crystalline structure, defects and gas sensor response

Crystalline structure, defects and gas sensor response to NO 2 and H 2 S of tungsten trioxide nanopowders I. Jime ´ nez * , J. Arbiol, G. Dezanneau, A. Cornet, J.R. Morante Departament d’Electro ` nica, Enginyeria i Materials Electro ` nics, Universitat de Barcelona, Barcelona 08028, Spain Abstract Structural and NO 2 and H 2 S gas-sensing properties of nanocrystalline WO 3 powders are analysed in this work. Sensor response of thick- film gas sensors was studied in dry and humid air. Interesting differences were found on the sensor response between sensors based on 400 and 700 8C-annealed WO 3 , what motivated a structural study of these materials. Crystalline structure and defects were characterised by X-ray diffraction (XRD), Raman spectroscopy and transmission electron microscopy (TEM). Experimental results showed that both triclinic and monoclinic structures are present in the analysed materials, although their amount depends on the annealing treatment. Crystalline shear planes, which are defects associated to oxygen deficient tungsten trioxide, were found in 400 8C-annealed WO 3 and their influence on XRD spectra was analysed by XRD simulations. Moreover, XRD and Raman spectra were also acquired at normal metal oxide-based gas sensor working temperatures in order to relate both crystalline structure and sensor response. # 2003 Elsevier Science B.V. All rights reserved. Keywords: WO 3 ; Gas sensor; Structural characterisation; NO 2 ;H 2 S 1. Introduction Tungsten oxide is nowadays considered as one of the most interesting materials in the field of gas sensors based on metal oxides, as it is shown by the increasing number of publications appeared in recent years. Very good results in the detection of NO 2 and H 2 S by sensors based on this material have been reported. Most of them concern WO 3 thin films obtained by physical routes such as sputtering [1,2] or thermal evaporation [3,4]. Besides, thick-films technologies based on the use of nanopowders have been also presented [5,6]. Powder is mixed with an organic vehicle to form a paste, which is usually deposited on a substrate as a thick sensitive film, although compatibility with micromachined gas sensors is also possible [7–10]. Since gas sensors based on metal oxides must usually work at temperatures ranging from 200 to 400 8C, the sensing material has to be previously stabilised at higher tempera- tures. This annealing step will strongly affect the structural properties of the nanocrystalline material and thus its gas- sensing properties too. Regarding WO 3 , the pyrolysis of ammonium paratung- state has been one of the most used routes to obtain this material as powder with nanometric grain size, being a well- known technique in the field of gas sensors [11,12]. In the same way, dehydration of tungstic acid has revealed as an interesting route to obtain WO 3 with low impurities con- centration [13,14], although it is not so usual in the field of gas sensors. Gas sensors based on WO 3 obtained by this route showed a better sensor response to NO 2 than that of pyrolytic WO 3 -based gas sensors under the same test con- ditions, combined with a low response to CO and CH 4 [15]. The aim of this work is to study the evolution of structural properties of WO 3 nanocrystalline powders as a function of annealing temperature in order to understand the sensor response to NO 2 and H 2 S of gas sensors based on differently annealed WO 3 . Thick-film gas sensors based on differently annealed WO 3 nanopowders (400 and 700 8C) were pre- pared and their gas-sensing properties towards NO 2 and H 2 S in dry and humid air were compared. The differences found in gas-sensing motivated a study of the crystalline properties and their evolution with annealing temperature. The com- bination of characterisation techniques such as X-ray dif- fraction (XRD), Raman spectroscopy and transmission electron microscopy (TEM) allows to precise the structure of the synthesised compounds and the nature of possible defects. Furthermore, XRD and Raman spectra were also acquired at normal working sensor temperatures for the detection of our target gases (between room temperature and 300 8C) in order to better relate structural and gas- sensing properties. Sensors and Actuators B 93 (2003) 475–485 * Corresponding author. Tel.: þ34-934-021-146; fax: þ34-934-021-148. E-mail address: ijimenez@el.ub.es (I. Jime ´ nez). 