Nanocrystalline TiO2/SnO2 heterostructures for gas sensing

The aim of this research is to study the role of nanocrystalline TiO2/SnO2 n–n heterojunctions for hydrogen sensing. Nanopowders of pure SnO2, 90 mol % SnO2/10 mol % TiO2, 10 mol % SnO2/90 mol % TiO2 and pure TiO2 have been obtained using flame spray synthesis (FSS). The samples have been characterized by BET, XRD, SEM, HR-TEM, Mössbauer effect and impedance spectroscopy. Gas-sensing experiments have been performed for H2 concentrations of 1–3000 ppm at 200–400 °C. The nanomaterials are well-crystallized, anatase TiO2, rutile TiO2 and cassiterite SnO2 polymorphic forms are present depending on the chemical composition of the powders. The crystallite sizes from XRD peak analysis are within the range of 3–27 nm. Tin exhibits only the oxidation state 4+. The H2 detection threshold for the studied TiO2/SnO2 heterostructures is lower than 1 ppm especially in the case of SnO2-rich samples. The recovery time of SnO2-based heterostructures, despite their large responses over the whole measuring range, is much longer than that of TiO2-rich samples at higher H2 flows. TiO2/SnO2 heterostructures can be intentionally modified for the improved H2 detection within both the small (1–50 ppm) and the large (50–3000 ppm) concentration range. The temperature Tmax at which the semiconducting behavior begins to prevail upon water desorption/oxygen adsorption depends on the TiO2/SnO2 composition. The electrical resistance of sensing materials exhibits a power-law dependence on the H2 partial pressure. This allows us to draw a conclusion about the first step in the gas sensing mechanism related to the adsorption of oxygen ions at the surface of nanomaterials.

: Schematics illustrating the beneficial action of n-n heterojunctions for the sensitization of the gas sensor. (a) Electronic band diagram of an n-n heterojunction, b) electron transfer from a TiO 2 to a SnO 2 grain providing active gas adsorption sites. E F : Fermi energy, E VB : valence band maximum energy, E CB : conduction band minimum energy, E g : energy band gap, e − : electron, O − : singly ionized oxygen adatom.
In fact, we have found in our research that more than two cases are possible. Our previous experience with this system [1,4,6,7,13] indicates that four classes of materials can be obtained: A. a simple mixture of the constituents denoted as TiO 2 -SnO 2 B. perfect solid solutions Ti x Sn 1−x O 2 where 0 ≤ x ≤1 C. partially decomposed Ti x Sn 1−x O 2 -Sn y Ti 1−y O 2 where 0 ≤ x ≤ 1 and 0 ≤ y ≤ 1 D. decorated nano-heterostructures denoted as TiO 2 @SnO 2 , e.g., TiO 2 nanoflowers overcoated with SnO 2 nanoparticles Synergetic effects and catalytic reactions can be expected in the case of A) and C) while changes of the morphological and the electronic structure dominate in the case of B) and D). Surface phenomena determine the gas-sensor response in the case of decorated nano-heterostructures, D). As shown in [6], electron transfer over n-n-type heterojunctions can account for sensor sensitization in the cases of A) and D). The formation of n-ntype heterojunctions at the contact between SnO 2 and TiO 2 grains and its effect on the enhancement of the sensor response has been reviewed recently [14]. Figure 1 explains why the formation of n-n heterojunctions between TiO 2 and SnO 2 grains enhances the response of the gas sensor. In fact, the sensitization comes to an effect in the first step of reducing gas detection, namely the preadsorption of oxygen at the grain surface (in our case it is assumed to be in the form of O − as shown in Figure 1b). The efficiency of the O − adsorption process is greatly enhanced when a sufficient concentration of electrons is provided. It is usually assumed that SnO 2 grains are more suitable for oxygen adsorption, thus electron transfer from TiO 2 grains is necessary to increase the number of adsorption sites. Electron transfer from TiO 2 to SnO 2 is provided by an appropriate electronic configuration because both conduction (CB) and valence (VB) band edges of TiO 2 are above those of SnO 2 as shown in Figure 1a. The potential difference that is formed when TiO 2 and SnO 2 grains come to contact facilitates electron transport from TiO 2 to SnO 2 thus promoting oxygen preadsorption at the surface of SnO 2 grains.
