- Email: [email protected]
Influence of the doping method on the sensitivity of Pt-doped screen-printed SnO2 sensors
Sensors and Actuators B 97 (2004) 67–73
Influence of the doping method on the sensitivity of Pt-doped screen-printed SnO2 sensors C. Bittencourt a , E. Llobet b,∗ , P. Ivanov b , X. Correig b , X. Vilanova b , J. Brezmes b , J. Hubalek c , K. Malysz c , J.J. Pireaux a , J. Calderer d a
d
Laboratoire Interdisciplinaire de Spectroscopie Electronique, Institute for Studies in Interface Sciences, Falcultés Universitaires Notre Dame de la Paix, 61 rue de Bruxelles, B-5000 Namur, Belgium b Departament d’Enginyeria Electrònica, Universitat Rovira i Virgili, Avinguda Pa¨ısos Catalans, 26, 43007 Tarragona, Spain c Faculty of Electrical Engineering and Communication, Brno University of Technology. Údoln´ı 53, 602 00 Brno, Czech Republic Department d’Enginyeria Electrònica, Universitat Politècnica de Catalunya, Campus Nord, Gran Capità s/n, 08034 Barcelona, Spain Received 2 April 2003; received in revised form 25 July 2003; accepted 25 July 2003
Abstract In this work, we study the influence of the introduction method of Pt atoms on the sensitivity to traces of ethanol of Pt-doped SnO2 sensors. The tin oxide films were obtained by a screen-printing process. Two different methods were employed to introduce Pt atoms on SnO2 films. In the first one, the Pt atoms were added to the screen-printed tin oxide layer by using RF magnetron sputtering and a subsequent thermal treatment. The second method consisted of mixing SnO2 and Pt pastes before the screen-printing process. The different active layers (including un-doped tin oxide) were carefully examined relative to their sensitivity to ethanol at different working temperatures. Sensors prepared by the second method showed sensitivity to ethanol four times higher than one of the sensors prepared by the first method and 12 times higher than un-doped sensors. XPS and scanning electron microscopy (SEM) measurements showed that this behaviour could be associated with the spatial distribution of the doping elements within the tin oxide film. While in Pt-sputtered sensors most of the Pt atoms were found at the surface of the active layer, for the sensors made by mixing Pt and SnO2 pastes, a homogeneous distribution of the Pt atoms was observed. These sensors show high sensitivity and fast response time to ethanol vapours, with a detection limit in the ppb range. © 2003 Elsevier B.V. All rights reserved. Keywords: SnO2 sensors; Screen-printing; Pt loading; Detection limit
1. Introduction Metal-oxide semiconductor (MOS) gas sensors are a low-cost option in the continuous monitoring of emissions of small amounts of hazardous gases into the atmosphere, for the detection of gases related to food quality [1] (e.g. ethanol in fermentation processes, ethylene in fruit ripeness determination/monitoring), or to diagnose illness through breath [2]. The ability of a MOS sensor to detect the presence of chemicals relies on the interaction between gas molecules and the surface of the sensing film. This interaction is affected by many factors such as the temperature of operation, the gas being analysed, the sensor geometry and packaging [3]. Gas detection is enabled by a change in the electric resistance arising from a surface phenomenon. The reactivity of a surface is dependent on its characteristics
∗ Corresponding author. Tel.: +34-977-558502; fax: +34-977-559605. E-mail address: [email protected] (E. Llobet).
0925-4005/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0925-4005(03)00648-8
that is, on its elemental composition including any doping, or impurity constituents, on its electronic and defect structure and on its microstructure [4–8]. Changes in the surface characteristics of the active layer can induce a change on the sensor performance. It has been reported that improvement of the sensing properties (selectivity and sensitivity) of metal oxides can be achieved by the addition of a small amount of noble metals to the active layer [4]. Metal additives, such as Pd and Pt, are dispersed on the oxide as activators or sensitizers to improve the gas selectivity and to lower the operating temperature [5,9,10]. The addition of a noble metal results in changes in the electronic states of the active layer and can also modify the microstructure of the base material, and the grain growth mechanism. Among the various metal oxides that can be used as gas-sensitive materials, SnO2 has been shown to possess characteristics (such as sticking coefficient) that can be tailored (by adding small amount of noble metals that have higher sticking coefficient) to improve its sensing properties [11]. Therefore, the use of SnO2 as an active material for gas sensing is
68
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
widely related to its combined use with catalytic metals. Under noble metal loading, it is expected that clusters form at the surface of SnO2 , such as those observed in the case of high loading Pt and Pd [9,12]. These clusters will be in metallic or oxidised forms depending on the noble metal, the loading process, the gases reacting at the surface and the sensor operating temperature. Anyway, the contact of the noble metal additive with the semi-conducting oxide changes the depletion region at the semiconductor and modifies surface barrier height. This barrier is fully characterised by electron affinity, the work function values and the density of surface states that are located inside the energy band gap, this quantities strongly influence the sensor performance [9]. It has been claimed that the presence of metals such as Pt or Pd allows for a significant decrease in the operating temperature of the sensors, the enhancement of sensitivity to different gases and to reduce response time [9]. Two mechanisms have been suggested to explain the observed results: chemical and electronic sensitisation [11]. In chemical sensitisation (typical for Pt), the action of noble metal atoms, which is expected to form metallic clusters at the surface of the SnO2 , is to improve the gas-semiconductor reaction by a catalytic effect on the Pt clusters. These clusters on the semiconductor surface have a higher sticking coefficient to gases than SnO2 , and dissociate nearly all the gas molecules, spilling the products over the semiconductor surface. In electronic sensitisation (typically attributed to Pd), there is no mass transfer between the cluster and the semiconductor. Instead, the chemical state of the cluster changes on contact with the gas, inducing the corresponding change in the electronic state of the semiconductor [11]. An important aspect that has to be taken into account when working with metal oxide loaded with noble metal atoms is their distribution through the active layer. It is expected that a homogeneous distribution through the metal oxide film would be advantageous for gas sensing [13]. However, the actual distribution of additives heavily depends on the method used to introduce them and on the subsequent thermal treatments performed [9]. This work presents a comparison of two procedures used to obtain tin oxide films loaded with Pt atoms. The influence of these procedures in the structure of the tin oxide layers and their performance as ethanol sensors is investigated and discussed. For this purpose, sensors were fabricated based on un-loaded and Pt-loaded tin oxide phases deposited onto alumina substrates by a screen-printing technique. Morphological and composition studies were conducted by scanning electron microscopy (SEM), energy dispersive X-ray (EDX), and XPS and gas sensitivity studies to ethanol vapours and ambient moisture were performed. The fabricated sensors show high sensitivity to ethanol, moderate water vapour cross-sensitivity and have the potential of becoming a low-cost detector for traces of ethanol at the ppb level (e.g. the early-stage detection of unwanted fermentation processes in food storage chambers) [1].
