A vailable online at www.sciencedirect.com Sensors and Actuators B 131 (2008) 174–182 Novel hybrid materials for gas sensing applications made of metal-decorated MWCNTs dispersed on nano-particle metal oxides R. Ionescu a , E.H. Espinosa a , R. Leghrib a , A. Felten b , J.J. Pireaux b , R. Erni c , G. Van Tendeloo c , C. Bittencourt d ,N.Ca ˜ nellas a , E. Llobet a,∗ a MINOS, Universitat Rovira i Virgili, ETSE-DEEEA, Av. Pa¨ısos Catalans 26, E-43007 Tarragona, Spain b LISE, Falcult´es Universitaires Notre Dame de la Paix, 61 rue de Bruxelles, B-5000 Namur, Belgium c EMAT, University of Antwerp, 171 Groenenborgerlaan, B-2020 Antwerp, Belgium d LCIA, Materia Nova, Parc Initialis, Av. Copernic, 1, B-7000 Mons, Belgium Received 5 July 2007; received in revised form 1 November 2007; accepted 2 November 2007 Available online 31 December 2007 Abstract Novel hybrid gas-sensitive materials were fabricated by means of metal-decorated multiwall carbon nanotubes (MWCNT) dispersed on nano- particle metal oxides. The MWCNT were initially functionalized in an oxygen plasma for improving their dispersion and surface reactivity, and then they were decorated with metal nano-clusters by thermally evaporating gold or silver on the MWCNT. Active layers for gas sensing applications were obtained by adding a small amount of metal-decorated MWCNT to two different types of metal oxides (SnO 2 and WO 3 ). The hybrid materials have been analyzed by means of XPS, TEM and SEM. The gas sensing potential of the fabricated hybrid materials has been tested upon exposure to different hazardous species, specifically NO 2 , CO, C 6 H 6 and NH 3 , at low operating temperature. © 2007 Elsevier B.V. All rights reserved. Keywords: Gas sensors; Metal-decorated MWCNT; Hybrid metal oxides/MWCNT; Low temperature operation 1. Introduction Carbon nanotubes (CNT) are receiving nowadays more and more attention from the gas sensor community [1–4]. A special characteristic that makes them interesting as active materials for gas sensors is given by their huge surface area that can be exposed to gases [5]. Experimental reports have shown that upon exposure to toxic gases such as NO 2 ,NH 3 , CO, ben- zene or ethanol, the electrical conductance of semiconducting CNT changes, even when they are operated at low temperatures [6–8], thus reducing considerably the power consumption of the sensing device. However, the agglomeration of CNT into bundles during their synthesis appears as a drawback for forming a well-dispersed active layer. To overcome this inconvenience, a plasma func- ∗ Corresponding author. Tel.: +34 977558502; fax: +34 977559605. E-mail address: eduard.llobet@urv.cat (E. Llobet). tionalization process applied to the CNT has proved to be efficient [9]. This treatment gives, furthermore, rise to functional groups attached to the surface of the nanotubes, which modifies the CNT-surface reactivity and can further improve gas detec- tion. So far, sensors fabricated with multiwall CNT (MWCNT) functionalized with oxygen have proved to give good results when operated at ambient temperature, above all showing good responsiveness to low concentrations of NO 2 [8]. Anyway, in spite of the observed potential of either untreated or functionalized MWCNT to detect gases, they show quite low sensitivities to the hazardous species detected, that can- not be improved even if the sensor operating temperature is increased. In order to overcome this drawback, we considered it worthy to investigate the approach of doping the carbon nan- otubes with metallic nanoclusters, and we recently reported a significant improvement of sensitivity to NO 2 when employing MWCNT decorated with Au or Ag nanoclusters as compared with the response obtained by un-doped MWCNT [10]. A pre- vious functionalization process of the carbon nanotubes in an 0925-4005/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2007.11.001 R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 175 oxygen plasma provided a more homogeneous distribution of the metal nanoclusters on the CNT surface as compared with cluster dispersion on non-treated CNT [11]. On the other hand, metal oxides are well-known materials suitable for detecting a wide spectrum of gases with enough sensitivity. Among these materials, SnO 2 and WO 3 have proved to be very suitable candidates, but when the detection of toxic species is devised they typically need to be operated at temper- atures ranging between 200 and 500 ◦ C [12,13]. Furthermore, it is generally known that, in practice, the sensitivity of metal oxide gas sensors can be enhanced by using bulk dopants or by the addition of metal clusters to the sensing material [14]. When metal oxide sensors are operated at high tempera- tures, changes in the microstructure of the gas-sensitive film are likely to occur (i.e. structural changes or coalescence [15]). At higher temperatures, the mobility of oxygen vacancies becomes appreciable and the mechanism of conduction becomes mixed ionic–electronic. The diffusion of oxygen vacancies is