MEMS: AN ULTIMATE CONQUEST
MEMS are poised to become the most important technology of the 21st century, but only if packaging issues can be resolved. Substantial progress continues in the key areas of lower-cost ceramics, molded plastic cavities and wafer-level packaging.
MEMS are the ultimate enabling technology for integrating virtually any phenomena — motion, light, sound, radio, chemistry and computation — all on a single chip. High-volume commercial products sense complex motion, measure the slightest pressures, and propel fluids. Newer products receive and send light beams, others detect specific molecules, and some deal with several "senses" simultaneously. If the logic circuit is the brain, MEMS add the eyes, nose, ears, touch, taste buds and super-human sensing ability. MEMS also bring control and manipulation, the hands and fingers that enable chips to operate on matter. This merging of motion, sensing, control and computation represents a gigantic leap in technology.
MEMS technology is essentially the grand unification of mechanical, photonic, electrical and electronic subcomponents into an integrated system that could be the most important technologies of the century. Developers are also adding optics and photonics to expand the increasingly popular subset known as "MOEMS," or optical MEMS. We now have a potent and versatile convergence of the world's technologies onto and into a single remarkable chip. We can build incredibly small machines that perform the same mechanical, optical, electrical and electronic functions found in massive equipment from our macro world. It's like shrinking everyday machinery onto tiny chips. MEMS are the ultimate system-on-a-chip (SoC), the new embodiment of the old cliché, "Shrink the world onto a chip.
MEMS present a "new" or "old" paradox. MEMS remain a hot "new" technology that achieved notoriety during this decade. So is it new? Surprisingly, MEMS technology is as mature as the IC. But MEMS are as modern as nanotechnology. And the two, which are often grouped together, may converge to become nanoelectromechanical systems (NEMS). The nanotech concept is often attributed to the Nobel-prize-winning physicist Richard Feynman and his famous 1959 lecture, "There's Plenty of Room at the Bottom." But Feynman was also talking about MEMS when he offered up the first MEMS challenge and prize to the person who could make an operating electric motor only 1/64 in.3. Less than a decade later, patents were filed that detailed the basic MEMS principles.
Modern MEMS and ICs travel in parallel universes. Most MEMS are a clever extension of semiconductor technology, where modified fab processes create mechanical and macro world structures on wafers, along with integrated circuitry. MEMS build on the huge semiconductor industry, tapping into that massive knowledge base, synergistically incorporating solid-state electronics where appropriate. But MEMS take us far beyond electronics, with chips that handle multiple energy forms: mechanical forces, light waves and radio waves, as well as conventional electrical energy. MEMS technology can provide the highest level of integration and functionality yet devised. Motion, light, sound, heat, wireless, reactive molecules and electrons can all converge on these unusual chips. MEMS are "convergence central" for all science and technology. The key is extreme integration of dissimilar systems in a miniature form factor to deliver unique, valuable and practical functionality. Most importantly, MEMS use the massively parallel productivity of wafer-level manufacturing. Because MEMS can shrink motors, generators, pumps and even turbines down to the semiconductor scale, it can literally build a factory on a chip.
Motion sensors remain the fastest growing commercial segment, with frequent new product launches by both established and new suppliers. Versatile MEMS accelerometers are trusted with the life-and-death situation of airbag deployment. These chips sense sudden deceleration to instantly "decide" when to deploy an airbag. Newer MEMS gyroscopes even prevent rollovers. Inertial sensors continue to find new applications in a wide range of industries. While airbag systems remain the primary market, many newer applications, such as disk drive free-fall detectors and innovative consumer products, are gaining marketshare. MEMS motion sensors have movedinto consumer applications, including cell phone pedometers, game controllers and the iPhone formotion-based input.
The addition of the photon brings us to MOEMS, which have added versatility.
MEMS pose the greatest packaging challenge, but there is a paradox. Early products such as inkjets and airbag sensors use reasonably simple packaging solutions. But emerging areas (i.e., bio-MEMS) have special requirements that need novel approaches. These packages will need to handle fluids. But, let's begin with the general requirements for MEMS, with the understanding that packaging can be device specific. Unfortunately, device-specific solutions mean that progress is slower and standardization is challenging.
