US20020096421A1 - MEMS device with integral packaging - Google Patents

Abstract

A MEMS device and method of making same is disclosed. In one embodiment, a micro-switch includes a base assembly comprising a movable structure bearing a contact pad. The base assembly is wafer-scale bonded to a lid assembly comprising an activator and a signal path. The movable structure moves within a sealed cavity formed during the bonding process. The signal path includes an input line and an output line separated by a gap, which prevents signals from propagating through the micro-switch when the switch is deactivated. In operation, a signal is launched into the signal path. When the micro-switch is activated, a force is established by the actuator, which pulls a portion of the movable structure upwards towards the gap in the signal path, until the contact pad bridges the gap between the input line and output line, allowing the signal to propagate through the micro-switch. Prior to bonding, the MEMS structures are annealed on a first wafer and the conductive traces and other metals are annealed on a second wafer to allow each wafer to be processed separately using different processes, e.g., different annealing temperatures.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS
• This application claims the benefit of U.S. Provisional Application No. 60/253,851, filed Nov. 29, 2000, which is incorporated by reference herein.[0001]
BACKGROUND
• 1. Field of the Invention[0002]
• The present invention relates to the packaging of semiconductor devices and, more particularly, to the use of wafer-scale bonding techniques in the packaging of microelectromechanical systems (MEMS).[0003]
• 2. Background[0004]
• MEMS devices have been manufactured in the past using integrated circuit (IC) fabrication techniques. Although MEMS devices are generally fabricated on silicon wafers, MEMS devices are different from most ICs in that they possess moving elements. These moving elements make MEMS devices vulnerable to contamination and moisture because they cannot be encapsulated in a typical injection-molded plastic package, i.e., at least one element of every MEMS device must remain free to move.[0005]
• Several specialized methods, which are directed to different stages of the fabrication process, have been employed for the sealing and packaging of MEMS. One early method directed to the final packaging stage employs a ceramic package with a co-fired lead frame, a cavity to contain a complete MEMS device and a lid to cover the package and cavity. With this chip-scale packaging method, the MEMS device is placed inside the cavity and the lid is soldered in place to form a hermetic seal. One problem with this chip-scale packaging approach is the expense associated with handling the individual components (e.g., the packages, lids, etc.) and assembling them into a single package. An additional problem with this chip-scale packaging approach is that the moving elements in MEMS are extremely fragile and tend to be damaged by any handling, especially near the end of the fabrication process. Therefore, packaging yields are typically low. A further problem with this packaging approach is that the solder seal on the lid tends to outgas when re-flowed during the final sealing process. If a vacuum is needed inside the package, a getter must be added, at additional cost. In such cases, packaging costs can constitute 80% of the total device cost. In sum, the use of ceramic packages to seal and package MEMS devices is inordinately expensive due to the need to individually handle components, the need to use getters and the considerable opportunity for damage and contamination of the MEMS devices.[0006]
• A wafer-scale pre-packaging technique of bonding two wafers together has been used to bond a wafer of silicon or glass onto a MEMS wafer. In this back-end process, a non-conductive glass frit is screen-printed onto a lid wafer in the shape of rings or gaskets. The wafers are then placed together to form cavities containing MEMS devices and bonded by the application of heat in excess of 400° C. After bonding, the two-wafer stack is diced and the resulting hybrid structures are packaged in the standard manner: by encapsulation in plastic packages such as DIPs, SOICs, etc. Although this process has been marginally successful, it suffers from the deficiency that the glass frit cannot be patterned into fine features. Typically, gasket pattern line widths are between 100-200 microns. Given that MEMS are currently made with dimensions or features in the one-micron size range, significant real estate must be sacrificed, thereby significantly reducing the number of devices that can be fabricated on a single wafer. Attempts have been made to improve the glass frit patterning resolution, using lithographic methods, but these have not shown success. Further, gasket patterns thinner than 100-200 microns in width do not appear to provide reliable hermetic seals, possibly because of the inherently porous nature of glass frits. Thus, it is currently believed that glass frit techniques simply cannot be applied on a smaller scale where hermetic seals are desired.[0007]
• Glass frit techniques also require high temperatures of between about 450-600 Celsius for sealing to occur. One drawback of using such a high firing temperatures is that it effectively precludes the incorporation of integrated circuitry on either of the wafers bonded together. These temperatures also preclude the use of most anti-stiction coatings and hydrophobic coatings, which are desirable for improving manufacturing yields and the MEMS's tolerance to moisture and shock. Such high temperatures also damage the MEMS themselves and, particularly, certain structural films deposited at lower temperatures and structures in which the film stress must be carefully controlled.[0008]
• Another deficiency of glass frit techniques is that the glass frit itself employs organic binders that outgas into the sealed cavity during firing. Since getters are not available to combat this problem, this fact precludes the use of glass frits for vacuum applications. Still another limitation of the glass frit techniques is that the sealing materials, being non-conductive, cannot be used to establish electrical coupling between the devices disposed on the bonded substrates.[0009]
