Research article A review of gas sensors employed in electronic nose applications

K. Arshak E. Moore G.M. Lyons J. Harris and S. Clifford

The authors K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford are all based at the College of Informatics and Electronics, University of Limerick, Limerick, Ireland. Keywords Sensors, Gases Abstract This paper reviews the range of sensors used in electronic nose (e-nose) systems to date. It outlines the operating principles and fabrication methods of each sensor type as well as the applications in which the different sensors have been utilised. It also outlines the advantages and disadvantages of each sensor for application in a cost-effective low-power handheld e-nose system.

Introduction
The human nose has been used as an analytical tool in many industries to measure the quality of food, drinks, perfumes and also cosmetic and chemical products. It is commonly used for assessing quality through odour and this is carried out using sensory panels where a group of people lls out questionnaires on the smells associated with the substance being analysed. These sensory panels are extremely subjective as human smell assessment is affected by many factors. Individual variations occur and may be affected by physical and mental health as well as fatigue (Pearce et al., 2003). For this reason, gas chromatography and mass spectrometry have been employed to aid human panels to assess the quality of products through odour evaluation and identication and also to obtain more consistent results. However, these assistive techniques are not portable, they tend to be expensive and their performance is relatively slow (Nagle et al., 1998). The solution to the shortcomings of sensory panels and the associated analytical techniques is the electronic nose (e-nose). E-nose systems utilize an array of sensors to give a ngerprint response to a given odour, and pattern recognition software then performs odour identication and discrimination. The e-nose is a cost-effective solution to the problems associated with sensory panels and with chromatographic and mass-spectrometric techniques and can accommodate real time performance in the eld when implemented in portable form.

Electronic access The Emerald Research Register for this journal is available at www.emeraldinsight.com/researchregister The current issue and full text archive of this journal is available at www.emeraldinsight.com/0260-2288.htm

Principle of operation of e-nose systems

The e-nose attempts to emulate the mammalian nose by using an array of sensors that can simulate mammalian olfactory responses to aromas. The odour molecules are drawn into the e-nose using sampling techniques such as headspace sampling, diffusion methods, bubblers or pre-concentrators
This work was conducted as part of a collaborative project between AMT Ireland, University of Limerick and University College Cork, and is funded by Enterprise Ireland under project ref. no. ATRP/ 2002/427 (Intelli-SceNT).

Sensor Review Volume 24 Number 2 2004 pp. 181198 q Emerald Group Publishing Limited ISSN 0260-2288 DOI 10.1108/02602280410525977

181

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

(Pearce et al., 2003). The odour sample is drawn across the sensor array and induces a reversible physical and/or chemical change in the sensing material, which causes an associated change in electrical properties, such as nyi, 2000). Each cell in conductivity (Harsa the array can behave like a receptor by responding to different odours to varying degrees (Shurmer and Gardner, 1992). These changes are transduced into electrical signals, which are preprocessed and conditioned before identication by a pattern recognition system as shown in Figure 1. The e-nose system is designed so that the overall response pattern from the array is unique for a given odour in a family of odours to be considered by the system.

referred to as the response time while the time it takes the sensor to return to its baseline resistance is called the recovery time. The next stage in analysing the odour is sensor response manipulation with respect to the baseline. This process compensates for noise, drift and also for inherently large or small signals (Pearce et al., 2003). The three most commonly used methods as dened by Pearce et al. (2003) are as follows. (1) Differential: the baseline xs(0) is subtracted from the sensor response xs(t) to remove any noise or drift dA present. The baseline manipulated response ys(t) is determined by: ys t xs t dA 2 xs 0 dA

E-nose sensor response to odorants

The response of e-nose sensors to odorants is generally regarded as a rst order time response. The rst stage in odour analysis is to ush a reference gas through the sensor to obtain a baseline. The sensor is exposed to the odorant, which causes changes in its output signal until the sensor reaches steady-state. The odorant is nally ushed out of the sensor using the reference gas and the sensor returns back to its baseline as shown in Figure 2. The time during which the sensor is exposed to the odorant is
Figure 1 Comparison of the mammalian olfactory system and the e-nose system

xs t 2 xs 0

(2) Relative: the sensor response is divided by the baseline. This process eliminates multiplicative drift dM and a dimensionless response ys(t) is obtained. y s t xs t 1 dM xs t xs 01 dM xs 0 2

(3) Fractional: the baseline is subtracted from the response xs(t) and then divided by the baseline xs(0) from the sensor response which provides a dimensionless, normalised

182

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Figure 2 E-nose sensor response to an odorant

with thin or thick lms of the sensor material. The sensor device detects the physical and/or chemical changes incurred by these processes and these changes are measured as an electrical signal. The most common types of changes utilised in e-nose sensor systems are shown in Table I along with the classes of sensor devices used to detect these changes. Some aspects of these classes of sensors, along with several examples from each class, are explored in the following sections of this review. Conductivity sensors Conducting polymer composites, intrinsically conducting polymers and metal oxides are three of the most commonly utilised classes of sensing materials in conductivity sensors. These materials work on the principle that a change in some property of the material resulting from interaction with a gas/odour leads to a change in resistance in the sensor. The mechanisms that lead to these resistance changes are different for each material type; however, the structure and layout of conductivity sensors prepared using these materials are essentially the same. A schematic of a typical conductivity sensor design is shown in Figure 3. The sensing material is deposited over interdigitated or two parallel electrodes, which form the electrical connections through which the relative resistance change is measured. The heater is required when metal oxides are used as the sensing material because very high temperatures are required for effective operation of metal oxide sensors.

response ys(t) that can compensate for inherently large or small signals. ys t xs t 2 xs 0 xs 0 3

The choice of baseline manipulation depends on the sensor type being used, the sensor applications and also the researchers preference. However, certain manipulation techniques have been shown to be more suitable to certain sensor types and also variations in the manipulation techniques can occur in the literature, which will be discussed later in this paper. Sensitivity is the measure of the change in output of a sensor for a change in the input. This is the standard denition given for the sensitivity of a sensor in several texts on sensor related topics (Fraden, 1996; Gopel et al., 1989; Johnson, 1997). In the case of e-nose sensors, the sensitivity of the sensor (S ) to the odorant is the change in the sensor output parameter ( y), i.e. resistance for a change in the concentration of the odorant (x) as shown in equation (4). Dy S 4 Dx However, in the literature several authors use different values to measure sensitivity, usually calculated from baseline-manipulated data.

Conducting polymer composite sensors

Conducting polymer composites consist of conducting particles such as polypyrrole and carbon black interspersed in an insulating polymer matrix (Albert and Lewis, 2000).
Table I Physical changes in the sensor active lm and the sensor devices used to transduce them into electrical signals Physical changes Sensor devices Conductivity sensors Piezoelectric sensors Optical sensors MOSFET sensors

Sensors employed in e-nose systems

Gas molecules interact with solid-state sensors by absorption, adsorption or chemical reactions

Conductivity Mass Optical Work function

183

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Figure 3 Typical structure of a conductivity sensor