0925-4005/03/$ – see front matter # 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0925-4005(03)00198-9 2. Experimental WO 3 nanocrystalline powders were obtained by a soft chemistry route based on tungstic acid. Tungstic acid was dissolved in a 50:50 volumic mixture of methanol and water with a tungsten over water molar ratio of 25. This solution was heated at 80 8C for 24 h under stirring in air and dried by further heating at 110 8C in air, leading to tungsten oxide hydrate. This material was annealed in a furnace between 400 and 700 8C for 5 h under a flow of synthetic air to obtain nanocrystalline WO 3 . Gas sensors were obtained by screen-printing of a paste based on WO 3 over alumina substrates, which had already printed platinum electrodes on the front side and a platinum heater on the backside to control the operating temperature. These gas sensor devices were placed in a stainless steel test chamber (200 ml) where a controlled atmosphere was pro- vided by means of mass flow controllers connected to a computer. The sensor response was acquired for different concentrations of H 2 S and NO 2 in synthetic dry and humid air at a flow of 200 ml min À1 . Humidity was controlled by mixing the appropriate quantities of dry air with water- bubbling air, monitoring the relative humidity with a com- mercial capacitive humidity sensor. Gas sensor response was calculated as the resistance ratio R gas /R air for both gases. Operating temperature of the sensor devices was varied between 200 and 300 8C. Sensor response at lower tempera- tures was not studied in order to avoid too high sensor resistance, which may lead to not reliable sensor response measurements. XRD patterns of the nanopowders were obtained with a Siemens D-500 X-ray diffractometer using Cu Ka radiation, with operating voltage of 40 kV and current of 30 mA. Raman scattering measurements were obtained in back- scattering geometry with a Jobin-Yvon T64000 spectro- meter coupled to an Olympus metallographic microscope. Excitation was provided by an argon-ion laser operating at a wavelength of 488.0 nm with a low incident power to avoid thermal effects. Raman shifts were corrected by using silicon reference spectra after each measurement. Trans- mission electron microscopy was carried out on a Phillips CM30 SuperTwin electron microscope operated at 300 keV with 0.19 nm point resolution. For TEM observa- tions, WO 3 nanopowders were ultrasonically dispersed in ethanol and deposited on amorphous holey carbon membranes. 3. Results and discussion 3.1. Gas sensor response Fig. 1a shows sensor response to 1 ppm of NO 2 and 20 ppm of H 2 S in synthetic dry air of gas sensors based on 400 and 700 8C-annealed WO 3 as a function of sensor working temperature. In the range of working temperatures studied, from 200 to 300 8C, sensor response increased when temperature decreased for both gases. A similar behaviour has already been described for NO 2 and H 2 S detection by WO 3 [3,6], although their sensing mechanisms are comple- tely different. The detection of NO 2 is usually based on the formation of absorbed surface-trap states NO À 2 ads, whereas H 2 S molecules react with surface oxygen. In our case, it was found that WO 3 annealed at 700 8C had a higher sensor response to both gases than 400 8C-annealed WO 3 in this operation temperature range. Sensor response to NO 2 and H 2 S in wet air was analysed at fixed operation temperatures (200 8C for H 2 S and 225 8C for NO 2 detection). These temperatures were selected so as to achieve a compromise between sensor response and recovery time, which decrease when operating temperature increases in the case of NO 2 detection. Fig. 1b shows the Fig. 1. (a) Sensor response to H 2 S and NO 2 of gas sensors based on 400 and 700 8C-annealed WO 3 as functions of working temperature, (b) Sensor response to H 2 S (200 8C) and NO 2 (225 8C) of gas sensors based on 400 and 700 8C-annealed WO 3 as functions of relative humidity. 