The Future of MEMS Dr. Thomas F. Marinis Draper Laboratory 555 Technology Square Cambridge, MA 02139 Abstract The future of MEMS is multifaceted, complex, and subject to change, in response to the prevailing winds of investment by government and commercial entities, somewhat like New England weather. Undaunted by these challenges, this author will present a forecast of the future from the perspectives of technology development, business growth and system integration. In many respects, MEMS technology development parallels that of solid state electronics. However, it lacks the definitive pedigree of that more extensive topic, which started with the invention of the transistor in 1947, followed by the integrated circuit in 1959. The closest analogy for MEMS was the 1954 discovery of the piezoelectric effect in silicon, which enabled strain in micromechanical structures to be measured. Subsequent developments in high gain, low noise amplifier technology made capacitance based sensors and actuators feasible. MEMS products based on piezoelectric and capacitance sensing now include pressure and flow sensors, accelerometers, gyroscopes, microphones, digital light projectors, oscillators, and RF switches. Development continues on a plethora of new devices, but significant effort is also being devoted to improving the performance and packaging of existing devices. MEMS based products produced in 2005 had a value of $8 billion, 40% of which was sensors. The balance was for products that included micromachined features, such as ink jet print heads, catheters, and RF IC chips with embedded inductors. Growth projections follow a hockey stick curve, with the value of products rising to $40 billion in 2015 and $200 billion in 2025! Growth to date has come from a combination of technology displacement, as exemplified by automotive pressure sensors and airbag accelerometers, and new products, such as miniaturized guidance systems for military applications and wireless tire pressure sensors. Much of the growth in MEMS business is expected to come from products that are in early stages of development or yet to be invented. Some of these devices include disposable chips for performing assays on blood and tissue samples, which are now performed in hospital laboratories, integrated optical switching and processing chips, and various RF communication and remote sensing products. The key to enabling the projected 25 fold growth in MEMS products is development of appropriate technologies for integrating multiple devices with electronics on a single chip. At present, there are two approaches to integrating MEMS devices with electronics. Either the MEMS device is fabricated in polysilicon, as part of the CMOS wafer fabrication sequence, or a discrete MEMS device is packaged with a separate ASIC chip. Neither of these approaches is entirely satisfactory, though, for building the high value, system-on-chip products that are envisioned. It is this author’s opinion that a combination of self-assembly techniques in conjunction with wafer stacking, offer a viable path to realizing ubiquitous, complex MEMS systems. Introduction Since its inception, the MEMS industry has burgeoned from a few product offerings to a preponderance of hundreds, which are produced today. MEMS devices are used in virtually all areas of industrial activity, health care, consumer products, construction, and military and space hardware. A vibrant, growing MEMS industry exists throughout the world’s industrialized countries. This diversification of product offerings and globalization of the industry has been accompanied by strong revenue growth and growing private investment. In 2004, the value of all MEMS devices shipped was $4.55 billion, which is projected to grow to $12.52 billion, by 2010. Today, profit from sales of MEMS products is beginning to displace government investment as the driving force for innovation and product development. This paper will attempt to present a perspective on the future development of MEMS by first summarizing the current state of the industry in terms of device applications in various markets and how revenues from those markets are projected to grow. Next, a sampling of the various types of MEMS devices will be presented to demonstrate the wide diversity of operating principles employed as well as the means and materials of their construction. Then, the technology for packaging MEMS devices and integrating them into useful systems will be reviewed. Finally, some likely developments in packaging and integration will be discussed. Markets for MEM Devices A report by J. Eloy, of Yole Development, breaks down revenues from various MEMS devices for years 2002 and 2005. A chart from that report is reproduced in Figure (1). Three products continue to dominate the industry in 2005, ink-jet print heads and pressure sensors at roughly $1.5 billion each, and digital light projectors at $1.0 billion. Between 2002 and 2005, the largest growth occurred in inertial instruments, accelerometers and gyroscopes, which together had revenues of $1.0 billion in 2005. 