On exposure to gases, these materials change resistance and this change is due to percolation effects or more complex mechanisms in the case of polypyrrole lled composites. When the polymer composite sensor is exposed to a vapour, some of the vapour permeates into the polymer and causes the polymer lm to expand. The vapour-induced expansion of the polymer composite causes an increase in the electrical resistance of the polymer composite because the polymer expansion reduces the number of conducting pathways for charge carriers (Munoz et al., 1999). This increase in resistance is consistent with percolation theory. Polypyrrole-based composites have a more complex transduction mechanism because the odour molecules can cause expansion of both the insulating polymer and the polypyrrole particles. Changes in the intrinsic conductivity of the polypyrrole particles can also occur if the odour interacts chemically with the conducting polymer backbone. Therefore, resistance changes in the polypyrrole-based composites are more difcult to predict (Albert and Lewis, 2000). Carbon black based composites were prepared by suspending the carbon black in a solution of the insulating polymer in a suitable solvent. The overall composition of this solution is usually 80 per cent insulating polymer20 per cent carbon black by weight. Different techniques have been used to apply the active material onto the substrate, which are shown in Table II. The transducer device is usually a at substrate with either two parallel electrodes or

interdigitated electrodes deposited onto the substrate surface as shown in Figure 3. However, both bulk and surface micromachining have also been used to produce wells on silicon substrates using patterned silicon nitride/oxide as insulation and gold electrodes as the metal contacts. The polymer is dropped into the well using a syringe as shown in Figure 4 (Zee and Judy, 2001). Examples of the types of substrates and electrodes used are given in Table III along with the electrode dimensions given in the literature. Polypyrrole-based composites are usually prepared by chemically polymerizing pyrrole using phosphomolybdic acid in a solution containing the insulating polymer. A thin lm coating (40-100 nm thick) is then applied across interdigitated electrodes (typical electrode gaps were 15 mm) using dip coating (Freund and Lewis, 1995). The fractional baseline manipulation response is the most common method used to record the sensor response from conducting gas sensors (Pearce et al., 2003). This fractional baseline manipulation is referred to as the maximum relative differential resistance change, DR/Rb. DR is the maximum change in the resistance of the sensor during exposure to the odorant and Rb is the baseline resistance before exposure. Response times may vary from seconds to minutes, as shown in Table IV, and, in some cases, milliseconds (Munoz et al., 1999). However, typically a set of exposure cycles is quoted for each analyte, i.e. initial reference gas, analyte and nal reference gas exposure times. The response time depends on the rate of diffusion of the permeant into the polymer. The diffusion rate mainly depends on the nature of the polymer, permeant and the crosslinking, the concentration of the permeant and thickness of the lm, and on the effects of llers, plasticisers and temperature (George and Thomas, 2001). General response times for conducting polymer composites are given in Table IV. The transport properties depend on the fractional free volume (FFV) in the polymer and on the mobility of the segments of the polymer chains. The segmental mobility of the polymer chains depends on the extent of unsaturation. Therefore, segmental mobility is signicantly

184

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Table II Application techniques used to deposit carbon black based composites Reference Matthews et al. (2002) Severin et al. (1998) Severin et al. (2000) Polymer Poly(alkylacrylate) Poly(co-vinyl-acetate) Poly(vinyl butyral) Coating thickness Thin lm Thin lm Thin lm 0.86 mm 1 mm N/A Coating application technique Spray coating Spin coating Dip coating

Figure 4 Bulk micromachined well as used by Zee and Judy (2001)

reduced with increased numbers of double and triple bonds on the carbon-carbon backbone. The presence of crystalline domains can have two different effects on the permeability of a polymer. Crystallites are generally impermeable to vapours at room temperature and will reduce the diffusion coefcient. However, crystallites act as large crosslinking regions in terms of the chains entering and leaving the areas around the crystallite in which diffusion occurs (Comyn, 1985). Increased defect formation and a reduction in interlamellar links predominate and overshadow the effect of geometric impedance to such an extent that diffusivity increases despite increasing crystallinity (Vieth, 1991). For example, in PE-grafted carbon black chemiresistors, heat treatment of the composite improved the crystallinity of the matrix PE and resulted in a ve-fold increase in the response of the sensor to cyclohexane vapour compared to the untreated sensor (Chen et al., 2002). The size and shape of the penetrating molecule affects the rate of uptake of the vapour

by the polymer where increased size of the molecule decreases the diffusion coefcient. Flattened or elongated molecules diffuse faster than spherical-shaped molecules (George and Thomas, 2001). The partial pressure, i.e. the concentration, of the penetrant gas at the gas polymer interface affects the response of the sensor. The response is inversely related to the vapour pressure of the analyte at the surface. Low vapour pressure compounds can be detected in the low ppb range while high vapour pressure compounds need to be in the high ppm range to be detected. This is due to the polymer/gas partition coefcient where low vapour pressure gas molecules have a higher tendency to inhabit the polymer and thus can be detected at much lower concentrations (Munoz et al., 1999). The equilibrium partition coefcient is essentially the solubility coefcient of the vapour in the polymer (Vieth, 1991). Sensors with constant thickness but with different surface areas have the same response when exposed to analytes with moderate partition coefcients. However, analytes with higher partition coefcients have higher afnity to sensors with smaller area. In this case, reducing the sensor area increases the sensitivity of the sensor towards particular analytes. Therefore the relationship between the partition co-efcient and the sensor geometry is an important factor in optimising polymer composite sensor response (Briglin et al., 2002).

Table III Electrodes and substrates used in conducting polymer composite sensors Electrode material Gold Gold Gold, Chromium Gold Gold, Chromium Electrode dimensions (mm) N/A 10 10 20 10 0.1 in width 1.8 in length Electrode gap (mm) 5 5 5 0.5 0.4 Electrode thickness (nm) 50-100 50 30, 20 N/A 50, 15-30

Substrate Glass microscope slides Glass microscope slides Glass microscope slides Micromachined silicon wafer Glass microscope slide

Reference Lonergan et al. (1996) Doleman et al. (1998a) Doleman et al. (1998a, b) Zee and Judy (2001) Briglin et al. (2002)

185

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Table IV General response times for conducting polymer composites Response time (s) < 2-4 60 180-240 20-200 Reference Lonergan et al. (1996) Partridge (2000) Corcoran (1993b) Doleman et al. (1998)

which leads to sensor drift, and also these materials are unsuitable for detecting certain gases, for example carbon-polymer composites are not sensitive to trimethylamine (TMA) for sh odour applications.

Intrinsically conducting polymers

Intrinsically conducting polymers (ICP) have linear backbones composed of unsaturated monomers, i.e. alternating double and single bonds along the backbone, that can be doped as semiconductors or conductors (Heeger, 2001). Yasufuku (2001) describes them as p electron conjugated polymers where the p symbol relates to the unsaturated structure of the monomer containing an unpaired carbon electron. These conducting polymers can be n-doped or p-doped depending on the doping materials used. Conducting polymers such as polypyrrole, polythiophene and polyaniline, as shown in Figure 5, are typically used for e-nose sensing. The doping of these materials generates charge carriers and also alters their band structure, which both induce increased mobility of holes or electrons in the polymer depending on the type of doping used (Albert and Lewis, 2000; Dickinson et al., 1998). The principle of operation for ICP e-nose sensors is that the odorant is absorbed into the polymer and alters the conductivity of the polymer (Albert and Lewis, 2000; Dickinson et al., 1998). Three types of conductivity are affected in intrinsically conducting polymers. (1) The intrachain conductivity in which the conductivity along the backbone is altered; (2) The intermolecular conductivity which is due to electron hopping to different chains because of analyte sorption (Charlesworth et al., 1997); and (3) The ionic conductivity which is affected by proton tunneling induced by hydrogen bond interaction at the backbone and also by ion migration through the polymer (Albert and Lewis, 2000). The physical structure of the polymer also has a major inuence on the conductivity (Yasufuku, 2001). Albert and Lewis (2000) described how the conductivity of doped polyaniline was greatly increased on interaction with ethanol, caused by the hydrogen bonds

