256 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001 Chemical Sensors for Portable, Handheld Field Instruments Denise Michele Wilson, Member, IEEE, Sean Hoyt, Jiri Janata, Karl Booksh, and Louis Obando

Abstract—A review of three commonly used classes of chemical transduction mechanisms or reactions. This broad range of sensor technologies as applicable to implementation in portable, stimuli and transduction mechanisms is a result of both an handheld field instruments is presented. Solid-state gas and chem- inherent characteristic of most chemical sensor technologies ical sensors have long been heralded as the solution to a wide va- riety of portable chemical sensing system applications. However, and an engineered characteristic that is used to produce a signal advances in optical sensing technology have reduced the size of sup- representing the stimulus of interest significantly stronger than porting infrastructure to be competitive with their solid-state coun- those of interfering stimuli. terparts. Optical, solid-state, and hybrid arrays of sensors have Sensitivity is a term often used to refer either to the lowest application for portable instruments, but issues of insufficient se- lectivity and sensitivity continue to hamper the widespread intro- level of chemical concentration that can be detected or to the duction of these miniaturized sensors for solving chemical sensing smallest increment of concentration that can be detected in problems in environments outside the laboratory. In this article, the sensing environment. In typical laboratory applications for we evaluate three of the major classes of compact chemical sen- chemical sensing, it is necessary to know exactly how much sors for portable applications: (solid-state) chemiresistors, (solid- of each chemical component is present in a complex mixture state) CHEMFETs, and (optical) surface plasmon resonance sen- sors (SPR). These sensors are evaluated and reviewed, according across a wide dynamic range. In many portable chemical to the current state of research, in terms of their ability to operate sensing applications, however, such as landmine detection, at low-power, small-size, and relatively low-cost in environments, environmental (ground water and air pollution) monitoring, with numerous interferents and variable ambient conditions. and toxic chemical agent detection, it is necessary only to Index Terms—CHEMFET, chemical sensors, chemiresistors, know when the concentration of a single or small number conducting polymer sensors, gas sensor, ISFET, metal–oxide of chemicals has exceeded certain alarm levels. This change sensors, surface plasmon resonance (SPR). in focus from complex composition analysis to alarm level detection changes the requirements of the sensing system. In I. INTRODUCTION the laboratory applications, both sensitivity associated with dynamic range and resolution as well as sensitivity associated INIATURIZED chemical sensors have yet to achieve with the lower detection limit are important. In many portable their full potential for commercialization. The reasons M applications, however, the lower detection limit, rather than for this delay in widespread commercialization are complex. resolution or dynamic range becomes a dominant factor in However, insufficiently robust selectivity and sensitivity determining the usefulness of a chemical sensor technology. of chemical sensor technologies after miniaturization are a It is worth noting that many concentration limits dictated common limitation to the progressive development of chemical by chemical sensor applications in the field are extremely sensors toward handheld commercial products. For portable low, well beyond the capability of the human nose to detect field-based instruments, high noise levels in the chemical and approaching or exceeding the capability of the canine composition of the field environment as well as the potential nose to detect as well. Engineering systems to achieve these impact of highly variable ambient conditions have a profound stringent concentration detection limits require mature and effect on the noise level which chemical sensor technologies well-understood science in the chemical sensor technology must accommodate while producing a robust and reliable as well as the elegant and creative design of measurement output signal. Unlike imaging (visual sensing) technologies and signal processing surrounding these sensor technologies. where only one type of light (e.g., infrared or visible range) Hence, miniaturization of sensor technologies in the design is detected and sound (auditory sensing) technologies where of the portable instrument can often sacrifice some of the only one type of pressure wave is detected, chemical sensors dynamic range and resolution of the sensor, but must not cause must transduce a variety of input stimuli using a variety of significant increase in noise levels and hence degradation of the detection threshold. Unfortunately, miniaturization often Manuscript received September 18, 2000; revised October 17, 2001. The as- generates an increase in noise, thereby necessitating further sociate editor coordinating the review of this paper and approving it for publi- cation was Prof. Julian W. Gardner. stringency in the remainder of the design parameters, including D. M. Wilson and S. Hoyt are with the Department of Electrical Engi- precise control of manufacturing variations and micro and neering, University of Washington, Seattle, WA 98195-2500 USA (e-mail: nanoscale characteristics of these sensors. [email protected]). J. Janata is with the Department of Chemistry, Georgia Institute of Tech- Sensor selectivity does not typically suffer adverse degrada- nology, Atlanta, GA 30332 USA. tion upon miniaturization but nevertheless presents many bar- K. Booksh and L. Obando are with the Department of Chemistry and Bio- chemistry, Arizona State University, Tempe, AZ 85287-1604 USA. riers in tailoring chemical sensing technologies to particular ap- Publisher Item Identifier S 1530-437X(01)11120-6. plications. Selectivity refers to the ratio of the sensor’s ability

1530–437X/01$10.00 © 2001 IEEE WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 257 to detect what is of interest over the sensor’s ability to detect stability variations. Unfortunately, the parameters of a chemical what is not of interest (the interferents). For the detection of sensor technology that can be adjusted to improve robustness only a single type of molecule, the ideal chemical sensor selec- are often the same parameters that result in a decrease in sensi- tivity is infinity, reacting only to the one molecule of interest and tivity and selectivity. For example, the most sensitive chemical transducing it into an electrical signal that contains no funda- sensor is often one whose reactions are totally irreversible; in mental noise from environmental interferents. For the detection this situation, every reaction with every molecule of interest of multiple types of single molecules, the ideal chemical sensor in the environment is captured and can be measured, naturally system contains an array of sensors, each infinitely selective to leading to maximum sensitivity; in semi-reversible sensors, one type of molecule. For the detection of complex mixtures of the kinetic balance between outgoing transduction events molecules, the ideal chemical sensor system is not as easily de- and incoming transduction events relevant to measuring the fined and depends on a wide variety of parameters, not the least stimulus of interest inherently results in reduced sensitivity of which is the composition and range of chemical variations in over the sensor whose reactions are completely irreversible. A the ambient environment. Even in relatively “simple” applica- similar design dilemma occurs when increasing the sensitivity tions, such as the detection of a single molecule, the ideal chem- of a chemical sensor improves the response to the stimulus ical sensor is typically not available. Selectivity then, is often of interest but also, frequently increases the sensitivity of the tailored and adjusted to the analyte(s) of interest by varying a chemical sensor to interferents (e.g., “noise” in the ambient wide range of parameters including dopants, grain sizes, cat- environment). Achieving a fine balance between robustness and alysts, external filters, operating temperature, and many other system sensitivity (e.g., performance to design specifications) factors. Miniaturization of the chemical sensor technology for is a delicate effort involving both engineering and science. portable applications often involves examination of new sensor While this balance can be achieved, it involves engineering technologies as many macroscale technologies cannot be scaled design methodologies and scientific optimization of chemical to miniaturized technologies using the same fabrication pro- sensors that must cross disciplines in a coherent, methodical, cesses. After a shift in fabrication processes, the tunability of and efficient manner. In addition to the inherent difficulty the selectivity to certain chemicals is often re-examined to es- of designing chemical sensors for desired sensitivity and tablish optimal ways to achieve selectivity under the constraints selectivity, adding robustness to the sensor for system level of the fabrication process. This re-evaluation is often not ac- functionality is a complex process, which has subsequently led complished elegantly, since the quickest conversion to minia- to slow development of portable systems that push the envelope turized sensors is often done by scaling down the macroscale of power, space, cost, and signal/noise levels of many chemical processes and design methodologies. After this direct path to sensor technologies. miniaturization is accomplished and found to be wanting, then, In this article, we review three major classes of chemical engineers and scientists turn to the redesign of the sensor itself. sensor technologies for portable field instrument development. This process can be slow, especially due to the complexity of These sensors, by no means, represent all available sensor transduction mechanisms that must be considered with every technologies; rather, they serve as promising, representative chemical sensor technology and its associated parameter vari- technologies for miniaturization into portable instruments. ations for tuning selectivity both at macro and microscale. The first class, the chemiresistors, are the simplest of the Again, unlike other sensing technologies such as those as- chemical sensor technologies available for miniaturization sociated with visual sensing (imagers), commercial, miniatur- into field instruments. Chemiresistors possess advantages ized chemical sensing products are out of the question until a of compact size, simple fabrication, low cost, and simple certain stringent level of precision (sensitivity and selectivity) measurement electronics but are hampered by limitations in is achieved with the component chemical sensor technologies. signal-to-noise ratio. The second class of sensor technologies Whereas the miniaturized visual imagers (CCD and CMOS) reviewed herein, the potentiometric CHEMFETs, are more were commercialized first at low resolutions and progressed to complex solid-state sensors. The CHEMFETs, in general, are higher resolutions in subsequent generations of products, most capable of higher signal-to-noise ratios than their chemiresistor miniaturized chemical sensor technologies have not had this counterparts but are inherently more complex to fabricate and luxury. This inability to use existing, manufacturing-scale fabri- require more extensive control and measurement electronics. cation processes to evolve subsequent generations of higher res- These characteristics increase their cost and limit the number olution, sensitivity, and selectivity chemical sensor products is a of CHEMFETs that can be placed inside a single device for profound limitation to the commercialization of portable chem- portable deployment. Several other major classes of solid-state ical sensing systems. With this limitation in mind, it is perhaps chemical sensor technologies are practical for field portable easier to understand the (relatively) slow evolution in miniatur- instruments, such as the mass sensors (surface acoustic wave, ization of chemical sensor technologies. bulk acoustic wave, etc.). Due to space limitations, however, For those applications where chemical sensor technologies we limit our discussion to two classes of solid-state sensor have been available for portable instrument development, technologies representative of very simple (chemiresistors) robustness has been the last major stumbling block toward and more complex (CHEMFET) fabrication structures and commercial product development. Robustness, at the sensor transduction mechanisms. For completeness, we also review in level, refers to the ability of the sensor to perform its function detail a common mode of optical chemical sensor, the surface over a range of ambient conditions (e.g., humidity, temperature, plasmon resonance-based sensor, which is capable of using etc.) and over a range of times in the presence of drift and a white light source as its optical input and can use highly 258 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001

