MEMS reliability

INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING J. Micromech. Microeng. 16 (2006) 676–683 doi:10.1088/0960-1317/16/4/002 Wafer level hermetic package and device testing of a SOI-MEMS switch for biomedical applications Rogier A M Receveur1, Michael Zickar2, Cornel Marxer2, Vincent Larik1 and Nicolaas F de Rooij2 1 Medtronic Bakken Research Center, Endepolsdomein 5, 6229GW Maastricht, The Netherlands 2 University de Neuchˆatel, Institute of Microtechnology, Jaquet-Droz 1, CH-2007, Neuchˆatel, Switzerland E-mail: rogier.receveur@medtronic.com Received 8 November 2005, in final form 28 December 2005 Published 27 February 2006 Online at stacks.iop.org/JMM/16/676 Abstract We have designed a wafer level chip scale package for a bi-stable SOI-MEMS dc switch using a silicon-glass hermetic seal with through the lid feedthroughs. Bonded at 365 ◦C, 230 V and 250 kg, they pass the fine/gross leak test after thermal cycling and mechanical shock/vibration according to MIL-STD-833, fulfilling the requirements for biomedical applications. The measured shear strength is 114 ± 26 N in correspondence with the theoretically expected 100 N. Ruthenium microcontacts are a factor of 100 more robust than gold microcontacts, being stable over 106 cycles measured in a N2 atmosphere inside the package presented here. Future work will include a more extensive bond quality assessment and continued microcontact reliability measurements. 1. Introduction The packaging of microelectromechanical system (MEMS) devices is one of the most difficult parts of the product development process due to its many requirements and functions [1]. It is more complex than conventional integrated circuit (IC) packaging due to a difference in nature between a MEMSand IC [2]. In many cases, a special environment needs to be created in cavities in the package to improve the function of the packaged MEMS device [1, 3–10]. This package redistributes signals, is a mechanical support, needs to be able to handle the power and takes care of thermal management. It needs to be compatible with other mounting techniques [11] and solves the temperature coefficient of expansion (TCE) mismatch between the device and substrate [4, 12]. Conventional packaging techniques take up between 75 and90%of the device costs [2, 7, 13] and as much as99%of the size [7]. As a consequence, wafer level packaging is increasing in importance [1] and still largely under development [2, 7]. Wafer level costs low due to the batch character of the process [4, 5, 12, 13] and small size [10–13]. Structures are already protected early in the process [5, 7], and testing can be done on the wafer [12]. Smaller packages could enable entirely new applications [13], and lead to better reproducibility [4] and reliability [5, 13] and to better electrical properties due to the shorter paths [4]. We have designed a bi-stable silicon on insulator (SOI)-MEMS direct current (dc) switch for biomedical applications [14]. A schematic drawing of the working principle of the bi-stable MEMS switch is shown in figure 1. This switch will be used to select electrical stimulation electrodes located on leads implanted in the human body. The wafer level hermetic package presented in this paper protects the fragile movable microstructures of the switch against dust. In addition, it provides a cavity in which a controlled environment can be maintained, improving the long-term microcontact reliability. It allows for batch processing of multiple packages in parallel leading to low cost per package. We also present here a hermetic feedthrough technique using a through the lid approach that allows contact with two control pads and two signal pads needed to operate the bi-stable switch. The requirements on the package are presented in table 1. These are part of the requirements that a 0960-1317/06/040676+08$30.00 © 2006 IOP Publishing Ltd Printed in the UK 676Package and testing of a MEMS switch for biomedical applications (a) (b) (c) Figure 1. Schematic of the bi-stable switch, not to scale, shown in the OFF position (a) and ON position (b). x is the displacement of the moving parts, where x = 0 is defined at the initial position of the central beam (as fabricated). Fx is the mechanical force exerted on the