• Front-end wafer-scale bonding techniques which can be used for sealing MEMS have also been developed, including anodic bonding, silicon fusion bonding, and other wafer-bonding methods employing combinations of silicon, silicon dioxide, and silicon nitride. These techniques can provide vacuum and hermetic seals and can do so with improved use of device real estate. These techniques, however, have low throughput and require high temperatures for high-quality bonds because their bonding mechanisms involve solid diffusion across the bond interface. Additionally, these techniques place strict constraints on the materials and processes used to form the wafers to be bonded together. Wafer-scale bonding techniques are exceptionally sensitive to the presence of small particulate contaminants, wafer warpage, scratches, and other imperfections commonly encountered during semiconductor fabrication because such contaminants can cause large defects in the bond and reduce yield substantially. Moreover, normal variations in surface topology due to the presence of circuitry can preclude proper bonding and/or sealing using such techniques. All of these factors make it difficult to employ wafer-bonding techniques with wafers that have undergone a significant amount of processing (e.g., wafers bearing finished microstructures).[0010]
• One particular MEMS device, the fabrication of which is disclosed herein, is a micro-switch. The micro-switch can be used in, for example, mobile phones for switching RF signals between transmit and receive modes and other related functions. Additional applications include micro-relays, which can be used in, for example, automated test equipment (ATE). As will be described below with respect to FIGS. 1[0011]aand1b, fabricating a hermetically sealed micro-switch using materials having good mechanical qualities is difficult to achieve using prior art front-end, wafer-scale bonding techniques.
• Referring to FIGS. 1[0012]aand2b, there are shown cross-sectional views of a priorart MEMS device100 configured to operate as a micro-switch (hereinafter also referred to as “micro-switch100”). FIG. 1ais a cross-sectional view of the micro-switch100 in its “deactivated” state (i.e., noelectric field118 present) and FIG. 1bis a cross-sectional view of the micro-switch100 in its “activated” state (i.e.,electric field118 present).
• The micro-switch[0013]100 includessubstrate102,pedestal104,actuator106,cantilever108,contact pad110, and signal path112. Thesubstrate102 is typically made of silicon and provides a support base for thepedestal102 and thecantilever108. Thepedestal102 and thecantilever108 are formed in different layers upon thesubstrate102 using known IC fabrication processes (e.g., CMOS). Thepedestal104 is disposed on thesubstrate102. It provides support for thecantilever108 and also provides a sidewall114, which together with thecantilever108 and thesubstrate102 defines agap116. Thegap116 is typically very small (e.g., on the order of 2 microns in the vertical dimension). It provides thecantilever108 with at least one degree of freedom.
• The[0014]contact pad110 is disposed on a free end (distal end) of thecantilever108 and is typically made of metal. The location of thecontact pad110 on thecantilever108 is selected to ensure that it makes electrical contact with the signal path112, when theactuator106 is activated and thecantilever108 is pulled downward towards thesubstrate102 in response to an electric ormagnetic field118, as shown in FIG. 1b.
• The signal path[0015]112 is disposed on thesubstrate100 in aligned opposition with thecontact pad110. It is typically a metal trace or strip, which serves as a path for signals propagating through themicro-switch100. The signal path112 includes a gap (not shown) in a location corresponding to the location of thecontact pad110 on thecantilever108. If theactuator106 is inactive, a signal launched into the signal path112 cannot propagate through themicro-switch100 due to the gap.
• The[0016]actuator106 is disposed on thesubstrate100. It establishes a field118 (shown in FIG. 1b) when supplied with a drive voltage. Thefield118 provides a force that pulls thecantilever108 downward toward thesubstrate102, until thecontact pad110 bridges the gap in the signal path112. In this “activated” state, a signal launched into the signal path112 will propagate through themicro-switch100 due to the bridgedgap116 provided by thecontact pad110. FIG. 1bshows thecantilever108 pulled into electrical contact with the signal path112, thereby bridging the gap to allow signals to propagate through themicro-switch100.
• It would be desirable to fabricate the[0017]prior art cantilever108 from materials that exhibit good mechanical qualities. One example of such material is polycrystalline silicon (“polysilicon’), which has been extensively studied and is known to be a desirable material for fabricating micromechanical devices. Polysilicon, however, is not generally used to fabricate prior art switches (e.g., micro-switch100), because thecantilever108 is designed to lie over thecontact pad110, and therefore must be deposited on thesubstrate102 after thecontact pad110 in the fabrication process flow. The high temperature of the polysilicon anneal (e.g., above 400° C.) may cause thecontact pad110 to melt, delaminate, or interdiffuse with other material layers, thus making polysilicon an unsuitable material for thecantilever108.
• This problem was addressed in the past by using low temperature deposited materials, such as PECVD silicon nitride. The disadvantage with this prior art technique is that a high temperature anneal is typically needed to control warping of the[0018]cantilever108 due to stress gradients. The degree of warping (curling of thecantilever108 up or down with respect to the substrate102) is not easily controlled and leads to wide variations in the voltage required to actuate thecantilever108. This is typically not acceptable in a mass-produced device.
• Accordingly, there is a need for a novel wafer-scale bonding technique that can employ materials having desirable mechanical qualities. Such a technique should enable massive, parallel assembly of hermetically sealed, low-cost MEMS devices, and thus should provide high yields. To realize these advantages, the technique should not require extreme wafer planarity or cleanliness. Additionally, there is a need for an improved micro-switch/relay device, fabricated using the novel wafer-scale bonding technique from materials having desirable mechanical qualities, to provide improved performance and reliability over prior art MEMS devices.[0019]
SUMMARY
• The present invention eliminates the foregoing difficulties through the use of a novel wafer-scale bonding technique that enables massively parallel assembly of sealed, low-cost, reliable MEMS devices using materials having desirable mechanical qualities, which technique does not require extreme wafer planarity or cleanliness.[0020]