known to be a mechanism responsible for long-term drift in metal oxide gas sensors [16,17]. Therefore, a strategy to avoid long-term changes in their response could consist in operating the sen- sors at temperatures low enough so that appreciable structural variation never occurs, provided that gas reactions occur at a reasonable rate. Recently, hybrid films based on tin or tungsten oxide and carbon nanotubes have been introduced as new gas-sensitive materials with improved sensitivity [10,18–21]. These works indicated that the detection at ambient temperature of toxic gases such as nitrogen dioxide, carbon monoxide and ammonia or ethanol vapors can be improved by dispersing an adequate quantity of carbon nanotubes into a metal oxide matrix. In this paper, we study and discuss the performance in gas sensing of hybrid materials consisting of Au or Ag decorated multiwall carbon nanotubes dispersed in a metal oxide matrix (either SnO 2 or WO 3 is considered for this purpose). 2. Experimental 2.1. Carbon nanotubes functionalization and doping The MWCNT used in the experiment were obtained from Mercorp [22]. They were synthesized by arc discharge without use of catalysts. The MWCNT powder presents 99% of car- bon with 30–40% nanotube content. They have 8–30 graphene layers, are 6–20 nm in diameter and 1–5 ␮m in length. A uniform functionalization with oxygen was applied to the as-provided carbon nanotubes in order to improve their disper- sion and surface reactivity. For this activation step the MWCNT powder was placed inside a glass vessel and a magnet, exter- nally controlled from the plasma chamber, was used to stir the nanotubes powder during the plasma treatment. Inductively cou- pled plasma at a RF frequency of 13.56 MHz was used during the process [23]. Once the MWCNT powder was placed inside the plasma glow discharge, the treatment was performed at a pressure of 0.1 Torr, using a power of 15 W, while the process- ing time was adjusted to 1 min. A controlled flow of oxygen was introduced inside the chamber, which gave rise to oxygen functional groups grafted at the carbon nanotube surface. In the second processing step, the oxygen functionalized nan- otubes were decorated with metal nanoclusters by thermally evaporating gold or silver atoms onto the MWCNT surface from a gold or silver wire, respectively. The processing parameters were adjusted to be sufficient enough in order to obtain a fair dispersion of few metal nanoclusters decorating the MWCNT, but at the same time to avoid the formation of a metallic layer covering the carbon nanotubes [11]. 2.2. Active layers Sensing layers were prepared using commercially avail- able SnO 2 and WO 3 nanopowders (Sigma–Aldrich). Hybrid materials were obtained by adding two different amounts of metal-decorated MWCNT to 70 mg of metal oxide, thus obtain- ing two different proportions of MWCNT embedded into the metal oxide matrix (1/500 and 1/250 wt%, respectively). An adequate mixture of the components was obtained by dissolving them in glycerol (employed as organic vehicle), and stirring the resulting solution in an ultrasonic bath at 75 ◦ C for 2 h. The amount of nanotubes to be added to the metal oxide matrix was based on a previous study [20]. The pastes obtained were dropped onto the electrode area of micro-hotplate transduc- ers (fabricated at the Centre Nacional de Microelectr ` onica, Barcelona, Spain) using a micro-injector (JBE1113 Dispenser, I&J FISNAR Inc., USA). The as-deposited sensing films were firstly dried at 140 ◦ C during 2 h in order to burn out the organic vehicle, using a slow temperature ramp of 2.5 ◦ C/min for reaching this temperature in order to avoid the occurrence of cracks in the films. We used for this process a lower temperature than the boiling point of glycerol (i.e., 182 ◦ C [24]) in order to avoid producing cracks in the films deposited. The drying time was however sufficiently long for the complete evaporation of the organic vehicle. Finally, the films were annealed in situ at 450 ◦ C during 3 h in ambi- ent atmosphere (this also ensures the complete removal of the organic vehicle). During the annealing process, the tempera- ture was raised from ambient to 450 ◦ C using again a ramp of 2.5 ◦ C/min. The mean grain size of the metal oxides particles, determined in a previous study, was near 40 nm [25]. 2.3. Material characterization The chemical composition of the samples’ surfaces was investigated by means of X-ray photoelectron spectroscopy (XPS). XPS analyses were performed using an ESCA-300 (SCI- ENTA, Sweden) photoelectron spectrometer equipped with a monochromatized Al K␣ = 1486.7 eV. A high-resolution hemi- spherical electrostatic analyzer of 600 mm diameter and 75 eV pass energy was used. The angle between the incident X-ray and the photoelectron take-off direction was 45 ◦ , with the latter nor- mal to the sample surface. The overall resolution of the system (source + analyzer) was 0.6 eV. The background pressure during experiment was better than 10 −9 Torr). 