6000 5000 RF Bio & Fluidics 4000 MOEMS $ Millions Gyroscope 3000 Accelerometer 2000 Digital Light Proj Pressure 1000 Inkjet Head 0 2002 2005 Figure 1. MEMS Product Revenue as Compiled by Yole Development Most of these products were supplied by a handful of companies, as summarized in Table (1) and Figure (2). In 2004, there were 230 manufacturers of MEMS products. Most of these were developing new products, some of which are just starting to significantly impact the market. Analog Devices’ line of accelerometers is a notable example. Table 1. 2005 Sales of MEMS Products Company Sales $M Company Sales $M Texas Instruments 788 ST Microelectronics 200 Hewlett Packard 750 Seiko 199 Bosch 325 Cannon 184 Lexmark 230 Freescale 182 Between 2000 and 2004 the largest growth in MEMS applications occurred in the automotive market. Sales in this period climbed from $1.26 billion to $2.35 billion, while applications expanded to the point that a typical automobile contains between 25 and 70 MEMS devices. Table (2) and Figure (3) illustrate the various automotive applications of MEMS. Even though the quantity of MEMS devices sold into the automotive market is expected to continue climbing steadily, the automotive percentage of the total MEMS market is actually expected to decline, as consumer product applications expand as illustrated in Figure (4). Figure 2. Ranking of MEMS Manufacturers Figure 3. Distribution of MEMS Fabs Table 2. Automotive Applications for MEMS Devices Application Sensor Type Drive Train Torque Engine Timing Position Antilock Brakes Accelerometer, Position Engine Management Mass Air Flow, Temperature, Position, Pressure Automatic Headlight Sun / Light Seat Control Temperature, Load / Force Emissions Control Oxygen Transmission Position Air Conditioning Temperature, Humidity, Sun / Light Active Suspension Accelerometer, Position, Speed, Pressure, Load / Force Crash Protection Accelerometer Skid Control Accelerometer, Gyroscope Tire Inflation Pressure Figure 4. Automotive Applications for MEMS Devices Figure 5. Current and Predicted Composition of Markets for MEMS Products Demand for ink jet print heads is expected to remain strong, but relatively constant. The large growth in the consumer products area is expected from demand in three areas, large screen, high definition television, read write heads for high density storage systems, and mobile phone sets. The first two of these applications are essentially single products, digital light projectors and high precision positioning devices. Mobile phones are expected to incorporate an increasing number of MEMS devices, which is reflected in the chart of emerging applications, shown in Figure (6). The two most significant of these devices, in terms of dollar value, are clocks and microphones. The projected sales of MEMS components, into the cellular phone market, is shown in Figure (7), which is taken from a report prepared by Yole Development. The medical market for MEMS devices is expected to more than double from $660M in 2004 to $1.6B in 2009. Much of the current volume is in pressure sensors, such as shown in Figure (8), but many new applications are emerging such as blood analysis chips. There is considerable potential for revolutionary products in which MEMS sensors are integrated with highly miniaturized electronics for in-vivo remote monitoring. Two such examples are a capsule camera shown in Figure (9) and injectable capsules, for monitoring the health of livestock, shown in Figure (10). 