Aerosol assisted chemical vapour deposition of gas sensitive SnO2 and Au-functionalised SnO2 nanorods via a non-catalysed vapour solid (VS) mechanism

Tin oxide nanorods (NRs) are vapour synthesised at relatively lower temperatures than previously reported and without the need for substrate pre-treatment, via a vapour-solid mechanism enabled using an aerosol-assisted chemical vapour deposition method. Results demonstrate that the growth of SnO2 NRs is promoted by a compression of the nucleation rate parallel to the substrate and a decrease of the energy barrier for growth perpendicular to the substrate, which are controlled via the deposition conditions. This method provides both single-step formation of the SnO2 NRs and their integration with silicon micromachined platforms, but also allows for in-situ functionalization of the NRs with gold nanoparticles via co-deposition with a gold precursor. The functional properties are demonstrated for gas sensing, with microsensors using functionalised NRs demonstrating enhanced sensing properties towards H2 compared to those based on non-functionalised NRs.

) nanoparticles segregated at the surface of tungsten oxide nanostructures.
There are very few reports on synthesis of tin oxide or tin oxide composites via AACVD, with most referring to the formation of thin films [33][34][35][36][37][38][39][40] rather than nanostructures. The precursors used for the AACVD of tin oxide include common precursors used in traditional CVD 41 , such as monobutyltin trichloride (C 4 H 9 SnCl 3 ) 38,40 , and tin(II) chloride dihydrate (SnCl 2 ·2H 2 O) 33 Here we demonstrate the AACVD of SnO 2 nanostructures, in the form of NRs, from a simple commercial tin precursor (SnCl 4 ·5H 2 O) at atmospheric pressure and exceptionally reduced process temperatures compared to existing VS-based CVD methods. In addition, to enhance the functionality of tin oxide we use the co-deposition opportunities afforded by AACVD to incorporate, in the same single-step, Au NPs at the surface of tin oxide NRs. These non-functionalised (SnO 2 ) and functionalised (Au@SnO 2 ) nanostructures are directly integrated into silicon micromachined platforms (fabricated using Micro-Electro-Mechanical-System (MEMS) technology) and validated for gas sensing, with sensors based on Au-functionalised SnO 2 NRs demonstrating enhanced properties.

Results
AACVD of SnO 2 . AACVD of SnCl 4 ·5H 2 O dissolved in acetone or methanol at temperatures between 300 and 675 °C resulted in the formation of adherent uniform greyish films on silicon wafers. XRD analysis confirmed the presence of tetragonal SnO 2 (P42/mnm, a = 4.7380 Å, c = 3.1870 Å, literature P42/mnm space group, a = 4.7382 Å, c = 3.1871 Å; ICCD card no. 41-1445) for all the films (Fig. 1), except for those deposited from a methanol-based solution at temperatures exceeding 500 °C (Fig. S1), which showed extra diffraction peaks corresponding to metallic tin (I41/amd, a = 5.8310 Å, c = 3.1820 Å, ICCD card no. 4-673). The morphology of the films was strongly dependent on the deposition temperature, solution concentration and solvent. For instance (Fig. 2), the films deposited from acetone-based solution (12.5 mM) displayed uneven morphology at 500 °C and porous structures with twinned crystallites at 600 °C, whereas the films deposited from lower solution concentration (8.5 mM) showed the formation of porous films with defined crystallites at 500 °C, a high density of leaf-like structures at 550 °C, and the formation of a high density of NRs at 600 °C. An increase in temperature from 600 °C to 620 °C improved the definition of the NRs, and above this temperature (up to 675 °C) the structures suffered deformations with less material deposited on the substrate, most likely due to homogeneous gas-phase reaction resulting in the formation of powder and the action of thermophoretic forces. A further decrease of solution concentration to 4.3 mM ended-up without visible film deposition. Similar test were carried out for methanol-based solutions, which produced porous films for solution concentrations of 12.5 and 8.5 mM in the temperature range of 300 °C and 675 °C (Fig. S2), and no visible deposition when the concentration was decreased to 4.3 mM.
