2.1 Film preparation and processing

Mesoporous Ti – Si mixed oxides thin films were prepared by EISA [4], under controlled deposition conditions, as previously reported for pure oxide films [20]. Non-ionic triblock copolymer Pluronic F127 ([EO]₁₀₆[PO]₇₀[EO]_106; EO = ethylene oxide, PO: propylene oxide) that gives rise to pore diameters in the 8 to 10 nm range was used as a template.

The sol composition was TiCl₄:Si(OEt)₄:F127:H₂O:EtOH 1-x:x:0.005:10:40 with x varying between 0 and 1. Direct mixing of all the components led to mesoporous films with homogeneous walls albeit locally ordered mesopores. In the Ti-rich range 0 ≤ x ≤ 0.2, the solution was prepared by mixing two precursor solutions A and B, in the appropriate molar ratio. The composition of A was TiCl₄:F127:H₂O:EtOH 1:0.005:10:40. The composition of B was Si(OEt)₄:F127:H₂O:EtOH 1:0.005:10:40, where Si(OEt)₄ was previously hydrolyzed in acidic conditions described elsewhere; [21] solution B was aged for at least 2 days before mixing with A. These conditions ensure homogeneous materials with highly ordered mesopores. A label PH (for pre-hydrolysis) was added to these samples.

All sols were transparent and stable at ambient temperature, and can be re-used several times if conserved in a freezer at −18 °C, and gently restored to room temperature prior to dip-coating.

Precursor solutions were used to produce mesoporous thin films by dip-coating onto silicon wafers or glass substrates under 50% relative humidity at 25 °C, with a withdrawing rate of 2 mm s⁻¹. As-prepared films were submitted to subsequent 24-hour treatments at 50% RH, 60 °C and 130 °C to improve cross-linking of the inorganic network, and favor the formation of ordered micellar arrays. Thermal treatment of these stabilized coatings (up to 350 °C, 2 hours) was performed in a tubular oven, under still air, using 1 °C min⁻¹ temperature ramp.

The films were called TSab were ab represents the molar proportion between Ti and Si in the mixed oxide. For example, a film containing 80% of Ti and 20% of Si was called TS82.

In order to compare with the porous films, non porous thin films were prepared in the same conditions described before, but without the template agent.

2.2 Mesostructure and porosity characterization

The film mesostructure was characterized by small angle X-ray scattering (2D SAXS) at the D02A-SAXS2 line at the Laboratório Nacional de Luz Síncrotron LNLS (Campinas, SP, Brazil) [22], using λ = 1.608 Å, a sample-detector distance of ∼690 mm, and a CCD detector (3° incidence).

XRR measurements were performed at the D10A-XRD2 line at LNLS, using λ = 1.5498 Å. Two main parameters are obtained from the XRR data: Δθ between maxima or minima of the reflection and the critical angle (θ_c), defined as the angle in which the intensity is half of the maxima [23].

Thicknesses (t) were obtained from Δθ (with θ > 4θ_c) using (1):

Film densities (δ_m) were calculated from the critical angle using (2):

where: λ = X Ray wavelength, N_A = Avogadro's number, r₀ = electron classical radius, c_j = j element fraction in the material, Z_j and A_j = j element mass and atomic number, respectively.

Densities of porous (δ_m) and non porous (δ_mnp) materials can be obtained with (2). Thus, the materials porosity can be calculated using (3):

Eq. (3) considers a two-phase system composed of the matrix (with δ_m), and void pores.

Ellipsometric porosimetry analysis (EPA) [24] of selected samples permitted to obtain refractive index, thickness and pore size distributions. EPA was performed in a SOPRA GESE5 spectrometric ellipsometer with a controlled vapor pressure chamber. Film thickness and refractive index spectra for single layer thin films were obtained from adjusting the ellipsometric parameters Ψ(λ) and Δ(λ) at each P/P_s value using Win Elli II software. A Cauchy model including Lorentzian peaks in order to model light absorption in the UV spectrum was used to fit the complex refractive index in the 200–800 nm range. Film porosity was evaluated by adjusting a three-medium (air, water, oxide) Bruggeman effective medium approximation (BEMA). Pore size distributions were estimated according to a simple Kelvin model. Pore radii were obtained from the adsorption branches of the EEP; pore necks were obtained from desorption curves [25]. Thickness values obtained by ellipsometry were cross-checked with FE-SEM and XRR measurements; an agreement within 10% was obtained in all cases.

