Influence of Mono- and Bimetallic PtOx, PdOx, PtPdOx Clusters on CO Sensing by SnO2 Based Gas Sensors

To obtain a nanocrystalline SnO2 matrix and mono- and bimetallic nanocomposites SnO2/Pd, SnO2/Pt, and SnO2/PtPd, a flame spray pyrolysis with subsequent impregnation was used. The materials were characterized using X-ray diffraction (XRD), a single-point BET method, transmission electron microscopy (TEM), and high angle annular dark field scanning transmission electron microscopy (HAADF-STEM) with energy dispersive X-ray (EDX) mapping. The electronic state of the metals in mono- and bimetallic clusters was determined using X-ray photoelectron spectroscopy (XPS). The active surface sites were investigated using the Fourier Transform infrared spectroscopy (FTIR) and thermo-programmed reduction with hydrogen (TPR-H2) methods. The sensor response of blank SnO2 and nanocomposites had a carbon monoxide (CO) level of 6.7 ppm and was determined in the temperature range 60–300 °C in dry (Relative Humidity (RH) = 0%) and humid (RH = 20%) air. The sensor properties of the mono- and bimetallic nanocomposites were analyzed on the basis of information on the electronic state, the distribution of modifiers in SnO2 matrix, and active surface centers. For SnO2/PtPd, the combined effect of the modifiers on the electrophysical properties of SnO2 explained the inversion of sensor response from n- to p-types observed in dry conditions.


Introduction
Because SnO 2 is a wide-bandgap oxygen-deficient n-type semiconductor with optical transparency, electron conductivity, and a high specific surface area, it is suitable for a large range of applications, including in solar cells, as catalytic support, and as solid state gas sensors [1]. However, the use of bare SnO 2 is often limited by lack of selectivity and a high operating temperature [2]. Chemical modification is a well-established practice intended to solve those problems [2][3][4][5]. This involves the creation of new active sites, with specific adsorptivity and reactivity toward target gases (i.e., carbon monoxide), on the surface of a semiconductor matrix.
Carbon monoxide (CO) is a colorless, odorless, and tasteless toxic gas, produced by automotive emissions, natural gas manufacturing, industrial activities, and the incomplete burning of fuels [6]. In the  (a) crystallite size (X-ray diffraction, XRD); (b) particle size (transmission electron microscopy, TEM); (c) obtained using X-ray fluorescence (XRF) analysis. Figure 1 shows the change in the resistance of the samples in the temperature range 60-300 • C under periodic change of the gas phase composition: air (15 min), 6.7 ppm CO in air (15 min). The measurements were effectuated in dry air (Relative Humidity (RH) = 0%, Figure 1a) and humid air (RH = 20%, Figure 1b). The decrease in the electrical resistance in the presence of CO (n-type response) was due to the oxidation of carbon monoxide by chemisorbed oxygen:

Results and Discussion
where CO (gas) represents the carbon monoxide molecule in the gas phase, O −α β(ads) is chemisorbed oxygen, CO 2(gas) is the reaction product desorbed to the gas phase, and e is an electron injected into the conduction band of the n-type semiconductor. The sensor signal for n-type response was calculated for each temperature as S = ∆G G air = G gas −G air G air , where G air is the sample's conductance in air and G gas is the sample's conductance in the presence of 6.7 ppm CO in air ( Figure 2). For SnO 2 /PtPd nanocomposite the increase in the electrical resistance in the presence of CO (p-type response) was observed in the temperature range 150-300 • C at RH = 0% (Figure 1c). In these cases, the sensor signal was calculated as S = ∆R R air = R gas −R air R air , where R air is the sample's resistance in air and R gas is the sample's resistance in the presence of 6.7 ppm CO in air.  Figure 1 shows the change in the resistance of the samples in the temperature range 60-300 °C under periodic change of the gas phase composition: air (15 min), 6.7 ppm CO in air (15 min). The measurements were effectuated in dry air (Relative Humidity (RH) = 0%, Figure 1a) and humid air (RH = 20%, Figure 1b). The decrease in the electrical resistance in the presence of CO (n-type response) was due to the oxidation of carbon monoxide by chemisorbed oxygen:

Results and Discussion