Estimation of the energy activation for the perpendicular growth (E a d ) of the films 43 via the Arrhenius equation indicated lower E a d for films grown from acetone-based solutions (16.2 kJ/mol) compared to films grown from the methanol-based solutions (27.4 kJ/mol), the latter being in agreement with the value reported in the literature (24.12 kJ/mol) 44 for the CVD of planar SnO 2 films from SnCl 4 ·5H 2 O. SnO 2 NRs. Detailed examination of the NRs formed at 620 °C via AACVD suggest that the structures have a prism-like morphology terminated in a pyramidal cap (Fig. 3a), with a total length of ~700 nm and apparently wider sides at the highest part of the prism (~100 nm) in contrast to the base (Fig. 3b), corresponding to an aspect ratio of approximately 7. TEM imaging of the NRs end-cap indicates a similar morphology to that observed by SEM with an apex and lateral angle of 133.8° and 113.3°, respectively (Fig. 3c). These geometrical features are consistent with the models reported previously on the study of octahedral SnO 2 structures synthesized from  SnCl 4 ·5H 2 O via a hydrothermal process 45 , and suggest that the pyramidal caps and the prism-like base might be enclosed by the {111} and {110} facets, respectively.
The Sn 3d region of the XPS spectra for the SnO 2 NRs (Fig. 4a) exhibited Sn 3d 5/2 and Sn 3d 3/2 doublets for binding energies of ~486.3 and ~494.7 eV, respectively, confirming the presence of the Sn 4+ oxidation state 46,47 . In addition, the good symmetry of the peaks, showing no sub-components, suggests the absence of Sn 2+ . The O 1s core level (Fig. 4b), in contrast, indicates the presence of two components, with the main peak at ~531 eV, assigned to the lattice oxygen, and a shoulder at ~532.5 eV, which is assigned to contamination (organic fragments) on the samples in light of the lack of additional structures on the Sn 3d core level 46  Au@SnO 2 NRs. The films synthesised at 620 °C by co-deposition via AACVD of tin and gold precursors dissolved in acetone resulted in the formation of adherent uniform greyish films on silicon substrates, with similar appearance to those deposited from only the tin precursor. SEM imaging of the Au@SnO 2 films revealed a high density of NRs ( Fig. 5a) with similar features to those observed for intrinsic SnO 2 films deposited at 620 °C ( Fig. 2), though with apparently small particles dispersed at the surface. Similarly, TEM imaging of the functionalised NRs displayed NPs along the NR surface (Fig. 5b,c), which indicate the incorporation of Au NPs as noticed previously when co-depositing gold with tungsten oxide nanoneedles 29 . The NPs showed spherical morphologies and sizes up to 35 nm; an analysis of the size distribution of these particles via TEM was complex due to NRs tending to agglomerate on the TEM grids and the relatively high thickness of NRs.
XRD of the Au@SnO 2 NRs (Fig. 6a) revealed similar pattern than that observed for intrinsic SnO 2 NRs (Fig. 1) showing a tetragonal phase (P42mnm space group, a = 4.7382 Å, c = 3.1871 Å; ICCD card no. 411-1445), with an extra weak diffraction at 44.3 degrees corresponding to the (200) reflections of gold face centred cubic phase (Fm3m space group, a = 4.07860 Å; ICCD card no. 04-0784). XPS of the films indicated a (0.9 at.%) 3.7 wt.% Au in the films with the characteristics of Au 4f core level spectra being in agreement with that reported for gold metal 29 , which suggests the gold NPs incorporated at the surface of the tin oxide NRs are in the metallic state. The Au:Sn ratio determined by XPS (found: Au:Sn 3.22 at.% (5.23 wt.%)) and compared to the initial ratio present in the precursor solution used for AACVD (anal. calcd for Au:Sn 23.2 wt.%) showed the incorporation of gold NPs is about 23% efficient, which is higher compared to the efficiency (5%) obtained previously for the co-deposition of tungsten oxide and gold 30 , likely due to the higher temperature of deposition. SnO 2 and Au@SnO 2 NRs integrated into micromachined platforms. The SnO 2 and Au@SnO 2 films grown directly into the micromachined platforms ( Fig. 7a) via AACVD at 620 °C, showed similar diffraction patterns ( Fig. 7b) with NR-like morphology (Fig. 7c,d) compared to that observed for the same deposition conditions on silicon wafers (section SnO 2 NRs and Au@SnO 2 NRs). Measurements of the electrical resistance of the films using the microelectrodes confirmed a good electrical contact, with the resistance at different sensor operating temperatures showing a direct dependency of the conductivity to the temperature, as expected for an n-type semiconductor. The apparent energy activation for electrical conduction (E a c ) estimated for temperatures between 250 °C and 350 °C yielded a value of 0.35 eV for the SnO 2 NRs, which is consistent with the E a c for porous SnO 2 (between 0.28 eV and 1.1 eV) reported previously 49 . The E a c estimated for the Au@SnO 2 NRs, in contrast, yielded a  much lower value (0.11 eV), approximately 69% less than the obtained with the SnO 2 NRs. This change is much higher compared to that recorded previously for the co-deposited Au-functionalised tungsten oxide (Au@WO 3 ) structures, which showed only 9% lower E a c compared to the non-functionalised tungsten oxide structures 30 . These numbers show a close relation with the theoretical differences in the work functions (W F ) of these materials, as the W F of Au (4.8 eV) 50 is 16% and 2% lower than that of SnO 2 (5.7 eV) 2 and WO 3 (4.9 eV) 51 , respectively, and the relative value of these differences (i.e. 8) is similar to that found for the E a c of Au@SnO 2 and Au@WO 3 (i.e. 7.7).