• In one embodiment of the present invention, a micro-switch includes a base assembly having a movable structure bearing a contact pad, and a lid assembly having an actuator and a signal path. A plurality of lid assemblies are fabricated in parallel on a first wafer and a plurality of base assemblies are fabricated in parallel on a second wafer, thus allowing each wafer to be processed separately using different processes (e.g., different annealing temperatures). The first wafer is then bonded to the second wafer causing each lid/base assembly pair to form a sealed cavity. The sealed cavity is adapted to hold the movable structure, which structure can move in at least one degree of freedom within the sealed cavity.[0021]
• In one embodiment of the present invention, the signal path in the lid assembly includes an input line and an output line separated by a gap, which prevents signals from propagating through the micro-switch. In operation, a force is established by the actuator, which pulls a portion of the movable structure upwards towards the gap in the signal path, until the contact pad bridges the gap between the input line and output line, allowing signals to propagate through the micro-switch.[0022]
• In one embodiment of the present invention, the lid assembly includes an insulation ring disposed on the first substrate and a metal seal ring substantially aligned with and overlying the insulation ring, such that the metal seal ring forms a seal around the movable structure and the insulating ring electrically insulates the metal seal ring from the signal path. The signal path runs under the insulation ring to connect with an external contact point, such as a wire bond pad.[0023]
• An additional advantage of the present invention is that the improved planarity of the movable structure leads to improved planarity of the contact pad. This, in turn, is believed to enable longer lifetime of the micro-switch.[0024]
• A further advantage of the present invention is that the bonding enables hermetic sealing. Hermetic sealing has been shown to significantly improve device lifetime.[0025]
• Another advantage of the present invention is the use of high-temperature materials in the movable structure, allowing the use of higher temperatures in the metal deposition steps that form the contacting surfaces. As a general principle, materials that are formed or processed at higher temperatures display greater reliability and wear resistance. These are critical factors in a micro-switch, because of the metal-to-metal contact, which may be repeated up to about 10[0026]¹²times in the life of the device.
• A further advantage of the present invention is that a wider choice of materials is available for the metal contacts. This is because they are the last materials deposited on their respective wafers.[0027]
• An additional advantage is that the gap between the metal contacts can be controlled directly, by varying the thickness of the seal ring. This thickness can be controlled to within about 5-10%.[0028]
• Yet another advantage of the present invention is improved yield. The present invention enables all of the microstructures on a given wafer to be protected immediately after release. Thus, they are protected during dicing. In the past, MEMS devices were generally diced before packaging. Particulates and water present during the dicing process would damage devices, drastically lowering yield.[0029]
• A further advantage of the present invention is that very small electrical connections can be formed between the two substrates. This allows, for example, high-frequency or noise-sensitive signals to be coupled between them with a minimum of attenuation or degradation. A serious challenge in the design of MEMS, for example, is coupling signals from a mechanical microstructure to an integrated circuit, when the two are formed on separate substrates. Wire bond pads, or solder bumps, with a typical size of 100×100 um, comprise a significant parasitic capacitance. The interconnection approach of the present invention, for example, using metal micro-bumps in the 10×10 um range, enable a hundred-fold reduction of parasitic capacitance.[0030]
BRIEF DESCRIPTION OF THE DRAWINGS
• These and other features, aspects, and advantages of the present invention will become better understood with regard to the following description, appended claims, and accompanying drawings where:[0031]
• FIG. 1[0032]ais a cross-sectional view of aprior art micro-switch100 shown in its “deactivated” state;
• FIG. 1[0033]bis a cross-sectional view of theprior art micro-switch100 shown in FIG. 1ain its “activated” state;
• FIG. 2[0034]ais a cross-sectional view of alid assembly200 of a micro-switch400, fabricated in accordance with the present invention;
• FIG. 2[0035]bis a cross-sectional view of abase assembly220 of a micro-switch400, fabricated in accordance with the present invention;
• FIG. 3[0036]ais a top plan view of thelid assembly200 shown in FIG. 2a, fabricated in accordance with the present invention;
• FIG. 3[0037]bis a top plan view of thebase assembly220 shown in FIG. 2b, fabricated in accordance with the present invention;
• FIG. 4[0038]ais a cross-sectional view of an assembledmicro-switch400, fabricated in accordance with the present invention;
• FIG. 4[0039]bis a cross-sectional view of the micro-switch400 shown in FIG. 4ain its “activated” state;
• FIG. 5 is a cross-sectional view of the micro-switch[0040]400 shown in FIG. 4a, including dielectric spacers;
• FIG. 6[0041]ais a cross-sectional view of the micro-switch400 shown in FIG. 4a, including upper and lowermechanical stops317,328;
• FIG. 6[0042]bis a cross-sectional view of the micro-switch400 shown in FIG. 6aafter final bonding;
• FIG. 7 is a flow diagram showing the process flow steps used in the fabrication of the[0043]lid assembly200 shown in FIG. 2a;
• FIGS. 8[0044]a-1 are cross-sectional views showing the process flow steps used in the fabrication of thelid assembly200 shown in FIG. 2a;
• FIG. 9 is flow diagrams showing the process flow steps used in the fabrication of the[0045]base assembly220 shown in FIG. 2b;
• FIGS. 10[0046]a-iare cross-sectional views showing the process flow steps used in the fabrication of thebase assembly220 shown in FIG. 2b; and
• FIGS. 11[0047]a-billustrate a wafer-scale bonding technique used to fabricate the micro-switch400 shown in FIG. 4a.
DETAILED DESCRIPTION OF EMBODIMENTS
• Although the MEMS device described below is a micro-switch, it should be understood by the reader that the present invention is not limited to a micro-switch, but also applies to other types of MEMS devices, including but not limited to optical switches, micro-relays, microdisplays, optical attenuators, variable optical attenuators, variable capacitors, phase shifters, surface acoustic wave devices, film bulk acoustic resonator (FBAR) devices, inductors, gyroscopes, accelerometers, pressure sensors, fluidic channels, chambers, resonators, strain sensors, rotary accelerometers, fiber optic aligners, optical micro-benches, scratch drives, and actuators. As used herein, the terms “micro-switch” and “micro-relay” are interchangeable.[0048]
