The promise of microelectromechanical systems (MEMS), — building devices on the micron scale — was presaged in 1959 by Richard Feynman, a wellknown physicist. He gave a talk titled "There's Plenty of Room at the Bottom" to a gathering of the American Physical Society at the California Institute of Technology. He mentioned tiny motors, writing the 24 volumes of the Encyclopedia Brittanica on the head of a pin, and rearranging individual atoms to build useful materials.
Today MEMS is a reality, thanks, in large part, to fabrication technology borrowed from the semiconductor industry. Most new cars, for example, use tiny accelerometers to detect crashes and initiate airbag inflation. Companies such as Hewlett-Packard use MEMS to build printer heads for inkjet printers. And every few months, a national lab, university, or technology-based company sets a new benchmark in "getting small." Back in 1997, for instance, Cornell University bragged about creating the world's smallest guitar, a 10-micron-long instrument sculpted of silicon with six strings, each 50 nanometers in diameter. And IBM once touted their feat of writing out "IBM" in xenon atoms with each letter five atoms tall.
Still a relatively immature technology, MEMS is receiving intense scrutiny in the medical industry. Researchers foresee it ushering in a new age of medical diagnostic and analytical equipment.
Many of the advantages MEMS devices bring to medical applications stem from their most obvious characteristic, their small size. "Medical research these days centers around small things and tight places," says Jerome Jakubczak II, MEMS science and technology manager at Sandia National Laboratory. "Researchers want to get inside arteries and organs, manipulating cells, proteins, and other compound molecules. With MEMS, scientists will be able to interact with individual cells using micro-sized actuators, or monitor and sample specific cells or regions inside cells."
Smaller devices are also minimally invasive, so they do less harm when used inside patients. "Micron-sized diagnostic and monitoring tools have the potential to safely take more localized measurements," notes Dean Winter, director of bio-engineering at Southwest Research Institute in San Antonio. "Right now, if you want to study cells, you have to remove them from their natural environment. With MEMS, you will be able to study them in their home territory, getting down to the individual cell level, and that's where the frontiers of medical knowledge are right now."
The smaller you can make diagnostic devices, such as blood analyzers, the smaller the fluid or tissue sample has to be and less reagents are needed. Reagents can be expensive, and no one wants to take half a pint of blood from a premature baby when only a drop will do. Chemical reactions are completed faster in smaller volumes, so using small samples might also incrementally speed up analysis, an important factor for drug companies screening thousands of compounds.
These are major issues for the biochemical equipment being used to study and map the human genome. Some incorporate MEMS devices to examine picoliter-sized specimens (10-12 liters) of DNA segments. For example, Nanogen, a San Diego company in the forefront of DNA and gene analysis, builds a NanoChip workstation that puts 99 samples on a 2-mm-square array, and each sample need only contain 40 microliters for an accurate assay. Much of the technology behind MEMS, such as photolithography, sputtering, and epitaxy, was developed and refined in the semiconductor industry to build ever smaller ICs. So today's MEMS factories are often no more than modified wafer fabs. This lets MEMS devices take advantage of the scaling factor that has helped keep Moore's Law in force (i.e., the number of transistors on a chip keep doubling about every 18 to 24 months). And as in the IC business, the price for raw materials, mostly silicon, becomes insignificant on a per part basis.
"As devices get smaller, researchers will be able to create more per batch, thus bringing down the price of individual parts," says Marc Madou, vice president of advanced technology at Nanogen. "You can actually make the devices so small and cheaply that you could design and install an array of identical components. That way, if one breaks, you just use the next one, a form of built-in redundancy."
Semiconductor manufacturing techniques, together with some ingenuity, lets MEMS designers put a lot of functionality into a small package. Take the example of an implantable insulin pump. "MEMS lets us build the pumps, valves, and mixers, as well as the electrodes to move and separate compounds and sophisticated electronic controls, on a single, small substrate," points out Michael Huff, a researcher and executive with the MEMS Exchange, a network of MEMS fabrication and design facilities based in Reston, Va. "The whole system could be contained in a volume of only a few cubic millimeters, leaving more room for insulin storage or reducing the overall size of the device."
The ultimate goal of many MEMS scientists is to pack an entire lab's worth of functions onto a single chip. This "lab-on-a-chip" or micro-TAS (total analytical system), could be used in portable, "on-the-nurse's-hip" devices to provide instant blood analysis and genetic fingerprinting. Or, as Sandia's Jakubczak foresees, "MEMS' high degree of functionality could lead to home devices that sample saliva and detect subtle protein changes, early signs of disease. Similar devices could make it easier and faster to check water and soil for levels of lead, mercury, or other toxic compounds."