Generally, composite polymers have operating resistances of approximately 1 to 1,000 kV with sensitivities of up to 100 per cent DR/Rb per mg/ l gas depending on the ller type, particle loading and vapour transport properties (Partridge et al., 2000). Sensitivity to hydrophobic analytes for conducting polymer composites were found to be one to two orders of magnitude higher than those recorded for intrinsically conducting polymers (Partridge et al., 2000). The polymer composites show linear responses to concentration for various analytes and also good repeatability after several exposures (Severin, 2000; Zee and Judy, 2001). From the above discussion, it can be seen that for high sensitivity, fast response and short recovery times it is essential that the sensor geometry and all the associated properties of the polymer sensing material be highly optimised. Conducting polymer composites offer many advantages over other materials when utilised as gas sensors. High discrimination in array sensors can be easily achieved using these materials due to the wide range of polymeric materials available on the market. This is due to the fact that different polymers give different levels of response to a given odour. Conducting polymer composites are also relatively inexpensive and easy to prepare. Sensors prepared from these materials can operate in conditions of high relative humidity and also show highly linear responses for a wide range of gases (Munoz et al., 1999). No heater is required as sensors prepared from these materials operate at room temperature. This is an important advantage with portable battery powered e-nose systems, as a heater would signicantly increase the power consumption of the system. The signal conditioning circuitry required for these sensors is relatively simple as only a resistance change is being measured. The main drawbacks of using conducting polymer composites as e-nose sensors are aging,

186

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Figure 5 Chemical structures of (a) polyaniline; (b) polypyrrole; and (c) polythiophene

tightening or restructuring the polymer into a more crystalline shape. Intrinsically conducting polymers are generally deposited onto a substrate, with interdigitated electrodes, using electrochemical techniques or by chemical polymerisation, (Freund and Lewis, 1995; Yasufuku, 2001). Sensors with a more simple structure can be prepared where the polymer is deposited between two conducting electrodes (Guadarrama et al., 2000). Electrochemical polymerisation is carried out using a three-electrode electrochemical cell with the electrodes on the substrate used as the working electrodes (Gardner and Bartlett, 1995). A potential is applied across the electrode that initiates the polymerisation of the polymer onto the substrate. The polymer is initially deposited onto the electrodes and then grows between them thus producing a complete thin lm across the electrode arrangement as the polymerisation reactions progress. Electrode thickness can range from 1 to 10 mm and the electrode gap is typically 10-50 mm (Albert and Lewis, 2000; Guadarrama et al., 2000). The total charge applied during the polymerisation process determines the thickness of the resultant lm, while the nal applied voltage determines the doping concentration. Partridge et al. (2000) described the sensor characteristics of intrinsically conducting polymers produced by pulsing the potential during polymerisation and observed a 1-50 per cent relative differential resistance change (DR/Rb) for saturated gas. The response of ICP sensors depends on the sorption of the vapour into the sensing material, causing swelling, and this affects the electron density on the polymeric chains

(Albert and Lewis, 2000). The sorption properties of these materials essentially depend on the diffusion rate of the permeant into the polymer matrix as explained earlier for composite conducting polymers. Generally, the reported response times for ICPs vary considerably from seconds to minutes, e.g. 30 s (Sotzing et al., 2000), 60 s (Pearce et al., 2003) and 180-240 s (Corcoran, 1993a). Polythiophene and poly(dodecylthiophene) sensors have sensitivities from approximately 0.2 up to 1.8 (DR/Rb for 300 ppm gas for 10 min) for gases such as CH4, CHCl3 and NH4 (Sakurai et al., 2002). Polypyrrole gas sensors were reported to have sensitivities (mV/ppm) ranging from 0.26 to 5.01 depending on the counter ions used in the polymers lms (Fang et al., 2002). These response times and sensitivities are reported for an extremely vast variety of materials prepared by various techniques, sensor geometries and odour molecules, rendering direct comparisons between sensors prepared by different researchers difcult if not impractical. Intrinsically conducting polymers have a number of advantages when used in e-nose systems. Increased discrimination when developing sensor arrays can easily be achieved with these materials as a wide range of intrinsically conducting polymers are available on the market (Albert and Lewis, 2000). ICP sensors operate at room temperature (Shurmer and Gardner, 1992), thereby simplifying the required system electronics. Conducting polymers show a good response to a wide range of analytes and have fast response and recovery times especially for polar compounds. Problems related to intrinsically conducting polymer sensors include poorly understood signal transduction mechanisms, difculties in

187

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

resolving some types of analytes, high sensitivity to humidity (Albert and Lewis, 2000) and the sensor response can drift with time. The fabrication techniques for these sensors can also be difcult and time-consuming (Nagle et al., 1998), and large variations in the properties of the sensors can occur from batch to batch (Nagle et al., 1998). Finally, the lifetime of these sensors can be quite short, typically 9-18 months, due to oxidation of the polymer (Schaller et al., 1998).

Metal oxide sensors

The principle of operation of metal oxide sensors is based on the change in conductance of the oxide on interaction with a gas and the change is usually proportional to the concentration of the gas. There are two types of metal oxide sensors; n-type (zinc oxide, tin dioxide, titanium dioxide or iron (III) oxide) which respond to reducing gases and p-type (nickel oxide, cobalt oxide) which respond to oxidising gases (Pearce et al., 2003). The n-type sensor operates as follows: oxygen in the air reacts with the surface of the sensor and traps any free electrons on the surface or at the grain boundaries of the oxide grains. This produces large resistance in these areas due to the lack of carriers and the resulting potential barriers produced between the grains inhibit the carrier mobility. However, if the sensor is introduced to a reducing gas like H2, CH4, CO, C2H5 or H2S the resistance drops because the gas reacts with the oxygen and releases an electron. This lowers the potential barrier and allows the electrons to ow, thereby increasing the conductivity. P-type sensors respond to oxidising gases like O2, NO2, and Cl2 as these gases remove electrons and produce holes, i.e. producing charge carriers. Equations (1) and (2) describe the reactions occurring at the surface: 1 O2 e2 ! O2 s 2 R g O2 s ! ROg e 5 6

Where e is an electron from the oxide. R(g) is the reducing gas and g and s are the surface and gas, respectively (Albert and Lewis, 2000; Pearce et al., 2003).

Thick and thin lm fabrication methods have been used to produce metal oxide gas sensors. The metal oxide lms are deposited using screenprinting (Schmid et al., 2003), spincoating (Hamakawa et al., 2002), RF sputtering (Tao and Tsai, 2002) or chemical vapour deposition (Dai, 1998) onto a at or tube type substrate made of alumina, glass, silicon or some other ceramic. Gold, platinum, silver or aluminium electrodes are deposited onto the substrate using the same methods. There are various electrode designs but the interdigitated structure appears to be the most common approach. A heating element is printed onto the back of the substrate to provide the high temperatures required for metal oxides to operate as gas sensors, typically 200-5008C. Film thickness ranges from 10 to 300 mm for thick lm and 6-1,000 nm for thin lm, (Schaller et al., 1998). Catalytic metals are sometimes applied on top of the oxides to improve sensitivity to certain gases. The same preparative methods are used to apply the catalytic metal (Albert and Lewis, 2000; Penza et al., 2001b, c). The sensitivity of oxide gas sensors DR=Rb =cgas; is calculated with DR R 2 Rb for oxidising gases and as DR Rb 2 R for reducing gases where Rb is the baseline resistance and R is the resistance on exposure to the odour and c(gas) is the concentration of the gas (Steffes et al., 2001). The sensitivity of the metal oxide sensor depends on the lm thickness and the temperature at which the sensor is operated, with thinner lms being more sensitive to gases (Schaller et al., 1998). The sensitivity can be improved by adding a catalytic metal to the oxide; however excessive doping can reduce sensitivity. The effect of doping SnO2 sensors with Cu, for example, demonstrated signicant increases in sensitivity (Galdikas et al., 1995). The grain size of the oxide also affects the sensitivity and selectivity to particular gases as the grain boundaries act as scattering centres for the electrons (Schaller et al., 1998). The ratio of the grain size (D) to the electron depletion layer thickness (L), ranging below D=2L 1, i.e. the depletion region extends over the entire grain, governs the sensitivity of the sensor (Behr and Fliegel, 1995). Thin lm metal oxide sensors saturate quickly, which can reduce the sensitivity range

