MEMS and compound semiconductors: Ready to take off?
11/01/2003
Overview
As the semiconductor industry appears poised to come out of its worst slump ever, two promising markets may be approaching the time when their potential is finally realized.
Global markets for microelectromechanical systems (MEMS) and compound semiconductors have offered great promise to chipmakers for years. Both categories of products face market and processing challenges, and the global economic woes of the past couple of years slowed down research that is needed to move processes forward, especially with MEMS.
 Figure 1. Global MEMS market is forecast to grow at a CAGR of nearly 16% through 2007, topping $8 billion. (Source: In-Stat/MDR) |
null
These systems, which combine a mechanical component such as a sensor with a traditional IC in the same device, have existing and potential functions in a host of applications that span major markets. While MEMS manufacturing shares processes to a great degree with IC manufacturing, significant differences and challenges also exist and must be overcome.
Growth of MEMS stalls
Market researchers peg the current overall MEMS market at somewhere between $3–$5 billion, depending on the source. The most mature market segment — accelerometers used in automotive air bags — may comprise up to $1 billion of that total.
Last year, the worldwide market for MEMS saw impressive unit growth of nearly 48%, but overall revenue growth of only 4.4%, according to In-Stat/MDR, which tracks MEMS markets. The Scottsdale, AZ-based market research firm is bullish on future prospects for MEMS growth; it forecasts MEMS revenues to grow at a compound annual growth rate (CAGR) of nearly 16% through 2007, with unit shipments growing at a CAGR of 26% (Fig. 1).
null
Data from other market research organizations differ, mostly due to inconsistent definitions of what is and is not a MEMS device. But all firms that study MEMS agree that the global market will show strong growth for years to come.
Market researchers have been predicting rapid MEMS growth for quite a while, but most markets and most individual applications have failed to develop as rapidly as expected. This is due to several factors that have combined to make the relative cost of most MEMS devices significantly higher than those of competing products, as well as the global economic slowdown of the past couple of years.
Automotive: Largest of many markets
Automotive, easily the most established market for MEMS devices, also has great growth potential due to its size and the many possible applications that have been identified (Fig. 2). Roger Grace, president, Roger Grace Associates, Naples, FL, estimates that more than 70 potential automotive applications exist.
Air bag accelerometers, which combine a sensor with an IC to first sense a collision and then deploy a protective airbag, are the largest single automotive application for MEMS. Analog Devices, Norwood, MA, supplies the majority of airbag accelerometers to auto manufacturers.
"Accelerometers will continue to be the growth driver" for automotive MEMS, says Marlene Bourne, senior analyst for MEMS markets at In-Stat/MDR. Bourne points out that growth in the accelerometer market may be spurred by tests conducted in June by the National Highway Traffic Safety Administration (NHTSA), which examined the results of side impacts to a vehicle being struck by an SUV. Bourne believes test results could produce a mandate that side "curtain" air bags be installed in new vehicles.
Another developing automotive application for MEMS is tire pressure monitoring. A MEMS device can directly measure tire pressure and report any instance of under-inflation to the driver.
Though automotive is the single largest potential market, many other opportunities exist for MEMS in biotechnology, wired and wireless communications, military applications, information technology, microfluidics for chemical analysis, consumer products, and various industrial applications.
Processing challenges
Manufacturing technologies for various MEMS devices differ just as much as MEMS end markets. "If you compare MEMS devices for inkjet heads and accelerometers, there's almost no common manufacturing steps, and the end functionality is completely different," said Andy McQuary, executive VP of marketing and general manager of US operations for STS, which supplies plasma etch systems used in MEMS manufacture. "But both are labeled MEMS devices."
Another factor that has held back broad use of MEMS devices is a design mindset that has focused on feasibility rather than manufacturability. Thus, MEMS producers are able to demonstrate quickly the feasibility of a device, but ramping up production to the point of economic viability is very difficult and costly. Large volumes often do not develop, and viable markets simply don't materialize.
Significant MEMS manufacturing challenges include:
- Low volumes, low level of automation. Because the vast majority of MEMS applications are not yet mature, with many far from being ready for commercialization, manufacturing costs are still far too high.
- Inconsistency of substrates. Substrates used for MEMS devices vary greatly in composition, size, weight, and other characteristics. MEMS devices can require the processing of not only silicon but also compound semiconductor materials such as gallium arsenide (GaAs), indium phosphate (InP), and other materials. The enormous variation in substrates "imposes challenges for equipment suppliers," says Peter ten Berge, product manager at ASM Lithography. "If there's a chance to use silicon, then people will do so," even if silicon may not always be the optimum choice, says ten Berge.
- Handling. In addition to variations in substrates, MEMS devices also present handling challenges because many contain device layers on both sides, unlike standard ICs. Also, MEMS are three-dimensional, often with delicate micro-bridges, creating further handling challenges.
