preperation and characterization of sno2 and wox - sno2

Preparation and characterization of SnO 2 and WO x ±SnO 2 nanosized powders and thick ®lms for gas sensing A. Chiorino a , G. Ghiotti a,* , F. Prinetto a , M.C. Carotta b , C. Malagu Á b , G. Martinelli b a Dip. di Chimica, I.F.M. Universita Á di Torino, Via P. Giuria 7, 10125 Torino, Italy b INFM-Dip. di Fisica, Universita Á di Ferrara, Via Paradiso 12, 44100 Ferrara, Italy Abstract SnO 2 powders pure and added with W 6 at two different loadings (1 and 5 W mol%), were prepared via a sol±gel route. Thick ®lms prepared from the powders were used as CO and NO 2 gas sensors. The morphology of the powders was analyzed by TEM, HRTEM and that of ®lms by SEM. The goal of obtaining powders and ®lms made by nanosized particles, even after thermal treatments at 8508C was attained. The effect of W on the response of powders and ®lms towards CO and NO 2 was studied by FT-IR and conductance measurements, respectively. W markedly lowered the response of SnO 2 towards CO and markedly enhanced its ability to sense NO 2 . Surface species formed by CO and NO 2 interaction were investigated. # 2001 Elsevier Science B.V. All rights reserved. Keywords: Gas sensors; WO x ±SnO 2 ; CO; NO 2 1. Introduction Research on gas sensors is aimed at obtaining new materials to achieve highly sensitive and selective devices. Grain size reduction is one of the main factors enhancing the gas sensing properties of semiconducting oxides, while addition of different catalysts has been demonstrated to increase the sensitivity towards speci®c gases. Preparation of oxidic powders via sol±gel methods has been proved to be one of the best method to obtain nanosized materials. The aim of this work is to obtain nanosized WO x ±SnO 2 materials and to test them as CO and NO 2 sensors. Indeed, WO 3 -doped SnO 2 [1] or Sn x WO 3x mixed oxides [2] have been proposed as materials for gas sensors, on the basis of the well-known outstanding sensitivity of WO 3 to NO x [3]. 2. Experimental 2.1. Powders Pure SnO 2 was prepared by a sol±gel process previously described [4]. The resulting gel was dried overnight at 958C, giving a powder subsequently calcined at 5508C, thereafter named SN. The W 6 added materials were prepared by impregnation of SN powder with given amounts of aqueous solutions of ammonium tungstate (Merck) to obtain two different W nominal loadings: 1 and 5 mol%. The impreg- nated powders, dried 3 h at 1208C, are thereafter named W1 and W5. TEM and HRTEM analysis of SN, W1 and W5 powders was performed with a Jeol 2000 EX electron microscope equipped with a top entry stage. The powders were compressed in self-supporting pellets of about 50 mg cm À2 , and put in an IR cell, which allowed thermal treatments in vacuo and in controlled atmospheres. They were initially submitted to alternate outgassing-oxi- dizing treatments in dry oxygen at 650 or 8508C, then cooled in oxygen at the chosen temperature (room temperature (RT), 150 or 3508C). FT-IR spectra were run before in O 2 , and then in CO, NO 2 , CO/O 2 or NO 2 /O 2 mixtures on a FT-IR spectrophotometer (Perkin-Elmer System 2000), working in transmission/absorption mode, with resolution of 2cm À1 . The ratio NO 2 /O 2 and CO/O 2 of the mixtures used was 1/5. High purity gases (from Praxair) were used: O 2 and CO without further puri®cation, while NO 2 was prepared in laboratory, by contacting NO, freshly distilled, with O 2 during 4 weeks at RT. The outgassing-oxidizing pre-treat- ment temperature is thereafter reported in the pellet label (i.e. SN-650, W1-650, etc.). BET surface areas of SN, SN-650, SN-850, W1-650 and W5-650 pellets were determined by N 2 physisorption using a Micromeritics ASAP (10 À4 Pa) apparatus. Sensors and Actuators B 78 (2001) 89±97 * Corresponding author. Tel.: 39-11-670-7539; fax: 39-11-670-7855. E-mail address: ghiotti@ch.unito.it (G. Ghiotti). 