where CO ( ) represents the carbon monoxide molecule in the gas phase, O ( ) is chemisorbed oxygen, CO ( ) is the reaction product desorbed to the gas phase, and e is an electron injected into the conduction band of the n-type semiconductor. The sensor signal for n-type response was calculated for each temperature as = ∆ = , where Gair is the sample's conductance in air and Ggas is the sample's conductance in the presence of 6.7 ppm CO in air ( Figure 2). For SnO2/PtPd nanocomposite the increase in the electrical resistance in the presence of CO (p-type response) was observed in the temperature range 150-300 °C at RH = 0% (Figure 1c). In these cases, the sensor signal was calculated as = ∆ = , where Rair is the sample's resistance in air and Rgas is the sample's resistance in the presence of 6.7 ppm CO in air.     Figure 1 shows the change in the resistance of the samples in the temperature range 60-300 °C under periodic change of the gas phase composition: air (15 min), 6.7 ppm CO in air (15 min). The measurements were effectuated in dry air (Relative Humidity (RH) = 0%, Figure 1a) and humid air (RH = 20%, Figure 1b). The decrease in the electrical resistance in the presence of CO (n-type response) was due to the oxidation of carbon monoxide by chemisorbed oxygen:

Results and Discussion
where CO ( ) represents the carbon monoxide molecule in the gas phase, O ( ) is chemisorbed oxygen, CO ( ) is the reaction product desorbed to the gas phase, and e is an electron injected into the conduction band of the n-type semiconductor. The sensor signal for n-type response was calculated for each temperature as = ∆ = , where Gair is the sample's conductance in air and Ggas is the sample's conductance in the presence of 6.7 ppm CO in air ( Figure 2). For SnO2/PtPd nanocomposite the increase in the electrical resistance in the presence of CO (p-type response) was observed in the temperature range 150-300 °C at RH = 0% (Figure 1c). In these cases, the sensor signal was calculated as = ∆ = , where Rair is the sample's resistance in air and Rgas is the sample's resistance in the presence of 6.7 ppm CO in air. (1) SnO2, (2) SnO2/Pd, (3) SnO2/Pt, (4) SnO2/PtPd. Pale blue areas correspond to exposure in CO.

Figure 2.
Sensor signal to 6.7 ppm CO of blank SnO2 and mono-and bimetallic nanocomposites in the temperature range 60-300 °C at relative humidity RH = 0% (a) and RH = 20% (b).

Figure 2.
Sensor signal to 6.7 ppm CO of blank SnO 2 and mono-and bimetallic nanocomposites in the temperature range 60-300 • C at relative humidity RH = 0% (a) and RH = 20% (b). The data presented in Figures 1 and 2, revealed the following trends.
(i) Comparison of the resistance values show that the introduction of modifiers caused an increase in the resistance of tin dioxide. This effect was most pronounced for nanocomposites containing platinum. (ii) For the non-modified SnO 2 , the value of the n-type sensor response toward CO increased with increasing measurement temperature and reached a maximum at 270-300 • C. The increase in humidity almost completely suppressed the sensor response of SnO 2 sample. (iii) For the SnO 2 /Pd nanocomposite, only the n-type response was observed. Two maxima can be distinguished on the temperature dependence of the sensor signal: one in the temperature range 240-270 • C and one in the range 60-90 • C. The sensor signal values at 240-270 • C decreased slightly with increasing air humidity from RH = 0% to RH = 20%. In the low-temperature interval, an increase in humidity led to a significant decrease in the sensor signal. (iv) The SnO 2 /Pt nanocomposite exhibited a low sensor response. However, in the low-temperature range, its response to CO in dry air exceeded the analogous value for unmodified SnO 2 , while in the high-temperature region its sensor signal turns out to be lower than for SnO 2. The inversion of the sensor response from the n-type to the p-type is observed only when measured in dry air at T = 240 • C. (v) When performing the measurements in dry air, the inversion of the sensor response was characteristic for bimetallic nanocomposite SnO 2 /PtPd over a wide temperature range. The maximum of the p-type response was observed at 210-240 • C. The increase in air humidity led to the disappearance of the inversion of the response. The observed n-type response in the whole temperature range was lower than for palladium containing monometallic nanocomposite SnO 2 /Pd.
To determine the factors responsible for the formation of the sensor response of mono-and bimetallic nanocomposites, the phase composition, the electronic state of platinum and palladium, and their distribution in the SnO 2 matrix were investigated. The effect of the presence of various modifiers on the surface composition of nanocrystalline SnO 2 was studied in detail.
The X-ray diffraction (XRD) pattern of SnO 2 corresponds to the cassiterite phase (ICDD 41-1445, Figure 3). The introduction of catalytic modifiers did not led to a change in the phase composition of the samples. The reflections corresponding to Pt-or Pd-containing phases do not appear on the diffractograms of the nanocomposites. The data presented in Figures 1 and 2, revealed the following trends.