MEMS chips can be sensitive to mechanical shock before they are packaged, and are especially vulnerable to particulate contamination from wafer-sawing residue. Some foundries, therefore, do some or all of the packaging. The MEMS chip "motion zone" must be protected during singulation with a mask or a wafer-l
evel technique. The most universal MEMS packaging requirement is protection without restricting mechanical action. However, we cannot simply adopt solutions developed during the past 50 years for "pure" electronic chips. Electronic devices are overmolded with encapsulant contacting the chip face, so this approach would "freeze" moving parts in MEMS.
MEMS have adopted one type of electronic package — the hermetic cavity. Hermetic packages, however, add substantial cost. While early MEMS chips were packaged in somewhat standard ceramic packages, cost impeded success. Progress is being made in reducing the cost of ceramic packages such as QFN. Although full hermeticity can ensure reliability, the added cost can torpedo a product and sink a company. So the hermeticity debate rages on. Which MEMS need full hermeticity? Not all. MEMS devices that handle fluids, especially aqueous, certainly do not need a hermetic enclosure. In fact, inkjet chips have most of their active surface exposed to the environment to "fire" ink at the target.
Another issue is the need for more than "electron plumbing." Even simple inkjets chips have two inputs — electrons and ink fluid — and other chips have even more. But MEMS inertial sensors are somewhat unique, only requiring electrical I/Os, which makes the interface simpler. These chips still need motion, space or "head room." Although early accelerometer packages were hermetic cavities, device makers and packaging partners came up with more cost-effective designs.
Full hermetic — The full hermetic package has admirably served the electronics and optoelectonics industries. This package class began as the cathode ray tube (CRT), then the vacuum tube and, finally, the semiconductor package. Early opto- and electronic devices could not operate in air, but very few systems require a vacuum today. The hermetic packages evolved from glass to ceramic and metal, and are still used for MEMS.
Non-hermetic plastic package — Plastic packages became mainstream with the breakthrough introduction of the dual inline package (DIP). The chip is typically attached to a metal leadframe by wire bonding, followed by epoxy overmolding. Molding compound — a blend of solid epoxy resins, hardeners, fillers and additives — is forced into a mold that hold the leadframe assembly and is heated to polymerize the material. The molding compound makes direct contact with the die, wire bonds and leadframe. The ball grid array (BGA) uses a similar method, but a different substrate. Overmolding cannot be used on unprotected MEMS chips, but capped devices can be packaged using this method.
Near-hermetic package — The component packaging industry offered two extremes: the full hermetic and non-hermetic packages. Package designers have suggested an intermediate design called the near-hermetic package (NHP). NHPs seek designs that are "good enough" and "cheap enough" for MEMS and other cavity-requiring devices. The concept is still evolving and can have product-specific performance.
Wafer-level partial packaging — This type of packaging is formed while devices are still in wafer format. A considerable number of wafer-level packaging (WLP) methods have emerged for electronic devices, but MEMS are just getting started. The dominant MEMS wafer-level process is capping, a partial packaging process that requires subsequent steps. Several companies have been capping accelerometers for many years , but the caps are passive (no electrical paths). Passive caps must not block chip bond pads, so the most common solution is to singulate caps first using a multi-step process, and later singulate the MEMS wafer. Double singulation, however, adds cost.
Full WLP — The alternative is to fully cap the MEMS device and provide an interconnect structure through the chip or cap. This allows the cap layer to be singulated during the MEMS sawing step. The patent area shows that considerable attention is being expended to create an interconnect structure through the cap or chip. Many inventions have been filed for full WLP MEMS, and this high patent activity indicates WLP will soon emerge.
Although newer ceramic materials and designs have reduced cost, plastic packaging remains the low-cost leader. Plastic packaging is primarily based on epoxy that accounts for most of that market. Advantages are lowest cost, high versatility and ease of automation. The area-molded QFN has reduced cost even further, making the five-cent package a reality. The MEMS industry has sought ways to adapt cost-reducing plastics with modest success. Two basic strategies are used: a molded cavity design or protection over the mechanical zone.
While overmolding can't be used on bare MEMS chips, it is acceptable for some capped devices. Injection molding, however, is ideal for producing 3-D shapes and is used to make cavity packages. Injection molding uses thermoplastic resins rather than thermoset epoxies. Several high-temperature thermoplastics, with properties superior to epoxies, are available. Liquid crystal polymer, for example, has a good moisture barrier and low flammability without flame-retardants. Thermoplastics are also "no waste" materials that can be remelted, reused and recycled, but withstand soldering
Plastic packages for MEMS
A pioneer in MEMS and MEMS packaging, Analog Devices (Norwood, Mass.) explored plastic cavity packages for its inertial sensors. Chips are first capped at the wafer level so the motion zone is hermetically enclosed. The capped wafer is protected from both particulate contamination and moisture, so package hermeticity is less critical. The capped MEMS devices can be overmolded with epoxy. But products requiring higher performance are placed into molded cavity packages because overmolded stress, caused by shrinkage, detunes the MEMS sensor. The loaded plastic package is completed by sealing a lid with adhesive, ultrasonic welding or a laser. Although wafer capping adds cost, it solves several problems, including contamination from sawing.