0925-4005/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved. PII: S 0925-4005(01)00795-X 2.2. Films The sensors were obtained starting from miniaturized laser pre-cut 96% alumina substrates (2:5mm 2:5mm Â0:25 mm for each device) provided with a heater element on the backside, a Pt-100 resistor for the control of the sensor operating temperature and a gold front interdigitated con- tacts. The thick ®lms were prepared starting from pastes obtained by adding to the above powders an organic vehicle together with a small percentage of glass fritt for improving the adhesion of the layers to the substrates. The ®lms were then ®red for 1 h at 650 or 8508C in air. The thickness of the deposited layers was in the 20±30 mm. The conductance measurements were performed in a closed test chamber at a ¯ow rate of 0.5 l min À1 , at temperature up to 4008C in wet (40% RH) air and in 100 ppm of CO or in 10 ppm of NO 2 in wet air. The sensors were biased at 5 V constant voltage and the conductance was obtained measuring the voltage drop on calibrated resistors. Six sample series have been tested, obtained from the SN, W1 and W5 powders ®red at 650 and 8508C. For sake of brevity, the labels used for powder pellets were also used for the ®red thick ®lms (SN-650, W1- 650, etc.). The morphology of all ®lms ®red at 8508C was analyzed by a 360-Cambridge scanning electron microscope. 3. Results and discussion 3.1. Morphology and texture TEM images of SN powder showed aggregates of parti- cles, rounded shaped and homogeneous in size (diameter 10±20 nm). W1 and W5 samples showed more densely packed aggregates of particles with very indented borders and homogeneous in size (20 nm). HRTEM images of the three powders showed particles with fringe distances of the cassiterite (1 1 0) and (1 0 1) planes. However, small spots of 0.5±1 nm in size could be seen on some particles of W5 sample. They could be due to the presence of small aggre- gates of tungsten-oxide units, even if too small and/or too poorly crystalline to give speci®c re¯exes in the electron diffraction pattern. In Fig. 1a and b HRTEM images of SN and W5 powders, respectively, are shown. On powders treated at 650 and 8508C no HRTEM images were done. Only scanning electron microscopies of SN-850, W1-850 and W5-850 ®lms were performed. SN-850 ®lms showed densely packed aggregates of rounded particles homogeneous in size (diameter 40±60 nm) (see Fig. 2a). At variance, W1-850 and W5-850 ®lms showed aggregates of particles with irregular shape, less homogeneous in size, their diameter ranging between 40 and 90 nm (see Fig. 2b and c). SN, SN-650 and SN-850 powders showed speci®c surface areas of 31, 31 and 17 m 2 g À1 , respectively. W1-650 and W5-650 pellets showed speci®c surface area of 25 and 27 m 2 g À1 , respectively. 3.2. FT-IR characterization The FT-IR spectra of both W1-650/850 and W5-650/850 in dry O 2 differed from those of SN-650/850 by two main features: (i) the overall IR transmittance was higher after W addition, as shown in Fig. 3 for SN-650 and W1-650 samples and (ii) new peaks, of medium and low intensity in the low frequency side of the spectrum (1150±780 cm À1 region) appeared. These peaks are well evident in the inset of Fig. 3, where the spectra in absorbance of SN-650, W1- 650 and W5-650 samples are compared (curves 1±3, respec- tively) in the more restricted 1150±780 cm À1 region. The increase in transmittance can be easily interpreted as due to a decreased concentration of free electrons and of electrons trapped in oxygen vacancies, the W 6 ions acting evidently as deeper acceptor levels. The new peaks, whose intensity increased with W loading, fall in a region where the vibra- tion modes of W=O, (W±O±W) n and (Sn±O±W) n are expected to fall [5±7]. Concerning the CO interaction at RT with the materials, the results can be resumed as follows: 1. The response to pure CO of the SN-650/850 pellets was that already reported in previous papers for SnO 2 prepared in different way [4,8]: CO reacted with oxygen surface ions (O 2 À ,O À ,O 2À ) giving carbonate-like species, which partially decomposed to CO 2 releasing electrons to the tin oxide; the electrons released repopulated the conduction band (CB) and the oxygen vacancies (V O 2 ). The electronic repopulation of the V O 2 could be revealed by the intensity increase of a very broad absorption extending all over near and medium IR regions, previously assigned to the photo- ionization of the monoionized oxygen vacancies, V O   hn 3 V O 2  e À (CB) [7±9]. The surface re- duction increased increasing the contact times and/or the equilibrium pressure. Taking the integrated intensity of the broad electronic absorption as a measure of the material response to CO, the SN-650 pellet appeared more sensitive than the SN-850 one. 2. As for the W-added materials, W1-650/850 and W5- 650/850 showed a behavior similar to that of SN-650/ 850 ones. However, within the same pre-treatment temperature, SN appeared the most sensitive material, the sensitivity being progressively lower increasing the W loading. Furthermore, the CO interaction caused the intensity decrease of the peaks