Page 92
Appendix C
Chemical Sensing Using MEMS Devices
Of all the microelectromechanical systems (MEMS) devices being developed, chemical sensors are probably the most difficult but also the most needed (Wise, 1996). Whereas mechanical sensors commonly provide levels of resolution and accuracy exceeding 12 bits (0.02 percent), the demands placed on MEMS-based chemical sensors are more exacting and, in some instances, may not be realizable. Indeed, chemical sensors must provide high sensitivity, stability, selectivity, specificity, and speed.
Some of the MEMS devices being explored do not have sufficient sensitivity or dynamic range for use in the field, especially over extended periods of time. Worse still, most chemical sensors are not stable and drift randomly at levels that are significant with respect to the overall dynamic range. Many also lack selectivity and respond similarly to a number of agents, only some of which may be hazardous. They also lack specificity and are unable to pick one (harmful) gas from other (benign) gases that may be present. Finally, response times are usually many seconds and can be many minutes, far too slow for wearable use. The human body would, unfortunately, respond faster.
Chemical sensors must be very sensitive (at the partsper-billion level), stable, and robust enough for use in the field, selective and specific to gases of interest (yet generic in approach), and fast enough to protect nearby personnel. The sensor should also be low enough in operating power to be deployed in the field in a small, lightweight package (similar in size to a credit card or wristwatch).
Several different types of MEMS devices are being explored as chemical sensors: devices based on gas absorption in bulk films; devices based on surface adsorption and microcalorimetry; conductivity-based devices (“microhotplates”); and full microinstruments, such as integrated gaschromatography systems. Devices based on the selective
~ enlarge ~
Page 93
~ enlarge ~
absorption of the chemical of interest in a thin bulk film usually detect the increase in mass that results from that absorption (Harsanyi, 1995). An early version of such a device was the vapor sensor of Howe (Howe and Muller, 1986) (see Figure C-1). It consists of a beam (typically polysilicon) which is deposited over a sacrificial layer (often phosphosilicate glass [PSG]). The beam is coated with a polymer that selectively absorbs the gas of interest; the sacrificial layer is then etched away, releasing the beam. In operation, the beam is driven electrostatically across the 1 to 2µ air gap left by removing the sacrificial layer, and the beam vibration is sensed capacitively. As the mass of the polymer changes, the resonance frequency and vibration amplitude shift, providing a measure of the vapor concentration. Such devices can be relatively sensitive (the frequency shift due to saturated xylene vapor in the Howe device was about 0.3Hz/ ppm), but most polymers are not very selective or very specific. Temperature sensitivity and stress effects are also difficult to predict and control. Time responses are set by diffusion into the sensing film and are measured in tens of seconds to minutes. (The response time of the Howe device was about 7 min.)
Surface acoustic wave (SAW) devices (Ballantine et al., 1997; Ricco et al., 1998) represent another implementation of structures in which changes in absorptive mass lead to a frequency shift in a surface wave, which can be detected
Page 94
differentially. Excellent work on SAWs has been carried out at Sandia National Laboratories. Chemiresistor arrays are another sensing approach worth watching that is currently being explored (Hughes et al., 1998).
Devices based on microcalorimetry (developed by IBM) have demonstrated that it is feasible to measure very small quantities of specific gases (Gimzewski et al., 1994). Using gold-coated, silicon cantilever arrays, the stress-induced bending of the cantilevers with molecular adsorption can be detected optically, and arrays of such structures are being developed as the equivalent of an artificial nose (Lang et al., 1999). By operating the sensor in several different modes, including cantilever bending (static) or resonance frequency shift tracking (dynamic), “orthogonal” responses can be obtained to distinguish between similar compounds. These devices are still in development but are worth watching for near-term (less than five years) commercial availability. Specificity and robustness in the field remain areas of concern for military applications.
Conductivity-based devices rely on the change in the resistance of a thin conducting film with adsorption or absorption of specific gases (Najafi et al., 1994). If metal- or metal-oxide-detecting films are used, they are supported on dielectric platforms that are thermally isolated from the body of the chip and its package (see Figure C-2). Because these devices typically operate at elevated temperatures (200–400°C), the on-state power requirements are tens of milliwatts or more for a single element. Response times are a few seconds. Conductivity-based devices can be very useful for detecting trace quantities of gases (a few parts per million or less), but they become saturated when a few monolayers of surface coverage are obtained. Even if the power requirements for such devices are acceptable, high specificity and selectivity are difficult to obtain. Selectively permeable membranes and/or detector arrays coated with many different films are being explored to improve performance. Microprocessor-based detection algorithms and/or neural nets to deconvolve the array signature into information on specific gases in a gaseous mixture will then have to be developed (Osbourn et al., 1994).
Temperature-programmed desorption can also be very helpful in identifying different species. Although it is possible to “program” completed arrays of microhot-plates with different detecting films using chemical vapor deposition on selectively heated dielectric windows, achieving acceptable selectivity, stability, speed, and sensitivity with these devices remains a formidable challenge (Majoo et al., 1995). Unless there are breakthroughs in detecting
~ enlarge ~
Page 95
films, these devices are unlikely to meet the requirements for detecting biohazards. Intermediate bioreceptors may make such breakthroughs possible. Thus, active-surface chemistries may greatly expand the utility of chip-based sensing platforms for both chemical and biological agents.
Chromatography, the most common laboratory technique for analyzing gaseous mixtures, is based on the fact that different molecular species spend different amounts of time adsorbed to surfaces as a sample of the gas is passed through a long tube (column). Thus, if a sample containing a mixture of gases is injected at one end of a long tube, the gases will emerge from the distant end of the tube separated in time. By calibrating the column delays for different known gases, the constituents of an unknown mixture can be identified. Chromatographic devices are sensitive, stable, specific, and selective.
Sensitivity can be enhanced by slowly absorbing the gas mixture of interest in a substance in front of the column (i.e., a preconcentrator) and then rapidly releasing it thermally into the column. If these systems can be miniaturized, they can also be relatively fast. Tabletop gas chromatographs (GCs) have column delays of several tens of minutes, but integrated miniature GCs, first proposed in 1971 and first reported by Terry et al. in 1979, can work much faster (a few minutes or less) (see Figure C-3). However, these microinstruments require an integrated gas-sampling system, which is only now becoming feasible. They also require a source of carrier gas or a pumping system, which has not yet been realized at the microscale.
Miniature GCs are now under development at Sandia and Lawrence Livermore National Laboratories. Calculations support the feasibility of achieving a sensitivity of 10 to 100ppb in a miniature GC, detecting atmospheric pollutants, and operating at less than 2mW in as little as 2 cm3 of space. This technology has, perhaps, a 10-year horizon to availability. A minaturized system could detect a variety of hazardous gases and could be deployed as a wrist-mounted sensor in the field. Detecting multiple biological agents is much harder, however; it will require a true chemical laboratory-on-a-chip based on microfluidics.