476 I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 dependence of the sensor response to NO 2 and H 2 Son humidity for 400 and 700 8C-annealed WO 3 , whereas Fig. 2a and b show a comparison of the dynamic sensor responses to these gases in dry and humidified air (50% of relative humidity). This sensor response was evaluated taking sensor resistance in dry or humidified synthetic air as the reference. Response and recovery times, for both gases, have a low dependence on humidity, as sensor response to NO 2 . On the other hand, sensor response to H 2 S is highly dependent on humidity, especially for 700 8C- annealed WO 3 . In fact, only sensor resistance in H 2 S atmo- sphere is highly dependent on humidity, as it will be shown later. According to sensor response and humidity depen- dency, 400 8C-annealed WO 3 was chosen for further H 2 S detection studies, whereas 700 8C-annealed WO 3 was cho- sen for NO 2 detection. Sensor responses of this materials to different concentrations of H 2 S(1–10 ppm) and NO 2 (0.2–2 ppm) under different humidified ambiences are shown in Fig. 3, which still clearly shows a greater influence of humidity on H 2 S detection than on NO 2 detection. Finally, in order to study this great influence of humidity on H 2 S sensor response, pulses of humidity (1 h) were introduced in atmospheres of synthetic air and H 2 S (2 ppm) in synthetic air, keeping the concentration of this gas constant. Fig. 4 shows the results for the gas sensor based on 400 8C-annealed WO 3 . This experimental procedure, the study of the dynamical response of metal oxide gas sensors to pulses of humidity in atmospheres containing a target gas, Fig. 2. Dynamic sensor response of WO 3 -based gas sensors to (a) 1ppm of NO 2 in dry and 50% relative humidity air, (b) 5 ppm of H 2 S in dry and 50% relative humidity air. Fig. 3. Sensor response to different concentrations of NO 2 (400 8C-annealed WO 3 ) and H 2 S (700 8C-annealed). I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 477 has revealed as a very useful method to investigate the interaction of not only water, but also that of the target gas with the sensing material [16]. When humidified air was introduced in a background of synthetic air, resistance decreased very fast and afterwards increased slowly, reach- ing a final value very close to the resistance value in dry synthetic air. Similar dynamic responses have been already described for SnO 2 ,In 2 O 3 and ZnO [17], although the reason is not completely understood yet. The decrease in sensor resistance was attributed to the dissociatively reaction of water with lattice oxygen, which leads to the formation of oxygen vacancies and so to a resistance decrease. The following slow resistance increase could be due to the recombination of the OH À ions with the lattice oxygen vacancies previously formed. As it is shown in our case, humidity has a low effect on base resistance in synthetic air in the case of these WO 3 -based gas sensors. However, it is interesting to notice that this behaviour is completely dif- ferent when the background is H 2 S in air. In this case, there is no fast decay of resistance but an important increase when humidity is introduced, which shows that water is probably competing with H 2 S molecules and preventing some oxygen ions from reacting with H 2 S. When dry air is reintroduced again, previous sensor resistance value is recovered, although response time is slower, probably due to a slow desorption of adsorbed water. Similar dynamic results were found for 700 8C-annealed WO 3 , although a greater influ- ence of humidity on resistance in the presence of H 2 S was found. Therefore, this may show not only that water is competing with H 2 S, but also that there are more than one reactive oxygen species participating on the detection of this gas, since 400 and 700 8C-annealed WO 3 are differ- ently affected. Therefore, the reactive site blocked by water would be much more abundant in 700 8C-annealed than in 400 8C-annealed WO 3 . 3.2. Structural characterisation Annealed powder was identified by XRD as nanocrystal- line WO 3 . Crystalline WO 3 presents a pseudo-cubic struc- ture with a slight distortion of the cubic ReO 3 -type lattice, where W atoms occupy the centre of oxygen octahedra linked by their corners. At room temperature, monoclinic and triclinic are the most common structures [18]. Mono- clinic is described in the P21/n space with cell parameters a ¼ 0:7301 nm, b ¼ 0:7539 nm, c ¼ 0:7689 nm and b ¼ 90:89  . The triclinic structure is described in the P-1 space group with cell parameters a ¼ 0:7310 nm, b ¼ 0:7524 nm, c ¼ 0:7686 nm, a ¼ 88:85  , b ¼ 90:91  , g ¼ 90:935  . Fig. 5 presents XRD patterns from 2y ¼ 20 to 408 of WO 3 samples annealed at different temperatures (400, 500, 600 and 700 8C). This figure also includes theoretical diffraction diagrams of the triclinic and monoclinic compounds. These diagrams have been calculated using the program FULL- PROF [19], taking the cell parameters and atomic positions of [18] and an artificial crystallite size of 50 nm. Due to the slight distortion of the lattice, the main reflection (1 0 0) of the ideal cubic