• Although the description below refers to a single micro-switch, it should be understood that a plurality of such devices are typically fabricated (as opposed to a single device) simultaneously by bonding two wafers together in accordance with the present invention. Such techniques provide massive parallel assembly of sealed MEMS devices with high device yields.[0049]
• Lid Assembly[0050]
• Referring to FIG. 2[0051]a, there is shown a cross-sectional view of one embodiment of alid assembly200 for a micro-switch400 shown in FIGS. 4aand4b, which switch is fabricated in accordance with the present invention. Thelid assembly200 is fabricated by depositing layers of materials in a predetermined hierarchy using a process described with respect to FIG. 7. As will be described below, thelid assembly200 is bonded to a base assembly220 (FIG. 2b) to form a hermetically sealedmicro-switch400, as shown in FIGS. 4aand4b. While a hermetic seal is preferred, other seals can be employed with the present invention, including but not limited to vacuum seals and gross-leak seals. FIG. 3ashows a top plan view of thelid assembly200. The description of thelid assembly200 will refer to both FIGS. 2aand3a.
• In one embodiment of the present invention, the[0052]lid assembly200 includessubstrate202, insulatingfilm204, seal ring206 (316), signal path208 (300,302), actuator210 (310), anchor contact bump212 (312), and insulating ring214 (314). Thesubstrate202 can be made of any silicon, but is preferably made of high-resistivity silicon, glass or quartz and provides a common support structure for the microstructures inlid assembly200. Other materials having differing resistivities can be used for thesubstrate202, including but not limited to gallium arsenide, indium phosphide, indium gallium aresenide, sapphire, silicon-on-insulator, silicon-on-sapphire, polysilicon, amorphous silicon, plastic, or any combination of the these materials. Thesubstrate202 is coated with the insulatingfilm204, which is preferably made of silicon nitride, but other materials can be used, such as silicon oxide or ceramic. In this particular embodiment, the insulatingfilm204 acts as a diffusion barrier between thesubstrate202 and a plurality ofconductive traces300,302,304,306, as described below with respect to FIG. 3a.
• Referring to FIG. 3[0053]a, thesignal path208 preferably comprises aninput line300 and an output line302 separated by agap308, which serves as a barrier to signals (e.g., RF signals) transmitted on theinput line300. To improve performance of the micro-switch400, a co-planar waveguide (CPW) structure can be used as theinput line300 and the output line302. Although only onesignal path208 is shown in this embodiment, the present invention can employ one or more signal paths, each having one or more gaps. The one ormore signal paths208 can be disposed in thelid assembly200 or thebase assembly220 or both. Thesignal path208 is preferably made of metal, including but not limited to gold, titanium, titanium nitride, tantalum, tantalum nitride, copper, aluminum, tungsten, solders, metal amalgams, gallium, mercury, tin, lead, indium, palladium, thallium, bismuth, or any combination of these metals.
• Also shown in FIG. 3[0054]a, isactuator driveline304, which terminates in the actuator310 (210) andanchor ground line306, which terminates in the anchor contact bump312 (212). Theactuator driveline304 provides the actuator310 with an AC or DC actuation voltage in either positive or negative form. The actuation voltage causes theactuator310 to apply aforce402 on themovable structure230, as described with respect to FIG. 4b. Theanchor ground line306 provides an electrical ground for the actuation voltage. Preferably, theactuator310 is an electrode capable of establishing an electric field, but one or more actuators of various types may be used with the present invention, including but not limited to electrostatic, capacitive, electromagnetic, magnetic, scratch drive, membrane capacitive, thermal, electro-thermal, shape memory alloy, mechanical and any combination of the above actuators.
• The conductive traces[0055]300,302,304,306, are coupled to one or more wire bond pads (not shown) located outside the insulating ring314 (214) using wire bonding or tab bonding techniques. Preferably, the connections to the wire bond pads run under the insulating ring314 (214), so that they do not short themetal seal ring316, as shown in FIG. 3a. The insulating ring314 (214) overlies theconductive traces300,302,304,306, in a ring-shaped pattern. This pattern is substantially a mirror image of a seal-landing ring320, which is described with respect tobase assembly220. The insulating ring314 (214) pattern can be any closed curve, but is preferably a square pattern as shown in FIG. 3a. The insulating ring314 (214) is preferably formed of silicon dioxide, since it has a relatively low dielectric constant, but is thermally and mechanically robust. The insulating ring314 (214) is preferably about2 microns in thickness, to ensure adequate electrical isolation between theconductive traces300,302,304,306, and the seal ring314 (214). In this particular embodiment, the seal ring316 (206) overlies and is substantially aligned with the insulating ring314 (214), as shown in FIG. 3a. Alternatively, the insulation ring314 (214) and the seal ring316 (206) can be formed in thebase assembly220 instead of thelid assembly200.
• Although the seal-ring[0056]316 (206) is preferably made of metal with little or no surface oxide (e.g., a pure, soft, bright gold), other electrically conductive materials having differing resistivities can be used for the otherconductive elements300,302,304,306,310,312. These materials include, but are not limited to metals such as gold, titanium, titanium nitride, tantalum, tantalum nitride, copper, aluminum, tungsten, solders, metal amalgams, gallium, mercury, tin, lead, indium, palladium, thallium, bismuth, or any combination of these metals.
• Although the insulation ring[0057]314 (214) and the seal ring316 (206) are shown as part of thelid assembly200, the insulation ring314 (214) and the seal ring316 (206) can be part of thebase assembly220 or part of both the lid andbase assemblies200,220.
• Base Assembly[0058]
• Referring to FIG. 2[0059]b, there is shown a cross-sectional view of one embodiment of abase assembly220 for the micro-switch400, fabricated in accordance with the present invention. Thebase assembly220 is fabricated by depositing layers of materials in a predetermined hierarchy using a process described with respect to FIG. 8. As will be described below, thebase assembly220 is bonded with the lid assembly200 (FIG. 2a) to form the hermetically sealed micro-switch400 shown in FIGS. 4aand4b. FIG. 3bshows a top plan view of thebase assembly220. The description of thebase assembly220 will refer to both FIGS. 2band3b.