TABLE I COMPARISON OF THREE TYPES OF CHEMICAL SENSOR TECHNOLOGIES FOR PORTABLE APPLICATIONS

integrated compact optics for output signal measurement, sensitivity, selectivity, or other desired characteristic of the making it suitable for portable sensing applications. All three portable instrument. classes of sensors have in common successful products on the macroscale, which make them likely candidates for successful A. Nonoptical Chemical Sensors miniaturization in commercial products. Each method has its Most forms of chemical sensors that are both suitable for particular disadvantages and advantages and we fully anticipate miniaturization and do not rely on optical means for signal that hybrid systems, consisting of combinations of optical and transduction depend on only a single transduction step to solid-state chemical sensing technologies will be developed to convert a chemical signal or concentration, in vapor or liquid meet the needs of many practical applications. phase, to an electrical signal. Typically, the chemical of interest interacts or adsorbs onto the surface of a chemically sensitive material, thereby inducing a change in the electrical II. CHEMICAL SENSORS FOR PORTABLE APPLICATIONS characteristics of the material that represents chemical concen- tration. The electrical parameters extracted can be of a wide Table I summarizes the chemical sensor technologies re- variety, including dc resistance, ac resistance, ac impedance, viewed in this paper in terms of their suitability to the design FET threshold voltage, phase change of a wave propagated of portable instruments. High sensitivity and selectivity are across the sensor surface (acoustic wave sensors), resonant desirable characteristics of any chemical sensor technology frequency of cantilever beams, and so on. In most cases, in order to meet the specifications of the sensing application. the primary advantage of many of these chemical sensors is However, advances in array processing have increased the se- that they require minimal transduction effort and therefore, lectivity and sensitivity at the system level beyond the potential upon miniaturization, incur minimal electronics overhead, of individual sensors within a chemical sensing system. To fabrication complexity, process variability, and supporting avoid confusion, however, Table I focuses only on the chemical infrastructure. Minimal electronics overhead and process sensor technology itself and the capacity and attractiveness of variability in turn result in the introduction of less noise to any one type of chemical sensor to meet the needs of portable the signal during transduction and readout. However, many instrument applications. The importance of sensitivity, selec- of the same factors that make these sensors attractive for tivity, and robustness for portable instruments is undisputed portable instrument applications also make them problematic. and has already been discussed. Supplemental to these basic For example, minimal supporting infrastructure and compact requirements, however, portable instruments also require sensor size usually coincide with packaging difficulties where the technologies that consume low power, are compact in size, sensing electronics and supporting infrastructure are not easily and incur low electronics overhead. Control and measurement shielded from aggressive ambient environments. In addition, electronics overhead can play a major role in increasing the miniaturization of many of these chemical sensor technologies size and power consumption of the overall system beyond what often causes increases in noise rather than improvements in is practical; in many sensing systems, overhead electronics can signal to noise ratios. Chemiresistors are a classic example consume 40% or more of the total space consumed and often where miniaturization increases noise significantly and gen- contribute at least this much to the overall power consumption erates the need for new electrode geometries and other clever as well. Comparisons between the CHEMFET, chemiresistor, solutions to improving signal-to-noise behavior for small and SPR sensor indicate that, in general, no single sensor (micron-scale) devices. CHEMFETs, while not hindered technology is ideal by itself for portable instrument design. by this noise scalability issue, are limited to certain device Comparisons among other chemical sensing technologies structures for microsensor-scale sizes, again complicating would delineate similar trade-offs. With these trade-offs in the miniaturization process. Acoustic wave devices require mind, however, it is possible to direct research and development higher frequencies of operation at smaller sizes, complicating effort for a particular chemical sensing application into the the readout electronics and incurring stringent constraints on weakest links in the system, whether they be related to power, impedance matching and noise management at these high WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 259 frequencies. The development of scalability methodologies Also, luminescence methods are sensitive to temperature fluctu- for miniaturizing most chemical sensor technologies is not ations and other environmental factors that quench fluorescence. straightforward and this complexity has hindered the progress There are insufficient examples of selective transducers that fa- of the chemical sensing microsystems development. cilitate broad deployment of UV/Vis or luminescence sensors. Perhaps even more significant are the specific requirements Infrared and Raman sensors exploit the vibrational structures on the choice of materials used for chemical microsensors. of potential target analytes. IR spectroscopy is difficult to im- While conventional integrated electronics use aluminum, plement for field use because water strongly absorbs across the polysilicon and now copper, the conductors (metallization) IR spectrum. Near-IR and Mid range IR have niche applica- on chemical microsensors often must be gold or platinum tions but lack the selectivity and sensitivity of IR spectroscopy. for sufficient inertness and electrochemical quality. Also, the The selectivity problem can be overcome with multivariate data quality of the final passivation layer of the devices that are analysis; however, the power requirements for NIR and MIR exposed to aggressive environment has to be higher than that light sources and detectors are relatively large compared to other of conventional integrated electronics. High density silicon spectroscopic techniques. Raman spectroscopy is very selective nitride is the material of choice for this purpose, because it with sharp spectral features, and relatively insensitive to water, is compatible with integrated circuit fabrication processes unlike IR spectroscopy. Therefore, analysis can be readily per- and provides immunity of the underlying electronics to the formed in aqueous environments. While Raman spectroscopy is ambient sensing environment. Finally, a high aspect ratio in perhaps the least sensitive of the optical methods, sensitivity can the z-direction is often required for chemical sensors in order be boosted by employing surface enhanced Raman spectroscopy to achieve optimum performance of the selective layers. High (SERS). The downside of SERS is the inability to reproducibly aspect ratios are not typical for integrated circuit fabrication create the small gold (or silver) particles required to induce the processes and incur further fabrication complexity into the surface enhancement and the long term stability of these parti- design and fabrication of chemical sensing microsystems. cles. Thus, SERS is often considered more of a qualitative than It is clear that specific and unusual tools are needed for quantitative technique. fabrication of chemical sensing arrays and the term chemical A relatively new optical method that promises sufficient electronics, coined in early seventies, truly describes this selectivity and sensitivity for environmental sensing appli- separate class of devices and their fabrication. Unfortunately, cations but can be readily miniaturized for minimal size and these specialized fabrication tools are usually not available in power consumption is surface plasmon resonance (SPR) spec- commercial silicon foundries. Thus, the lack of suitable fabri- troscopy. SPR is sensitive to minute refractive index changes at cation facilities for chemical electronics has hindered the entire a metal-dielectric surface. The physics of this effect are closely chemical sensing field in miniaturizing devices for portable related to the energy transfer that generates the enhanced sensing applications. However, progress in high-volume sensitivity of SERS while demanding only the instrumental microsensor applications, such as for monitoring carbon requirements of UV/Vis sensors. The one drawback of SPR monoxide and ozone, have generated commercial fabrication is the intrinsic lack of selectivity for molecular recognition. processes that will encourage continued progress toward more However, this lack of selectivity can be mitigated by uniquely diverse, multi-functional chemical sensing microsystems for employing antibodies as selective transductive mechanisms. many applications where portability and long battery life are SPR saves a transduction step compared to fluorescence or required. ELISA methods; the binding of the analyte to the sensor surface is directly detected by changes in the surface plasmon wave. B. Optical Sensors III. CHEMIRESISTORS Many classes of optical sensors are applicable to field portable chemical sensing. Naturally, each of the classes have The chemiresistor is, for reasons of its simplicity, perhaps intrinsic benefits and limitations. With the advent of diode the most attractive chemical sensor type for portable applica- lasers, light emitting diodes, and spectrographs with diode array tions. The conductivity (resistance) of the chemiresistor is pro- detectors that fit into the PCMCIA slot of a laptop computer, portional to the analyte concentration in the ambient environ- most spectroscopic techniques can be made field portable. ment. Fast response time in these sensors typically depends on a UV/Vis and luminescence spectroscopies exploit the inter- thin film of chemiresistive material applied between two contact action of light with the electronic structure of a potential an- or measurement points (electrodes). Under thin-film conditions alyte or perturbations in the electronic structure of a molec- (or alternatively: grains or pellets with high surface area), the ular transducer that the analyte induces. UV/Vis spectroscopy electrical resistance of the chemiresistor is dominated by surface is simple to construct and requires little power, but it lacks the rather than bulk characteristics, allowing response times to ap- sensitivity or selectivity for direct measurements. Luminescence proach a limit dominated by reaction rather than diffusion times. sensors are extremely sensitive to small quantities of many or- When dominated by surface, reaction-limited interactions, the ganic analytes of interest such as polycyclic aromatic hydrocar- chemiresistive film response can approach sub-second response bons (PAHs) and other compounds found in fossil fuels. How- and recovery times. ever, the broadband character of luminescence spectra and the Simplicity of fabrication and fast response times in the ubiquitous presence of naturally fluorescent compounds leads chemiresistor, however, are offset by inherent limitations in luminescence sensors to suffer from a critical lack of selectivity. selectivity and sensitivity. Chemiresistors tend to be broadly 260 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001 selective, responding to a large family of chemicals, such as all reducing gases in the case of the metal–oxide semiconductor chemiresistor. Relative response to these large families of chemicals can be tuned through the addition of catalysts and external filters or through the adjustment of operating param- eters such as temperature; however, tuning selectivity cannot entirely eliminate the influence of interferent chemicals but can only reduce their influence on the sensor response (resistance (a) change). The noise contributed by interferents in the ambient environment is compounded by the inherently noisy behavior of the resistor, where in nonideal electrode geometries, 1/ noise (flicker) in dc resistance measurements can often approach the desired sensitivity threshold of the sensor. ac resistance measurements are one way to overcome prohibitive 1/ noise, but they incur more complex measurement electronics and cal- ibration/sensor reproducibility issues. Capacitance is another attractive parameter to extract from the basic chemiresistive (b) sensor in an attempt to meet sensitivity requirements. Several Fig. 1. Behavior of the metal-oxide semiconductor chemical microsensor. efforts have explored the measurement of capacitance and po- When (a) no stimulus is present in the ambient environment, oxygen extracts electrons from the metal–oxide film thereby decreasing the sensor conductivity. larity changes in conducting polymers during interaction with When (b) a reducing agent is present, it combines (reduces) with the oxygen on chemical species. Ultimately, the influence of many potential the surface of the sensor and electrons are reinjected into the sensor material, interferents and the inherently noisy behavior of the resistor thereby incresing conductivity. restrict the sensitivity of the current miniaturized chemiresistor to relatively high thresholds in the sub-ppm range. Since the complex mixture evaluation have also become commercially chemiresistor response is dominated by surface interactions, available in the 1990s through companies such as AromaScan the batch reproducibility of these films is also highly variable (OsmeTech) and AlphaMOS [5], but their high price (tens of due to the inherent random variations in surface that occur thousands of dollars) and size prohibit them from being used in during fabrication processes. portable chemical sensing applications. Despite these drawbacks, the simplicity and low-cost of the Metal–oxide chemical sensors have enjoyed substantial chemiresistor, combined with the proper sensor array archi- popularity in both the research and commercial community tectures and signal conditioning and preprocessing, continue because of their ease of fabrication and stability relative to other to make this type of chemical sensor an attractive choice for solid-state sensors, including other chemiresistive materials chemical sensing microsystems. The two dominant types of such as polymers and organic films. Since the 1970s, the chemiresistors are fabricated using metal–oxide semiconduc- number of publications in technical journals regarding the use tors and polymers, and are addressed and reviewed herein. of metal–oxides for sensing materials has skyrocketed, with a majority of the research effort occurring in Europe and Japan. A. Metal–Oxide (Semiconductor) Chemiresistors A variety of metal–oxide films, in various film geometries, Metal–oxide semiconductor chemiresistors, based on thickness, structures, and configurations have been constructed tin–oxide, have been commercially available as discrete in research efforts to sense, with desired specificity, specific (thick-film or pellet) products, since the early 1970s through reducing gases among the many to which metal–oxide is Figaro Engineering [1] for carbon monoxide, combustible inherently sensitive. By the year 2000, investigation of the gas, and related toxic gas detection applications. Discrete basic properties of common metal–oxide films (e.g., tin–oxide, (thick-film) versions of mixed-metal oxide chemiresistor zinc–oxide) have been largely exhausted, but have given rise to sensors have been commercially available since the mid 1990s efforts targeted at optimizing selectivity, sensitivity, stability, through Capteur Sensors and Analyzers Ltd. for applications and integratibility for small or single-chip chemical sensors and in environmental monitoring, automotive, medical, industrial, microsystems. A summary of the most common metal–oxide aerospace, ozone, air quality, and similar vapor phase moni- films and a sampling of relevant (recent) publications regarding toring. The Capteur product line senses a wide variety of gases these films is outlined in Table II. In the design of portable with the higher selectivity and resistance to aging inherent in instruments, power consumption (directly related to tempera- mixed metal–oxide over single metal–oxide sensors; vapors ture of operation); selectivity/sensitivity, and stability typically that can be sensed effectively include ammonia, hydrogen, have the most influence on the choice of metal–oxide films hydrogen sulphide, ozone, propane, sulphur dioxide, hexane, for a particular application. The operation of the metal–oxide carbon monoxide, and other reducing gases [2]. Recently, Mi- chemical sensors is straightforward [Fig. 1(a) and (b)]. croChemical Systems [3] has taken over the Motorola carbon When oxygen in the ambient, vapor-phase environment monoxide sensor (based on tin–oxide and fabricated on a silicon reaches a certain level, the amount of oxygen on the sensor substrate, but sold as a discrete sensor). Arrays of chemire- surface is constant (saturated) and its oxidizing effect results sistive sensors containing a combination of metal–oxide and in the removal of electrons from the bulk of the semiconductor other types of chemiresistive sensors for “electronic nose” or and the subsequent reduction in free carrier concentration (and WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 261