central beam by the double clamped beam and the hinges, which is plotted as a function of displacement in (c). Bi-stable behavior is generated by the double clamped beam and hinges. x0 is the offset distance and xc is the contact distance. The central beam can be moved in a positive or negative x-direction by the electrostatic comb actuators. In the ON state, the central beam is moved to the right to contact the opposing contact member. When the beam is displaced from its initial stable position (OFF state) in the positive x-direction the force becomes negative, meaning it is opposing the motion. At the offset position x0, the force changes direction and is now promoting further motion. The fixed contact member is placed at position xc for maximum contact force (Fc) (stable ON state). Table 1. Sealing requirements. MIL-STDItem 883 method Description Fine leak 1014 Fine He BOM Gross leak 1014 Gross Bubble test Shear strength 2019 Destructive Thermal cycling 1010 Cond B −55 ◦C/+125 ◦C, 10 cycles Mechanical shock 1010 Cond B 5 shocks, ×6 directions 1500 g, 0.5 ms half sine Mechanical vibration 2007 Cond A var. frequency 150–2000 Hz 20 g component inside an implantable medical device must fulfill. This paper will not address the requirements that have to be fulfilled when exposing this package to the body. This paper starts with a short review of hermetic wafer level packages and feedthroughs as reported in the literature. The design, experimental and theoretical methods are described. Section 4 contains the results and discussion, including the results of device testing performed inside the wafer level package. In the last section, we conclude till what extent the package meets the requirements for use in biomedical application. 2. Packaging and interconnect We present a short overview of ways to hermetically seal MEMS devices and electrically interconnect them from outside to place the design presented in this paper in the proper context. 2.1. Hermetic seal Direct bonding creates a bond between two silicon wafers that are pressed together at high temperature (1000 ◦C). A very low surface roughness and no particles between the wafers are required. With a special surface activation method, a low temperature (200–400 ◦C) variant was developed, creating bond strengths up to 2 J m−2 [15]. This technique makes it possible to create very small gaps between the two wafers, has no TCE mismatch and is compatible with complementary metal oxide semiconductor (CMOS). The anodic (or electrostatic) bond between glass and silicon is one of the most applied techniques in MEMS, for example, in pressure sensors and accelerometers [16]. It is believed that a covalent bond is formed between the surface atoms of the glass and the silicon when both wafers are pressed together under an application of a voltage at elevated temperatures [7], creating a hermetic seal [6]. The surfaces need to be clean [1] and planar, with a surface roughness of ∼100 A° [17]. A glass with a TCE close to that of silicon (like Pyrex) should be used [16], where the silicon can also be a fine-grain poly silicon layer [17] or an SOIwafer [1, 12]. Applied temperatures are in the range of 300–450 ◦C [1, 12, 17, 18] with optimum bond quality achieved at uniform distribution and slow cooling down. Voltages can range from 200 to 2000 V [1, 12, 17, 18] with an importance for the uniformity. The required bond width is 250 μm [17]. The processing time is in the range of 10–20 min [1, 12] and can be done in an arbitrary environment. The resulting bond strength can be >20 MPa [1]. Outgassing of oxygen from the glass during bonding [3, 16] can be limited by pre-baking prior to bonding [17]. The mobile alkali ions in the glass (required for anodic bonding) can diffuse to other parts of the microstructure and/or electronics and cause failures [16]. The process can be performed locally using a Nd:YAG laser with 300 μm beam and 10–30 MPa applied pressure at temperatures of <250 ◦C [16].For bonding using an intermediate glass layer, a paste filled with glass particles is screen printed on the lid with a thickness of 20 μm and a width of 500–600 μm. Both wafers are pressed together and bonded at a temperature of ∼450 ◦C [9, 12]. Alternatively a metal intermediate layer can be used. SnPb/NiAu [5, 19] and SnAgCu/Pd [5] material combinations have been demonstrated. The metal system is deposited, patterned, plated and/or printed on both wafers and reflowed at temperatures in the range of 180–320 ◦C. Minimum frame width is in the range of 80 μm [5]. Polymer intermediate layers are classified as non-hermetic seal [4, 9]. Benzocyclobutene (BCB) is spin coated and patterned on one of the two wafers and cured at temperatures <250 ◦C [20, 21]. An entirely different class of protection is to provide a thin film encapsulation. In general, a seal layer is deposited over a sacrificial layer and released through etch holes that are then hermetically sealed. It can be a poly silicon layer over an oxide layer, HF vapor released and hermetically sealed by a passivation layer [11] or a 40 μm electroplated Ni layer over photoresist sealed with a solder using a mold and transfer technique [10]. A variation on this process is presented in [8] using a thick epipoly layer with DRIE-etched release holes. 677R A M Receveur et al 2.2. Hermiticity When fabricating wafer level packaged devices, one would like to know the hermiticity level or more in general the quality of the bond to assess compliancewith the requirements. Often MIL-STD-883 is used to objectively define various hermeticity levels [5]. General bond quality test methods employed include gross leak test, fine leak test, mechanical shear test and fracture [1, 5]. Stressing the package is done by temperature cycling, high T high RH storing, drop test and thermal shock, autoclave or saline soaking [4, 10, 17]. In situ test structures used to determine changes occurring inside the sealed cavity include pressure sensors, moisture sensors, dew-point sensors, tensiometric bridge and pirani gauges [10, 12, 17, 22]. Packages can be inspected using a scanning electron microscope (SEM), scanning acoustic microscopy or ultrasound [1, 18]. 2.3. Hermetic feedthrough In most cases, it should be possible to electrically connect structures inside the package without disturbing the longterm stability of the environment created inside the cavity. A development decreasing the need for feedthroughs is a 3D system design and CMOS-MEMS integration, but a review of these possibilities is beyond the scope of this paper. ‘Under the lid’ interconnects extend out of the package in a lateral direction, in the plane of the substrate and lid. Filling the trenches in the SOI device layer would allow an ‘under the lid’ approach. A sealing ring can then be constructed over these areas without causing a leakage path to the inside of the cavity. A silicon nitride [23, 24] or silicon dioxide [25] layer can be used to form a mechanical connection but an electrical isolation. LPCVD deposited lowstress silicon nitride has good conformal step coverage (completely filling the trenches), and a smooth and flat surface remains after removal of the top layer with CMP [24]. Buried conductor traces in the substrate are another possibility to electrically get out of the hermetic cavity in a lateral way [6, 17, 26]. The addition of electrical interconnect between the lid and MEMS wafer can decrease the need for feedthroughs [19]. This has been demonstrated using the same metal as is used for the sealing [19] and with the use of mechanical press contacts between metalization on the lid and substrate [6, 26]. ‘Through the lid’ interconnects extend out of the package in a vertical direction, perpendicular to the plane of the substrate and lid. A silicon cap wafer wet etched through holes eutectic bonded to a silicon substrate is described in [1]. A small remaining gap between metalization on the substrate wafer (deposited prior to the eutectic bonding) and the cap wafer is closed by sputtering a thick (1 μm) aluminum layer. Deep reactive ion etched through holes of 40–60μmin 130μm Pyrex glass filled with Ni by pulsed electroplating are described in [27]. Through the lid wet etched holes in glass have been shown in [26], but they are used as a pressure inlet only. (a) (b) Figure 2. Schematic cross section of the packaged microswitch. The SOI carrier layer, silicon dioxide, SOI device layer, metalization and glass lid are indicated. (a) After removal of the temporary shadow mask used for selective metal deposition on the contact members and bondpads and (b) the final packaged switch. 