MEMS Technology Adds Manufacturing Diversity
One of the most exciting technologies of the new century, silicon microelectro-mechanical system (MEMS) technology is turning out to be much more versatile than most people expected, going beyond simple sensors to tackle a variety of roles in a diversity of markets.
At the heart of the new technology is the development of commercial technologies for micromachining silicon - the basic production techniques that are the basis for various device types. Though details vary, the fundamental process steps are simple to explain. Typically we begin with a silicon wafer onto which a layer of sacrificial oxide is grown. Into this layer we etch holes at points corresponding to the supports for fixed elements and anchors for moving elements. A thicker polysilicon epitaxial layer is grown on top of this and into this third layer we etch the structures for the moving and fixed elements of the device. Finally the sacrificial oxide layer beneath the structures is removed by an isotropic etching operation to free the moving parts. The open space around the structures is filled with gas, normally dry nitrogen, to avoid effects caused by humidity or variations in gas density, which would affect resonant frequencies.
MEMS technology is attractive because it brings to the manufacture of components the same economies of scale that made microelectronics such a success. Instead of manufacturing microactuators and sensors one at a time, they can now be made hundreds at a time on silicon wafers and they can be manufactured using proven manufacturing techniques pioneered and developed for silicon chip manufacture. Most of the production steps used in MEMS manufacture are borrowed from silicon chip manufacturing technologies, and the relatively large geometries of current MEMS devices means that they can be made in older wafer fabs which are already amortized. This gives these fabs an unexpected new lease of life when they are no longer suitable for leading-edge chip manufacture.
One of the simplest, commonest MEMS devices is the accelerometer. Linear accelerometers are widely used in the automotive sector for airbags and are also finding new applications in disk drives where they help to compensate for vibration. They can also be used in videogames to make controllers that sense movements in three dimensions, in computers where they can be used to make a pointing device that does not need to be placed on a desk and in new generation robotic toys.
Typically a MEMS accelerometer consists of interlocking fingers which are alternately moving and fixed. Acceleration is sensed by measuring the capacitance of the structure, which varies in proportion to changes in acceleration. The elements can be arranged like combs to make a linear accelerometer - the type used in airbags - or like spokes of a wheel to make a rotational accelerometer (Fig 1).
Rotational accelerometers are useful in automotive stability systems, for example, to sense when the vehicle is moving around the yaw or Z axis. An evolution of conventional anti-skid braking systems (ABS), they sense the actual movement of the car and not just the blocking of wheels. These stability systems greatly enhance the safety when driving on snow, ice and other difficult surfaces where one or more wheels may have little adherence. They also add an extra margin of safety to the newer vehicles that have a higher center of gravity and hence less passive stability.
In the automotive sector similar structures can also be used to make Coriolis effect gyroscope devices, which are important components for navigation systems. Satellite based position-sensing systems like GPS do not work well in cities because the satellite signals are frequently blocked by shadows from buildings. For this reason vehicle navigation systems also apply a "deduced reckoning" approach where a sensor on a wheel measures the distance traveled while a gyroscope senses the direction of motion. The gyroscope/speed sensor combination makes it possible to continue tracking movements during the time when the satellite signals are not visible or where they are not sufficiently precise.
Disk Drive Applications
In addition to their automotive uses, rotational accelerometers are also useful in hard disk drives (HDD) to sense rotational movements because they affect the voice coil head positioner, causing the drive head to lose tracking. By building a rotational sensor into the drive it is possible to compensate for rotational movements, ensuring that the head stays put on the track it is reading or writing. This results in better drive performance because more time is spent reading and writing and less time is spent restoring head tracking after a shock.
Designed specifically for this application, the L6670 rotational accelerometer system shows how a rotational accelerometer sensor is used in practice. In this case the MEMS sensor device is packaged together with an interface chip (Fig 2) that measures the minute changes in capacitance, generating a clear, consistent and precise output signal.