The performance of a resistive-type gas sensor is inherently related to the form and number of oxygen species adsorbed at the surface of the sensing material in the first step [18]. The equation describing the oxygen chemisorption can be written as [19]: (1) Table 1: TiO 2 -SnO 2 systems for gas sensing. The best response is defined either as: R 0 /R (for n-type material + reducing gas and p-type material + oxidizing gas) or R/R 0 (for n-type material + oxidizing gas and p-type material + reducing gas), where R 0 denotes the electrical resistance in the reference atmosphere and R is the electrical resistance under exposure to the detected gas. where is an oxygen molecule in the ambient atmosphere, e − is an electron that can reach the surface, S is an unoccupied chemisorption site, and represents chemisorbed oxygen species with α = 1 for singly ionized forms, α = 2 for doubly ionized forms, β = 1 for atomic forms and β = 2 for molecular oxygen. Table 2 presents possible oxygen species that can be chemisorbed at the surface of the gas sensing material. Hydrogen is considered to react in a second step, at the surface of the oxides, with preadsorbed or lattice oxygen, which, in consequence, increases the electronic conduction. The surface reaction between hydrogen and oxygen can be described in general by the following equation: It has been observed that as the result of the two step interaction described above, the electrical resistance, R, of the sensor for any reducing gas can be expressed as [1,20]: where p gas is the partial pressure of the reducing gas while the power coefficient n is specific to the kind of the target gas and particular reaction with oxygen species preadsorbed at the surface of the semiconductor.
In our previous work [1] one can find an analysis of TiO 2 -SnO 2 in the form of polycrystalline ceramics and rf-sputtered thin films upon interaction with H 2 but not much is known in the case of nanomaterials. Since that time we have focused on commercial TiO 2 and SnO 2 starting nanomaterials for the sensing of H 2 and NH 3 [4,6,7]. One of the main conclusion from our latest research on commercial materials is that a small addition of TiO 2 to SnO 2 affects gas sensing characteristics to a large extent [4,6]. In this work, for the first time, we intend to demonstrate nano-heterostructures of the TiO 2 -SnO 2 system prepared by flame spray synthesis with application to gas sensing.
Flame spray synthesis is a well-known and efficient method for the synthesis of crystallized metal oxide nanopowders with particular morphology, e.g., spherical, monodispersed nanoparticles of TiO 2 with good photocatalytic properties [21][22][23]. However, its application to nano-heterostructures for gas sensing is not known.
The aim of the current work is to study the role of nanocrystalline TiO 2 /SnO 2 n-n heterojunctions for hydrogen sensing.
Within this work the detailed study on crystallographic structure, morphology, electrical properties, H 2 sensing behavior and the power-law nature of the electrical resistance of TiO 2 /SnO 2 heterostructures is presented. The influence of water adsorption and desorption on the electrical properties of TiO 2 /SnO 2 is also taken into account. The detection threshold is studied for the first time as a function of the chemical composition of TiO 2 / SnO 2 heterostructures.
The required composition and specific surface area (SSA) were obtained by adjusting the ratio of the precursors in the precursor mixture, the total flow rate of which was kept constant at 12.64 cm 3 ·min −1 . The total precursor concentration in the flame was kept constant at 1.5 mol·kg −1 . The precursor solution was fed by a syringe pump and was atomized with oxygen (583 cm 3 ·s −1 ) in a gas-assisted external mixing nozzle. The combustible aerosol was ignited by six oxygen-acetylene flamelets (C 2 H 2 , 217 cm 3 ·s −1 ; O 2 , 283 cm 3 ·s −1 ) and the produced particles were collected on glass-fiber filters (GF/A 150, Whatman) using vacuum pumps. The nanopowders of TiO 2 -SnO 2 did not require any post-synthesis heat treatment since the technique provides well crystallized nanostructures.
The specific surface area (SSA) was determined using Brunauer-Emmett-Teller (BET) nitrogen-adsorption isotherms obtained with a Beckman-Coulter SA3100 apparatus.
The crystallographic structure was analyzed on the basis of XRD patterns recorded in Bragg-Brentano configuration with the help of a Philips X'Pert Pro diffractometer. Based on Rietveld refinement it was possible to determine the weight fractions of cassiterite SnO 2 , rutile TiO 2 and anatase TiO 2 , the lattice constants and the crystallite sizes, d XRD .