(i) Comparison of the resistance values show that the introduction of modifiers caused an increase in the resistance of tin dioxide. This effect was most pronounced for nanocomposites containing platinum. (ii) For the non-modified SnO2, the value of the n-type sensor response toward CO increased with increasing measurement temperature and reached a maximum at 270-300 °C. The increase in humidity almost completely suppressed the sensor response of SnO2 sample. (iii) For the SnO2/Pd nanocomposite, only the n-type response was observed. Two maxima can be distinguished on the temperature dependence of the sensor signal: one in the temperature range 240-270 °C and one in the range 60-90 °C. The sensor signal values at 240-270 °C decreased slightly with increasing air humidity from RH = 0% to RH = 20%. In the low-temperature interval, an increase in humidity led to a significant decrease in the sensor signal. (iv) The SnO2/Pt nanocomposite exhibited a low sensor response. However, in the low-temperature range, its response to CO in dry air exceeded the analogous value for unmodified SnO2, while in the high-temperature region its sensor signal turns out to be lower than for SnO2. The inversion of the sensor response from the n-type to the p-type is observed only when measured in dry air at T = 240 °C. (v) When performing the measurements in dry air, the inversion of the sensor response was characteristic for bimetallic nanocomposite SnO2/PtPd over a wide temperature range. The maximum of the p-type response was observed at 210-240 °C. The increase in air humidity led to the disappearance of the inversion of the response. The observed n-type response in the whole temperature range was lower than for palladium containing monometallic nanocomposite SnO2/Pd.
To determine the factors responsible for the formation of the sensor response of mono-and bimetallic nanocomposites, the phase composition, the electronic state of platinum and palladium, and their distribution in the SnO2 matrix were investigated. The effect of the presence of various modifiers on the surface composition of nanocrystalline SnO2 was studied in detail.
The X-ray diffraction (XRD) pattern of SnO2 corresponds to the cassiterite phase (ICDD 41-1445, Figure 3). The introduction of catalytic modifiers did not led to a change in the phase composition of the samples. The reflections corresponding to Pt-or Pd-containing phases do not appear on the diffractograms of the nanocomposites. According to the high angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images, all the materials were composed of agglomerated crystalline SnO2 nanoparticles, with sizes varying from approximately 5-35 nm, with an average size of 10.7 ± 4.9 nm ( Figure 4). The introduction of modifiers does not affect the particle size of SnO2. In SnO2/Pd nanocomposite ( Figure 5) several Pd particles with a size 8-20 nm were found among the SnO2 matrix According to the high angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images, all the materials were composed of agglomerated crystalline SnO 2 nanoparticles, with sizes varying from approximately 5-35 nm, with an average size of 10.7 ± 4.9 nm ( Figure 4). The introduction of modifiers does not affect the particle size of SnO 2 . In SnO 2 /Pd nanocomposite ( Figure 5) several Pd particles with a size 8-20 nm were found among the SnO 2 matrix particles ( Figure 5b). Also small Pd particles with a size of 2 nm could be seen on the images and energy dispersive X-ray (EDX) maps ( Figure 5b).
In the SnO 2 /Pt nanocomposite, the Pt particles were very easy to see on the HAADF-STEM images ( Figure 6a) because they are large and bright. Their size varied in the range of 25-100 nm and the average size was 49.1 ± 19.5 nm (Figure 6c).
The nanocomposite SnO 2 /PtPd contained bimetallic nanoparticles particles, which were present in a form of agglomerates with the size 17-64 nm (Figure 7). According to the EDX maps, the particles contained both metals (Figure 7b), however their ratio varied from particle to particle and was independent from the particle size ( Figure 7c). The distribution of Pt and Pd inside the particle was not uniform. Besides the PtPd particles, small Pd particles with a size about 2 nm could be seen on the maps (Figure 7b). particles ( Figure 5b). Also small Pd particles with a size of 2 nm could be seen on the images and energy dispersive X-ray (EDX) maps ( Figure 5b).
In the SnO2/Pt nanocomposite, the Pt particles were very easy to see on the HAADF-STEM images ( Figure 6a) because they are large and bright. Their size varied in the range of 25-100 nm and the average size was 49.1 ± 19.5 nm (Figure 6c).
The nanocomposite SnO2/PtPd contained bimetallic nanoparticles particles, which were present in a form of agglomerates with the size 17-64 nm (Figure 7). According to the EDX maps, the particles contained both metals (Figure 7b), however their ratio varied from particle to particle and was independent from the particle size (Figure 7c). The distribution of Pt and Pd inside the particle was not uniform. Besides the PtPd particles, small Pd particles with a size about 2 nm could be seen on the maps (Figure 7b).    particles (Figure 5b). Also small Pd particles with a size of 2 nm could be seen on the images and energy dispersive X-ray (EDX) maps (Figure 5b). In the SnO2/Pt nanocomposite, the Pt particles were very easy to see on the HAADF-STEM images (Figure 6a) because they are large and bright. Their size varied in the range of 25-100 nm and the average size was 49.1 ± 19.5 nm (Figure 6c).