TEM images were collected using a Philips EM 301 transmission microscope (CMA, FCEyN, UBA) operated at 60 kV or a Philips CM 200 Super Twin (UAM, CAC, CNEA) operated at 200 kV and equipped with EDAX^®. Samples were obtained by scratching the films from the substrate and depositing them on carbon coated copper grids.

2.3 Framework homogeneity characterization

FTIR were performed on a Nicolet Magna 560 instrument, equipped with a liquid nitrogen cooled MCT-A detector. Films were deposited onto Si wafers (University Wafer, South Boston, MA), transparent in the IR region.

UV-visible measurements were performed on a Hewlett Packard 8453 spectrophotometer equipped with a diode array detector, in the 200–1000 nm range. For these measurements, films were deposited on Pyrex 7740 glass (University Wafer, South Boston, MA), transparent in the UV-visible region.

Ti K-XANES spectra were acquired using a fluorescence detector and a Si(1 1 1) monochromator at the D04B-XAS beamline of the LNLS [26]; the resolution was about 0.8 eV. The beam intensity was measured using an ionization chamber filled with air at room temperature. Photon energies were calibrated during the experiments in transmission mode by using a Ti foil placed between the two ion chambers located at both sides of the reference sample and setting the first inflection point at the known absorption edge energy of Ti⁰ (4966 eV).

The region of the absorption spectra below 4960 eV was fitted with a Victoreen function to subtract the background contributions throughout the entire energy range. Absorption intensities were normalized by their average value between 5050 and 5200 eV [27]. Four features in the pre-edge region were identified, in agreement with previous studies [28–30]. After background subtraction and normalization, this region was fitted with four Gaussian peaks [31], labelled in ascending energy as A1, A2, A3, and B [29,32] (Supplementary information, SI).

Si K-XANES spectra were acquired in total electron yield mode (TEY) and using a InSb(1 1 1) monochromator at the D04A-SXS beamline of the LNLS [33]. The beam focalization was performed using a Rh mirror. The incident photon energy was in the range of the Si K-edge for the experiments. The resolution was about 1 eV. The I₀ beam intensity was measured using a thin foil of Al located before the main chamber. The photon energies were calibrated in TEY mode using a Si foil and setting the first inflection point to the energy of the K absorption edge of Si⁰ (1839 eV). Two polynomials functions were fitted to discount the background and normalize the spectrum, one for the pre-edge region and the other for the region beyond the edge. After background subtraction and normalization, the white line (WL) region was fitted using Gaussian peaks (SI).

2.4 Ag nanoparticle synthesis

Ag NPs were synthesized within the mesopore systems of the thin films as follows: calcined mesoporous films were immersed in a 1:1 water:ethanol mixture of AgNO₃ 0.05 mol dm⁻³ for 15 minutes. Then, 7% w/w formaldehyde (HCHO) solution was added until a 1% w/w final concentration was reached and left under stirring for 3 hours. After 15 to 20 minutes, a noticeable change in color was observed in the mesoporous films; coloration was more intense in films containing higher titania loadings. Silver loading (i.e., the pore volume fraction occupied by the nanoparticles) was obtained from the analysis of the critical angle in XRR measurements, as previously described [34]. In order to obtain accurate density values, measurements were done under dry nitrogen flux, to avoid atmospheric water condensation within the pores that leads to the underestimation of mesoporosity.

3.1 Synthesis and structural characterization

Mesoporous thin films were obtained by dip-coating under controlled conditions followed by careful post-processing. Two important factors were kept in mind, taking into account the sol-gel chemistry of the precursor solutions. The hydrolysis and condensation behaviour of both ions have to be matched in order to control order at two length scales: (a) molecular, i.e., walls with homogeneous composition, (b) mesoscopic, i.e. ordered arrays of monodisperse pores. Film samples were prepared either by a direct mixture of TiCl₄ and TEOS (all the Ti:Si ratio were explored) or by combining TiCl₄ and a prehydrolized solution of Si(OEt)₄, in the Ti-rich samples (0 ≤ x ≤ 0.2, x being the Si fraction in the mixed oxide).

