MEMS: A View from Aerospace -- The potential of MEMS in mil/aero applications appears almost limitless, but, first, the dreaded ilities must be overcome.

Micro-electro-mechanical systems (MEMS) are the next big technology push. The corresponding military/aerospace (mil/aero) market pull targets space, the battlefield, surveillance, communications, and weapons management and guidance. Size, function and potential low costs are powerful drivers, but barriers to MEMS' acceptance do exist.

Mil/aero users are conservative and demand evidence of producibility, testability, functional quality, reliability and several other ilities. Their products must not fail in service; no returns, no refunds, no replacements...people die. New product development and demonstration projects must satisfactorily resolve all of these ilities.

Many MEMS processes, each with unique capabilities and limitations, are currently being developed. These processes include: wafer-scale silicon fab; deep-x-ray lithography for high-aspect-ratio; LIGA processing to create and release a molded shape; multi-layer diffusion bonding; powder metal or ceramic sintering; micromachining (Figure 1); and micro-pen direct-write of thick-film conductive inks.

MEMS include sensors for temperature, pressure, humidity, chemical and medical applications; fluidics such as pumps, dispensers, flow-meters and valves; inertial devices such as accelerometers, position sensors and gyroscopes; actuators of all types including rotational, steppers, drivers and positioners; electrical switches; optics such as mirrors, lenses, gratings, filters, choppers, detectors, switches and wave-guides; and thermal management devices such as heaters, coolants and heat pipes. These MEMS will be deployed in missiles, satellites, spacecraft, ships, ground vehicles, and on battle warriors (Table 1).1

Barriers to MEMS

Obstacles to deployment of MEMS products in mil/aero applications do exist. Most of the obstacles are technical, some involve cost, some are political, all are interrelated and all are real.

MEMS Novelty

MEMS is bleeding-edge technology, but new is not necessarily better. Heritage is crucial in mil/aero applications and essentially non-existent with MEMS; almost no similar products exist to build on, ruggedize, redesign or shrink. Few military specifications and standards exist for MEMS. Industry specifications, including IPC, SEMI, ASTM and JEDEC, are only recently being developed.

MEMS devices and concepts are truly alien. Their mechanical performance and reliability are influenced more by surface chemistry and atomic forces than by gravity and bulk strength. We cannot simply scale down the tolerances, take a hit in yield, pay our suppliers a bit more, and buy high-powered microscopes. MEMS prototype fab shops are enthusiastic but are not on any qualified-supplier list. Certain MEMS semiconductor processes have a reassuring mil/aero heritage, but the specific MEMS device will be beyond any qualification limits.

Performance

That the mil/aero product must perform is obvious, but no texts, databases or experienced teams are available to rely on. Further, MEMS performance depends on unique properties and interactions, far beyond well-known principles and data. Gears, sliders, hinges, latches, detents and reflective surfaces all depend on elements (burrs, planarity, taper, roughness, freedom from in-service warpage) that we long ago considered manageable no-brainers.

One example of a "simple" issue is the wear of rubbing surfaces, which can be a major MEMS failure mechanism (Figure 2). The measurement and principles of stiction,2 adhesion, molecular lubrication, surface degradation and wear-resistant materials are critical elements in any MEMS actuator or rotating device.

For any mil/aero application, the product must demonstrate very high likelihood of nominal performance at the most severe limits of its environmental history and conditions, over its design lifetime, with safety factors and performance margins. No repair or replacement is allowed.

Generally, any MEMS must endure some extreme conditions (Table 1). The specific requirements will depend on various factors such as mounting on a vehicle, shielding, isolation or thermal management. All requirements must be detailed early-on to ensure that the actual deployed MEMS will demonstrate the necessary reliable performance.

The Dreaded ilities

The easy part is getting something to perform; the hard part is to nail the lurking ilities. The mil/aero user needs data on reliability, procurability, producibility, availability, maintainability, verifiability and testability for the specific device, package, mounting and application. But, these ilities may not be directly specified. The requirements may simply be that each and every unit, today and 10 years from now, must perform properly, no matter what.

Producibility and manufacturability are the banes of infant technologies. A complex system, a MEMS will function properly only if many factors are tightly controlled at several points along the supply/process chain. A robust design is crucial. For MEMS especially, DFX means design for all of the ilities.

Depending on the design and application, every MEMS requires a certain flatness or specific curvature, bending force/deflection, hysteresis, actuation speed and repeatability, rest positioning, response times, optical alignment, finish, resistance and capacitance, and, most important, the uniformity of all elements in the device over a wide range of conditions, in all devices within the lot and from lot to lot. These factors depend on dimensional control, spacing and gaps, polymer cladding, deposition, intermetallics; surface and interface effects; polishing and etching effects; plating and deposition; surface oxides and roughness; residual stresses, creep, relaxation, warpage, thermal deformation; surface tension, stiction, friction, resonance; particulate and molecular contamination; opticals such as refractive indices, transmissivity and reflectivity; and electrostatic forces.

