) is much higher than that of Unity (-1OOC). TABLE II MEASURED PROPERTIES OF AVATREL [15]
|Equipment
Electrical
Permitivity at 1GHz
hp network analyzer
2.57
Mechanical
Young's modulus (at 250C) |2GPa Poisson's ratio 0.3
Hysitron nanoindenter Hysitron nanoindenter
Thermal Glass transition temperature
3000C
KLA Tencor profiler
1.5-1.6
Woollam ellipsometer
Optical Index of refraction
(b)
The reduced-thickness of Avatrel and its stress after curing was measured using a contact profiler, as shown in Figure 3.a A 10ptm thick Avatrel overcoat can have a tensile stress of 4.5GPa after curing at 200°C (a typical decomposition temperature used for Unity).
Fig.4. Scanning electron microscope view of a resonator: a) before and b) after packaging by photo-defining Unity.
The titanium/gold was deposited in a DC sputterer in a chamber pressure of about 6mTorr. Etching of gold was performed in a Transene gold etcher. Finally the oxide isolation film was removed in buffered oxide etch (BOE). Figure 5 shows the microscope view of the cavity after decomposition and prior to metalization. Avatrel is transparent to visible light, it can be seen that the cavity is free of sacrificial polymer residues.
6
0
2
4
6
8
10 12 14 16 18 20
Avatrel Thickness, pim Fig.3. Avatrel stress versus thickness after curing at 2000C.
The packaging technique was applied to package a variety of MEMS devices. This includes (but not limited to) inertial sensors, resonators, and RF passives. Characterization of the Packaged Resonators Figure 4 shows the packaging of a 2.6MHz beam resonator. Figure 3.a. shows the viewgraph of the resonator with 1 ptm transduction gap before packaging. Figure 3 .b is the same device after packaging. The thickness of sacrificial polymer should be at least 2pm to properly cover the air gap. Thickness of Unity, Avatrel, and gold are 9ptm, 10ptm, and 2.5ptm, respectively. The polymer processing steps are the same as listed in Table 1.
Besides photo-defining and dispensing, Unity can be created by reactive ion etching (RIE) or molding. RIE of nonphotodefinable Unity requires using titanium or oxide to be used as the mask. A 50sccm, 200W oxygen plasma at 5OmTorr pressure, can etch Unity with a rate of 0.2tm/min. Besides RIE, this kind of Unity can be also patterned by molding. Figure 6 shows a package created by molding 10pQm Unity. In the molding method, the packaging sequence starts with deposition of an oxide mold using plasma enhanced CVD (PECVD), to overbridge the trenches. This is followed by patterning the oxide with a negative resist, and removing of oxide in RIE. Then Unity is spin coated and cured on top of the oxide. Finally oxygen plasma removes the Unity on top of the mold and leaves some Unity on the channel. The oxide mold is removed in BOE. The metal-organic package in this
1-4244-0261-1/06/$20.oo A©2006 IEEE.
141
device consists of 2ptm/1Oim chromium/Avatrel layers. Chromium has a close CTE to silicon.
The reduction in Q is due to presence of small residues on the beam surface that increase the surface loss. As shown in Figure 8, Q after packaging (Qp) is always lower than unpackaged Q (Qu). Therefore, the vibration amplitude of the packaged resonator (QpAo) and resonance frequency (fp) are always lower than the unpackaged amplitude (QuAo) and resonance frequency (fu), respectively. Several batches of packaged resonators were examined and it was observed that in average, Q was degraded by 8% and the frequency change (A4) was within 0.2%.
Af
A' Fig.6. Scanning electron microscope view of a packaged resonator by molding Unity.
A packaged resonator was tested at wafer-level inside a vacuum probe station. A DC polarization voltage in the range of 70-80V was applied to substrate, while the electrodes were directly connected to the network analyzer. Figure 7(a) and (b) show the frequency response of the resonator in vacuum before and after packaging. The high Q factor of -4500 did not change significantly for this device, proving that thermal decomposition does not affect the performance of the device.
fp fu
Fig.8. Conceptual demonstration of degradation in Q and output power after packaging.