2. Sensor fabrication Sensors were fabricated by thick-film technology onto alumina substrates. Fig. 1 shows the two sides of a sensor. On the front side, the sensitive film coats interdigited gold electrodes. A Pt heater and a temperature sensitive meander are printed onto the backside of the substrate. The procedure for the fabrication of sensors is described below. In the first step, interdigited gold electrodes were deposited by using a commercial conductive paste (ESL 8884). Once printed, the substrates were left at room temperature for levelling, in order to eliminate defects in the structure before firing. After levelling, the substrates were dried (15 min at 125 ◦ C) and fired in a belt-furnace during 1 h with a peak temperature of 950 ◦ C. In the second step, a heating element and a temperature sensitive meander were printed using a commercial platinum paste (Heraeus C3657). After levelling at room temperature, the substrates were dried (20 min at 150 ◦ C). Gold pads were printed and the firing of the Pt-heater and sensing resistance and gold pads was performed during 1 h. The peak temperature was set to 875 ◦ C. In the third step, the gas-sensitive layers were deposited onto the electrodes. The sensors were prepared by using a commercial SnO2 powder (Sigma–Aldrich) mixed with 1% (w/w) of glass frit. Glass frit ensures good connection between the tin oxide particles and also between the particles and the substrate. In addition an organic binder (methylmetacrylate) and an organic solvent (therpineol) were used to achieve good rheology properties and viscosity, respectively [14]. Two different active layers were printed on the substrates. Two-thirds of the substrates were printed with the paste described above (un-loaded tin oxide). One-third of the substrates were printed with a paste prepared as follows: the original paste (un-loaded tin oxide) was mixed with 3% (w/w) of a Pt paste, which resulted in Pt-loaded
Fig. 1. Sensor design: (a) front side (b) backside.
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
tin oxide films. After the deposition of the active layer, all the sensors were left 20 min at room temperature (25 ◦ C) for levelling. They were subsequently dried (5 min at 90 ◦ C followed by 10 min at 150 ◦ C) and, then fired in a belt-furnace (1.5 h at 600 ◦ C). After that, the sensors were slowly cooled to ambient temperature. Finally, half of the un-loaded tin oxide sensors had a thin layer (15 nm) of Pt deposited by RF magnetron sputtering on top of the tin oxide layer. After Pt deposition, the sensors were kept at 450 ◦ C for 1 h to promote diffusion of the Pt inside the tin oxide film. Therefore, three sets of sensors were available after the fabrication process. Set I consisted of eight Pt-loaded tin oxide sensors (Pt added by sputtering after printing), set II consisted of eight Pt-loaded tin oxide sensors (Pt added before printing) and set III consisted of eight un-loaded tin oxide sensors.
3. Experimental The chemical composition of the active layers was determined by XPS. The XPS measurements were performed with a system equipped with a hemispherical electron energy analyser. The photon source was a monochromatised Al K␣ line (hν = 1486.6 eV). The resolution of the system (source + analyser) was 0.6 eV. Charging effects were neutralised using a flood gun operated at 2 eV kinetic energy. The surface morphology and homogeneity of the samples were investigated with a JSM6400 SEM operated at 30 keV, equipped with an energy dispersive X-ray spectroscopy (EDX) detector. To investigate the gas sensing properties of the fabricated samples, the sensors were kept into a thermally
controlled test chamber (±1 ◦ C). The operating temperature of the sensors was successively set to 250, 300 and 350 ◦ C. This was done by applying a voltage to the sensors’ heating element. The ethanol and water needed to create the desired concentrations of ethanol vapours and moisture were introduced into the test chamber by using high precision chromatographic syringes. Dry air was used as a reference gas. Ethanol was injected in the liquid phase (in successive steps) to create concentrations of 1, 10, 100 and 1000 ppm when volatilised inside the sensor chamber (a fan was used to homogenise the gas mixture inside the chamber). The first injection of ethanol was made 50 s after the data acquisition started. The following injections were made at 150, 250 and 350 s, respectively. Hundred seconds after the last injection, the test chamber was cleaned by flushing it with dry air. Between measurements, a pause of 30 min was made for the sensors to recover their baseline resistance. The effect of moisture on the response of sensors was investigated by performing measurements at different humidity levels (RH varied between 15 and 85% at 30 ◦ C). The electrical resistance of the sensors was monitored using a resistance acquisition system HP 34970A, and stored and processed in a PC. During all the measurements a commercially available Taguchi TGS 822 sensor (Figaro Inc., Japan) was used as reference. This sensor is designed to detect volatile organic compounds (e.g. ethanol). Fig. 2 shows a typical response curve of our Pt-doped SnO2 sensors (set II), working at 300 ◦ C to successive injections of ethanol vapours. It can be seen that the Pt-doped tin oxide sensors are far more sensitive to ethanol than the commercial sensor. The time needed to reach 90% of the steady state response is around 15 s.
1.0E+05
10 ppm
Cleaning of the test chamber with dry air
100 ppm Commecial TGS 822
Resistance
1.0E+04
1000 ppm 8 Pt-doped (set II)
1.0E+03
1.0E+02 1
51
69
101 151 201 251 301 351 401 451 501 Time (s)
Fig. 2. Response of the eight Pt-doped SnO2 sensors from set II working at 300 ◦ C to successive injections of ethanol.