3. Materials and methods 3.1. Design and fabrication process We select anodic bonding since it is a well-known process capable of forming a hermetic seal. It can be performed at the wafer level of which we expect a positive effect on the packaging costs. We present here a new process for creating hermetic feedthroughs through the lid. Each contact is electrically isolated in the SOI device layer. The structures and lid are designed using the Expert system. A 200 μm sealing ring surrounds the complete structure and each individual bond pad. The switch is fabricated out of a SOIwafer using a singlemask step [14]. The DRIE process is tuned to create a slight negative angle, causing the contact position to be at the top side of the device layer where we expect the most metal. We now use a shadow mask technique to selectively deposit metal on the contact members and bond pads while keeping the glass silicon bond area clean (figure 2(a)). A silicon wafer with KOH etched through holes is manually attached to the MEMS wafer. A chromium– gold–nickel–gold layer (Cr/Au/Ni/Au, 10/250/30/200 nm) is deposited using a special rotating evaporation setup, and the shadow mask is removed. As a comparison, ruthenium (Ti/Ru, 10/1000 nm) was also deposited using a conventional sputtering setup with no special measures to deposit on the vertical side walls. Ruthenium has a higher Vickers hardness than gold (220 versus 25) so we expect a longer lifetime for switches coatedwith this metal. Glass lids (Pyrex, 500μm) are pre-etched in 50% HF from both sides through a 0.5 μm poly silicon etch mask deposited by LPCVD (temperature is 570 ◦C and the deposition rate is 2.7 nm min−1). After removal of the poly silicon, glass lid and MEMS wafer are aligned (Electronic Visions) and moved to the wafer bonder (EVG). Anodic bonding is done at two different conditions, group I 365 ◦C, 350 V, 200 kg and Au metalization and group II 365 ◦C, 230 V, 250 kg and no metalization, both in an N2 atmosphere. Bond areas and scribe lines in the glass lid are opened by a final 50% HF etch of the complete wafer stack. The wafer can then be diced into individual capped switches (figure 2(b)). 678Package and testing of a MEMS switch for biomedical applications Figure 3. SEM picture of a packaged switch with feedthroughs. The etched holes in the glass lid open the way to contact the metal on the pads (visible as the light square areas). On the side, the SOI carrier layer, silicon dioxide and device layer are visible. Figure 4. Cross section through a feedthrough encapsulated in epoxy. 3.2. Experiment, theory and simulation The packaged switches are inspected under the scanning electron microscope. Cross sections of the feedthroughs are made by encapsulating the complete die in epoxy, dicing and inspection under an optical microscope. Bond quality and environmental stress tests are carried out by Zarlink (Caldicot, SouthWales) according to the standards as described in table 1. Finite element analysis is done by Infinite (The Netherlands) using ANSYS. Wafer level packaged individually diced switches are placed inside a 8 pin Dual Inline Package (DIL-8) and electrically contacted by wirebonding to the metal pads inside the through the lid holes. Switch contact properties are then measured as reported earlier [14] with the exception that we now switch off the current through the contact before opening and closing the switch (cold switching). 4. Results and discussion A SEM picture of a sealed chip is shown in figure 3 and a cross section in figure 4. The details of the lid are shown in figure 5. 