A room-temperature operated hydrogen leak sensor H. Nakagawa a,* , N. Yamamoto b , S. Okazaki b , T. Chinzei a , S. Asakura b a Research Center for Advanced Science and Technology, University of Tokyo, Komaba, Meguro-ku, Tokyo 153-8904, Japan b Graduate School of Engineering, Yokohama National University, Tokiwadai, Hodogaya-ku, Yokohama 240-8501, Japan Abstract A new chemi-resister type sensor for hydrogen leak detection is suggested. Tungsten trioxide (WO 3 ) with Pt was used as sensing material. The sensor was fabricated by a sol–gel method. Tungstic acid sol with chloroplatinic acid was spread on a quartz plate with a spinner and calcined in atmosphere to form a WO 3 film. Pt was expected to act as catalyst for hydrogen reduction of WO 3 . The conductivity of the sensor was less than 0.001 mS in oxidizing atmosphere, and more than 10 6 times conductivity increase was observed upon exposure to (1% H 2 /99% N 2 ) gas. Transient characteristics of the reduction process and oxidation process were not identical. The reduction process exhibited super-linear nature, whereas oxidation process may be approximated by a simple exponential decay. The sensitivity was susceptible to humidity. Not only the response was faster and more sensitive in humid environment than in dry one, it was also affected by the previous exposure history. Even in dry environment the sensitivity increases if the device was exposed to hydrogen, several times. A new detection scheme to explain these observations is suggested. # 2003 Elsevier Science B.V. All rights reserved. Keywords: Gas sensor; Tungsten trioxide; Hydrogen; Double injection 1. Introduction With the increasing concern about the global climate change, more attention is paid to hydrogen as a clean energy source. Hydrogen burns into water and no global warming gas is produced. Although pure hydrogen fuel may not be utilized in near future, fuel cells may be brought in a wide usage to automobile and home within a few years. Some precautions are required, however, for the safe use of hydrogen. Hydrogen has a large diffusion coefficient (0.61 cm 2 /s in air, methane’s value is 0.15 cm 2 /s) and wide combustion range (4–75%) and small ignition energy (0.02 mJ in air, methane’s value is 0.3 mJ). Continuous monitoring of hydrogen leak at storage or usage sites is indispensable for safe operation. Two types of sensitive sensors are widely used for this purpose. A high-temperature operated oxide–semiconductor gas sensor has high sensitiv- ity, reliability, as well as maintenance-free nature, but con- sumes relatively large power for device heating. The other type, an electrochemical gas sensor consumes very little power, but electrolyte liquid within the cell has to be replaced every year for the reliable operation. Development of a new hydrogen sensor that consumes negligible electrical energy with negligible maintenance is highly desirable. Tungsten trioxide (WO 3 ) is known to interact with hydro- gen and other alkaline metal ions in a unique manner. This interaction seduces the development of hydrogen sensor [1,2]. In the present work, we report sensing characteristics of a resistance-sensing hydrogen sensor that is operated at ambient temperature. WO 3 film with platinum catalyst derived from a sol–gel method was utilized. The sensor of this type is particularly suitable for hydrogen leak mon- itoring because it consumes negligible electrical power due to the insulating characteristics in air. It had been known that WO 3 changes its color to blue upon partial reduction. The reduction may be achieved either by electrochemical reac- tion in liquid (electrochromism) [3] or by gas-phase reaction in reducing atmosphere (gasochromism) [4].WO 3 is classi- fied as oxide semiconductor with a band gap of about 3.2 eV. Its electrical resistance is very high due to wide band gap in oxidized state, but the resistance becomes low upon reduc- tion due to generated free electrons [5].AWO 3 hydrogen gas sensor that detects the resistance change was reported more than 30 years ago [1], but the detection mechanism is still of some controversy. Traditionally, the double injection model [3], in which both a proton (hydrogen ion) and an electron are simultaneously supplied to the film, is widely accepted. Not all the observations were in accordance with Sensors and Actuators B 93 (2003) 468–474 * Corresponding author. Tel.: þ81-3-5452-5241; fax: þ81-3-5452-5241. E-mail address: nakagawa@bme.rcast.u-tokyo.ac.jp (H. Nakagawa). 