The nanocomposite SnO2/PtPd contained bimetallic nanoparticles particles, which were present in a form of agglomerates with the size 17-64 nm (Figure 7). According to the EDX maps, the particles contained both metals (Figure 7b), however their ratio varied from particle to particle and was independent from the particle size (Figure 7c). The distribution of Pt and Pd inside the particle was not uniform. Besides the PtPd particles, small Pd particles with a size about 2 nm could be seen on the maps (Figure 7b).    particles (Figure 5b). Also small Pd particles with a size of 2 nm could be seen on the images and energy dispersive X-ray (EDX) maps (Figure 5b). In the SnO2/Pt nanocomposite, the Pt particles were very easy to see on the HAADF-STEM images (Figure 6a) because they are large and bright. Their size varied in the range of 25-100 nm and the average size was 49.1 ± 19.5 nm (Figure 6c).
The nanocomposite SnO2/PtPd contained bimetallic nanoparticles particles, which were present in a form of agglomerates with the size 17-64 nm (Figure 7). According to the EDX maps, the particles contained both metals (Figure 7b), however their ratio varied from particle to particle and was independent from the particle size (Figure 7c). The distribution of Pt and Pd inside the particle was not uniform. Besides the PtPd particles, small Pd particles with a size about 2 nm could be seen on the maps (Figure 7b).    The Pd3d and Pt4f X-ray photoelectron spectroscopy (XPS) signals of the nanocomposites SnO2/Pd, SnO2/Pt, and SnO2/PtPd could be fitted by only one doublet component (Figure 8a,b). Table  2 presents the Pd 3d5/2 and Pt 4f7/2 XPS spectral assignments. When comparing the estimated Pd 3d5/2 binding energies with reference data [65], we concluded that in SnO2/Pd nanocomposite, palladium was present in the +2 oxidation state corresponding to the PdO. The Pd3d XP spectrum of the nanocomposite SnO2/PtPd was shifted toward lower energies (Figure 8a), indicating a partial reduction of Pd. However, it was not possible to decompose the spectrum into two components that corresponded to different palladium oxidation states (0, +2). The Pt4f XP spectra of both SnO2/Pt and SnO2/PtPd nanocomposites corresponded to the Pt + 2 oxidation state in PtO [66]. When taking into account how the depth of the XPS analysis was determined by the mean free path of electrons with respect to inelastic collisions, and is 0.5-2.5 nm for metals and 4-10 nm for organic substances, it was impossible to exclude the presence of Pd 0 and especially, Pt 0 inside the particles (under the oxide layer).
The O1s XP spectra consisted of two components (Figure 8c) corresponding to the oxygen anions in the SnO2 lattice (530.7-530.9 eV) and to different forms of chemisorbed oxygen and hydroxyl groups on the SnO2 surface (531.8-532.0 eV). The impact of the higher energy component is 23-27% and did not depend significantly on the type of modifier.  The Pd3d and Pt4f X-ray photoelectron spectroscopy (XPS) signals of the nanocomposites SnO 2 /Pd, SnO 2 /Pt, and SnO 2 /PtPd could be fitted by only one doublet component (Figure 8a,b). Table 2 presents the Pd 3d 5/2 and Pt 4f 7/2 XPS spectral assignments. When comparing the estimated Pd 3d 5/2 binding energies with reference data [65], we concluded that in SnO 2 /Pd nanocomposite, palladium was present in the +2 oxidation state corresponding to the PdO. The Pd3d XP spectrum of the nanocomposite SnO 2 /PtPd was shifted toward lower energies (Figure 8a), indicating a partial reduction of Pd. However, it was not possible to decompose the spectrum into two components that corresponded to different palladium oxidation states (0, +2). The Pt4f XP spectra of both SnO 2 /Pt and SnO 2 /PtPd nanocomposites corresponded to the Pt + 2 oxidation state in PtO [66]. When taking into account how the depth of the XPS analysis was determined by the mean free path of electrons with respect to inelastic collisions, and is 0.5-2.5 nm for metals and 4-10 nm for organic substances, it was impossible to exclude the presence of Pd 0 and especially, Pt 0 inside the particles (under the oxide layer).