In some batches, a lower decomposition temperature was used (150°C) and the packaging did not result working samples. After breaking the polymer cap, the Unity residues were analyzed using a VEECO atomic force microscope (AFM) in tapping mode configuration. As Figure 9 shows, the thickness of residues was measured from the maximum height difference and it was about 200A°. Heating the packaged resonators up to higher temperature (-250°C) resolved the residue issue.
Fig.9. Residue analysis using a VEECO AFM in tapping mode.
Characterization of the Packaged HARPSS Varactors A variable capacitor (varactor) was created using the high aspect ratio single crystal silicon and polysilicon (HARPSS) technology [16]. Figure 10 shows the HARPSS varactor before and after packaging. The varactor is made by using a 1-2pm gap between polysilicon and silicon electrodes and the series resistance has been reduced by evaporation of gold to increase the electrical Q factor. The 60ptm thick, 2pF HARPSS capacitor (lmmx1.5mm area) shows a tuning range of 2:1 by applying a voltage of about 2V [16]. The measured electrical Q is about 50 at 1GHz. The metal-organic package does not change the DC performance of the device, although it can change the RF performance. Packaging was done by
Fig.7. Q versus frequency for a beam resonator: a) before and b) patterning 15ptm thick unity, and 20pm thick Avatrel. The after packaging using an input signal level of -5dBm. hermiticity was added to the polymer package by evaporation 142 1-4244-0261-1/06/$20.oo A©2006 IEEE.
of 1 [tm gold overcoat, eliminating the extra mask needed to pattern the metal (step 'e' in Figure 2). Moreover, evaporation of metal overcoat instead of deposition and etching metal, eliminates the need for deposition and etching the extra insulator layer before Avatrel patterning (step 'b' in Figure 2). The only limitation is that the thickness of evaporated metal should be less than thickness of the polymer overcoat.
Characterization of the Packaged HARPSS Accelerometer HARPSS technology was used to fabricate micro-gravity silicon accelerometer [18]. Fig. 12.a shows a 50pim thick, 1.8mmxl.5mm X-axis accelerometer. Fig. 12.b shows the
accelerometer, packaged by dispensing Unity and overcoating a 120im thick Avatrel overcoat. The static sensitivity of the packaged accelerometer is measured to be about 0.27pF/g [10].
Fig. 10. Scanning electron microscope view of a HARPSS varactor: a) before and b) after packaging by photo-defining Unity.
The gold-Avatrel package was charactFerized using an Agilent 8517 vector network analyzer. The Ipackage adds an insertion loss of about 1.4dB at 1 GHz and 1..5dB at 5GHz to the MEMS device, as shown in Figure 11. The package is only 20% larger than the device. The horizon[tal feedthroughs in this packaging method do not add a signifrican size to me package. In the wafer bonding methods, tihe feedthroughs form a large component of the final MEMS pzackage size [17].
IlI Fig.12.Scanning electron microscope view of a HARPSS accelerometer: a) nonpackaged and b) packaged by dispensing Unity [10].
Conclusions A low-cost low-temperature packaging technique for wafer-level encapsulation of MEMS devices is presented. The packaging process does not involve wafer bonding, high temperature deposition of LPCVD or epitaxial films, or wet etching of sacrificial thin-films. It utilizes decomposition of a sacrificial polymer through a photo-definable polymer overcoat. Cavities as small as 0.00015mm3 (for resonators) m and as large as 1mm3 (for inertial sensors) were created on U,0 MEMS devices using 10-15m thick gold-Avatrel or re Packaging -J chromium-Avatrel package. The packaging has been Packaging 0 successfully applied to encapsulate surface/bulk oU, micromachined structures including beam resonators on SOI substrate with reduced-size feedthrough, HARPSS varactors and HARPSS accelerometers. For the 2.6 MHz beam resonators, a quality factor of about 4500 was measured before and after packaging. For the HARPSS varactor, the gold-Avatrel package adds about 1.5dB of loss at 5GHz as a Frequency, GHz result of capacitive loading of the package. For the Fig. 11. Measured insertion loss of the HARP' SS varactor before microgravity accelerometer, the static sensitivity after and after low-loss packaging polymer packaging is about 0.27pF/g. Stress measurement of 143 1-4244-0261-1/06/$20.oo A©2006 IEEE.