The downside of small
Unfortunately, what makes MEMS possible, also makes it expensive, at least for now. The technology behind semiconductors and MEMS is capital intensive. It costs millions of dollars to set up and maintain a modern fab, complete with the clean rooms and hardware needed to build MEMS devices. Few companies have the resources to do it.
"Right now, price is not an advantage for potential medical applications for MEMS," says Winter. "Most people use technology from the semiconductor industry, and that's very capital intensive. It takes a huge investment before ever building your device. That's why we don't build them here at the Institute."
Of course, one person's problem is another person's opportunity. Companies such as Micralyne, Edmonton, Canada, work with companies to refine and manufacture MEMS designs. Micralyne, with its 43,000 sq-ft fab, has worked with companies building mass-spectroscopy tools and genetic analysis instruments, as well as optical switches and other devices for increasing the bandwidth of telecom transmission lines. But with much of MEMS work outsourced from parent companies, it can take longer than usual to prototype and test products that use MEMS components.
The technology used for ICs is also centered mostly around silicon, which isn't always the best material for biomedical applications. "It might turn out that we need to use nonsilicon material for certain medical applications," says C. C. Liu, director of the Electronics Design Center at Case Western Reserve University in Cleveland. The Center focuses on MEMS technology applied to chemical and biological research. Medical tools such as tiny scalpels or other cutting instruments, for example, will probably not be made of silicon alone because it lacks the necessary mechanical strength. "So we will have to modify silicon-based processes we now use, or develop new ones."
Though based on 20 years or more of IC manufacturing, MEMS is a relatively immature technology, so researchers will have to devise standard ways to build products and invent new manufacturing techniques. "We will also need to develop new qualityassurance methods to characterize and investigate the reliability of various subcomponents such as pumps, seals, and liquid/gas interfaces," says Murat Okandan, member of the MEMS technical staff at Sandia.
There's also a problem in that many MEMS researchers come from the semiconductor industry. "We need to educate MEMS engineers and get them to think beyond silicon," notes Harry Stephanou, director of the Center for Automation Technologies at Rensselaer Polytechnic Institute. "They also have to be thinking of automating mass production and how to integrate microcomponents into macrodevices. For example, electrical engineers from the IC side might know how to get power and control signals to a device, but how do you get liquids or gases into such tiny devices? It becomes an assembly and packaging problem."
"MEMS also lacks a set of fixed processes," says Huff. "There's a diverse set of manufacturing techniques for MEMS, but the process seems to be different for each MEMS component. And although there is a diverse set of manufacturing techniques, they are scattered around the globe. No single fab site has everything, and most use only a small subset of MEMS processes and materials."
Another drawback for MEMS is that the manufacturing processes only make financial sense for large volumes. "Even in widespread applications like airbag accelerometers, the volumes are relatively small compared to ICs," says Huff. "Although there are 40 to 50 million cars sold each year, maybe half have four or five MEMS accelerometers on them. At a few dollars apiece, that translates into a market size of less than $500 million."
Of course the major question is how to manufacture MEMS for small markets and get an acceptable return on the investment when every device uses a different set of processes and the cost of the manufacturing equipment is high, according to Huff.
Reading MEMS' future
Despite the technological hurdles, MEMS seems to have a bright future. Several software companies are selling programs to help engineers design in the micron world where friction is more important than gravity and maintenance consists of replacement. At least one company, Micralyne, a MEMS fab house, sells a microfluidic toolkit to interested companies. It includes optics, electronics, software, and a few simple fluidic chips, enough to let companies find out what MEMS might be able to do for them.
There's also a perceived need for MEMS devices pushing development. "The Genome Project gave us the code that the machinery of life runs on, but now we have to figure out how that machinery runs. Researchers hope MEMS tools will help, building up from cells, to tissues, to organs, and then organisms," says Okandan.
MEMS is also a logical stepping stone to nanotechnology, or building devices atom by atom on the nanometer scale. (A micron is a millionth of a meter, a nanometer is a billionth of a meter.) As Winter points out, "It's just a matter or time until the problems that plague MEMS today will either be solved or bypassed. Someone might even make a technological leap beyond MEMS and straight to nanotechnology, discovering a totally new way to do molecular manipulation rather than bulk manipulation."