About TCAW - Subscription Info
May 2002
Vol. 11, No. 5
pp 26–31.
Table of Contents
Electronic Reader Service
Contact Us
Focus : Gas Chromatography


GC is in the Chips

The marriage of lab instruments and microengineering may transform the way we do analysis.

Laboratory equipment has been getting smaller for decades, but in the words of a famous song, “you ain’t seen nothing yet”. Conrad Yu, a researcher at Lawrence Livermore National Laboratory, may literally hold the future of chromatography in his hand—a prototype portable gas chromatograph. Yu’s GC uses technology called microelectromechanical systems (MEMS) to achieve its small size. If you want a portable GC this small, you will have to wait, although a larger portable GC based on MEMS components is available commercially.

Why is there a race for the really small? Three words: portability, speed, and cost. Portability enables testing in environments that were previously impossible or impractical. Of particular interest to the researchers at Lawrence Livermore, a compact GC could help U.S. soldiers determine if chemical warfare agents are being used against them. However, on-line GC testing at chemical plants and other locations could also benefit from the faster analysis times of smaller instruments.

For many, the most important consideration is cost. MEMS, like many new technologies, are expensive to design and develop; however, once developed, the cost savings can be tremendous. Many MEMS units can be produced on one silicon wafer, which, like computer chips produced on a single wafer, reduces costs tremendously. In addition to costing less to produce, MEMS GCs use less sample, carrier gas, and energy (for heating the coil), thus reducing operational costs.

One compelling example of the benefits of going small can be seen in the MEMS component you may already own: the accelerometer. These devices, installed in automobile bumpers, sense the sudden slowing of a car during a crash and trigger air bags to inflate. Conventional accelerometers took up valuable space and cost auto manufacturers around $50. MEMS accelerometers on most new cars are extremely small and cost $5–10. The reduction in cost and size has encouraged some manufacturers to introduce side air bags.

MEMS History
Interest in the micromechanical world began in 1959, when Richard Feynman gave a remarkably insightful lecture, “There is plenty of room at the bottom”. His idea of operation at the micro scale was partially realized with the introduction of silicon transistors and integrated circuits (better known as microchips). Although microchips were developed in the 1970s, research on micromechanical devices did not begin in earnest until the late 1980s, when fabrication techniques used to make microchips were applied to making mechanical structures.

Because micromechanics began with techniques used in microchip manufacture, researchers used silicon and other microchip materials. Silicon devices still dominate the MEMS arena because of extensive experience with its fabrication, but research into using other materials such as glass and plastic has produced alternatives to silicon MEMS.

Making MEMS
Fabricating MEMS and microchips is a complex process that requires the combination of numerous techniques and ultraclean conditions. The production techniques can be separated into three categories based on what they accomplish: thin-film deposition, pattern transfer, and etching processes.

One of the most well-known microchip processes, lithography, consists of a particular deposition technique (casting) followed by pattern transfer and etching. However, fabricating a MEMS or microchip device often requires a series of processes to make the complex shapes.

Thin-film deposition techniques, used to deposit small amounts of materials, from a few nanometers to about 100 µm, have been of interest to chemists and chip makers for decades. Several methods are widely used to deposit thin films, depending on what materials and conditions are needed.

Chemical vapor deposition. A gas–gas reaction deposits a uniform film on a silicon wafer or other surface. A high-temperature (600 °C), low-pressure process results in excellent uniformity. The low-temperature, gas plasma process is far quicker but results in more imperfections.

Electrodeposition. The same as electroplating. Restricted to deposition of electrically conductive materials.

Epitaxy. A gas–gas reaction, but the deposited film (usually silicon) continues the crystal pattern of the existing surface (i.e., amorphous, polycrystalline, or single crystal). Wafers are heated to half the melting point of the deposit material.

Thermal oxidation. A pseudodeposition technique. Reacts the surface (typically silicon) with oxygen at 800–1100 °C, producing a thin layer (<100nm) of oxidized material (silicon oxide).

Evaporation. Under a vacuum, the deposit material is evaporated (by electron gun or heating) and condenses on the wafer and the rest of the reactor. Evaporation is less uniform than chemical vapor deposition but more economical.

Sputtering. Similar to evaporation (similar quality at lower temperature), but uses a gas plasma (usually argon) to evaporate the deposition material.