The 119 Sn Mössbauer effect measurements were performed in transmission geometry using an MS-4 RENON spectrometer and CaSnO 3 as source. The Mössbauer spectra were fitted using a transmission integral in order to take into account the absorber thickness effects. The spectra were refined with quadrupole doublets of Lorentzian lines assuming a non-zero value of the electric field gradient at the tin site. Hyperfine parameters, the isomer shift, IS, and quadrupole splitting, QS, as well as the full width at half maximum of the Sn peaks, G, were found. The values of isomer shift are given relative to the CaSnO 3 source kept at room temperature.
Morphology of the synthesized TiO 2 -SnO 2 nanomaterials was studied by means of scanning electron microscopy (SEM) performed with a FEI Nova Nano SEM 200 apparatus. High-resolution transmission electron microscopy (HR-TEM) images were obtained using a FEI Tecnai TF 20 X-TWIN microscope.
Mapping of chemical elements and diffraction patterns were provided.
The electrical properties were investigated by impedance spectroscopy (IS) in the temperature range from 20 to 550 °C in air.
The impedance spectroscopy measurements were performed with a Solatron system (Fra 1260 + dielectric interface 1294). Experimental parameters and data acquisitions were controlled with the FRA software. A frequency range from 1 to 10 6 Hz was covered, with 10 mV amplitude. The impedance spectra were analyzed using the ZView software. An equivalent circuit containing one resistor and a constant phase element (CPE) was used for fitting.
In order to perform gas sensing measurements, the nanosensors were prepared in the form of tablets that were pressed from powders under a pressure of 25 MPa, then annealed at 400 °C and covered with planar silver electrodes. The detailed description of the experimental setup used for the H 2 -sensing measurements can be found in [26,27]. The desired hydrogen concentration was obtained by using mass flowmeters mixing synthetic air (reference gas) with H 2 (0.01% H 2 , 0.1% H 2 , 1% H 2 + Ar depending on the concentration range, i.e., 1-30 ppm, 5-300 ppm and 50-3000 ppm H 2 , respectively). The total gas (hydrogen mixture + air) flow rate was kept constant at 120 sccm. The measurements were carried out in dry atmosphere. The synthetic air contained less than 1 ppm of water vapor while that of hydrogen + argon mixture had less than 10 ppm of contaminants. The relative humidity level was verified to be of about 0-1% RH at room temperature.
Dynamic changes in the electrical resistance upon hydrogen exposure have been detected over a low-to-medium concentration range of 1-3000 ppm at a constant temperature between 200 and 400 °C. Measurements within an interval of 1-50 ppm H 2 were performed to determine the hydrogen detection limit. The sensor response S was defined as the ratio between the electrical resistance in the reference atmosphere, R 0 , and the electrical resistance upon interaction with hydrogen, R: (4) Results and Discussion  Table 3 recapitulates the results of XRD Rietveld refinement performed for TiO 2 /SnO 2 nanopowders, and the SSA values determined from BET measurements. The FSS parameters were chosen intentionally in order to obtain approximately the same SSA, and according to the expectation the SSA were found to be within a range of 54-62 m 2 ·g −1 , independent of the chemical composition.
As it can be concluded from Table 3  The main conclusion from Figure 2a is that one can observe a systematic shift of all cassiterite SnO 2 XRD peaks towards higher diffraction angles resulting from a decrease in the lattice constants a and c (Table 3) of the 90 mol % SnO 2 / 10 mol % TiO 2 nanopowder compared with 100% SnO 2 . This effect is typical and is usually interpreted as Ti substitution at Sn lattice sites [1]. The absence of non-identified peaks belonging to TiO 2 phases supports the conclusion that under these conditions a solid solution is formed (case B). However, the presence of heterojunctions between the small amount of TiO 2 grains well dispersed within the primary SnO 2 cassiterite phase cannot be excluded.