Mesoporous oxide thin films with different degrees of order are obtained under the studied conditions, as can be seen in the TEM pictures and 2D-SAXS patterns showed in Fig. 1. Results are summarized and compared with the results obtained for the pure oxides in Table 1. All samples produced under the reported synthesis conditions appear homogeneous under TEM observation. No sharp contrast variation corresponding to massive phase separation (“islands”) [9] was observed in the micrographs, and metal ratios estimated by EDS presented homogeneous values along the samples (Fig. 1c). This is a first hint of framework homogeneity, at least at the mesoscopic level. These observations were confirmed by spectroscopic techniques, as detailed below.

Mesostructure, interpore distances and mesostructure contraction for samples heated at 200 °C were obtained from 2D-SAXS patterns. Samples derived from a direct mixture of the precursors lead to mesoporous oxides with local pore order throughout all the mixture range. Interpore distances in these samples shown in Table 1 were calculated from 3° or 90° patterns. For silica-rich systems (0.5 ≤ x ≤ 0.9), some influence of the Ti(IV) is observed, and pore size becomes smaller than in pure mesoporous silica with increasing Ti content; the obtained interpore distance values tend to 12.5 nm, similar to those observed in PH systems or pure titania. For x ≤ 0.5, interpore distances are significantly higher (ca. 14 nm), and correspond roughly to worm-like systems such as those observed in F127 templated worm-like titania obtained in conditions where condensation is favored (high temperature) [35]. In the case of titania-rich prehydrolyzed samples (PH, x ≤ 0.5), the analysis of 2D-SAXS patterns reveals that the interplanar (110) distances for cubic Im3¯m samples are in the 12.5 to 12.7 nm range, thus the cubic pattern a = 17.5 to 18 nm, in agreement with previously reported mesoporous titania [20]. In all cases, whether in the case of prehydrolyzed or direct mixing samples, mesostructure contraction for samples heated at 350 °C does not strongly depend on the composition of the oxide (all in the 40 to 50% range, Table 1). It has to be mentioned that samples heated at 350 °C are amorphous, and no wall crystallization process is evident until at least 700 °C in the examined samples; a similar behaviour has been reported in other mixed mesoporous oxide systems containing Si(IV) [9].

These results are the consequence of complex behaviour of the reacting systems at several length scales. Pore ordering and compositional homogeneity of metal centers depend strongly on the compatibility of the hydrolysis–condensation processes of the inorganic precursors [36]. In particular, it has been observed that conditions where condensation is favored lead to poorly ordered mesophases, due to the increased viscosity in the freshly formed liquid film that hampers reorganization from an initially locally ordered mesophase [37]. In the case of direct mixture systems, an increase in the condensation degree of the silica precursors with increasing Ti quantities could be due to the well established fact that small amounts of transition metal catalyze Si(IV) hydrolysis and condensation [38–40]. The higher the titanium contents, the larger the silica condensation rate should be, leading to more disordered pore arrays. On the other hand, the use of pre-hydrolyzed silica can lead to smaller Si-oxo clusters with diminished reactivity in the presence of Ti(IV) centers, permitting a certain flexibility of the inorganic networks. This could explain the more extended organization observed in PH systems.

In the following, only Si-rich films formed by a direct mixture of the precursors (0.5 ≤ x ≤ 0.9) and Ti-rich films formed under PH conditions (x ≤ 0.5) will be discussed; these were the most reproducible conditions to obtain homogeneous mesoporous samples.