The O1s XP spectra consisted of two components (Figure 8c) corresponding to the oxygen anions in the SnO 2 lattice (530.7-530.9 eV) and to different forms of chemisorbed oxygen and hydroxyl groups on the SnO 2 surface (531.8-532.0 eV). The impact of the higher energy component is 23-27% and did not depend significantly on the type of modifier.  The infra red (IR) spectra of blank SnO2 samples and nanocomposites are compared in Figure  9a. The intense absorption band at 400-800 cm −1 corresponded to the oscillations of Sn-O-Sn bridges (670 cm −1 ), Sn-OH terminal bonds (590 cm −1 ), and surface (540 cm −1 ) and bulk (480-460 cm −1 ) phonon vibrations of SnO2 [67]. The other absorption bands in the spectrum, apparently, were due to adsorbates. For a clear comparison of their concentration, the baselines were subtracted from the spectra, and the transmission in the whole range was normalized to the peak of the lattice vibrations at 670 cm −1 . The broad absorption band at 3300-3650 cm −1 , apparently referred to the stretching vibrations of O-H adsorbed water derivatives [68]. The modification of tin dioxide with palladium led to a significant increase in the concentration of surface hydroxyl groups, as evidenced by the increase in absorption in the range of 3300-3650 cm −1 . On the contrary, the introduction of platinum did not have any effect on the concentration of hydroxyl groups on the SnO2 surface. For the SnO2/PtPd nanocomposite, a nonadditive increase in the concentration of OH groups was observed in comparison with SnO2/Pt and SnO2/Pd nanocomposites.   [67]. The other absorption bands in the spectrum, apparently, were due to adsorbates. For a clear comparison of their concentration, the baselines were subtracted from the spectra, and the transmission in the whole range was normalized to the peak of the lattice vibrations at 670 cm −1 . The broad absorption band at 3300-3650 cm −1 , apparently referred to the stretching vibrations of O-H adsorbed water derivatives [68]. The modification of tin dioxide with palladium led to a significant increase in the concentration of surface hydroxyl groups, as evidenced by the increase in absorption in the range of 3300-3650 cm −1 . On the contrary, the introduction of platinum did not have any effect on the concentration of hydroxyl groups on the SnO 2 surface. For the SnO 2 /PtPd nanocomposite, a nonadditive increase in the concentration of OH groups was observed in comparison with SnO 2 /Pt and SnO 2 /Pd nanocomposites. The infra red (IR) spectra of blank SnO2 samples and nanocomposites are compared in Figure  9a. The intense absorption band at 400-800 cm −1 corresponded to the oscillations of Sn-O-Sn bridges (670 cm −1 ), Sn-OH terminal bonds (590 cm −1 ), and surface (540 cm −1 ) and bulk (480-460 cm −1 ) phonon vibrations of SnO2 [67]. The other absorption bands in the spectrum, apparently, were due to adsorbates. For a clear comparison of their concentration, the baselines were subtracted from the spectra, and the transmission in the whole range was normalized to the peak of the lattice vibrations at 670 cm −1 . The broad absorption band at 3300-3650 cm −1 , apparently referred to the stretching vibrations of O-H adsorbed water derivatives [68]. The modification of tin dioxide with palladium led to a significant increase in the concentration of surface hydroxyl groups, as evidenced by the increase in absorption in the range of 3300-3650 cm −1 . On the contrary, the introduction of platinum did not have any effect on the concentration of hydroxyl groups on the SnO2 surface. For the SnO2/PtPd nanocomposite, a nonadditive increase in the concentration of OH groups was observed in comparison with SnO2/Pt and SnO2/Pd nanocomposites.  The results of the TPR-H 2 experiments are shown in Figure 9b and in Table 3. The high-temperature (350-850 • C) peak corresponded to the hydrogen consumption because of SnO 2 reduction to metallic tin: For nanocomposites SnO 2 /Pd, SnO 2 /Pt, and SnO 2 /PtPd, a shift was observed from the high-temperature maximum of hydrogen consumption to the lower temperature region. This may be due to the catalytic activity of noble metal clusters in the matrix of nanocrystalline tin dioxide. The most probable mechanism was the hydrogen and oxygen joint spillover [69] of the SnO 2 crystal lattice through the clusters of modifiers. The reduction of SnO 2 at a lower temperature became possible due to the dissociation of hydrogen molecules on noble metal clusters. Examples illustrating such a mechanism of interaction with the gas phase for different "noble metal/metal oxide" systems were provided in the review [69]. By the oxygen isotopic exchange method [70], it was established that the modification of nanocrystalline tin dioxide with palladium and ruthenium resulted in the realization of a multistage heteroexchange mechanism, involving the dissociation of the O 2 molecule on the surface of the clusters of platinum-group metals, spillover of atomic oxygen from the clusters of modifiers to the SnO 2 surface, and its rapid exchange with the oxygen of the crystal lattice of tin dioxide. According to the analysis of the oxygen isotopic exchange data presented in the review [71], platinum was more active in this process than palladium, explaining the more significant decrease in the temperature of the complete reduction of the SnO 2 /Pt and SnO 2 /PtPd nanocomposites compared to SnO 2 /Pd.
The processes responsible for the H 2 consumption in low temperature region 100-300 • C can be expressed as follows The amount of oxygen adsorbed on the surface of SnO 2 , estimated from TPR data under the assumption of interaction (3)  ions can be estimated at 0.5-1 monolayer. This is 2-3 orders of magnitude greater than the Weisz limitation for coverage of a semiconductor with charged adsorbates (10 −2 -10 −3 monolayer [72]), indicating that the real composition of oxidative adsorbates on the surface of materials was much more diverse than this assumption.