188

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

in which the sensor can operate. The sensitivity of metal oxide sensors is very dependent on the operating temperature, for example, the sensitivity, DR=Rb (per 10 ppm NO2), for In2O3 gas sensors range from approximately 1.2 to 1.6 with the operating temperature varying from 350 to 4508C (Steffes et al., 2001). For SnO2 gas sensors doped with Al2O3, the sensitivity per 500 ppm CH4 ranges from 0.2 to 0.7 between 250 and 4008C (Saha et al., 2001). The response times for tin oxides tend to be very fast and can reach steady-state in less than 7 s, (Doleman et al., 1998a, b). The general response times for tin oxide sensors, with gas concentrations between 0 and 400 ppm and at temperatures between 250 and 5008C, are 5 to 35 s and the recovery times vary from 15 to 70 s (Albert and Lewis, 2000). Corcoran gave values of 20 s for thick lm metal oxide sensors and 12 s for thin lm metal oxide sensors (Corcoran, 1993b). The main advantages of metal oxide sensors are fast response and recovery times, which mainly depend on the temperature and the level of interaction between the sensor and gas (Penza et al., 2001c). Thin lm metal oxide sensors are small, and relatively inexpensive to fabricate, have lower power consumption than thick lm sensors and can be integrated directly into the measurement circuitry (Dai, 1998). However, they have many disadvantages due to their high operating temperatures, which results in increased power consumption over sensors fabricated from materials other than metal oxides. As a result, no handheld e-nose system has been fabricated utilising sensors prepared from metal oxides (Pearce et al., 2003). They also suffer from sulphur poisoning due to irreversible binding of compounds that contain sulphur to the sensor oxide (Dickinson et al., 1998) and ethanol can also blind the sensor from other volatile organic compound (VOC), gases (Schaller et al., 1998). Piezoelectric sensors used in e-nose arrays There are two types of piezoelectric sensors used in gas sensing, the surface acoustic wave (SAW) device and the quartz crystal microbalance (QCM). The SAW device produces a surface wave that travels along the surface of the sensor while the QCM produces a wave that travels through the bulk of the sensor.

Both types of devices work on the principle that a change in the mass of the piezoelectric sensor coating due to gas absorption results in a change in the resonant frequency on exposure to a vapour (Albert and Lewis, 2000).

Surface acoustic wave sensor

The SAW device is composed of a piezoelectric substrate with an input (transmitting) and output (receiving) interdigital transducer deposited on top of the substrate (Khlebarov et al., 1992). The sensitive membrane is placed between the transducers, Figure 6, and an ac signal is applied across the input transducer creating an acoustic twodimensional wave that propagates along the surface of the crystal at a depth of one wavelength at operating frequencies between 100 and 400 MHz (Pearce et al., 2003). The mass of the gas sensitive membrane of the SAW device is changed on interaction with a compatible analyte and causes the frequency of the wave to be altered. The change in frequency with sorption of a vapour is given by Df Df p cv K p =rp 7

where Dfp is the change in frequency caused by the membrane, cv is the vapour concentration, Kp is the partition coefcient, rp is the density of the polymer membrane used (Albert and Lewis, 2000; Nagle et al., 1998; Pearce et al., 2003). The substrates are normally prepared from ZnO, lithium niobate or quartz, which are piezoelectric in nature (Pearce et al., 2003). The sensitive membrane is usually polymeric or
Figure 6 SAW sensor

189

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

liquid crystal, however, phospholipids and fatty acids deposited using Langmuir-Blodgett techniques have also been used (Albert and Lewis, 2000). Generally, SAW devices are produced commercially by photolithography, where airbrush techniques are often used to deposit the thin lms ( Nagle et al., 1998; Schaller et al., 1998) approximately 20-30 nm thick (Groves and Zellers, 2001). SAW devices have also been produced by utilising screen printing techniques (White and Turner, 1997), dip coating and spin coating. The sensitivity of the SAW device to a particular odour depends on the type sensitive membrane and the uptake of vapours by polymers are governed by the same factors as those for composite conducting polymer sensors as detailed earlier in this paper. The sensitivity is normally quoted as differential response, i.e. Dfp/ppm of odour (Hierlemann et al., 1995; Pearce et al., 2003). Polymer-coated SAW devices have quite low detection limits, for example tetrachloroethylene, trichloroethylene and methoxyurane have been detected at concentrations as low as 0.7, 0.6 and 4 ppm, respectively (Albert and Lewis, 2000). Groves et al. used an array of SAW devices with polymer coatings 20-30 nm thick to detect 16 different solvents using four different polymers. Sensitivities ranging from 0.5 up to 12 Hz/mg/m3 have been achieved with these devices (Groves et al., 2001). Organophosphorous compounds have also been detected using these types of devices at concentrations from 10 to 100 ppm at temperatures between 25 and 508C where a sensitivity of 27 Hz/ppm at 308C (Dejous et al., 1995). As can be seen above, SAW devices can detect a broad spectrum of odours due to the wide range of gas sensitive coatings available (Carey et al., 1986; Grate and Abraham, 1991) and they also offer high sensitivity and fast response times and their fabrication is compatible with current planar IC technologies (Nagle et al., 1998). However, SAW devices suffer from poor signal to noise performance because of the high frequencies at which they operate, (Schaller et al., 1998) and the circuitry required to operate them is complex and expensive (Pearce et al., 2003). Batch to batch reproducibility is

difcult to achieve and the replacement of damaged sensors was also found to be problematic (Schaller et al., 1998).

Quartz crystal microbalance sensor

Quartz crystal microbalance (QCM) gas sensors operate on the same principle as SAW devices. When an ac voltage is applied across the piezoelectric quartz crystal the material oscillates at its resonant frequency, normally between 10 and 30 MHz (Schaller et al., 1998). The three-dimensional wave produced, travels through the entire bulk of the crystal. A membrane is deposited onto the surface of the crystal and this layer adsorbs gas when exposed to the vapour, which results in an increase in its mass. This increase in mass alters the resonant frequency of the quartz crystal and this change in resonant frequency is therefore used for the detection of the vapour (Albert and Lewis, 2000). The sensor, shown in Figure 7, is composed of a quartz disc coated with the absorbing polymer layer and a set of gold electrodes evaporated onto either side of the polymer/ quartz structure. Micromachining is used to fabricate the QCM devices, which makes the fabrication of small sensor structures possible. Coatings can be between 10 nm and 1 mm and are applied using spincoating, airbrushing, inkjet printing or dip coating, (Schaller et al., 1998). The sensitivity of QCMs is given by Df 22:3 1026 f 2 Dc A 8