Casting. Solvent and deposition material (typically polymers) are sprayed or spun (rapid spinning of the wafer) onto the wafer, and the solvent is evaporated. This process is key to lithography because most photosensitive materials are cast.

Pattern Transfer
Transferring a pattern onto a small chip is difficult. Most MEMS and microchip patterns are produced by exposing photosensitive material to light through a transparency (a mask). The photosensitive material reacts to the light, thereby transferring the image. Some materials become stronger when exposed, making it possible to wash away the unexposed regions with a solvent; other compounds are weakened by the radiation, so the exposed regions can be washed away.

The end result is the exposure of the original surface (below the photosensitive material) in only certain places that correspond to the mask. These exposed areas can then be modified, most commonly by etching.

Etching is the process of removing material from an exposed surface. Two types of etching are used: wet etching and dry etching. Wet etching uses a liquid solution of a compound that will etch the desired surface without dissolving the remaining photosensitive material.

There are three dry etching processes, all of which cost more than wet etching. HF is used in chemical vapor phase etching, and XeF2 is used to etch silicon and create isotropic etches. Sputtering, similar to the deposition process of the same name, etches the surface by bombarding it with ions.

Reactive Ions
But the most flexible etching technology is reactive-ion etching. This technique uses plasma gas to not only physically knock atoms off the surface but also chemically react with the surface. The features of the chip can be shaped by adjusting the process: Chemical etching is isotropic (etches evenly in all directions, including parallel to the surface under the mask), and physical etching is anisotropic (etches only perpendicular to the surface). A special adaptation to the process, deep reactive-ion etching, allows the etching of vertical-walled channels (with aspect ratios of 50 to 1) that can be hundreds of micrometers deep.

Using these basic techniques, researchers can produce an amazing variety of moving parts, channels, valves, and other structures on chips. The difficulty in making three-dimensional parts on this extremely small scale is complicated by the fact that all of the processes are basically two-dimensional. Certainly the channels are three-dimensional, but a bridge or pipe can’t be made with these techniques.

However, nonsilicon-based materials are being developed for MEMS. Plastics have been of keen interest because of their cost and ease of handling (Anal. Chem. 2002, 74, 78A). Laser ablation has been used to make channels and other features, skipping the lithography step and creating patterns directly.

Figure 1. Transfer pf pattern onto photosensitive material
Figure 1. Transfer of pattern onto photosensitive material
Figure 2. Reacted photosensitive material removed
Figure 2. Reacted photosensitive material removed
Figure 3. Silicon nitride layer is anisotropically etched
Figure 3. Silicon nitride layer is anisotropically etched
Figure 4. A pattern of exposed siliconnitride surface
Figure 4. Pattern of exposed silicon nitride surface
Figure 5. Channel is formed in silicon by wet etching.
Figure 5. Channel is formed in silicon by wet etching
Figure 6. Channel after silicon nitride is removed
Figure 6. Channel after silicon nitride is removed
Figure 7. A cross-section photo of the completed column
Figure 7. A cross-section photo of the completed column
MEMS in the GC
Conrad Yu’s portable GC uses a column etched into a silicon wafer, instead of the typical copper or glass tubing. Previous attempts to make etched columns on silicon were unsuccessful for several reasons. Some attempts used lithography and anisotropic etching that produced square channels, which were then capped by bonding a different material to the top of the silicon wafer. This proved difficult because the materials had different heating coefficients and expanded and conducted heat differently. In addition, the stationary phase was unevenly distributed in a square channel.

Yu used lithography, but instead of making a square channel with anisotropic etching, he used isotropic etching to create a semicircular channel. The process works by first depositing a layer of silicon nitride or boron nitride onto the silicon substrate. Next, a photosensitive material is cast on top of the nitride surface and exposed to light shown through a mask containing the pattern of the column to be etched (the result is shown in Figure 1). The exposed photosensitive material is then washed away, leaving a small amount of nitride exposed (Figure 2). Plasma etching then anisotropicly etches away the exposed nitrate while the photosensitive layer protects the rest of the nitrate (Figure 3). The photosensitive layer is washed away, leaving the shape of the column in exposed silicon (Figure 4).

The exposed silicon is then etched using an isotropic wet etchant such as HF, HNO3, or CH3COOH. The silicon is eaten away equally in all directions, even underneath the silicon nitride or boron nitride top layer (Figure 5). Another etchant is used to remove the nitride layer (Figure 6).