the organic cap was performed for different thicknesses and the different parameters of the polymer overcoat was extracted. Acknowledgments
The authors would like to thank STMicrolectronics for supporting this project and Promerus LLC for supplying the materials, especially especially Jeff Krotine and Ed Elce for the valuable discussions. References [1] Najafi K, "Micropackaging Technologies for Integrated Microsystems: Applications to MEMS and MOEMS," SPIE MOEMSIMEMS (San Jose, CA) pp.1-12, 2003. [2] Lee B, Seok S, Chun K, "A Study on Wafer-Level Vacuum Packaging for MEMS Devices," J. Micromech Microengineering, Vol. 13, pp. 663-669, 2002. [3] Mitchel J, Lahiji G R and Najafi K, "Encapsulation of Sensors in a Wafer-Level Package using a Gold-Silicon Eutectic," JIEEE Transducers (Seoul, Korea) pp.663-669, 2005. [4] Pan C T, Yang H, Shen S C, Chou M C and Chou H P, "A Low-Temperature Wafer Bonding Technique using Patternable Materials," J. Micromech. Microeng. Vol.12, 611-615, 2002. [5] Sparks D, Massoud-Ansari S and Najafi N, "Long-Term Evaluation of Hermetically Glass Frit Sealed Silicon to Pyrex Wafers with Feedthroughs," J. Micromech. Microeng. Vol.15, pp.1560-1564, 2005. [6] Lin L, "MEMS Post-Packaging by Localized Heating and Bonding," IEEE Trans. Adv. Packaging Vol.23, pp.608616, 2004. [7] Candler R N, Park W T, Li H, Yama G, Partridge A, Lutz M and Kenny T W, "Single Wafer Encapsulation of MEMS Devices," IEEE Trans. Adv. Pack. Vol.26, No.3, pp.227-232, 2003. [8] Stark B H and Najafi K, "A Low-Temperature Thin- Film Electroplated Metal Vacuum Package," IEEE J. Microelectromech. Syst. Vol.13, No.2, pp.147-157, 2004. [9] Joseph P J, Reed H A, Hongshi Z, Rhodes L F, Henderson CL, Allen S A.B and Kohl PA, "Air-Channel Fabrication
[13] Senturia S D, "Microsystem Design," Kluwer Academic Publishers., (New York, 2001). [14] Joseph P J, Reed H A, Allen, S A.B.; Kohl, P A, "Improved Fabrication of Micro Air- Channels by Incorporation of a Structural Barrier," J. Micromech. Microeng. Vol.15, pp. 35-42, 2005. [15] Patel K S, Kohl PA and Allen S A.B "Three-Dimensional Dielectric Characterization of Polymer Films," J. Applied polymer Science Vol.80, pp.2328-2334, 2001. [16] Monajemi P and Ayazi F, "A High-Q Low-Voltage HARPSS Tunable Capacitor," IEEE International Microwave Symposium (Long Beach, CA) pp.749-752, 2005. [17] Muldavin J, Bozler C, Rabe S, Keast C, "Wide-Band Low-Loss MEMS Packaging Technology," IEEE International Microwave Symposium (Long Beach, CA) pp.765-768, 2005. [18] Monajemi P and Ayazi F, "Analytical Design and Implementation of a Micro-Gravity HARPSS Accelerometer," Proc. Sensors Journal, Vol.6, No. 1, pp.39-47, 2006.
for Microelectromechanical Systems via Sacrificial J. Photosensitive IEEE Polycarbonates," Microelectromech. Syst. Vol.12, No.2, pp. 147-157, 2003. [10] Monajemi P, Jospeh P, Kohl P, Ayazi F "A Low-Cost Wafer-Level MEMS Packaging Technology," Proc. IEEE MEMS, Miami, FL, 2005, pp.634-637. [11] Hamett C, Satyalakshmi K, Coates G and Craighead H, "Direct Electron-Beam Patterning of Surface Coatings and Sacrificial Layers for Micro-Total Analysis Systems," J. ofPhotopolymer Sci. Tech., Vol. 15, pp. 493-496, 2002. [12] Chan K, Casserly T, Loo L, Gleason K, "Air Dielectric Fabrication via CVD Sacrificial Materials," Proc. MRS Symp. Vol. 766 (Materials, Technology, and Reliabilityfor Advanced Interconnects and Low-k Dielectrics), San Francisco, CA, April 2003. 144
1-4244-0261-1/06/$20.oo A©2006 IEEE.