0925-4005/03/$ – see front matter # 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0925-4005(03)00201-6 this model, however, and other models based on the oxygen deficiency [6,7] were suggested. Our observations may be interpreted by the double injection/surface oxidation model. 2. Experimental AWO 3 sensor was fabricated on a quartz plate by a sol–gel method and spin coating. An amount equal to 0.5 M sodium tungstate (Na 2 WO 4 ) solution was converted into tungstate (H 2 WO 4 ) sol solution by passing through cationic ion exchange column (Amberlite IB120, Organo Co.). Appro- priate amounts of hexachroloplatinum (H 2 PtCl 6 ) solution and ethyl alcohol were added. Addition of ethyl alcohol prolonged the gelation time. This solution was spread on a quartz plate and formed a thin film by a home-made spinner (900 rpm). The film was dried for a few days and calcined for 1 h at 200 8C in air to remove crystalline water. Final calcination time was 1 h in air with calcination temperatures ranging 300–700 8C. The sensor was placed in a flow- through chamber and the electrical conductance was mea- sured with a LCR meter (HP 4263B) under various gas compositions and temperatures. Five hundred millivolts rms ac signal of 1 kHz was applied during the measurements. Relatively large applied bias eased the conductance mea- surement of semi-insulating oxidized WO 3 . Electrical con- tacts were achieved by physical contact of two parallel copper electrodes about 0.5 mm apart. Good Ohmic contacts were ascertained by the linear current–voltage characteris- tics. Sensitivity measurements were performed at room temperature unless otherwise stated. Test gases used in the sensitivity measurements were drawn from a bottle and the test gases were passed through a water-filled bub- bling vessel when humid gases were required. The relative humidity (RH) of humid gas was about 85% at room temperature. 3. Results and discussions Knowledge of crystalline structure is essential for the accurate interpretation of the observed result. Tungsten trioxides exist in several polymorphic forms, i.e. monoclinic [8], hexagonal [9], and pyrochlore [10] forms around room temperature. The basic unit of these three crystals are octahedral unit in which W atom stays at the center and oxygen atoms form every corners. Monoclinic crystal struc- ture may be considered as slightly-warped ReO 3 type, or of warped perovskite (ABO 3 ) with vacant A sites. Corners of every octahedrons are shared with neighboring octahedrons and none of the edges are shared in monoclinic form. Monoclinic WO 3 exhibits six phase transitions with tem- perature [11]. It is monoclinic below À40 8C and triclinic between À40 and 20 8C, and another monoclinic between 20 and 325 8C. It then changes to an orthorhombic at 325 8C, and succession of tetragonal at 725, 900, and 1225 8C. The actual change associated with these transitions is slight change of bond length and all crystal phases of monoclinic family may be considered as a modification of cubic ReO 3 structure in the first order approximation. Monoclinic phase is the most stable and both hexagonal and pyrochlore crystals switch to monoclinic phase if it is heated to more than 500 8C and cooled down to room temperature. Both hexagonal and pyrochlore are obtained as polycrystalline powder and no single crystal of appreciable size had not been obtained. Several crystal forms contain lattice water. WO 3 Á2H 2 O and WO 3 ÁH 2 O [12] are believed to be mono- clinic and WO 3 Á(1/3)H 2 O is hexagonal [13]. The initial state of WO 3 films obtained by