6 Camera Lenses 5 Memories Finger Print 4 Pressure Sensors $ Billions RF Clocks 3 Microphones 2 Accelerometers Gyroscopes 1 RW Heads Microdisplays 0 2004 2009 Figure 6. Emerging Consumer Applications for MEMS Figure 7. Pressure Sensors for Medical Applications Figure 8. Wireless Capsul Endoscope Figure 9. Injectable Sensor Capsules for Temperature and Activity Monitoring in Animals As the various application markets develop, the growth in diversity of MEMS devices continues. The most significant classes of devices, in terms of current and projected dollar value, are listed in Table (3). It is reasonable to presume that the broad range of commercial processes and materials, which supports these product offerings, will facilitate development and rapid commercialization of entirely new products. Table 3. Forecasted Growth of MEMS Devices Application 2005 2015 Pressure Sensors $3.0B $6.0B In-Vitro Diagnostics $0.01B $5.0B Read / Write Heads $2.0B $4.0B Ink Jet Print Heads $2.0B $3.5B Optical Displays $1.0B $3.0B Gyroscopes $0.1B $2.0B Lab-On-a-Chip $0.01B $2.0B Drug Delivery Systems $0.0 $1.5B Inertial Sensors $0.2B $1.5B Chemical Sensors $0.1B $1.0B Optical Switches $0.1B $1.0B RF Devices $0.1B $1.0B Microspectrometers $0.02B $0.4B Survey of MEMS Devices The following is organized into four classes of devices, namely fluidics, optical, electrical and inertial. It is not intended to be a comprehensive survey, but rather to demonstrate the broad range of design innovation and application of MEM devices. Most of these devices are used as discrete packaged elements in systems that may include multiple MEMS components. Some applications, though, such as lab-on-a-chip and digital radio, require diverse sets of components that must be integrated at the microscale in order to meet performance requirements. Towards this end, packaging is becoming an integral part of design and fabrication of devices. Two noteworthy examples of this are the SiTime precision oscillator and Akustica’s digital microphone. Traditional packaging technology is also being utilized to provide MEMS scale integration as well as functionality. For example, microfluidic devices have been fabricated in low temperature co-fired ceramic technology and IR imaging bolometer arrays have been fabricated using Kapton films. Fluidic Devices Inkjet Print Head – represents an early application of MEMS technology, which quickly grew to become the most economically significant application because of its use in low cost computer printers. Its design has not changed significantly since its inception. It has evolved such that the number of nozzles in print heads has grown, while the pitch between them has decreased to improve the quality of print. The basic operation is depicted in Figure (10). Figure 10. Operation of a Single Print Head Nozzle Bulk Micromachined Pressure Sensor – represents the first important commercial device. It utilized the silicon piezoelectric effect to measure the deflection of the sense diaphragm. The basic device and its fabrication steps are illustrated in Figure (11). Figure 11. Fabrication of a Bulk Micromachined Pressure Gauge Blinking Bubble Pump – is shown schematically in Figure (12). It consists of a capillary tube that connects two reservoirs, and a set of heating wires in contact with the capillary, which are positioned close to the left-hand reservoir. The pump operates by rapidly heating the fluid inside the capillary until a bubble nucleates and grows to fill the heated region. Then the heater power is turned off, the bubble collapses, and the cycle is repeated at a rate on the order of 200 Hertz. Heater Wires PS PL LS LL Figure 12. Bubble Pump Fabricated as a capillary with Embedded Heater Pirani Vacuum Gauge – uses a heated wire, with a high temperature coefficient of resistance, to measure pressure from atmosphere down to 10-5 torr. The wire forms one leg of a Wheatstone bridge circuit. Gas molecules collide with the wire sense element, which transfer heat away from it. The power that is required to maintain the wire at a constant temperature and hence resistance, is proportional to the gas pressure. In addition to being easier to fabricate than a conventional Pirani gauge, the MEMS device has significantly better performance. Its small size eliminates convection heat transfer, which limits the accuracy of larger devices. A schematic representation of the MEMS device appears in Figure (13). Heater Si3N4/SiO2 Membrane Sensor Silicon Cap Silicon Base Figure 13. MEMS Pirani Vacuum Gauge Microchannel Resonant Mass Sensor – is intended for detecting biomolecular substances in liquids. It consists of a microchannel fabricated on a suspended cantilevered beam. The inside wall of the channel is treated to bond to the biomolecular substance of interest. An electrostatic drive causes the cantilever beam to oscillate in a vacuum at its resonant frequency. As biomolecular material accumulates in the microchannel, its mass increases, thus lowering the resonant frequency. A schematic illustration of this device appears in Figure (14). Inlet Microfluidic Channel Electrostatic Drive Connections Oscillating Cantilever Outlet Drive Electrode Figure 14. Vacuum-Packaged Suspended Microchannel Resonant Mass Sensor for Detecting Biomolecular Materials in Fluid Streams Enzyme Based