XRR permits one to accurately evaluate film density and thickness [23]. Figure 2 shows the dependence of the mass density of non-mesoporous and mesoporous films, obtained from the critical reflection angle, with the silicon content, x. Film composition was considered close to the nominal composition values, as confirmed by EDS measurements. Film thicknesses, densities and porosities calculated from XRR measurements are displayed in Table 2. According to XRR measurements, the thicknesses are slightly higher for the films containing higher concentrations of Ti, due to the increase in sol viscosity [5]. Film density remains relatively constant up to x ≤ 0.2 and decreases more markedly as x increases for x ≥ 0.5. These trends coincide well with the densities of the non-mesoporous mixed oxides, as shown in Fig. 2. Substitution of Ti by Si leads to a change in the inorganic matrix density. In this situation, there is little density variation when the less dense oxide is incorporated in the denser one and more notorious variations in the opposite case.

The density changes are closely related to refractive index variations, a material feature that is crucial for many of the projected applications of the mixed oxide thin films. Furthermore, in all the compositional range the porosity of the films, reflected in the difference between both curves in Fig. 2, remains almost constant. Therefore, it is possible to vary the refractive index of the thin film by changing the composition without losing porosity.

The XRR-derived porosity results were checked by EPA. Fig. 3 shows the EPA analysis of a TS82PH sample, and the resulting narrow pore size distribution, coherent with a highly ordered pore array. Pore diameters and interpore openings can be estimated at 7 and 3.7 nm, from the analysis of the adsorption desorption curves, respectively. In the case of a Si-rich worm-like sample such as TS19, pore distributions are somewhat less well-defined, probably due to a higher pore size dispersion. Both analysis techniques yield comparable results; indeed, XRR estimations constitute an excellent seed to initial EPA calculations. Overall, these two techniques demonstrated that mixed mesoporous oxides with a narrow distribution of accessible pores and a gradient of density (therefore, of refractive index) can be produced.

3.2 Spectroscopic characterization

FTIR, UV visible and XANES spectroscopic techniques were used to evaluate the interdispersion of the mixed oxide and the local environment of the different ions that compose the oxide framework.

FTIR analysis allows a straightforward determination of important information about mesoporous oxides, in particular, the presence of the template, and the existence of bands attributed to mixed oxides. All the spectra were acquired before and after thermal treatment and thoroughly analyzed in the 1500–650 cm⁻¹ region. Fig. 4a displays the FTIR spectra of all the oxides synthesized in this work, treated at 350 °C, with band assignation according to literature [41]. Three bands associated to the SiOSi bond are observed: LO₃ at ∼1200 cm⁻¹, ν_{as SiOSi} (TO₃) at ∼1050 cm⁻¹ and ν_{sym SiOSi} (TO₂) at ∼800 cm⁻¹. A medium intensity band that can be associated to superposition of ν_SiOH and titania-silica vibrations is also observed at 950 cm⁻¹ (see discussion below). The results for the films treated at 130 °C (Figure S4 in supplementary information) are analogous but harder to interpret due to the presence of intense bands of the template molecule (ν_COC around 1100 cm⁻¹, SI) that hinder a proper observation of the framework bands.

As the Si proportion decreases, two main trends are observed in the spectra: a decrease in the intensity of the ν_{sym SiOSi} and a displacement of the ν_{as SiOSi} to lower wavenumbers.

The first trend has been attributed to the disruption of the order in the SiO₂ framework [9,42] as the Ti concentration increases. This signal is indicative of an extended interface between Si and Ti, which implies a good interdispersion in the mixed oxide.

The second observation, the displacement of the ν_{as SiOSi} to lower wavenumbers as Ti concentration increases, is coincident with previous observation in Zr-Si mixed oxides [43–45]. It has been proposed that the position of the SiOSi asymmetric vibration is an indicator of the oxide dispersion. The displacement is attributed to an increase in the oscillator reduced mass (that changes from SiOSi to TiOSi) and the corresponding bond vibration frequency reduction. In fact, this is reflected in the continuous relationship between the band position and the oxide composition displayed in Fig. 4b.

