Emerging MEMS and Sensor Technologies
Researchers continue to innovate with MEMS. But will low cost sensors demanded by IoT drive MEMS off of silicon substrates? What’s next?
This article is adapted from the presentation “Emerging MEMS & Sensors Technologies to Watch” presented by Alissa Fitzgerald of AMFitzgerald & Associates on Nov. 1, 2017, at the MSIG Executive Congress 2017.
Microelectromechanical systems (MEMS) is a technique for creating electromechanical components historically fabricated using materials (e.g., silicon) and processes similar to those for fabricating integrated chips. The first MEMS device was a strain gauge that became commercially available in 1958. MEMS gained traction in the 60s, and by the late 70s, MEMS saw vital technologies emerge such as wet chemical etching with alkaline etchants, enabling precise management of dimensions. The industry went on to create miniscule nozzles, shapes, and pockets in the bulk of silicon. The alkaline etchant process technology, coupled with a deeper understanding of the piezoresistive effect of silicon, facilitated creating products such as pressure sensors and inkjet print nozzles. Although wet etching is almost a 40-year-old technology, these products are still used today.
Researchers are using SAW technology to create a battery-free temperature sensor.
The 90s saw experimentation with CMOS thin-film processes and the beginning of semi-planar (a.k.a. 2.5D) structures using standard CMOS processes. MEMS engineers selectively removed materials in the CMOS stack to create freestanding, thin-film structures. In turn, 2.5D technology gave rise to Texas Instruments’ Digital Mirror Device (DMD) and subsequent Digital Light Projection (DLP), which is ubiquitous in modern projectors. In 1991, Analog Devices used the CMOS process to create the first fully integrated MEMS accelerometer on a single chip.[i]
Continuing through the 2000s, the Bosch Deep Reactive Ion Etching (DRIE) become a significant process in MEMS, enabling the consistent creation of deep trenches, low incidence of non-uniformity across a wafer, and excellent profile control without sacrificing high selectivity in masking material. DRIE ushered in new device architectures, including inter-digitated, comb-like capacitive sensors. At the same time, AVAGO perfected thin-film deposition of the piezoelectric material Aluminum Nitride (AlN). From these technologies emerged microphones, gyros, and accelerometers. Many MEMS devices produced today leverage 2000s-era deep silicon etch. The debut of Broadcom (formerly Avago) Film Bulk Acoustic Resonator (FBAR) filters offered superior performance to Surface Acoustic Wave (SAW) filters, improving sensitivity and battery life at a smaller size; a boon to smartphones.
By the 2010s, one significant technology to emerge and be perfected is aligned, eutectic wafer bonding that enables a hermetically sealed vacuum cavity within microstructures. The technology also forms electrical interconnects through the bond interface between wafers, enabling stacked wafer products. Separate CMOS and MEMS wafers are combined at the wafer level. Operating under different sensing modalities, the two are brought together to create dense, small-footprint devices with sophisticated capabilities. The InvenSense gyro is an early example that embodies this technology.
The MEMS Industry Today
How does MEMS intersect with markets?
Figure 2 describes what the MEMS industry has looked like for many years. A few large companies with high revenues are the major players. “The long tail” is where another 400 or more MEMS companies exist in the market. The top few are making 50% of the revenue, with the top thirty making 80% of the revenue worldwide. The long tail includes MEMS businesses where new process technologies are being incubated, explored, and heading to commercialization. History has shown companies in the long tail jump to the head of the market.
Figure 3 shows the top players in the MEMS industry by 2006. Various technologies that were being used are designated by color. The top of the market is using 15- to 25-year-old technology. The iPhone was introduced in June 2007 and the Nintendo Wii in November 2006. InvenSense, established in 2003, is not yet on this chart. Seven years later, smartphones, game consoles, and other devices have transformed the market. In Figure 4, companies making top revenue in MEMS have next-generation technology, having invested in process and device technology. Disruptive products cause those still using older generation technology to slip from market leadership. Epson is now last place, and others are no longer on the chart.
NASA Method for Comparing Technologies, Translated to MEMS
NASA has a Technology Readiness Level (TRL) scale by which they measure the maturity level of a particular technology, which is used below to describe the evolution of a technology from initial idea to full commercialization.[ii] Few companies own labs focused on basic research. The MEMS industry mostly depends on universities and government-funded research laboratories, who do early-stage research TRL 1 through TRL 4. TRL 1 – 4 indicate emerging technologies. Proof of concept exists; however, the technology is far from high-volume manufacturing. Companies like AMFitzgerald are development service providers who bridge the gap from early concept to completing engineering work to get a technology ready for commercialization and production. Development service providers take the process from TRL 4 to about TRL 7.