176 R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 The size of the metal clusters and their dispersion on the CNT walls were studied by means of high-resolution transmis- sion electron microscopy (HRTEM) carried out using a Philips CM30 FEG instrument operated at 300 kV. In order to reduce potential knock-on radiation damage caused by the 300 keV electron beam, the electron dose was significantly decreased such that during the entire electron-beam exposure no changes in the nanotubes were observable. The morphology of the hybrid films deposited onto the microhotplate substrates was investigated by means of scan- ning electron microscopy (SEM). SEM measurements were performed using a Joel JSM 6400 equipment, with a resolution of 0.3 nm. The magnification during this study was set to values varying between 30,000 and 50,000. Accelerating voltages of 25 or 30 kV were employed. The system allows for sample rotation (360 ◦ ) and sample inclination (90 ◦ ). 2.4. Gas sensing measurements The gas sensing properties of the different hybrid active mate- rials produced were tested in the presence of different hazardous species, such as CO (carbon monoxide), NO 2 (nitrogen diox- ide), NH 3 (ammonia) and C 6 H 6 (benzene). The sensors were operated at three different temperatures: 25 (i.e., ambient), 150 and 250 ◦ C. To perform the measurements, the gas sensors were placed inside a 5.3 dm 3 test chamber, and the desired concen- trations of each contaminant under study were introduced by the direct injection method using a gas-tight chromatographic syringe. Specifically, these concentrations were: 100, 200, 500 and 1000 ppb for NO 2 ; 10 and 50 ppm for CO; 2, 5 and 10 ppm for NH 3 and 50 and 150 ppm for C 6 H 6 . To assess the repro- ducibility of the results, each measurement was replicated 4 times. An Agilent 34970A multimeter was used for continu- ously monitoring the electrical resistance of the sensors during the measurement process. The data acquired were stored in a PC for further analysis. The measurement process was as follows (identical for any species tested): data acquisition started 10 min before injecting into the measurement chamber the required volume correspond- ing to the lowest concentration of the contaminant measured. After reaching a steady state, a new amount of the same con- taminant was injected, so that the second concentration to be tested was reached. Successive injections were repeated until all desired concentrations of the gas were measured. After each series of successive injections, the sensor chamber was flushed using pure dry air for 2 h, which ensured the cleaning of both the chamber and the sensor surface. Finally, the airflow was interrupted and the sensors were left to recover their baseline resistance before performing a new set of measurements. In addition to the previously mentioned measurements, con- trol gas sensing experiments were carried out employing both pure WO 3 and pure SnO 2 materials. Furthermore, we reported recently gas sensing results obtained with Au and Ag deco- rated MWCNT [10]. The present results are also compared with previous works performed under the same experimental con- ditions with WO 3 and SnO 2 films activated with Au and Ag [25–27]. Fig. 1. XPS spectra recorded before and after oxygen functionalization of MWCNT. 3. Results and discussion 3.1. Characterization Fig. 1 shows the XPS survey spectra of the carbon nanotubes recorded before and after the plasma treatment. The peak at 284.6 eV, observed in both spectra, is generated by photoelec- trons emitted from the C 1s core level. In the XPS spectrum recorded after oxygen plasma treatment a peak near 535 eV can be observed. This peak is generated by photoelectrons emitted from the O 1s core level. The relative atomic oxygen concentra- tion evaluated by XPS was found to be 6%. Fig. 2 shows the XPS spectra recorded on the metal-decorated carbon nanotubes. After the evaporation of the metals, the XPS survey spectra showed, in addition to the mean peak situated at 284.6 eV generated by photoelectrons emitted from the C 1s peak, a doublet peak near 85.0 eV that was generated by pho- toelectrons emitted from the Au 4f core levels (Fig. 2a), or one peak at 370.0 eV generated by photoelectrons emitted from the Ag 3d core level (Fig. 2b). The interaction between the metal clusters and the CNT surface can be studied by XPS. If there is a chemical reaction at the interface, then the new chemical environment of the atoms at the interface will show in the XPS spectra by the appearance of new features. In the core level spectra recorded (Fig. 2, inset) on these samples no additional features beyond the Ag 3d and Au 4f doublet were observed, thus no chemical reaction between the carbon and gold or silver atoms occurs. The relative atomic concentration of each element, evaluated by XPS after plasma functionalization and metal evaporation, is summarized in Table 1. The decrease in the oxygen atomic con- centration can be associated to the increase in the CNT surface area that is covered by gold