The influence of 10 mol % SnO 2 in TiO 2 on the XRD pattern ( Figure 2b) is much more pronounced. As observed previously for polycrystalline ceramics and thin films [1], even a relatively small amount of SnO 2 in TiO 2 results in a dramatic reconstruction of the crystallographic structure. In order to study the possible tin oxidation states, Mössbauer spectroscopy was applied. Figure 3 demonstrates transmission spectra of: a) SnO 2 ; b) 90 mol % SnO 2 /10 mol % TiO 2 and c) 90 mol % TiO 2 /10 mol % SnO 2 nanopowders.     [29][30][31]. Figure 4 shows the dynamic responses of the electrical resistance of 90 mol % SnO 2 /10 mol % TiO 2 and 10 mol % SnO 2 / 90 mol % TiO 2 heterostructures upon interaction with hydrogen at a constant temperature of 400 °C. As one can see in Figure 4a and Figure 4c, the electrical resistance decreases upon admission of reducing gas (hydrogen). Thus we can conclude that globally both heterostructures (SnO 2 -rich and TiO 2 -rich) exhibit n-type conductivity. This is not surprising because usually SnO 2 and TiO 2 are treated as n-type semiconductors [32,33]. Moreover, from the comparison of the gas sensing responses given in Figure 4a and Figure 4c it is easily seen that the heterostructure of 90 mol % SnO 2 /10 mol % TiO 2 is very sensitive even to small H 2 concentrations (5-300 ppm H 2 , Figure 4a) while the TiO 2 -rich composition, i.e., 10 mol % SnO 2 /90 mol % TiO 2 requires higher hydrogen concentrations (50-3000 ppm H 2 , Figure 4c).
SEM and HR-TEM images, as well as the results of selected area electron diffraction (SAED) and mapping of elements are given in Figure 4b and Figure 4d for the gas sensing materials given in Figure 4a and Figure 4c, respectively. There are some differences between the microstructure of 90 mol % SnO 2 / 10 mol % TiO 2 and 10 mol % SnO 2 /90 mol % TiO 2 . In the case of TiO 2 -rich heterostructures the grains are larger and spherical (Figure 4d), while for SnO 2 -rich compositions grains are smaller, more irregular in shape and elongated (Figure 4b). The spherical nanograins of TiO 2 are probably composed of smaller crystallites while separate SnO 2 grains were not identified by SAED for 10 mol % SnO 2 /90 mol % TiO 2 . Element mapping suggests that a small amount of Sn (Figure 4d) is finely dispersed within the TiO 2 matrix. In the 90 mol % SnO 2 / 10 mol % TiO 2 heterostructures Ti is well incorporated into SnO 2 building blocks (Figure 4b). Step changes in hydrogen concentrations are given on the right hand scale (a, c). Driven by the promising sensor signal for the step changes in H 2 concentration (Figure 4), we decided to perform additional measurements in order to determine the hydrogen detection threshold for the studied TiO 2 /SnO 2 heterostructures. Figure 5a and Figure 5b present dynamic changes in the electrical resistance of 90 mol % SnO 2 /10 mol % TiO 2 and 10 mol % SnO 2 / 90 mol % TiO 2 , respectively, upon interaction with 1 and 2 ppm H 2 . The H 2 detection threshold for the studied TiO 2 / SnO 2 heterostructures is lower than 1 ppm, especially in the case of SnO 2 -rich composition ( Figure 5a). As one discusses 10 mol % SnO 2 /90 mol % TiO 2 it appears that at 1 ppm H 2 the signal-to-noise ratio becomes much worse. However, the sensor signal is still discernible.
Within the studied temperature range SnO 2 -rich nanomaterials exhibit better gas-sensing performance (Figure 5c,d). The larger sensor response, R 0 /R (by about 20 times) for SnO 2 -rich heterostructures compared to TiO 2 -rich ones is typical as titanium dioxide requires higher temperatures for improved sensing characteristics.
In Figure 5c and Figure 5d one can also analyze the influence of the formation of heterostructures on the sensor response, R 0 /R. In both cases the sensor response increases compared to the pure oxides. The improvement in gas-sensing by a small addition of TiO 2 to SnO 2 was reported previously [5]. The explanation of this phenomenon is based on the charge transfer be- tween TiO 2 and SnO 2 due to the differences in the positions of the conduction and valence band edges of both oxides (Figure 1). A similar effect was reported in our previous work for 2 mol % TiO 2 /98 mol % SnO 2 nanocomposites working as H 2 sensors [4].
In this work, for the first time, based on Figure 5e, we can conclude that the addition of SnO 2 to TiO 2 (10 mol % SnO 2 / 90 mol % TiO 2 ) has a much more pronounced effect than the addition of TiO 2 to SnO 2 (90 mol % SnO 2 /10 mol % TiO 2 ). Moreover, one should also take into account the kinetics of interaction described by response and recovery times.