70
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
4. Results and discussion 4.1. XPS analysis The first step when working with tin oxides is to determine the type of oxide under analysis, i.e. to distinguish between SnO and SnO2 . This can be made only by an analysis of the valence-band (VB) region [15], and by measuring the energy difference between the Sn 4d peak and the first peak in the VB toward low EB . Fig. 3 shows the typical XPS spectra used in the identification. The energy difference between the Sn 4d peak and the O 2p—derived structure near EB = 4.5 eV measured on all analysed samples ranged from 21.23 to 21.3 eV. This result is in close agreement with the reported value (21.1 eV) for SnO2 [16]. It indicates that the doping procedures used did not modify the type of tin oxide. The Sn 3d core level spectra recorded on the samples showed two components associated with the Sn 3d5/2 and Sn 3d3/2 spin-orbit doublet separated 8.41 eV, in good agreement with the tabulated value [17] (Fig. 4). A close inspection of the Sn 3d core level spectra recorded on the samples showed that the spectra could be fit using a single Voigt profile per component centred respectively near 486 and 494.5 eV. This suggests that the films are mainly formed by Sn4+ species [9]. The presence of +2 and +4 metallic oxide states of the Pt atom was inferred by XPS spectra (not shown) on samples of both sets I and II. It has been reported, although there is no definitive evidence, that metals atoms fixed at the surface and surrounded by absorbed oxygen atoms are the origin of the +4 chemical state of the introduced noble metal additive and the +2 chemical states correspond to metal atoms surrounded by Sn and O atoms in the interior of the SnO2 grains [9].
Fig. 4. Typical XPS spectrum of Sn 3d recorded on the studied samples.
In order to verify the homogeneity of the in-depth Pt concentration, XPS analysis of sets I and II was carried in two steps. In the first step, the samples were analysed as they were deposited. In the second step, samples were analysed after a thin layer had been mechanically removed with a razorblade. Sputtering could not be used for the removal of thin layers as it is known to produce preferential sputtering, thus altering film composition. After scraping the sensor surface, XPS shows that the concentration (1.2%) of the additive of the active layer of set II samples did not change; we inferred that the sensor preparation method lead to a homogeneous distribution of Pt atoms through the base material. On the other hand, for samples in set I, it was observed that the concentration of the Pt atoms changed while the surface layer was removed, i.e. this preparation procedure lead to a non-homogeneously doped active layer. The concentration of Pt in the first layer surface (12.5%) was found to be nearly 10 times higher than on the bottom layers (1.1%). The concentration analysis of the tin and oxygen atoms confirmed the stoichiometry of the prepared SnO2 films. 4.2. SEM and EDX analysis
Fig. 3. Typical XPS spectrum of valence-band and near-core region recorded on the studied samples. The binding energy scale refers to the Fermi level. The VB region has been magnified with respect to the Sn 4d region.
To understand how the doping affects the morphology of the SnO2 active layers, a structural characterisation based on scanning electron microscopy analysis was performed. The micrographs recorded for the different sensors showed that the films are essentially in-homogeneous, and made up of grains (grain size is around 60 nm) and voids (see Fig. 5). The voids within the film structure provide direct conduits for gas molecules to flow in from the environment. An EDX analysis showed that the grains and the bottom of the voids contained tin (the equipment used was not suitable for detecting the presence of oxygen). Fig. 5a–c shows the surface of Pt-loaded (by sputtering), Pt-loaded (before printing) and un-loaded tin oxide, respectively. From visual inspection of the SEM micrographs, it
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
71
Fig. 5. SEM analysis of the samples: (a) Pt-doped tin oxide from set I, (b) Pt-doped tin oxide from set II, (c) un-doped tin oxide from set III.
is clear that no significant changes in the morphology of the samples can be detected. Therefore, the two different procedures used to obtain Pt-loaded tin oxide films did not significantly alter the morphology at the surface. To analyse the homogeneity of the Pt distribution within the doped films, images using back-scattering electrons were used. The samples analysed were cut and their in-depth morphology was studied. Film thickness varied between 20 and 25 m. In back-scattered electron imaging, greyscale intensity is characteristic of atomic number, and the distribution of Pt in the oxide layer is revealed (Fig. 6). The homogeneous aspect of the image that corresponds to films doped by adding a Pt paste to the tin oxide paste (set II) indicates the Pt is homogeneously distributed within the tin oxide
Fig. 6. SEM analysis (back-scattered electrons image) of a section of Pt-doped samples. (a) Sample from set I where the layer near the surface containing Pt is clearly visible, (b) sample from set II.
72
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
layer. On the other hand, the films that were doped by Pt sputtering (set I) show a sharp contrast in intensity, which suggests that Pt is mainly concentrated in a layer (around 3 m thick) at the surface of the tin oxide film. These results are in good agreement with the XPS analysis. R/R_15%
1.40
4.3. Sensing properties The ethanol sensing properties of the screen-printed SnO2 gas sensors were studied. Ethanol was chosen to run sensitivity tests because it is well known that tin oxide is very sensitive to this species. The sensitivities of the eight sensors in each set were characterised. Sensitivity was defined as the ratio between Ra /Rg , where Ra is the resistance of the sensor in the presence of clean air and Rg is the resistance of the sensor, when either ethanol or water were introduced into the sensor chamber. Since tin oxide behaves as an n-type semiconductor and ethanol vapour is a reducing species, Rg is lower than Ra . Therefore, sensitivity is always higher than or equal to unity. For each set of sensors, the ratio between the standard deviation of sensitivity and the mean of sensitivity (over eight sensors in each group and five replicate measurements) was computed. This ratio was found to be below 3.8 × 10−3 , which shows that sensors within a given set behaved similarly (i.e. sensors showed a fairly good reproducibility). Fig. 7 reports the sensitivity results, averaged over the eight sensors within each set. Sensitivity heavily depends on the working temperature of the sensors. The sensors were operated at 250, 300 and 350 ◦ C, exception made of the commercial sensor, which was always operated as recommended by the manufacturer (i.e. applying 5 V to its heating resistor). The sensors (no matter the set to which they belonged) reached the highest sensitivity to ethanol at the working temperature of 300 ◦ C.
Set I Set II Set III TGS822 1.20
1.00 15%
50%
85%
Relative humidity
Fig. 8. Response to water vapour of the SnO2 sensors operated at 300 ◦ C. The sensitivity to moisture of Pt-loaded sensors (sets II and II) is small, especially compared with the sensitivity of the commercial sensor.
Tin oxide films containing Pt (sets I and II) appear more sensitive to ethanol than un-loaded tin oxide films (set III). However, the sensors that were loaded by mixing the tin oxide paste with a Pt paste before printing (set II) show even higher sensitivities (typically more than four times higher) than sensors loaded by Pt sputtering. These results prove (according to XPS results) that a homogeneous distribution of Pt in the tin oxide layer, rather than a high concentration of Pt near the tin oxide surface, is advantageous to enhance sensitivity to ethanol. Pt-loaded sensors operated at 300 ◦ C are far more sensitive to ethanol than the commercial sensor. For example, the sensitivity of set II sensors is higher than the sensitivity of the TGS 822 sensor by a factor that ranges between 30 and 60, depending on ethanol concentration. Furthermore, since the resistance of set II sensors in the
1000
250”C 300”C 350”C
Set II Set I
Rg/Ra
100
10
TGS 822 Set III
1 1
10 Concentration, ppm
100
1000
Fig. 7. Sensitivity to ethanol of the different sensors when operated at 250, 300 and 350 ◦ C. Set I: tin oxide loaded with Pt by sputtering, set II: tin oxide loaded with Pt before printing, set III: un-loaded tin oxide.