cell splits in three in the range 20–308 [20]; (1 0 0), (0 1 0) and (0 0 1) pseudo-cubic reflections. Although these reflections are referred as (2 0 0), (0 2 0) and (0 0 2) if the monoclinic or triclinic unit cells are con- sidered, indexations will be referred herein to the pseudo- cubic representation with cell parameter a % 0:38 nm. It is clear from Fig. 5 these three main reflections can not be used to determine if the crystalline structure is triclinic or mono- clinic, since their position and relative intensity is very Fig. 4. Sensor resistance variation (400 8C-annealed WO 3 ) to pulses of humidity (30, 50 and 80% relative humidity) in a background of synthetic air and H 2 S (2 ppm) in synthetic air. 478 I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 similar for both structures when dealing with nanometric grain sizes. However, it is interesting to notice the evolution of the full width at half maximum (FWHM) and maximum intensity with annealing temperature of these three main peaks, shown in Fig. 6. Whereas relative intensities of the three main reflections should be almost identical, experi- mental spectra show that diffraction peak corresponding to (0 0 1) reflection (at 23.128) is broader and not so well defined as the other two for low annealing temperatures. This is reflected by the evolution of FWHM of (0 0 1) reflection in Fig. 6, as it only approaches the values of the other two peaks after a 700 8C annealing. A similar behaviour is presented by the maximum peak intensity as a function of annealing temperature. Since these character- istics should be similar for these three reflections, either in the case of monoclinic or triclinic hypothesis, the nature of this difference should be attributed to the presence of some bulk defects that would mainly affect the (0 0 1) reflection peak. This fact will be further discussed according to TEM, selected area electron diffraction (SAED) observations and XRD simulations. Regarding crystalline structure identification by XRD, the distribution of diffraction peak intensities in the range 32– 358 has been used in literature to distinguish between Fig. 5. XRD spectra of obtained WO 3 (annealed at indicated temperature). Triclinic and monoclinic simulated spectra are also shown. Fig. 6. Evolution of FWHM and intensity of the three main XRD reflections with annealing temperature. I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 479 triclinic and monoclinic structures [21], as simulations of triclinic and monoclinic structures suggest in Fig. 5. In our case, intensities agree better with the hypothesis of a mono- clinic structure. Nevertheless, due to the small mean crystal- lite size, between 30 and 70 nm [15],reflections are badly resolved and the possibility of a mixture of both monoclinic and triclinic phases should be considered, as Raman inves- tigations will show. As regards TEM investigations, diffuse ring patterns obtained by selected area electron diffraction (SAED) (not shown) were not sufficient to determine the amount of triclinic/monoclinic crystalline structure of the samples either. On the other hand, a detailed analysis performed by high- resolution electron microscopy (HRTEM) showed some wide fringes next to the borders of some of the 400 8 C- annealed WO 3 nanoparticles, as it is marked with black arrows in Fig. 7. Detailed SAED pattern showed large reflections corresponding to typical WO 3 and short reflec- tions that have been assigned to Magneli phases (SAED pattern inset in Fig. 7). These phases correspond to oxygen deficient tungsten trioxide with formulas W n O 3nÀ2 [22]. Since these wide fringes were superposed to the WO 3 atomic planes, a detailed digital image analysis was carried out in order to separate these phases and study them. Firstly, a representative squared area from Fig. 7 was selected (Fig. 8a) and a FFT image of this squared region was obtained (Fig. 8b). Afterwards, spots corresponding to the WO 3 planes and to the Magneli Phase on the FFT were selected and filtered by using a mask filter, in order to obtain their representation in separate images, as shown in Fig. 8c and d. From these images, it can be concluded that the wide planes observed correspond to shear planes, in good agreement with those {1 0 3} R crystallographic shear (CS) defects observed by Sloan et al. [23], where {1 0 3} R refer to the family of equivalent directions expressed in the ideal ReO 3 cubic cell. Visible between the CS planes, 0.38 nm lattice fringes corresponding to the (0 1 0) WO 3 