Table 1: NASA’s Technology Readiness Level scale as applied to the MEMS industry. Funding amounts are conservative minimum estimates that are required to move from stage to stage. Sources: AMFitzgerald and NASA.
Mining for MEMS Gold
Is it possible that the next billion-dollar product is lurking somewhere in academic literature? AMFitzgerald and Associates combed over 500 papers in the proceedings from Transducers 2017, one of the largest international conferences on transducers, sensors, and MEMS technology. AMFitzgerald filtered proceedings for techniques that appeared to offer commercial viability, a path to manufacturing, or some technology game changer.
The following technologies are shown in order of most to least mature, followed by representative examples:
- FBAR and SAW-based sensors
- Near-zero power or event-driven sensors
- Piezoelectric devices
- CMOS+ sensors
- Novel piezoelectric materials
- Paper sensors
FBAR and SAW-based Sensors
Researchers are finding new uses for existing technology. FBAR and Surface Acoustic Wave (SAW) filters are Radio Frequency (RF) components that are in every cell phone for wireless RF signal acquisition and filtering. Researchers are using SAW technology to create a battery-free temperature sensor. Tsinghua University shows that a passive SAW device, when interrogated with a radio signal, results in an acoustic standing wave in the device. The center frequency of that resonance is highly sensitive to temperature. Tsinghua University has demonstrated a very sensitive, passive temperature sensor that can be woken up with a remote wireless signal and return the temperature measurement. Researchers built a SAW sensor using silicon carbide, which can be used in very high-temperature applications. Using resonant detection, which is the highly sensitive readout mechanism, results in part per million resolutions. The SAW structure enables a battery-free inductive readout. Other examples exist that use the same architecture for sensing pressure. Gas sensing can be accomplished with an overlay on top of the sensitizing layer, over the SAW device. This technology is at TRL 4; ripe for further investment to make it scalable.
Near Zero Power, Actual Zero Power, and Event Driven Sensors
Northeastern University is researching event-driven sensors (a.k.a. near zero power, actual zero power) sponsored by DARPA in a clever implementation that couples physics phenomena. A sensor with a surface plasmonic IR absorber surface has a special pattern on its surface that is configured to detect certain wavelengths of infrared light. DARPA is interested in the exhaust signature of vehicles that give off excess heat. There’s a spectrum of infrared signatures coming out of the exhaust.
The group at Northeastern University found a way to create a wavelength-sensitive device such that when a specific wavelength of infrared energy is absorbed, it causes the stress state of a mechanical structure to change. The bimorph structure, upon changing shape, closes an electrical switch. The result is an energy-free means to take optical or thermal energy input and produce a mechanical output to actuate an electrical switch. This architecture produces a device with no leakage current, as it is an open circuit until an external event occurs. When the event occurs, the sensor wakes up and actuates its purpose in a system. These sensors can be sentinels; useful for vast area arrays, security, or any application where batteries cannot be changed often. These examples are new embodiments of existing processes and technologies and ready for further development for mass production.
There are many piezoelectric products on the market today, but trends are emerging for more sensors based on piezoelectric transduction instead of the more common capacitive electrostatic comb figure design. For example, the MEMS market has seen the Vespers piezoelectric microphone (versus the older capacitive MEMS microphones). Thin-film Lead Zirconate Titanate (PZT) is becoming more mainstream; the market is displacing electrostatic designs as they re-architect existing sensors. There appears to be a shift from DRIE sensors to piezoelectric materials. One example is a micro-speaker designed by Fraunhofer where the inner square of the MEMS device is a tweeter, and the surrounding area is a woofer, resulting in a near 1 cm² micro-speaker. Piezoelectric devices are sensitive and low power, which is of high interest for consumer electronics and the Internet of Things (IoT). Whereas some of the shift to piezoelectric materials leverages mature processes, others need more work. Reimagining sensors in piezoelectric elements are at TRL 3 to 4.
AMFitzgerald found research that repurposes existing and well-established CMOS technology for novel applications. One biotech application from National Tsing Hua University in Taiwan includes CMOS CCD arrays coated with reactive or sensitized materials that enable an electrically capacitive, rather than optical, readout of DNA. Presently, when DNA is detected, samples are washed with a chemical that fluoresces. Technicians need a huge fluorescent detector to be able to read out whether DNA is present. Several groups are figuring out how to repurpose CCD technologies so that the DNA can be read out directly using electrical detection methods. These applications are suitable for point of care diagnostics and biotechnology in general. The CMOS technology is mature, the materials that are applied to the surface of the CMOS to sensitize it for certain pathogens need more engineering. This example rates a TRL 3 or 4.