• In one particular embodiment of the present invention, the[0060]base assembly220 includessubstrate222, seal-landing ring224 (320), anchor contact226 (318),pedestal228, movable structure230 (326),isolation pad232, and contact pad234 (322). Thesubstrate222 is preferably formed of single-crystal silicon and provides a common support structure for the microstructures in thebase assembly220. Other materials can be used for thesubstrate222, including those materials described with respect tosubstrate202. Thepedestal228 overlies thesubstrate222 and is preferably made of silicon dioxide but other materials may be used, such as germanium or graphite. Thepedestal228 provides support for themovable structure230 and asidewall236, which together with themovable structure230 and thesubstrate222 defines agap238. Thegap238 is typically very small (e.g., on the order of about 2 um in the vertical dimension). It provides themovable structure230 with at least one degree of freedom within a sealedcavity404, as shown in FIG. 4b.
• Although the embodiment shown in FIG. 2[0061]bemploys a single cantilever as themovable structure230, one or moremovable structures230 can be employed, together with one or more signal paths, to create variety of MEMS switches for switching one or more signals, including but not limited to single pole single throw, single pole double throw, single pole multiple throw, and multiple pole multiple throw configurations. Such configurations may require one or more additional substrates and movable structures. Moreover, themovable structure230 need not be supported at one end by a pedestal, but can also be supported at several locations by one or more pedestals or no pedestals (e.g., fixed directly to substrate222). For example, thecantilever beam230 shown in FIG. 2acan be supported in its center by a single pedestal, which acts as a pivot for the cantilever beam to move in a “see-saw” fashion, each free end of the cantilever beam bearing one ormore contact pads234 for bridging one ormore gaps308 in one ormore signal paths208 located onsubstrate202 and/or andsubstrate222.
• The movable structure[0062]230 (326) is preferably formed of polysilicon and doped with a dopant so that is it conductive (e.g., phosphorus), but can also be formed of monocrystalline silicon. The anchor contact226 (318) preferably is deposited at the anchored end of the movable structure230 (326) and provides an electrical ground for the actuation voltage. The free end (distal end) of the movable structure230 (326) bearsisolation pad232, which is preferably composed of a silicon nitride layer, for isolating the actuation voltage from the signals. The contact pad234 (322) overlies theisolation pad232 and is used to bridge thegap308 between theinput line300 and the output line302. The location of the isolation pad232 (and therefore the contact pad234) on themovable structure230 is selected to ensure that the contact pad234 (322) bridges thegap308 when the actuator210 (310) is activated. Preferably, the shape and width of the contact pad234 (322) is selected to achieve this purpose.
• In this particular embodiment, the movable structure[0063]230 (326) lies under the contact pad234 (322), and therefore is deposited on thesubstrate222 before the contact pad234 (322) in the fabrication process flow, as illustrated in FIGS. 8A and 8B. This preferred configuration allows the polysilicon to be annealed at high temperatures (above 400° C.) for the movable structure230 (326), without causing the contact pad234 (322) to melt, delaminate, or interdiffuse with other material layers.
• The seal-landing ring[0064]224 (320) surrounds the movable structure230 (326) and is preferably about 1.5 microns thick in its vertical dimension and has a wall thickness in the range of about 5-20 microns. The seal landing ring224 (320), together with the seal ring206 (316) bonds thelid assembly200 to thebase assembly220, forming a hermetic seal around the movable structure230 (326). The movable structure230 (326) includes etch access holes324, which facilitate removal of material underlying the movable structure230 (326), as described with respect to FIGS. 8A and 8B.
• Assembled Micro-Switch[0065]
• Referring to FIG. 4[0066]a, there is shown cross-sectional views of one embodiment of an assembledmicro-switch400 fabricated in accordance with the present invention. Preferably, a plurality oflid assemblies200 are processed on a first wafer and a plurality ofbase assemblies220 are processed on a second, separate wafer. This preferred configuration enables the conductive traces and other metal elements to be processed on the first wafer using the appropriate annealing temperatures for such metals (e.g., below 300° C.) and the MEMS structures (e.g., movable structure230) to be processed on the second wafer using the appropriate annealing temperatures for structures (e.g., above 400° C. for polysilicon).
• Although the[0067]micro-switch400 is shown with a cantilever beam as themovable structure230, the present invention is not limited to this preferred configuration. For example, themovable structure230 can be any type of structure (e.g., T-shaped structure, membrane structure). Moreover, the micro-switch400 can be configured to operate in an electrically closed state when in its deactivated state (i.e., normally closed) or in an electrically opened state when in its deactivated state (i.e., normally open), as shown in FIGS. 4aand4b. Alternatively, themovable structure230 can include a signal path for transmitting signals from thelid assembly200 to thebase assembly220 or vice-versa depending on the configuration of themicro-switch400. The location of theactuator310 on thesubstrate202 can vary between the contact pad234 (322) and the anchor contact226 (318).
• Operation of Micro-Switch[0068]
• Referring to FIG. 4[0069]b, there is shown the assembledmicro-switch400 in its “activated” state. In operation, an RF signal is launched into theinput line300 of themicro-switch400. Because of thegap308, the signal cannot reach the output line302. If an actuation voltage is applied to the actuator210 (310), aforce402 is established, which pulls themovable structure230 upward towards the substrate202 (i.e., themovable structure230 flexes upwards), until the contact pad234 (322) bridges thegap308 between theinput line300 and the output line302, thereby allowing the signal to propagate through themicro-switch400.
• Applications for the assembled[0070]micro-switch400 include but are not limited to, mobile phones, portable one-way and two-way radios, satellite phones, base stations, mobile Internet access devices, fixed location Internet access devices, radar collision sensing, Bluetooth compliant devices, wireless networks, or any combination of these applications.