4.1. Bond quality With the small cavity volumes of typical MEMS packages, special care has to be taken with the interpretation of leak rate results [20]. Our cavity volume is approximately 150 nL. Following the same approach as in [20], this leads to an upper limit for the fine leak rate (Rfl) of 1.5 × 10−6 mbar l s−1 and a lower limit for the gross leak rate (Rgl) of 1 × 10−4 mbar l s−1. Our hermiticity criterion for passing the fine leak test is 2 × 10−8 mbar l s−1, well within the range of detectable rates (a) (b) Figure 5. SEM photographs of the glass lid. (a) Bottom view showing the recessed area that is positioned above the moving structures and the feedthroughs. (b) Detail of the feedthrough seen from the bottom side. (the lower limit fine leak rate Rfl is 0.4 × 10−9 mbar l s−1). The hermiticity of anodic bond as reported in the literature is limited by the diffusion of gas through the glass. This leads to an expected leak rate of 10 × 10−15 mbar l s−1. With this in mind, we have interpreted samples with zero leak rate as having a leak rate below the detection limit, and classified them as pass. We are working on measurements on the MEMS devices itself inside the cavity to detect leak rates in the undefined regime. From the 37 as-fabricated dies, 33 and 31 pass both seal tests (groups I and II, respectively). The overall bond quality and stress test results are given in tables 2 and 3 and will be described in more detail in the following sections. For the shear test, a custom holder was made to fix a single packaged chip on the SOI side and for pressing the glass lid sideways with an automated tool while measuring the force according to MIL-STD-883 method 2019 (destructive test). All shear strength tests (before and after thermal/mechanical stress tests) are passed. The actual value was 137 ± 37 N (group I) and 114 ± 26 N (group II). Reported bond strengths are larger than 20 MPa [1]. The bond area for our devices as estimated from the mask sets is 5 × 10−6 m2, giving an expected bond force of 100 N. The test samples are fixated to a holder using glue for mechanical stress tests. The glue is carefully removed afterward in order to prevent negative side effects in the leakage tests. Mechanical shock is carried out according to the MIL-STD-883 method 1010 condition B. The package is subjected to five 1500 g shocks in six directions. Mechanical vibration is carried out according to MIL-STD-883 method 2007 condition A. The package is subjected to a varying frequency between 150 and 2000 Hz, with an acceleration of 20 g. Both mechanical tests are survived by 14 out of 15 and 13 out of 15, respectively (groups I and II). For thermal stress tests, the samples are placed in a chip carrier that is placed inside an oven. This oven can provide the thermal profile as prescribed by MIL-STD-833 method 1010 condition B. The temperature is cycled ten times between −55 ◦C and +125 ◦C with a rise and fall time of less than 1 min and a stable time of 10 min. Thermal cycling is survived by 18 out of 18 and 14 out of 16 samples, respectively (groups I and II). In order to assess the thermally induced stresses caused by the difference in TCE between glass and silicon, numerical calculations were carried out. Table 4 gives the 679R A M Receveur et al Figure 6. Equivalent stress in the Pyrex lid (looking at the surface that interfaces with the SOI device layer) at a temperature of −65 ◦C calculated using finite element analysis. Maximum stress in the order of 0.252 MPa occurs at the edges and around the holes. Table 2. Test results for group I 200 N, 350 V, 365 ◦C, Au. The table is to be read from left to right. Numbers mean pass (and in brackets failed) samples. All passed samples proceed to the test(s) in the next column. Note that the shear test is destructive. The samples that pass the fine leak test proceed to the gross leak test. Initial quantity Bond quality Stress Bond quality Stress Bond quality 40 Shear 3(0) Thermal 18 Fine 18(0) Fine 33(4) Gross 18(0) Gross 33(0) Shear 3(0) Mech. shock Fine 14(1) Mech. vibr. Fine 14(0) 15 Gross 14(0) 14 Gross 14(0) Shear 3(0) Table 3. Test results for group II 250 N, 230 V, 365 ◦C, no metal. For more explanation, see table 2. Initial quantity Bond quality Stress Bond quality Stress Bond quality 40 Shear 3(0) Thermal 16 Fine 14(2) Fine 32(5) Gross 14(0) Gross 31(1)a Shear 3(0) Mech. shock Fine 14(2) Mech. Vibr. Fine 13(1) 15 Gross 14(0) 14 Gross 13(0) Shear 3(0) a Operator error. Table 4. Material properties used for finite element analysis [7, 28, 29]. Property Unit Symbol Si SiO2 Pyrex Coefficient of thermal expansion 10−6 ◦C−1 α 2.6a 0.55 3.2 Thermal conductivity at 300 K W cm−1 K−1 λ 1.57 0.014 0.011 Young’s modulus GPa Y 160 73 0.65 Poisson’s ratio – ν 0.22 0.17 0.2 Yield strength GPa σ 7 8.4 ? Specific heat J g−1 K−1 cp 0.7 1.0 0.75 Density g cm−3 ρ 2.4 2.3 2.23 a Depends on temperature. material properties that were used to numerically calculate the thermally induced stresses. Figure 6 gives a picture of the induced stresses at the glass silicon interface. The maximum stress is in the order of 0.12–1 MPa and occurs at the edges and around the holes in the lid. Dynamic calculations do not pose any significant additional stress. 