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
presence of 1 ppm of ethanol is 37 times lower than their resistance in clean air, these sensors are well suited for the detection of ethanol at the ppb level. The resistance value of a metal oxide gas sensor has a strong dependency with ambient moisture. This causes interference when measuring the presence of other gases. To study the effect of humidity, its level inside the sensor chamber was varied from 15 to 85% and the drift in the sensor baseline resistance recorded. The sensors were operated at 300 ◦ C, which was found to be the optimal temperature for ethanol detection (see above). The results of this study are shown in Fig. 8, where R/R 15% represents the ratio between the sensor resistance at a given moisture level and the resistance at 15% RH. All the sensors behave similarly in the presence of water vapour, the effect of which can be considered as moderate and, significantly lower than for the commercial sensor.
5. Conclusions A comparison of two different procedures to load tin oxide with Pt and their influence in the morphology of the films and in the performance of the resulting gas sensors has been presented. Pt-loaded tin oxide films significantly outperformed un-loaded films in terms of sensitivity to ethanol vapours. In one set of samples, the screen-printed tin oxide layer was loaded with Pt by using RF magnetron sputtering and a subsequent thermal treatment to promote the diffusion of Pt into the tin oxide layer. In the second set of samples, a Pt paste was added to the tin oxide paste before screen-printing onto the sensor substrate. The sensors prepared by this second method showed sensitivity to ethanol that was four times higher than the sensitivity of the sensors prepared by the first method and 12 times higher than the sensitivity of un-loaded sensors. XPS and SEM measurement showed that this behaviour can be associated with the distribution of the Pt atoms within the tin oxide layer. For the sensors doped by sputtering, the Pt atoms were found to mostly remain at the surface of the active layer. On the other hand, the sensors produced by mixing the Pt and the SnO2 pastes showed a more homogeneous distribution of the doping Pt atoms in the resulting films. The Pt-loaded SnO2 material shows high sensitivity and fast response to ethanol compared with pure SnO2 . The method that allows to homogeneously load the films is preferred since it leads to the highest sensitivity to ethanol, while keeping moderate the cross-sensitivity to water vapour. Therefore, these
73
sensors can be used in applications where there is a need for low-cost, real time monitoring of ethanol at the sub-ppm level.
References [1] B.P.J. de Lacy Costello, R. Ewen, N. Guernion, N. Ratcliffe, Highly sensitive mixed oxide sensors for the detection of ethanol, Sens. Actuators B 87 (2002) 207–210. [2] J.W. Gardner, H.W. Shin, E.L. Hines, An electronic nose system to diagnose illness, Sens. Actuators B 70 (2000) 19–24. [3] C.A. Papadopoulos, D.S. Vlachos, J.N. Avaritsiotis, Comparative study of various metal-oxide-based gas-sensor architectures, Sens. Actuators B 32 (1996) 61–69. [4] M. Sauvan, C. Pijolat, Selectivity improvement of SnO2 films by superficial metallic films, Sens. Actuators B 58 (1999) 295–301. [5] K.D. Schierbaum, J. Geiger, U. Weimar, W. Göpel, Specific palladium and platinum doping for SnO2 -based thin film sensor arrays, Sens. Actuators B 13 (1993) 143–147. [6] M.A. Ponce, C.M. Aldao, M.S. Castro, Influence of particle size on the conductance of SnO2 thick films, J. Eur. Ceramic Soc. 23 (2003) 2105–2111, available online 18 March 2003. [7] M. de la, L. Olvera, R. Asomoza, SnO2 and SnO2 : Pt thin films used as gas sensors, Sens. Actuators B 45 (1997) 49–53. [8] T. Maosong, D. Guorui, G. Dingsan, Surface modification of oxide thin film and its gas-sensing properties, Appl. Surf. Sci. 171 (2001) 226–230. [9] A. Diéguez, A. Vilà, A. Cabot, A. Romano-Rodr´ıguez, J.R. Morante, J. Kappler, N. Bˆarsan, U. Weimar, W. Göpel, Influence on the gas sensor performances of the metal chemical states introduced by impregnation of calcinated SnO2 sol–gel nanocrystal, Sens. Actuators B 68 (2000) 94–99. [10] A. Heilig, N. Barsan, U. Weimar, W. Göpel, Selectivity enhancement of SnO2 gas sensors: simultaneous monitoring of resistances and temperatures, Sens. Actuators B 58 (1999) 302–309. [11] A. Cabot, J. Arbiol, J.R. Morante, U. Weimar, N. Bˆarsan, W. Göpel, Analysis of noble metal catalytic additives introduced by impregnation of as obtained SnO2 sol–gel nanocrystals for gas sensors, Sens. Actuators B 70 (2000) 87–100. [12] A. Diéguez, A. Romano-Rodr´ıguez, J.R. Morante, U. Weimar, M. Schweizer-Berberich, W. Göpel, Morphological analysis of nanocrystalline SnO2 for gas sensor application, Sens. Actuators B 31 (1996) 1–8. [13] I. Kocemba, S. Szafran, J. Rynkowski, T. Paryjczak, The properties of strongly pressed tin oxide-based gas sensors, Sens. Actuators B 79 (1) (2001) 28–32. [14] M.R. Haskard, R. Malcolm, Thick-film Technology and Applications, Port Erin, Electrochemical Publications, 1997. [15] J.-M. Themlin, M. Chta¨ıb, L. Henrard, P. Lambin, J. Darville, J.M. Gilles, Characterization of tin oxides by X-ray-photoemission spectroscopy, Phys. Rev. B 46 (4) (1992) 2460–2466. [16] P.M.A. Sherwood, Valence-band spectra of tin oxides interpreted by X␣ calculations, Phys. Rev. B 41 (14) (1990) 10151–10154. [17] C.L. Lau, G.K. Werthein, Oxidation of tin: an ESCA study, J. Vaccum Sci. Technol. 15 (2) (1978) 622–624.