sol–gel method or vacuum deposition was considered to be amorphous with some coordinated and adsorbed water, but the amorphous state mainly consists of nanocrystals and short range order of the crystal structure is conserved [14]. Calcination procedure coagulates and nanocrystals to form polycrystalline film and desorbs water. The effect of calcination temperature on the sensitivity was investigated. The conductance values of 2 min after exposure to humid (1% H 2 /99% N 2 gas) were plotted against calcination temperature in Fig. 1. As will be shown in Fig. 5, humidity affects the sensing characteristics. The molar ratio of tungstate and platinum (W/Pt) was chosen to be the best sensitivity value of 13. Good sensitivity was obtained when the film was sintered above 400 8C. Although this tempera- ture coincides with the phase change to the orthorhombic phase (which changes to monoclinic at room temperature), the major reason would be the reduction of platinum ions. Platinum has to be in the form of metal particles to function as efficient catalysts. If we use K 2 PtCl 4 instead of H 2 PtCl 6 , then some sensitivity appears even without any sintering. As the temperature was further increased, the sensitivity increased further and eventually decreased slightly at 700 8C. As temperature increases, crystal size becomes larger and quality of crystal improves. But if the crystal size becomes too large, then surface area would decrease and s- e- Fig. 1. Sensitivity vs. calcination temperature relation. Conductivity of 2 min after exposure to humid (1% H 2 /99% N 2 ) gas was plotted. W/Pt ratio was 13. H. Nakagawa et al. / Sensors and Actuators B 93 (2003) 468–474 469 nsitivity reduces. Fig. 2 presents the effect of catalyst amount to the sensitivity. The conditions of the measurements are as same as those for Fig. 1. The conductance values of 2 min after exposure to humid (1% H 2 /99% N 2 ) gas were plotted as a function of W/Pt ratio. The calcinations temperature was 600 8C. The highest sensitivity was achieved when W/Pt ratio was 13. The sensitivity decreased as W/Pt increases in the large W/Pt region. This was expected since the con- centration of the tungstate solution was constant (0.5 M), larger the W/Pt ratio implies lower amount of catalyst. The sensitivity also decreased in high Pt region (low W/Pt range). If the platinum concentration is too high, then the platinum particles coagulate each other. This coagulation might have reduced the effective catalyst surface area and spill-over probability. The transient responses of the sensor at various tempera- tures were plotted in Fig. 3. The sensor was exposed to humid (1% H 2 /99% N 2 ) gas for 1200 s and humid 100% O 2 afterwards. Higher temperature resulted faster response. At 13 8C, the sensor did not reached the stable state within 1200 s. But more than 100 times increase in conductance was obtained within 120 s. The conductance ratios of more than 10 6 were obtained at every temperature except 13 8C. It should be mentioned that the transient response to hydrogen is super-linear, and it is almost exponential at the beginning. Since ordinate is scaled in logarithm, linear portion of the curve meant exponential response. Although the rising part of hydrogen response is super-linear in all temperatures, they cannot be approximated by a simple power law or a exponential function. Falling part or oxygen response may be divided into two distinct regions: a fast decaying portion at the beginning and a slow decaying portion of the follow- ing part. Both portions may be crudely approximated by a simple exponential decay with temperature dependent time constants. The complicated nature of the reduction and oxidation processes were confirmed by the separate optical measurement [15]. To investigate the nature of hydrogen response further, maximum response speed of response curve was plotted in Arrhenius format in Fig. 4. The data point of the oxidation velocity at 50 and 13 8C is likely to be underestimated than the intrinsic value, since the