MEMS makers: machinists who work in silicon

The versatile and amazing silicon devices that are microelectromechanical systems (MEMS), such as microphones and inertial, light and pressure sensors, have already found their way into a multitude of consumer electronics devices and automobiles, among other products. As exemplified by the proliferation of iPhone apps that utilize motion sensing, there seems to be no end to what one can do with MEMS devices, from the sublime (lifesaving automotive stability control systems) to the ridiculous (the iBeer app).

An example of MEMS machining capabilities: free-standing silicon cylinders, 12µm in diameter, 40µm Z-axis height and arrayed on a 40µm pitch, formed by plasma etching into a silicon wafer. Photo courtesy A.M. Fitzgerald & Associates.

MEMS technology was pioneered nearly 40 years ago by mechanical engineers who wanted to make tiny 3-D structures out of silicon—not transistors. They used the tools at hand, borrowed from integrated circuit (IC) manufacturers, and shaped silicon with light, chemicals and plasmas in the cleanroom facilities known as “fabs.” Once regarded as exotic, crazy stuff, MEMS has become a mainstream technology with annual global sales of $8 billion.

Now that bigger markets are at stake, particularly in consumer electronics, attention is focused on how to make MEMS cheaper, faster and better. The drumbeat for standardization has started. The loudest call is coming from those who believe that MEMS, being made from silicon wafers, ought to conform to the successful IC industry model: process standardization. In the IC industry, wafers are manufactured using set process flows with rigid design rules controlled by the foundries.

There is a good reason for the IC system. Standardized, highly controlled processes allow foundries to run a few processes very well and provide their customers with low-cost, high-yielding chips through economies of scale.

Those with an IC perspective often lament that MEMS manufacturing is stuck in the trap of “one product, one process” and is therefore doomed to high production costs until it can conform to the IC model. This commonly held belief, however, is borne from the flawed assumption that MEMS devices are just like ICs. They are not.

MEMS devices are mechanical: They are freestanding bridges, diaphragms, hinges and trusses, shrunk down to microscopic size. Like any mechanical structure, their features are intentionally selected to serve a particular mechanical function, whether deflecting under certain loads or resonating at a specified frequency. The height, or Z-axis, of these features is created by the process steps of etching (sometimes deeply into the silicon wafer) or by material deposition. Different MEMS designs must use different process flows because it is not possible to create two different Z-axis features with the same process.

ICs, on the other hand, are planar devices. A cross-sectional slice of an IC chip reveals a Z-axis structure that forms transistors, the common building block of all integrated circuitry. An IC designer creates new designs by linking transistors on the wafer’s X-Y plane into different circuits. Many IC designs can be manufactured using the same process, as long as they share the same basic transistor architecture.

The planar nature of ICs facilitates process standardization. The IC manufacturing model consists of the foundries that own the Z-axis design—and therefore the process flow—and the chip companies that own the X-Y axes design, such as the circuit. Only with a common Z-axis building block, similar to the transistor, could MEMS ever conform to a standardized process flow.

So what now?

Just because MEMS cannot conform to the IC model does not mean we should abandon the quest for standardization. MEMS devices do need to be made cheaper and faster in order to enter additional markets. But we need to look elsewhere for a transformative manufacturing business model.

As a mechanical engineer who once shepherded parts through a machine shop and now does custom MEMS development for dozens of different applications, I believe the MEMS industry—despite its use of silicon wafers—has far more in common with producers of custom-machined parts. A machine shop, like a MEMS fab, uses a suite of machines, such as mills, lathes, saws and benders, to perform specific functions. The same shop might make products as diverse as door hinges or engine blocks. No two mechanical products will move through the shop the same way because their widely varying features—the very features that make them door hinges or engine blocks—demand different manufacturing steps. Sound familiar?

MEMS manufacturing is micromachining. We use plasma and silicon like a shop uses drill bits and workpiece materials. To reduce manufacturing costs and product development time, we should be looking to machine shops for ideas and emulating their manufacturing models.

In a shop, one finds standardization not in the process flow, but in the tools and methods that accompany each machine tool, such as drill bits, saw blades, cutting speeds, sheet metal gages and fastener sizes. Mechanical engineers have learned to work within these machine-specific standards when designing products. They specify 3.0mm holes, for example, instead of 3.023mm holes because the former can be drilled with a standard tool, which makes the part cheaper to produce.

To improve our manufacturing efficiency and lower costs, the MEMS industry must start thinking about how to standardize at the tool and/or recipe level, similar to how a standardized set of drill bits accompanies the drill press. Standardization of MEMS silicon wafer specifications such as for layer thicknesses of silicon-on-insulator devices, silicon-etch recipes to achieve specific depths or aspect ratios, and commonly used film deposition thicknesses are all within easy reach.

The potential benefits are many:

• standard material specs would enable material suppliers to plan production and inventory more effectively;
• standardized tool recipes would allow foundries to avoid costly process development and operate more efficiently; and
• MEMS designers could select from known process modules to ensure design success and repeatability.

Customization would still be available—but, like a machine shop, the supplied product would be priced accordingly. µ

About the author: Dr. Alissa M. Fitzgerald is founder and managing member of A.M. Fitzgerald & Associates, a MEMS product development firm located in Burlingame, Calif. Phone: 650-347-MEMS. E-Mail: info@amfitzgerald.com.