Gas sensing characterization. Gas sensing tests were carried out to H 2 and CO at various operating temperatures between 250 and 390 °C using dc resistance measurements. The temperature dependency of the sensor response for each analyte and concentration is displayed in Fig. 8a and S3a. Results for H 2 suggest a similar trend for sensors composed of SnO 2 and Au@SnO 2 NRs, with slight changes of the response by increasing the operating temperature, whereas results for CO display a tendency to increase the sensor response by increasing the operating temperature, particularly for sensors comprised of Au@SnO 2 NRs. Figure 8a and S3a also reveal greater sensor responses for the Au-functionalised NRs compared to the intrinsic NRs, with higher increments (12-fold) for H 2 than for CO (2-fold), which reduces the cross-response of these analytes and in turn improves the selectivity of the device. The typical change of resistance recorded for each type of sensor is shown in Fig. 8b and S3b. Overall, the sensor responses displayed an n-type behaviour, i.e. decreasing electrical resistance when exposed to H 2 or CO. During the testing period (i.e. 100 h) the sensor response showed good reproducibility with standard errors below ± 1.5%, and little variation of the baseline resistance at each operating temperature (below ± 6%), with the SnO 2 NRs showing lower baseline resistances compared to the Au@SnO 2 NRs (e.g. 4.5 kΩ for SnO 2 and 50 kΩ for Au/SnO 2 at 290 °C). With the injection of humidity (90% RH) in the system, at operating temperature of 290 °C, the baseline resistance of the sensors increased up to 5.5 kΩ for sensors based on SnO 2 NRs and 120 kΩ for those based on Au@SnO 2 NRs. Under these conditions the electrical resistance changes to H 2 remained stable and reproducible, but the sensor response was higher compared to the response obtained in dry air (Fig. 8), showing an increase of the sensor response to H 2 up to 18% for the SnO 2 NRs and 140% for the Au@SnO 2 NRs, indicating decreased humidity tolerance of the SnO 2 NRs after decoration with gold NPs. A detailed view of the normalised response to H 2 at 290 °C for the sensors based on SnO 2 and Au@SnO 2 NRs show the characteristics of the response (t R ) and recovery (t rec ) time, and the time needed to reach stationary state when exposed to H 2 (Fig. S4). The overall view of the response and recovery times as function of the temperature for each type of sensor towards H 2 (Fig. 9) suggests a decrease in the response and recovery time of the sensor as the operating temperature increases. Results in Fig. 9 also show faster response times for the Au@SnO 2 NRs compared to intrinsic NRs, and an inverse relationship of the recovery time with respect to the response time, i.e. longer recovery times for Au@SnO 2 as opposed to SnO 2 . Similar comparison between the non-functionalised and Au-functionalised tungsten oxide structures studied previously showed a similar tendency of the response/ recovery time 30 .
After 100 h of testing the sensor alternately to H 2 and CO in dry and humid environment and at different operating temperatures, the gas sensitive nanostructures where examined again using SEM and XRD. In comparison to the initial samples the morphology of the SnO 2 and Au@SnO 2 NRs was unchanged and the diffraction patterns were identical, indicating a good stability of the devices.