Despite the fact that SnO 2 -rich heterostructures exhibit larger responses to gases over the whole measuring range, it appears that their recovery time, τ, for the sensor to reach 90% of the initial electrical resistance, R 0 , is much longer than that of TiO 2rich heterostructures at higher H 2 concentrations ( Figure 6). In the case of 90 mol % SnO 2 /10 mol % TiO 2 (1100 ppm H 2 ), τ is about 2500 s, whereas for 10 mol % SnO 2 /90 mol % TiO 2 (1100 ppm H 2 ), τ is less than 30 s. The longer recovery time of SnO 2 -rich sensors can be attributed to a constricted gas desorption that probably results from the differences in the microstructure evidenced by SEM (Figure 4).
A fast desorption process is a prerequisite for the reproducible response and from this point of view TiO 2 -rich heterostructures exhibit better performance at higher H 2 concentrations (1000-3000 ppm). From the analysis presented in Figure 5 and Figure 6 one can make the conclusion that TiO 2 /SnO 2 nanoheterostructures can be intentionally modified by changing the chemical composition in order to meet requirements for successful detection of both small (SnO 2 -rich content) and large H 2 concentrations (TiO 2 -rich compositions). Figure 7 demonstrates the temperature variation of the electrical resistance in the reference gas (air), R 0 , its value upon interaction with 100 ppm H 2 , R, as well as the sensor response defined as a ratio of R 0 /R for the two compositions of 90 mol % SnO 2 /10 mol % TiO 2 (Figure 7a,b) as well as 10 mol % SnO 2 /90 mol % TiO 2 (Figure 7c,d).
It can be seen that R 0 decreases with increasing operating temperature. As for R, it seems that this effect is more pronounced for the TiO 2 -rich sample. In the case of the SnO 2 -rich composite the electrical resistance R upon interaction with 100 ppm H 2 seems to be independent of the temperature. For 90 mol % SnO 2 /10 mol % TiO 2 , the temperature dependence of R 0 /R follows R 0 vs temperature, because R is almost constant.
On the other hand, in the case of the TiO 2 -rich sample both R 0 and R exhibit a similar temperature dependence, which leads to a gas response R 0 /R almost independent of the temperature.
In order to study the electrical properties of TiO 2 /SnO 2 , impedance spectroscopy was applied. Figure 8 presents: a) the impedance spectra obtained at 400 °C as well as the electrical resistance as a function of the temperature for: b) 90 mol % SnO 2 / 10 mol % TiO 2 and c) 10 mol % SnO 2 /90 mol % TiO 2 .
The impedance spectra (IS) in Nyquist representation (Figure 8a) consist of a well-developed semicircle, followed by a deformed semicircle at lower frequencies. The equivalent circuit fitted to all spectra is a loop composed of one resistor R in parallel with a constant phase element CPE. Resistor R and CPE represent the bulk/surface process and their values have been determined by fitting (Figure 8a). The CPE in the majority of cases resembles a Debye capacitor C. The resulting electrical resistance R as a function of temperature exhibits a maximum, the position of which T max depends on the competing processes water desorption, oxygen adsorption and semiconducting behavior at higher temperatures. As can be seen in Figure 8b, for SnO 2 -rich heterostructures T max is about 100-125 °C, while for TiO 2 -rich heterostructures as shown in Figure 8c, T max is much higher within the range of 200-250 °C.
From the thermodynamics of chemical reactions it is well known that oxygen adsorption (described, e.g., by the coverage degree Γ) is an exothermic process and decreases with temperature [34]. Under the experimental conditions this situation is given when the adsorption processes remain in thermodynamic equilibrium, i.e., at temperatures larger than a characteristic value T eq . In the case of oxygen adsorption at the surface of oxides, T eq is of the order of 400 °C [35]. At temperatures smaller than T eq the coverage degree Γ increases with temperature as described by the laws of chemical kinetics.
The interpretation of the results given in Figure 8, assuming that the resistance changes are related only to the gas adsorption, is based on the fact that the experimental T max is much smaller than the theoretically predicted T eq . At these relatively low temperatures water desorption is believed to predominate over oxygen adsorption. However, both processes are possible. In the literature one can find three types of mechanisms explaining the increase in the surface conductivity in the presence of water vapor as in all these cases the electron concentration is increased [36]. Water adsorption becomes important at temperatures below T max and certainly at room temperature.