Regarding the bands located around 950 cm⁻¹, the analysis is less straightforward. These bands are clearly observed in samples treated at 130 (SI) and 350 °C (Fig. 4a). Qualitative comparison of these bands with the silica framework TO₃ signal (ca. 1050 cm⁻¹) for each sample suggests that the 950 cm⁻¹ signals are intense for high silica content samples. It is not possible to extract more accurate quantitative information, due to the fact that spectra have been obtained at ambient conditions, around 40 % relative humidity, and they have not been normalized. According to the literature, absorption in this region can be mainly attributed to two processes:

• (a) sol-gel derived silica films presenting high surface area, or before full densification, exhibit a band of medium intensity centered near 940–960 cm⁻¹ due to Si–OH stretching vibrations, overlapped, or coincident with Si–O⁻ groups [41];
• (b) previous work on mixed Ti-Si oxides attributes the band to Si-O-Ti vibrations [46]. This assignment has been confirmed with isotopic exchange between ¹⁶O and ¹⁸O in dehydrated samples of a Ti-zeolite composite [47].

Under our explored conditions, the most likely hypothesis is that the infrared absorption observed in the 950 cm⁻¹ region could be ascribed to the superposition of both kinds of bands. This can explain the noticeable band at x = 0, as well as the qualitative absorption increase for mixed oxide bands with increasing Ti concentration for low titania contents (x < 0.5), where ν_SiOH due to surface groups can be important.

UV – visible absorption has been proven to be a useful tool to determine the coordination of transition metals such as Ti in solids or complexes. UV-visible spectra of mixed and pure oxides prepared in this work are presented in Fig. 5. Depending on Ti concentration, the observed bands can be attributed to Ligand to Metal Charge Transfer (LMCT) from O²⁻ to Ti⁴⁺ isolated centers, or to electron transfer between valence and conduction bands in the mixed oxide.

A continuous behaviour is clearly observed: those oxides having a high concentration of Ti, with spectra closely related to pure TiO₂, and those oxides having high concentration of Si, with spectra closely related to pure SiO₂, TS55 being an intermediate case. Spectra of TS91 and TS82 present the same three absorption peaks attributable to the TiO₂ band structure (206, 235 and 267 nm), with a slight shift in the position of the absorption border (263 nm in TS82) to lower λ. Less intense absorption bands in TS55 are observed at 206, 231 and 258 nm. On the other hand, in the case of TS28 and TS19 two or one band of lower intensity can be distinguished at 205 and 220 nm, clearly displaced to lower λ values respect to pure TiO₂. This band shift to lower wavelengths as Ti concentration decreases (or x increases) has been previously observed by UV diffuse reflectance in several Ti-Si mixed oxides prepared by sol gel method [48–51]. Moreover, the band position has been related to different local environment of the Ti center surrounded by O atoms, as shown in Table 3. A clear tendency is deduced from the literature data: isolated tetrahedrically coordinated Ti centers appear at lower λ; as the coordination number increases from 4 to 6 (octahedral coordination), the band is displaced to higher wavelengths.

According to this correlation between coordination number and position of the absorption band, it can be proposed that when Ti is the minor component of the mixed oxide, its coordination is lower and closer to the one of Si in SiO₂ (4). In contrast, when Ti is the major component, its coordination number is close to the one of Ti in the TiO₂ matrix (6). The Ti capability to act as a network modifier of silica, having a coordination around 4, has been previously demonstrated [5,7b]. These observations about the coordination number of the Ti are another proof of the high degree of mixing that present the mixed oxides prepared in this work. The band shift to higher energies (lower λ) with increasing Si contents can be attributed to two different phenomena:

• • formation of TiO₂ clusters of lower size as Si concentration increases. The lower size of the aggregates is correlated with an increase in the transition energy, due to quantum size effect of semiconductors [50,51];
• • a continuous modification of the band structure of the oxide and a separation of the bands as going through SiO₂ (higher E_gap) to TiO₂ (lower E_gap).

The first hypothesis implies that the mixed oxides are not really mixed at the atomic scale, (but are still mixed at TEM-EDS level). The second one implies the formation of mixed oxide at the atomic level.

In order to confirm the high interdispersion of the inorganic centres in the matrices, XANES experiments at the Ti and Si K-edges were performed. Fig. 6 shows the Ti K-XANES spectra for the extreme samples: a. TS19, b. TS28, c. TS82, and d. TS91, and amorphous TiO₂ (am-TiO₂) and anatase TiO₂ (a-TiO₂) reference compounds.