assigned to tungstenyls (W=O) groups, revealing they were at the surface. The behavior above described is illustrated in Fig. 4a±c for SN-650, W1-650 and W5-650 materials (taken as an example): the reported curves are the differences between absorption spectra after and before the interaction with CO at different contact times and/or equilibrium pressures and they clearly show the increase of the very broad absorption previously discussed. The species formed at the surface were 90 A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 also detected, the sharp peaks superimposed to the broad electronic absorption are, indeed, the vibration modes of the surface species formed by CO interaction. On all materials a weak band at 2203 cm À1 is present, assigned to CO linearly bound to coordinatively unsaturated surface Sn 4 ions. The various peaks observed in the 1800±1000 cm À1 spectral region could be assigned to different surface carbonate-like species formed by the reductive adsorption of CO on different sites. The main species formed, in absolute and relative amounts depending on the materials and on the Fig. 1. HRTEM images of (a) SN and (b) W5 powders. A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 91 pre-treatment temperatures, were as follows: (i) chelating and bridging bidentates carbonates (n(C=O) at 1650± 1550 cm À1 , n asym (OCO) and n sym (OCO) at 1270 and 1020 cm À1 , respectively), (ii) ``strongly perturbed CO 2 '' or ``carboxylate species'' (n asym (OCO) and n sym (OCO) at 1750±1730 and 1270 cm À1 , respectively) and (iii) CO 2 linearly coordinated (n asym (OCO) at 2330 cm À1 ). On W1 and W5 the negative peaks at 1040±1030 cm À1 account for the erosion of the (W=O) surface groups. At 150 or 3508C in CO/O 2 mixture all the samples responded to CO. However, while it was possible to compare the spectroscopic response at 1508C of the different materi- als, the SN-650/850 being again the most active samples, this was not possible at 3508C because they completely loose their transparency to the radiation. We could only conclude that at 3508C all materials better respond to CO than at 1508C. No surface species could be detected at 1508C, and obviously, at 3508C. As for NO 2 or NO 2 /O 2 mixture, we only examined the interaction with SN-650, W1-650 and W5-650 pellets. Concerning interaction at RT we examined the response to 5 mbar of NO 2 , following the spectra evolution with the contact time. The results can be resumed as follows: 1. By interaction with SN-650 pellet, a quick reduction process at the surface immediately occurred with electron release to the CB and the subsequent V O 2 repopulation to V O  , revealed by the quick increase of the very broad absorption already described extending all over the near and medium IR regions (see Fig. 5a), which reached its maximum intensity in 2 s. Sharp vibrational bands appeared, but of very weak intensity, the most intense being a peak at 1178 cm À1 , assignable to surface n asym (ONO) mode of chelating nitrites. At higher contact time, the electron population of the V O  Fig. 2. SEM images of (a) SN-850, (b) W1-850 and (c) W5-850 films. Fig. 3. FT-IR transmittance spectra of SN-650 (solid line) and W1-650 (dotted line) samples in dry oxygen. Inset: FT-IR absorbance spectra of SN-650, W1- 650 and W5-650 samples (curves 1±3, respectively) in the 1150±750 cm À1 region; the spectra have been translated along the absorbance axis to allow a better comparison. 92 A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 started to decrease revealing an oxidation process with a kinetic slower than the reductive one. The electron population decrease was paralleled by the growth of sharp peaks related to the formation a variety of NO x surface species: chelating nitrites (n asym (ONO) mode at 1178 cm À1 ), several types of chelating and bridging nitrates (n(N=O) modes in the range 1700±1500 cm À1 , n asym (ONO) modes in the range 1250±1180 cm À1 and n sym (ONO) modes in the range 1000±900 cm À1 , respec- tively), NO dÀ and NO 2 d weakly adsorbed species (bands in the range 2050±1800 cm À1 ). When the reactions stopped after 1 min, the sample was slightly oxidized, that is the electron concentration in the V O  was lower than before the NO 2 admission: in fact, taking as reference the dotted line representing the line of absorbance zero in Fig. 5a, a broad, weak, and negative absorption is evident, showing the same shape of the electronic absorption assigned to V O   hn 3 V O 2  e À (CB). 