Survivability describes the device's function during and after an external environmental exposure or event. Often the nominal device will survive the event, but a borderline device that passes original performance tests will not survive. Specific testing protocols, such as Highly Accelerated Life Testing (HALT), could reveal the failure, its environmental trigger and its design weak link.

Testability and verifiability of nano-scale functions and materials properties often are critical in MEMS. The subtle, but troublesome, attribute must be sensed, visually, by scanning electron microscope (SEM), x-ray or other means. Often, final properties necessitate up-stream, in-process measurement and control. Is there a "known-good-MEMS" protocol? If a particular device is bad, can it or should it be fixed? Does the inspection itself jeopardize the device's quality?

Procurability and availability

From prototype to production is a huge step, especially for infant technologies like MEMS. The supply chain must be defined, committed, exercised and controlled. Distressingly, major mil/aero projects often amount to only tiny supply-chain line items. The mil/aero user must see a clear multi-year commitment to a specific materials and processing supply chain.

Reliability is a critical MEMS issue. The user must be assured that the specified projected failure rate over thousands, perhaps millions, of cycles and units will be satisfied. The mil/aero market and recently the high-end commercial market have evolved test protocols-HALT, environmental stress screening (ESS), highly accelerated stress screening (HASS), databases, failure mode and effects analysis (FMEA) and models-to detect failure modes and weak links, to define failure rates of nominal and extreme products, to correlate failure modes and times with dimensional and other signature characteristics.

Another reliability element involves maintainability, in-service, over time. Credible cost-benefit trade studies on preventive maintenance, retesting, re-placement and recalibration are necessary. Burn-in and screening tests can weed out workmanship-defective units, without subtracting significant life. This approach demands clear reasoning, requires significant experience with the technology and application, and probably will not be a major protocol for high-volume, low-cost MEMS products. The focus must be robust design, not weeding out marginal units.

Life test statistics

Hypothetically, suppose an application requires no more than one failure in a population of 10,000 units at 100,000 cycles. You report a test of five prototypes, ranging in life from 110,000 to 200,000 cycles. Sounds promising, but the results show that we can be 95 percent confident that, at 100,000 cycles, 200 to 6,000 units will have failed (Figure 3). Not nearly good enough. This result is based on a Monte Carlo run, using a reasonable assumption of a beta = 5 Weibull distribution.

Suppose you return six months later with data from a N=10 run, an even tighter distribution, citing a mean life of 200,000 cycles, twice the life target. Figure 4 confirms the F50=200,000, but also shows that 20 units out of 10,000 will have failed at 100,000 cycles. Still not good enough! These analyses, enabled by good software with confidence interval and Monte Carlo capabilities, can show what this life data really means. Present all sample and life data and reduce the data properly, using the available software, including confidence intervals extended to the desired ppm or target life levels.

Costs

Business cost elements and models are rudimentary. Predictions are shaky. Development will be expensive, and invention costs, typically impossible to predict, will be incurred. As a result, funding, cost-benefit trade studies and business partnerships will be hampered.

The individual user might not be aware of the importance of the ilities issue. Often a nominal demonstration or two is deemed enough to set the MEMS lab and user onto the primrose path. Careful! Listen early on to the ilities folks (supply-chain, material and properties engineers, test engineers, prototype techs and mission statisticians). Develop parallel efforts, back-up plans and contingency funding.

Resources: The Good News

Many commercial MEMS, such as air-bag accelerometers and ink-jet printers, are clearly more reliable than earlier large electromechanical designs. Recent literature describes extended life testing of actuation and switching cycles and robust designs surviving extreme shock/environmental conditions. The semiconductor industry offers important process information, reliability analysis and FMEA modeling. Many metrics are phrased in terms of electrical parametric deviations, but the process effects are relevant to MEMS physical performance. In his book, Madou reviews nano-scale principles and fundamentals.3

Sophisticated measurement tools are being developed. Laser interferometry documents the shape and position of mirrors and accelerometer structures at varying frequencies, as well as out-of-plane displacement of cantilever beams, revealing unexpected bending modes and hinge stresses and deformations. Stroboscopic video microscopy characterizes rocking and out-of-plane motions of MEMS gyro structures. Raman spectroscopy detects and measures stresses in silicon membranes.