Figure 7 Quartz crystal microbalance with polymer coating

190

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

where f is the fundamental frequency, c is the concentration and A is the area of the sensitive lm (Carey and Kowalski, 1986). Higher frequencies and smaller surface areas of sensitive coatings give rise to higher sensitivity (Carey, 1987). A quartz crystal microbalance gas sensor was developed and had a linear frequency shift on interaction with ethanol, n-heptane and acetone and showed a 16 Hz shift for 2.5 ppm of ethanol and , 1.5 Hz shift for 1 ppm of n-heptane, (Kim, 2002) demonstrating their high sensitivity to organic vapours. QCM devices are sensitive to a diverse range of analytes and are also very selective (Pearce et al., 2003). Polyvinylpyrrolidone modied QCMs have been used to detect various organic vapours and had sensitivities in the range of 7.5-48.2 Hz/mg/l (Mirmohseni and Oladegaragoze, 2003). The advantages of using QCMs are fast response times, typically 10 s (Haug et al., 1993), however, response times of 30 s to 1 min have been reported (Carey, 1987). However, QCM gas sensors have many disadvantages which include complex fabrication processes and interface circuitry, (Nagle et al., 1998), and poor signal to noise performance due to surface interferences and the size of the crystal (Nagle et al., 1998). As with SAW devices, batch-tobatch reproducibility of QCM gas sensors and the replacement of damaged sensors are difcult (Dickinson et al., 1998). Optical sensors used in e-nose arrays Optical bre sensor arrays are yet another approach to odour identication in e-nose systems. The sides or tips of the optic bres (thickness, 2 mm) are coated with a uorescent dye encapsulated in a polymer matrix as shown in Figure 8. Polarity alterations in the

uorescent dye, on interaction with the vapour, changes the dyes optical properties such as intensity change, spectrum change, lifetime change or wavelength shift in uorescence, (Nagle et al., 1998; Pearce et al., 2003). These optical changes are used as the response mechanism for odour detection (Grattan and Sun, 2000; Gopel, 1992). The sensitivity depends on the type of uorescent dye or mixture of dyes and the type of polymer used to support the dye (Nagle et al., 1998). The nature of the polymer controls the response and the most important factors are polarity, hydrophobicity, porosity and swelling tendency (Nagle et al., 1998; Pearce et al., 2003). Adsorbants, such as alumina, can be added to the polymer to improve the response by lowering the detection limits of the sensor (Walt et al., 1998). Polyanaline-coated optical sensors have been used to detect ammonia at concentrations as low as 1 ppm and the linear dynamic range was between 180 and 18,000 ppm, ( Jin, 2001). Optical gas sensors have very fast response times, less than 10 s for sampling and analysis ( Walt et al., 1998). These compact, lightweight optical gas sensors can be multiplexed on a single bre network, immune to electromagnetic interference (EMI) and can operate in high radiation areas due to Bragg and other grating based optical sensors (Gopel, 1992; Walt et al., 1998). However, there are several disadvantages of these types of sensors. The associated electronics and software are very complex, leading to increased cost, and the sensors have quite a short lifetime due to photobleaching (Nagle et al., 1998). However, this problem was overcome by measuring the temporal

Figure 8 A gas optical bre sensor. The odorant interacts with the material and causes a shift in the wavelength of the propagating light wave

191

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

responses because these remain consistent despite photobleaching (Walt et al., 1998). Metal-oxide-semiconductor eld-effect transistor sensor and its use in the e-nose The metal-oxide-semiconductor eld-effect transistor (MOSFET) is a transducer device used in e-nose to transform a physical/chemical change into an electrical signal. The MOSFET sensor is a metal-insulator-semiconductor (MIS) device and its structure is shown in Figure 9 (Eisele et al., 2001). This particular sensor works on the principle that the threshold voltage of the MOSFET sensor changes on interaction of the gate material, usually a catalytic metal, with certain gases, such as hydrogen, due to corresponding changes in the work functions of the metal and the oxide layers (Pearce et al., 2003). The changes in the work functions occur due to the polarization of the surface and interface of the catalytic metal and oxide layer when the gas interacts with the catalytically active surface (Kalman et al., 2000). In order for the physical changes in the sensor to occur, the metalinsulator interface has to be accessible to the gas. Therefore, a porous gas sensitive gate material is used to facilitate diffusion of gas into the material (Eisele et al., 2001). It has been observed that the change in the threshold voltage is proportional to the concentration of the analyte and is used as the response mechanism for the gas. Changes in the drainsource current and the gate voltage have also been used as the response mechanisms for the MOSFET gas sensors as they are also affected by changes in the work function (Albert and Lewis, 2000; Nagle et al., 1998). Gas sensing MOSFETs are produced by standard microfabrication techniques, which
Figure 9 MOSFET gas sensor with gas sensitive membrane deposited on top of SiO2

incorporate the deposition of gas sensitive catalytic metals onto the silicon dioxide gate layer (Nagle et al., 1998). In the case of catalytic metals, such as platinum (Pt), palladium (Pd) and iridium (Ir), the gate material is thermally evaporated onto the gate oxide surface through a mask forming 100-400 nm thick lms or 3-30 nm thin lms depending on the application (Albert and Lewis, 2000). Thick lms have been used to detect hydrogen and hydrogen sulphide while thin lms are more porous and can detect amines, alcohols and aldehydes (Kalman et al., 2000). Polymers have also been used as the gate material for MOSFET gas sensors (Hateld et al., 2000). Covington et al. 2001 used three polymers (poly(ethylene-co-vinyl acetate, poly(styrene-co-butadiene), poly(9-vinylcarbazole)) mixed with 20 per cent carbon black, which were deposited onto the FETs using a spray system. The polymer thickness was between 1.9 and 3.7 mm, respectively (Covington et al., 2001). These MOSFETs, which use polymers as the sensitive membrane are more commonly called PolFETs and can operate at room temperature. Apart from the standard MOSFET gas sensor architecture, a hybrid suspended gate FET (HSGFET) gas sensor can also be fabricated by micromachining (Figure 10). The HSGFET is a metal air-gap insulator (MAIS) device. The air-gap allows easy access to both the gate material and the insulator, so diffusion is not necessary and therefore a wider choice of gas sensitive materials can be used (Eisele et al., 2001; Pearce et al., 2003). The catalytic metal thickness and the operating temperatures of the MOSFETs in
Figure 10 Hybrid suspended gate eld effect transistor

192

Table V Characteristics of commercial e-noses Pattern recognition/ comments www.airsense.com/ www.alpha-mos.com Website/reference/university background

Manufacturer MOS CP, MOS, QCM, SAW 22 6-24 10 Food evaluation; avour and fragrance testing

Sensor type Applications

No. of sensors

Airsense analysis GmbH

Alpha MOS-Multi Organoleptic Systems

Applied Sensor

Analysis of food types; quality control of food storage, fresh sh, and petrochemical products; packaging evaluation; analysis of dairy products, alcoholic beverages and perfumes Identication of purity, process and quality control, environmental analysis, medical diagnosis ANN Desktop

ANN, DC, PCA Laptop ANN, DFA, PCA Desktop ANN, PCA Laptop

www.appliedsensor.com University of Linkopings, Sweden

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

AromaScan PLC QCM CP Food evaluation, avour and fragrance testing microbiology, environmental monitoring, degradation detection Food quality, chemical analysis, freshness, spoilage, contamination detection, consistency in foods and beverages 14 8

IR, MOS, MOSFET, QCM CP 32 Environmental monitoring, chemical quality control, pharmaceutical product evaluation Diagnosing lung cancer, food analysis

A review of gas sensors employed in electronic nose applications

Array Tech

University of Manchester Institute of science and technology, UK University of Rome University of Leeds, UK ANN, CA, PCA, DA Laptop

Bloodhound sensors

193
CP QCM, CP, MOS, SAW GC, SAW 1 40, 8 MOS, SAW 8-28 32 QCM 6 FO

Cyrano Science Inc.