MEMS devices cannot be capped using today's technology because they require gas or fluid input. These may be candidates for injection-molded plastic cavity packages. Thermoplastic packages may have higher barrier performance than non-hermetic epoxies, but they don't achieve full hermeticity, even though they can pass the fine leak test.
Other packaging issues
A vacuum may not be required or even acceptable for some devices, but in-package atmosphere control may be essential. A MEMS package enclosure and adhesives can release materials (outgas), and this affects MEMS performance. Phenomena associated with chip mechanics can also be a problem. Getters, anti-stiction additives and lubricants can help. Getters trap moisture and particles, while lubricants and anti-stiction agents can improve performance and reliability. Stiction, a combination of sticking and friction, occurs when smooth planar surfaces make contact and become permanently locked together by short-range atomic forces — a problem for most MEMS chips. Coatings with low surface energy and hydrophobicity can help. The film must be very thin and is applied by vacuum coating or other methods. Many patents deal with stiction and typically involve low surface energy coatings.
Bio-MEMS, fluidic packaging
Bio-MEMS devices are being developed and tested at universities, clinics and companies. They can sense in-body parameters, analyze body motion, measure tiny forces, identify bio-agents, pump and control fluids, administer drugs, and perform other actions that have great value for the medical and biological fields. Bio-MEMS of the future may become the most important MEMS class, with "medical practitioners" becoming permanent inhabitants of your body. Bio-MEMS packaging requirements will be very challenging because some devices will handle fluids that require fluid channels and couplings. Work is well underway to develop fluidic couplings for packages. These couplings can be fabricated from silicon or plastics.
The fully hermetic package was the original "box" for MEMS, but it was costly. Lower-cost ceramic QFNs have replaced the larger, more expensive traditional packages in many products. Wafer-level capping is a strong trend for inertial sensors, and the technology is evolving. Capped chips are now used in plastic packages, both overmolded and injection-molded cavity styles. Injection-molded packages have key advantages, but still lack full hermeticity. They offer low cost and extraordinary precision at small dimensions. Micro-molding can produce extremely small, precise and complex structures, even nanoscale features. These attributes may not yet be needed, but they will prove valuable for future advanced MEMS, especially biomedical products.
Full WLP is another important direction for MEMS, but is still emerging. Recent patents and applications indicate the MEMS industry is getting ready to launch full WLPs. Capping processes are already well established for passive caps (protection-only caps, no vias), but connector feed-through packages are on the way. Connection strategies include vias-through-chip or vias-through-cap, and such packages could be rolling out of the foundries soon. This brings up supply chain logistics, such as where MEMS will be packaged. At the foundry? Most capping is already done at the fab.
Plastic cavity packaging, which represents only a small segment today, will evolve for specialty applications. A lack of full hermeticity will remain an impediment to wide-scale use unless hermetic barrier properties can be achieved, which is a definite possibility. Future packaging development will concentrate on both plastic molded packages and full WLPs, but will the two methods compete? There may be some competition, but the technologies are so different that each can solve problems for different products and coexist peacefully.
There is also a place for commercial MEMS packaging. Several companies already offer MEMS packages. Some companies, like RJR Polymers (Oakland, Calif.) and Quantum Leap Packaging (Wilmington, Mass.), focus on injection-molded plastic cavity packages. Others, such as Hymite (Copenhagen, Denmark), offer full hermetic systems. Larger companies, like Amkor (Chandler, Ariz.), also provide MEMS packaging services. IP-centric companies, such as Tessera Technologies (San Jose), offer solutions for MEMS and MOEMS. But it remains to be seen how the industry will solve the myriad of packaging issues that relate to both technology and business.
MEMS have the potential to become a hallmark technology of the 21st century. The ability to sense, analyze, compute, control and synthesize materials at the chip level can provide new and powerful products. Sensors are the big market today, but many other areas are quickly evolving, including bio-MEMS and other advanced areas.