680Package and testing of a MEMS switch for biomedical applications (a) (b) Figure 7. Top view of the glass lid of the failed samples inspected under an optical microscope (scale approximately 500 μm). (a) Visible crack in the the top of the glass lid. (b) Pit in the top of the glass lid, due to a hole in the etch mask. The pit clearly does not penetrate all the way through the glass. Similar defects could be observed on the passed samples. These observations do not explain the failures. All failed samples were visually inspected under a microscope (see figure 7). Sometimes defects could be observed, but not on all samples. In addition, similar types of defects could also be observed on passed samples. Other inspection methods as described in [1, 18] have not yet been developed in our group. Therefore, we cannot be more specific on why the samples failed based on these inspections. Our anodic bonding process parameters are at the low side of the spectrum reported in the literature and not at the optimum [18]. However, the measured bonding strength is close to the expected value, and the largest part of the samples passes the leakage tests after thermal and mechanical stress. We conclude that our process creates an acceptable bond quality for our purposes. 4.2. Microcontact properties As an alternative to the gold metalization described above, we have also fabricated samples with a titanium (Ti)–ruthenium (Ru) layer (1000 °A/8000 A° ) deposited using a conventional sputtering setup. Themain advantage of Ru overAu is that it is much harder and can withstand higher temperatures. The latter means that we have less constraints on the temperatures we can use during the anodic bonding. Since this metal is much harder, we expect less contact damage to occur in comparison to Au [14]. Ruthenium coverage on top of the SOI device layer as measured with the alpha stepper is 1 μm. EDX measurements were performed at the top of the structures and on the vertical side walls of the device layer close to the top and the bottom. Since the same angle of incidence was used at all positions, we can make relative comparisons between Ru coverage. From these measurements, we estimate approximately 0.39 μm Ru at the top of the side wall. Figure 8 shows a SEM picture with a close-up of the two contact members. Inspection of the contacts before the duration tests shows contact damage (see figure 10). We attribute this to electrical effects on the released structures during the anodic bonding process. Unfortunately, we used high resistive silicon for these samples so we cannot determine the influence on the initial contact resistance. Therefore, we will only investigate the relative changes in contact resistance over time in the remainder of this paper. (a) (b) Figure 8. SEM pictures showing the contact position and the gaps between the contact members ((b) is a detail of (a)). The right part with the semispherical bump on it is the movable part, touching the thin opposing contact member that is fixed. Note that there is a small gap at the top of the contacts that can be influenced by the deep reactive ion etching. To assess the difference in contact properties between Ru and Au metalized switches, we measured the open and closed resistance while repeatedly cold switching the contact (see figure 9). Switches were packaged as described in this paper. The ruthenium-covered switches show a marked change in behavior after roughly 106 cycles, compared to 104 for the gold-covered switches. Switches will only have to change state at the time of implant and in exceptional cases during the lifetime of the device so we expect that we can apply both these switches in this case. A SEM picture of the contact surface before and after the test for both metalizations is shown in figure 10. From these pictures, a clear difference in the type of damage between Ru and Au can be seen as well. The contact damage in the ruthenium-coated switches is much more localized than in the gold-coated switches. Both these results correspond with what we expected because of the higher hardness of ruthenium compared to gold. In summary, given the difference between the two metals described above, ruthenium would be the preferred choice. 