hydrogen response at 13 8C did not reached the stable state (Fig. 3). For the hydrogen response or reduction reaction, the curve may be divided into two regions. The dominating process that limits the response may differ in different temperature range. The Arrhenius energy in the low temperature range was $75 kJ/mol, whereas that in the high-temperature range was $23 kJ/mol. Two different mechanisms may be involved as a rate-determining process. The Arrhenius energy in the low temperature range is within the range of reaction limited process, whereas that of the high-tem- perature range may be in the higher range of the diffusion- limited processes. Complex rise-time characteristics, how- ever, suggest the many other possibilities such as simulta- neous contributions from two sequential processes. Fig. 2. Sensitivity vs. W/Pt molar ratio relation. Conductivity of 2 min after exposure to humid (1% H 2 /99% N 2 ) gas was plotted. Calcination temperature was 600 8C. Fig. 3. Transient characteristics at several temperatures. Humid (1% H 2 /99% N 2 ) gas was flown for the initial 1200 s, and humid air was flown afterwards. 470 H. Nakagawa et al. / Sensors and Actuators B 93 (2003) 468–474 Temperature range for the oxidation reaction may seem to be divided into two regions, but this could be superficial, because oxidation process at 13 8C had not started from the stable states and oxidation velocity at this temperature is likely to be underestimated. In the high-temperature region of the oxidation process, Arrhenius energy was $5 kJ/mol. We attributed this value to the surface diffusion of oxygen. It has been known that water affects the sensitivity from the early stage of the sensor development [1,30]. Fig. 5 shows transient response to repeated exposures of dry and humid (1% H 2 /99% N 2 ) gas for 300 s, and dry and humid (100% O 2 ) gas for 300 s. The device was kept in dry or humid nitrogen atmosphere for several hours before the measurements. In the dry atmosphere the both reduction and oxidation responses were weak and slow, but both responses improved with the repeated exposure to hydrogen. The improvement in response could be interpreted by the accumulation of the water in the film generated by hydrogen exposure. The response to the humid gas is high and fast. The response slightly decreased in the second exposure, but decrease was small in comparison to the dry gas response. The calibration curve for hydrogen in air was plotted in Fig. 6. The sensitivity is not linear with concentration. The response in air is roughly three orders of magnitude smaller than that in nitrogen. The sensor response was obtained as a result of the competing reactions: hydrogen reduction and oxygen oxidation. Faughnan et al. [3] proposed a double injection model for electrochromic coloration of tungsten trioxide. Protons and electrons are simultaneously supplied to keep the charge neutrality, Protons are supposedly intercalated to form a tungsten bronze. An injected electron reduces a W 6þ ion to a W 5þ ion, and the polaron transition between a W 5þ ion and nearby a W 6þ ion is responsible for the blue coloration [16]. The model nicely explained the coloration and electrical characteristics. This model may be easily expanded to reduction of WO 3 by hydrogen if one assumes the double injection of a proton and an electron from a catalyst metal. Fig. 4. Arrhenius plots of the reduction and oxidation speeds. Slopes of the transient response of the data in Fig. 3 and of additional data were normalized by 1 (mS/s) and their logarithmic value is plotted. Fig. 5. Transient responses to dry and humid (1% H 2 /99% N 2 ) gases with repeated exposures. Relative humidity of humid gas was 85% RH. Fig. 6. A calibration curve for humid gas. Balance gas was humid air and conductance values 5 min after the exposure to hydrogen was plotted. H. Nakagawa et al. / Sensors and Actuators B 93 (2003) 468–474 471 However, there are arguments that hydrogen does not form a bronze as the other alkaline metals such as lithium and