TABLE II METAL–OXIDE SENSING FILMS AND THEIR COMPATIBILITY WITH PORTABLE INSTRUMENT DESIGN

increase in depletion region at the surface) reduces the film on the surface of the sensor. A variety of efforts have tuned this conductivity. A reducing gas removes oxygen from the surface selectivity to improve the quantitative resistance change to a (in proportion to its concentration), thereby reinjecting electrons single or smaller group of reducing vapors over other vapors. into the bulk of the semiconductor and decreasing the depletion A sampling of recent efforts to improve selectivity of these sen- region as well as the resistance. When the oxygen in the ambient sors is summarized in Table III. is below the saturation level, the metal–oxide sensor can be used Research efforts that focus on development of metal–oxide as an oxygen sensor. The sensitivity of the metal–oxide sensor is sensors have been widespread since the 1970s. For portable directly influenced by the size of the oxygen-induced depletion applications, metal–oxide semiconductors are largely hampered layer at the film surface relative to the thickness of the bulk by their power consumption demands. Typical metal–oxide semiconductor. In Figaro-type sensors (based on powders rather semiconductors operate at temperatures in excess of 200 C than films of tin–oxide), grain boundary effects are dominant and in order to achieve an appreciable response to the vapor of conductivity changes in response to the presence of an analyte interest. Commercially available products that use metal–oxide are largely a result of interactions at the grain boundaries rather semiconductors require that they be operated at these elevated than at the surface of the sensor. temperatures continuously to avoid instabilities that occur Unfortunately, the selectivity of the metal–oxide sensor is during cooling and heating. Efforts in the last decade by broad, responding to all reducing gases that interact with oxygen researchers at NIST [61], [62] and other groups have used 262 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001

TABLE III METHODS OF MODIFYING SELECTIVITY AND SENSITIVITY IN METAL–OXIDE FILMS

instabilities in cooling and heating of these sensors as indi- These and other efforts [61]–[63] have also isolated the sensors cators of gas type to improve selectivity and chemical vapor from the surrounding substrate through the use of hotplates discrimination. Cooling and heating sequences are determined and similar structures fabricated with both surface and bulk to generate signatures that are optimized for distinguishing micromachining techniques. Thermal isolation and intermittent the vapors of interest from closely related interferent vapors. operation of the heaters reduce the power consumption of the WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 263 sensors themselves to facilitate their use in portable applica- for other polymers (e.g., [76]–[78]) as has the benefit of ac over tions. dc measurement techniques [79]. Significant effort focusing on polymer films fabricated into B. Composite and Conducting Polymer Chemiresistors composite, conductive films have been conducted at the Cali- Unlike metal–oxide chemiresistors, which rely on a surface fornia Institute of Technology by Lewis et al. Various combina- interaction with oxygen to induce a resistance change in the bulk tions of the conductive agent, carbon black, and essentially in- film, composite and conducting polymers rely on the adsorption sulating polymers, are assembled into composite chemiresistor of an analyte into the polymer film to induce swelling that in arrays in large numbers to obtain both wide discrimination capa- turn generates a resistance change. In order to measure this re- bility and wide dynamic range (number of components to which sistance change, the polymer film has to be conductive or com- the array is sensitive) [80]–[82]. An important point regarding posite (combined conductor/insulator), a requirement that has the research at CalTech and related efforts elsewhere is that the limited the types of polymers that can be practically used for use of a composite film with a chemically insensitive conductor resistance measurements. Among the “electronic nose” instru- and chemically sensitive insulator opens the way to using poly- ments that are known to use conducting polymer films and com- mers for chemiresistors that were previously disregarded be- posite polymer films as part of their core sensor arrays are Aro- cause of their insulating properties. mascan and Cyrano Sciences, respectively. Unlike metal–oxide In the future, it is expected that the emphasis on polymer sensors, the development of purely conductive polymer-based chemiresistors will be on the optimization of films and mea- chemiresistors has been restricted to relatively few types of base surement techniques for generating low noise signals from these polymer layers including polypyrrole and polyaniline for their devices. Compromises between simple measurements (e.g., dc conductive characteristics. A larger variety is available in com- resistance) and more complex measurements such as ac resis- posite film chemiresistors that combine the characteristics of a tance, impedance and capacitance will be achieved to overcome chemically sensitive insulating polymer with a conductive ele- noise limitations in current polymer films. It is also expected ment such as carbon black. that these films will become highly integrated onto monolithic Humidity has a significant effect on the baseline structure arrays for vapor-phase sensing, making them promising candi- of these polymers and therefore on the swelling behavior in dates for portable instrument development. response to analytes in the sensing environment [64]. Noise in these layers has been a significant obstacle in using these IV. CHEMFET SENSOR TECHNOLOGIES polymers for practical chemical sensing applications [65]. Considerable noise in the dc measurement of polymer resis- The terms CHEMFET and potentiometric sensor have come tance has generated significant interest in ac measurements to be used interchangeably in the applications of miniaturized of the resistance as well as in impedance and capacitance solid-state chemical sensors. As the name implies, potentio- measurements. It is unclear how much additional useful infor- metric sensors derive their useful (analytical) information from mation is provided in capacitance measurements in terms of an explicit relationship between the potential of the indicator the polarization of molecules adsorbed into the polymer films. electrode and concentration in the analyte (vapor or liquid). Be- Polypyrrole films, by far, have dominated research efforts in cause the potential of a single phase cannot be measured, a polymer chemiresistors. Single sensors using polypyrrole films second or reference electrode is introduced to enable a voltage have been used to discriminate among single molecules such (or potential difference) to be measured between the indicator or as methanol, carbon tetrachloride, and methanol [66], toluene measurement electrode and the reference electrode. The need and ethanol [64], methanol [67], [68], and ethanol [69], as for a reference electrode is similar to the need for a ground well as complex mixtures including vintages of wine [64] and (floating, signal or earth) in any electrical circuit and as in elec- other vapors. Recent publications in the area include the use of trical circuits, may be constructed to reference the signal of in- ac measurement techniques that demonstrate desirable ohmic terest at the measurement electrode in a variety of ways. Using behavior of these films at greater than 10 MHz frequencies a reference electrode to enable a voltage measurement that is of operation [65]; improved selectivity provided by obtaining representative of changes in charge distribution on the measure- measurements at lower frequencies (20–10 kHz) for discrim- ment electrode uses established sensing principles. Reference inating among methanol, acetone, ethyl acetate, and ethanol and measurement electrodes are used in a variety of potentio- [70]; performance improvements induced by microelectrode metric ion-selective electrodes for both aqueous (liquid-phase) geometry optimization [71]; and improved selectivity to am- and vapor (gas-phase) measurements and macroscale ion-selec- monia obtained via doping with ClO and TsO [72]. The impact tive electrodes now represent the largest group among all chem- of humidity in degrading the quality of polypyrrole response ical sensors in commercial use today. to vapors of interest has also been studied extensively [64]. Potentiometric sensors can be miniaturized into a FET-based Other polymers studied for chemiresistor applications include architecture, called a CHEMFET, and are particularly suitable polystyrene in the form of beads cross-linked for mechanical for miniaturization because the integrity of the transduced integrity and used for measuring pH [73]; PVA (polyvinyl al- signal that carries chemical information does not depend on the cohol) for measuring humidity [74]; and polyaniline for sensing sensing area. Unlike chemiresistors, noise and S/N ratio do not ammonia across a wide dynamic (resistance) range from 10 degrade significantly with miniaturization. The trade-off to the to 10 [75]; The impact of humidity and temperature on chemiresistor, however, is that the miniaturized potentiometric hysteresis and stability has also been extensively investigated sensor is often larger than its chemiresistor counterpart and 264 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001 requires more complex measurement overhead. This increase in size can be partially compensated by the noise advantages of this type of device. Another disadvantage of the miniaturized potentiometric sensor is that the power of the measured signal containing relevant analyte information is very small and its measurement requires significant amplification at high input impedance before the signal can be further processed for concentration and discrimination information. Embedding the analyte-sensitive electrode into the field-effect transistor (a) (FET) structure is an attractive solution to the amplification problem. FETs, of course, can be fabricated in standard CMOS and other microfabrication processes and conversion of the electronic FET to an effective chemical sensor requires, in theory, only replacement of the gate with a suitable chemically sensitive material. The impact of the environment, materials compatibility, and other fabrication issues, however, make the integration of the CHEMFET with standard FET architectures significantly more complicated than the theory and has gen- erated a broad range of CHEMFET architectures and design (b) methodologies directed at solving practical, portable chemical Fig. 2. CHEMFET structure. Most CHEMFETs are fabricated on top of sensing problems. Because of the convenience and compat- a semiconductor substrate, where drain, source, and conductive channel are housed. The insulator layer directly on top of the substrate is typically ibility of the FET structure with standard microfabrication a traditional insulator layers such as SiO . On top of this layer is either a processes, miniaturized potentiometric sensors (microsensors) conductive gate of chemically sensitive material (the basic CHEMFET) or have become synonymous with chemically sensitive field-ef- a chemically sensitive materials to which ohmic and conductive contact is not possible (the ISFET or ENFET). Show is a general schematic diagram of fect transistors or CHEMFETs and continue to be pursued the chemically sensitive field-effect transistor operated either as (a) a basic as a viable solution to chemical sensing problems requiring CHEMFET or as (b) an ion-selective sensor (ISFET). portability and real-time operation. A schematic diagram of chemically sensitive field-effect tran- tifacts originating from the composition of the matrix of the sistor is shown in Fig. 2. Unlike a conventional MOSFET, the sample [86]–[88]. The same limitation applies to the ENFETs or threshold voltage of the CHEMFET can be chemically modu- miniaturizations of the potentiometric enzyme electrode which lated. The nature of this modulation has been reviewed in de- until now remain only a laboratory curiosity. The ISFET, on tail [83], [84] and it defines three major types of devices in the the other hand, despite its need for a reference point or elec- CHEMFET family. trode, has met with substantial progress in the past decade, in part due to the fact that its macroscale counterpart, the ion se- • ISFET (ion selective FET): the ion sensitive field effect lective electrode, has already matured into a variety of commer- transistor (ISFET), which is a direct miniature equivalent cial products. The situation is substantially better with the work of the ion selective electrode, is a CHEMFET structure function CHEMFET than with both the ENFET and ISFET in without a conductive gate; the ion selective layer is placed terms of compatibility with miniaturization for portable instru- on top of the insulator layer of the FET structure. ment development. The basic CHEMFET does not suffer from • ENFET (enzyme FET): enzymatically selective field ef- the problem of requiring a large reference electrode because the fect transistor, which is equivalent to a potentiometric en- silicon substrate in the FET structure serves as the stable ref- zyme electrode, is another CHEMFET structure without erence point in the circuit. Closing the circuit, however, is still a conductive gate; the chemically sensitive enzyme layer necessary (as with all circuits) to extract an electrical signal that forms part or all of the entire insulator layer of the FET is representative of the chemical stimuli in the environment. The structure. first and foremost example of a work function CHEMFET is a • Basic, or “work function” CHEMFET: this field effect MOSFET structure in which the palladium is used as the gate transistor has its macroscopic counterpart in the Kelvin metal. This hydrogen sensitive chamfer has been extensively in- Probe [85] and is a FET structure with a conductive gate; vestigated by I. Lundstrom [89]. Just as with chemiresistors and the conductive gate is the chemically selective area layer. other types of chemical sensors, it is obvious that partial, lim- The insulator can be a traditional material such as SiO , ited selectivity of individual sensors can be enhanced by assem- a special layer chosen for its chemically selective interac- bling them into arrays in which the individual sensors give a tion with the gas in generating a work function change, or more or less orthogonal response to multiple species in a real a sandwich of custom and traditional insulator layers. sample. This concept is not new, but it has become practical only The primary disadvantage of the ENFET and ISFET structure in the last ten years when computational capability dedicated is the inescapable need for a large reference electrode which to a chemical sensor has become physically and economically somewhat diminishes the practical appeal of individual small feasible. Since field-effect transistors are particularly suitable indicator devices. Potentiometric enzyme electrodes have only for miniaturization, it is no surprise that the first CHEMFET limited usefulness due to their vulnerability to experimental ar- arrays have already been reported [90]. Further contributing to WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 265 the attractiveness of the CHEMFET family of chemical sensor CHEMFETs are such that, like the chemiresistor, measurement technologies for miniaturized and portable field instruments is and transduction electronics can be easily designed to process the fact that, relative to other multi-stage transduction structures signals at least as fast as these reaction time limits, making such as the acoustic wave devices and optical devices for chem- real-time operation straightforward in integrated systems. ical sensing, CHEMFET supporting electronics are relatively The ISFET and basic CHEMFET, for the reasons discussed simple and can be accommodated on the same silicon substrate above, remain promising candidates for miniaturized portable as the sensors themselves without appreciably increasing the chemical sensing systems and are discussed and reviewed overall size of the sensing system. It has been suggested that in further detail. The ISFET has achieved maturity in the some signal processing circuitry could be also incorporated on late eighties and research efforts in the area of ISFETs have the same chip, although advantages of full, single-chip integra- been recently reviewed. One particularly notable development tion are not always clear-cut for chemical sensing applications affects not only ISFETs but the entire class of potentiometric due to the need to protect circuits from the sensing environment. ion sensors that use polymeric membranes. Harrison et al. have Multi-chip modules, where the circuits are connected in near investigated in detail the distribution of water in fully hydrated proximity to the CHEMFET devices, but encapsulated as a sep- polymeric ion-selective membranes and found that there is arate entity to protect them from the sensing environment, are a sharp gradient of water concentration inside the polymer more practical than monolithic chips and developments in the at its interface with water. These membranes have long been integration of application specific circuits (ASICs) into such hy- considered uniformly hydrophobic and the interface between brid chemical sensing packages can be anticipated. the membrane and the solution was considered to be planar. Also adding to the importance of the CHEMFET in the de- The existence of a hydrated layer of finite thickness has been velopment of future portable chemical sensing systems is the established by light scattering [91] and by NMR [92]. This continued widening in the gap between fabrication of conven- finding substantially changes the understanding of kinetics, tional electronics and fabrication of integrated chemical sensors. interpretation of impedance behavior and selectivity and sets While the dimensions of conventional electronic elements are the practical limits on the minimum optimal thickness of now on the order of a tenth of a micron, the dimensions of the ion-selective membranes on ISFETS. It also underscores other gate area of transistors forming the basis of CHEMFETs have issues related to water ingress into the membrane, namely the remained the same as they were in mid seventies (on the order of effect of osmotic pressure in ISFET structures with hydrogel tens of microns). The widening of this gap is important because intermediate layer [93], thereby providing an advance in it effectively decreases the size disadvantage of the CHEMFET the fundamental science of these membranes that is critical devices over simpler structures such as the widely used chemire- to formalized design methodologies for assembling these sistor. Both devices in practical terms consume space equal to membranes into practical and useful structures for commercial the chemical sensing area plus the measurement and control applications. electronics. Since the size of chemiresistors and CHEMFETs The basic CHEMFET is a miniaturization of the Kelvin Probe has remained effectively the same in the past decade on the mi- (vibrating capacitor). The heart of a basic CHEMFET is the croscale, the reduction in size of the electronics required in the gate capacitor in which silicon forms one plate while the chem- CHEMFET infrastructure effectively makes it more competitive ically selective layer forms the other. When the two chemi- with the real-estate consumed by the chemiresistor and its asso- cally different plates are electronically connected, the equaliza- ciated support electronics. Continued advances in integrated cir- tion of Fermi levels leads to formation of electric field in the cuit miniaturization will continue to improve the attractiveness dielectric. This field is proportional to the difference of work of the CHEMFET structure for miniaturized chemical sensing function of the two plate materials [94]. A remarkable prop- applications. erty of the silicon-based devices is the fact that silicon is her- Unfortunately, continued advances in integrated circuit elec- metically sealed by silicon dioxide (and silicon nitride). There- tronics have also negatively impacted the compatibility of the fore, the work function of silicon does not change over the en- CHEMFET structure with microfabricated architectures. While tire operating range of temperatures and thus serves as a stable the tendency in conventional electronics is to decrease the gate reference electrode. This aspect of conventional silicon elec- oxide thickness to dimensions on the order of 10 s and 100 s tronics is normally taken for granted and generally overlooked, of angstroms, the thickness of the composite oxide/nitride in- but is a tremendous advantage of Kelvin probe miniaturization sulator in chemical electronics is typically 1200–1600 Å. This into integrated CHEMFET structures. Direct translation of the increase in gate oxide thickness is driven by the need to pro- Kelvin probe leads to the suspended gate field-effect transistor vide passivation integrity to the chemical sensing electronics (SGFET) which was also reviewed previously [85]. The SGFET lying underneath the ChemFET structures. While this difference Fig. 3 is a form of CHEMFET where the conductive gate is sus- in gate-oxide thickness is unfortunate, it is not prohibitive and pended above the insulator layer so that solution or vapor can thicker gate oxides are possible in order to provide passivation directly interact with both gate and insulator layer to produce a and protection of underlying support electronics. change in work function of the device. The main motivation for As a final general comment on the usefulness of CHEMFETs fabricating this type of transistor is the fact that it is possible to for portable instrument development, no special requirement on modify electrochemically the suspended gate and thus change the speed of CHEMFETs is anticipated, since the response time its selectivity for gases of interest. The second advantage of this is governed by the kinetics of the chemical interactions or by the sensor is that the gate can be operated at different temperatures, mass transport. The fundamental limits in response time of the thus achieving a certain additional selectivity [95]. The SGFET 266 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001