The relationship between position and intensity of pre-peaks provides information on the coordination number of the absorbent atom [52]. In particular, it has been observed that the peak intensity decreases while the corresponding energy increases with the center-symmetry degree of the absorbent. In the Ti K edge pre-peaks region, a very intense peak is observed, which is associated with tetrahedrical symmetry, shifted 3 eV towards lower energy than in the case of less intense pre-peaks associated with octahedral symmetry [52].

The pre-edge features in Ti K-edge XANES spectra of a-TiO₂ consist in four peaks (labeled as A1, A2, A3 and B [53], SI). A1 corresponds to 1s → 3d quadrupolar transitions [53,54], while A2 plus A3 and B peaks correspond to transitions from the Ti 1s to t_2g and e_g bands respectively [29]. Even if the definitive description of the origin of the pre-peaks is still open, it is known that A2 has higher intensity for XANES spectra of Ti compounds with low coordination [53–55], directly reflecting the very small size of nano-domains of TiO₂ through the increment of the A2 intensity [56]. The dipolar transitions 1s-np are contained in a region between 4980 and 5020 eV [52].

The spectra in Fig. 6 can be quantitatively described by adjusting the positions and areas of four peaks at the pre-edge region, which was fitted with Gaussian curves A1, A2, A3, and B, and a Gaussian shape for background. Taking into account the spectra of am-TiO₂ and a-TiO₂, it can be stated that even the sample with higher Ti contents (TS91) presents amorphous characteristics.

In previous papers [31,57] we proposed a methodology to determine the relative fraction of Ti with varying coordination number. The local symmetry of Ti with six near neighbours (oxygen atoms) can be associated with Ti in crystalline TiO₂ anatase [32]. The same type of analysis developed here on the average coordination number N, indicates the presence of a fraction of crystalline (local Ti environment) TiO₂ less than 45% in the TS91 film. Therefore, even the presence of small silica quantities disrupts the Ti-O-Ti network.

A numerical fitting of the peaks in the pre-peak region with Gaussian curves shows a broadening and shifting position towards higher energies for A3 in the samples TS82PH and TS91PH (Fig. 6). As reported by Farges et al. [52], this corresponds to an increase in the coordination number (from n = 4 to 5). The increase in the coordination number should correspond to the coexistence of am-TiO₂ (with a fraction of Ti sites with n = 4) with crystalline a-TiO₂ (where Ti sites have n = 6). Mountjoy et al. [58] have shown that the presence of small amounts of Ti in SiO₂ induce the decrease of the Ti coordination number because of the lower coordination of Si. This fact is evidenced by present XANES results obtained at the pre-peak region.

As it was mentioned, peaks at A2 plus A3 and B regions correspond to transitions from the Ti 1s orbital to t_2g and e_g bands, respectively. Fig. 6 shows a change of position and intensity of peak B, in addition to changes in intensity of A2 and A3 peaks. These correspond to variations at t_2g and e_g bands due to the change in the structural configuration of Ti environment. The position of B in the TS28 and TS19 samples is similar to the corresponding position in Ti K-XANES spectra of a-TiO₂. In samples with low percentage of Ti the transitions to the e_g band are partially forbidden. This is in close agreement with UV-visible results, which indicates the presence of two different types of samples: those samples having a high concentration of Ti, with spectra closely related to pure TiO₂, and those having high concentration of SiO₂, with spectra closely related to Ti included in SiO₂.

On the other hand, Natoli's rule [59] allows a simple correlation between the energy difference ΔE of the resonance peaks [60] (Fig. 7) with the distance r between scatter neighbours and their ligands, expression (4). The corresponding values obtained for Ti – O distances in our samples using this rule are shown in Table 4.

Taking into account the comparison of the Ti-O distances between am-TiO₂ and a-TiO₂ (Table 4), it can be noted that the Ti-O first neighbour distance in the a-TiO₂ compound is shorter than in am-TiO₂. This is consistent with a simple electrostatic consideration, which indicates that a more orderly electrostatic ion configuration is adopted for the minimum energy. This energy is precisely reached by minimizing the Ti-O distance in the a-TiO₂ compound. On this basis, it can be assumed that the shorter Ti-O distance indicates a greater degree of local crystalline order. This is in agreement with the results reported by Mountjoy et al. [61] (and reference therein), where it is also reported that the coordination number of Ti increases with the increase of the molar concentration of TiO₂ in TiO₂-SiO₂ mixed oxide xerogels.