Blood Analysis – sensor is constructed by first depositing silver and platinum electrode arrays on a silicon wafer as shown in Figure (15a). Then a layer of SU-8 photo-Definable Epoxy is spun on the wafer and processed to create arrays of microchannels that are connected by a pair of fill channels. Figure (15b) shows how the microchannels are aligned with the array of metal electrodes. Next, enzyme solutions are introduced into the fill channels. Capillary action draws the solution into the microchannels, where the platinum electrodes are selectively coated as shown in Figure (15c). A molded wafer of Silgard, is bonded on top of the SU-8, as shown in Figure (15d). Channels in the Silgard allow blood samples to be introduced into multiple sets of electrode pairs. The sensor is completed, by dicing it out of the wafer array. The device shown in Figure (15e) can perform multiple enzyme tests on one blood sample. After the blood sample is introduced, a bias potential is applied across the electrodes, while the current is monitored. Figure (15f), shows the results of a test to measure blood glucose concentration. a. Electrode array deposited on Si wafer b. SU-8 micro-channels built over arrays c. Pt electrodes coated with enzymes d. PDMS used to form blood channels f. Demonstration of blood glucose determination e. Blood analysis chip diced from wafer using this sensor Figure 15. Enzyme Based Blood Analysis Sensor Paper Handler – for printer and scanner applications utilizes a very large array of air jets, each of which is controlled by an electrically actuated mylar flapper valve. The arrangement of a single air jet and control valve is illustrated in Figure (16). The backside of the mylar film is metallized and connected to one side of a voltage source. The copper seat of the valve is connected to the other terminal of the voltage source. When a voltage is applied, the field draws the metallized mylar flapper down on the valve seat, closing off the air flow. When the voltage is turned off, air pressure pushes open the flapper valve, turning on the air jet. One advantage of this MEMS based paper handler is that it can readily adapt to papers of different size and weight, or odd shaped outlines. Figure 16. MEMS Based Paper Handling System Flow Sensor – is fabricated by stacking five layers of low temperature cofired ceramic as shown in Figure (17). The heart of the sensor consists of a heater element located between two temperature-sensing resistors on layer 3 of the stack. Layers 2, 4, and 5 form a ductwork that directs fluid down from layer 1, across the resistor elements on layer 3 and back up to exhaust through layer 1. Electrodes on layer 1 are connected to the resistor elements through vias in layer 2. In operation, constant power is dissipated in the heater element. The difference in temperature of the two sense resistors, which are wired as two legs of a Weathstone bridge circuit, is proportional to the fluid flow rate. The electronic components needed to control power dissipation in the heater and measure the voltage generated by the resistor bridge circuit are not shown. They are mounted and interconnected on the cofired ceramic circuit board. Layer 1 Layer 2 Layer 3 Layer 1 Layer 5 Layer 4 Layer 5 Figure 17. Gas Flow Sensor Fabricated from Layers of Cofired Ceramic Optics Digital Light Projector – developed and produced by Texas Instruments remains one the most innovative, technologically impressive, and financially important MEMS devices built to date. A large array of movable aluminum mirrors is fabricated on top of a CMOS memory chip. Electrodes beneath each mirror element allow it to be toggled between to positions in response to the contents of its associated memory cell. The device is the enabling component in bright, high resolution projection displays for high definition television. Figure 18. Construction of TI Digital Light Projector Variable-Focus Liquid Lens – holds an electrically conductive fluid, which serves as the lens element, within a sealed glass cylinder. One electrode contacts the fluid, while a second, insulated electrode, is wrapped around the inside wall of the cylinder. The inside wall of the cylinder is coated with a hydrophobic coating to prevent the conductive liquid from wetting it. A non-conductive fluid fills the internal volume of the cylinder that is not occupied by the conductive fluid. The operation of the lens is illustrated in Figure (19). In the off state, shown on the left side of Figure (19), the conductive fluid has positive meniscus. When a voltage is applied across the two electrodes, the surface tension of the conductive