• Alternate Layouts[0071]
• Referring to FIG. 5 (with reference to FIGS. 1[0072]aand4b), there is shown a cross-sectional view of the micro-switch400, includingdielectric spacers500 located between theisolation pad232 and theanchor contact226. In the past, a contiguous dielectric layer (e.g., silicon nitride) was often used as an isolation layer between thecantilever108 and theactuator210. The two-substrate process flow described with respect to FIGS. 7 and 8, enables such dielectric layer (e.g., the same dielectric layer used to pattern the isolation pad232) to be patterned and etched into one or more discrete islands, hereinafter referred to asdielectric spacers500. This is advantageous because it reduces the total area of dielectric material. Dielectric materials tend to accumulate charge over time and through use. This accumulated charge can cause a shift in the required actuation voltage or sticking of themovable structure230. Reducing the area of dielectric material reduces these negative effects due to the accumulated charge.
• Referring to FIGS. 6[0073]aand6b, there are shown cross-sectional views of themicro-switch600. Themicro-switch600 is similar to themicro-switch400, but includes uppermechanical stops317 and lowermechanical stops328. FIG. 6bshows the micro-switch600 after the final bonding step, as described below with respect to FIGS. 7 and 8. The uppermechanical stops317 are part of thelid assembly200 and the lowermechanical stops328 are part of thebase assembly220, as show in FIGS. 3aand3b, respectively. The mechanical stops317,328, are used to provide a consistent gap between themovable structure230 and thesubstrate202. Thesemechanical stops317,328, are preferably comprised of the same layers used to fabricate themicro-switch400, as shown in FIG. 6a.
• Integration With Other Devices[0074]
• One or more micro-switches can be integrated with one or more passive and active devices, including but not limited to capacitors, inductors, filters, low-noise amplifiers, power amplifiers and phase shifters. For example, one or more micro-switches can be fabricated as described above with at least one passive component (e.g., one or more capacitors, inductors and/or phase shifters) or at least one active component (e.g., one or more power amplifiers), fabricated on at least one of the two[0075]substrates202,222, with at least one electrical interconnection (e.g., via conductive layers, wire bonding, tab bonding) connecting at least one of the fabricated components (e.g., anchor contact226) with one or more passive components. Alternatively, the passive components do not need to be fabricated on either of the twosubstrates202,222, but can be electrically interconnected to one or more micro-switches (e.g., via conductive layers).
• Process Flow for Lid Assembly[0076]
• Referring to FIG. 7 (with reference to FIGS. 8[0077]a-1 showing corresponding cross-sectional views of the process flow), there is shown a flow diagram of one embodiment of a process flow for fabricating a lid assembly (e.g., lid assembly200). The process starts700 with a glass, quartz, or high-resistivity silicon substrate (e.g., substrate202). The substrate is coated702 with an insulating film (e.g., insulating film204) comprising about 1 micron of silicon dioxide (oxide 1.5). Deposited704 on the insulating film is about 1000 A of sputtered TiW, about 1-2 um of gold and about 1000 A of TiW. The TiW acts as a diffusion barrier, and also provides adhesion between the gold and the adjoining layers above and below. Hereinafter, this tri-layer is also referred to as metal2.
• Metal[0078]2 is wet etched706 to define conductive traces (e.g.,input line300, output line302) and wire bond pads. About 2 microns of silicon dioxide (ox2) is deposited708 over the conductive traces. Ox2 is patterned710 to form a insulating ring (e.g., insulating ring214). Step710 can be performed using a wet etchant, such as 5:1 BOE, or a dry etch, such as CH4/CHF3/He. The insulating ring can have any desired width, but is preferably in the range of about 20 to about 50 microns.
• Next, the wafer is coated[0079]712 with a seed layer for electroplating. The seed layer preferably consists of about 1000A of TiW, followed by about 1000A of Au, preferably sputter deposited. A stencil mask is formed714, preferably using a planarizing resist such as SJR5740. The stencil mask is formed714 with an opening in the shape of a ring, nominally overlying the insulating ring. An opening in the shape of a small square, roughly 10×10 um in size, is also provided, nominally overlying theanchor contact226. The resist is exposed and developed716 to prepare the surface for plating. A deformable metal is plated718 to a thickness of about 8 microns. This is preferably a pure, soft, bright gold. Hereinafter, this layer is also referred to as metal3. Finally, the resist is stripped720 and the exposed seed layer is stripped722 to finish the process flow.
• Process Flow for Base Assembly[0080]
• Referring to FIG. 9 (with reference to corresponding FIGS. 10[0081]a-ishowing cross-sectional views of the process flow), there is shown a flow diagram of one embodiment of a process flow for fabricating a base assembly (e.g., base assembly220). The process starts900 with a silicon substrate (e.g., substrate222). The substrate is coated902 with about 2-3 microns of silicon dioxide (ox1). The silicon dioxide is then doped904 with about 2 microns of polycrystalline silicon (poly1), either in-situ, or from symmetrical (upper and lower) layers of dopant source material, such as thin layers of doped polysilicon. The poly1 layer, together with the ox1 layer, provide an intermediate structure from which a cantilever (e.g., movable structure230) will be formed in later process flow steps. The polysilicon is annealed for about 60 minutes at about 1000° C., in nitrogen, to distribute the dopant and reduce stress. About a 60 second RTA is performed in nitrogen to further reduce stress. About 0.5 microns of silicon nitride (nitride1; preferably low-stress) is deposited906 over the polysilicon. The nitride2 layer is patterned908 to form an insulating pad (e.g., insulating pad232) at what will become the free end (distal end) of a cantilever. Additionally, one or more small pads (e.g., dielectric spacers500) of nitride2 can be distributed over the area of the movable structure, to act as insulating spacers and mechanical stops (e.g., mechanical stops328) between the cantilever and the actuator, as shown in FIG. 5. The poly layer is patterned910 to define the cantilever. Preferably, one or more etch access holes (e.g., etch access holes324) are provided in the cantilever, except in the area that is to be supported by an anchor pedestal (e.g., pedestal228). The wafer is preferably sputter coated912 with about 1000 A of TiW, followed by about 1 um of Au. Hereinafter, this layer is also referred to as metal1.