Georgia Institute of Technology, Atlanta, GA 30332 Phone: 404-894-9496; e-mail: ayazigece.gatech.edu
Abstract
This paper presents characterization of a low-cost polymer-based technique for wafer-level packaging of microelectromechanical systems (MEMS). The packaging process does not impose any size limitation to the device and can be applied to a wide variety of MEMS devices regardless of substrate type. Our technique utilizes thermal decomposition of a sacrificial polymer through a polymer overcoat cap, followed by metal deposition to create hermiticity. The method has been applied to surface and bulk micromachined silicon structures including resonators, tunable capacitors and accelerometers. Mechanical and electrical characterization of the packaged devices is reported to be very close to the corresponding values before packaging. Keywords: MEMS, HARPSS, thermal decomposition, sacrificial polymer, polymer overcoat. Introduction MEMS packaging adds complexity and significant cost to the microsystem manufacturing. In many applications, hermiticity is required to isolate the microelectromechanical systems (MEMS) from the environment [1]. Several techniques have been reported for wafer-level packaging of MEMS devices, including a variety of wafer bonding [2-6] and sacrificial-film-based [7, 8] methods. Most of the reported techniques are either costly, have size limits, require high-temperature, or are device-specific. This work reports on characterization of a low-cost waferlevel packaging process that is applicable to both surface and bulk-micromachined structures, after their fabrication is completed. Our technique is based on thermal decomposition of a polymeric sacrificial material, Unity 2000 (Promerus, LLC) [9, 10], through a spin-coated solid polymer overcoat, Avatrel (Promerus, LLC). The sacrificial material undergoes thermolytic degradation in an oven to create an air cavity on top of the active parts of the device. Thermal decomposition of sacrificial polymers happens at low temperatures below 250°C. This packaging scheme has certain advantages over current packaging methods. First, thermal heating is stictionfree and does not require etching holes in the overcoat. This simplifies the process. Second, the cap formation is done by regular photolithography. This eliminates the errors associated with aligning a cap with cavity to the MEMS wafer. Also, the package is resizable. In other words, thickness of the sacrificial polymer and the overcoat can be arbitrarily tailored based on the device geometry, size and application. Cavities as thin as 1 im can be created by
lithography. The sacrificial polymer can be dispensed to create cavities as thick as few millimeters (which is impossible by wafer-to-wafer bonding or by deposition of sacrificial inorganic films). The polymer overcoat can be spin coated to create capsules as thin as few micrometers or as thick as few 100 micrometers. Moreover, thermal release of MEMS and package can be done simultaneously by using the sacrificial polymer in the device fabrication and packaging. Thermally depolymerizable polycarbonates have been used to create microfluidic channels on silicon [11]. Patterning these polymers requires electron beam lithography and the depolymerization starts above 300°C. Also creation of air channels has been reported in [12], where fluorocarbons deposited by chemical vapor deposition (CVD) are thermally depolymerized at about 300°C. In our method, the sacrificial polymer is created by spin coating or dispensing and can be photodefined using regular optical exposure tools. The polymer encapsulation is cost-effective and requires common processing tools including spinners, mask aligners, and ovens. Hermiticity is obtained by sputtering, evaporation or electroplating of a metal having a thermal coefficient of expansion (CTE) close to that of substrate. Metal-Organic Package Description Figure l.a shows the schematics of a metal-organic packaged beam resonator in silicon-on-insulator (SOI) substrate. Figure l.b is the electron micrograph of a packaged beam resonator. The cavity is only 80imx 120imx 15pim.
(a)
(b
........
(b) . . .
Fig. 1. a) Schematics of a packaged SOI beam resonator (15,um thick), b) Electron microscope view of a packaged resonator [10]
1-4244-0261-1/06/$20.oo A©2006 IEEE.