681R A M Receveur et al Figure 9. Resistance as a function of number of cycles for a ruthenium-covered switch in closed and open position (cold cycled). The absolute resistance is very high since we used undoped silicon for these tests. However, the difference between closed and open resistance, and in particular the change of this difference over time is attributed to the microcontact properties. It is clear that at approximately 12 × 105, the contact is damaged and the difference between open and closed resistance disappears. For this application, the switch only has to change state a couple of times. Figure 10. SEM pictures of the movable contact part, all at same magnification. Au and Ru coated switches (top and bottom row, respectively) and unused and used switches (left and right column, respectively). Note that the unused switches already show contact damage attributed to electrical effects on the released structures during anodic bonding. The used switches have gone through duration testing. We only investigate relative changes in microcontact properties as a function of duration test cycles. Note the difference in morphology of the contact damage between gold and ruthenium coated switches due to the differences in material properties. 5. Conclusion A short review of the available packaging and feedthrough techniques has been presented. We selected anodic bonding since it is a well-known process capable of forming a hermetic seal that will fulfill the requirements for our biomedical application of the MEMS switch as given in the introduction. It can be performed at the wafer level of which we expect a positive effect on the packaging costs. We have presented a process for creating hermetic feedthroughs through the lid. Each contact is electrically isolated in the SOI device layer and individually surrounded by a sealing ring. Structures bonded at 250 N, 230 V, 365 ◦C pass thermal and mechanical stress tests according to MIL-STD-883 and meet the requirements for biomedical applications for these aspects. Numerical analysis shows that the thermal stresses in the glass silicon stack are a factor of 20 lower than the bond strength, which is confirmed by shear strength measurements to be in the order of 20 MPa. We have demonstrated that ruthenium-covered contacts are robust for 106 cycles, which is a factor of 100 more than gold-coated contacts. These measurements were carried out in an N2 atmosphere inside the wafer level package described in this paper. In summary, the anodic bonding of a glass lid with through the lid feedthroughs to our MEMS device presented in this paper passes our initial requirements for biomedical application. Ruthenium-metalized switches tested inside the wafer level hermetic cavities show longer lifetime than gold-coated switches. These results form the basis for further advancements toward application in medical devices. Future work will include a more extensive bond quality assessment and continued microcontact reliability measurements. 682Package and testing of a MEMS switch for biomedical applications Acknowledgment The authors would like to thank L Nygren from Medtronic for ruthenium deposition. References [1] Wong C K, Wei J, Qi G J, Wang Z F, Jin Y F and Lim P C 2003 A wafer level packaging for pressure sensors MEMS Micro System Technologies 2003 (Poing, Germany: Franzis Verlag GmbH) pp 123-30 [2] Najafi K 2003 Micropackaging technologies for integrated microsystems: applications to MEMS and MOEMS Proc. SPIE 4982 9–27 [3] Caplet S, Sillon N, Delaye M T and Berruyer P 2003 Vacuum wafer-level packaging for MEMS applications Proc. SPIE 4979 271–8 [4] Feiertag G, Kruger H and Selmeier P 2003 Advanced packages for surface acoustic wave components Micro System Technologies (Poing, Germany: Franzis Verlag GmbH) pp 44–51 [5] Harder T, Schimanski H, Bertels L and Zoberbier M 2003 Wafer-level encapsulation of MEMS using solder sealing Micro System Technologies (Poing, Germany: Franzis Verlag