• The metal[0082]1 layer is etched912, preferably using a wet etchant such as KI, to define a contact pad (e.g., contact pad234), the cantilever's ground contact (e.g., anchor contact226), and a metal seal ring (e.g., seal ring206). Preferably, a timed wet etch of ox1 is performed in HF. The etch access holes allow the HF to undercut the main body of the cantilever. The wafer is removed from the etch and rinsed in DI water, before the pedestal area is etched916 away. This results in the finished cantilever.
• Bonding Procedure[0083]
• One embodiment of a bonding procedure is described below in accordance with the present invention. Other bonding processes, however, may be employed with the present invention, including but not limited to, use of different temperatures, different pressures, different ambients, ultrasonic agitations, megasonic agitation, surface cleaning or any combination of the above techniques.[0084]
• Referring to FIGS. 11[0085]a-c,there is shown a wafer-scale bonding technique in accordance with the present invention. The first wafer (the wafer bearing the lid assemblies) and the second wafer (bearing the base assemblies) are aligned1100 and tack bonded1102, using, for example, a Karl Suss MA/BA-6 bond aligner and SB-6 wafer bonder. Preferably, a gold thermocompression recipe is used for tack bonding, e.g. about 300° C. in inert gas or vacuum, approximately a 1 minute bond time, at full pressure (about 2-3 bar). To achieve afinal bond1104, the wafers are preferably transferred to a hydraulic press apparatus, where approximately about 100-200 bar is applied, at a temperature of about 270°-300° C., for about 10-20 minutes. This is a relatively long “soak,” which enables the gold to plastically deform and create a hermetic seal where the two rings bond (e.g.,seal ring206 and seal-landing ring224). The long soak time also enables a suitably large degree of plastic deformation without excessive applied pressure. If a shorter soak time is used, higher pressures are required, which leads to cracking of the wafers. The wafers are preferably constrained to prevent lateral shearing or shifting of one wafer relative to the other, while this higher level of pressure is applied. This may be achieved by placing the wafer between two metal plates, of diameter larger than the wafers'. The two plates are bolted or otherwise secured together at a number of points around their perimeters. Spacers are used to create a gap between the plates, equal to the combined thickness of the two wafers, plus about 25-100 um. The thickness and diameter of the plates is chosen so that they can flex to allow compression of the wafers, coming into contact against the wafers with relatively little applied pressure (e.g. about 10% of the full applied pressure of about 200 bar). In this way, the two metal plates act to prevent lateral motion of the wafers during the compression step. The soak time is preferably adjusted to achieve a compression of about 50% in the electroplated gold layer. This leads to a gap of about 2 um between the contact pad (e.g., contact pad234) and the signal path (e.g., signal path208).
• An advantage of the techniques described above is that each lid/base assembly pair is individually bonded together to form a sealed[0086]cavity404 adapted to house the movable structure230 (FIG. 4a) so that it can move freely under the influence of theactuator210. The sealedcavity404 can significantly increase the lifetime of the micro-switch400 by sealing off the microstructures from harmful contaminants in the environment.
• Although the bonding procedure described above employs wafer-scale bonding techniques, the present invention is equally applicable to die-scale bonding techniques, including the bonding of any subset of devices on a wafer or other substrate (e.g., n×m array of devices).[0087]
• Dicing Procedure[0088]
• Once the[0089]micro-switches400 are assembled, a dicing saw is used to expose the bond pads patterned on the first wafer (the wafer bearing the lid assemblies). This is achieved by making cuts in the second wafer (the wafer bearing the base assemblies). Themicro-switches400 are then singulated by cutting with the dicing saw completely through the bonded wafers. Themicro-switches400 are protected during dicing by their respective hermetic seals during the wafer-scale bonding step. In the past, MEMS devices were generally diced before packaging. Particulates and water present during the dicing process would damage devices, drastically lowering yield.
• Second-level Packaging[0090]
• The singulated devices are packaged in a conventional manner using standard IC chip packaging techniques. Preferably, this is done using a surface-mount plastic package, or a wafer-scale package, such as used for stacked-chip devices such as DRAMs.[0091]
• The above description is included to illustrate the operation of the preferred embodiments and is not meant to limit the scope of the invention. The scope of the invention is to be limited only by the following claims. From the above discussion, many variations will be apparent to one skilled in the relevant art that would yet be encompassed by the spirit and scope of the invention. For example, to ensure that at least about a 2 um gap between the contact pad and signal path is achieved in a repeatable, uniform manner, a two-layer structure may be substituted for the electroplated gold ring. In this case, the first layer would be formed of nickel or other relatively hard material. Nickel may be conveniently electroplated in this case. Subsequently, gold-tin alloy or other solder material would be electroplated on top of the nickel. The nickel layer here serves as a well-defined hard stop, determining the final thickness of the plated layer. In the bonding step, the gold-tin would melt and reflow, sealing the ring structure and forming electrical contacts, but taking on a negligible thickness. As an alternative to gold-tin, pure gold could be used. With adequate pressure and soak time, it may be expected that the gold layer would be compressed to negligible thickness. An additional alternative embodiment might employ polysilicon or other ceramic material as the mechanical stop. This material could be patterned to lie outside the active area of the device, but could conceivably occupy a larger surface area than the nickel in the embodiment above. The larger surface area would lead to reduced stress, and more reliable establishment of the desired vertical gap during the bond process.[0092]