139
In Figure l.a, the metal-organic square overcoat has a width, L, polymer thickness, tpolymer, and metal thickness, tMetal. The air cavity is created by thickness of the sacrificial polymer, tCaviy. The maximum cap bending due to the pressure difference between inside and outside the cavity, Ap, should be less than tCaviy.. The package design should predict the proper thickness of the sacrificial polymer to completely cover the MEMS holes. Also, the minimum overcoat thickness, required to tolerate the pressure difference, Ap should be calculated. For the rectangular cap of Figure l.a., the package bending, Z, is formulated in (1) [13]. Here tc, E and, v are the cap effective thickness, Young's modulus, and Poisson ratio, respectively: Ap-
2 ° t Z+
2L2
2
4E
t3CZ +
4C 3(1- v )L
(
2
+
1674E t3Z+
3(l1_V2 )r4
(7-v)tCz3 3(l1_V)r4
(2)
The three terms in equations (1, 2) represent thin film residual stress, plate bending due to flexural rigidity, and inplane stretching, respectively. For thin-films, the second term is usually the largest term. For the square cap with a cavity height, tCavily, tc should be designed so that the maximum plate bending (Zmax) is less than tcavity:
3(1
V2 )L4
(a) Insulator
2)
128(l1-v)L4CtLZ3
For a circular cap with radius, r, equation (1) can be rewritten as (2):
Ap4o-t r
and expedite the evacuation. The metal film is deposited as in Figure 2.d. Finally to get access to the pads, the metal film and consequently the isolation film are patterned and etched (Figure 2.e).
(b) Polymer overcoat `
(c) Metal overcoat >
Ap3
Zmax - 2 Et (3) For the metal-organic package, the Young's modulus of the metal film is orders of magnitude larger than that of the polymer. For gold, E=79GPa and vo0.44. Using a 1I m thick gold overcoat in equation (3) and the mentioned values for E, v, and Ap=latm yields Zmax =1.57[tm. Therefore a sacrificial polymer with minimum thickness of 2pm is required to ensure that the MEMS package will survive the pressure difference without collapsing. The polymer overcoat can be selected among variety of photo-definable polymers with high glass transition temperature.
(d)
(e) Fig.2. Polymer-based packaging process.
Table 1 lists the processing steps in detail for Unity and Avatrel. For a Unity formulated with photo-acid generator MEMS Packaging Process Figure 2 describes the packaging steps. In Figure 2.a., the (PAG), the development involves PAG-assisted sacrificial material is first spin-coated or dispensed to cover decomposition of the exposed area at around 110°C on a the active MEMS components (such as beam in a beam hotplate followed by a short dip in isopropanol alcohol. TABLE I resonator). In Figure 2.b, a layer of insulator such as silicon PROCESSING PARAMETERS POLYMER dioxide is deposited prior to polymer overcoat formation. This Parameter Unity (9pm) Avatrel (10im) layer increases the rigidity of the package [14] and creates Spin coating speed 3800rpm 4200rpm isolation between MEMS bond pads and metal package. The Soft bake l0min/110°C l0min/1000C polymer overcoat is spin coated and patterned to cover the MEMS except the bond pads. The sacrificial material is then IJ/cm2 (240nm) 0.2J/cm2 (365nm) Exposure Post-exposure bake l0min/1000C decomposed at 200-250°C. This is the highest temperature step in this process. The by-products of thermal Development Spray 1min -1 0min/1 10°C decomposition (oxygen, hydrogen, carbon monoxide, etc.) Decomposition 1°C/min to 1600C, hold for 1 hour permeate through the polymer overcoat (Figure 2.c). The 1°C/min to 2500C, hold for 2 hours organic package is left in a metal deposition chamber such as an electron beam evaporator or a DC sputterer. The package Dispensing of sacrificial polymer can be done using a syringe should be left for enough time prior to deposition to create to create a dome-shaped capsule. This is suitable for bulk vacuum inside the cavity. Heating can increase gas diffusivity 140 1-4244-0261-1/06/$20.00 A©2006 IEEE. -
micromachined structures with wide and deep cavities. This method is especially suitable for MEMS devices with delicate components that may break during sacrificial material spin coating. The examples include MEMS accelerometers and gyroscopes.