Marconi Applied Technologies

PCA Palmtop ANN, DA, PCA SPR Desktop PCA

www.cyranosciences.com California institute of Technology, USA University of Warwick, UK www.estcal.com www.fzk.de/FZK2/english

Electronic Sensor Technology Inc. Forschungszentrum Karlsruhe

Sensor Review

Volume 24 Number 2 2004 181198

HKR-Sensorsysteme GmbH

Food and beverage quality, bacteria identication, explosives and drug detection, environmental monitoring Environmental protection, industrial process control, air monitoring in textile mills, re alarms. Quality control in food production. Automotive applications Food and beverages, cosmetics and perfumes, organic materials, pharmaceutical industry Life sciences, food processing, agriculture, chemical detection

Illumina

ANN, CA, DFA, PCA Desktop ANN

www.hkr-sensor.de/ Technical University of Munich, Germany www.illumina.com Tufts University, USA (Continued )

Table V Pattern recognition/ comments ANN, PCA Desktop Website/reference/university background

Manufacturer MOS, QCM 16-40 Food and beverage processing, perfume oils, agricultural odours and chemical analysis, process control Sniffs palms for personal identication

Sensor type Applications

No. of sensors

Lennartz Electronic GmbH

A review of gas sensors employed in electronic nose applications

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Marconi technologies Mastiff Electronic Systems Ltd. Microsensor Systems SAW CO MOS, QCM, SAW MOS 6 Diagnosing tuberculosis, bowel cancers, infections in wounds and the urinary tract 6-10 2 Chemical agent detection, chemical warfare (CW) agents and toxic industrial chemicals (TICs) analysis Automotive applications, food evaluation, packaging evaluation Early recognition of res, warning in the event of escape of hazardous substances, leak detection, workplace monitoring

CP

16

www.lennartz-electronic.de/ PDF_documents University of Tubingen www.mastiff.co.uk/

Palmtop

194

OligoSense RST Rostock Raum-fahrt und Umweltschatz GmbH Shimadzu Co. Diag-Nose

ANN, PCA Desktop PCA, Desktop

www.microsensorsystems.com/pdf/ hazmatcad http://www.oligosense.be Antwerp University, Belgium www.rst-rostock.de/ www.nose-network.org/ Craneld University, Bedfordshire, England

Sensor Review

Volume 24 Number 2 2004 181198

Notes: MOS metal oxide semiconductor; CP conducting polymer; QCM quartz crystal monitor; SAW surface acoustic wave; IR infra red; MOSFET metal oxide semiconductor eld effect transistor; GC gas chromatography; FO bre optic; CO conductive oligomer; ANN articial neural network; DC distance classiers; PCA principle component analysis; DFA discriminant function analysis; CA cluster analysis; DA discriminant analysis

Table VI Summary of the properties of each sensor type reviewed Fabrication ppb for HVPG ppm for LVPG; 1 per cent DR/Rb/ppm; DL # 0.1-5 ppm 0.1-100 ppm Operate at room temperature, cheap, diverse range of coatings Sensitive to temperature and humidity Examples of sensitivity/detection range/detection limits (DL) Advantages Disadvantages Reference Munoz et al. (1999), Gardner and Dyer (1997) and Ryan et al. (1999) Nagle et al. (1998)

Sensor type

Measurand

Polymer composites

Conductivity

Conductivity

Screenprinting, spincoating, dipcoating, spray coating, microfabrication Electrochemical, chemical polymerisation 5-500 ppm Sensitive to polar analytes, cheap, good response times, operate at room temperature Fast response and recovery times, cheap Sensitive to temperature and humidity, suffer from baseline drift High operating temperatures, suffer from sulphur poisoning, limited range of coatings Complex interface circuitry, difcult to reproduce

A review of gas sensors employed in electronic nose applications

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Intrinsically conducting polymers Metal oxides

Conductivity

Nagle et al. (1998)

SAW

Piezoelectricity

195
Low ppb; DL (NH3) 1 ppm with polyaniline coating

QCM

Piezoelectricity

Sensor Review

Optical devices

Intensity/ spectrum

Screenprinting, RF sputtering, thermal evaporation, microfabrication Photolithography, airbrushing, screenprinting, dipcoating, spincoating Micromachining, spincoating, airbrushing, inkjet printing, dipcoating Dipcoating 1 pg to 1 mg of vapour 1 pg mass change; DL 2 ppm for octane and 1 ppm NO2 and 1 ppm H2S with polymer membranes 1.5 Hz/ppm; 1 ng mass change Diverse range of coatings, high sensitivity, good response times, IC integratable Diverse range of coatings, good batch to batch reproducibility

Nagle et al. (1998), Albert and Lewin (2000) and Penza et al. (2001a) Poor signal-to-noise ratio, complex circuitry Nagle et al. (1998) and Kim and Choi (2002)

Jin et al. (2001) and Nagle et al. (1998) Suffer from photobleaching, complex interface circuitry, restricted light sources Baseline drift, need controlled environment Covington et al. (2001) and Kalman et al. (2000)

Volume 24 Number 2 2004 181198

MOSFET

Threshold voltage change

Microfabrication, thermal evaporation

Immune to electromagnetic interference, fast response times, cheap, light weight Small, low cost sensors, CMOS 2.8 mV/ppm for toluene; DL (amines, Sulphides) 0.1ppm; integratible and reproducible Maximum response 200 mV especially for amines

Notes: HVPG high vapor pressure gas; LVPG low vapor pressure gas; DR/Rb relative differential resistance change

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

the e-nose arrays are altered to provide different selectivity patterns to different gases (Schaller et al., 1998), and operating temperatures are usually between 50 and 1708C (Kalman et al., 2000). The factors that affect the sensitivity of FET-based devices are operating temperature, composition and structure of the catalytic metal (Lundstrom et al., 1995; Schaller et al., 1998). The temperature is increased to decrease the response and the recovery times (Dickinson et al., 1998). Typical sensitivities of ChemFET sensors with carbon black-polymer composite sensitive lms was found to be approximately 2.8 mV/ppm for toluene vapour (Covington et al., 2001). The detection limit of MOSFET devices for amines and sulphides was 0.1 ppm with Pt, Ir and Pd gates and the maximum response was approximately 200 mV especially for amines (Kalman et al., 2000). Response time varies from device to device and response times from milliseconds up to 300 s have been reported in the literature (Covington et al., 2001; Eisele et al., 2001; Wingbrant et al., 2003). MOSFET sensors have a number of advantages and disadvantages when used in e-nose arrays. Gas sensing MOSFETs are produced by microfabrication, therefore reproducibility is quite good and the sensor can be incorporated into CMOS technology resulting in small, low cost sensors (Dickinson et al., 1998; Gu et al., 1998; Nagle et al., 1998; Pearce et al., 2003; Schaller et al., 1998; ). The sensors can suffer from baseline drift and instability depending on the sensing material used (Nagle et al., 1998). If CMOS is used, the electronic components of the chip have to be sealed because the sensor needs a gas inlet so it can penetrate the gate (Nagle et al., 1998). The gas ows across the sensor and also the operating temperature have serious effects on the sensitivity and selectivity of the sensor, so control of the surrounding environment is important (Eklov et al., 1997; Nagle et al., 1998). This does not suit a handheld e-nose.

different sensor types depending mainly on the applications. Many of these systems are reviewed in Table V.

Conclusions
Conducting polymer composite, intrinsically conducting polymer and metal oxide conductivity gas sensors, SAW and QCM piezoelectric gas sensors, optical gas sensors and MOSFET gas sensors have been reviewed in this paper. These systems offer excellent discrimination and lead the way for a new generation of smart sensors which will mould the future commercial markets for gas sensors. The principle of operation, fabrication techniques, advantages, disadvantages and applications of each sensor type in e-nose systems have been clearly outlined and a summary of this information is given in Table VI.