planes were found. In tungsten trioxide, it is possible to shear the structure in such a way that oxygen vacancies are eliminated and some tungsten atoms remain more closely spaced, so pairs of W 5þ atoms are found in order to compensate the charge left by the oxygen deficiency. These phases, called Magneli phases, correspond to oxygen deficient tungsten trioxide. All of them present a crystalline structure based on WO 3 zones (corner sharing WO 6 octahedra) linked by units of edge sharing octahedra in the CS phases [23]. These bulk defects influence the electrical transport properties: carrier concen- tration increases, as each missing oxygen atom contributes two carriers, and carrier mobility decreases [24]. This kind of defect was not found by HRTEM in samples annealed at temperatures over 400 8C. Although these substoichiometric regions seemed to be very localised, as diffraction patterns generally observed were those of typical stoichiometric WO 3 , it is reasonable to think that some shear planes could be also present inside the bulk material. These planes may affect the intensities of XRD reflections and this effect has been studied by XRD simulations with the software Diffax [25]. Two kinds of layers were defined: the first one corresponds to an unfaulted WO 3 derived structure (cell no. 1), whereas the other one (cell no. 2) corresponds to the planar defect. Both cells are represented in Fig. 9a, where vectors defining the unit cell 1, 2 and cubic unit cell are also shown. In this representation, W atoms occupy the centre of oxygen octahedra. For the sake of simplicity, the atomic positions have been first described in a cubic representation, the unit cells being Fig. 7. HRTEM micrograph from a 400 8C-annealed WO 3 nanoparticle. The inset corresponds to the SAED pattern of this nanoparticle. Wide fringes have been marked with black arrows. The squared region has been digitally analyzed in Fig. 4. 480 I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 artificially distorted at the end (orthorhombic unit cells) to reflect the distortion of WO 3 structure. The cells parameters used for the pseudo-cubic unit cell were a 0 ¼ 0:365 nm, b 0 ¼ 0:377 nm, c 0 ¼ 0:384 nm and the shear plane direction was [0 1 3], according to HRTEM investigations. Simula- tions have been performed supposing a random stacking of layer 1 and layer 2 with a given probability associated to each layer. On Fig. 9b, simulations for different probabilities of shear planes are shown. It can be seen that for a zero percent probability for layer 2, i.e. a stacking of unfaulted WO 3 layers, the distribution of intensities for the three main diffraction peaks remain similar to that observed for mono- clinic or triclinic structures, confirming the validity of the approximation done on atomic positions. When the prob- ability of layer 2 presence increase, the intensities of calcu- lated (0 0 1) reflection diminishes while its width increases. A similar tendency is observed for the (0 1 0) reflection but to a lower extent. These results agree with the evolution of Fig. 8. (a) Magnified detail of the squared region in Fig. 7, (b) FFT of image 4.a, (c) 0.38 nm WO 3 lattice fringes corresponding to the {1 0 0} R planes obtained after selecting the large reflections in 4.b, (d) {1 0 3} R crystallographic shear (CS) defects from Magneli phases obtained after selecting the short reflections in 4.b. Fig. 9. (a) Representation of the layers used for XRD simulation: the first one corresponds to an unfaulted WO 3 derived structure (cell no. 1), whereas the other one (cell no. 2) corresponds to the planar defect, (b) XRD simulations considering different probabilities of shear planes. I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 481 experimental XRD diffraction patterns of samples annealed between 400 and 700 8C(Fig. 5), confirming that the TEM observed CS planes could be responsible for the anomalous XRD pattern of the (0 0 1) reflection after a 400 8C-anneal- ing. Therefore, this anomalous reflection behaviour can be used as an indirect proof of the presence of bulk oxygen deficiencies. Higher annealing temperatures would reduce the presence of these defects, according to experimental a simulated XRD patterns and TEM observations. Regarding Raman spectroscopy, it is able to give a clearer evidence of the monoclinic or triclinic nature of the WO 3 phase, since the lowest frequency phonon modes are dif- ferent for both structures. Cazzanelli et al. [21] reported that monoclinic structure presents a peak