sodium do [17]. Gerard et al. [18] observed no increase of hydrogen on coloration on their evaporated film colorization by the nuclear reaction analysis. They observed some hydro- gen increase in WO 3 upon colorization in their sputtered film, but hydrogen content did not decreased upon disco- loration. Wagner et al. [19] used the same technique and reported that the hydrogen concentration did not change in WO 3 films during electrochemical reduction and oxidation. Their WO 3 film was, however, inserted between ITO elec- trode and SiO 2 or Ta 2 O 5 material. In their film, hydrogen might not be able to move in or out easily, and electro- chemical oxidation or reduction within film might have taken place. They have observed the increase and decrease of hydrogen contents by gas reduction/oxidation in their earlier paper [20]. The accumulation of hydrogen in the film with repeated exposure was observed in their paper. Further several works on WO 3 film based on vacuum deposition reported transparent films despite oxygen defi- ciency [21,22]. Zhang et al. [23] assumed the existence of W 4þ and postulated that the polaron transition takes place between W 4þ and W 5þ ions, not between W 5þ and W 6þ as widely believed. Later, Lee et al. [24] established a new model in which both W 4þ –W 5þ and W 5þ –W 6þ transitions contribute to polaron transitions, but the coloration effi- ciency of the former is larger based on the Raman spectro- scopy. This model seems to reconcile previous contradicting observations in terms of tungsten valency. Apart from these studies, Georg et al. [7,25,26] published several works based on oxygen deficiencies instead of hydrogen intercalation. Lattice oxygen is reduced to water by hydrogen and removed from the film in their model and effect of water was explained. Recent work by Lee et al. [27] denied the creation of oxygen vacancies from their Raman study. Genin et al. [28] investigated the crystal structures of various structures and reported that structure changes to cubic symmetry with increasing lattice constants as amounts of hydrogen is increased. This observation supports hydrogen bronze model, rather than oxygen deficiency model. None of the reported model seemed to explain our observed data as well as observation of other researchers, a new scheme, double injection/surface oxidation model, is suggested to interpret these phenomena in a unified way. Since the present sensor was fabricated by a sol–gel method and high-temperature calcination in air, the film was likely oxidized completely. To exclude possible implications asso- ciated with oxygen deficiencies, we limit the discussion to the gas reduction/oxidation of WO 3 . The basic concept is the formation of tungsten bronze with the intercalation of hydrogen ions. But the oxygen removal by hydrogen reduc- tion and water formation on (1 0 0) plain of ReO 3 crystal structure [29] is included. Further, it was assumed that an intercalated proton is removed as a form of water from the film on oxidation. Therefore, reduction and oxidation take different path and they are not reversible processes. This is in accordance with our observed data of in Fig. 3 where shapes of rising transient and falling transient have quite different nature. Different values of the reduction and oxidation velocities and their temperature dependence in Fig. 4 further support this argument. Two-dimensional view of our model crystal was presented in Fig. 7. Here, we treat the phenomena within a single crystal to avoid the implications associated with grain boundaries. It Fig. 7. A mechanism model for the hydrogen reduction. 