(a)
Characterization of the Polymer Package First, the electrical, mechanical, thermal, and optical parameters of Avatrel were measured, as listed in table 2. The small dielectric constant is suitable for low-loss RF MEMS packaging. The glass transition temperature of Avatrel (-3000C) is much higher than that of Unity (-1OOC). TABLE II MEASURED PROPERTIES OF AVATREL [15]
|Equipment
Electrical
Permitivity at 1GHz
hp network analyzer
2.57
Mechanical
Young's modulus (at 250C) |2GPa Poisson's ratio 0.3
Hysitron nanoindenter Hysitron nanoindenter
Thermal Glass transition temperature
3000C
KLA Tencor profiler
1.5-1.6
Woollam ellipsometer
Optical Index of refraction
(b)
The reduced-thickness of Avatrel and its stress after curing was measured using a contact profiler, as shown in Figure 3.a A 10ptm thick Avatrel overcoat can have a tensile stress of 4.5GPa after curing at 200°C (a typical decomposition temperature used for Unity).
Fig.4. Scanning electron microscope view of a resonator: a) before and b) after packaging by photo-defining Unity.
The titanium/gold was deposited in a DC sputterer in a chamber pressure of about 6mTorr. Etching of gold was performed in a Transene gold etcher. Finally the oxide isolation film was removed in buffered oxide etch (BOE). Figure 5 shows the microscope view of the cavity after decomposition and prior to metalization. Avatrel is transparent to visible light, it can be seen that the cavity is free of sacrificial polymer residues.
6
0
2
4
6
8
10 12 14 16 18 20
Avatrel Thickness, pim Fig.3. Avatrel stress versus thickness after curing at 2000C.
The packaging technique was applied to package a variety of MEMS devices. This includes (but not limited to) inertial sensors, resonators, and RF passives. Characterization of the Packaged Resonators Figure 4 shows the packaging of a 2.6MHz beam resonator. Figure 3.a. shows the viewgraph of the resonator with 1 ptm transduction gap before packaging. Figure 3 .b is the same device after packaging. The thickness of sacrificial polymer should be at least 2pm to properly cover the air gap. Thickness of Unity, Avatrel, and gold are 9ptm, 10ptm, and 2.5ptm, respectively. The polymer processing steps are the same as listed in Table 1.
Besides photo-defining and dispensing, Unity can be created by reactive ion etching (RIE) or molding. RIE of nonphotodefinable Unity requires using titanium or oxide to be used as the mask. A 50sccm, 200W oxygen plasma at 5OmTorr pressure, can etch Unity with a rate of 0.2tm/min. Besides RIE, this kind of Unity can be also patterned by molding. Figure 6 shows a package created by molding 10pQm Unity. In the molding method, the packaging sequence starts with deposition of an oxide mold using plasma enhanced CVD (PECVD), to overbridge the trenches. This is followed by patterning the oxide with a negative resist, and removing of oxide in RIE. Then Unity is spin coated and cured on top of the oxide. Finally oxygen plasma removes the Unity on top of the mold and leaves some Unity on the channel. The oxide mold is removed in BOE. The metal-organic package in this
1-4244-0261-1/06/$20.oo A©2006 IEEE.
141
device consists of 2ptm/1Oim chromium/Avatrel layers. Chromium has a close CTE to silicon.
The reduction in Q is due to presence of small residues on the beam surface that increase the surface loss. As shown in Figure 8, Q after packaging (Qp) is always lower than unpackaged Q (Qu). Therefore, the vibration amplitude of the packaged resonator (QpAo) and resonance frequency (fp) are always lower than the unpackaged amplitude (QuAo) and resonance frequency (fu), respectively. Several batches of packaged resonators were examined and it was observed that in average, Q was degraded by 8% and the frequency change (A4) was within 0.2%.
Af
A' Fig.6. Scanning electron microscope view of a packaged resonator by molding Unity.
A packaged resonator was tested at wafer-level inside a vacuum probe station. A DC polarization voltage in the range of 70-80V was applied to substrate, while the electrodes were directly connected to the network analyzer. Figure 7(a) and (b) show the frequency response of the resonator in vacuum before and after packaging. The high Q factor of -4500 did not change significantly for this device, proving that thermal decomposition does not affect the performance of the device.
fp fu
Fig.8. Conceptual demonstration of degradation in Q and output power after packaging.
In some batches, a lower decomposition temperature was used (150°C) and the packaging did not result working samples. After breaking the polymer cap, the Unity residues were analyzed using a VEECO atomic force microscope (AFM) in tapping mode configuration. As Figure 9 shows, the thickness of residues was measured from the maximum height difference and it was about 200A°. Heating the packaged resonators up to higher temperature (-250°C) resolved the residue issue.
