MEMS: One-of-a-kind challenges
07/01/2003
Solid State Technology asked industry experts for their views on the unique challenges and opportunities in the MEMS-manufacturing arena.
Process challenges for MEMS device manufacturing
David A. Markle, senior VP and CTO, Ultratech Stepper Inc., San Jose, California
Microelectromechanical systems (MEMS) are highly miniaturized devices that integrate diverse functions — including possibly fluidics, optics, mechanical elements, sensors, and electronics — on a single silicon substrate. Because the manufacturing methods are similar to those techniques used to manufacture traditional ICs, a number of chipmakers and equipment providers have moved into the MEMS arena, and it's clear that opportunities abound. Market research firm In-Stat/MDR estimates revenues from MEMS products, which were $3.9 billion in 2001, will more than double by 2006, reaching $8 billion. Broad adoption, however, hinges on overcoming some key process challenges.
 David A. Markle |
Often, the development of even the most mundane MEMS device requires a dedicated research effort to find a suitable process sequence. This leads to the industry's foremost challenge that prevents large-scale manufacturing of MEMS devices: the lack of standardization in MEMS fabrication process technologies due to the high degree of customization. For each new device, the fabrication process employed is dependent on the particular device and its intended application, so the wheel must be continually reinvented, making economies of scale difficult to attain.
The delicate and sometimes esoteric materials required to manufacture MEMS devices pose another processing conundrum. In creating MEMS, it is common to process not only silicon, but also quartz, sapphire, indium phosphide (InP), gallium arsenide (GaAs), and other compound semiconductor materials. Some are expensive and difficult to grow into large crystals, and can require handling of small (2- or 3-in.) substrates and even broken pieces. No rules apply when it comes to photoresist, which can be thick or thin, negative, positive, g- or i-line sensitive, or even dual-side coated. The latter must be handled very carefully to protect the backside coating.
Packaging of MEMS devices and systems also needs to evolve considerably. The diversity of MEMS devices, and their (typically) continuous contact with the environment, makes MEMS packaging more challenging than IC packaging. Also, because MEMS devices are not planar, but have dimensional, topographical characteristics, the use of pick-and-place systems for assembly and packaging becomes implausible. As with the fabrication process, the packaging approach must be customized for each new device.
As these issues are addressed and MEMS manufacturing shifts toward production volumes, foundries and companies planning for higher production volumes and lower product average selling prices will need to confront process yield improvement concerns. For example, as critical dimension and alignment specifications dip below the capabilities of off-contact aligners, stepper (projection) lithography systems are required to meet new product design rules. Contact aligners can adversely affect product yields by making physical contact and introducing defects onto the wafer. Stepper technology will eliminate the high mask-cleaning and replacement costs associated with contact aligners.
Recognizing that the manufacturing industry needs to address these challenges in a more formalized way, Semiconductor Equipment and Materials International (Semi) has recently formed a special interest group (the International MEMS Steering Group) focused on MEMS and microsystem technology. Some key tasks within the group's charter include disseminating market information and statistics; promoting customer awareness of MEMS technical, regulatory, and business issues; solving common problems and eliminating duplication of resources; and promoting focused and efficient international standards development efforts.
Clearly, the industry has some formidable challenges to overcome before high-volume manufacturing of MEMS for broad applications becomes viable. Equipment and materials suppliers, as well as designers, chipmakers, packaging/assembly providers, and others throughout the supply chain, must collaborate to remove these barriers and ensure that the promise of MEMS technology is fulfilled.
For more information, contact David Markle at [email protected].
Metrology challenges in MEMS manufacturing
F. Michael Serry, applications scientist, Veeco Instruments, Santa Barbara, California
MEMS manufacturing is benefiting from the many solutions developed to measure miniature parts across the industrial spectrum. In particular, benefits accrue from the mature tools that address metrology problems in microelectronics. The diversity of MEMS devices creates an abundance of new manufacturing and metrology challenges requiring novel solutions. At least two types of MEMS metrology pose new hurdles: detecting/measuring voids and characterizing/measuring motion.
 F. Michael Serry |
Manufacturers must be able to detect the presence and measure the dimensions of voids beneath the surface nondestructively, that is, without removing the surface layer. This is not a new challenge. With MEMS, however, mechanical defects take on an elevated importance because they can significantly affect the performance of moving parts, like the hinges that support free-standing structures. Also, while in other miniature technology, including microelectronics, voids are most often mechanical defects in MEMS, some voids are created by design. For example, surface micromachining of MEMS relies on sacrificial layer removal to create space (voids), often beneath moving parts.