Modification of the SnO 2 surface with mono-and bimetallic clusters led to a decrease in the amount of hydrogen consumed in the low-temperature range. This indirectly indicated an increase in the fraction of the monatomic form of chemisorbed oxygen. Indeed, when comparing the reactions (3) and (4) surfaces with the same negative charge, predetermined by the Weitz limitation, it can be concluded that in the latter case, a smaller hydrogen consumption should be observed. In the case of nanocomposites containing palladium, the decrease in hydrogen absorption may be due to an increase in the concentration of hydroxyl groups detected by the IR spectroscopy and consequently, an increase in the contribution of the process (5). The obtained information on the structure and surface composition of SnO 2 and nanocomposites SnO 2 /Pd, SnO 2 /Pt, and SnO 2 /PtPd allowed us to explain the differences in their sensor properties to CO in dry (RH = 0%) and humid (RH = 20%) air.

SnO 2 /Pd nanocomposite
For the SnO 2 /Pd nanocomposite, the observed sensor response to CO in dry air was not unexpected and can be described within the framework of the concepts of electronic and chemical sensitization [2,5,11].
(i) Direct oxidation of CO gas molecules on the surface of PdO x clusters resulted in a partial reduction of the modifier, while the fraction of Pd 0 increases: This effect corresponded to the mechanism of electronic sensitization. The electron work function ϕ for reduced palladium surface was smaller than in the case of PdO x and is ϕ = 4.8 eV [73], which was close to the work function for SnO 2 (ϕ = 4.9 eV, [74]). Thus, as a result of the reduction of PdO x clusters, a barrier was removed at the Pd/SnO 2 interface, which led to a decrease in material resistance and the sensor response appearance.
(ii) Strong chemisorption of CO molecules on Pd 0 was accompanied by a weakening of the intramolecular bond in CO and facilitated its break and further transformations of chemisorbed molecules: (iii) Modification with PdO x clusters led to an increase in the SnO 2 concentration of paramagnetic centers ·OH and rooted hydroxyl groups OH···OH that participated in the low temperature oxidation of chemisorbed CO molecules: The increase in humidity led to a decrease in the sensitivity of SnO 2 /Pd in the low temperature region due to competitive adsorption of water molecules and blocking of active sites on the surface of SnO 2 and PdO x clusters [75].

SnO 2 /Pt nanocomposite
The low response of nanocomposites SnO 2 /Pt was caused by direct CO oxidation on PtO x clusters. Based on the results of complementary investigations by in situ and operando X-ray absorption spectroscopy and operando FT-IR spectroscopy, D. Degler et al. found that CO oxidation mainly occurs at the PtO x surface (Pt-O-Pt sites) [76]. This led to a decrease in the quantity of CO molecules that can react with the oxygen chemisorbed on SnO 2 surface. The authors assumed that if platinum is introduced by impregnation after the calcination of the SnO 2 matrix, it forms a separate oxide phase, which creates additional reaction sites not electronically coupled to the SnO 2 . In our investigation, this assumption could not be used since the introduction of PtO x clusters led to the important (~10 3 times) increase in SnO 2 resistance in dry air (Figure 1a). The work function of metallic Pt was sufficiently high (ϕ = 5.65 eV [77]). Covering platinum with a full monolayer of oxygen led to the further increase in the work function by 1.19 eV (ϕ~6.8 eV) [78]. When a contact was formed between PtO x nanoparticles and SnO 2 (ϕ = 4.9 eV), the Fermi level of the semiconductor oxide was pinned to the Pt 2+ /Pt 0 potential. This led to the formation of an electron depleted space charge region and to an increase in the SnO 2 resistivity. As in the case of PdO x , it was possible that PtO x clusters reduced upon interaction with CO [79]. However, the work function of metallic Pt significantly exceeded the corresponding value for SnO 2 . As a result, the band-bending at the Pt/SnO 2 interface persisted in the presence of CO, which led to a low sensor response of SnO 2 /Pt nanocomposite when CO was detected in dry air.
In humid air, the Pt-O-Pt sites were deactivated [76] because of strong water adsorption on PtO x clusters [75]. This should have led to a decrease in the amount of CO oxidized directly on the clusters of the modifier, and consequently, to an increase in the sensor response due to an increase in the amount of CO oxidized by oxygen chemisorbed on the SnO 2 surface. However, in humid air, the surface of tin dioxide also became inactive due to the partial replacement of chemisorbed oxygen by hydroxyl groups according to the following reaction [80] The combination of these factors suggests that CO oxidation can take place at the Pt-O-Sn sites at the three-phase boundary between PtO x and SnO 2 [76], providing a very slight increase in sensor response as compared with detection in dry air.