Influence of the doping method on the sensitivity of Pt-doped screen-printed SnO2 sensors C. Bittencourt a , E. Llobet b,∗ , P. Ivanov b , X. Correig b , X. Vilanova b , J. Brezmes b , J. Hubalek c , K. Malysz c , J.J. Pireaux a , J. Calderer d a
d
Laboratoire Interdisciplinaire de Spectroscopie Electronique, Institute for Studies in Interface Sciences, Falcultés Universitaires Notre Dame de la Paix, 61 rue de Bruxelles, B-5000 Namur, Belgium b Departament d’Enginyeria Electrònica, Universitat Rovira i Virgili, Avinguda Pa¨ısos Catalans, 26, 43007 Tarragona, Spain c Faculty of Electrical Engineering and Communication, Brno University of Technology. Údoln´ı 53, 602 00 Brno, Czech Republic Department d’Enginyeria Electrònica, Universitat Politècnica de Catalunya, Campus Nord, Gran Capità s/n, 08034 Barcelona, Spain Received 2 April 2003; received in revised form 25 July 2003; accepted 25 July 2003
Abstract In this work, we study the influence of the introduction method of Pt atoms on the sensitivity to traces of ethanol of Pt-doped SnO2 sensors. The tin oxide films were obtained by a screen-printing process. Two different methods were employed to introduce Pt atoms on SnO2 films. In the first one, the Pt atoms were added to the screen-printed tin oxide layer by using RF magnetron sputtering and a subsequent thermal treatment. The second method consisted of mixing SnO2 and Pt pastes before the screen-printing process. The different active layers (including un-doped tin oxide) were carefully examined relative to their sensitivity to ethanol at different working temperatures. Sensors prepared by the second method showed sensitivity to ethanol four times higher than one of the sensors prepared by the first method and 12 times higher than un-doped sensors. XPS and scanning electron microscopy (SEM) measurements showed that this behaviour could be associated with the spatial distribution of the doping elements within the tin oxide film. While in Pt-sputtered sensors most of the Pt atoms were found at the surface of the active layer, for the sensors made by mixing Pt and SnO2 pastes, a homogeneous distribution of the Pt atoms was observed. These sensors show high sensitivity and fast response time to ethanol vapours, with a detection limit in the ppb range. © 2003 Elsevier B.V. All rights reserved. Keywords: SnO2 sensors; Screen-printing; Pt loading; Detection limit
1. Introduction Metal-oxide semiconductor (MOS) gas sensors are a low-cost option in the continuous monitoring of emissions of small amounts of hazardous gases into the atmosphere, for the detection of gases related to food quality [1] (e.g. ethanol in fermentation processes, ethylene in fruit ripeness determination/monitoring), or to diagnose illness through breath [2]. The ability of a MOS sensor to detect the presence of chemicals relies on the interaction between gas molecules and the surface of the sensing film. This interaction is affected by many factors such as the temperature of operation, the gas being analysed, the sensor geometry and packaging [3]. Gas detection is enabled by a change in the electric resistance arising from a surface phenomenon. The reactivity of a surface is dependent on its characteristics
∗ Corresponding author. Tel.: +34-977-558502; fax: +34-977-559605. E-mail address: [email protected] (E. Llobet).
0925-4005/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0925-4005(03)00648-8
that is, on its elemental composition including any doping, or impurity constituents, on its electronic and defect structure and on its microstructure [4–8]. Changes in the surface characteristics of the active layer can induce a change on the sensor performance. It has been reported that improvement of the sensing properties (selectivity and sensitivity) of metal oxides can be achieved by the addition of a small amount of noble metals to the active layer [4]. Metal additives, such as Pd and Pt, are dispersed on the oxide as activators or sensitizers to improve the gas selectivity and to lower the operating temperature [5,9,10]. The addition of a noble metal results in changes in the electronic states of the active layer and can also modify the microstructure of the base material, and the grain growth mechanism. Among the various metal oxides that can be used as gas-sensitive materials, SnO2 has been shown to possess characteristics (such as sticking coefficient) that can be tailored (by adding small amount of noble metals that have higher sticking coefficient) to improve its sensing properties [11]. Therefore, the use of SnO2 as an active material for gas sensing is
68
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
widely related to its combined use with catalytic metals. Under noble metal loading, it is expected that clusters form at the surface of SnO2 , such as those observed in the case of high loading Pt and Pd [9,12]. These clusters will be in metallic or oxidised forms depending on the noble metal, the loading process, the gases reacting at the surface and the sensor operating temperature. Anyway, the contact of the noble metal additive with the semi-conducting oxide changes the depletion region at the semiconductor and modifies surface barrier height. This barrier is fully characterised by electron affinity, the work function values and the density of surface states that are located inside the energy band gap, this quantities strongly influence the sensor performance [9]. It has been claimed that the presence of metals such as Pt or Pd allows for a significant decrease in the operating temperature of the sensors, the enhancement of sensitivity to different gases and to reduce response time [9]. Two mechanisms have been suggested to explain the observed results: chemical and electronic sensitisation [11]. In chemical sensitisation (typical for Pt), the action of noble metal atoms, which is expected to form metallic clusters at the surface of the SnO2 , is to improve the gas-semiconductor reaction by a catalytic effect on the Pt clusters. These clusters on the semiconductor surface have a higher sticking coefficient to gases than SnO2 , and dissociate nearly all the gas molecules, spilling the products over the semiconductor surface. In electronic sensitisation (typically attributed to Pd), there is no mass transfer between the cluster and the semiconductor. Instead, the chemical state of the cluster changes on contact with the gas, inducing the corresponding change in the electronic state of the semiconductor [11]. An important aspect that has to be taken into account when working with metal oxide loaded with noble metal atoms is their distribution through the active layer. It is expected that a homogeneous distribution through the metal oxide film would be advantageous for gas sensing [13]. However, the actual distribution of additives heavily depends on the method used to introduce them and on the subsequent thermal treatments performed [9]. This work presents a comparison of two procedures used to obtain tin oxide films loaded with Pt atoms. The influence of these procedures in the structure of the tin oxide layers and their performance as ethanol sensors is investigated and discussed. For this purpose, sensors were fabricated based on un-loaded and Pt-loaded tin oxide phases deposited onto alumina substrates by a screen-printing technique. Morphological and composition studies were conducted by scanning electron microscopy (SEM), energy dispersive X-ray (EDX), and XPS and gas sensitivity studies to ethanol vapours and ambient moisture were performed. The fabricated sensors show high sensitivity to ethanol, moderate water vapour cross-sensitivity and have the potential of becoming a low-cost detector for traces of ethanol at the ppb level (e.g. the early-stage detection of unwanted fermentation processes in food storage chambers) [1].