atoms. This reduces the effective number of C 1s and O 1s photoelectrons passing through the “gold or silver overlayer” and contributing to the XPS spectrum, which explains the decrease in the curve for the total number of photoelectrons emitted. R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 177 Fig. 2. XPS spectra recorded on metal-decorated MWCNT: (a) Au-decorated; (b) Ag-decorated. Fig. 3 shows a typical TEM image recorded on oxygen plasma treated MWCNT decorated with gold clusters. The presence of homogeneously dispersed quasi-spherical gold clusters at the CNT surface can be observed. In the inset, a detailed view of a gold cluster sitting on the CNT surface is shown. The preserved structural characteristics of the graphene layer under the gold cluster are a strong indication of the absence of an Au–C phase formation. Similar results (not shown here) were obtained for Ag clusters evaporated on the CNT surface. Table 1 Relative atomic concentrations obtained by XPS [C] (%) [O] (%) [Ag] (%) [Au] (%) Oxygen plasma treated MWCNTs 94.0 6.0 Au-MWCNTs 92.0 2.7 5.3 Ag-MWCNTs 95.0 3.2 1.8 Fig. 3. TEM image recorded on MWCNT decorated with gold metallic nan- oclusters deposited by thermal evaporation. Considering that the interaction between Au or Ag atoms and the CNT surface was reported to be weak and presumably of Van der Waals in nature [11], an inspection to evaluate if the gas sensing layer preparation method employed (i.e., the soni- cation process applied for obtaining a good mixture of the active layer components) could affect the dispersion of the metal nan- oclusters at the CNT surface was performed. For this evaluation, MWCNTs decorated with metal nanoclusters dispersed in glyc- erol were analyzed. In this case no metal oxide was used and the same procedure as the one employed for the fabrication of the hybrid materials (i.e., stirring the solution in an ultrasonic bath for 2 h at 75 ◦ C) was implemented. Fig. 4 shows a TEM image of a metal-decorated nanotube having undergone this process. The presence of metal clusters laying isolated within the deposited film reveals that the stirring process implemented to obtain a good dispersion of the metal-decorated MWCNT in the glyc- erol solution affects the decoration of MWCNT, removing some of the clusters from the CNT surface. From Fig. 4 it can be seen Fig. 4. TEM image recorded on Au-decorated MWCNT after a sonication pro- cess performed during the preparation of the sensing materials. 178 R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 Fig. 5. SEM images recorded on different hybrid films: (a) Ag-MWCNT/WO 3 (1/500 wt%); (b) Ag-MWCNT/SnO 2 (1/250 wt%). that mainly big clusters are removed from the CNT surface. Fur- ther studies will be performed to estimate if the cluster removal is size-selective and if clusters are removed intact or only in part. It is worth noting that the clusters removed from the CNT surface during the fabrication of hybrid metal oxides/metal- decorated CNT will be embedded in the metal oxide matrix. However, improvements in the sensing properties (e.g., enhanc- ing their sensitivity, decreasing the operating temperature or making them more selective to a given target species) have been reported to be achieved by adding small amounts of noble metals to the metal oxide active layer [28]. Thus, it can be suggested that besides the presence of CNTs, the presence of noble metal Table 2 Responsiveness to NO 2 and CO of the different sensors studied as calculated from Eq. (1) Film Layer NO 2 (250 ◦ C) NO 2 (150 ◦ C) NO 2 (25 ◦ C) 0.1 ppm 0.5 ppm 1 ppm 0.1 ppm 0.5 ppm 1 ppm 0.1 ppm 0.5 ppm 1 ppm WO 3 1.98 5.73 7.14 0.19 0.41 0.50 0 0 0 SnO 2 0.81 11.04 15.72 1.01 4.47 4.66 0 0.10 0.10 Au-MWCNT/WO 3 (1:500) 1.39 8.96 21.44 0.35 1.66 2.07 0 0 0 Au-MWCNT/WO 3 (1:250) 0 0 0 −0.03 −0.06 −0.08 −0.03 −0.07 −0.08 Ag-MWCNT/WO 3 (1:500) 1.40 8.00 24.44 0.29 1.51 2.78 0 0 0 Ag-MWCNT/WO 3 (1:250) 0 0 0 −0.01 −0.04 −0.05 −0.03 −0.07 −0.09 Au-MWCNT/SnO 2 (1:500) 47.36 238.56 243.92 10.67 32.34 32.34 0.96 0.98 0.98 Au-MWCNT/SnO 2 (1:250) 73.21 471.21 485.83 30.29 88.91 88.91 1.14 1.14 1.14 Ag-MWCNT/SnO 2 (1:500) 15.24 72.02 96.69 10.34 31.40 31.40 1.04 2.64 2.69 Ag-MWCNT/SnO 2 (1:250) 19.30 98.59 135.79 9.01 16.42 16.42 0.35 0.35 0.35 Au-MWCNT 0 −0.01 −0.02 0 −0.04 −0.07 −0.01 −0.07 −0.09 Ag-MWCNT 0 −0.02 −0.03 −0.01 −0.03 −0.06 −0.01 −0.04 −0.07 Film Layer CO (250 ◦ C) CO (150 ◦ C) CO (25 ◦ C) 10 ppm 50 ppm 10 ppm 50 ppm 10 ppm 50 ppm WO 3 0.73 0.79 0.27 0.72 0 0 SnO 2 0.07 0.25 0.14 0.34 0 0 Au-MWCNT/WO 3 (1:500) 000000 Au-MWCNT/WO 3 (1:250) 000000 Ag-MWCNT/WO 3 (1:500) 2.34 16.11 0.48 2.51 0 0 Ag-MWCNT/WO 3 (1:250) 0.06 0.05 0 0 −0.02 −0.04 Au-MWCNT/SnO 2 (1:500) 9.32 31.10 4.37 7.50 0 0 Au-MWCNT/SnO 2 (1:250) 25.05 74.59 11.29 17.52 0 0 Ag-MWCNT/SnO 2 (1:500) 15.03 36.76 10.12 16.10 0 0 Ag-MWCNT/SnO 2 (1:250) 14.39 45.18 10.96 16.44 0 0 Au-MWCNT 000000 Ag-MWCNT 000000 R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 179 clusters inside the metal oxide matrices can act improving the sensing properties. Fig. 5 shows SEM images recorded on the different hybrid carbon nanotubes/metal oxide sensing films deposited over the sensors substrates. The micrographs show the presence of WO 3 and SnO 2 metal oxide grains, whereas carbon nanotubes are observed only in WO 3 /CNT hybrids. This can be associated to the low proportion of carbon nanotubes embedded in the metal oxide matrix (1/500 or 1/250 wt%) as well as to the difference in density between WO 3 and SnO 2 (7.16 and 6.95 g/cm 3 , respec- tively [29]); in order to obtain the desired weight proportions between the metal oxide and CNTs a higher amount of SnO 2 than WO 3 is present in the hybrid materials. 