Editor's note: This article originally appeared in the Winter 2009 issue of MICROmanufacturing magazine.
Microelectromechanical systems are very small, which is one reason they’re a big challenge to design. Designers are meeting the challenge in a variety of ways, including application of new engineered materials, software tools and testing techniques.
A MEMS-based mirror produced by Microvision Inc. Photo courtesy Microvision Inc.
What’s more, designers are employing strategies to expedite the process of producing new MEMS devices, such as minimizing component interactions, making use of generic designs and modifying existing MEMS designs to improve manufacturability.
Some MEMS designers are also benefiting from a broader view of what can be a narrowly focused process. These designers are paying attention to critical areas like assembly and packaging, as well as reaching out for important input from colleagues outside the design office.
Material advances
Perhaps nothing affecting MEMS design is experiencing more rapid
and profound change than materials. New engineered silicon-on-insulator
materials include through-silicon vias (TSVs) to significantly reduce device
footprints. TSVs connect to external devices via solder bumps, eliminating the need to make electrical connections with wires and bond pads that can clutter chips, explained David Buckley, vice president of sales and marketing for Micralyne Inc., a MEMS foundry in Edmonton, Alberta.
Micralyne also works with proprietary engineered materials known as Glancing Angle Deposition films. GLAD is a patented, thin-film fabrication process developed by Micralyne and the GLAD Laboratory at the University of Alberta. GLAD differs from traditional thin film deposition techniques by using highly oblique deposition angles that produce tilted-column nanostructures on substrates. By creating millions of these columns on a substrate, GLAD greatly increases the active surface area, making it more sensitive and, thereby, increasing its ability to detect various substances. This can be an advantage in life sciences and other applications, Buckley noted.
HT Micro impact switch device, a passive omnidirectional acceleration switch that measures 2mm × 2mm × 1mm, was developed in conjunction with the U.S. Army for fuzing applications. Photo courtesy HT Micro Analytical Inc.
MEMS material options have also expanded to include silicon carbide and diamond, which can withstand higher temperatures than regular silicon, according to Keith Ortiz, manager of MEMS technologies for Sandia National Laboratories in Albuquerque, N.M. This makes them suitable for devices such as sensors that measure temperatures in the hot compartments of jet engines.
On the downside, Ortiz added, silicon carbide and diamond are less reactive than silicon, making them difficult to etch during a MEMS-fabrication process.
As for metals, MEMS designers are moving away from traditional monolithic materials and toward alloys, noted Paul Rubel, vice president of design and product development for Innovative Micro Technology (IMT), which offers contract MEMS manufacturing and design services in Santa Barbara, Calif. Unlike monolithic materials, gold, copper and nickel alloys offer long-term resistance to creep, as well as greater conductivity and hardness, added Rubel.
At HT MicroAnalytical Inc., Albuquerque, an electrodeposition process is used to produce metal alloys for MEMS devices. Research has shown that electrodeposition can alter the microstructures of materials in order to change mechanical characteristics such as yield strength and fatigue resistance, said Todd Christenson, HT Micro’s vice president and chief technology officer.
Christenson said that one metal alloy produced in this manner offers yield strength of more than 1GPa and can handle more than 10 million cycles at a stress of more than 1,000 MPa. HT Micro uses the alloy to make microminiature flexures for impact switches.
In addition to metals and silicon, MEMS designers are taking advantage of the special properties of polymers. For example, Rubel said designers are beginning to use polyimides for structural components of complex 3-D MEMS devices that would be more difficult to make using silicon or metal. In addition, he said, SU-8 is becoming more common in the fabrication of MEMS parts. This photo-imageable polymer becomes a hard, glass-like structure once cured, making it an effective structural material.
Other polymers have what it takes to be good actuator materials.
Electro-active polymers can undergo large amounts of deformation, making them suitable for artificial muscle actuators, according to Dan Popa, an assistant professor of electrical engineering at The University of Texas at Arlington.
Another new actuator material is PZT, or lead zirconate titanate. A piezoelectric material, PZT changes shape when an external electric field is applied. For micro-actuator use, PZT offers high precision and force output but very small displacement, Popa said. It also allows conversion of displacement into voltage for energy-harvesting applications.
Software and MEMS design
Like their counterparts in other fields, MEMS designers rely heavily on software for a variety of crucial tasks. Among other things, the latest MEMS software can help engineers optimize designs to achieve their goals.