Claims (22)

What is claimed is:

1. A micro-switch, comprising:

a first substrate having a signal path, the signal path including a gap portion for preventing signal propagation; and

a second substrate bonded to the first substrate to form a sealed cavity, the sealed cavity adapted to house a movable structure, the movable structure having a contact pad for bridging the gap portion of the signal path when the movable structure is pulled toward the first substrate in response to a force provided by an actuator.

2. The micro-switch ofclaim 1, further including:

a seal ring, forming a seal around the movable structure.

3. The micro-switch ofclaim 1, further including:

an insulation ring disposed on the first substrate; and

a metal seal ring substantially aligned with and overlying the insulation ring, the metal seal ring forming a seal around the movable structure, the insulating ring electrically insulating the metal seal ring from the signal path.

4. The micro-switch ofclaim 1, wherein the movable structure is made of polysilicon.

5. The micro-switch ofclaim 1, wherein the movable structure is a cantilever beam.

6. The micro-switch ofclaim 1, further comprising:

a passive element disposed on one of the first and second substrates and coupled to the signal path.

7. The micro-switch ofclaim 1, further comprising:

an active element disposed on one of the first and second substrates and coupled to the signal path.

8. The micro-switch inclaim 1, further comprising:

a plurality of spacers formed in an insulating layer disposed between the contact pad and the movable structure.

9. The micro-switch ofclaim 1, further comprising:

a passive element disposed on the first substrate and electrically coupled to the signal path.

10. The micro-switch ofclaim 1, wherein at least one of the first and second substrates includes a mechanical stop.

11. The micro-switch ofclaim 1, further including an isolation pad disposed between the contact pad and the movable structure.

12. The micro-switch ofclaim 11, wherein the isolation pad is made of silicon nitride.

13. The micro-switch ofclaim 2, wherein the seal ring is made of a deformable metal.

14. The micro-switch ofclaim 2, wherein the deformable metal comprises gold.

15. The micro-switch ofclaim 2, further comprising:

a seal-landing ring disposed on the second substrate, the seal-landing ring for mating with the metal seal ring to form a hermetic seal.

16. The micro-switch ofclaim 2, wherein the signal path runs under the insulation ring to connect to an external contact point outside the insulation ring.

17. The micro-switch ofclaim 2, wherein the contact point is a wire bond pad.

18. A lid assembly for a micro-switch having a lid assembly and a base assembly, comprising:

a substrate;

a first insulating layer formed on the substrate;

a first conductive layer formed on the first insulating layer;

a signal path formed in the conductive layer and having a gap for preventing signals from propagating on the signal path;

an actuator formed in the first conductive layer for providing a force on a movable structure in the base assembly when the actuator is activated;

a second insulting layer formed on the first conductive layer;

an insulating ring formed in the second insulting layer;

a second conductive layer formed on the second insulating layer; and

a conductive ring formed in the second conductive layer, the conductive ring substantially aligned with and overlying the insulating ring, the conductive ring forming a seal around the movable structure when the lid assembly is bonded to the base assembly.

19. A base assembly for a micro-switch having a lid assembly and a base assembly, comprising:

a substrate;

a sacrificial layer formed on the substrate;

a structural layer formed on the sacrificial layer;

an insulating layer formed on the structural layer;

an insulating pad formed in the insulating layer;

a movable structure formed in the structural layer and the sacrificial layer, the movable structure bearing the insulating pad; and

a contact pad formed on the insulating pad, the contact pad for bridging a gap in a signal path on the lid assembly, when the movable structure is pulled towards the lid assembly in response to a force provided by an actuator in the lid assembly.

20. A method of making a lid assembly for a micro-switch having a lid assembly and a base assembly, the method comprising:

forming a first insulating layer on a substrate;

forming a first conductive layer on the first insulating layer;

forming a signal path and an actuator in the first conductive layer;

the signal path having a gap for preventing signals from propagating on the signal path, the actuator for providing a force on a movable structure in the base assembly when the actuator is activated;

forming a second insulating layer on the first conductive layer;

forming an insulating ring in the second insulating layer;

forming a second conductive layer on the second insulating layer; and

forming a conductive ring in the second conductive layer, the conductive ring substantially aligned with and overlying the insulating ring, the conductive ring for making a seal between the lid assembly and the base assembly.

21. A method of making a base assembly for a micro-switch having a lid assembly and a base assembly, the method comprising:

forming a sacrificial layer on a substrate;

forming a structural layer on the sacrificial layer;

forming an insulating layer on the structural layer;

forming an insulating pad in the insulating layer;

forming an intermediate movable structure in the structural layer and the sacrificial layer;

forming a conductive layer on the structural layer and sacrificial layer;

forming a contact pad on the conductive layer, wherein the contact pad overlies the intermediate structure; and

removing a portion of the intermediate structure to form a movable structure, the movable structure bearing the contact pad, the contact pad for bridging a gap in a signal path in the lid assembly when the movable structure is pulled towards the lid assembly in response to a force provided by an actuator in the lid assembly.

22. The lid assembly inclaim 18, wherein the signal path extends under the insulation ring to connect with a contact point external to the insulation ring.

Images