3.2. Gas response analysis The gas sensing properties of the different hybrid materials produced were evaluated in terms of the ratio between the change experienced by the sensor resistance after its exposure to pollu- tant species and the sensor baseline resistance in air (see Eq. (1), where S defines sensor’s responsiveness, R air sensor’s resistance in air and R gas sensor’s resistance in the presence of the pollutant after reaching a steady state). Responsiveness results for NO 2 and CO are summarized in Table 2; because no response was obtained to NH 3 and C 6 H 6 , for space reasons these were not included in the responsiveness table. S = R air − R gas R air (1) Sensors based on SnO 2 and Au-decorated MWCNT hybrids were the most sensitive to NO 2 among the different materi- als studied, outperforming the responsiveness of either pure SnO 2 or pure Au-decorated MWCNT materials when operating both at 250 and 150 ◦ C. Typical responses of metal-decorated MWCNT/SnO 2 gas sensors to NO 2 are shown in Fig. 6. The fluctuation in the response signal that occurs at high NO 2 con- centrations is due to the increased effect of noise when sensor resistance becomes very high (i.e. comparable in magnitude to the input impedance of the multimeter employed to acquire it). The quantity of nanotubes dispersed in the SnO 2 matrix was found to play a determinant role in the responsiveness of the hybrid materials to NO 2 . At a concentration ratio of 1/250 wt%, the responsiveness of the hybrids made of metal- decorated MWCNT (either using Au or Ag as dopants) and SnO 2 was significantly superior to that with the 1/500 wt% ratio when the detection of NO 2 at 250 ◦ C was aimed at. When the operat- ing temperature of sensors was lowered to 150 ◦ C, the particular type of metal decorating the carbon nanotubes played also an important role in the NO 2 detection. Thus, at 150 ◦ C a concen- tration ratio of 1/250 wt% of carbon nanotubes dispersed in the SnO 2 matrix was the most appropriate when Au was used as a dopant, while 1/500 wt% of Ag-decorated MWCNT added to SnO 2 was the most suitable for this latter case. On the other hand, similar values of responsiveness were found at 150 ◦ C for both Au-MWCNT/SnO 2 and Ag-MWCNT/SnO 2 (1/500 wt%) materials. Furthermore, it is worth mentioning that the response of the hybrid films based on SnO 2 became already saturated after Fig. 6. Response to different hazardous species obtained by gas sensors employing different hybrid materials: (a) NO 2 detection at 250 ◦ C with Au-MWCNT/SnO 2 (1/250 wt%) sensor; (b) NO 2 detection at 150 ◦ C with Au- MWCNT/SnO 2 (1/250 wt%). the injection of only 500 ppb of NO 2 at the working temperature of 150 ◦ C (see Fig. 6b). Regarding the responsiveness towards NO 2 of the hybrid materials based on WO 3 (see Fig. 7), it was at least one order of magnitude below the one obtained by the hybrids based on SnO 2 . The quantity of nanotubes embedded in WO 3 was of impor- tant relevance. Thus, when the sensors were operated at 250 ◦ C, only metal-decorated MWCNT dispersed in the WO 3 matrix in a concentration ratio of 1/500 wt% were able to detect NO 2 (the metal used as a dopant did not change sensor performance in this case), while at 150 ◦ C the semiconducting behaviour of the metal-decorated MWCNT/WO 3 hybrids changed from n-type at a concentration ratio of 1/500 wt% to p-type at 1/250 wt%. For the measurements performed at 250 ◦ C, the response time to 100 ppb and to 500 ppb of NO 2 was around 2 min both for the hybrid materials based on WO 3 and for the pure WO 3 sensors. It rose to 3 min in the case of hybrids based on SnO 2 , but in this latter case it compared very favourably to the one of pure SnO 2 sensors, which did not reach completely a steady state regime 15 min after gas injection. When the sensor operating tempera- ture was lowered and NO 2 test measurements were performed, the sensors needed a longer time to reach the steady state. The response times varied between 6 and 10 min for hybrid sensors and were over 15 min for pure metal oxide sensors. Recovery 180 R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 Fig. 7. (a) NO 2 detection at 150 ◦ C with Ag-MWCNT/WO 3 (1/500 wt%) sensor; (b) NO 2 detection at 150 ◦ C with Ag-MWCNT/WO 3 (1/250 wt%) sensor. time varied between 10 and 20 min for Au-based hybrid sensors (see Fig. 6) and