References
Albert, K.J. and Lewis, N.S. (2000), Cross reactive chemical sensor arrays, Chem. Rev., Vol. 100, pp. 2595-626. Behr, G. and Fliegel, W. (1995), Electrical properties and improvement of the gas sensitivity in multiple-doped sno2, Sensors and Actuators B: Chemical, Vol. 26 Nos 1-3, pp. 33-7. Briglin, S.M., Freund, M.S., Tokumaru, P. and Lewis, N.S. (2002), Exploitation of spatiotemporal information and geometric optimization of signal/noise performance using arrays of carbon black-polymer composite vapor detectors, Sensors and Actuators B: Chemical, Vol. 82 No. 1, pp. 54-74. Carey, P.W. and Kowalski, B.R. (1986), Chemical piezoelectric sensor and sensor array characterisation, Analytical chemistry, Vol. 58, pp. 3077-84. Carey, P.W., Beebe, K.R. and Kowalski, B.R. (1986), Selection of adsorbates for chemical sensor arrays by pattern recognition, Analytical chemistry, Vol. 58, pp. 149-53. Carey, P.W., Beebe, K.R. and Kowalski, B.R. (1987), Multicomponent analysis using an array of piezoelectric crystal sensors, Analytical chemistry, Vol. 59, pp. 1529-34. Charlesworth, J.M., Partridge, A.C. and Garrard, N. (1997), Mechanistic studies on the interactions between poly(pyrrole) and orgaic vapors, J. Phys. Chem., Vol. 97, pp. 5418-23. Chen, J.H., Iwata, H., Tsubokawa, N., Maekawa, Y. and Yoshida, M. (2002), Novel vapor sensor from polymergrafted carbon black: effects of heat-treatment and gamma-ray radiation-treatment on the response of

Commercial e-nose systems

Manufacturers of commercial e-nose systems have become plentiful in recent times and produce a wide range of products utilising

196

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

sensor material in cyclohexane vapor, Polymer, Vol. 43 No. 8, pp. 2201-6. Comyn, J. (1985), Polymer permeability, Elsevier Applied Science, Amsterdam. Corcoran, P. (1993), Electronic odor sensing systems, Electronics and Communication Engineering Journal, Vol. 5 No. 5, pp. 303-8. Covington, J.A., Gardner, J.W., Briand, D. and de Rooij, N.F. (2001), A polymer gate fet sensor array for detecting organic vapours, Sensors and Actuators B: Chemical, Vol. 77 Nos 1-2, pp. 155-62. Dai, G. (1998), A study of the sensing properties of thin lm sensor to trimethylamine, Sensors and Actuators B: Chemical, Vol. 53 Nos 1-2, pp. 8-12. Dejous, C., Rebiere, D., Pistre, J., Tiret, C. and Planade, R. (1995), A surface acoustic wave gas sensor: detection of organophosphorus compounds, Sensors and Actuators B: Chemical, Vol. 24 Nos 1-3, pp. 58-61. Dickinson, T.A., White, J., Kauer, J.S. and Walt, D.R. (1998), Current trends in articial-nose technology, Trends in Biotechnology, Vol. 16 No. 6, pp. 250-8. Doleman, B.J., Lonergan, M.C., Severin, E.J., Vaid, T.P. and Lewis, N.S. (1998a), Quantitative study of the resolving power of arrays of carbon black-polymer composites in various vapor-sensing tasks, Analytical Chemistry, Vol. 70 No. 19, pp. 4177-90. Doleman, B.J., Sanner, R.D., Severin, E.J., Grubbs, R.H. and Lewis, N.S. (1998b), Use of compatible polymer blends to fabricate arrays of carbon black-polymer composite vapor detectors, Analytical chemistry, Vol. 70, pp. 2560-4. Eisele, I., Doll, T. and Burgmair, M. (2001), Low power gas detection with fet sensors, Sensors and Actuators B: Chemical, Vol. 78 No. 1-3, pp. 19-25. Eklov, T., Sundgren, H. and Lundstrom, I. (1997), Distributed chemical sensing, Sensors and Actuators B: Chemical, Vol. 45 No. 1, pp. 71-7. Fang, Q., Chetwynd, D.G., Covington, J.A., Toh, C-S. and Gardner, J.W. (2002), Micro-gas-sensor with conducting polymers, Sensors and Actuators B: Chemical, Vol. 84 No. 1, pp. 66-71. Fraden, J. (1996), Aip Handbook of Modern Sensors, AIP, New York. Freund, M.S. and Lewis, N.S. (1995), A chemically diverse conducting polymer based electronic nose, Proc. Natl. Acad. Sci. USA, Vol. 92, pp. 2652-6. Galdikas, A., Mironas, A. and Setkus, A. (1995), Copperdoping level effect on sensitivity and selectivity of tin oxide thin-lm gas sensor, Sensors and Actuators B: Chemical, Vol. 26 No. 1-3, pp. 29-32. Gardner, J.W. and Bartlett, P.N. (1995), Application of conducting polymer technology in microsystems, Sensors and Actuators A: Physical, Vol. 51 No. 1, pp. 57-66. Gardner, J.W. and Dyer, D.C. (1997), High-precision intelligent interface for a hybrid electronic nose, Sensors and Actuators A: Physical, Vol. 62 No. 1-3, pp. 724-8. George, S.C. and Thomas, S. (2001), Transport phenomena through polymeric systems, Progress in Polymer Science, Vol. 26 No. 6, pp. 985-1017.

Gopel, W., Hesse, J. and Zemel, J.N. (Eds) (1989), Magnetic Sensors, VCH, Weinheim. Gopel, W., Hesse, J. and Zemel, J.N. (Eds) (1992), Optical Sensors, VCH, Weinheim. Grate, W.J. and Abraham, M.H. (1991), Solubility interactions and design of chemically selective sorbant coatings for chemical sensors and arrays, Sensors and Actuators B, Vol. 3, pp. 85-111. Grattan, K.T.V. and Sun, T. (2000), Fiber optic sensor technology: an overview, Sensors and Actuators A: Physical, Vol. 82 Nos 1-3, pp. 40-61. Groves, W.A. and Zellers, E.T. (2001), Analysis of solvent vapors in breath and ambient air with a surface acoustic wave sensor array, The Annals of Occupational Hygiene, Vol. 45 No. 8, pp. 609-23. Gu, C., Sun, L., Zhang, T., Li, T. and Zhang, X. (1998), Highsensitivity phthalocyanine lb lm gas sensor based on eld effect transistors, Thin Solid Films, Vol. 327-329, pp. 383-6. Guadarrama, A., Fernandez, J.A., Iniguez, M., Souto, J. and de Saja, J.A. (2000), Array of conducting polymer sensors for the characterisation of wines, Analytica Chimica Acta, Vol. 411 No. 1-2, pp. 193-200. Hamakawa, S., Li, L., Li, A. and Iglesia, E. (2002), Synthesis and hydrogen permeation properties of membranes based on dense srce0.95yb0.05o3-[alpha] thin lms, Solid State Ionics, Vol. 148 No. 1-2, pp. 71-81. nyi, G. (2000), Polymer lms in sensor applications: Harsa a review of present uses and future possibilities, Sensor Review, Vol. 20 No. 2, pp. 98-105. Hateld, J.V., Covington, J.A. and Gardner, J.W. (2000), Gasfets incorporating conducting polymers as gate materials, Sensors and Actuators B: Chemical, Vol. 65 No. 1-3, pp. 253-6. Haug, M., Schierbaum, K.D., Gauglitz, G. and Gopel, W. (1993), Chemical sensors based upon polysiloxanes: Comparison between optical, quartz microbalance, calorimetric, and capacitance sensors, Sensors and Actuators B: Chemical, Vol. 11 No. 1-3, pp. 383-91. Heeger, A.J. (2001), Semiconducting and metallic polymers: the fourth generation of polymeric materials, Current Applied Physics, Vol. 1, pp. 247-67. Hierlemann, A., Weimar, U., Kraus, G., Schweizer-Berberich, M. and Gopel, W. (1995), Polymer-based sensor arrays and multicomponent analysis for the detection of hazardous organic vapours in the environment, Sensors and Actuators B: Chemical, Vol. 26 Nos 1-3, pp. 126-34. Jin, Z., Su, Y. and Duan, Y. (2001), Development of a polyaniline-based optical ammonia sensor, Sensors and Actuators B: Chemical, Vol. 72 No. 1, pp. 75-9. Johnson, D.J. (1997), Process control instrumentation technology, Prentice-Hall. Kalman, E-L., Lofvendahl, A., Winquist, F. and Lundstrom, I. (2000), Classication of complex gas mixtures from automotive leather using an electronic nose, Analytica Chimica Acta, Vol. 403 No. 1-2, pp. 31-8. Khlebarov, Z.P., Stoyanova, A.I. and Topalova, D.I. (1992), Surface acoustic wave gas sensors, Sensors and Actuators B: Chemical, Vol. 8 No. 1, pp. 33-40.