2. At variance, W1-650 pellet was immediately oxidized by NO 2 (see Fig. 5b), with decrease of the electron population of the V O  : a broad, weak, and negative absorption is already evident after a contact time of 2 s, at the same time weak vibrational peaks related to the formation of a variety of NO x surface species appeared. At higher contact times the electronic population remained unchanged, while the vibrational peaks related to the NO x surface species continued to increase during 1 min. The surface species formed were similar to those present at the surface of the SN-650 material but in different amounts. In particular chelating nitrites, NO dÀ and NO 2 d weakly adsorbed species were markedly Fig. 4. Effects of CO at RT on FT-IR spectra of (a) SN-650, (b) W1-650 and (c) W5-650 samples. (a) Equilibrium pressures: <10 À2 mbar, 1  10 À2 mbar and 2 or 7 mbar (curves 1±3, respectively). (b) and (c) Equilibrium pressures: 0.1, 2 and 7 mbar (curves 1±3, respectively). The dotted line represents the zero of absorbance. Fig. 5. Time resolved effects of 5 mbar of NO 2 at RT on FT-IR spectra of (a) SN-650, (b) W1-650 and (c) W5-650 pellets. Spectra after 2 s, 10 s and 1 min (curves 1±3, respectively). The dotted line represents the zero of absorbance. A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 93 decreased. Chelating and bridging nitrates were the species predominant at the surface, their vibration frequencies being slightly different from those present on SN-650, indicating that tungsten ions were involved in the coordination of these species. Concerning W5- 650 sample (Fig. 5c), again the electron population of the V O  started to decrease immediately after NO 2 interaction, revealing oxidation processes, that stopped after 10 s. Also in this case the electron concentration decrease was paralleled by the growth of sharp peaks related to the formation of NO x surface species. However, chelating nitrites and NO dÀ or NO 2 d weakly adsorbed species were completely absent, only one type of bridging nitrate was still present (n(N=O) mode at 1700 cm À1 , n asym (ONO) modes at 1300 cm À1 and n sym (ONO) at 1000 cm À1 , respectively), while the main surface species were nitrate of ionic types not detected on the other two materials (n asym (ONO) mode at 1450 cm À1 and n sym (O- NO) mode at 1370 cm À1 ). This indicates that the surface chemistry was markedly changed. At the equilibrium (after 1 min), the values of the inte- grated negative intensity of the broad electronic absorption were in the following order: W5-650 < W1-650 < SN-650 sample (compare the intensities of the broad negative absorption of the curve 3 in Fig. 5). Taking this integrated negative intensity as a measure of the electronic response to NO 2 , W5 appeared the most sensitive one, the sensitivity decreasing with the W loading. Concerning the NO 2 /O 2 interaction at 150 or 3508C, we only studied SN-650 and W5-650 pellets. By interaction with the gas mixture both the materials were quickly oxi- dized, the oxidation degree being higher for the W5-650 than for SN-650. The highest oxidation degree was obtained at 1508C. At this temperature, taking the integrated negative intensity of the broad electronic absorption as a measure of the material response, W5 appeared more sensitive than SN material (see Fig. 6). At variance, no marked difference could be detected between the two samples at 3508 C. 3.3. Electrical characterization Carbon monoxide and nitrogen dioxide are two of the most dangerous gases polluting the atmosphere in the urban areas. For obtaining actual responses to CO and NO 2 ,we tested the sensors in conditions as much as possible close to the operating ones. With this aim in view, the in¯uence of tungsten addition onto the electrical properties of tin oxide as host material was investigated performing conductance measurements as a function of temperature in wet air at 40% RH. Fig. 7 shows the Arrhenius plots of pure and tungsten- added SnO 2 thick ®lms ®red at 8508C (SN-850, W1-850 and W5-850) obtained changing the temperature from 100 up to 5008C at the heating rate of 3 K min À1 . The samples exhib- ited three regions of conductivity; this behavior is quite usual in thick ®lm sensors of SnO 2 differently synthesized as reported in previous works [10,11]. Moreover, an addition of 5% of tungsten caused a signi®cant decrease in conductance, till about one order of magnitude at the temperature of 3008C; besides the W1-850 samples showed a behavior in between the SN-850 and W5-850 