“If you have a device that heats something, [software] can help you figure out where to place the heater to get maximum efficiency,” said Mary Ann Maher, CEO of SoftMEMS LLC, a MEMS software developer in Santa Clara, Calif.
MEMS Xplorer from SoftMEMS is said to have an intuitive user interface and provides MEMS-specific capabilities that reduce mask layout time. Photo courtesy SoftMEMS LLC.
No software task is more important to the MEMS design process than simulation, which is complicated by the fact that even relatively simple MEMS designs can involve a number of different physical domains. MEMS designers deal with these situations by using so-called “multiphysics” simulation software from companies such as ANSYS, Comsol, Coventor and IntelliSense.
Even with this special software, however, it can be difficult to perform a multiphysics simulation. In some cases, “you would need a Cray computer to do all the calculations,” IMT’s Rubel said. To simplify the process, designers sometimes break up the job into smaller tasks. When designing some electromagnetic devices, for example, IMT does separate magnetic simulations and then inputs that data into actuator simulations.
“If you start by trying to do everything at once, you’ll get really confused. Start with a simple model,” advised Micralynes’ Buckley, who defines a simple model as one that takes into account just one aspect of the design. MEMS devices are essentially “big mechanical structures made very small,” he added, so one way to start is by performing a relatively simple mechanical analysis of the structure using finite element analysis (FEA). The mechanical analysis can be followed by separately modeling other physical aspects of the design. Finally, all the separate models can be combined into the multiphysics model.
To simplify matters further, MEMS designers may choose not to start with FEA, which can involve complex meshing and yield detailed simulation results. Instead, University of Texas’ Popa said, designers can make “lumped model approximations” that don’t involve FEA. Lumped models break complex systems into discrete components and approximate the behavior of the components. These approximations can then be tied together to show how the entire system will work.
The important difference between lumped model approximations and FEA is the number of nodes in the models. For example, an FEA model of a cantilever might have 100 or even 1,000 nodes, Popa said. In such cases, “if you wanted to look at 1,000 cantilevers with FEA, it would be prohibitively complicated and time consuming.” By contrast, he noted, a lumped model can approximate a cantilever with just one node, greatly simplifying the task of studying the behavior of 1,000 cantilevers at the same time.
Physical testing
Whatever method is employed, the key to a good simulation is using accurate material properties. But in many cases, obtaining these properties is not simply a matter of looking them up somewhere.
“In the MEMS industry, there are a lot of thin films and other materials being used that may not be well characterized,” SoftMEMS’ Maher said. “So designers need good test structures to make sure they’re putting accurate material information into their simulations.”
Shown is a MEMS optical mirror produced by Micralyne using Micragem, which is a four-mask lithography MEMS process. The starting point is a 500µm-thick glass wafer. Photo courtesy Micralyne Inc.
Test structures help designers evaluate both the structural behavior and the reliability of new materials. These test structures are much simpler than the MEMS devices being designed. For example, Rubel said, a designer might make a cantilever beam out of a certain material and then subject it to electrostatic force to see how much the beam deforms. Creating a simple beam to obtain this information is much easier and faster than building a prototype of the entire device in order to run the test.
Testing also can be the best way to obtain other types of information critical to MEMS design. Consider a designer who must determine the damping effects in an impact switch in order to produce a certain response time. One way to get the information is to use a computer simulation. But modeling an entire complex device structure “can take a long time,” Christenson said.
So instead of going through a lengthy simulation process, he and his colleagues at HT Micro create cells where they use tests to obtain empirical information about damping behavior. If the fabrication process is relatively short and simple, test cells can yield key design data faster than simulation, he said.
Similarly, Micralyne uses simple “short loop” experiments to help eliminate uncertainties. For example, suppose a MEMS design calls for an embedded structure 1.5µm wide, with an accuracy of ±50nm. To see if this design is manufacturable, Buckley said, “you can run the full fab process, which may take 8 weeks. Or you can just do one layer to see if you can hit those tolerances, which may take 2 days and allows you to provide rapid feedback to the designer.”
Simplifying the process
Just as MEMS devices can be complex, so can the processes used to design and manufacture them. But there are steps designers can take to simplify these processes, thereby saving time and money—not to mention trouble. For instance, they can minimize or eliminate component interactions such as pneumatic cross-talk so that the components can be considered independent of each other, which simplifies design.