197

A review of gas sensors employed in electronic nose applications

Sensor Review Volume 24 Number 2 2004 181198

K. Arshak, E. Moore, G.M. Lyons, J. Harris and S. Clifford

Kim, Y.H. and Choi, K.J. (2002), Fabrication and application of an activated carbon coated quartz crystal sensor, Sensors and Actuators B, Vol. 87, pp. 196-200. Lonergan, M.C., Severin, E.J., Doleman, B.J., Beaber, S.A., Grubb, R.H. and Lewis, N.S. (1996), Array-based vapor sensing using chemically sensitive, carbon blackpolymer resistors, Chemistry of Materials, Vol. 8 No. 9, pp. 2298-312. Lundstrom, I., Ederth, T., Kariis, H., Sundgren, H., Spetz, A. and Winquist, F. (1995), Recent developments in eld-effect gas sensors, Sensors and Actuators B: Chemical, Vol. 23 No. 2-3, pp. 127-33. Matthews, B., Jing, L., Sunshine, S., Lerner, L. and Judy, J.W. (2002), Effects of electrode conguration on polymer carbon-black composite chemical vapor sensor performance, IEEE Sensors Journal, Vol. 2 No. 3, pp. 160-8. Mirmohseni, A. and Oladegaragoze, A. (2003), Construction of a sensor for determination of ammonia and aliphatic amines using polyvinylpyrrolidone coated quartz crystal microbalance, Sensors and Actuators B: Chemical, Vol. 89 No. 1-2, pp. 164-72. Munoz, B.C., Steinthal, G. and Sunshine, S. (1999), Conductive polymer-carbon black composites-based sensor arrays for use in an electronic nose, Sensor Review, Vol. 19 No. 4, pp. 300-5. Nagle, H.T., Gutierrez-Osuna, R. and Schiffman, S.S. (1998), The how and why of electronic hoses, IEEE Spectrum, Vol. 35 No. 9, pp. 22-31. Partridge, A.C., Jansen, M.L. and Arnold, W.M. (2000), Conducting polymer-based sensors, Materials Science and Engineering: C, Vol. 12 No. 1-2, pp. 37-42. Pearce, T.C., Schiffman, S.S., Nagle, H.T. and Gardner, J.W. (2003), Handbook of Machine Olfaction, Wiley-VCH, Weinheim. Penza, M., Cassano, G., Sergi, A., Lo Sterzo, C. and Russo, M.V. (2001a), Saw chemical sensing using poly-ynes and organometallic polymer lms, Sensors and Actuators B: Chemical, Vol. 81 No. 1, pp. 88-98. Penza, M., Cassano, G. and Tortorella, F. (2001b), Gas recognition by activated wo3 thin-lm sensors array, Sensors and Actuators B: Chemical, Vol. 81 No. 1, pp. 115-21. Penza, M., Cassano, G., Tortorella, F. and Zaccaria, G. (2001c), Classication of food, beverages and perfumes by wo3 thin-lm sensors array and pattern recognition techniques, Sensors and Actuators B: Chemical, Vol. 73 No. 1, pp. 76-87. Ryan, M.A., Buehler, M.G., Homer, M.L., Mannatt, K.S., Lau, B., Jackson, S. and Zhou, H. (1999), The Second International Conference on Integrated Micro/ Nanotechnology for Space Applications-Enabling Technologies for New Space Systems, Pasadena, CA, USA. Saha, M., Banerjee, A., Halder, A.K., Mondal, J., Sen, A. and Maiti, H.S. (2001), Effect of alumina addition on methane sensitivity of tin dioxide thick lms, Sensors and Actuators B: Chemical, Vol. 79 No. 2-3, pp. 192-5.

Sakurai, Y., Jung, H-S., Shimanouchi, T., Inoguchi, T., Morita, S., Kuboi, R. and Natsukawa, K. (2002), Novel arraytype gas sensors using conducting polymers, and their performance for gas identication, Sensors and Actuators B: Chemical, Vol. 83 No. 1-3, pp. 270-5. Schaller, E., Bosset, J.O. and Escher, F. (1998), Electronic noses and their application to food, LebensmittelWissenschaft und-Technologie, Vol. 31 No. 4, pp. 305-16. Schmid, W., Barsan, N. and Weimar, U. (2003), Sensing of hydrocarbons with tin oxide sensors: possible reaction path as revealed by consumption measurements, Sensors and Actuators B: Chemical, Vol. 89 No. 3, pp. 232-6. Severin, E.J., Doleman, B.J. and Lewis, N.S. (2000), An investigation of the concentration dependence and response to analyte mixtures of carbon black/insulating organic polymer composite vapor detectors, Analytical chemistry, Vol. 72, pp. 658-68. Severin, E.J., Sanner, R.D., Doleman, B.J. and Lewis, N.S. (1998), Differential detection of enantiomeric gaseous analytes using carbon black-chiral polymer composite, chemically sensitive resistors, Analytical chemistry, Vol. 70, pp. 1440-3. Shurmer, H.V. and Gardner, J.W. (1992), Odour discrimination with an electronic nose, Sensors and Actuators B, Vol. 8, pp. 1-11. Sotzing, G.A., Phend, J.N. and Lewis, N.S. (2000), Highly sensitive detective and discrimination of biogenic amines ulitizing arrays of polyanaline/carbon black composite vapor detectors, Chem. Mater., Vol. 12, pp. 593-5. Steffes, H., Imawan, C., Solzbacher, F. and Obermeier, E. (2001), Enhancement of no2 sensing properties of in2o3-based thin lms using an au or ti surface modication, Sensors and Actuators B: Chemical, Vol. 78 No. 1-3, pp. 106-12. Tao, W-H. and Tsai, C-H. (2002), H2s sensing properties of noble metal doped wo3 thin lm sensor fabricated by micromachining, Sensors and Actuators B: Chemical, Vol. 81 No. 2-3, pp. 237-47. Vieth, W.R. (1991), Diffusion in and Through Polymers: Principles and Applications, Hanser Publishers, Munich. Walt, D.R., Dickinson, T., White, J., Kauer, J., Johnson, S., Engelhardt, H., Sutter, J. and Jurs, P. (1998), Optical sensor arrays for odor recognition, Biosensors and Bioelectronics, Vol. 13 No. 6, pp. 697-9. White, N.M. and Turner, J.D. (1997), Thick-lm sensors: past, present and future, Meas Sci. Technology, Vol. 8, pp. 1-20. Wingbrant, H., Lundstrom, I. and Lloyd Spetz, A. (2003), The speed of response of MISiCFET devices, Sensors and Actuators B: Chemical, Vol. 93 Nos. 1-3, pp. 286-94. Yasufuku (2001), Electroconductive polymers and their applications in Japan, IEEE Electrical Insulation Magazine, Vol. 17 Nos. 5, pp. 14-24. Zee, F. and Judy, J.W. (2001), Micromachined polymer-based chemical gas sensor array, Sensors and Actuators B: Chemical, Vol. 72 No. 2, pp. 120-8.

198