ones. All the electrical characteristics investigated showed the same trend being in very good agreement with FT-IR analyses. At the same time the W loading increased the intergrain energy barriers as a function of temperature, as it is shown in Fig. 8 (the experimental method to determine them is described in [12]). Since the difference between the mini- mum and the maximum values of the energy barrier (DE)is lower in the case of W-added materials, the above result cannot be associated with a greater amount of oxygen ions Fig. 6. Effects of 5 mbar of the NO 2 /O 2 mixture at 1508C on FT-IR spectra of SN-650 and W5-650 pellets (curves 1 and 2, respectively). The dotted line represents the zero of absorbance. Fig. 7. Temperature dependence of the conductance in wet (40% RH) air of SN-850, W1-850 and W5-850 thick films heated to 8508C. 94 A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 adsorbed which is at contrary expected to increase DE [13]. In fact from the Fig. 8 it is possible to evaluate DE as 0.15 eV in the case of SN-850, 0.13 eV for W1-850 and ®nally 0.11 eV for W5-850. Therefore, the decreasing conductance has to be attributed to the increased number of holes in the valence band injected by the acceptor-like W ions. As suggested by the behavior of the energy barrier, the temperature interval in which the oxygen adsorption may occur is ranging from 150 up to 3608C, as well as the gas±surface reactions [12]. Moreover, in n-type semicon- ductor oxides as SnO 2 , at low temperature, when the energy barrier is low, we expect the oxidizing gases more reactive; an opposite behavior occurs in the case of reducing gases. According to the above observation, the SnO 2 and W- added SnO 2 -based ®lms were tested as gas sensor versus carbon monoxide and nitrogen dioxide in the 150±3508C range of temperature. In agreement with previous works of the same authors, the maximum response to NO 2 has been found at 1508C, whilst the CO presented its maximum response at about 3008C (see for instance [4]). Concerning the SN-850, W1-850, W5-850 samples, the tests showed that the ability of the thick ®lms to detect CO was lowered increasing the W loading, in agreement with the barrier energy behavior. In fact it has been demonstrated that DE is strongly related to the sensitivity to reducing gases, like carbon monoxide [13]. The W content also affected the temperature of the maximum response. In particular W addition stabilized the response at the different working temperatures: W5 thick ®lms showed a weak response almost unaltered in all the examined range (200±3508C). Concerning NO 2 , conductance measurements showed that the ability of the material to detect the gas was enhanced increasing the W loading. The enhancement was particularly marked at 2508C (three times more passing from SN-850 to W5-850). It is worthy to note that at this working tempera- ture the SN-850 samples showed the maximum response to CO, being more than three times higher than that of the W5 ®lms (see Fig. 9). Furthermore, all three materials showed the maximum response to NO 2 at 1508C, temperature at which the response of the three ®lms to CO was very low. Fig. 8. Temperature dependence of the barrier energy in wet (40% RH) air of SN-850, W1-850 and W5-850 thick films heated to 8508C. Fig. 9. Electrical response to (a) 100 ppm of CO (G gas /G air ) and (b) 10 ppm of NO 2 (G air /G gas ) in wet air (40% RH) of SN-850, W1-850 and W5-850 thick films, measured at 2508C. A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 95 The SN-650, W1-650, W5-650 samples behaved in the same way as the samples ®red at 8508C; the only difference was the temperature of the maximum response to CO (300 instead of 2508C). More generally the samples ®red at 6508C showed a higher response to the tested gases probably because of a greater speci®c surface area. As an example, the tests at the working temperatures of 3008C for 100 ppm of CO (Fig. 10) and at 2008C for 10 ppm of NO 2 (Fig. 11) in wet air (40% RH) are reported. We have shown these tests (300 instead of 2508C for CO and 200 instead of 1508C for NO 2 ) to highlight the differences between the behavior of various samples taking also in to account that higher temperatures are always to prefer both to improve the response time and to avoid poisoning phenomena. 