at 34 cm À1 , while triclinic presents a peak at 43 cm À1 . Fig. 10 shows that only the peak corresponding to triclinic phase was present in powders annealed at 4008. When annealing temperature is increased, the peak at 34 cm À1 appears and both peaks seem to coexist in the range of annealing temperatures studied. Similar results concerning the evidence of triclinic-mono- clinic coexistence in WO 3 by Raman have also been reported by Souza-Filho et al. [26]. Therefore, powders annealed at 400 8C have mainly a triclinic phase and both monoclinic and triclinic phases coexist at higher annealing temperatures. Finally, XRD and Raman spectra of 400 and 700 8C- annealed powders were also acquired under controlled temperature (from room temperature to 300 8C for both techniques) and ambient conditions (synthetic dry air only for Raman spectra). The aim of this study was to determine if the previously shown structural differences could also be present at normal sensor operating temperatures. Concern- ing XRD (not shown), displacements lower than 0.18 respect room temperature were found and they were attributed to thermal effects. However, apart from these displacements, no appreciable changes on spectra were found. Therefore, taking into account the previous results, this would mean that CS defects are also present at normal sensor working temperatures in 400 8C-annealed WO 3 . Regarding Raman spectra, no trace of the 34 cm À1 monoclinic vibration was found on the 400 8C-annealed sample in the range of temperatures studied (Fig. 11a). This fact is remarkable, since transition temperature between triclinic and monocli- nic temperature is set around 20 8C [18]. In the case of 700 8C-annealed WO 3 (Fig. 11b), 34 cm À1 monoclinic vibration became broader and collapsed over 100 8C with the peak at 43 cm À1 , typical of triclinic structure. Therefore, this would lead to think that 400 8C-annealed WO 3 is able to maintain its triclinic structure at least up to 100 8C. Over this temperature, unfortunately, it is not possible to determine the predominant crystalline structure since 700 8C-annealed WO 3 showed only a single broad peak over 100 8C. Similar situations have been already described for WO 3 , such as a tetragonal metastable phase at room temperature [27], while it is considered to appear over 1000 8K. Basically, Raman spectroscopy showed that both triclinic and monoclinic structures are present in WO 3 nanocrystal- line powder obtained from tungstic acid. Their abundance depends on annealing temperature and seems to be stable at least up to 100 8 C. TEM showed the presence of CS planes, associated to oxygen deficiencies, in 400 8C-annealed WO 3 , whereas they were not observed in materials annealed at higher temperatures. XRD simulations based on the pre- sence of this defect showed that it might be responsible for the anomalous XRD pattern of the 0 0 1 reflection, confirm- ing its disappearance with annealing over 400 8C. However, Fig. 10. Low-frequency Raman spectra of obtained WO 3 (annealed at indicated temperature). 482 I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 no change of the 400 8C-annealed WO 3 XRD pattern was observed when XRD measuring temperature was changed between room temperature and 300 8C. As a result, this defect may be present at usual working sensor temperatures. 3.3. Discussion In addition to an obvious difference of mean grain size and crystalline quality between materials annealed at 400 and 700 8C [15], previously reported structural differences have to be taken into account to explain differences on sensor response. Some influences of structural properties on sensor response to NO 2 and H 2 S have been already discussed in literature. For instance, it has been reported that high annealing temperature leads to better NO 2 sensor response, despite grain size increase, in screen-printed gas sensors based on SnO 2 nanopowders [28]. This was attributed to the improvement of the crystalline quality and the faceting of nanograins, which improved NO 2 adsorption. Additionally, sensor response of WO 3 to H 2 S has been reported to be highly dependent on crystalline structure, with especially good results in the case of tetragonal structure [23], although there is no reference about influence of triclinic or mono- clinic structure influence, to the best of our knowledge. Fig. 11. (a) evolution of the low-frequency Raman spectra of 400 8C-annealed WO 3 with measurement temperature, (b) evolution of the low-frequency Raman spectra of 700 8C-annealed