472 H. Nakagawa et al. / Sensors and Actuators B 93 (2003) 468–474 should be noted that the surface of the nanocrystal may not be necessarily a film surface. But most of amorphous WO 3 films are porous and whether the model surface is real film surface, or inner crystal grain boundary of the film, is irrelevant. Arrows schematically guide the reduction processes. They may be summarized as following three processes: (1) Dissociation of H 2 on Pt and spill-over to WO 3 surface (double injection). (2) Surface Diffusion of a proton and a parallel slow water formation reaction on (1 0 0) plane. (3) Internal diffusion of a proton from a surface through favorable plane or sites. The first process may not need arguments, since both double injection model [3] and oxygen deficiency model [7] accepted spill-over phenomenon. The second process may need some explanation. It had been known that the existence of water molecules increased the hydrogen reduction pro- cess considerably and the increase was interpreted as the acceleration of hydrogen diffusion by water molecule [30]. Our observation of the sensitivity increase with repeated exposure to dry (1% H 2 ) gas in Fig. 5 suggested the water generation. Henrich and Cox [29] mentioned that hydrogen atoms cannot diffuse into the bulk on (1 0 0), but slowly react with surface oxygen to form water. The water gener- ated on the surface of internal nanocrystal may stay for prolonged period since the water has to pass through rela- tively long crevasse-like narrow grain boundaries before leaving the film. Their (1 0 0) plane would mean equivalent (1 0 0) of cubic ReO 3 structure, not of monoclinic. The accumulation of generated water is responsible for the sensitivity increase in repeated exposure. Protons that dif- fused out of (1 0 0) plane may now diffuse into bulk as stated in the third process. Of course there should be some pre- ference in surface orientations for the easiness of bulk diffusion and surface lattice defects or kinks may be a preferential site for the start of bulk diffusion, but the topic is out of scope of the present work. The phenomena may be summarized by a familiar che- mical equation with explicit valence of double injection model: W 6þ O 3 2À þ xH þ þ xe À ! H x þ W 1Àx 6þ W x 5þ O 3 2À : (1) The surface reaction may be expressed as: H þ þ W 6þ O 3 2À þ e À ! 1 2 H 2 þ O 2À þ W 5þ O 2:5 2À : (2) This reaction creates the oxygen lattice defect which might cause some adverse effects on the sensing characteristics. The water formation reaction (2), however, is limited at the outermost layer and the reaction is reported to be slow. Furthermore, we placed Pt catalyst on (1 0 0) surface for the explanation purpose, but the tungsten film consists from many nanocrystals with randomly placed Pt catalyst parti- cles, and only limited portion of catalyst particles were placed on (1 0 0) surfaces. Therefore, implication of the oxygen reaction, except water generation, may be neglected in first order approximation. The oxidation mechanism of a reduced WO 3 is different from the standard double injection model. Instead of hydro- gen leaving the film as hydrogen molecules, it is oxidized at the nanocrystal surface. The oxidation processes may be summarized as: (1) Dissociation of O 2 on Pt and spill-over. (2) Surface diffusion of an O 2À ion and simultaneous bulk diffusion of a proton to a surface. (3) Hydrogen oxidation at a surface by a O 2À ion. The first process is similar to the hydrogen injection process. Oxygen molecules are dissociated at Pt surface and oxygen ion (O 2À ) diffuses to the WO 3 surface with removal of two electrons from WO 3 bulk. Oxygen ions at the surface attract the intercalated hydrogen atoms and the hydrogen atoms diffuse to the surface and form water at the surface. This water may stayed for prolonged time due to the crevasse-like grain boundary nature as explained for the water creation on the (1 0 0) surface associated with the hydrogen reduction process. We would not deny the possi- bility of bulk diffusion of oxygen ions and water formation within bulk, but the probability of this reaction could be small due to the molecular size difference. There is also good possibility for water molecules to diffuse into the bulk, although the diffusion velocity may be small. In any case, the chemical equation with explicit valence may be expressed: 1 2 xO 2À þ H x þ W 1Àx 6þ W x 5þ O 3 2À ! 