fluid is reduced along the cylinder wall. This causes it to wet more of the wall area and consequently change the meniscus shape. Incident Light Hydrophobic Coating Insulating Fluid Insulator V Electrodes Conducting Fluid Figure 19. Operation of Variable Focus Liquid Lens Tunable Fabry-Perot Filter – developed by NASA for infrared spectrometry applications is shown in the left side of Figure (20). A schematic cross section of the device is shown in the right side of Figure (20). The spacing between the mirror elements, in the center of the disk, is controlled by a pair of capacitor plates on the perimeter of the disk. Figure 20. Photograph and Schematic Representation of a Fabry-Perot Filter for IR Spectrometry Adaptive Mirror – shown in Figure (21), is comprised of two micromachined silicon parts. The top chip is a carrier for an aluminum metallized film. The bottom chip carries the array of electrodes which deform the film in proportion to the quantity of charge placed on them. Figure 21. Adaptive Mirror Optical Fiber Switches – are built in many different configurations for a variety of networking functions. Figure (22a) shows a 2 X 2 switch in which the optical path of two intersecting beams are enabled or blocked by a single slider. A large electrostatic comb drive actuator is used to quickly position the slider. The image in Figure (22b) is of a 1 X 1 switch in which the shutter is mounted on the end of a teeter-totter arm. A parallel plate capacitor structure on the other end of the teeter-totter is used to drive the shutter up and down. Figure 22. (a) 2 x 2 Optical Switch Based on Moving Slider. (b) 1 x 1 Optical Switch with Vertical Actuated Shutter. Bolometer Imaging Array – constructed by patterning an array of thermally sensitive resistors on a Kapton film carrier. Electrical connections to the resistors are made through the film from the bottom side to minimize space between imaging pixels. Figure 23. Imaging Bolometer Constructed From Array of Thermally Sensitive Resistors on Kapton Film RF Devices 1 GHz Resonator – shown in Figure (24) is fabricated using a silicon-on-insulator wafer, which has several advantages for this application. The thickness of the silicon layer is precisely controlled over the entire wafer. This dimensional control, combined with precise etching of the structure geometry minimizes variations in the eigen vibration modes. The use of single crystal silicon for the resonator minimizes energy losses to obtain a high Q device. Finally, oxide attachment of the device silicon to the handle wafer minimizes residual stress and its associated degradation of the device oscillation. Figure 24. 1 Ghz Resonator Design Switch – operating modes cab be classified as either conductive or capacitive. An example of a conductive switch is shown in Figure (25). In this device the conductive element is attached to the ends of two compliant beams. It is moved up and down by a parallel plate capacitor drive structure to make or break contact between two electrodes. The other type of switch, not shown, uses a grounded conductor element coated with dielectric. It is moved up or down to contact a conductor line below it to change its impedance. The use of this type of switch is prevalent in RF applications. Figure 25. Example of Conductive Element Switch Silicon Oscillator – developed by SiTime is targeted at traditional quartz crystal oscillator applications. The MEMS device incorporates packaging as an integral part of device fabrication to yield a product, which is significantly smaller and of lower cost than competing quartz products. A top view of the main components of the MEMS resonator structure is shown in Figure (26). The device is built on a silicon-on-insulator wafer using standard CMOS processes. The first step is to deep trench etch the resonator structure and drive and pickoff electrodes in the device layer of the wafer as shown in Figure (27). Drive Electrode Anchor Resonator Beam Pickoff Electrode Figure 26. Top View of Resonator Structure and Drive and Pickoff Electrodes Silicon Oxide Silicon Figure 27. Cross Section View After Deep Trench Etching of Resonator and Electrodes Figure 28. Oxide Deposited Over Wafer and Patterned Figure (28) shows a cross section of the wafer after oxide is deposited and patterned to provide access to electrode. A two microns thick layer of silicon is deposited and etched with vent hole to provide access to the underlying oxide. Resonator Beam Figure 29. After Removal of Oxide by Etching in HF Vapor The thickness of the top silicon layer is built up using chemical vapor deposition process and deep etched to isolate