GmbH) pp 52–9 [6] Hedenstierna N, Habibi S, Nilsen S M and Kvisteroy T 2003 Bulk micromachined angular rate sensor based on the ‘butterfly’—gyro structure Micro Mechanics Europe Workshop (Delft, The Netherlands: Delft University of Technology) p 175 [7] Madou M 1997 Fundamentals of Microfabrication (New York: CRC Press) [8] Partridge A 2003 A Lateral Piezoresistive Accelerometer with Epipoly Encapsulation (CA: Stanford University) [9] Reimann M, Ulm M, Buck T, Schobel J and Dechow J 2003 RF MEMS Glass Frit Packaging Micro System Technologies (Poing, Germany: Franzis Verlag GmbH) pp 60–7 [10] Stark B H and Najafi K 2004 A mold and transfer technique for lead-free fluxless soldering and applications to wafer-level low-temperature thin film packages Int. Conf. on Micro Electro Mechanical Systems (Piscataway, NJ: IEEE) vol 17 pp 13–6 [11] Aigner R, Opperman K-G, Kapels H and Kolb S 2001 Cavity-micromachining technology: zero-package solution for inertial sensors 11th Int. Conf. on Solid-State Sensors and Actuators vol 11 (Piscataway, NJ: IEEE) pp 186–9 [12] Iliescu C, Miao J and Tay F E H 2003 Chip scale packaging in glass of pressure sensors Micro System Technologies (Poing, Germany: Franzis Verlag GmbH) pp 110–6 [13] Hierold C 2003 Micro-and nanosystems: review and outlook MicroMechanics Europe Workshop (Delft, The Netherlands: Delft University of Technology) pp 101–4 [14] Receveur R A M, Marxer C, Duport F, Woering R, Larik V and De Rooij N F 2004 A laterally moving bi-stable MEMS DC switch for biomedical applications Int. Conf. on Micro Electro Mechanical Systems (Piscataway, NJ: IEEE) pp 854–6 [15] Hiller K, Kurth S, Neumann N, Hahn R and Kaufmann C 2003 Application of low temperature direct bonding in optical devices and integrated systems Micro System Technologies (Poing, Germany: Franzis Verlag GmbH) pp 102–9 [16] Wild M J, Gillner A and Poprawe R 2001 Locally selective bonding of silicon and glass with laser Sensors Actuators A 93 63–9 [17] Ziaie B, Von Arx A, Dokmeci M R and Najafi K 1996 A hermetic glass-silicon micropackage with high-density on-chip feedthroughs for sensors and actuators IEEE J. Solid-State Circuits 5 166–79 [18] Cao Z, Chen H, Xue J and Wang Y 2005 Evaluation of mechanical quality of field-assisted diffusion bonding by ultrasonic nondestructive method Sensors Actuators A 118 44–8 [19] Tilmans H A C, Van De, Peer M and Beyne E 2000 The indent reflow sealing (IRS) technique—a method for the fabrication of sealed cavities for MEMS devices J. Microelectromech. Syst. 9 206–17 [20] Jourdain A, De Moor P, Pamidighantam S and Tilmans H A C 2002 Investigation of the hermeticity of BCB-sealed cavities for housing (RF-) MEMS devices Int. Conf. on Micro Electro Mechanical Systems (Piscataway, NJ: IEEE) pp 677–80 [21] Jourdain A, Ziad H, De Moor P and Tilmans H A C 2003 Wafer-scale 0-level packaging of (RF-) MEMS devices using BCB Design, Test, Integration and Packaging of MEMS/MOEMS (Grenoble, France: CNRS, TIMA Laboratory) pp 239–44 [22] Information Society Technologies Programme 2004 Long-term stability of vacuum-encapsulated MEMS devices using eutectic wafer bonding (VABOND) Information Society Technologies, http://www.cordis.lu/ist/[23] MEMSCAP 2002 MetalMUMP: plated metal multi user MEMS processes www.memscap.com [24] Sarajlic E, De Boer M J, Jansen H V, Arnal N, Puech M, Krijnen G and Elwenspoek M 2003 Integration of trench isolation technology and plasma release for advanced MEMS design on standard silicon wafers MicroMechanics Europe Workshop (Delft, The Netherlands: Delft University of Technology) pp 123–6 [25] Borwick R L, Stupar P A and DeNatale J 2003 A hybrid approach to low voltage MEMS switches Int. Conf. on Solid State Actuators and Microsystems (Piscataway, NJ: IEEE) pp 859–62 [26] Jakobsen H, Lapadatu A and Kittilsland G 2001 Anodic bonding for MEMS Symp. on Semiconductor Wafer Bonding (Pennington, NJ: The Electrochemical Society) [27] Li X, Abe T, Liu Y and Esashi M 2002 Fabrication of high-density electrical feed-throughs by deep-reactive-ion etching of Pyrex glass J. Microelectromech. Syst. 11 625–30 [28] Corning 2004 Glass silicon constraint substrates www.slam.corning.com [29] Maluf N 2000 An Introduction to Microelectromechanical Systems Engineering (Boston, MA: Artech House) 683