WO 3 with measurement temperature. I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 483 Finally, influence of crystalline shear planes and bulk oxy- gen deficiencies on electrical conduction (carrier concentra- tion and mobility) has also been reported [23]. In our case, as these defects disappear when annealing temperature increases over 400 8C, oxygen deficiencies may drift to the surface of the grain and become reactive sites for the adsorption of NO 2 and oxygen molecules [29,30], the latter being consumed by H 2 S, and this may thus improve sensor response. It is well known that the formation of chemical bonds between gaseous species and metal oxides depends on the presence of unsaturated bonds on the surface of the material, so the amount of chemisorbed species increases with surface defect concentration [31]. As revealed by the introduction of humidity pulses in the presence of H 2 S, it is clear that water is competing with H 2 S on the grain surface and some sites are blocked for H 2 S reaction. Since gas sensors based on 700 8C-annealed is much more affected, this may lead to think these sites are more abundant in 700 8C-annealed WO 3 , provided there is more than one reactive site. Nevertheless, more spectroscopic in situ mea- surements are in progress in order to confirm these hypoth- eses, which may explain reported differences on sensor response. 4. Conclusions Crystalline structure, defects and gas sensor response to NO 2 and H 2 S of nanocrystalline WO 3 were analysed in this work. Annealing temperature was varied between 400 and 700 8C. Gas sensors based on 700 8C-annealed tungsten trioxide showed a better response to NO 2 and H 2 S in dry air than the ones based on 400 8C-annealed WO 3 .Influence of humidity on the detection of these gases was also ana- lysed. Gas sensors based on 700 8C-annealed tungsten tri- oxide exhibited a lower influence of humidity on NO 2 response, whereas gas sensors based on 400 8C-annealed WO 3 showed a lower influence of humidity on H 2 S detec- tion. Regarding structural characterisation, two crystalline structures (triclinic and monoclinic) in WO 3 nanocrystalline powders obtained from tungstic acid were identified by Raman spectroscopy. The abundance of these structures was dependent on annealing temperature. Crystalline shear planes defects, related to oxygen deficiencies, were present in 400 8C-annealed WO 3 , as revealed by TEM investiga- tions. Comparison between XRD simulations and experi- mental data showed that the amount of these defects in bulk is decreasing when annealing temperature increases from 400 to 700 8C. This fact may mean that oxygen deficiencies are displaced to the outermost part of the grain, increasing thus reactive sites. These structural characteristics were also studied at normal sensor working temperatures by Raman and XRD. It was found that triclinic structure was stable, at least, up to 100 in 400 8C-annealed WO 3 . Besides, crystal- line shear plane defects were present at operating tempera- tures under 300 8C, according to XRD results. References [1] G. Sberveglieri, L. Depero, S. Groppelli, P. Nelli, WO 3 sputtered thin films for NO x monitoring, Sens. Actuators, B 26–27 (1995) 89–92. [2] S. Moulzolf, S. Ding, R. Lad, Stoichiometry and microstructure effects on tungsten oxide chemiresistive films, Sens. Actuators, B 77 (2001) 375–382. [3] C. Cantalini, H.T. Sun, M. Faccio, M. Pelino, S. Santucci, L. Lozzi, M. Passacantando, NO 2 sensitivity of WO 3 thin film obtained by high vacuum thermal evaporation, Sens. Actuators, B 31 (1996) 81–87. [4] D. Lee, J. Lim, S. Lee, J. Huh, D. Lee, Fabrication and characterization of micro-gas sensor for nitrogen oxides gas detection, Sens. Actuators, B 64 (2000) 31–36. [5] H. Lin, C. Hsu, H. Yang, P. Lee, C. Yang, Nanocrystalline WO 3 - based H 2 S sensors, Sens. Actuators, B 22 (1994) 63–68. [6] J. Solis, S. Saukko, L. Kish, C. Granqvist, V. Lantto, Semiconductor gas sensors based on nanostructured tungsten oxide, Thin Solid Films 391 (2001) 255–260. [7] D. Vincenzi, A. Butturi, V. Guidi, M. Carotta, G. Martinelli, V. Guarnieri, S. Brida, B. Margesin, F. Giaocemozzi, M. Zen, D. Giusti, G. Soncini, A. Vaisiliev, A. Pisliakov, Gas-sensing device imple- mented on a micromachined membrane: a combination of thick-film and very large scale integrated technologies, J. Vac. Sci. Technol., B 18 (2000) 2441–2445. [8] I. Simon, N. Barsan, M. Bauer, U. Weimar, Micormachined metal oxide gas sensors: opportunities to improve sensor performance, Sens. Actuators, B 