The L6670 is assembled in a standard, inexpensive molded plastic package, raising the question of how the moving parts are protected from the molding resin. The answer is that on top of the MEMS wafer a second wafer is bonded having a small cavity etched in correspondence with the MEMS part in the wafer below. This operation is carried out in a controlled, inert atmosphere such as dry nitrogen. The gas encapsulated in the MEMS structure must be dry because any humidity could cause moving parts to stick, though today MEMS devices are increasingly designed so that even after wetting they do not suffer these stiction effects.
It is also important that the density of the encapsulated gas be controlled because the resonance frequency of the moving structures depends on this density. This fact also provides a non-destructive test for measuring the density of the gas inside a structure since it can be measured by measuring the resonance frequency. By measuring this parameter over time or before and after, reliability tests also provides a method for evaluating the long-term integrity of the wafer-on-wafer bond.
A similar structure to the rotational accelerometer can be used to make a rotational microactuator. In this case instead of motion causing a change in capacitance it is the other way around: an applied voltage that causes movement. To increase the magnitude of the effect each "spoke" of the wheel-like structure has an interdigitated comb like structure that increases the effective area of the rotor and stator. Micromotors can be used for any micropositioning application, but one of immediate commercial value is in HDDs.
Conventional HDDs achieve bits-per-inch capacities along each track as much as ten times the tracks-per-inch radial capacity of the drive. This is because the voice-coil head positioner does not allow sufficiently precise positioning to track finer pitch tracks. But by adding a second fine positioning mechanism at the end of the normal head suspension it is possible to achieve a significant improvement of the tracks per inch - and thus the overall drive capacity - without changing other details of the drive.
Not all MEMS devices have moving silicon parts; sometimes the only moving part is a fluid (Fig 3). The most common example of this type of MEMS is the bubble-jet printer head chip used by one of the major printer manufacturers. In this chip there are hundreds of microscopic channels connected to ink-filled chambers. Each chamber can be heated very rapidly by a resistive heating element, vaporizing part of the ink and propelling the volume of the ink towards the paper in a tiny droplet. In this device the only moving "part" is the ink. The MEMS device in this case is clearly visible on the surface of the ink cartridge as a small rectangle of silicon connected by a flexible cable to the contact pads of the cartridge. Early devices for this application contained only the micromachined parts and the heating elements. In the latest generation of these devices some of the control circuitry is also included on the same chip so that a large number of channels can be managed with a relatively small number of connections from the die to the rest of the circuit.
This approach, combining MEMS and electronic components, is clearly beneficial in some cases because it reduced interconnections by keeping some of the electronics close to the actuators or sensors. However, it is unlikely that complete complex systems will be integrated on a MEMS device in the near future because it is less expensive to implement complex control circuits using standard CMOS technology. In most cases the optimal approach is to put a minimum of electronic content on the MEMS chip or, as in the case of the rotational accelerometer, put all of the electronics on a separate chip.
Agilent Technologies is applying the same basic thermomicrofluidic technology of the inkjet printer head for a completely different application - optical switching. In conventional optical networks the bottleneck is that to switch the signals they need to be first converted to an electrical signal, switched and then reconverted to optical signals. To bypass this problem the solution is an optical switch, which finally makes it possible to create an all optical network that is transparent from end to end (Fig 4).
There are several techniques for making optical switches, of which the most common is to use microscopic moving mirrors that are moved electrostatically. Agilent's Photonic Switching Platform adopts an innovative approach where the light beams to be switched flow through transparent channels. In these transparent channels there are small cavities filled with a liquid having similar optical properties as the surrounding material. When the liquid is present the light passed through the switch uninterrupted. But under each cavity there is a small heating element, like in the inkjet printer, and when this is activated the liquid is vaporized, leaving a small bubble that reflects the light beam, switching it to a separate output. This highly original approach demonstrates how MEMS technology can be applied in surprising ways to achieve new results.