electrodes. A second layer of oxide is deposited and patterned to expose electrodes, as shown in Figure (30). Figure 30. After Patterning of Second Oxide Layer Figure (31) shows the completed resonator depositing and patterning metal on the top oxide layer. Pickoff Drive Electrode Electrode Figure 31. Completed Resonator Neural Probes – of micro-arrays of electrically conductive needles, each of which is addressed individually, have shown considerable promise in medical experiments. Implanted devices have allowed motor-impaired patients to generate commands to a computer system. Implants in blind patients have allowed them to perceive images of simple geometric shapes. Figure 32. Neural Probes Data Storage Devices – fabricated using massive arrays of read write probes would have capacities commensurate with those of the largest hard drives used today. Their form factor, however, would approach that of single chip memory devices. Like hard drives, they are a non-volatile storage medium, but as mechanically robust as solid state memory. Figure 33. IBM Data Storage Array Inductor Coils – are ubiquitous in RF, power, sensing and actuator circuits and their size, geometry and method of fabrication are as varied as their applications. Figure 934) illustrates construction of a multi-turn coil, which is fabricated using low temperature cofired ceramic and printed silver conductors. Arrays of actuators, for micro- valves, have been fabricated this way as well as multi-pole filters, transformers and magnetic sensors. A key attraction of this technology is resistor and capacitor circuit elements, as well as cavities and fluid channels are easily fabricated as part of the process. Layer A Layer B 1 A 2 B 3 A 4 B 5 A Figure 34. Inductor Coil Fabricated by Stacking LTCC Sheets Inertial Sensors In-Plane Accelerometer – produced by Analog Devices Incorporated utilizes surface micromachining to fabricate a suspended proof mass with inter-digitated comb capacitor structure on top of a CMOS ASIC chip. The chip includes all of the electrical circuits needed to read the acceleration that acts on the MEMS structure. Figure 35. Analog Devices Accelerometer Tuning Fork Gyroscope – shown in Figure (36A) was invented at The Charles Stark Draper Laboratory. Its operation is depicted in the schematic illustration of Figure (36B). Electrostatic comb drives cause the two proof masses to move back and forth parallel to the plane of the substrate in oscillatory fashion. When the sensor is rotated at a rate ω, about an axis parallel to the suspension beams, the Coriolis forces displace the proof masses normal to the substrate plane. The sensor detects rotation by measuring the change in capacitance between the proof mass and its underlying electrode, which occurs as it moves along y. Motor Motor Motor Proof Proof Mass Mass Suspension Beam Anchor x2 x1 y2 y1 ω Electrode Glass Substrate (A) (b) Figure 36. (a) Scanning Electron Micrograph of a Tuning Fork Gyroscope. (b) Schematic Representation of Top and Cross Section Views of Tuning Fork Gyroscope. System Integration The number of different types of MEMS devices is large and steadily growing. Also, increasing numbers of MEMS devices are being incorporated in functional systems. For example, temperature, air flow, relative humidity and pressure sensors may be combined to control the heating and air conditioning system in a building or automobile. Inertial measurement units for military and aircraft navigation systems combine a magnetometer with three pairs of accelerometers and gyroscopes mounted on orthogonal axes. Figure (37) is a schematic diagram of a simple RF transceiver such as used in a mobile telephone handset. There are 14 separate MEMS devices shown highlighted in this circuit, but the number increases rapidly in more capable radio designs, such as those used for phone base stations. Figure 37. Diagram of Typical Radio Transceiver The current size of packaged MEMS devices is relatively large, primarily because of limitations with available packaging technology. This situation is changing, however, as wafer level capping processes mature and as packaging becomes an integral part of the device fabrication process. The looming challenge for MEMS and Microsystems in general, is how to assemble large collections of disparate components, with sub-micron scale alignment, at an affordable cost. In a paper by Morris et. al, they plotted the measured rate of assembly of various components, by the best available technology, as a function component size. Their plot is reproduced in Figure (38). The actual cases shown are: (a) individual atoms placed by