was over 30 min for Ag-based hybrids and pure metal oxide sensors. The second air pollutant tested was carbon monoxide. The highest responsiveness in the case of the CO tests was again achieved by the hybrid sensors based on Au-decorated MWCNT and SnO 2 in a concentration ratio of 1/250 wt%, operated at 250 ◦ C (see Fig. 8). Although lower, some responsiveness was also obtained at 150 ◦ C by the hybrid sensors containing SnO 2 . Fig. 8. CO detection at 250 ◦ C with Au-MWCNT/SnO 2 (1/250 wt%) sensor. Similarly to the detection of NO 2 , when CO was tested Au- based hybrid sensors showed response and recovery times of about 5 min (operated at 250 ◦ C), while the response and recov- ery times of pure metal oxides was higher than 15 min. When the metal oxide employed was WO 3 , the only hybrid based on this material that responded to CO was Ag-MWCNT/WO 3 in the concentration ratio of 1/500 wt% at an operating temperature of 250 ◦ C. When the Ag-MWCNT/WO 3 sensor (concentration ratio 1/250 wt%) was operated at 250 ◦ C, it behaved as an n- type semiconductor in the presence of CO; operated at 150 ◦ C it did not respond at all to CO; while at room temperature its behaviour was equivalent to an n-type semiconductor. This behaviour clearly suggests that not only the amount of carbon nanotubes present determines the semiconducting character of the resulting hybrid material, but also that the operating temper- ature can play an important role in the sensing mechanism. Unlike oxygen-functionalized MWCNT/SnO 2 hybrid mate- rials, which showed good responsiveness to NO 2 at room temperature [30], metal-decorated MWCNT/metal oxide hybrid materials were not responsive at room temperature. Regarding the response of the gas sensors to the other two pollutants tested (i.e., benzene and ammonia), they were not able to detect the presence of these two contaminants at a concentra- tion level up to 10 ppm in the case of NH 3 and up to 150 ppm in the case of C 6 H 6 , for the operating temperatures investigated. 3.3. Discussion On the basis of the images of the hybrid films recorded by SEM analyses, it can be derived that MWCNT are embed- ded within the metal oxide matrix. It has been reported that in hybrid films, two different depletion layers (and associ- ated potential barriers) co-exist [18,19]: one depletion layer is located at the surface of the grains of the metal oxide film and the other one at the interface between MWCNT and metal oxide films. Since SnO 2 or WO 3 films behave as n-type semiconductors and MWCNT films behave as p-type semi- conductors [8,31], it can be suggested that the hetero-structure n-SnO 2 /p-MWCNT (n-WO 3 /p-MWCNT) is formed at the inter- face between tin oxide (tungsten oxide) and carbon nanotubes. Furthermore, our results indicate that the addition of metal nanoclusters at the CNT surface plays a fundamental role in improving the sensing properties. Studies are being performed to establish if the metal clusters at the CNT surface act lowering the potential barrier of the depletion layers and/or enhancing specific gas adsorptions or promoting specific chemical reac- tions. Considering the sensitivity of the hybrid films (see Table 2), it can be derived that the adsorption of NO 2 or CO at the metal oxide modifies the depletion layer at the surface of its grains and also at the n-metal oxide/p-MWCNT hetero- structures. This combined effect may explain the improvement in responsiveness shown by tin or tungsten oxide-based hybrid sensors as compared with either pure metal oxides, metal- decorated CNT based gas sensors [10] or WO 3 or SnO 2 metal oxides doped with Au or Ag noble metals [25–27]. The results obtained indicate also that the number of CNT added to the R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 181 metal oxide matrix has to be extremely small. The best results were obtained with the SnO 2 /CNT hybrids, when the pres- ence of the carbon nanotubes was undetectable by normal SEM analyses. This is in concordance with the results pub- lished by Wei et al. [18]. We expect that keeping extremely low the number of CNT added to the metal oxide matrix, improved results in terms of sensitivity to NO 2 and CO can be obtained. The lack of responsiveness observed for NH 3 and C 6 H 6 by the new hybrid sensors can be associated to the non-appropriateness of the used combination of materials for detecting such species, as suggested by Penza et al. [32] who found that Au-CNT is an appropriate material for detecting NO 2 while Pt-CNT is more suitable for detecting benzene or ammonia. 