Another interesting application of fluidic MEMS technology is a prototype device for performing the Polymerase Chain Reaction (PCR) technique for amplifying minute samples of DNA. In this technique the sample is mixed with DNA building blocks called primers and an enzyme called Polymerase. The mixture is then cycled repeatedly through three temperatures - 95centigrade, 65centigrades and 72centigrades - which break up the original DNA and builds copies using the primers and the enzyme. Each repetition of the cycle doubles the DNA and can be repeated indefinitely, multiplying tiny samples until they are large enough for analysis. If we start with just one DNA molecule after just 20 cycles there will be a million copies and after 30 cycles there will be a billion copies.
In the conventional laboratory method this is done with bulky and costly equipment and can take several hours. The same PCR method can be implemented using a MEMS device in which channels buried in the silicon carry the mixture of sample and reagents while resistive heating elements perform the temperature cycling. With this MEMS device the PCR technique can be applied in a compact, portable analysis tool that also requires very small quantities of the costly reagents. Since there is no practical way to clean the channels after use these MEMS devices are inevitably for single use only.
Similar devices for chemical and biomedical analysis are likely to revolutionize tests today carried out in the laboratory, enabling new portable analysis devices that can be used in the field rather than gathering samples for subsequent analysis elsewhere. This is likely to have a major impact on the medical world, but could also enable new products for checking food safety, detecting bacteria, toxins or allergens before food is prepared or consumed. MEMS technology will be a major driving factor in this new business.
MEMS RF Components
MEMS technology is also finding interesting applications in wireless technology, which is becoming increasingly important as short and medium-range wired connections are being replaced by radio links. Cellular networks have already revolutionized telephones, and soon Bluetooth wireless technology will eliminate wired connections to peripherals. New standards such as HomeRF and IEEE802.11 will bring the same benefits to home networking. All of these applications demand inexpensive, stable and vibration resistant RF circuits and MEMS technology can contribute to meeting these requirements.
Among MEMS RF components now being developed are switches, variable capacitors, filters and so on. Today most RF switches are based on semiconductor devices but with MEMS technology we can build a replacement where electrostatic actuation moves a physical switch, creating a structure that has both high OFF state isolation and very high linearity in the ON state (Fig 5). Switches are useful to separate the transmit and receive paths in wireless applications, though they can also be used to make a variable tuning capacitor by arranging an array of fixed capacitors and switches. This kind of variable capacitor allows the capacitance to be varied in discrete steps but with the benefit of extremely robust resistance to vibration - a very useful benefit in mobile applications. It is also possible to build the equivalent of a moving plate variable capacitor in MEMS technology, and this allows continuous control but it is subject to vibration. Compared to the varicap diodes they replace, MEMS variable capacitors offer a much greater range of variation.
Filters can be constructed using MEMS devices where a resonator is excited electrostatically. Because of the physical properties of silicon and the small size of the moving parts it is possible to achieve resonance at very high frequencies. Because of the great precision of silicon micromachining technology it is possible to achieve very close tolerances in the manufacture of MEMS filter devices.
MEMS devices bring many new challenges to manufacturing, most of which are not directly related to the silicon micromachining operations but to the related areas of packaging and testing. In some cases like the accelerometer device described above the chip can be designed for assembly in a standard molded plastic package by bonding a second wafer on top of the MEMS wafer. There are many other cases, however, where packaging is complicated because contact is needed between the MEMS structure and other mechanical components - like in the disk drive micromotor - or with liquids, as is the case of the thermofluidic reactor for the Polymerase Chain Reaction.
Testing, too, involves a number of new aspects related to the mechanical movements, integrity of seals, resonance frequencies and so on. An accelerometer, for example, will require a test jig that generates movements to exercise the sensor.
Finally, long-term reliability is essential in many uses such as optical switches for telecom networks, though perhaps a little less critical in single-use parts.
The growth of MEMS technology in the first decade of the 2000s will create many new opportunities for business and also for employment, creating demand for a new kind of engineer with a broader, multidisplinary approach.
by Benedetto Vigna, Manager, MEMS Sensors and Microactuators,
Telecommunications and Automotive/Peripherals Groups, STMicroelectronics,
(August 2001 Issue, Nikkei Electronics Asia)