1 2 xH 2 þ O 2À þ W 6þ O 3 2À þ xe À : (3) There may be various other irregular processes, such as hydroxyl adsorption at some kinks or steps. But those discussions are out of scope in the present analysis and it may be emphasized that the present model explains our observations and previous data in a unified manner. 4. Conclusion A sensitive hydrogen sensor was fabricated by a sol–gel method and characterized. The sensor exhibited high sensi- tivity with six orders of conductance increase upon 1% H 2 detection. The effect of water was found to be large with some memory effects. The existing theories failed to inter- pret the observed data. The new, double injection/surface oxidation model is suggested to explain the observed data as well as previously reported data. The reduction mechanism is based on the double injection model with the addition of surface oxygen reaction on (1 0 0) crystal surfaces. Oxygen reaction generates surface water and this accelerates the sensor response, and super-linear characteristics would be observed. The oxidation was achieved by the water forma- tion of dissociated oxygen ions and intercalated hydrogen ions at the nanocrystal surfaces. The sensitivity increase in H. Nakagawa et al. / Sensors and Actuators B 93 (2003) 468–474 473 the repeated exposure to dry H 2 gas may be attributed to the trapped water. The suggested model successfully explained observed data. References [1] P.J. Shaver, Activated tungsten oxide gas detectors, Appl. Phys. Lett. 11 (1967) 255–257. [2] S. Sekimoto, H. Nakagawa, S. Okazaki, K. Fukuda, S. Asakura, T. Shigemori, S. Takahashi, A fiber-optic evanescent-wave hydrogen gas sensor using palladium-supported tungsten oxide, Sens. Actua- tors B 66 (2000) 142–145. [3] B.W. Faughnan, R.S. Crandall, P.M. Heyman, Electrochromism in WO 3 amorphous films, RCA Rev. 36 (1975) 177–197. [4] M. Zayat, R. Reisfeld, H. Minti, B. Orel, F. Svegl, Gasochromic effect in platinum doped tungsten trioxide films prepared by the sol– gel method, J. Sol Gel Sci. Technol. 11 (1998) 161–168. [5] S.K. Deb, Optical and photoelectric properties and colour centers in thin films of tungsten oxide, Philos. Mag. 27 (1980) 801–822. [6] J G. Zhang, D.K. Benson, C.E. Tracy, S.K. Deb, A.W. Czanderna, C. Bechinger, Chromic mechanism in amorphous WO 3 films, J. Electrochem. Soc. 144 (1997) 2022–2026. [7] A. Georg, W. Graf, R. Neumann, V. Wittwer, Mechanism of the gasochromic coloration of porous WO 3 films, Solid State Ionics 127 (2000) 319–328. [8] E. Salje, K. Viswanathan, Physical properties and phase transitions in WO 3 , Acta Crystallogr. A 31 (1975) 356–359. [9] B. Gerand, G. Nowogrocki, J. Guenot, M. Figlarz, Structure study of new hexagonal form of tungsten trioxide, J. Solid State Chem. 29 (1979) 429. [10] A. Coucou, M. Figlarz, A new tungsten oxide with 3D tunnels: WO 3 with the pyrochlore-type structure, Solid State Ionics 28–30 (1988) 1762–1765. [11] R. Diehl, G. Brandt, E. Salje, The crystal structure of triclinic WO 3 , Acta Crystallogr. B 34 (1978) 1105–1111. [12] J.T. Szymanski, A.C. Roberts, The crystal structure of tungstite, WO 3 ÁH 2 O, Can. Mineral. 22 (1984) 681–688. [13] M.F. Daniel, B. Desbat, J.G. Lassegues, B. Gerand, M. Figlarz, Infrared and Raman study of WO 3 tungsten trioxide and WO 3 ÁxH 2 O tungsten trioxide hydrates, J. Solid State Chem. 67 (1987) 235–247. [14] N.N. Greenwood, A. Earnshaw, Chemistry of Elements, Butter- worths–Heinemann, London, 1997, p. 1008. [15] S. Okazaki, H. Nakagawa, S. Asakura, Y. Tomiuchi, N. Tsuji, H. Murayama, M. Yashiya, Sensing characteristics of an optical fiber sensor for hydrogen leak 93 (2003) 142–147. [16] O.F. Schirmer, V. Wittwer, G. Bauer, G. Brandt, Dependence of WO 3 electrochromic absorption on crystallinity, J. Electrochem. Soc. 124 (1977) 749–753. [17] R.S. Crandall, B.W. Faughnan, Comment on the cluster model of alkaline-metal tungsten bronzes, Phys. Rev. B 16 (1977) 1750–1752. [18] P. Gerard, A. Deneuville, R. Courths, Characterization of a-‘‘WO 3 ’’ thin films before and after coloration, Thin Solid Films 71 (1980) 