73 (2001) 1–26. [9] J. Cerda, A. Cirera, A. Vila, A. Cornet, J.R. Morante, Deposition on micromachined silicon substrates of gas sensitive layers obtained by a wet chemical route: a CO/CH 4 high performance sensor, Thin Solid Films 391 (2001) 265–269. [10] I. Jime ´ nez, A. Cirera, A. Cornet, J.R. Morante, Pulverisation method for active layer coating on microsystems, Sens. and Actuators, B 84 (2002) 78–82. [11] M. Akiyama, J. Tamaki, N. Miura, N. Yamazoe, Tungsten oxide- based semiconductor sensor highly sensitive to NO and NO 2 , Chem. Lett. (1991) 1611–1614. [12] J. Tamaki, Z. Zhang, K. Fujimori, M. Akiyama, T. Harada, N. Miura, N. Yamazoe, Grain-size effects in tungsten oxide-based sensor for nitrogen oxides, J. Electrochem. Soc. 141 (1994) 2207–2210. [13] J.H. Kim, K.L. Kim, A study of preparation of tungsten nitride catalysts with high surface area, Appl. Catal. A 181 (1999) 103–111. [14] C. Bala ´ zsi, M. Farkas-Jahnke, I. Kotsis, L. Petra ´ s, J. Pfeifer, The observation of cubic tungsten trioxide at high-temperature dehydra- tion of tungstic acid hydrate, Solid State Ionics 141-142 (2001) 411–416. [15] I. Jime ´ nez, J. Arbiol, A. Cornet, J.R. Morante, Structural and gas- sensing properties of WO 3 nanocrystalline powders obtained by a sol-gel method from tungstic acid, IEEE Sens. J. 2 (4) (2002) 329– 335. [16] R. Ionescu, A. Vancu, C. Moise, A. Tomescu, Role of water vapour in the interaction of SnO 2 gas sensors with CO and CH 4 , Sens. Actuators, B 61 (1999) 39–42. [17] T. Kuse, S. Takahashi, transitional behavior of tin oxide semicon- ductor under a step-like humidity change, Sens. Actuators, B 67 (2000) 36–42. [18] P. Woodward, A. Sleight, T. Vogt, Structure refinement of triclinic tungsten trioxide, J. Phys. Chem. Solids 56 (1995) 1035– 1315. [19] J. Rodriguez-Carvajal, FULLPROF: a program for Rietveld refine- ment and pattern matching analysis, Abstracts of the Satellite Meeting on Powder Diffraction of the XV Congress of the IUCr, Toulouse, France, 1990, p. 127. [20] L.E. Depero, S. Groppelli, I. Natali-Sora, L. Sangaletti, G. Sberveglieri, E. Tondello, Structural studies of tungsten-titanium oxide thin films, Solid State Chem. 121 (1996) 379–387. 484 I. Jime ´ nez et al. / Sensors and Actuators B 93 (2003) 475–485 [...]... Parameter optimisation in SnO2 gas sensors for NO2 detection with low cross-sensitivity to CO: sol-gel preparation, film preparation, powder calcination, doping and grinding, Sens Actuators, B 65 (2000) 166–168 [29] S.R Aliwell, J.F Halsall, K.F.E Pratt, J O’Sullivan, R.L Jones, R.A Cox, S.R Utembe, G.M Hansford, D.E Williams, Ozone sensors based on WO3: a model for sensor drift and a measurement correction... triclinic and monoclinic phases in WO3 ceramics, J Raman Spect 31 (2000) 451–454 485 [27] J Solis, S Saukko, L Kish, C Granqvist, V Lantto, Nanocrystalline tungsten oxide thick-films with high sensitivity to H2S at room temperature, Sens Actuators, B 77 (2001) 316–321 ´ ´ [28] A Dieguez, A Romano-Rodrıguez, J.L Alay, J.R Morante, N ˆ ¨ Barsan, J Kappler, U Weimar, W Gopel, Parameter optimisation in SnO2 gas. ..´ I Jimenez et al / Sensors and Actuators B 93 (2003) 475–485 [21] E Cazzanelli, C Vinegoni, G Mariotto, A Kuzmin, J Purans, Lowtemperature polymorphism in tungsten trioxide powders and its dependence on mechanical treatments, J Solid State Chem 143 (1999) 24–32 [22] K Kosuge, Chemistry of non-stoichiometric... Ozone sensors based on WO3: a model for sensor drift and a measurement correction method, Meas Sci Technol 12 (2001) 684–690 [30] M Ivanovskaya, A Gurlo, P Bogdanov, Mechanism of O3 and NO2 detection and selectivity of In2O3 sensors, Sens Actuators, B 77 (2001) 264–267 [31] V.E Henrich, P.A Cox, The surface Science of Metal Oxides, Cambridge University Press, Cambridge, 1994, p 257 ... Chem 143 (1999) 24–32 [22] K Kosuge, Chemistry of non-stoichiometric compounds, Oxford University Press, Oxford, 1994, p 120 [23] J Sloan, J Hutchison, R Tenne, Y Feldman, T Tsirlina, M Homyonfer, Defect and ordered tungsten oxides encapsuled inside 2H-WX2 fullerene-related structures, J Solid State Chem 144 (1999) 100–117 [24] J.G Allpress, R.J.D Tilley, M.J Sienko, Examination of substoichiometric WO3Àx . (a) Sensor response to H 2 S and NO 2 of gas sensors based on 400 and 700 8C-annealed WO 3 as functions of working temperature, (b) Sensor response to H 2 S. reference. Response and recovery times, for both gases, have a low dependence on humidity, as sensor response to NO 2 . On the other hand, sensor response to H 2 S