4. Conclusions In this paper we have shown that the addition of a small quantity of metal-decorated MWCNT to metal oxides can sig- nificantly improve the detection capability of metal oxide based sensors and lower the operating temperature. In particular, micro-sensors based on Au-MWCNT/SnO 2 hybrid films in a concentration ratio of 1/250 wt% showed the highest sensitivity towards NO 2 and CO, among the different materials studied. The response mechanism is fully reversible, since the sensors can recover their baseline resistance after each exposure to pollutant gases. Our results suggest that there is an optimum amount of carbon nanotubes to be added to each particular metal oxide in order to enhance the responsiveness. Material characteriza- tion analyses (performed by SEM and TEM) showed that the nanotubes endured the process of deposition and subsequent annealing at 450 ◦ C in air, but at the same time part of the metal nanoclusters decorating the nanotube surface were deta- ched. Based on these results, the modulation of the width of two depletion layers existing at the surface of metal oxide grains and at the interface of metal oxide grains and MWCNT, respec- tively, is postulated as the mechanism that could explain the enhanced performance of hybrid metal oxide/MWCNT sen- sors in comparison with pure metal oxide or pure MWCNT sensors. Acknowledgements This work was funded in part by the Spanish Commis- sion for Science and Technology (CICYT) under grant no. TIC2006-03671/MIC. E.H. Espinosa and R. Leghrib gratefully acknowledge Ph.D. studentships from Rovira i Virgili Univer- sity. R. Ionescu holds a ‘Juan de la Cierva’ research fellowship funded by the Spanish Ministry for Science and Education. Parts of this work are directly connected to the Belgian Program on Interuniversity Attraction Poles (PAI5/1/1) on Quantum Size Effects in Nanostructured Materials, sponsored by the Commu- naut ´ e Franc¸aise de Belgique. References [1] S. Chopra, A.M. Rao, Selective gas detection using a carbon nanotube sensor, Appl. Phys. Lett. 83 (2003) 2280–2282. [2] C. Cantalini, L. Valentini, I. Armentano, L. Lozzi, J.M. Kenny, S. San- tucci, Sensitivity to NO 2 and cross-sensitivity analysis to NH 3 , ethanol and humidity of carbon nanotubes thin film prepared by PECVD, Sens. Actuators B 95 (2003) 195–202. [3] J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, M. Meyyappan, Carbon nanotube sensors for gas and organic vapor detection, Nano Lett. 3 (2003) 929–933. [4] Y. Lu, C. Partridge, M. Meyyappan, J. Li, A carbon nanotube sensor array for sensitive gas discrimination using principal component analysis, J. Electroanal. Chem. 593 (2006) 105–110. [5] O.K. Varghese, P.D. Kichambre, D. Gong, K.G. Ong, E.C. Dickey, C.A. Grimes, Gas sensing characteristics of multi-wall carbon nanotubes, Sens. Actuators B 81 (2001) 32–41. [6] L. Valentini, C. Cantalini, I. Armentano, J.M. Kenny, L. Lozzi, S. Santucci, Highly sensitive and selective sensors based on carbon nanotubes thin films for molecular detection, Diamond Relat. Mater. 13 (2004) 1301–1305. [7] S.G. Wang, Q. Zhang, D.J. Yang, P.J. Sellin, G.F. Zhong, Multi-walled car- bon nanotube-based gas sensor for NH 3 detection, Diamond Relat. Mater. 13 (2004) 1327–1332. [8] R. Ionescu, E.H. Espinosa, E. Sotter, E. Llobet, X. Vilanova, X. Correig, A. Felten, C. Bittencourt, G. Van Lier, J C. Charlier, J.J. Pireaux, Oxygen functionalisation of MWNT and their use as gas sensitive thick-film layers, Sens. Actuators B 113 (2006) 36–46. [9] N.O.V. Plank, R. Cheung, Functionalisation of carbon nanotubes for molec- ular electronics, Microelectron. Eng. 73/74 (2004) 578–582. [10] E.H. Espinosa, R. Ionescu, C. Bittencourt, A. Felten, R. Erni, G. Vantende- loo, J J. Pireaux, E. Llobet, Metal-decorated multi-wall carbon nanotubes for low temperature gas sensing, Thin Solid Films 515 (2007) 8322–8327. [11] A. Felten, C. Bittencourt, J F. Colomer, G. Van Tendeloo, J J. Pireaux, Nucleation of metal clusters on plasma treated multi wall carbon nanotubes, Carbon 45 (2007) 110–116. [12] I. Simon, N. Barsan, M. Bauer, U. Weimar, Micromachined metal oxide gas sensors: opportunities to improve sensor performance, Sens. Actuators B 73 (2001) 1–26. [13] M. Graf, D. Barrettino, K U. Kirstein, A. Hierlemann, CMOS micro- hotplate sensor system for operating temperatures up to 500 ◦ C, Sens. Actuators B 117 (2006) 346–352. [14] W. G ¨ opel, New materials and transducers for chemical sensors, Sens. Actu- ators B 18 (1994) 1–21. [15] E. Comino, Metal oxide next term nano-crystals for gas sensing, Anal. Chim. Acta 568 (2006) 28–40. [16] P.K. Clifford, D.T. Tuma, Characteristics of semiconductor gas sensors. 1. Steady-state gas response, Sens. Actuators 3 (1983) 233–254. [17] P.K. Clifford, D.T. Tuma, Characteristics of semiconductor gas sensors. 2.Transient-response to temperature-change, Sens. Actuators 3 (1983) 255–281. [18] B.Y. Wei, M.C. Hsu, P.G. Su, H.M. Lin, R.J. Wu, H.J. Lai, A novel SnO 2 gas sensor doped with carbon nanotubes operating at room temperature, Sens. Actuators B 101 (2004) 81–89. [19] Y. Chen, C. Zhu, T. Wang, The enhanced ethanol sensing properties of multi-walled carbon nanotubes/SnO 2 core/shell nanostructures, Nanotech- nology 17 (2006) 3012–3017. [20] C. Bittencourt, A. Felten, E.H. Espinosa, R. Ionescu, E. Llobet, X. Correig, J J. Pireaux, WO 3 films modified with functionalised multi-wall carbon nanotubes: Morphological, compositional and gas response studies, Sens. Actuators B 115 (2006) 33–41. [21] E.H. Espinosa, R. Ionescu, E. Llobet, A. Felten, C. Bittencourt, E. Sotter, Z. Topalian, P. Heszler, C.G. Granqvist, J.J. Pireaux, X. Correig, Highly selective NO 2 gas sensors made of MWCNTs and WO 3 hybrid layers, J. Electrochem. Soc. 154 (2007) J141–J149. [22] http://www.mercorp.com. [23] C. Lefebvre, J. Verbist, Characterization of acrylonitrile plasma- polymerized on carbon fibers, Adv. Comp. Mat. Lett. 1 (1992) 34–37. [24] http://www.sigmaaldrich.com/catalog/search/ProductDetail/FLUKA /49770?LastFive. 