MEMS provide an alternative route to scaling technology beyond Moore's Law that could have benefits in markets such as the Internet of Things.
Many industry pundits have declared that the cost-reduction aspect of Moore’s Law has already ended, even as linewidths continue to shrink. It is hard to imagine a world where annual step increases in processing power do not come as naturally as the changing seasons, but that time may be fast approaching. MEMS devices, however, and piezoelectric MEMS in particular, are a new technology, and 2015 may be the start of a Moore’s Law for MEMS.
Jean Hoerni at Fairchild Semiconductor invented the planar transistor in 1959. Industry pioneers Robert Noyce and Gordon Moore, founders of both Fairchild Semiconductor and Intel, led the commercialization of this device, which defined the standard manufacturing process for integrated circuits until the commercialization of the FinFET by Intel in 2012.
The beauty of the planar transistor is that it was a robust device that enabled a relatively straightforward scaling path. Since Moore’s time, the photolithography source has moved from visible light to extreme ultraviolet, single patterning has been replaced by multipatterning, immersion lithography has become commonplace, and 3D FinFets have replaced 2D planar transistors. All of these efforts have kept Moore’s Law progressing for much longer than Moore envisioned, but the cost per transistor is no longer going down.
Micro electromechanical systems (MEMS) are microscale devices that use silicon wafer-based manufacturing processes similar to those used in integrated circuit manufacture. Unlike classic ICs, which only operate in the electrical domain, MEMS devices are both mechanical and electrical systems that interact directly with the real world. Some of the most common applications include accelerometers, gyroscopes, microphones, pressure sensors, oscillators, energy harvesters and RF filters.
At their heart, MEMS devices are energy transducers: they take mechanical or chemical energy and transform it into electrical energy that is processed as information by ICs. It may be more intuitive to think of MEMS devices as information, rather than energy, transducers. A gyroscope or microphone, for example, takes information from the real physical world and transduces it into electrically represented information that can be processed by an integrated circuit, stored in digital media or transmitted as data.
Traditionally, this transduction is done via electrostatic sensing sometimes called capacitive transduction. In electrostatic devices, the moving mechanical structure forms a portion of an air gap capacitor. The efficiency of the energy transduction is governed by the characteristics of this capacitor; the larger the surface area and the closer the capacitive plates, the higher the transduction efficiency. This is why many electrostatic MEMS devices have large comb-drive arrays that maximize the surface area of the opposing plates.
Decreasing the distance between the plates also will improve performance, but this creates another problem that has plagued electrostatic MEMS -- stiction. If the plates are too close together, they will get stuck and possibly never come unstuck. The physics of electrostatic energy transduction puts a hard limit on the size, performance, cost and reliability of MEMS devices.
Piezoelectricity is an alternative transduction mechanism for MEMS that has been gaining ground. Piezoelectrics are materials that generate a charge when stressed or conversely move in response to an electric charge.
Common industrial piezoelectric materials include lead zirconium tantalite (PZT), lithium niobate, lithium tantalate, aluminum nitride and quartz. Traditionally, piezoelectric materials were difficult to manufacture and often required growing a large high-purity ingot over the course of weeks or months that was then sliced into wafers for processing. This process was not well suited to advanced MEMS processing, and many materials were contaminants for standard CMOS IC processing.
A breakthrough in piezoelectric MEMS occurred when RF filter manufacturers pioneered the use of sputtered AlN thin films to make FBAR- and BAW-type resonators for use in RF filtering. AlN RF filters have been spectacularly successful products for companies like Avago and Qorvo (the former TriQuint) and kickstarted the development of the piezoelectric MEMS industry.
AlN has many advantages: it is CMOS-compatible, it is deposited using low-cost sputtering, is straightforward to etch, and has many beneficial mechanical properties. Many academic research centers have been working with AlN for over a decade, and at present, high-quality sputtered AlN films are available at several high-volume MEMS foundries.
Piezoelectric MEMS devices use the piezoelectric effect instead of electrostatic force, giving them a key advantage in energy transduction. Piezoelectric MEMS do not use an air gap capacitor and are therefore immune from stiction and are extremely reliable.
The only limitation on piezoelectric performance is the quality of the piezoelectric film itself, primarily the piezoelectric coefficient that governs how efficiently mechanical strain is transduced into electric charge. In many circumstances, piezoelectric transduction is more efficient than electrostatic.
Next page: Seeking benefits in scaling MEMS