scanning probe microscope, (b) polymer memory storage device, (c) optical tweezers assembly of molecules, (d) 3-D microassembly, (e) robotic pick and place of circuit boards, (f) robotic assembly of grease gun, (g) robotic assembly of automobile wheel bearing, (h) robotic assembly of automobile body. Their data clearly illustrates the challenge posed by assembly of micro- systems and the need for new innovative assembly methods. Figure 38. Measured Rate of Assembly Versus Component Size One approach to the problem of precisely aligning and interconnecting MEMS devices and associated integrated circuit devices is to employ micromachining technology to fabricate the interconnect structure. A paper by Chowdhury et. al. describes such an approach for building a multi-element microphone array with integrated signal processing capability, which is designed to be implanted in the inner ear. The beam forming capability of this system offers vastly superior performance with respect to noise suppression and echo cancellation, as compared to more conventional hearing aids. The 6 millimeter diameter stack of microphones, amplifiers, digital converters and signal processors is held together by micron scale pins and sockets. Electrical connections between the stack layers are made through rows of cantilever beam spring contacts. A sketch of the assembly appears in Figure (39). Figure 39. Assembly of Microphone Array The idea standardizing MEMS sensor interfaces and support electronics by using carriers fabricated in low temperature cofired ceramic or high density circuit board technology has been promoted in a paper by Schuenemann et. al. Figure (40) from their paper illustrates their vision of building complex systems by stacking standardized components. Packages are fabricated from low temperature cofired ceramic, with standardized I/O patterns to allow stacking. Standardization of package styles, supply voltages, etc. has been an essential element in adaptation of integrated circuit technology to modern electronic systems. This approach is reasonable for many sensor and control applications, in which the number of devices is small and signal path lengths are not critical. Figure 40. Scheme for Packaging and Integrating MEMS Devices Assembly of larger quantities of smaller devices has been investigated by using several different techniques. Flip chip technology has been successfully used to assemble silicon IC chips onto multichip module substrates. Advantage is taken of the high surface energy of molten solder to precisely align thousands of chip I/O pads to their counterparts on a substrate. The desired state of precise alignment is designed to be that which minimizes the solder surface area. Misalignment between the chip and its desired location generates forces, proportional to the degree of offset, which pull the chip into position. Alignment using solder wetting has been used to successfully align and connect over a thousand light emitting diodes in a cylindrical structure. They were wired to be individually addressable as shown in Figure (41). Figure 41. Cylindrical Structural with Over 1000 Self-Assembled LEDs In an adaptation of this basic process, Chung et. al. utilized a substrate with a grid of heating wires embedded in it, which allowed them to attach several different planar devices in sequence.19 Heaters at the desired placement sites were turned on so that the appropriate devices could be attached from a fluid bath as they were carried over the site by induced flow of the fluid. The use of self-assembled monolayers (SAM), which can render a surface hydrophobic or hydrophilic and which can be modified by exposure to ultraviolet light have also been used in self-assembly demonstrations. These materials offer more flexibility than the solder approaches, avoids heating parts, and is thought to be extendable to the nanoscale. In a process utilized by Srinivasan et al, hydrophobic binding sites are created on the parts and substrate by evaporating gold films on them, which are photo patterned and coated with a alkanethiol self- assembled monolayer precursor. The substrate is then passed through a film of hydrophobic adhesive on water. The adhesive selectively coats the gold pads of the substrate. The parts are then directed towards the substrate under water. When the hydrophobic pattern on a part comes into contact with and adhesive coated site on the substrate, self-alignment occurs as a result of minimization of the adhesive-water and SAM-water interfaces. Summary The thesis of this paper is that the future of MEMS is integration of larger numbers of diverse components into