SnO 2 /PtPd nanocomposite
A feature of the SnO 2 /PtPd nanocomposite was a two-level distribution of the modifiers over the surface of the semiconductor matrix: platinum only existed as a part of large (20-60 nm) bimetallic particles PtPdO x with a Pt content generally exceeding 50 mol%, while palladium was distributed between these large bimetallic particles PtPdO x and small (~2 nm) particles PdO x . As a result of this distribution of modifiers, the active sites on the surface of SnO 2 /PtPd were a combination of the centers characteristic for SnO 2 /Pt and SnO 2 /Pd. Thus, the temperature of SnO 2 reduction in the SnO 2 /PtPd nanocomposite coincided with that of SnO 2 /Pt (Figure 9b), which was due to the presence of particles enriched in platinum. At the same time, an increase in the concentration of hydroxyl groups on the surface of SnO 2 /PtPd (Figure 9a) was determined by the presence of small PdO x particles.
However, when comparing the sensor response, such an additive picture was not observed. To explain the inversion of sensor response from nto p-type observed in dry conditions, it was necessary to analyze the combined effect of the modifiers on the electrophysical properties of SnO 2 . Since metals Pt, Pd, and, to a greater extent, their oxides, are characterized by a higher work function than SnO 2 , in the contact areas SnO 2 -PtPdO x and SnO 2 -PdO x , an electron-depleted layer formed in SnO 2 . The depth of this layer was determined using the height of the energy barrier-i.e., the difference in the work function values of the contacting materials-and the lateral length along the surface was determined by the area of the contact between SnO 2 and clusters of modifiers. In the SnO 2 /PtPd nanocomposite, bimetallic particles enriched in platinum form on the surface of the SnO 2 agglomerates regions with a deep depleted layer, while small PdO x particles formed extended space charge regions of the surface of SnO 2 grains. As a result, it can be expected that the concentration of electrons in SnO 2 near the surface layer became so low that they ceased to be the main charge carriers. Such a change in the response type was reported for various oxides: from nto p-type for MoO 3 [81], In 2 O 3 [82], SnO 2 (Fe) [83], ZnO [84], WO 3 [85], WO 3 nanorods [86], TiO 2 nanofibers [87], and from pto n-type conductivity for α-Fe 2 O 3 [88]. The inversion of the sensor response was explained by a change in the type of main charge carriers in semiconductor oxide due to either the surface reactions under certain conditions, or because of the effect of impurities.
In dry air, in the high temperature region, the p-type response of the SnO 2 /PtPd nanocomposite had a similar temperature dependence as the n-type response of the SnO 2 /Pd nanocomposite. It indicated that under these conditions, the PdO x small nanoparticles determined the reactivity of SnO 2 /PtPd sample through reaction (6). The reduced amplitude of the p-type response of SnO 2 /PtPd nanocomposite compared with the n-type response of SnO 2 /Pd was because of the direct oxidation of part of the CO molecules on the surface of large bimetallic PtPdO x particles. This process did not alter the concentration of charge carriers in the SnO 2 semiconductor matrix.
In humid air, the adsorption of water vapor on SnO 2 , in addition to blocking the active centers, led to an increase in the concentration of electrons and in the conductivity of SnO 2 (reaction (9)). Comparison of the data presented in Figure 1a,b, clearly demonstrates that an increase in the air relative humidity (25 • C) from RH = 0% to RH = 20% reduced the resistance of all samples by about 10 times at each measurement temperature. Thus, in a humid atmosphere, electrons remained the main charge carriers in the SnO 2 /PtPd nanocomposite and no inversion of sensor response from nto p-type was observed.

Materials and Methods
The synthesis of nanocrystalline SnO 2 was carried out using flame spray pyrolysis (FSP). First, 20 mL of tin (II) 2-ethylhexanoate was dissolved in 60 mL of toluene; the resulting solution was divided into four equal parts, each of them then slowly injected into the FSP reactor. After the completion of each injection, the apparatus was dismantled, tin dioxide powder was collected, and a clean filter was installed. The obtained portions of the powder were combined and annealed in air at 400 • C for 24 h.
For the modification by impregnation method, the solutions of Pt (II) acetylacetonate and Pd (II) acetylacetonate in ethanol were used as metal precursors. The calculated volume of the precursor solution was added to the weighed SnO 2 powder to obtain 1 wt.% total metal content (1 wt.% M for monometallic nanocomposites and 0.5 wt.% Pd + 0.5 wt.% Pt for bimetallic one) and ethanol was allowed to evaporate. All impregnated samples were then annealed at 300 • C for 24 h for decomposition of acetylacetonates. These annealing conditions corresponded to the lowest temperature, which ensured complete decomposition of both Pd(acac) 2 and Pt(acac) 2 that was proven using thermal analysis. A reference sample of SnO 2 was created using an annealing undoped matrix material at 300 • C for 24 h.