The commercial off-the-shelf (COTS) revolution has enabled up-screen and burn-in testing, FMEA, modeling software, accelerated tests and ruggedizing, all transferable to mil-aero MEMS deployment. Commercial MEMS sensors, actuators and controllers are being targeted for demanding under-the-hood, down-hole and medical applications. These applications increase awareness of harsh no-fail requirements, facilitating transition to mil/aero deployment. Personal electronics applications have provided an extensive infrastructure of miniature-scale packaging, verification tools, mounting and shielding concepts, plus supply-chain, personnel and design resources.

Useful military specifications (mil-specs) are available. MIL-STD-883 is basic for microelectronics testing. MIL-HDBK 217F covers reliability modeling, statistics and mean time between failure data on electromechanical devices. MIL-45208 and -9858 describe inspection and control. MIL-STD-810 describes accelerated life tests, scaling and life-cycle environmental histories for military hardware. MIL-E-5400T covers manned aircraft requirements. MIL-STD-1546B and MIL-STD 1540D cover space/launch vehicles. MIL-HDBK-343 classifies risks, and MIL-M-28787 provides test protocols for electronic modules. NASA NHB 5300 is the core quality/reliability document.

Design bureaus and MEMS software are proliferating. MEMS prototype shops, spun off from technology centers, provide specialized design, process and application services. These shops can help work around process limitations and shorten prototype cycles. Further, these start-ups help identify development activities, sequences and milestones, plus projected times and costs. Characterization and analytical labs that arose to handle integrated circuit (IC) and micro-packaging now support MEMS.

Many academic and government groups and consortia are emerging to support MEMS development; most are Web-accessible:

- Sandia National Laboratory, www.sandia.gov

- DARPA, www.darpa.mil

- Pacific Northwest National Laboratory, www.pnl.gov

- Lawrence Livermore National Laboratory, www.llnl.gov

- Stanford University, www.stanford.edu.

Further information on MEMS may be found by viewing the Websites of the following organizations: Ardesta Co. (www.ardesta.com), MEMX Inc. (www. memx.org), Jet Propulsion Laboratory (www.jpl.nasa.gov), Langley Research Center (www.larc.nasa.gov) and Aberdeen Test Center (www.atc.army.mil).

Conclusion

Technology push is evident. MEMS labs and start-ups are demonstrating attractive concepts and prototypes. Consortia are being formed to provide synergism and critical mass. Trade publications and societies are covering MEMS. Researchers are documenting fundamental and applied process development, materials characterization and performance testing.

The mil/aero community is flexing its market pull. MEMS' potential appears almost limitless. Scores of U.S. Department of Defense, Air Force, Army and Navy contracting authorities actively support MEMS concepts, planning, partnering and funding.

Barriers to rapid deployment include the absence of heritage and a dearth of demonstrated ilities. Life-test data interpretation is a key element. These realities rarely appear in product performance requirements. Robust designs and clear requirements are required. The user and MEMS developer must highlight the ilities in project and funding plans.

Resources are becoming available. HALT and HASS protocols are in place. Fundamental characteristics are being identified and explored in the labs. Databases of nano-scale materials' properties and environmental effects are being created. NASA and military specifications on methodology, test protocols, reliability modeling and background data are still available. COTS implementation increases awareness within the supply-chain and test labs of reliability and ruggedizing practices. Test labs show growing competence in wafer-scale metrology and chemical/electrical characterization. Design tools and analytical models are being developed and proven.

The best news is that many MEMS products, particularly in demanding commercial applications, are already showing better performance, envelope, reliability and cost than their older counterparts. We now see millions of cycles and lifetimes of excellent performance in fluidic, optical and inertial applications.

Mil/aero designers and managers must clearly describe performance requirements, including long-term reliability and producibility. Close consultation with the customer is crucial. Successful deployment of MEMS products in mil-aero is at hand. The electronics industry has been successful with ICs; the same success can be had with MEMS.

Acknowledgments

The author acknowledges the support of the reliability staff at Lockheed Martin Sunnyvale, especially Dave Himmelblau and Grant Inouye; as well as Thomas George of JPL and Dannelle Tanner of Sandia.

References

1. NRL. Optics and MEMS. NRL report #NRL/MR/6336-99-7975. Washington, D.C.; Naval Research Laboratory.

2. Tanner, D. (2000). Reliability of surface micromachined microelectromechanical actuators. Proceedings of the 22nd International Microelectronics Conference. Yugoslavia, May.

3. Madou, M. (1997). Fundamentals of Microfabrication. Boca Raton, FL: CRC Press.

Tom Clifford is a process engineer with Lockheed Martin Space Systems Co., Sunnyvale, CA; e-mail: [e-mail].

http://www.circuitsassembly.com

Copyright [copyright] 2001 CMP Media LLC