(a) (b) (c) Fig. 3. Common implementations of the basic CHEMFET. The (a) suspended gate field effect transistor (SGFET) is the most straightforward miniaturization of the Kelvin Probe and consists of a regular MOSFET structure, with the gate suspended over the insulator layer so that both chemically sensitive gate and insulator are exposed to the sensing environment. Other common structures include the (b) ordinary CHEMFET which is identical to the MOSFET structure except the gate is chemically sensitive and the (c) extended gate FET (EGFET) EGFET where the electronic gate and most of the FET are housed in a location that is relatively remote from the chemically sensitive area where signal transduction occurs. and other common basic CHEMFET structures which are modi- remain the size of the reference electrode. Enzyme sensors have fications of this direct miniaturization of the Kelvin probe struc- their recognized inherent limitations and it is not surprising ture are depicted in Fig. 3. that there has been no major development or application of A different approach has been introduced by using con- these types of microfabricated devices. On the other hand, the ducting polymers dissolved in and cast from organic solvents prospects for potentiometric gas sensors based on modulation for the chemically sensitive material in the basic CHEMFET. of work function are good. These sensors do not suffer from A typical example of such a polymer is polyaniline (PANI) the need for a large reference electrode since the fully enclosed which can be prepared chemically or electrochemically and silicon is an excellent reference. The lack of a reference elec- then dissolved and cast from such compounds as formic acid. trode is a major advantage of the basic CHEMFET over other This technique makes it possible to deposit the chemically types of miniaturized potentiometric sensors. Furthermore, the sensitive and electroactive gate materials and to pattern them range of gas-solid interactions suitable for this type of sensing photolithographically directly on top of the gate dielectric [96]. is diverse. With introduction of lithographically processable Subsequent chemical or electrochemical modification of this organic semiconductors and their tunable selectivity, the field material can be then performed [97] which opens a possibility of potentiometric gas microsensors will continue to grow. for construction of a wide range of chemical sensors for detec- tion of electrically neutral species in nonconducting samples, V. O PTICAL SENSORS:SURFACE PLASMON RESONANCE (SPR) such as gases or dielectric liquids. This technique and other Surface plasmon resonance (SPR) is achieved by optically ex- extensive research in all three major classes of CHEMFETs citing a standing charge density wave on a thin dielectric (metal) continue to be published in the technical literature. In addition surface. The physics of the SPR effect has been extensively de- to research reviews published in the 1990s noted previously, a scribed [117], [118]. The surface plasmon wave (SPW) propa- sampling and summary of recent efforts in CHEMFET research gates along the metal-substrate interface a wave vector ( ) is provided in Table IV to provide the reader with a basic perspective on the ongoing distribution of research effort in this (1) area of chemical sensor technologies. The term FE refers to the functionalizing element that enables the chemical sensitivity in where and are the wavelength dependent complex di- the topmost CHEMFET layer. electric permittivities of the (m)etal and (s)ample. is the free Although less prevalent than in chemiresistor development, space wave vector equal to . When exciting an SPW, the research efforts in the CHEMFET area have also been com- component of an incident photon that is plane polarized perpen- mitted to the development of suitable packaging, chemical dicular to the glass–metal interface must equal the value of sensing electronics, and integration levels for the construction to allow energy transfer from the incident light to the SPW. The of portable systems. Efforts to integrate all signal processing wave vector of TM polarized component of an incident photon into a microcontroller-based system have been reported [115] as ( )is have efforts to develop circuit models of CHEMFET structures in SPICE [116] for future chemical sensing microsystem and (2) electronics development. These efforts are gaining momentum as the demand to turn the vast amount of research attention where is the refractive index of the substrate supporting onto the CHEMFET structure and materials into viable systems the metal surface and is the angle of incidence. Thus, the continues to increase. performance of the SPW is determined by The last ten years have not introduced any major fundamental • the angle of the incident photon; changes or improvements in the ion and enzyme based poten- • the energy of the incident photon; tiometric microsensors (the CHEMFET family of chemical • the complex dielectric constant of the metal film; sensor technologies). The limiting factor on relative usefulness • the refractive index of the substrate on which the metal of ISFETs, as compared to their macroscopic counterpart, will film is deposited; WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 267

TABLE IV RECENT RESEARCH EFFORTS IN THE AREA OF CHEMFET DESIGN

• the refractive index of the sample at the metal surface. is transferred from the SPW to the medium. Consequently, the If the wave function of the SPW matches the wave function of resonant photon, in contrast to the nonresonant photons, is not the environment at the metal surface, the energy of the photon reflected off the metal interface. 268 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001