Fig.9. Residue analysis using a VEECO AFM in tapping mode.
Characterization of the Packaged HARPSS Varactors A variable capacitor (varactor) was created using the high aspect ratio single crystal silicon and polysilicon (HARPSS) technology [16]. Figure 10 shows the HARPSS varactor before and after packaging. The varactor is made by using a 1-2pm gap between polysilicon and silicon electrodes and the series resistance has been reduced by evaporation of gold to increase the electrical Q factor. The 60ptm thick, 2pF HARPSS capacitor (lmmx1.5mm area) shows a tuning range of 2:1 by applying a voltage of about 2V [16]. The measured electrical Q is about 50 at 1GHz. The metal-organic package does not change the DC performance of the device, although it can change the RF performance. Packaging was done by
Fig.7. Q versus frequency for a beam resonator: a) before and b) patterning 15ptm thick unity, and 20pm thick Avatrel. The after packaging using an input signal level of -5dBm. hermiticity was added to the polymer package by evaporation 142 1-4244-0261-1/06/$20.oo A©2006 IEEE.
of 1 [tm gold overcoat, eliminating the extra mask needed to pattern the metal (step 'e' in Figure 2). Moreover, evaporation of metal overcoat instead of deposition and etching metal, eliminates the need for deposition and etching the extra insulator layer before Avatrel patterning (step 'b' in Figure 2). The only limitation is that the thickness of evaporated metal should be less than thickness of the polymer overcoat.
Characterization of the Packaged HARPSS Accelerometer HARPSS technology was used to fabricate micro-gravity silicon accelerometer [18]. Fig. 12.a shows a 50pim thick, 1.8mmxl.5mm X-axis accelerometer. Fig. 12.b shows the
accelerometer, packaged by dispensing Unity and overcoating a 120im thick Avatrel overcoat. The static sensitivity of the packaged accelerometer is measured to be about 0.27pF/g [10].
Fig. 10. Scanning electron microscope view of a HARPSS varactor: a) before and b) after packaging by photo-defining Unity.
The gold-Avatrel package was charactFerized using an Agilent 8517 vector network analyzer. The Ipackage adds an insertion loss of about 1.4dB at 1 GHz and 1..5dB at 5GHz to the MEMS device, as shown in Figure 11. The package is only 20% larger than the device. The horizon[tal feedthroughs in this packaging method do not add a signifrican size to me package. In the wafer bonding methods, tihe feedthroughs form a large component of the final MEMS pzackage size [17].
IlI Fig.12.Scanning electron microscope view of a HARPSS accelerometer: a) nonpackaged and b) packaged by dispensing Unity [10].
Conclusions A low-cost low-temperature packaging technique for wafer-level encapsulation of MEMS devices is presented. The packaging process does not involve wafer bonding, high temperature deposition of LPCVD or epitaxial films, or wet etching of sacrificial thin-films. It utilizes decomposition of a sacrificial polymer through a photo-definable polymer overcoat. Cavities as small as 0.00015mm3 (for resonators) m and as large as 1mm3 (for inertial sensors) were created on U,0 MEMS devices using 10-15m thick gold-Avatrel or re Packaging -J chromium-Avatrel package. The packaging has been Packaging 0 successfully applied to encapsulate surface/bulk oU, micromachined structures including beam resonators on SOI substrate with reduced-size feedthrough, HARPSS varactors and HARPSS accelerometers. For the 2.6 MHz beam resonators, a quality factor of about 4500 was measured before and after packaging. For the HARPSS varactor, the gold-Avatrel package adds about 1.5dB of loss at 5GHz as a Frequency, GHz result of capacitive loading of the package. For the Fig. 11. Measured insertion loss of the HARP' SS varactor before microgravity accelerometer, the static sensitivity after and after low-loss packaging polymer packaging is about 0.27pF/g. Stress measurement of 143 1-4244-0261-1/06/$20.oo A©2006 IEEE.