4. Conclusions SEM, TEM and HRTEM microscopies show that the goal of obtaining powders and ®lms made by nanosized particles even after thermal treatments at 8508C, is attained. FT-IR spectroscopic and electrical measurements have been employed on powders and ®lms, respectively, for obtaining information on the electronic effect due to the tungsten addition. FT-IR results show that W addition increases the sample transmittance and this phenomenon has been attributed to a decreased concentration of free electrons and of electrons trapped in oxygen vacancies, the W 6 ions acting as deeper acceptor levels. According to this, the electrical data show that W markedly lowers the con- ductance of the ®lms in air. Electrical measurements show that W lowers the response of tin oxide to CO and also affects the temperature of the maximum response. At variance the W addition enhances its ability to sense NO 2 , leaving unaltered the temperature of the maximum response: 1508C. FT-IR transmission changes of the three materials, pre- treated either at 650 or 8508C, contacted with pure CO at RT or with CO/O 2 mixture at 150 and 3508C, are qualitatively in good agreement with the electrical measurements. The agreement between FT-IR and conductance measurements has been also con®rmed for NO 2 and NO 2 /O 2 mixtures, at least for the samples studied with both the techniques. FT-IR spectroscopy was also employed to obtain information on the nature of the surface species formed by interaction with the two examined gases. Acknowledgements Financial support was provided by the Italian CNR (Progetti Finalizzati MADESS II). References [1] D.H. Yun, C.H. Kwon, K. Lee, Abnormal current±voltage character- istic of WO 3 -doped SnO 2 semiconductors and their applications to gas sensors, Sens. Actuators B 35 (1996) 48±52. [2] J.L. Solis, V. Lannto, Gas sensing properties of Sn x WO 3x mixed oxide thick films, Sens. Actuators B 48 (1998) 322±326. [3] D S. Lee, S. Han, J. Huh, D D. Lee, Nitrogen oxides-sensing characteristics of WO 3 -based nanocrystalline thick film gas sensor, Sens. Actuators B 60 (1999) 57±63. [4] A. Chiorino, G. Ghiotti, F. Prinetto, M.C. Carotta, D. Gnani, G. Martinelli, Preparation and characterization of SnO 2 and MoO x ± SnO 2 nanosized powders for thick film gas sensors, Sens. Actuators B 58 (1999) 338±349. [5] J. Leyer, R. Margraf, E. Taglauer, H. Kno È zinger, Solid±solid wetting and formation of monolayers in supported oxide systems, Surf. Sci. 201 (1988) 603±623. Fig. 10. Electrical response to CO (100 ppm) in wet air (40% RH) of all samples measured at 3008C. Fig. 11. SN-850 is shown in the figure. Electrical response to NO 2 (10 ppm) in wet air (40% RH) of all samples measured at 2008C. 96 A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 [6] J.A. Horsley, I.E. Wachs, J.M. Brown, J.H. Via, F.D. Hardcastle, Structure of surface tungsten oxide species in the WO 3 /Al 2 O 3 supported oxide system from X-ray absorption near-edge spectro- scopy and Raman spectroscopy, J. Phys. Chem. 91 (1987) 4014. [7] I.E. Wachs, F.D. Hardcastle, S.S. Chan, Raman spectroscopy of supported metaloxide catalysts, Spectroscopy (Springfield) 1 (1986) 30. [8] G. Ghiotti, A. Chiorino, F. Boccuzzi, Infrared study of surface chemistry and electronic effects of different atmosphere on SnO 2 , Sens. Actuators B 19 (1989) 151±157. [9] A. Chiorino, G. Ghiotti, F. Prinetto, M.C. Carotta, G. Martinelli, Characterization of SnO 2 -based gas sensors. A spectroscopic and electrical study of thick films from commercial and laboratory samples, Sens. Actuators B 44 (1997) 474±482. [10] M.C. Carotta, C. Dallara, G. Martinelli, L. Passari, A. Camanzi, CH 4 thick-film gas sensors: characterisation method and theoretical explanation, Sens. Actuators B 3 (1991) 191±196. [11] G. Ghiotti, A. Chiorino, G. Martinelli, M.C. Carotta, Moisture effects on pure and Pd-doped SnO 2 thick films analyzed by FT-IR spectroscopy and conductance measurements, Sens. Actuators B 24/25 (1995) 520±524. [12] V. Lantto, P. Romppainen, S. Leppa È vuori, A study of the temperature dependence of the barrier energy in porous tin dioxide, Sens. Actuators B 14 (1988) 149±163. [13] G. Martinelli, M.C. Carotta, Sensitivity to reducing gas as a function of energy barrier in SnO 2 thick-film gas sensor, Sens. Actuators B 7 (1±3) (1992) 717±720. A. Chiorino et al. / Sensors and Actuators B 78 (2001) 89±97 97 . used as CO and NO 2 gas sensors. The morphology of the powders was analyzed by TEM, HRTEM and that of ®lms by SEM. The goal of obtaining powders and ®lms. area of 25 and 27 m 2 g À1 , respectively. 3.2. FT-IR characterization The FT-IR spectra of both W1-650/850 and W5-650/850 in dry O 2 differed from those of