A more in depth analysis of the SPR physics leads to four optical length for resolution of angular dispersion, would be salient features that govern the geometry and application of SPR prohibitively large for any in-situ applications. sensors. A miniature Kretschmann configuration sensor is com- mercially available through Texas Instruments [123]. The • The SPW cannot be excited by direct illumination. Thus, approximately 3″ 2″ 1″ Spreeta sensor head contains an the exciting photon and the sample must be on opposite LED source, optics, and a diode array detector [124]. While the sides of the metal layer. sensor achieves 5 10 RI precision that is comparable to • The evanescent field of the incident photon excites SPW. bench-top systems, the dynamic range of the sensor is limited Thus, the refractive index of the substrate supporting the to 1.320 to 1.368 RI. For field applications, the sensor is metal layer ultimately limits the dynamic range of any incapable of in-situ analyzes; the sample must be brought to the SPR sensor sensor head with a flow cell attachment. However, a potential • The field strength of the SPW decays exponentially into modification to the system is to build the system around a the sample media. Thus, SPR is primarily a surface tech- replaceable cuvette [125]. This would eliminate power con- nique; however, SPR sensors respond to bulk RI changes. sumption of the flow cell and associated problems stemming • The real component of the refractive index (complex per- from fluid handling. The Spreeta sensor requires connection mittivity) of the sample has a significantly greater influ- to a computer controller/power supply via a standard serial ence on the SPW than does the imaginary component of cable. Field sensing application of the Spreeta system has been the refractive index (complex permittivity) of the sample. demonstrated for analysis of trinitrotoluene (TNT) from soil Thus, while SPR can be employed in highly absorbing extracts [126]. media, bulk absorbance changes of a sample will induce Modification of the Kretschmann configuration for white minor changes in SPR reflectance profiles. light SPR sensors affords a more compact sensor head that can be coupled with fiber optic light guides for remote sensing ap- Many of the references discussed in this section specifically ad- plications. Collimated, polarized, white light is reflected off the dress mitigating the limitations of one or more of these fea- gold coated prism surface at a predetermined angle. Fiber optic tures. The three key components of an SPR sensor are the op- jumpers can be employed to minimize the optics at the probe tical train, the choice of metal and substrate, and analyte spe- head; Oonly collimating lenses and a broadband polarizer are cific receptor. The optical train of an SPR sensor is designed needed near the prism. Stemmler et al. compared 5 geometries around one of two sensing philosophies. Constant wavelength for miniaturizing prism-based white light SPR sensors [127]. sensors relate the SPR angle of incidence of maximum attenua- With collimating lenses at the prism/fiber interface, a detection tion to analyte concentration. Constant angle sensors relate the limit of 1 10 Refractive Index Units (RIU) was observed. wavelength of maximum attenuation to analyte concentration. Removing the lenses and mounting the fibers flush with the Coinage metal layers between 30 nm to 50 nm thick support prism induced shallower SPR dips. However, detection limits the SPW. Silver yields the sharpest and most defined attenua- between 1.0 10 RIU and 1.7 10 RIU were still tion dip of all metals; however, gold is more often employed observed. Further compacting the probe head design by placing for sensor applications due the intrinsic resistance to corrosion the excitation and collection fibers on the same side of the and ability to functionalize the surface with specific receptors. prism and adhering a mirror to the reverse side of the prism, Because, SPR is ultimately sensitive to only refractive index yielded shallower and broader SPR spectra while degrading changes at the metal-sample surface, all selectivity is derived the detection limit by a factor of 4. The effect of prism size on from the inclusion of analyte receptors on the sensor. sensitivity and precision of analysis was inconclusive. Benchtop, SPR instrumentation traditionally employs the An alternative to prism based designs is to perform SPR anal- constant wavelength–multiple angle sensing paradigm. This ysis on optical fibers. Single mode optical fibers are often em- is the basis of the commercially available BIAcore [119] and ployed because the single mode fibers support sharp SPR tran- IAsys [120] SPR systems. These “Kretschmann configuration” sitions as only one angle of light (mode) propagates down the sensors focus a laser through a BK7 glass prism. A replacable, fiber at a given wavelength and the fibers preserve the polar- metal-coated glass slide is used for the sensing area. The ization of the incident light. The sensors are constructed by pol- normalized intensity of reflected light (versus an air or water ishing through the cladding of the fiber and metal coating the ex- reference) is plotted against the incident angle of the photon posed core. There are two disadvantages associated with single [121], [122]. While bench-top variable angle SPR sensors are mode fibers for optical sensors. The difficulty of affixing a metal arguably more precise than bench-top white light SPR sensors, mirror to the distal end of a single mode fiber prohibits con- the cumbersome optical train of the Kretschmann configuration structing the sensor in a “dip probe” configuration. Hence, the limits the ability to miniaturize the sensor head. Assuming excitation and collection optics must be on opposite ends of the that the ideal field sensor is to have no moving parts, a high fiber which leads to a large probe head at the sample interface. resolution array detector must be securely mounted with a rigid Also, the dynamic range, in RIU, of the sensors is very lim- optical path in line with the glass prism. Fiber optic connectors ited. The sensors only respond over a 0.01 RIU range. Thus, can be employed to guide light to the sensor probe head, but the dynamic range must be shifted to lower refractive indices cannot be employed to return the reflected light to the detector by increasing the thickness of the metal layer [128], [129] or because all angular information would be lost in the fiber. The shifted to a higher refractive index by adding a high Refrac- probe head, containing an array type detector and sufficient tive Index (RI) overlay such as SiO [130] or Ta O [129]. The WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 269 problem with adding overlays is that the sensitivity of the sensor While all SPR sensor geometries have demonstrated detec- decreases with increasing overlay thickness. A flow injection tion limits between 10 RIU and 10 RIU, the degree of in- sensor system employing a dual-channel, single mode wave- teraction between the analyte and sensor ultimately determines guide (one channel for reference, one channel for analysis) was the sensitivity and selectivity of the sensor system. Antibodies constructed for determination of simazine, a triazine class herbi- are often employed to achieve a high degree of specificity for the cide. Detection limits of 1 10 , that correspond to 0.11 mg/L sensor surface. The chemistry for covalently binding antibodies simazine, have been reported for this particular sensor and ap- to the gold surface has been well documented. A self-assem- plication [131]. bled monolayer of alkanethiol links is usually employed where The smallest probe head dimensions are achieved by the density of antibodies at the surface may be increased by performing SPR analysis on multimode optical fibers. This binding the antibodies to an intermediate dextran layer [137], configuration offers the simplest optical train consisting of two [138]. For large molecule detection, this sensing scheme is sen- collimating lenses and a bifurcated fiber. The interchangeable sitive to fractions of a monolayer of coverage (picograms per sensor tips are attached with a standard fiber-optic connector. square millimeter) [139]. However, the binding of smaller ana- The overall power requirements of this configuration are low lytes, such as pesticides, do not induce a significant change in as use of lap-top computer based diode array detectors [132], the surface RI. Thus, competitive binding assays are employed [133] for spectral collection has been reported and white LED’s to increase the sensitivity of the sensor. The antibody-coated provide sufficient illumination to rapidly collect SPR spectra. probe is pretreated with mass-tagged analytes such as bovine The practical dynamic range for silica fibers extends from 1.36 serum albumin (BSA). The presence of untagged analytes will RIU to 1.45 RIU; analysis can be performed as low as 1.30 displace the BSA-tagged analytes. Alternately, the analyte will RIU although the sensitivity is quite low. The dynamic range be bound to the sensor surface prior to pretreating the probe of multimode sensors can be extended beyond the 1.42 RI of with antibodies; free antigens will compete with the bound anti- silicon by employing high RI core fibers such as sapphire [134]. gens for the antibodies [140]. In either case, displacement in- The practical dynamic range can be tuned to lower RIU by duces a blue shift in the SPR spectrum. The competitive binding beveling the tip of the fiber where the SPW occurs. However, assay sensing strategy is limited to point monitoring because multimode fiber based SPR spectral dips are neither as sharp the probe tip must be regenerated or replaced after each mea- nor as deep as prism based SPR spectra. Multimode fibers surement. Continuous field monitoring can be performed if a are not polarization preserving so the maximum attenuation is reversible coating is developed for specific applications. For ex- 50% due to the constant reflected background from parallel ample, changes in RI of a polymer coating with absorption of polarized light that does not excite an SPW. Because the fibers nonpolar organics has been employed as a transduction layer for support a wide range of angles of total internally reflected SPR based vapor detection [141]. light, photons of the same wavelength will have a wide range The major bane to SPR analysis is non specific binding on of electric field wave vectors as predicted by equation (2). the sensor surface and temperature fluctuations of the sample. Wide range serves to broaden features on the observed SPR Fouling of the SPR probe tip and the bulk refractive index reflection spectra. The broader features degrade the precision changes of water (approximately 10 RIU per degree Kelvin of commonly employed calibration methods that are based at room temperature) can impair the accuracy of SPR analysis on the location of the spectral minimum. Consequently, when [139]. To mitigate the bias derived from nonspecific binding these calibration methods are employed, multimode fibers will and temperature fluctuations, dual sensing area SPR probes demonstrate detection limits an order of magnitude (or more) have been designed for angular [142] and white light-prism worse than prism based SPR sensors. [143] systems. One of the sensing areas responds to analyte Accurate and precise calibration of multimode SPR sensors is binding as well as bulk and temperature effects; the other facilitatedbyemployingmultivariatecalibrationmethodssuchas reference area responds to just bulk and temperature changes. those based on principal components regression. With traditional The estimated refractive index at the first area is adjusted calibration, the precision of analysis is determined by the preci- based on changes in the signal from the reference area. The sion and accuracy of which the true minimum of the reflectance reference area may be coated with a high RI overlay to shift spectrum can be determined. Thus, the broad dips from multi- the SPR signal to higher angles or wavelengths; this permits mode fiber optic sensors are difficult to calibrate based on the lo- both the analyte and reference signals to be collected simul- cationof thespectral minimum. Multivariatecalibrationemploys taneously on an array detector. Like the prism based sensors, the whole spectrum for calibration; this results in a more favor- tapered fiber optic probes can also be constructed to generate ableutilizationoftheSPRsignal.Multivariatecalibrationhasper- wavelength resolved signals from spatially resolved sensing formed equivalently to minima hunt calibration when the spectra areas [144]. The limitation to this strategy is the assumption employed with multivariate calibration data had 1/64 the resolu- that nonspecific binding occurs equivalently on the analyte and tion as the minimum hunt data [132], [133]. The ability to cali- reference areas. However, novel coatings that are resistant to brate broad SPR dips from tapered fibers has been further demon- nonspecific binding of proteins [144] should also prove helpful strated where five different multivariate calibration methods are in mitigating this limitation. comparedonspectracollectedwithtaperedfibers[135].Truepre- Many of the pieces needed for constructing a practical, field diction errors for weight-percent sucrose in solution were deter- portable SPR sensor have been developed. However, the pieces mined to be equivalentto the precision of the constructed samples have yet to be assembled into a working system. Prism-based [136]. sensors can be employed for applications where sampling is ac- 270 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001 ceptable and “dip probe” fiber optic sensors can be constructed the field. Overall, the tremendous research effort dedicated to for in situ analyzes. Multivariate calibration methods overcome tuning and optimizing sensor materials and structures is now in many of the limitations associate with the broad spectral fea- the phase where systems integration can be fruitful and flexible tures induced by miniaturization of SPR sensor geometries. Fur- in taking advantage of this large body of research and available thermore, with multivariate calibration, the need for high res- chemical sensor technologies. olution spectra is minimized, thus permitting low power-low Optical sensors, which are often dismissed for portable sys- spectral resolution sensor systems with sufficient accuracy and tems due to size and weight issues, are now at the stage where precision for field analysis. The number of commercially avail- portable, in-situ instruments for chemical sensing using opti- able antibodies is still growing. Therefore, once a stable, field cally-based sensors as well as nonoptical methods can be ef- portable SPR platform is developed, the technology should ex- ficiently designed and implemented as follow-up systems to pand rapidly into numerous applications. The limitation of SPR first generation portable optical systems. As with nonoptical sensors needing surface regeneration may be overcome by mi- chemical sensors, issues of integration, packaging, and calibra- crofluidics, otherwise SPR sensors will be mostly restricted to tion must be carefully considered and streamlined in the design point monitoring. process to meet stringent size, weight, and power constraints. Optical sensors offer the additional advantage of remote sensing which is useful both as a precursor to on-site sensing and as an VI. CONCLUSIONS alternative to autonomous vehicle sensing in toxic sites not ac- In this paper, we have reviewed three representative families cessible by human personnel. of chemical sensor technologies that are suitable for implemen- From the current state-of-the-art in miniaturizing both tation in portable instrument design. The chemiresistor repre- nonoptical and optical chemical sensor technologies, future sents the simplest of miniaturizable technologies for portable developments in portable instruments are anticipated to chemical sensing instruments; however, it also presents signif- progress more rapidly than in previous years. It is expected icant obstacles in terms of noise, drift, aging, and sensitivity to that introduction of products into the medical or automotive environmental parameters. The CHEMFET, while a more com- markets (both high volume) will provide an infrastructure plex yet still miniaturizable solid-state chemical sensor tech- for chemical sensing microsystems fabrication processes that nology, does not exhibit many of the fluctuations and instability is currently not available on a widespread basis. A mature of the chemiresistor, but also requires more complex signal pro- fabrication process will naturally lead to increased attention cessing. The SPR sensor holds great promise for miniaturizable to packaging, reproducibility, and to the optimal level of inte- optical sensors since it does not require the use of a laser in a gration that is currently not a focus of the research community portable system (the SPR probes can operate using an LED as in this area. Because of the complex, highly variable, and the light source). These three technologies have been chosen for interdisciplinary nature of chemical sensing microsystems and review since they represent a spectrum of a) simple to complex; chemical sensing electronics, a synergy between research and b) solid-state to optical; c) high to low noise; and d) specific commercial product development communities is essential to to broadly selective set of sensor technologies that can all be realizing practical nonlaboratory chemical sensing products miniaturized for portable instrument development. from the wealth of research dedicated to chemical sensors and A great deal of research effort has been expended on the their support systems. development and optimization of sensor films and materials, yet comparatively little effort has been committed to assembling these technologies into viable compact, low-power systems REFERENCES that are suitable for handheld sensing instruments. Attention to [1] Figaro Engineering Inc. [Online]. Available: http://www.fi- instrument design at the system level requires more attention garosensor.com [2] Capteur Sensors (recently acquired by First Technology). [Online]. to packaging for aggressive environments, to sampling for Available: http://www.capteur.co.uk. reproducible measurements, to optimization of electronic [3] MicroChemical Systems. [Online]. Available: http://www.microchem- infrastructure and overhead, and to improvements in overall ical.com. [4] OsmeTech plc. [Online]. Available: http://www.aromascan.com. signal-to-noise performance. Integration and miniaturization, [5] Aroma Scan. [Online]. Available: http://www.alpha-mos.com. in and of themselves, are not always advantageous in chemical [6] D. H. Dawson, G. S. Henshaw, and D. E. Williams, “Description sensing systems. Miniaturization often enhances noise issues, and characterization of a hydrogen sulfide gas sensor based on Cr Ti O ,” Sens. Actuators B, vol. B26, no. 1–3, pp. 76–80, subsequently requiring alternative design methodologies and 1995. sensor structures to obtain desired sensitivity levels. Protection [7] M. Fleischer and H. Meixner, “Gallium oxide thin films: A new mate- of the chemical sensing electronics and incompatibilities rial for high temperature oxygen sensors,” Sens. Actuators B, vol. 4, pp. 437–441, 1992. between sensor and integrated circuit materials can often [8] M. Fleisher, M. Seth, C. D. Kohl, and H. Meixner, “CO-sensor for do- counteract the advantages of full integration onto a monolithic mestic use based on high temperature Ga O thin films,” Sens. Actua- substrate. Hybrid structures such as multi-chip modules then tors B, vol. 4, pp. 437–441, 1992. [9] J. Frank, M. Fleischer, H. Meixner, and A. Feltz, “Enhancement of sensi- become practical alternatives to full integration while providing tivity and conductivity of semiconducting Ga O gas sensors by doping similar improvements in size and power consumption as their with SnO ,” Sens. Actuators B, vol. B49, no. 1–2, pp. 110–114, 1998. fully integrated monolithic counterparts. Auto calibration [10] Gurlo, M. Ivanoskaya, N. Barsan, M. Schweizer-Berberich, U. Weimar, W. Goepel, and A. Dieguez, “Grain size control in nanocrystalline In O methods embedded into the signal processing are also desired semiconductor gas sensors,” Sens. Actuators B, vol. 44, pp. 327–333, for many of these systems to operate at acceptable lifetimes in 1997. WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 271