the organic cap was performed for different thicknesses and the different parameters of the polymer overcoat was extracted. Acknowledgments
The authors would like to thank STMicrolectronics for supporting this project and Promerus LLC for supplying the materials, especially especially Jeff Krotine and Ed Elce for the valuable discussions. References [1] Najafi K, "Micropackaging Technologies for Integrated Microsystems: Applications to MEMS and MOEMS," SPIE MOEMSIMEMS (San Jose, CA) pp.1-12, 2003. [2] Lee B, Seok S, Chun K, "A Study on Wafer-Level Vacuum Packaging for MEMS Devices," J. Micromech Microengineering, Vol. 13, pp. 663-669, 2002. [3] Mitchel J, Lahiji G R and Najafi K, "Encapsulation of Sensors in a Wafer-Level Package using a Gold-Silicon Eutectic," JIEEE Transducers (Seoul, Korea) pp.663-669, 2005. [4] Pan C T, Yang H, Shen S C, Chou M C and Chou H P, "A Low-Temperature Wafer Bonding Technique using Patternable Materials," J. Micromech. Microeng. Vol.12, 611-615, 2002. [5] Sparks D, Massoud-Ansari S and Najafi N, "Long-Term Evaluation of Hermetically Glass Frit Sealed Silicon to Pyrex Wafers with Feedthroughs," J. Micromech. Microeng. Vol.15, pp.1560-1564, 2005. [6] Lin L, "MEMS Post-Packaging by Localized Heating and Bonding," IEEE Trans. Adv. Packaging Vol.23, pp.608616, 2004. [7] Candler R N, Park W T, Li H, Yama G, Partridge A, Lutz M and Kenny T W, "Single Wafer Encapsulation of MEMS Devices," IEEE Trans. Adv. Pack. Vol.26, No.3, pp.227-232, 2003. [8] Stark B H and Najafi K, "A Low-Temperature Thin- Film Electroplated Metal Vacuum Package," IEEE J. Microelectromech. Syst. Vol.13, No.2, pp.147-157, 2004. [9] Joseph P J, Reed H A, Hongshi Z, Rhodes L F, Henderson CL, Allen S A.B and Kohl PA, "Air-Channel Fabrication
[13] Senturia S D, "Microsystem Design," Kluwer Academic Publishers., (New York, 2001). [14] Joseph P J, Reed H A, Allen, S A.B.; Kohl, P A, "Improved Fabrication of Micro Air- Channels by Incorporation of a Structural Barrier," J. Micromech. Microeng. Vol.15, pp. 35-42, 2005. [15] Patel K S, Kohl PA and Allen S A.B "Three-Dimensional Dielectric Characterization of Polymer Films," J. Applied polymer Science Vol.80, pp.2328-2334, 2001. [16] Monajemi P and Ayazi F, "A High-Q Low-Voltage HARPSS Tunable Capacitor," IEEE International Microwave Symposium (Long Beach, CA) pp.749-752, 2005. [17] Muldavin J, Bozler C, Rabe S, Keast C, "Wide-Band Low-Loss MEMS Packaging Technology," IEEE International Microwave Symposium (Long Beach, CA) pp.765-768, 2005. [18] Monajemi P and Ayazi F, "Analytical Design and Implementation of a Micro-Gravity HARPSS Accelerometer," Proc. Sensors Journal, Vol.6, No. 1, pp.39-47, 2006.
for Microelectromechanical Systems via Sacrificial J. Photosensitive IEEE Polycarbonates," Microelectromech. Syst. Vol.12, No.2, pp. 147-157, 2003. [10] Monajemi P, Jospeh P, Kohl P, Ayazi F "A Low-Cost Wafer-Level MEMS Packaging Technology," Proc. IEEE MEMS, Miami, FL, 2005, pp.634-637. [11] Hamett C, Satyalakshmi K, Coates G and Craighead H, "Direct Electron-Beam Patterning of Surface Coatings and Sacrificial Layers for Micro-Total Analysis Systems," J. ofPhotopolymer Sci. Tech., Vol. 15, pp. 493-496, 2002. [12] Chan K, Casserly T, Loo L, Gleason K, "Air Dielectric Fabrication via CVD Sacrificial Materials," Proc. MRS Symp. Vol. 766 (Materials, Technology, and Reliabilityfor Advanced Interconnects and Low-k Dielectrics), San Francisco, CA, April 2003. 144
1-4244-0261-1/06/$20.oo A©2006 IEEE.