Discussion
The AACVD of SnCl 4 ·5H 2 O allowed for the synthesis of nanostructured SnO 2 films by adjusting the deposition temperature, the precursor concentration and solvent used to produce the aerosol. The ideal conditions for growing SnO 2 NRs were found using a solution 8.5 mM of SnCl 4 ·5H 2 O dissolved in acetone at 620 °C (Fig. 2), a much lower temperature compared to other CVD methods based on VS mechanism which require temperatures exceeding 850 °C 10,11,13,14,25 . The AACVD of SnO 2 NRs via VS mechanism at this relatively low temperature can be connected to two factors: a drop of the energy activation for the perpendicular growth of the film registered when using acetone (16.2 kJ/mol) instead of methanol as carrier solvent, and a decrease of the number of precursor molecules per volume induced by reducing the initial solution concentration about 30%. This is consistent with previous analysis of the growth mechanism of tungsten oxide nanorods via AACVD, which indicated that the transition from planar to nanorod films (characterised by the Frank-van der Merwe and Volmer-Weber growth mode, respectively) at a fixed temperature requires the attenuation of these factors (i.e. E a d and density of precursor molecules) 43 . As the same deposition temperatures and AACVD parameters (e.g. flow, gas carrier, solution concentration) for the different systems were kept constant with the exception of the 'carrier' solvent, it is therefore reasonable to suggest that the use of solvents such as acetone and methanol, which can decompose via radical mechanism at these temperatures [52][53][54] , might also add active reactive products to the reaction in AACVD. This has a direct influence on the decomposition profile of the tin precursor and in turn modifies the energy activation for the perpendicular growth rate of the film promoting formation of nanostructured growth. The presence of metallic tin at 600 °C when using a methanol solution (Fig. S1) corroborates this hypothesis, as this suggests a different chemical species, with a different decomposition profile to the initial precursor, is involved in the AACVD process when this solvent was used. Acetone may also play a similar role in the decomposition mechanism of the tin precursor but results suggest it is less reactive than methanol. The change in reactive chemical species is likely to affect the reaction time and hence the ratio between this and the reactor residence time, parameters which have previously been shown to be involved in the transition from planar to columnar-like structures 55 .
The incorporation of a second precursor (HAuCl 4 ·3H 2 O) during the AACVD of SnO 2 NRs allowed for the co-deposition of Au@SnO 2 NRs, similar to our previous results for the co-deposition of gold NPs with tungsten oxide nanostructures 29,30 , although requiring use of a different carrier solvent and higher deposition temperatures (acetone at 620 °C instead of methanol at 400 °C for tungsten oxide). Overall, the analysis of these structures showed that the gold NPs incorporated at the tin oxide NRs are in metallic state (Fig. 6b) with no evidence of modification of either the morphology (Fig. 5) or the crystalline structure (Fig. 6a) of the NRs. However, the The direct integration of these structures (SnO 2 and Au@SnO 2 ) with micromachined platforms was achieved, demonstrating this process and the associated temperature are compatible with the complementary electronics on these devices. The capability to integrate nanostructured materials with microsystems, such as those used in this work (i.e. fabricated by MEMS technology), represents a technological advantage for gas sensing, as MEMS sensor platforms provide low power consumption features and are suitable for integrating monolithic sensor arrays. Sensor testing showed an optimum operating temperature for SnO 2 of 300 °C, a temperature frequently reported for SnO 2 when sensing H 2 and CO, Table 1. At this temperature (i.e. 300 °C) the micromachined sensors registered 16-times less power consumption (32 mW) than the traditional planar ceramic sensors (525 mW) used in our previous works for tungsten oxide 29 .
Overall, validation of these devices towards the detection of H 2 and CO showed good performance with stable signal, very low drift of electrical resistance over the testing period and relatively fast response. These characteristics are highly enhanced compared to our previous sensors based on SnO 2 NPs synthesised via AACVD of tin complexes 34 ; which showed a very slow and low resistance change to NO 2 and non-sensitivity towards H 2 or CO. A direct comparison of our results and those of the literature is relatively complex, as the performance of the  sensor is not only linked to the material properties, but also in part to the characteristics of the transducers and test conditions (e.g., operating temperatures, flows, and gas concentrations). Despite this we believe that Table 1 can still be useful to evaluate the tendency of SnO 2 towards the analytes tested in this work. Consequently, this comparative table suggests that the SnO 2 NRs synthesised via AACVD provide higher sensitivity (S) to H 2 compared to that recorded in other works for this analyte. In contrast, the sensitivity obtained for CO reveals a much lower value with respect to that observed for hydrothermal synthesised grains, although marginally higher than that reported for electrospun tin oxide fibres. These differences could be linked to some technological characteristics of each sensor, for instance the use of platinum top electrodes in the sensor architecture of the electrospun films, or the use of transfer steps and most likely the introduction of impurities when integrating the tin oxide films synthesized via hydrothermal or thermal evaporation method into the sensor platforms. In fact, impurities such as chlorine (typically introduced from the precursors), or potassium and calcium (often introduced by the use of transfer steps) have been demonstrated to play a relevant role in the surface activity and sensing properties of tin oxide 56 ; most of the tin oxide films in Table 1 were synthesised from chlorine containing precursors (i.e. SnCl 4 and SnCl 2 ) or integrated using transfer steps. The analysis of our SnO 2 NRs integrated directly on the micromachined platform though showed no evidence of chlorine, likely due to easy elimination via HCl from the SnCl 4 ·5H 2 O precursor. The shape and geometry of the AACVD NRs showed similar characteristics to those reported previously in prism-like rods 45 , suggesting a surface likely dominated by the SnO 2 {110} facets. The presence of these facets, which have shown to be less favourable for CO adsorption due to the need of a particular adsorption geometry with the C-end orientated to the surface 2 , may be responsible for the attenuation of the response to CO compared to H 2 . No equivalent studies were found for H 2 , however as the adsorption of H 2 includes the formation of intermediate molecules as water, and these have shown more favourable adsorption on SnO 2 {110} surfaces 2 , the higher responses registered for H 2 in dry and humid air seem consistent.