182 R. Ionescu et al. / Sensors and Actuators B 131 (2008) 174–182 [25] P. Ivanov, E. Llobet, F. Blanco, A. Vergara, X. Vilanova, I. Gracia, C. Can ´ e, X. Correig, On the effects of the materials and the noble metal additives to NO 2 detection, Sens. Actuators B 118 (2006) 311–317. [26] M. Stankova, X. Vilanova, J. Calderer, E. Llobet, J. Brezmes, I. Gr ` acia, C. Can ´ e, X. Correig, Sensitivity and selectivity improvement of rf sputtered WO 3 microhotplate gas sensors, Sens. Actuators B 113 (2006) 241–248. [27] M. Penza, G. Cassano, F. Tortorella, Gas recognition by activated WO 3 thin-film sensors array, Sens. Actuators B 81 (2001) 115–121. [28] S. Santucci, C. Cantalini, M. Crivellari, L. Lozzi, L. Ottaviano, M. Pas- sacantando, X-ray photoemission spectroscopy and scanning tunneling spectroscopy study on the thermal stability of WO 3 thin films, J. Vac. Sci. Tech. A 18 (2000) 1077–1082. [29] http://www.sigmaaldrich.com. [30] E.H. Espinosa, R. Ionescu, B. Chambon, G. Bedis, E. Sotter, C. Bittencourt, A. Felten, J J. Pireaux, X. Correig, E. Llobet, Low tem- perature gas detection with hybrid metal oxides/MWCNTs, in: XX Eurosensors Conference Aniversary, G ¨ oteborg, Sweden, September 17–20, 2006. [31] L. Valentini, I. Armentano, J.M. Kenny, C. Cantalini, L. Lozzi, S. Santucci, Sensors for sub-ppm NO 2 gas detection based on carbon nanotube thin films, Appl. Phys. Lett. 82 (2003) 961–963. [32] M. Penza, G. Cassano, R. Rossi, M. Alvisi, A. Rizzo, M.A. Signore, Th. Dikonimos, E. Serra, R. Giorgi, Enhancement of sensitivity in gas chemire- sistors based on carbon nanotube surface functionalized with noble metals (Au, Pt) nanoclusters, Appl. Phys. Lett. 90 (2007) 173–176. Biographies Radu Ionescu is a postdoctoral research fellow at the Department of Electronics, Electrical and Automatic Engineering, Rovira i Virgili University, Tarragona, Spain. His main research interests are in the field of chemical gas sensors, carbon nanotubes and pattern recognition. Edwin Espinosa is a PhD student at the Department of Electronics, Electrical and Automatic Engineering, Rovira i Virgili University, Tarragona, Spain. His research topic consists of gas sensors based on plasma functionalised carbon nanotubes. Radouane Leghrib is a PhD student at the Department of Electronics, Electrical and Automatic Engineering, Rovira i Virgili University, Tarragona, Spain. His research topic consists of gas sensors based on carbon nanotubes decorated with metal nanoclusters. Alexandre Felten is a postdoctoral research fellow at the LISE laboratory, University of Namur, Belgium. One of his research interests is in the func- tionalisation of carbon nanotubes using cold plasmas. Jean Jacques Pireaux is the director of the LISE laboratory, University of Namur, Belgium. Professor Pireaux leads a project on the interface design of metal nanocluster-carbon nanotube hybrids via control of structural and chemi- cal defects in a plasma discharge. Rolf Erni is a researcher at the EMAT, University of Antwerp, Belgium. His main areas of interest are in Electron scattering and diffraction physics, aber- ration correction in TEM and STEM, low-loss and high-resolution electron energy-loss spectroscopy, functional materials such as optically active materials, nanomaterials such as quantum dots, quantum well structures and nanoparticles. Gustav Van Tendeloo is professor of physics at the University of Antwerp, Belgium. His main research interest are in superconducting and CMR materials, ceramic thin films, nanotubes, nanowires, nanobelts, nanoparticles and meso- porous materials, GaN and related semiconductors, solid state phase transitions and modulated structures. Carla Bittencourt is a senior researcher at the LCIA, University of Mons- Hainaut, Belgium. One of her research interest is in the development of metal oxide and carbon nanotube hybrid materials for sensing gases at low operating temperatures. Eduard Llobet is an associate professor of electronics at the University Rovira i Virgili (Tarragona, Spain). His main research interests are in the fabrication and modelling of semiconductor gas sensors and in the applications of intelligent systems to complex odour analysis. . www.sciencedirect.com Sensors and Actuators B 131 (2008) 174–182 Novel hybrid materials for gas sensing applications made of metal-decorated MWCNTs dispersed on nano-particle. SnO 2 than WO 3 is present in the hybrid materials. 3.2. Gas response analysis The gas sensing properties of the different hybrid materials produced were evaluated