[11] T. Doll, A. Fuchs, I. Eisele, G. Faglia, S. Groppelli, and G. Sberveglieri, [38] W. M. Sears, K. Colbow, and F. Consadori, “Algorithms to improve the “Conductivity and work function ozone sensor based on indium oxide,” selectivity of thermally-cycled tin–oxide gas sensors,” Sens. Actuators, Sens. Actuators B, vol. 49, pp. 63–67, 1998. vol. 19, pp. 333–349, 1989. [12] D. Mutschall, K. Holzner, and E. Obermeier, “Sputtered molybedenum [39] , “General characteristics of thermally cycled tin oxide gas sen- oxide thin films for NH detection,” Sens. Actuators B, vol. 35–36, pp. sors,” Semiconductor Sci. Technol., vol. 4, pp. 351–359, 1989. 320–324, 1996. [40] G. Kortochenkov, V. Brynzori, and S. Dmitriev, “Electrical behavior of [13] M. Ferroni, V. Guidi, G. Martinelli, M. Sacerdoti, P. Nielli, and G. SnO thin films in humid atmosphere,” Sens. Actuators B, vol. 54, no. Sberveglieri, “MoO -based sputtered thin films for fast NO detection,” 3, pp. 197–201, 1999. Sens. Actuators B, vol. 48, pp. 285–288, 1998. [41] S. G. Ansari, P. Boroojerdian, S. K. Kulkarni, S. R. Sainkar, R. N. [14] K. Ihokura, The Stannic Oxide Gas Sensor: Principles and Applica- Karekar, and R. C. Aiyer, “Effect of thickness on H gas sensitivity of tions. Boca Raton, FL: CRC, 1994. SnO nanoparticle-based thick film resistors,” J. Mater. Sci.: Mater. [15] G. Korotchenkov, V. Brynzari, and S. Dmitriev, “SnO films for thin Electron., vol. 7, pp. 267–270, 1996. film gas sensor design,” Mater. Sci. Eng., vol. B56, pp. 195–204, 1999. [16] V. Brynzari, S. Dmitriev, and G. Kortochenkov, “SnO based sensors [42] W. Hellmich, G. Mueller, Ch. Bosch-v Braunmuehl, T. Doll, and I. of smoke,” in Proc. Intl. Conf. Microelectron., Yugoslavia, 1997, pp. Eisele, “Field-effect-induced gas sensitivity changes in metal oxides,” 547–550. Sens. Actuators B, vol. 43, pp. 132–139, 1997. [17] G. Kortochenkov, V. Brynzari, and S. Dmmitriev, “SnO films for thin [43] Y. Kato, K. Yoshikawa, and M. Kitora, “Temperature-dependent dy- film gas sensor design,” Mater. Sci. Eng. B: Solid State Mater. Adv. namic response enables the qualification and quantification of gases by Technol., vol. 63, no. 3, pp. 195–204, 1999. a single sensor,” Sens. Actuators B, vol. 40, pp. 33–37, 1997. [18] G. Sberveglieri, L. Depero, S. Groppelli, and P. Nelli, “WO sputtered [44] P. H. Wei, G. B. Li, S. Y. Zhao, and L. R. Chen, “Gas-sensing properties thin films for NO monitoring,” Sens. Actuators B, vol. 26–27, pp. of Th/SnO thin-film gas sensor to Trimethylamine,” J. Electrochem. 89–92, 1995. Soc., vol. 146, pp. 3536–3537, 1999. [19] L. E. Depero, M. Ferroni, V. Guidi, G. Marca, G. Martinnelli, P. Nelli, L. [45] C. Xu, J. Tamaki, N. Miura, and N. Yamazoe, “Grain size effects on gas Sangaletti, and G. Sberveglieri, “Preparation and microstructural charac- sensitivity of porous SnO -based elements,” Sens. Actuators B, vol. B3, terization of nanosized thin films of TiO –WO as a novel material with high sensitivity to NO ,” Sens. Actuators B, vol. 35–36, pp. 381–383, no. 2, pp. 147–155, 1991. 1996. [46] N. Yamazoe, Y. Kurokawa, and T. Seiyama, “Effects of additives on [20] A. A. Tomchenko, I. L. Emilianov, and V.V.Khatko, “Tungsten trioxide- semiconductor gas sensors,” Sens. Actuators, vol. 4, no. 2, pp. 283–289, based thick-film NO sensors: Design and investigation,” Sens. Actuators 1991. B, vol. 57, no. 1–3, pp. 166–170, 1999. [47] V. A. Chaudhary, I. S. Mulla, and K. Vijayamohanan, “Selective hy- [21] P. Mitra, A. P. Chatterjee, and H. S. Maiti, “ZnO thin film sensor,” Mater. drogen sensing properties of surface functionalized tin oxide,” Sens. Ac- Lett., vol. 35, pp. 33–38, 1998. tuators B, vol. 55, no. 2, pp. 154–160, 1999. [22] A. Jones, T. A. Jones, B. Mann, and J. G. Firth, “Effect of the physical [48] C. V. G. Reddy, S. V. Manorama, and V. J. Rao, “Effect of Mo N on the form of the oxide on conductivity changes produced by CH CO, and gas-sensing characteristics of SnO -based sensors,” J. Mater. Sci. Lett., H O on ZnO,” Sens. Actuators, vol. 5, no. 1, pp. 75–78, 1984. vol. 18, no. 9, pp. 673–676, 1999. [23] B. Bott, T. A. Jones, and B. Mann, “Detection and measurement of CO [49] N.-L. Wu, S.-Y. Wang, and I. A. Rusakova, “Inhibition of crystallite using ZnO single crystals,” Sens. Actuators, vol. 5, no. 1, pp. 488–495, growth in the sol-gel synthesis of nanocrystalline metal oxides,” Sci- 1984. [24] S. Saito, M. Miyayama, K. Koumoto, and H. Yanagida, “Gas sensing ence, vol. 285, no. 5432, pp. 1375–1377, 1999. characteristics of porous ZnO and Pt/ZnO,” J. Amer. Ceram. Soc., vol. [50] C. V. Gopal Reddy and S. V. Manorama, “Room temperature hydrogen 68, no. 1, pp. 40–43, 1985. sensor based on SnO :La O ,” J. Electrochem. Soc., vol. 147, no. 1, [25] S. Pizzini, N. Butta, D. Norducci, and M. Palladino, “Thick film ZnO pp. 390–392, 2000. resistive gas sensors. Analysis of their kinetic behavior,” J. Electrochem. [51] B. Bhooloka Rao, “Zinc oxide ceramic semiconductor sensor for ethanol Soc., vol. 136, no. 7, pp. 1945–1948, 1989. vapor,” Mater. Chem. Phys., vol. 64, no. 1, pp. 62–65, 2000. [26] T. Minami, H. Sato, H. Nanto, and S. Takata, “Highly conductive and [52] Heilig, N. Barsan, U. Weimar, M. Schweizer-Berberich, J. W. Gardner, transparent silicon doped zinc oxide thin films prepared by RF mag- and W. Gopel, “Gas identification by modulating temperatures of netron sputtering,” Jpn. J. Appl. Phys., Part 2: Lett., vol. 25, no. 9, pp. SnO -based thick film sensors,” Sens. Actuators B, vol. 43, pp. 45–51, 776–779, 1986. 1997. [27] S. Takata, T. Minami, and H. Nanto, “Stability of aluminum-doped ZnO [53] Chu, D. Winnie, M. Jamal, and R. H. Hill, “Sensors for detecting transparent electrodes fabricated by sputtering,” Thin Solid Films, vol. sub-ppm NO using photochemically producted amorphous tungsten 135, no. 2, pp. 183–187, 1986. [28] S. Basu and A. Dutta, “Modified heterojunction based on zinc oxide thin oxide,” J. Electrochem. Soc., vol. 145, no. 12, pp. 4219–4225, 1998. film for hydrogen gas sensor application,” Sens. Actuators B, vol. 22, no. [54] M. Penza, C. Martucci, and G. Cassano, “NO gas sensing characteris- 2, pp. 83–87, 1994. tics of WO thin films activated by noble metals (Pd, Pt, Au) layers,” [29] P. Esser, R. Feierabend, and W. Gopel, “Kinetics of the catalytic co-ox- Sens. Actuators B, vol. B50, no. 1, pp. 52–59, 1998. idation on ZnO,” in Soc. Petroleum Eng. AIME Mtg., Cannes, France, [55] J. Tamaki, Z. Zhang, K. Fujimari, M. Akiyama, T. Harada, N. Miura, Sept. 22–26, 1980, pp. 436–439. and N. Yamazoe, “Grain size effects in tungsten oxide-based sensor for [30] M. Fleischer and H. Meixner, “Fast gas sensors based on metal oxides nitrogen oxides,” J. oElectrochem. Soc., vol. 141, no. 8, pp. 2207–2222, stable at high temperatures,” Sens. Actuators B, vol. 24–25, pp. 591–595, Aug. 1994. 1995. [56] A. P. Chatterjee, P. Mitra, and A. K. Mukhopadhyay, “Chemically de- [31] Dingquan, Xiao, P. Yu, R. Du, J. Zhu, S. Peng, P. Li, and Y. Zhuang, posited zinc oxide thin film gas sensor,” J. Mater. Sci., vol. 34, pp. “Investigation on new types of semiconducting ceramic gas sensors,” J. 4225–4331, 1999. Mater. Synthesis Process., vol. 6, no. 6, pp. 429–432, 1998. [57] J. H. Yu and G. M. Choi, “Electrical and CO gas sensing properties of [32] T. Schwebel, M. Fleischer, H. Mexiner, and C. D. Kohl, “CO-sensor for domestic use based on high temperature stable Ga O thin films,” Sens. ZnO–SnO composites,” Sens. Actuators B, vol. 52, pp. 251–256, 1998. Actuators B, vol. B49, pp. 46–51, 1997. [58] , “Electrical and CO gas-sensing properties of ZnO/SnO hetero- [33] Ratko, O. Babushkin, A. Bran, and S. Baran, “Sorption and gas sensitive contact,” Sens. Actuators B, vol. 61, pp. 59–67, 1999. properties of In O based ceramics doped with Ga O ,” Journal of the [59] Y. Yamada, Y. Seno, Y. Masuoka, and K. Yamashita, “Nitrogen oxides European Ceramic Society, vol. 18, no. 14, pp. 2227–2232, Dec. 1998. sensing characteristics of Zn SnO thin film,” Sens. Actuators B, vol. [34] J. Zhang and K. Colbow, “Surface silver clusters as oxidation catalysts B49, no. 3, pp. 248–252, 1998. on semiconductor gas sensors,” Sens. Actuators B, vol. 40, pp. 47–52, [60] D. S. Vlachos and A. C. Xenoulis, “Gas detection sensitivity and cluster 1997. size,” Nanostructured Mater., vol. 10, no. 8, pp. 1355–1361, 1998. [35] C. A. Papodopoulos and J. N. Avaritsiotis, “A model for the gas sensing [61] R. E. Cavicchi, J. S. Suehle, K. G. Kreider, M. Gaitan, and P. Chaparala, properties of tin oxide thin films with surface catalyst,” Sens. Actuators “Optimized temperature pulse sequences for the enhancement of chem- B, vol. 28, pp. 201–210, 1995. ically-specific response patterns from micro-hotplate gas sensors,” in [36] S. R. Davis, A. Wilson, and J. D. Wright, “Flammable gas sensors based Intl. Conf. Solid-State Sens. Actuators, Stockholm, Sweden, June 25–29, on sol-gel materials,” IEE Proc: Circuits, Devices, Syst., vol. 145, no. 5, 1995, pp. 823–826. pp. 379–382, 1998. [62] R. E. Cavicchi, S. Semancik, R. M. Walton, B. Panchapakesan, D. L. [37] N. Yamazoe and T. Seiyama, “Recent advances in gas sensors in DeVoe, M. Aquino-Class, J. D. Allen, and J. S. Suehle, “Microhotplate Japan,” in ACS Symposium, ACM Middle Atlantic Regional Meeting, gas sensor arrays,” in Proc of SPIE—Chem. Microsens. Applicat. II, vol. Washington, D.C., 1986, pp. 39–55. 3857, Boston, MA, Sept. 19–20, 1999, pp. 38–49. 272 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001