Once these fabrication techniques are completed, the tough part begins: lining up the two halves of the column. If alignment is off even by a millimeter, it is disastrous. To combat the alignment problem, Yu designed a set of guides that line up when the wafers are in perfect alignment. The guides are etched when the channels are etched, allowing a circular column to be formed (Figure 7).

Once constructed, silicone is forced through the column to provide a stationary phase. "Our column’s internal diameter is 100 microns [micrometers] and more than 5 meters long, which is equivalent to a [conventional column] 20 meters long with a 200 micron diameter," says Yu. He also reported that analysis times for simple hydrocarbons (C1–C6) were only 30–40 s and flow rates were 6 µL/s; conventional GCs would have flow rates of 5–10 mL/s and take 5–10 min.

Add a much more complex MEMS inlet valve, and either a MEMS thermal or glow discharge detector, and voilà, you have a handheld GC. Of course it is not that simple, especially at this scale.

While the GC column illustrates the amazing complexity of the steps used to fabricate micrometer-scale structures, designing MEMS is a trial-and-error activity. Problems with mechanical devices at this scale plagued early designs. Early gears and motors were quickly overcome by heat buildup, something easily handled in the macro world. Friction and wear are also complicated, as adding a lubricant is not as simple as finding the right oily substance.

More problems arise when MEMS are designed to handle fluids. Many of the engineering assumptions made for macroscale fluid flow are turned on their heads at the microscale. Surface effects (attraction between the pipe and the fluid) are virtually ignored in house and reactor plumbing, but at the microscale, surface effects cannot be ignored. In some cases, surface effects such as hydrophobicity can cause a liquid not to flow. In the macroscale world, hydrophobicity is not usually the cause of blocked sewage pipes.

Bubbles cause significant problems in some MEMS applications. Bubbles normally form in big pipes but cause little trouble except under unusual conditions. But a single bubble can clog a channel that is only 50 µm wide. To overcome this, MEMS designers have developed structures that can pop bubbles and channels that can safely hold them without blocking flow.

Particles cause some of the same problems as bubbles, blocking fluid flow and clogging valves. However, one of the best ways to prevent particle contamination is also a serious problem for MEMS designers. Packaging is what keeps particles out, since more MEMS are produced in clean rooms. Packaging for computer chips is fairly straightforward: Surround it completely in plastic with metal connections coming out. It’s not that simple for MEMS. MEMS usually need to have an area that is open to external contact—not just to electrical contacts. The GC MEMS parts need to connect to one another and to the outside world via the injection nozzle. Particles can enter through that route as well as through the packaging around the injection area. Almost all MEMS packaging must be designed specifically for that application, which adds an additional constraint.

Perhaps the most interesting problems that MEMS fluid designers grapple with are related to laminar flow at this very small scale. Laminar flow is the ordered, nonturbulent manner in which fluid flows. In the macro world, turbulent flow is the norm, but diameters in the micro world are small enough to virtually eliminate turbulence.

Laminar flow is easier to simulate than turbulent flow, but it does have some strange side effects. For example, when two different fluids are joined in a laminar flow pipe, they mix only through diffusion. In turbulent flow, the two fluids mix quickly. The waters of two rivers, one muddy and one sediment-free, mix in the course of several miles; a long distance, but considering the enormous volume of water, it is not very far at all. In microfluids, mixing does not occur in what would be hundreds of miles on a major river. This is not all bad, as it can be used as a means of separation based on diffusion between two fluids.

A MEMS Future
MEMS technology, like that of its cousin the microchip, will undoubtedly change the face of scientific instrumentation. The change will not happen overnight, as the curve is steep for learning how to handle fluids and motion in a strangely small world. However, the potential impact of MEMS and microfluidics on our lives and work could be as important as the personal computer revolution. Just imagine, one day, as strange as it sounds, we may have analytical equipment in the palm of our hand. But it may only happen if we up the ante with chips.

Further Reading
Introduction to microengineering: www.dbanks.demon.co.uk/ueng
Lawrence Livermore MEMS GC: www-eng.llnl.gov/GCwebsite/index.html
MEMS Clearinghouse: www.memsnet.org

Michael J. Felton is an associate editor of Today’s Chemist at Work. Send your comments or questions regarding this article to tcaw@acs.org or the Editorial Office, 1155 16th St N.W., Washington, DC 20036.

Return to Top || Table of Contents

 CASChemPortChemCenterPubs Page