The elemental composition of mono-and bimetallic nanocomposites was determined by X-ray fluorescence (XRF) analysis using a M1 Mistral (Bruker) micro-X-ray spectrometer. The phase composition of the samples was determined by XRD using a DRON-4-07 diffractometer (CuK α , λ = 1.5406 Å). The crystallite size of SnO 2 phase (d XRD ) was calculated by the Sherrer formula using 110 and 101 reflections.
The specific surface area was measured on Chemisorb 2750 instrument (Micromeritics) using a low-temperature nitrogen adsorption using single point BET model.
The microstructure of the samples and distribution of modifiers in SnO 2 matrix were investigated using transmission electron microscopy (TEM), high angle annular dark field scanning transmission electron microscopy (HAADF-STEM), and energy dispersive X-ray (EDX) mapping, all accomplished using a FEI Osiris microscope equipped with a Super-X detector operated at 200 kV.
The chemical state of the elements was studied by X-ray photoelectron spectroscopy (XPS). The measurements were effectuated on K-Alpha (Thermo Scientific) spectrometer equipped with a monochromatic Al K α X-ray source (E = 1486.7 eV). The positions of the peaks in the binding energy scale were adjusted with a C1s peak (285.0 eV) that corresponded to the carbon contamination of the surface with an accuracy of 0.1 eV. XP-spectra were fitted by Gaussian-Lorentzian convolution functions with simultaneous optimization of the background parameters.
Active surface sites with oxidizing properties were investigated using the thermo-programmed reduction with hydrogen (TPR-H 2 ) method. The experiments were carried out on Chemisorb 2750 (Micromeritics) in a quartz reactor at a gas mixture flow of 10% H 2 in argon at 50 mL/min and at a heating rate of 10 • C/min to 900 • C.
The molecules adsorbed on the surface of materials were studied using the Fourier Transform infrared spectroscopy (FTIR) method. The IR spectra of the samples were taken on a Spectrum One (Perkin Elmer) spectrometer in transmission mode within the wavenumber range 400-4000 cm −1 with 1 cm −1 steps. The powders (5 mg) were grinded with 100 mg of dried KBr (Aldrich, "for FTIR analysis") and pressed into tablets.
To perform the sensor tests, the powders were deposited onto microelectronic transducers equipped with Pt contacts and heaters in form of thick films. The sensors were placed into a gas flow chamber under conditions of a controlled gas flow of 100 ± 0.1 mL/min and operated by a resistance-measuring device connected to a PC. The DC-resistance was registered in situ under changing conditions in a temperature range of 60-300 • C. CO-air mixture containing 6.7 ppm CO was used as a test gas. The required level of humidity (RH = 0% and RH = 20%) was provided by mixing two streams of dry air and humid air using the membrane humidifier Cellkraft P-2.

Conclusions
The nanocrystalline SnO 2 was synthesized using flame spray pyrolysis and used as semiconductor matrix to obtain the mono-and bimetallic nanocomposites SnO 2 /Pd, SnO 2 /Pt and SnO 2 /PtPd. It was found that in monometallic nanocomposites SnO 2 /Pt and SnO 2 /Pd, platinum forms large (several tens of nanometers) particles, PtO x , while palladium was distributed as small (including less than 2 nm) PdO x nanoparticles on the SnO 2 surface. In bimetallic nanocomposite SnO 2 /PtPd, platinum was located in large bimetallic PtPdOx particles with a different Pt/Pd ratio, but palladium was also present in the form of small nanoparticles PdO x .
The surface characteristics of bimetallic SnO 2 /PtPd nanocomposite were not additive as compared with monometallic SnO 2 /Pt and SnO 2 /Pd samples. Thus, active surface sites with oxidizing properties were determined by the presence of Pt-containing particles, while PdO x nanoparticles were responsible for the increase in the surface hydroxyl concentration.
The sensor properties of bimetallic SnO 2 /PtPd nanocomposite can be explained by the combined effect of modifiers on the electrophysical properties of SnO 2 . The inversion of the sensor response from nto p-type observed for SnO 2 /PtPd nanocomposite in dry air was due to a change in the type of main charge carriers in the near-surface layer of SnO 2 . Furthermore, the chances was due to the formation of a deep and extended electron depleted layer in the area of SnO 2 contacts with platinum-enriched bimetallic PtPdO x particles and PdO x nanoparticles. In humid air, the adsorption of water vapor on SnO 2 led to an increase in the concentration of electrons. As a result, electrons remained the main charge carriers and no inversion of sensor response was observed.