[63] A. Pike and J. W. Gardner, “Thermal modeling and characterization [91] X. Li and J. D. Harrison, “Measurement of concentration profiles inside of micropower chemoresistive silicon sensors,” Sens. Actuators B, vol. a nitrite ion selective electrode membrane,” Anal. Chem., vol. 63, no. 19, B45, no. 1, pp. 19–26, Nov. 1997. pp. 2168–2174, 1991. [64] R. H. M. van de Leur and A. van der Waal, “Gas and vapor detection [92] A. D. C. Chan and J. D. Harrison, “NMR study of the state of water in using polypyrrole,” Synthetic Metals, vol. 102, pp. 1330–1331, 1999. ion-selective electrode membranes,” Anal. Chem., vol. 65, p. 32, 1993. [65] P. D. Harris, W. M. Arnold, M. K. Andres, and A. C. Partridge, “Resis- [93] Van Der Berg, P. Bergveld, D. N. Reinhoudt, and E. J. R. Sudholter, tance characteristics of conducting polymer films used in gas sensors,” “Sensitivity control of ISFETs by chemical surface modification,” Sens. Sens. Actuators B, vol. 42, pp. 177–184, 1997. Actuators, vol. 8, no. 2, pp. 129–148, 1985. [66] J. E. G. de Souza, B. B. Neto, F. L. dos Santos, C. P. de Melo, M. S. [94] J. Janata and M. Josowicz, “Chemical modulation for work function as Santos, and T. B. Ludermir, “Polypyrrole based aroma sensor,” Synthetic a transduction mechanism for chemical sensors,” Accounts Chem. Res., Metals, vol. 102, pp. 1296–1299, 1999. vol. 31, no. 5, pp. 241–248, 1998. [67] P. N. Bartlett, P. B. M. Archer, and S. K. Ling-Chung, “Conducting [95] T.-H. Zhang, D. Petelenz, and J. Janata, “Temperature-controlled Kelvin polymer gas sensors—Part 1: Fabrication and characterization,” Sens. microprobe,” Sens. Actuators B, vol. 12, no. 3, pp. 175–180, 1993. Actuators, vol. 19, no. 2, pp. 125–140, 1989. [96] M. Liess, D. Chinn, D. Petelenz, and J. Janata, “Properties of insulated [68] P. N. Bartlett and S. K. Ling-Chung, “Conducting polymer gas sen- gate field-effect transistors with a polyaniline gate electrode,” Thin Solid sors—Part II: Response of polypyrrole to methanol vapor,” Sens. Ac- Films, vol. 286, no. 1–2, pp. 252–255, 1996. tuators, vol. 19, no. 2, pp. 141–150, 1989. [97] K. Domansky, L. Jing, and J. Janata, “Selective doping of chemically [69] J. W. Gardner, M. Vidic, P. Ingleby, A. C. Pike, J. E. Brignell, P. Scivier, sensitive layers on a multisensing chip,” J. Electrochem. Soc., vol. 144, P. N. Bartlett, A. J. Duke, and J. M. Elliott, “Response of a poly(pyrrole) no. 4, pp. 75–78, 1997. resistive micro-bridge to ethanol vapor,” Sens. Actuators B, vol. B48, no. [98] N. Jaffrezic-Renault, A. Senillou, C. Martelet, K. Wan, and J. M. 1–3, pp. 289–295, 1998. Chovelon, “ISFET microsensors for the detection of pollutants in liquid [70] F. Musio and M. C. Ferrara, “Low frequency a.c. response of polypyrrole media,” Sens. Actuators B, vol. 59, pp. 154–164, 1999. gas sensors,” Sens. Actuators B, vol. 41, pp. 97–103, 1997. [99] Senillou, N. Jaffrezic-Renault, C. Martelet, and F. Griffe, “A miniatur- [71] P. Ingelby, J. W. Gardner, and P. N. Bartlett, “Effect of micro-electrode ized ammonium sensor based on the integration of both ammonium and geometry on response of thin-film polypyrrole and polyaniline chemire- reference FETS in a single chip,” Mater. Sci. Eng., vol. C6, pp. 59–63, sistive sensors,” Sens. Actuators B, vol. 57, pp. 17–27, 1999. 1998. [72] M. Brie, R. Turcu, C. Neamtu, and S. Pruneanu, “The effect of initial [100] M. B. Ali, R. Kalfat, H. Sfihi, J. M. Chovelon, H. B. Ouada, and N. Jaf- conductivity and doping anions on gas sensitivity of conducting polypyr- frezic-Renault, “Sensitive cylcodextrin-polysiloxane gel membrane on role films to NH ,” Sens. Actuators B, vol. 37, pp. 119–122, 1996. EIS structure and ISFET for heavy metal ion detection,” Sens. Actua- [73] V. L. Conway, K. P. Hassen, L. Zhang, W. R. Seitz, and T. S. Gross, “The tors B, vol. 62, pp. 233–237, 2000. influence of composition on the properties of pH-swellable polymers for [101] G. Taillades, O. Valls, A. Bratov, C. Dominguez, A. Pradel, and M. chemical sensors,” Sens. Actuators B, vol. 45, pp. 1–9, 1997. Ribes, “ISE and ISFET microsensors based on a sensitive chalcogenide [74] H. Wang, C.-D. Feng, S.-L. Sun, C. U. Segre, and J. R. Stetter, “Com- glass for copper ion detection in solution,” Sens. Actuators B, vol. 59, parison of conductiometric humidly-sensing polymers,” Sens. Actuators pp. 123–127, 1999. B, vol. 40, pp. 211–216, 1997. [102] V. Volotovsky and N. Kim, “Cyanide determination by an ISFET-based [75] S. K. Dhawan, D. Kumar, M. K. Ram, S. Chandra, and D. C. Trivedi, peroxidase biosensor,” Biosens. Bioelectron., vol. 13, pp. 1029–1033, “Application of conducting polyaniline as sensor material for am- 1998. monia,” Sens. Actuators B, vol. 40, pp. 99–103, 1997. [103] D. N. Reinhoudt, “Molecular recognition applied to sensors,” Sens. Ac- [76] N. T. Kemp, G. U. Flanagan, A. B. Kaiser, H. J. Trodahl, B. Chapman, A. tuators B, vol. 24–25, pp. 33–35, 1995. C. Parridge, and R. G. Buckley, “Temperature-dependent conductivity of [104] Z. Brzozka, H. A. J. Holterman, G. W. N. Honig, U. H. Verkerk, conducting polymers exposed to gases,” Synthetic Metals, vol. 101, pp. H. H. van den Vlekkert, J. F. J. Engbersen, and D. N. Reinhoudt, 434–435, 1999. “Enhanced performance of potassium CHEMFETs by optimization of [77] N. J. Pinto, P. D. Shah, P. K. Kahol, and B. J. McCormick, “Dielec- a polysiloxane membrane,” Sens. Actuators B, vol. 18–19, pp. 38–41, tric constant and ac conductivity in polyaniline derivatives,” Solid State 1994. Commun., vol. 97, no. 12, pp. 1029–1031, 1996. [105] D. N. Reinhoudt, “Durable chemical sensors based on field-effect tran- [78] N. J. Pinto, P. D. Shah, B. J. McCormick, and P. K. Kahol, “Dependence sistors,” Sens. Actuators B, vol. 24–25, pp. 197–200, 1995. of the conducting state of polyaniline films on moisture,” Solid State [106] Z. Brzozka, M. Dawgul, D. Pijanowska, and W. Trobicz, “Durable NH Commun., vol. 97, no. 11, pp. 931–934, 1996. sensitive CHEMFET,” Sens. Actuators B, vol. 44, pp. 527–531, 1997. [79] M. E. H. Amrani and P. A. Payne, “Multi-frequency interrogation tech- [107] K. Scharnagl, M. Bogner, A. Fuchs, R. Winter, T. Doll, and I. Eisele, nique applied to conducting polymer gas and odour sensors,” IEE Proc. “Enhanced room temperature gas sensing with metal oxides by means Sci. Meas. Technol., vol. 146, no. 2, pp. 95–101, Mar. 1999. of the electro-adsorptive effect in hybrid suspended gate FET,” Sens. [80] N. S. Lewis, M. C. Lonergan, E. J. Severin, B. J. Doleman, and R. H. Actuators B, vol. 57, pp. 35–38, 1999. Grubbs, “Array-based vapor sensing using chemically sensitive carbon [108] J. Hendrikse, W. Olthuis, and P.Bergveld, “The EMOSFET as an oxygen black-polymer resistors,” Proc. SPIE, vol. 3079, pp. 660–670, Apr. sensor: Constant current potentiometry,” Sens. Actuators B, vol. 59, pp. 21–24, 1997. 35–41, 1999. [81] N. Lewis, “Chemically diverse, conducting polymer-based electronic [109] H.-K. Liao, J.-C. Chou, W.-Y. Chung, T.-P. Sun, and S.-K. Hsiung, nose,” in IEEE Aerospace Applicat. Conf., vol. 3, Snowmass Village, “Study of amorphous tin oxide thin films for ISFET applications,” CO, Feb. 1997. Sens. Actuators B, vol. 50, pp. 104–109, 1998. [82] A. Matzger, T. P. Vaid, and N. S. Lewis, “Vapor sensing with arrays of [110] F. Shitthelm, K.-S. Rover, and R. Ferreti, “Electronic characterization of carbon black-polymer composites,” Proc. SPIE, vol. 3710, pp. 315–320, thin diamond like carbon films for pH sensor applications,” Proc. SPIE, April 1999. vol. 3539, pp. 139–146, Nov. 1998. [83] J. Janata, “Chemical modulation for the electron work function,” Anal. [111] J.-C. Cho and J.-L. Chiang, “Ion sensitive field effect transistor with Chem., vol. 63, no. 22, pp. 2546–2550, 1991. amorphous tungsten trioxide gate for pH sensing,” Sens. Actuators B, [84] , “Potentiometric microsensors,” Chem. Rev., vol. 90, pp. 691–703, vol. 2, pp. 81–87, 2000. 1990. [112] J. Bausells, J. Carrabina, A. Errachid, and A. Merlos, “Ion-sensitive [85] J. Janata and M. Josowicz, “A fresh look at some old principles: The field-effect transistors fabricated in a commercial CMOS technology,” Kelvin probe and the nernst equation,” Anal.Chem., vol. 69, no. 9–12, Sens. Actuators B, vol. 57, pp. 56–62, 1999. pp. 293A–296A, 1997. [113] H.-K. Liao, E.-S. Yang, J.-C. Chou, W.-Y. Chung, T.-P. Sun, and S.-K. [86] J. Janata, Principles of Chemical Sensors. New York: Plenum, 1989. Hsiung, “Temperature and optical characteristics of tin oxide membrane [87] , “Chemical sensors,” Anal. Chem., vol. 62, no. 9–12, pp. 33R–39R, gate ISFET,” IEEE Trans. Electron. Devices, vol. 46, pp. 2278–2281, 1990. Dec. 1999. [88] J. Janata, M. Josowicz, and M. DeVaney, “Chemical sensors,” Anal. [114] L.-L. Chi, J.-C. Chou, W.-Y.Chung, T.-P. Sun, and S.-K. Hsiung, “Study Chem., vol. 66, pp. 207–228, 1994. on extended gate field effect transistor with tin oxide sensing mem- [89] Lundstrom, “Palladium gate hydrogen sensors,” in Chemical Sensor brane,” Mater. Chem. Phys., vol. 63, pp. 19–23, 2000. Technology, T. Seiyama, Ed. New York: Kluwer, 1989, vol. 2, pp. [115] M. V. Ochoa, A. A. Guitierrez, and P. R. Hernandez, “Microcontroller- 1–20. based system for multiple ISFET characterization using a single ref- [90] K. Domansky, D. L. Baldwin, J. W. Grate, T. B. Hall, J. Li, M. Josowicz, erence electrode,” in Proc. IEEE/EMBS, Chicago, IL, Oct. 30–Nov. 2, and J. Janata, “Development and calibration of field-effect transistor- 1997, pp. 2361–2364. based sensor array for measurement of hydrogen and ammonia gas mix- [116] S. Martinoia and G. Massobrio, “A behavioral macromodel of the ISFET tures in humid air,” Anal. Chem., vol. 70, pp. 473–481, 1998. in SPICE,” Sens. Actuators B, vol. 62, pp. 182–189, 2000. WILSON et al.: CHEMICAL SENSORS FOR PORTABLE, HANDHELD FIELD INSTRUMENTS 273