2. Sensor fabrication Sensors were fabricated by thick-film technology onto alumina substrates. Fig. 1 shows the two sides of a sensor. On the front side, the sensitive film coats interdigited gold electrodes. A Pt heater and a temperature sensitive meander are printed onto the backside of the substrate. The procedure for the fabrication of sensors is described below. In the first step, interdigited gold electrodes were deposited by using a commercial conductive paste (ESL 8884). Once printed, the substrates were left at room temperature for levelling, in order to eliminate defects in the structure before firing. After levelling, the substrates were dried (15 min at 125 ◦ C) and fired in a belt-furnace during 1 h with a peak temperature of 950 ◦ C. In the second step, a heating element and a temperature sensitive meander were printed using a commercial platinum paste (Heraeus C3657). After levelling at room temperature, the substrates were dried (20 min at 150 ◦ C). Gold pads were printed and the firing of the Pt-heater and sensing resistance and gold pads was performed during 1 h. The peak temperature was set to 875 ◦ C. In the third step, the gas-sensitive layers were deposited onto the electrodes. The sensors were prepared by using a commercial SnO2 powder (Sigma–Aldrich) mixed with 1% (w/w) of glass frit. Glass frit ensures good connection between the tin oxide particles and also between the particles and the substrate. In addition an organic binder (methylmetacrylate) and an organic solvent (therpineol) were used to achieve good rheology properties and viscosity, respectively [14]. Two different active layers were printed on the substrates. Two-thirds of the substrates were printed with the paste described above (un-loaded tin oxide). One-third of the substrates were printed with a paste prepared as follows: the original paste (un-loaded tin oxide) was mixed with 3% (w/w) of a Pt paste, which resulted in Pt-loaded
Fig. 1. Sensor design: (a) front side (b) backside.
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
tin oxide films. After the deposition of the active layer, all the sensors were left 20 min at room temperature (25 ◦ C) for levelling. They were subsequently dried (5 min at 90 ◦ C followed by 10 min at 150 ◦ C) and, then fired in a belt-furnace (1.5 h at 600 ◦ C). After that, the sensors were slowly cooled to ambient temperature. Finally, half of the un-loaded tin oxide sensors had a thin layer (15 nm) of Pt deposited by RF magnetron sputtering on top of the tin oxide layer. After Pt deposition, the sensors were kept at 450 ◦ C for 1 h to promote diffusion of the Pt inside the tin oxide film. Therefore, three sets of sensors were available after the fabrication process. Set I consisted of eight Pt-loaded tin oxide sensors (Pt added by sputtering after printing), set II consisted of eight Pt-loaded tin oxide sensors (Pt added before printing) and set III consisted of eight un-loaded tin oxide sensors.
3. Experimental The chemical composition of the active layers was determined by XPS. The XPS measurements were performed with a system equipped with a hemispherical electron energy analyser. The photon source was a monochromatised Al K␣ line (hν = 1486.6 eV). The resolution of the system (source + analyser) was 0.6 eV. Charging effects were neutralised using a flood gun operated at 2 eV kinetic energy. The surface morphology and homogeneity of the samples were investigated with a JSM6400 SEM operated at 30 keV, equipped with an energy dispersive X-ray spectroscopy (EDX) detector. To investigate the gas sensing properties of the fabricated samples, the sensors were kept into a thermally
controlled test chamber (±1 ◦ C). The operating temperature of the sensors was successively set to 250, 300 and 350 ◦ C. This was done by applying a voltage to the sensors’ heating element. The ethanol and water needed to create the desired concentrations of ethanol vapours and moisture were introduced into the test chamber by using high precision chromatographic syringes. Dry air was used as a reference gas. Ethanol was injected in the liquid phase (in successive steps) to create concentrations of 1, 10, 100 and 1000 ppm when volatilised inside the sensor chamber (a fan was used to homogenise the gas mixture inside the chamber). The first injection of ethanol was made 50 s after the data acquisition started. The following injections were made at 150, 250 and 350 s, respectively. Hundred seconds after the last injection, the test chamber was cleaned by flushing it with dry air. Between measurements, a pause of 30 min was made for the sensors to recover their baseline resistance. The effect of moisture on the response of sensors was investigated by performing measurements at different humidity levels (RH varied between 15 and 85% at 30 ◦ C). The electrical resistance of the sensors was monitored using a resistance acquisition system HP 34970A, and stored and processed in a PC. During all the measurements a commercially available Taguchi TGS 822 sensor (Figaro Inc., Japan) was used as reference. This sensor is designed to detect volatile organic compounds (e.g. ethanol). Fig. 2 shows a typical response curve of our Pt-doped SnO2 sensors (set II), working at 300 ◦ C to successive injections of ethanol vapours. It can be seen that the Pt-doped tin oxide sensors are far more sensitive to ethanol than the commercial sensor. The time needed to reach 90% of the steady state response is around 15 s.
1.0E+05
10 ppm
Cleaning of the test chamber with dry air
100 ppm Commecial TGS 822
Resistance
1.0E+04
1000 ppm 8 Pt-doped (set II)
1.0E+03
1.0E+02 1
51
69
101 151 201 251 301 351 401 451 501 Time (s)
Fig. 2. Response of the eight Pt-doped SnO2 sensors from set II working at 300 ◦ C to successive injections of ethanol.
70
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
4. Results and discussion 4.1. XPS analysis The first step when working with tin oxides is to determine the type of oxide under analysis, i.e. to distinguish between SnO and SnO2 . This can be made only by an analysis of the valence-band (VB) region [15], and by measuring the energy difference between the Sn 4d peak and the first peak in the VB toward low EB . Fig. 3 shows the typical XPS spectra used in the identification. The energy difference between the Sn 4d peak and the O 2p—derived structure near EB = 4.5 eV measured on all analysed samples ranged from 21.23 to 21.3 eV. This result is in close agreement with the reported value (21.1 eV) for SnO2 [16]. It indicates that the doping procedures used did not modify the type of tin oxide. The Sn 3d core level spectra recorded on the samples showed two components associated with the Sn 3d5/2 and Sn 3d3/2 spin-orbit doublet separated 8.41 eV, in good agreement with the tabulated value [17] (Fig. 4). A close inspection of the Sn 3d core level spectra recorded on the samples showed that the spectra could be fit using a single Voigt profile per component centred respectively near 486 and 494.5 eV. This suggests that the films are mainly formed by Sn4+ species [9]. The presence of +2 and +4 metallic oxide states of the Pt atom was inferred by XPS spectra (not shown) on samples of both sets I and II. It has been reported, although there is no definitive evidence, that metals atoms fixed at the surface and surrounded by absorbed oxygen atoms are the origin of the +4 chemical state of the introduced noble metal additive and the +2 chemical states correspond to metal atoms surrounded by Sn and O atoms in the interior of the SnO2 grains [9].