systems of increasing complexity. This direction of development is motivated by both the economic benefit derived from these systems and the new technological capabilities that they enable. For example the operating efficiency of combustion engines, energy conversion equipment, and many manufacturing processes will be improved by more capable MEMS systems. More advanced MEMS systems offer significant potential for technological advances in medical diagnostics, autonomous robotic systems, communications, and computer systems. Three different, somewhat disparate topics were discussed to support this thesis. First, data from various market surveys and trade publications was presented to demonstrate that the sale of MEMS devices into commercial markets is significant and growing. Profits from the sale of these devices will provide the resources and incentive for continued technology development independent of government investment. Second, a sampling of the many types of MEMS devices available was presented. This survey attempted to illustrate the broad range of applications as well as the diversity in materials and processes used to produce MEMS devices. It also illustrated how recent progress of integrating packaging into the fabrication process has yielded smaller devices that can be produced in state of the art CMOS fabs at very low cost. Finally, a simple radio transceiver was used to illustrate the need to incorporate a plethora of small, discrete devices of diverse construction into a cost sensitive, consumer product. Experiments using a variety of self- assembly techniques have yielded encouraging results that suggest it will indeed be possible to assemble these miniature, complex systems at an affordable cost. References 1 EE Times, Feb, 2007 2 John Huggins, Berkeley sensor and Actuator Center, “BSAC First Half FY07 State of the Center Summary,” March, 2007 3 J. C. Eloy, Yole Development, 2004, “Status of the MEMS Industry” 4 Jeff Chappell, Electronic News, June 16, 2004 5 MEMS4Mobile, “Analysis of the Applications and Markets of MEMS in Mobile Communications,” Yole Development, Feb. 2005 6 Janusz Bryzek, et. al., IEEE Circuits and Devices Magazine, March/April, 2006, pp.8-28, “Marvelous MEMS” 7 http://www.SiTime.com 8 http://Akustica.com 9 Zhizhong Yin and Andrea Prosperetti, Journal of Micromechanics and microengineering, vol. 15, 2005, pp. 643- 651, “A Microfluidic “Blinking Bubble” Pump” 10 Paul Dozoretz, Chris Stone, and Ole Wenzel, Sensors Magazine, January, 2005, pp. 28-32, “Shrinking the Pirani Vacuum Gauge” 11 Thomas P. Burg, Amir R. Mirza, Nebojsa Milovic, Christine H. Tsau, George A. Popescu, John S. Foster, and Scott R. Manalis, Journal of Microelectromechanical Systems, vol. 15, no. 6, 2006, pp. 1466-1476, “Vacuum- Packaged Suspended Microchannel Resonant Mass Sensor for Biomolecular Detection” 12 Fan-Gang Tseng, Kun-Hua Lin, Hui-Ting Hsu, and Ching-Chang Chieng, Sensors and Actuators, vol. 111, 2004, pp. 107-117, “A Surface-Tension Driven Fluidic Network for Precise Enzyme Batch-Dispensing and Glucose Detection” 13 Photonics Spectra, October 2004, pp.32-35, “Liquid Used as Lens for Miniature Cameras” 14 G. Vdovin, PhD Thesis, Delft University of Technology, 1996, “Adaptive Mirror Micromachined in Silicon” 15 Ji-song Han, Zhi-yong Tan, K. Sato, and M. Shikida, J. Micromechanics and Microengineering, vol. 14, 2004, pp. 38-48, "Three-Dimensional Interconnect Technology on a Flexible Polyimide Film" 16 Rob N. Candler</A>, Matthew A. Hopcroft, Bongsang Kim, Woo-Tae Park, Renata Melamud, Manu Agarwal, Gary Yama, Aaron Partridge, Markus Lutz, and Thomas W. Kenny, J. Microelectromechanical Systems, vol. 15, no. 6, 2006, pp. 1446-1456, "Long-Term and Accelerated Life Testing of a Novel Single-Wafer Vacuum Encapsulation for MEMS Resonators" 17 D. Kewley, M. Hills, D. Borkholder, I. Opris, N. Maluf, C. Storment, J. Bower, and G. Kovacs, Sensors and Actuators, vol. A58, 1997, pp. 27-35, “Plasma Etched Neural Probes” 18 H. J. Yeh</A> and J. S. Smith, IEEE Photonics Technology Letters, vol. 6, 1994, pp. 706-708, "Fluidic Self- Assembly for the Integration of GaAs Light Emitting Diodes on Si Substrates" 19 Jaehoon Chung</A>, Wei Zheng, Thomas J. Hatch, and Heiko O. Jacobs, J. Microelectromechanical Systems, vol. 15, no. 3, 2006, pp. 457-464, "Programmable Reconfigurable Self-assembly: Parallel Heterogeneous Integration of Chip-Scale Components on Planar and Nonplanar Surfaces" 20 Uthara Srinivasan</A>, Dorian Liepmann, and Roger T. Howe, J. Microelectromechanical Systems, vol. 10, no. 1, 2001, pp.17-24, "Microstructure to Substrate Self-Assembly Using Capillary Forces"