Novel Piezoelectric Materials
AMFitzgerald found additional active research on innovative piezoelectric materials as thin-film AlN and PZT are becoming well-established. More researchers are working on thin-film Lithium niobate (LiNbO3), thin-film electroactive polymers, different doping levels of Germanium in AlN, and experimenting with Scandium Germanium. As materials mature, new architectures are imagined using these materials. Most new piezoelectric materials are not ready to implement, as processes need more development to make them scalable. Thus, this research trend rates at TRL 2 or 3. However, potential applications may eventually be found in consumer electronics and IoT. Research was identified in a piezoelectric transformer for wake-up functions at the University of Illinois at Urbana-Champaign. Another example for a novel use of piezoelectric materials is a flexible, printed micro-speaker, presented by CEA-LETI in France.
The Trend from Silicon to….Paper?!?
A research trend is apparently emerging in a shift away from silicon to paper, at least in the presentations at the Hilton Head Workshops from 2004 to 2016. In 2004, 88% of papers discussed silicon-based devices, with around 15% considering silicon-based devices in 2016. The majority of new papers presented involve paper or plastic. IoT is also driving a trend towards very low-cost sensors.
General precedence shows that when markets put pressure on technology for cheaper prices, then technology moves to lower cost materials. For example, microfluidics, which is the engineered movement of fluid to flow through channels of 1 mm or less, has moved from silicon, then to glass, and on to plastic and paper. RFID moved from silicon to paper and plastic. Market forces have pushed technologies onto lower-cost substrates. It is highly unlikely that anyone will put a gyroscope on paper, however. Some sensors require silicon for dimensional stability and its semiconductor properties.
Paper does not need an expensive clean room, has a very large format substrate, high throughput, and ultra-low cost. Devices potentially driven to paper or plastic include sensors for temperature, humidity, pressure, selective gases, and point-of-care diagnostics. One research example is paper that can be sensitized to react to the presence of bacteria. A whole class of devices is evolving for low resource point-of-care diagnostics. Such paper-based devices are early stage, and therefore rated at TRL 1 – 2, as there is no scalable process for this pathogen detector yet. This research example comes from National Tsing Hua University in Taiwan, where they are impregnating paper with detection materials by hand.
Another example from Clarkson University in the U.S. is a paper sensor that detects food spoilage; essentially a gas sensor. Researchers impregnate paper with inorganic nanostructures that attract designated molecules. The inorganic nanostructures react with a color change proportional to the concentration of the gases detected from decomposition. No power sources are needed. Results include an instant readout. Potential applications encompass quality control of consumables, smart labels for food packaging, vaccines, or other perishables. However, researchers are still building this very early-stage product by hand.
Trends indicate that in 2020 we will see a full expression of simple piezoelectric devices. From that process technology will come more event-driven and battery-free sensors and the re-engineering of lighting and motion sensors. Many existing sensors manufactured using DRIE are going to piezoelectric. The 2030s may see paper and plastic technologies in ultra-low-cost sensors with disposable packaging, biodegradable sensors, smart clothing, point-of-care diagnostics, and large-format arrays. Sensors may be embedded in vehicle wraps, wall coverings, or rooftops. We can expect a notable shift away from capacitive MEMS designs that are dependent on precision deep silicon etch. Silicon-based sensor technology is at risk of stagnation due to waning research effort, and MEMS’ academic engine of new ideas is moving away from silicon.
Lynnette Reese is Editor-in-Chief, Embedded Intel Solutions and Embedded Systems Engineering, and has been working in various roles as an electrical engineer for over two decades. She is interested in open source software and hardware, the maker movement, and in increasing the number of women working in STEM so she has a greater chance of talking about something other than football at the water cooler.
[i] Lee, Sunggyu, editor. “Volume 5.” Encyclopedia of Chemical Processing, vol. 5, Taylor & Francis, 2005, pp. 3049–3059.
[ii] Dunbar, Brian. “Technology Readiness Level.” NASA, NASA, 28 Oct. 2012, www.nasa.gov/directorates/heo/scan/engineering/technology/txt_accordion1.html.
All other references: Fitzgerald, Alissa M., and Keith M. Jackson. “Emerging MEMS & Sensor Technologies to Watch.” MEMS Executive World Congress 2017. MEMS Executive World Congress 2017, 1 Nov. 2017, San Jose, Ca, U.S.A.