Fig. 4. Typical XPS spectrum of Sn 3d recorded on the studied samples.
In order to verify the homogeneity of the in-depth Pt concentration, XPS analysis of sets I and II was carried in two steps. In the first step, the samples were analysed as they were deposited. In the second step, samples were analysed after a thin layer had been mechanically removed with a razorblade. Sputtering could not be used for the removal of thin layers as it is known to produce preferential sputtering, thus altering film composition. After scraping the sensor surface, XPS shows that the concentration (1.2%) of the additive of the active layer of set II samples did not change; we inferred that the sensor preparation method lead to a homogeneous distribution of Pt atoms through the base material. On the other hand, for samples in set I, it was observed that the concentration of the Pt atoms changed while the surface layer was removed, i.e. this preparation procedure lead to a non-homogeneously doped active layer. The concentration of Pt in the first layer surface (12.5%) was found to be nearly 10 times higher than on the bottom layers (1.1%). The concentration analysis of the tin and oxygen atoms confirmed the stoichiometry of the prepared SnO2 films. 4.2. SEM and EDX analysis
Fig. 3. Typical XPS spectrum of valence-band and near-core region recorded on the studied samples. The binding energy scale refers to the Fermi level. The VB region has been magnified with respect to the Sn 4d region.
To understand how the doping affects the morphology of the SnO2 active layers, a structural characterisation based on scanning electron microscopy analysis was performed. The micrographs recorded for the different sensors showed that the films are essentially in-homogeneous, and made up of grains (grain size is around 60 nm) and voids (see Fig. 5). The voids within the film structure provide direct conduits for gas molecules to flow in from the environment. An EDX analysis showed that the grains and the bottom of the voids contained tin (the equipment used was not suitable for detecting the presence of oxygen). Fig. 5a–c shows the surface of Pt-loaded (by sputtering), Pt-loaded (before printing) and un-loaded tin oxide, respectively. From visual inspection of the SEM micrographs, it
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
71
Fig. 5. SEM analysis of the samples: (a) Pt-doped tin oxide from set I, (b) Pt-doped tin oxide from set II, (c) un-doped tin oxide from set III.
is clear that no significant changes in the morphology of the samples can be detected. Therefore, the two different procedures used to obtain Pt-loaded tin oxide films did not significantly alter the morphology at the surface. To analyse the homogeneity of the Pt distribution within the doped films, images using back-scattering electrons were used. The samples analysed were cut and their in-depth morphology was studied. Film thickness varied between 20 and 25 m. In back-scattered electron imaging, greyscale intensity is characteristic of atomic number, and the distribution of Pt in the oxide layer is revealed (Fig. 6). The homogeneous aspect of the image that corresponds to films doped by adding a Pt paste to the tin oxide paste (set II) indicates the Pt is homogeneously distributed within the tin oxide
Fig. 6. SEM analysis (back-scattered electrons image) of a section of Pt-doped samples. (a) Sample from set I where the layer near the surface containing Pt is clearly visible, (b) sample from set II.
72
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
layer. On the other hand, the films that were doped by Pt sputtering (set I) show a sharp contrast in intensity, which suggests that Pt is mainly concentrated in a layer (around 3 m thick) at the surface of the tin oxide film. These results are in good agreement with the XPS analysis. R/R_15%
1.40
4.3. Sensing properties The ethanol sensing properties of the screen-printed SnO2 gas sensors were studied. Ethanol was chosen to run sensitivity tests because it is well known that tin oxide is very sensitive to this species. The sensitivities of the eight sensors in each set were characterised. Sensitivity was defined as the ratio between Ra /Rg , where Ra is the resistance of the sensor in the presence of clean air and Rg is the resistance of the sensor, when either ethanol or water were introduced into the sensor chamber. Since tin oxide behaves as an n-type semiconductor and ethanol vapour is a reducing species, Rg is lower than Ra . Therefore, sensitivity is always higher than or equal to unity. For each set of sensors, the ratio between the standard deviation of sensitivity and the mean of sensitivity (over eight sensors in each group and five replicate measurements) was computed. This ratio was found to be below 3.8 × 10−3 , which shows that sensors within a given set behaved similarly (i.e. sensors showed a fairly good reproducibility). Fig. 7 reports the sensitivity results, averaged over the eight sensors within each set. Sensitivity heavily depends on the working temperature of the sensors. The sensors were operated at 250, 300 and 350 ◦ C, exception made of the commercial sensor, which was always operated as recommended by the manufacturer (i.e. applying 5 V to its heating resistor). The sensors (no matter the set to which they belonged) reached the highest sensitivity to ethanol at the working temperature of 300 ◦ C.
Set I Set II Set III TGS822 1.20
1.00 15%
50%
85%
Relative humidity
Fig. 8. Response to water vapour of the SnO2 sensors operated at 300 ◦ C. The sensitivity to moisture of Pt-loaded sensors (sets II and II) is small, especially compared with the sensitivity of the commercial sensor.
Tin oxide films containing Pt (sets I and II) appear more sensitive to ethanol than un-loaded tin oxide films (set III). However, the sensors that were loaded by mixing the tin oxide paste with a Pt paste before printing (set II) show even higher sensitivities (typically more than four times higher) than sensors loaded by Pt sputtering. These results prove (according to XPS results) that a homogeneous distribution of Pt in the tin oxide layer, rather than a high concentration of Pt near the tin oxide surface, is advantageous to enhance sensitivity to ethanol. Pt-loaded sensors operated at 300 ◦ C are far more sensitive to ethanol than the commercial sensor. For example, the sensitivity of set II sensors is higher than the sensitivity of the TGS 822 sensor by a factor that ranges between 30 and 60, depending on ethanol concentration. Furthermore, since the resistance of set II sensors in the
1000
250”C 300”C 350”C
Set II Set I
Rg/Ra
100
10
TGS 822 Set III
1 1
10 Concentration, ppm
100
1000
Fig. 7. Sensitivity to ethanol of the different sensors when operated at 250, 300 and 350 ◦ C. Set I: tin oxide loaded with Pt by sputtering, set II: tin oxide loaded with Pt before printing, set III: un-loaded tin oxide.