On the other hand, an inspection of Si K XANES spectra (Fig. 8) shows that all spectra have a similar feature than those corresponding to Si in SiO₂. This region of the spectrum corresponds to the dipolar transition 1s → t₂ (Si 3p-3s) [62].

A complete replica of the XANES spectra can be obtained by fitting with four Gaussian curves [63]. The first one (A) centered at 1844.34 eV is attributed to the forbidden electronic transition 1s → a₁ (Si 3s-3p) (because for this transition Δl = 0). The second one corresponds to the white line (B), centered at 1847.05 eV, associated with the 1s → t₂ (Si 3p-3s) transition. A third one (C) at 1853.50 eV is due to the 1s → e (Si 3d-3p) transition and finally the fourth one (D) at 1863.66 eV corresponds to the 1s → t₂ (Si 3d-3p) transition.

The more noticeable difference between the spectrum of the reference compound and the samples is the lower intensity of the white line (WL) for the latter. The mentioned area is about the same for all samples and also with respect to the amorphous SiO₂. The only confident difference between areas, but minor, can be found in peak A. In effect this higher intensity for this forbidden transition in the spectra corresponding to samples with lower Si content suggests the lower symmetry for Si sites when this element is highly diluted in the sample. This is the only fact affected because the fraction of Ti present in the sample, being the remains aspects constant.

In conclusion, a set of spectroscopic techniques permits one to confirm the high interdispersion of Ti and Si centres attained by this synthesis path. Bonding of Si centres to Ti along all the composition axis has been evidenced by shifts in the Si-O-Si vibration modes. UV-Vis and XANES spectroscopy permitted to corroborate two limiting behaviors for Ti(IV) centres: at low concentration (x ≥ 0.8), Ti substitutes Si in tetrahedral sites, the coordination increases steadily for larger Ti:Si ratios. The conditions of sample preparation chosen here seem to preclude extended condensation of silica and titania species, thus leaving the opportunity to create flexible mixed building blocks that lead to non-segregated mixed mesoporous oxides.

3.3 Selective catalytic properties of mixed mesoporous thin films

One of the interests in high-area mixed oxides is the possibility to combine two metal centres with different acidity, in order to be able to regulate processes such as heterogeneous catalysis. The mesoporous mixed oxide films developed and characterized in the preceding sections present interesting properties regarding tuned reactivity towards metal reduction. It is well known that chemical reduction of silver is favored on the surface of transition metal oxides. Recently, this concept was applied in order to selectively deposit silver nanoparticles within mesoporous titania films [34]. The presence of a titania framework seemed decisive to trigger a fast formation of silver nanoparticles in the pore system, presumably by enhancing the reactivity of the reducing agent, formaldehyde. In order to test an eventual role of Ti(IV) centres in the reduction process, mesoporous mixed oxide films with variable composition were put in contact with solutions containing formaldehyde (1% w/w) and Ag⁺ (0.05 mol dm⁻³) for a fixed period of 3 hours. XRR measurements permitted one to determine the pore volume fraction loaded with metal. A typical XRR diagram of empty and loaded films, and the pore volume occupied by Ag nanoparticles in films with varying composition are displayed in Fig. 9. It can be observed that nanoparticle growth within the pore systems due to Ag⁺ reduction is notoriously favored in samples containing a higher Ti fraction, for a fixed reduction time. This fact demonstrates that mesoporous films with a tuned framework composition present a controlled catalytic behaviour for the production of nanoparticles. In addition, these films presenting a continuous variation of refractive index lead to the modification of the silver plasmon position and intensity (Supplementary information), due to matrix-nanoparticle interactions [64]. This is an example of the use of two different features of the same film (refractive index and catalysis) in order to obtain nanocomposites with tailored properties.