To nondestructively characterize and measure subsurface voids on a scale relevant to MEMS performance, new methods (or modifications to existing methods) are needed, and some are being developed, including various acoustic (ultrasonic) imaging techniques. But acoustic imaging lacks spatial resolution due to diffraction effects. As the device size and proximity of adjacent surfaces in MEMS decrease, the spatial resolution of void detection and measurement methods needs to improve. For example, a near-field acoustic imaging approach that can reduce or substantially eliminate diffraction effects may improve the spatial resolution. On the other hand, as the depth of voids and undercuts increases, the depth of field may need to increase, too — ideally without compromising the resolution.
The second, more urgent need is characterizing and measuring motion, which is ubiquitous in all MEMS. But motion is all but absent, and so is the corresponding metrology infrastructure to address it in microelectronics.
The nature of the motion and the moving object vary from system to system in MEMS — from the 3-D motion of a liquid-borne particle in a microfluidic channel, to the tilt of a micromirror and the in-plane translation of a comb-drive element. The frequencies and the dynamic ranges are also different from one system to another and vary by orders of magnitude. Measuring motion in MEMS requires new, diverse approaches.
Most commercial products available today for measuring motion in MEMS rely on purely far-field optical methods, meaning they shine light directly onto the moving device and detect the light in the far field reflected from the device. This is true regardless of the source of the light — coherent laser light, white light, or incoherent monochrome light — and independent of the nature and details of the detection method and technology.
As the in-plane dimensions of moving devices shrink to near and below the wavelength of light, purely far-field optical methods will be inadequate to measure the motion; near-field optical and hybrid techniques may then be necessary. Such an evolution in instrumentation has already occurred for measuring the static dimensions of devices, including MEMS. The AFM, for example, is a hybrid instrument in the sense that it uses far-field optical detection together with a highly localized mechanical probe — the AFM tip — to measure features with deep subwavelength resolution.
Two main classes of MEMS motion metrology tools exist in the market, one very different from the other: stroboscope-illuminated interferometry and laser Doppler vibrometry (LDV). Stroboscopic interferometry captures a temporal sequence of images that cover an area, or the whole, of a MEMS device as the device moves. LDVs measure the Doppler shift in the frequency of the (coherent) light reflected from a single point on a device as the device moves in such a way that the motion has a component perpendicular to the laser light's propagation direction.
In the current generation of tools, measuring the transient mechanical response of most devices takes longer using stroboscopic interferometry. On the other hand, detecting in-plane motion with LDV is all but impossible. Also, in order to capture data over an area — rather than on a single point of the device — LDV technology relies on in-plane scanning of the laser beam, making it slow to capture the motion of the whole device. Future generations of these tools may combine the two technologies to offer MEMS manufacturers tools with a broader range of applications.
MEMS moving parts are often in constant or intermittent contact with nearby surfaces. The nature of contact is, again, quite different (for example, a micromirror array vs. a comb-driven translation stage). Contact, adhesion, deformation, and friction in MEMS need to be measured and analyzed. Metrology solutions that address tribological issues exist, and some are already in use in MEMS research and development, and perhaps in manufacturing and production as well. These include instruments based on the AFM and micro-indenters.
Future tools for measuring and characterizing contact and contact-related phenomena will have to work with MEMS parts moving at high speeds, up to several tens or even hundreds of meters/sec. Such speeds will exist, for example, in the disk-drive sector, which will be incorporating new generations of MEMS elements into the head to help fine-position its read-write element for better control of track/in. density, thus improving recording density on magnetic media.
One area of MEMS motion metrology that needs significant investment (with great potential return) is interfacing measurement and data analysis software with modeling and parameter extraction software. Optimizing MEMS designs in a timely fashion will increasingly demand that measurement and analysis data be readily portable into modeling and parameter extraction software, which is often made by companies other than the measurement tool manufacturers. Measurement tool suppliers will need to identify and interface with the suppliers of modeling and parameter extraction software to ensure that sense can be made out of the measurement data. To this end, those companies that do both well (perhaps in partnerships) will have a competitive advantage.
Industry-wide roadmaps will play a much less important role in MEMS manufacturing, including in metrology, than they do in chip manufacturing. This is because no generic MEMS device is equivalent to the transistor in microelectronics. This is also the reason most MEMS manufacturers and their suppliers are likely to concentrate on submarkets within the much broader MEMS category. Unlike in microelectronics, many MEMS startups with new device designs and functions may not be able to find metrology tools that address their needs and may be forced to come up with their own technology solutions in-house first. Furthermore, that in-house technology may or may not be picked-up by a metrology solutions manufacturer, and introduced into the market as a commercial product.
For more information, contact F. Michael Serry at 112 Robin Hill Road, Santa Barbara, CA 93117; [email protected].