C. Bittencourt et al. / Sensors and Actuators B 97 (2004) 67–73
presence of 1 ppm of ethanol is 37 times lower than their resistance in clean air, these sensors are well suited for the detection of ethanol at the ppb level. The resistance value of a metal oxide gas sensor has a strong dependency with ambient moisture. This causes interference when measuring the presence of other gases. To study the effect of humidity, its level inside the sensor chamber was varied from 15 to 85% and the drift in the sensor baseline resistance recorded. The sensors were operated at 300 ◦ C, which was found to be the optimal temperature for ethanol detection (see above). The results of this study are shown in Fig. 8, where R/R 15% represents the ratio between the sensor resistance at a given moisture level and the resistance at 15% RH. All the sensors behave similarly in the presence of water vapour, the effect of which can be considered as moderate and, significantly lower than for the commercial sensor.
5. Conclusions A comparison of two different procedures to load tin oxide with Pt and their influence in the morphology of the films and in the performance of the resulting gas sensors has been presented. Pt-loaded tin oxide films significantly outperformed un-loaded films in terms of sensitivity to ethanol vapours. In one set of samples, the screen-printed tin oxide layer was loaded with Pt by using RF magnetron sputtering and a subsequent thermal treatment to promote the diffusion of Pt into the tin oxide layer. In the second set of samples, a Pt paste was added to the tin oxide paste before screen-printing onto the sensor substrate. The sensors prepared by this second method showed sensitivity to ethanol that was four times higher than the sensitivity of the sensors prepared by the first method and 12 times higher than the sensitivity of un-loaded sensors. XPS and SEM measurement showed that this behaviour can be associated with the distribution of the Pt atoms within the tin oxide layer. For the sensors doped by sputtering, the Pt atoms were found to mostly remain at the surface of the active layer. On the other hand, the sensors produced by mixing the Pt and the SnO2 pastes showed a more homogeneous distribution of the doping Pt atoms in the resulting films. The Pt-loaded SnO2 material shows high sensitivity and fast response to ethanol compared with pure SnO2 . The method that allows to homogeneously load the films is preferred since it leads to the highest sensitivity to ethanol, while keeping moderate the cross-sensitivity to water vapour. Therefore, these
73
sensors can be used in applications where there is a need for low-cost, real time monitoring of ethanol at the sub-ppm level.
References [1] B.P.J. de Lacy Costello, R. Ewen, N. Guernion, N. Ratcliffe, Highly sensitive mixed oxide sensors for the detection of ethanol, Sens. Actuators B 87 (2002) 207–210. [2] J.W. Gardner, H.W. Shin, E.L. Hines, An electronic nose system to diagnose illness, Sens. Actuators B 70 (2000) 19–24. [3] C.A. Papadopoulos, D.S. Vlachos, J.N. Avaritsiotis, Comparative study of various metal-oxide-based gas-sensor architectures, Sens. Actuators B 32 (1996) 61–69. [4] M. Sauvan, C. Pijolat, Selectivity improvement of SnO2 films by superficial metallic films, Sens. Actuators B 58 (1999) 295–301. [5] K.D. Schierbaum, J. Geiger, U. Weimar, W. Göpel, Specific palladium and platinum doping for SnO2 -based thin film sensor arrays, Sens. Actuators B 13 (1993) 143–147. [6] M.A. Ponce, C.M. Aldao, M.S. Castro, Influence of particle size on the conductance of SnO2 thick films, J. Eur. Ceramic Soc. 23 (2003) 2105–2111, available online 18 March 2003. [7] M. de la, L. Olvera, R. Asomoza, SnO2 and SnO2 : Pt thin films used as gas sensors, Sens. Actuators B 45 (1997) 49–53. [8] T. Maosong, D. Guorui, G. Dingsan, Surface modification of oxide thin film and its gas-sensing properties, Appl. Surf. Sci. 171 (2001) 226–230. [9] A. Diéguez, A. Vilà, A. Cabot, A. Romano-Rodr´ıguez, J.R. Morante, J. Kappler, N. Bˆarsan, U. Weimar, W. Göpel, Influence on the gas sensor performances of the metal chemical states introduced by impregnation of calcinated SnO2 sol–gel nanocrystal, Sens. Actuators B 68 (2000) 94–99. [10] A. Heilig, N. Barsan, U. Weimar, W. Göpel, Selectivity enhancement of SnO2 gas sensors: simultaneous monitoring of resistances and temperatures, Sens. Actuators B 58 (1999) 302–309. [11] A. Cabot, J. Arbiol, J.R. Morante, U. Weimar, N. Bˆarsan, W. Göpel, Analysis of noble metal catalytic additives introduced by impregnation of as obtained SnO2 sol–gel nanocrystals for gas sensors, Sens. Actuators B 70 (2000) 87–100. [12] A. Diéguez, A. Romano-Rodr´ıguez, J.R. Morante, U. Weimar, M. Schweizer-Berberich, W. Göpel, Morphological analysis of nanocrystalline SnO2 for gas sensor application, Sens. Actuators B 31 (1996) 1–8. [13] I. Kocemba, S. Szafran, J. Rynkowski, T. Paryjczak, The properties of strongly pressed tin oxide-based gas sensors, Sens. Actuators B 79 (1) (2001) 28–32. [14] M.R. Haskard, R. Malcolm, Thick-film Technology and Applications, Port Erin, Electrochemical Publications, 1997. [15] J.-M. Themlin, M. Chta¨ıb, L. Henrard, P. Lambin, J. Darville, J.M. Gilles, Characterization of tin oxides by X-ray-photoemission spectroscopy, Phys. Rev. B 46 (4) (1992) 2460–2466. [16] P.M.A. Sherwood, Valence-band spectra of tin oxides interpreted by X␣ calculations, Phys. Rev. B 41 (14) (1990) 10151–10154. [17] C.L. Lau, G.K. Werthein, Oxidation of tin: an ESCA study, J. Vaccum Sci. Technol. 15 (2) (1978) 622–624.