[117] H. Raether, Surface Plasmons on Smooth and Rough Surfaces and on [142] M. J. O’Brien, S. R. J. Brueck, V. H. Perez-Luna, L. M. Tender, and G. Gratings. Berlin, Germany: Springer-Verlag, 1988. P. Lopez, “SPR biosensors: Simultaneously removing thermal and bulk- [118] Liebsch, Electronic Excitations at Metal Surfaces. New York: composition effects,” Biosens. Bioelectron., vol. 14, no. 2, pp. 145–154, Plenum, 1997. 1999. [119] BIAcore, Inc. [Online]. Available: http://www.biacore.com. [143] J. Homola, H. B. Lu, and S. S. Yee, “Dual-channel surface plasmon res- [120] Affinity Sensors, Inc. [Online]. Available: http://www.affinity-sen- onance sensor with spectral discrimination of sensing channels using di- sors.com. electric overlayer,” Electron. Lett., vol. 35, no. 13, pp. 1105–1106, 1999. [121] A. G. Frutos and R. M. Corn, “SPR of ultrathin organic films,” Anal. [144] N. M. Lyndin, I. F. Salakhutdinov, V. A. Sychugov, B. A. Usievich, F. A. Chem., vol. 70, no. 13, pp. 449A–455A, 1998. Pudonin, and O. Parriaux, “Long-range surface plasmons in asymmetric [122] A. T. M. Lenferink, R. P. H. Kooyman, and J. Greve, “An improved op- layered metal-dielectric structures,” Sens. Actuators B, vol. 54, no. 1–2, tical method for surface plasmon resonance experiments,” Sens. Actua- pp. 37–42, 1999. tors B, vol. 3, pp. 261–265, 1991. [123] Texas Instruments, Inc. [Online]. Available: http://www.ti.com/spreeta. [124] J. Melendez, D. Bartholomew, R. Carr, M. Sciascia, P. Smith, A. Strong, and J. Elkind, “Integrated fault detection capability for Spreeta biosensor,” in Internal Standardization and Calibration Architectures for Chemical Sensors, R. E. Shaffer and R. A. Potyrailo, Eds: Proc. SPIE, 1999, vol. 3856, pp. 282–288. [125] B. Catimel, J. Weinstock, M. Nerrie, T. Domagala, and E. C. Nice, “Mi- Denise Michelle Wilson (M’89) was born in cropreparative ligand fishing with a cuvette-based optical mirror reso- Chicago, IL, in 1966. She received the B.S. degree nance biosensor,” J. Chromatography A, vol. 869, pp. 261–273, 2000. in mechanical engineering from Stanford University, [126] A. Strong, D. I. Stimpson, D. U. Bartholomew, T. F. Jenkins, and J. L. Stanford, CA, in 1988 and the M.S. and Ph.D. Elkind, “Detection of trinitrotoluene extracted from soil using a surface degrees in electrical engineering from the Georgia plasmon resonance based sensor platform,” in Proc. SPIE, 1999, vol. Institute of Technology, Atlanta, in 1995 and 1989, 3710, pp. 362–372. respectively. [127] I. Stemmler, A. Brecht, and G. Gauglitz, “Compact surface plasmon res- She is currently an Associate Professor in the onance-transducers with spectral readout for biosensing applications,” Department of Electrical Engineering, University Sens. Actuators B, vol. 54, no. 1–2, pp. 98–105, 1999. of Washington, Seattle, and worked previously at [128] R. Alonso, F. Villuendas, J. Tornos, and J. Pelayo, “New in-line optical- the University of Kentucky, Lexington, in a similar fiber sensor based on surface plasmon excitation,” Sens. Actuators A, position from 1996 to 1999. Her research interests focus on the development vol. 37–38, pp. 187–192, 1993. of signal processing architectures, array platforms, and other infrastructure [129] R. Slavik, J. Homola, and J. Ctyroky, “Single-mode optical fiber sur- for visual, auditory, and chemical sensing microsystems. She has also worked face plasmon resonance sensor,” Sens. Actuators B, vol. 54, no. 1–2, pp. for several years for Applied Materials, a semiconductor capital equipment 74–79, 1999. supplier, from 1990 to 1992. [130] W. J. Bender, R. E. Dessy, M. S. Miller, and R. O. Claus, “Feasibility of a chemical microsensor based of surface plasmon resonance on fiber optics modified by multi layer vapor deposition,” Anal. Chem., vol. 66, no. 7, pp. 963–970, 1994. [131] R. D. Harris, B. J. Luff, J. S. Wilkinson, J. Piehler, A. Brecht, G. Gauglitz, and R. A. Abuknesha, “Integrated optical surface plasmon resonance immunoprobe for simazine detection,” Biosens. Bioelectron., Sean Hoyt was born in Long Beach, CA, on Jan- vol. 14, no. 4, pp. 377–386, 1999. uary 21, 1977. He received the B.S. in electrical engi- [132] K. S. Johnston, S. S. Yee, and K. S. Booksh, “Calibration of surface neering from the University of Washington, Seattle, plasmon resonance refractometers using locally weighted parametric re- in 2000. He is currently pursuing the M.S. degree gression,” Anal. Chem., vol. 69, no. 10, pp. 1844–1851, 1997. in electrical engineering at the University of Wash- [133] K. S. Johnston, K. S. Booksh, T. M. Chinowsky, and S. S. Yee, “Perfor- ington in the Distributed Systems Laboratory. mance comparison between high and low resolution spectrophotometers His research includes sensors systems and radio used in a white light surface plasmon resonance sensor,” Sens. Actuators frequency identification. Other interests include B, vol. 54, no. 1–2, pp. 80–88, 1999. biking, drawing, and programming. [134] R. C. Jorgenson and S. S. Yee, “Control of the dynamic range and sen- sitivity of a surface plasmon resonance based fiber optic sensor,” Sens. Actuators A, vol. 43, pp. 44–48, 1994. [135] L. A. Obando and K. S. Booksh, “Tuning dynamic range and sensitivity of white-light, multimode, fiber-optic surface plasmon resonance sen- sors,” Anal. Chem., vol. 71, no. 22, pp. 5116–5122, 1999. [136] M. K. Boysworth, L. A. Obando, and K. S. Booksh, “Calibration systems for surface plasmon resonance spectroscopy,” Intern. Stand. Calibr. Ar- Jiri (Art) Janata received the Ph.D. degree in analytical chemistry from the chitect. Chem. Sens., vol. 3856, pp. 308–316, 1999. Charles University, Prague, Czechoslovakia, in 1965. [137] S. Lofas and B. Johnsson, “A novel hydrogel matrix on gold surfaces He is Georgia Research Alliance Eminent Scholar in the School of Chemistry in surface plasmon resonance sensors for fast and efficient covalent im- and Biochemistry, Georgia Institute of Technology, Atlanta. After postdoctoral mobilization of ligands,” J. Chem. Soc.—Chem. Commun., vol. 21, pp. studies at the University of Michigan, Ann Arbor, and six years in the Corpo- 1526–1528, 1990. rate Laboratory of ICI, Ltd., U.K., he joined the faculty of the University of [138] E. Brynda, J. Homola, M. Houska, P. Pfeifer, and J. Skvor, “Antibody Utah, Salt Lake City, in 1976. He was a Visiting Professor at the Universität der networks for surface plasmon resonance sensors,” Sens. Actuators B, Bundeswehr Munich, Munich, Germany, EPLF Lausanne, and at the Tokyo In- vol. 54, no. 1–2, pp. 132–136, 1999. stitute of Technology, Tokyo, Japan. Between 1992 and 1997, he was Associate [139] L. S. Jung, C. T. Cambell, T. Chinowsky, M. N. Mar, and S. S. Director of EnvironmentalMolecular Sciences Laboratory at the Pacific North- Yee, “Quantitative interpretation of the response of surface plasmon west National Laboratory. His interests include chemical sensors, electrochem- resonance sensors to adsorbed films,” Langmuir, vol. 14, no. 19, pp. istry, interphasial physical chemistry, and microfabrication. He has published 5636–5648, 1998. over 180 papers and book chapters and a textbook Principles of Chemical Sen- [140] Brecht and G. Gauglitz, “Label free optical immunoprobes for pesticide sors (1989), chaired three Gordon Research Conferences, and holds 17 U.S. and detection,” Anal. Chim. Acta, vol. 347, no. 1–2, pp. 219–233, 1997. International Patents. [141] R. Casalini, J. N. Wilde, J. Nagel, U. Oertel, and M. C. Petty, “Organic Dr. Janata received the Alexander von Humboldt Prize in 1988, Heyrovsky vapor sensing using thin films of a co-ordination polymer: Comparison Medal in 1990, and Outstanding Research Achievement Award from the Elec- of electrical and optical techniques,” Sens. Actuators B, vol. 57, no. 1–3, trochemical Society in 1994. In 2001, he was elected Honorary Member of the pp. 28–34, 1999. Czech Learned Society. 274 IEEE SENSORS JOURNAL, VOL. 1, NO. 4, DECEMBER 2001

Karl Booksh received the B.S. in chemistry from Louis Obando was born in Belize in 1974. He re- the University of Alaska, Fairbanks, in 1990 and the ceived the B.S. degree in chemistry from the Univer- Ph.D. from the University of Washington, Seattle, in sity of Illinois, Springfield, in 1997 and started doc- 1994. torate work with Dr. Karl Booksh at Arizona State Following two years as a National Science University (ASU), Tempe, later that summer. Foundation Postdoctoral Fellow at the University of His doctoral work focuses on developing reliable South Carolina, Columbia, he joined the Department and robust fiber-optic SPR sensors for integration of Chemistry and Biochemistry at Arizona State with instrumentation for analyzes of refractive index, University, Tempe. His research interests include the environmental pollutants, and antibody–antigen integrated design and calibration of optical sensors interactions. In addition to being a research assistant, for environmental and industrial process monitoring. Louis has also worked four semesters teaching chemistry to undergraduate engineers. He is the recipient of the 2001 AOAC International Harvey W. Wiley Award; the 2001 Outstanding Graduate Research Award from the ASU Department of Chemistry, and the 2000 ASU KOKO Award for outstanding analytical chem- istry graduate student.