The functionalization of the SnO 2 NRs with Au NPs showed enhanced sensing characteristics, compared to the intrinsic SnO 2 NRs, which include higher sensor response, (almost 12-and 2-times more for H 2 and CO respectively) and a reduction of the response time of about 6 times. However, the functionalization also showed a tendency to reduce the sensor tolerance to humidity, inducing a larger change of the response to hydrogen in humid ambient compared to the non-functionalised SnO 2 NRs. This characteristic is undesirable in gas sensing and seems to be related to both the propensity of SnO 2 to adsorb hydroxyl groups on the surface at the temperature used for this test 57 , and the amplification of the response by analogous mechanisms than those involved in the enhancement of H 2 and CO sensing. The mechanisms that lead the metal NPs functionalised SMOx to an enhanced gas sensing performance has been discussed in the literature previously, and generally involve surface dependent effects (i.e. chemical sensitization, which may include mechanisms such as spill-over and complementary decomposition) and/or interface dependent effects (electronic sensitization, which may include the modulation of potential barrier heights, carrier injection and conduction channel modulation) [58][59][60] . The faster response in the Au@SnO 2 compared to the SnO 2 NRs gives evidence of a chemical sensitization, with the Au NPs most likely accelerating the dissociation of hydrogen molecules into H atoms and simultaneously inducing a faster saturation of the active sites on the NR surface via spill-over. This saturation (also observed in the response (Fig. S4) of the sensors based on Au@SnO 2 NRs as opposed to those based on SnO 2 NRs) is responsible for the larger desorption time required to recover the baseline resistance at each temperature (Fig. 9). In contrast, the lower E a c obtained for the Au@SnO 2 NRs compared to the SnO 2 NRs indicates an electronic sensitization, with the Au NPs facilitating the carrier injection and thus modulating the conduction channel along the nanostructure.
The ratio of the response to H 2 and CO for each microsensor (Δ R = 2.5 for SnO 2 and Δ R = 20 for Au/SnO 2 ) indicates relatively higher cross-sensitivities for the microsensors comprised of intrinsic SnO 2 NRs, as opposed to those comprised of Au@SnO 2 NRs, which potentially improves selectivity to H 2 . A comparison of these ratios with those recorded for similar systems synthesised via sol-gel in the literature 61 revealed similar values for the SnO 2 NRs and nearly 7 times higher values for the Au@SnO 2 NRs, suggesting the functionalisation of SnO 2 with Au NPs via AACVD is effective for improving the selectivity of tin oxide towards H 2 and CO.
In summary, these results demonstrate the AACVD of SnO 2 NRs via VS mechanism at exceptionally reduced process temperature compared to existing CVD methods with no need for substrate pre-treatment. This allowed for the direct integration of these nanostructures into micromachined platforms and their use for sensing H 2 and CO. The incorporation of Au NPs at the surface of the SnO 2 NRs via co-deposition improved the functionality of SnO 2 , particularly to H 2 , reducing the cross-sensitivity of this analyte to CO.