The subsequent increase in the temperature above 100-200 °C (see Figure 8b,c) leads to a decrease in the electrical resistance, which is a typical effect for semiconductors and is related to the creation of additional charge carriers.
The temperature T max at which the semiconducting behavior begins to prevail over water desorption/oxygen adsorption depends on the TiO 2 -SnO 2 composition. The higher T max for TiO 2 -rich heterostructures can be explained on the basis of the higher ionic defect concentration (mainly oxygen vacancies) at the surface of TiO 2 . It is well known that oxygen vacancies act as water adsorption centers. Moreover, in the case of SnO 2 water adsorption takes place because of the formation of weak van der Waals bonds between water dipoles and lattice ions (Sn 4+ and O 2− ) [19]. This facilitates water desorption from the surface of SnO 2 -rich heterostructures at lower temperatures.
As one discusses the interaction between the gas phase and the semiconducting sensor, the two-step mechanism described in the Introduction section has to be taken into account. The second step given in a general form by Equation 2 is the surface reduction, which appears upon interaction with hydrogen and can be described in detail as follows [1,37]: Applying the law of mass action to Equations 5-7 yields: where n e denotes the concentration of electrons, and K 1 , K 2 , K 3 are the equilibrium constants of the reactions described by Equations 5-7. The concentration of adsorbed oxygen is assumed to remain constant during its interaction with hydrogen. This is justified by p O2 >> p H2 and a high rate of oxygen chemisorption under the experimental conditions [1].
As the electron mobility μ e is practically independent of the gas partial pressure, and the relationship for the electrical resistivity reduces to ρ = 1/(e·μ e ·n e ) for n-type semiconductors, the 1/R(p H2 ) dependence assumes the same form as n e (p H2 ) does (Equations 8-10). Thus, n = 1/2, 1 or 2 are theoretically predicted for different oxygen species preadsorbed on the surface of the semiconductor.
In the case of formation of oxygen vacancies V O , the following reaction could be proposed: The condition of lattice electroneutrality requires that: where k = 1 or 2 corresponds to singly or doubly ionized defects, respectively.
Applying the law of mass action to Equation 11 (with k = 1 or 2) gives power-law coefficients of n = 1/2 or 1/3 according to the relation:    (13) Figure 9 and Table 4 demonstrate the results of the power-law analysis of the sensor response for: a) 90 mol % SnO 2 / 10 mol % TiO 2 and b) 10 mol % SnO 2 /90 mol % TiO 2 . In the log-log plot the dependence can be fitted with a linear function. The values of the power coefficient n corresponding to the predominating form, along with the experimentally determined values (Table 4) can be attributed either to the specific oxygen form preadsorbed at the surface of the sensor or to the oxygen vacancies following the equations given above.
In our case the parameter n is around 1 for 90 mol % SnO 2 / 10 mol % TiO 2 at temperatures of 200-300 °C and slightly higher than 1 at 350 and 400 °C (Figure 9, Table 4). According to the literature [19]  Considering the 10 mol % SnO 2 /90 mol % TiO 2 nanomaterial, n is in the range of 0.25-0.40. The reduction of titanium dioxide leads to the formation of oxygen vacancies.
For TiO 2 -rich nanomaterials, the sensing properties cannot be explained within this simplified model. It appears that not only oxygen species preadsorbed on the surface of the semiconductor but also formation of point defects need to be considered. 2) The detection threshold is below 1 ppm H 2 for SnO 2 -rich heterostructures.
3) The addition of a small amount of SnO 2 to TiO 2 has a much more pronounced effect on the sensor response than the modification of SnO 2 by a small amount of TiO 2 .
4) The recovery time of SnO 2 -based heterostructures is longer than that of TiO 2 -rich samples at higher H 2 concentrations. 5) TiO 2 /SnO 2 heterostructures can be intentionally modified in order to meet the requirements for the successful detection of both small (SnO 2 -rich) and large H 2 concentrations (TiO 2 -rich).
6) The temperature T max at which the semiconducting behavior begins to prevail upon water desorption/oxygen adsorption depends on the TiO 2 /SnO 2 composition.