At IMT, complications result when customers insist on the use of materials and processes that are unfamiliar to the company. These customers would be better off opting for one of IMT’s many “platforms,” which are generic or base MEMS designs, suggested Rubel. “Even though we may not have designed a particular product for a customer before, we can use one of our platforms as a basis for creating the product, taking a general architecture and adapting it to what the customer wants,” he said. “To adapt that platform makes the design simpler, increases the probability of success and significantly reduces costs.”
By choosing a platform, MEMS designers also can make their devices easier to fabricate, because platforms are based on products that have already been successfully manufactured. This allows IMT to take advantage of previously acquired manufacturing knowledge, Rubel noted.
A broader view
MEMS devices are affected by many external factors, both in production and in use, so designers must be sure not to focus on a device operating in isolation. Such an approach fails to account for factors like electrostatic effects or energy loss, which can be caused by wiring or other devices that are in close proximity to a MEMS device.
Designers “have to simulate the environment of the device in addition to the device itself,” Maher said, adding that new software tools allow designers to evaluate so-called “proximity effects” on MEMS devices.
In addition to proximity effects, the MEMS designer may fail to account for the impact of assembly and packaging. According to some estimates, assembly and packaging can account for up to 90 percent of the cost of a MEMS device. They also can have a significant effect on the behavior of a device. “You may have a product that works great when it’s not packaged, but won’t work when it is packaged,” Popa said.
Still, he added, the traditional approach is to deal with assembly and packaging after the prototype is created. But he said these two processes should be considered early in the design phase.
New software can help designers do just that. In the past, separate software tools were used to design the MEMS device, the device packaging and the accompanying electronics. While this allowed designers to optimize their particular parts of the system, “they couldn’t really see how to optimize the whole system,” Maher said.
So SoftMEMS developed software that brings simulation up to the system level, linking FEA with electronics and packaging design. This allows people in different areas to work together on a co-design, which can reduce design time and overall system costs, Maher said.
Collaborative software tools may be helpful, but they can’t take the place of an old-fashioned meeting, according to one industry observer.
“Early in the design process, it’s imperative that you bring the whole team together,” said Roger Grace, a Florida-based MEMS marketing consultant. In addition to the designers, the MEMS team includes those responsible for packaging, testing and manufacturing the device.
During the meeting, all the attendees should provide their own special design input. For example, Grace said, “the testing people might tell the designers, ‘If we’re going to test this thing, we need these kinds of pads on these devices.’ Or the packaging people might provide practical footprints for the design.”
Of course, communication should be a continuing process throughout the design phase. But according to Grace, “the kickoff meeting is important, so everybody is on notice at the outset that this project is a team effort.” µ
___________
About the author: William Leventon is a New Jersey-based freelance writer. Telephone: (609) 926-6447. E-mail: wleventon@verizon.net.
------------------
The tiny swimmer
A “microswimmer” resembling a tiny fish was this year’s winner in the novel design category of the 5th annual MEMS University Alliance Design Competition, sponsored by Sandia National Laboratories. About the diameter of a human hair, the microswimmer has an aluminum tail that moves back and forth when heated and cooled by bursts of microwave radiation.
In the future, a medical device like the microswimmer could be used to travel in the human bloodstream, according to Tom Zipperian, Sandia senior microfabrication manager.
The novel design category is for innovative designs that take advantage of the strengths of the Sandia Ultra-planar Multi-level MEMS Technology 5 (SUMMiT V) fabrication process, which manufactures MEMS devices with five levels of polysilicon. SUMMiT V was used to make parts for the contest entrants. Sandia’s MEMS experts and university professors reviewed the designs and picked the winner.
Contestants must come from schools that are members of the MEMS University Alliance, which is open to all U.S. institutions of higher learning. The alliance provides classroom teaching materials and reasonably priced licenses for the use of Sandia’s SUMMiT V tools. This makes it possible for universities lacking their own microfabrication facilities to offer a MEMS curriculum.
—W. Leventon
--------------------------
Contributors
Roger Grace Associates
(239) 596-8738
www.rgrace.com
HT MicroAnalytical Inc.
(505) 341-0466
www.htmicro.com
Innovative Micro Technology (IMT)
(805) 681-2800
www.imtmems.com
Micralyne Inc.
(780) 431-4400
www.micralyne.com
Sandia National Laboratories
(505) 845-0011
www.sandia.gov
SoftMEMS LLC
(408) 426-4301
www.softmems.com
University of Texas, Arlington
Automation & Robotics
Research Institute
(817) 272-5900
www.arri.uta.edu
New directions: MEMS suppliers adopt new design strategies