Conclusions
SnO 2 NRs with an aspect ratio of approximately 7 were synthesized without catalyst seeds via AACVD of SnCl 4 ·5H 2 O at 620 °C, a much lower onset temperature compared to other CVD methods based on a VS mechanism, which typically requires temperatures exceeding 850 °C. The evolution of nanorod SnO 2 is linked to an increase in energy activation of the perpendicular growth of about 40% respect to that observed for non-nanostructured SnO 2 films, and is attributed to the use of acetone as solvent carrier, and a reduction of precursor concentration. Co-deposition of Au NPs (< 35 nm) and SnO 2 NRs via AACVD was also achieved at 620 °C. The gas microsensors based on the intrinsic SnO 2 and Au@SnO 2 NRs were validated towards H 2 and CO and show sensing properties that are in agreement with the literature, with notable enhancement of sensing properties for Au@SnO 2 NRs which showed 12-fold higher response with 6-fold faster response and improved selectivity to H 2 compared to the gas sensors based on intrinsic SnO 2 NRs.

Experimental Section
SnO 2 synthesis. SnO  15 ml of acetone. In order to vary the concentration of the solution, three different weights (45, 30 and 15 mg) of SnCl 4 ·5H 2 O were used for the deposition. A piezoelectric ultrasonic atomiser (Liquifog, Johnson Matthey) operating at 1.6 MHz was used to generate an aerosol of the solution. The aerosol droplets were transported to the heated substrate by a nitrogen (BOC, oxygen free) gas flow (200 cm 3 ·min −1 ), and the time taken to transport the entire volume of solution was typically 15 minutes. Au@SnO 2 synthesis. Nanocomposites composed of gold NPs and tin oxide NRs were synthesised at 620 °C by co-deposition, via AACVD, of tin (IV) chloride pentahydrate (30 mg, SnCl 4 .5H 2 O, Sigma-Aldrich, ≥ 98%) and tetrachloroauric acid trihydrate (4.2 mg, HAuCl 4 ·3H 2 O, Sigma-Aldrich, 99.9%) dissolved in acetone (15 ml, Sigma-Aldrich, ≥ 99.6%) using the same system described above and following the method reported in literature 29 . Substrate and micromachined platforms. Silicon wafers (10 mm × 10 mm × 0.37 mm) and KBr disks were used as substrates for film analysis, whereas micromachined platforms consisting of an array of four SiO 2 / Si 3 N 4 /SiO 2 membrane, each of them with isolated polysilicon heaters and platinum electrodes (gap: 50 μ m, thick: 0.2 μ m), were used for gas sensor fabrication. After deposition of the sensing active film the platforms were mounted on a TO8 package. The sensor technology was described in detail previously 62 .
Film analysis. The morphology of the samples was examined using Scanning Electron Microscopy (SEM -Jeol 6301F, 5 keV). The structure using X-Ray Diffraction (either XRD -Bruker, AXD D8-Discover for the films grown on silicon wafers or Rigaku Smartlab 9 kW for the films grown on micromachined platfoms) and the chemical composition using Wavelength Dispersive X-Ray (WDX -Philips, XL30ESEM) and X-ray Photoelectron spectroscopy (XPS) (Thermo Scientific K-Alpha, using Al Ka radiation operated at 0.6 eV with electron gun operating at 1 eV and argon-ion gun operated at 10 eV; the binding energies were calibrated to the C 1s peak at 284.5 eV). TEM (JEOL JEM-100CX II, 100 kV) images were carried out on samples prepared by deposition on KBr substrates followed by dissolution of the substrate in distilled water and suspended on Cu grids.
Gas sensing tests. Gas sensors were tested in a continuous flow (200 sccm) test chamber (280 cm 3 ) as previously described 62 . The sensors were exposed to 250 and 500 ppm of hydrogen and carbon monoxide for 10 min and subsequently the chamber purged with air until initial baseline resistance was recovered. The whole testing period comprised of 100 h during which sensors were tested to different hydrogen and carbon monoxide concentrations at operating temperatures between 250 and 390 °C in dry and humid ambient, performing 5 replicates for each condition. To obtain the desired analyte concentration calibrated cylinders of either hydrogen (Praxair, 1000 ppm) or carbon monoxide (Praxair, 1000 ppm) were mixed with pure synthetic air (Carburos Metálicos, 99.99%) by means of a mass flow system (Bronkhorst hi-tech 7.03.241). The sensor response was defined as R = R a /R g , where R a is the sensor resistance in air at stationary state and R g represents the sensor resistance after 10 min of the analyte exposure. The response time was defined as the time required for the sensor to reach 90% of the sensor response, and the recovery time as the time required to reach 10% of the initial baseline resistance after the analyte was purged.