221–236. [19] W. Wagner, F. Rauch, C. Ottermann, K. Bange, Hydrogen dynamics in electrochromic multilayer systems investigated by the 15 N technique, Nucl. Instrum. Methods B 50 (1990) 27–30. [20] F. Rauch, W. Wagner, K. Bange, Nuclear reaction analysis of optically active coating on glass, Nucl. Instrum. Methods B 50 (1990); F. Rauch, W. Wagner, K. Bange, Nuclear reaction analysis of optically active coating on glass, Nucl. Instrum. Methods B 42 (1989) 264–267. [21] H. Morita, H. Washida, Electrochromism of atmospheric evaporated tungsten oxide films (AETOF), Jpn. J. Appl. Phys. 23 (1984) 754–759. [22] C. Bechinger, M.S. Burdis, J G. Zhang, Comparison between electrochromic and photochromic coloration efficiency of tungsten oxide thin films, Solid State Commun. 101 (1997) 753–756. [23] J G. Zhang, D.K. Benson, C.E. Tracy, S.K. Deb, A.W. Czanderna, C. Bechinger, Chromic mechanism in amorphous WO 3 films, J. Electrochem. Soc. 144 (1997) 2022–2026. [24] S.H. Lee, H.M. Cheong, C.E. Tracy, A. Mascarenhas, D.K. Benson, S.K. Deb, Raman spectroscopic studies of electrochromic a-WO 3 , Electrochim. Acta 44 (1999) 3111–3115. [25] A. Georg, W. Graf, R. Neumann, V. Wittwer, The role of water in gasochromic WO 3 films, Thin Solid Films 384 (2001) 269–275. [26] A. Georg, W. Graf, V. Wittwer, The gasochromic coloration of sputtered WO 3 films with a low water content, Electrochim. Acta 46 (2001) 2001–2005. [27] S.H. Lee, H.M. Cheong, P. Liu, D. Smith, C.E. Tracy, A. Mascarenhas, J.R. Pitts, S.K. Deb, Raman spectroscopic studies of gasochromic a-WO 3 thin films, Electrochim. Acta 46 (2001) 1995–1999. [28] C. Genin, A. Driouuiche, B. Gerand, M. Figalarz, Hydrogen bronzes of new oxides of the WO 3 –MoO 3 system with hexagonal, Solid State Ionics 53–56 (1992) 315–323. [29] V.E. Henrich, P.A. Cox, The Surface Science of Metal Oxides, Cambridge University Press, Cambridge, 1994, p. 338. [30] J.E. Benson, H.W. Kohn, M. Boudart, On the reduction of tungsten accelerated by platinum and water, J. Catal. 5 (1966) 307–313. Biographies Hidemoto Nakagawa obtained his Masters degree in applied science and PhD degree from University of Toronto in 1974 and 1978, respectively. He had been a visiting associate professor at Yokohama National University from 1993 to 2001. He joined the Research Center for Advanced Science and Technology, University of Tokyo as visiting researcher in 2002. His research centers on various chemical and biological sensors and their applications. Nanako Yamamoto received her BEng in 2000 from Yokohama National University. And she obtained her MEng degree in 2002. Her research interest focuses gas sensors as well as medical sensors. Shinji Okazaki received his BEng and MEng degrees from Yokohama National University in 1991 and 1993, respectively. He joined Yokohama National University as a research associate in 1997. His major fields are electrochemistry and sensor engineering. Tsuneo Chinzei obtained his MD degree from Faculty of Medicine, University of Tokyo in 1982. He enrolled at Graduate School of Medicine, University of Tokyo in 1984 and became a research associate at the Research Center for Advanced Science and Technology, University of Tokyo in 1987 and promoted to an associate professor in 1999. He is specializing in artificial hearts and medical thermography. He is also interested in micromachining and medical sensors. Shukuji Asakura received his MEng and PhD degrees from University of Tokyo in 1965 and 1968, respectively. In 1972, he joined Yokohama National University, and became a professor in 1988. His fields of interest are safety engineering, corrosion science and chemical sensors. 474 H. Nakagawa et al. / Sensors and Actuators B 93 (2003) 468–474 . A room-temperature operated hydrogen leak sensor H. Nakagawa a, * , N. Yamamoto b , S. Okazaki b , T. Chinzei a , S. Asakura b a Research Center for Advanced. Butter- worths–Heinemann, London, 1997, p. 1008. [15] S. Okazaki, H. Nakagawa, S. Asakura, Y. Tomiuchi, N. Tsuji, H. Murayama, M. Yashiya, Sensing characteristics of an