- Metrology. Varying substrates and the need for double-side measurement present obvious metrology challenges compared to standard ICs.
- Lack of standards. While there has been some recent progress on international standards for MEMS, the lack of recognized standards has held back development and kept manufacturing costs high. To date, companies that have been successful in MEMS have developed their own standards to some degree, but MEMS manufacturing is still characterized by far too much custom work. "For mass production, we need standards that will let the industry evolve," says Chris Constantine, CTO of Unaxis Semiconductors, a supplier of deep silicon etching and piezo deposition equipment for MEMS.manufacturing.
- Packaging. MEMS pose packaging challenges that the IC industry simply has not seen before. In order to work, the mechanical component of a MEMS device must be exposed to its environment but the circuitry must be protected. Nearly all MEMS applications require specialized packaging, adding significantly to cost. Most estimates peg packaging as comprising at least half of the overall cost of a MEMS device.
- Processing. Special processes are often required for MEMS devices, including thick resists, deep etching, and etching of sacrificial material under microbridges. This may require tools modified from those used in conventional CMOS processing. (More processing details for MEMS will appear in the December issue of Solid State Technology.)
Kevin Fitzgerald, Executive Editor
Compound semiconductors
Compound semiconductors satisfy the need for faster ICs needed to improve performance in cellular communications and telecommunications, as well as other leading-edge applications, along with other uses as light emitters (LEDs or lasers), plus some specialized applications for the military.
Materials for compound semiconductors exhibit higher frequency, better signal reception, superior signal processing in congested bands, and greater power efficiency than pure silicon devices. Compound semiconductors also can be "bandgap engineered," enabling them to be optimized for a given application.
Compound semiconductors include gallium arsenide (GaAs), indium phosphate (InP), and gallium nitride (GaN). GaAs and other III-IV compounds dominate the overall compound-semiconductor market. Silicon-germanium (SiGe) also competes with these devices, and using strained crystal lattices further enhances their speed and frequency capabilities.
Like the overall IC market, the global market for compound semiconductors grew strongly in 2000, but slipped back in 2001 and 2002. IC Insights, a market research firm in Scottsdale, AZ, forecasts strong growth for compound semiconductors in coming years (see figure).
Cell phones drive GaAs growth
"Mobile phone applications will continue to provide strong growth for GaAs, especially in the transmitter function," says Vic Steel, VP, research and development, for RF Micro Devices, a manufacturer of radio-frequency ICs for wireless communications based in Greensboro, NC. "GaAs frontends also will be critical in dual-mode WLAN systems. As this market grows, so will demand for GaAs."
SiGe is forecast to gain market share from GaAs devices in coming years. SiGe offers some advantages over GaAs, according to IC Insights. SiGe is compatible with silicon IC processes, thereby offering the economic benefits that come with high-volume manufacturing. Also, SiGe offers the ability to combine the advantages of CMOS technology in low-power, high-density digital signal processing with the high speed offered by the SiGe heterojunction bipolar transistor (HBT) in a BiCMOS process.
Because of these advantages, IC Insights forecasts that revenue from sales of SiGe devices will increase from 13% of the overall compound semiconductor market in 2002 to 33% in 2007.
InP has attracted interest because it may be able to extend the performance of ICs beyond the capability of other semiconductor materials. InP requires a lower supply voltage than GaAs, and InP HBTs exhibit better noise and linearity performance compared to GaAs HBTs.
From the perspective of processing equipment, most GaAs production uses HBT technology. "The key to improving technology is in the epi," says RF Micro Devices' Steel. "Any improvement in molecular beam epitaxy or metal-organic chemical vapor deposition is quite valuable."
Compound semiconductor foundries
Over the past few years, a dedicated compound semiconductor industry has emerged.
Foundry services for SiGe devices — a newer segment of the overall compound semiconductor foundry market — include major foundry players such as IBM Microelectronics, Chartered, Jazz Semiconductor, TSMC, and UMC. All SiGe services offered are BiCMOS processes targeted at mixed-signal devices in communications applications, both wired and wireless.
Mixed feelings exist about the growth potential for compound semiconductor foundries. "GaAs fabs are relatively inexpensive compared to Si CMOS or BiCMOS," says Steel. "This reduces somewhat the value of the foundry to high-volume applications. Foundries provide opportunity for companies to experiment, but high-volume production will most likely go to captive suppliers." Steel adds, however, that compound semiconductor foundries will handle overflow production capacity when needed.
 The global market for compound semiconductors is forecast to surpass $4.5 billion by 2007. (Source: IC Insights) |
null
Another recent trend is the move from 4-in. wafer toward 6-in. wafers, in an effort to improve yield and manufacturing economics. Moving to 6-in. wafers more than doubles the amount of die/wafer, without significant increases in cost. Major producers are heading this way, including RF Micro Devices, which also continues to run 4-in. wafers.