Structure Sensitivity and Hydration Effects in Pt/TiO2 and Pt/TiO2–SiO2 Catalysts for NO and Propane Oxidation

The NO and propane oxidation activities of a series of 1%Pt/TiO2–SiO2 catalysts show different underlying trends as the support composition changes. Surface characterisation of the catalysts indicates that the trend for NO conversion is consistent with the oxidation rate being dependent on the degree of metallic character of the Pt nanoparticles, rather than their morphology. Although a similar correlation is expected for the total oxidation of propane, it is masked by the effects of adventitious ions originating during manufacture of the support materials. When residual chloride is present in the support, most of the exposed Pt is stabilised in its low-activity ionic form; while support materials containing W or oxidised-S ions give rise to catalysts with much higher activity than expected from their measured Pt0 content. When a Cl-containing, but SiO2-free, TiO2 support material is pre-treated hydrothermally, the propane-oxidation activity of the resultant Pt/TiO2 catalyst is substantially improved, so that it matches the performance of highly-metallic Pt supported on TiO2 containing 16 wt% SiO2. The hydrothermal pre-treatment removes residual chloride from the support material, but it also leaves the catalyst in a hydrated state. We show that, by controlling the metallic content of Pt nanoparticles, understanding the promoting and inhibiting effects of adventitious ions, and optimising the degree of catalyst hydration, the activity of 1%Pt/TiO2–SiO2 catalysts can be made to exceed that of a benchmark 2%Pt/γ-Al2O3 formulation for both NO and propane oxidation.


Introduction
Among the key functions of a diesel oxidation catalyst is its ability to convert both alkanes and NO [1][2][3]. The deep oxidation of alkanes allows hydrocarbon emission standards (e.g. [4] in EU) to be met by current diesel vehicles [5] and off-road machinery [6]. Although the conversion of NO to NO 2 can be a pivotal step in the passive oxidation of soot particulate in catalysed filters and during NO x adsorption catalysis, its most important role currently is to optimise the rate of selective catalytic reduction of NO x [7,8]. This dual alkane/NO functionality can be achieved, together with the other less demanding oxidation reactions of CO and alkenes, by using a relatively high loading of Pt or a Pt-rich bimetallic combination of Pt and Pd (typically 2-2.5% Pt by weight of the monolith washcoat) supported on γ-Al 2 O 3 [9,10]. TiO 2 is a potential alternative to γ-Al 2 O 3 as a support material for Pt-based oxidation catalysts [11]. It interacts strongly with platinum group metals, stabilising the metal nanoparticles against sintering, while also conferring chemical resistance and sulphur-tolerance to the catalysts [12]. Although commercial TiO 2 support materials (which are often bi-phasic mixtures of anatase and rutile [13]) have a moderate surface area, this can be increased and made thermally stable by the inclusion of SiO 2 [14]. As SiO 2 interacts very weakly with platinum group metals [15], its main role is thought to be in the control of the surface morphology of the TiO 2 [16], through which it can influence the dispersion [15] and oxidation state [17] of the Pt, as well as the extent of the interface between Pt nano-particles and the TiO 2 present in the support [18].
Here we show that, apart from increasing the surface area, the presence of SiO 2 in a TiO 2 support material can influence the surface Pt 2+ /Pt 0 ratio of the supported platinum, which is in effect a measure of exposed Pt 0 . The NO-oxidation activity in the kinetic regime, i.e. below the temperature at which the reaction becomes equilibrium limited (typically ~ 350 °C), shows an inverse correlation with the Pt 2+ /Pt 0 ratio, except when chloride is present. By contrast, deep oxidation of propane does not show this same simple dependence on the proportion of metallic Pt present at the surface. Instead, it is highly sensitive to the presence of Cl, W and oxidised-S ions, which can originate during the manufacture of the support materials. We also show that hydrothermal pre-treatment of a Cl-containing TiO 2 support material not only removes the adventitious ions, but results in the formation of a hydrated catalyst with high propane oxidation activity. In practical terms, knowledge of these effects enables the Pt loading to be reduced in catalysts used for both NO and alkane oxidation.

Catalyst Preparation
Proprietary support materials were sourced from two different specialist suppliers (TiO 2 and TiO 2 with 5 and 16 wt% SiO 2 from one supplier; TiO 2 and TiO 2 with 10 wt% SiO 2 from another supplier), which made it essential to start by comparing their bulk structures and surface compositions. Characterisation of the crystalline phases by X-ray diffraction (see below) showed that, depending on the source, each material contained either anatase alone or a mixture of anatase and rutile. A series of catalysts with 1wt% loading of Pt was prepared by impregnating these support materials with a non-aqueous solution of a chloride-free precursor: 'Pt(acac) 2 ' (Pt(II)2,4-pentanedionate, Alfa Aesar) dissolved in toluene (Fisher, HPLC grade). The suspensions were stirred for 24 h at room temperature. After removing the solvent in a rotary evaporator, the residue was dried at 100 °C for 16 h and calcined at 400 °C for 5 h under static air. Additionally, two benchmark catalysts were prepared using the same impregnation method to support 1 wt% or 2 wt% platinum on anhydrous γ-Al 2 O 3 (Merck Performance Materials).
Hydrated variants of the TiO 2 -containing catalysts were prepared by hydrothermally pre-treating the support materials. In simple hydration, suspensions were formed by adding deionised water (100 cm 3 ) to the support materials (4 g); while in basic hydration, aqueous NH 4 OH (1 mol dm − 3 ; 100 cm 3 ) was added to the support materials (4 g) to form suspensions with a pH of 11-12. The suspensions were heated at ≈ 110 °C under reflux for 24 h, before the solids were isolated (by filtration) and dried (16 h at ≈ 70 °C). The resulting hydrated materials were then impregnated with platinum precursor, dried and calcined following the method described above.

Catalyst Notation
Our notation indicates the wt% loading of Pt in the catalysts and of any SiO 2 present in the support materials. For example, 1%Pt/TiO 2 -10%SiO 2 represents a catalyst in which 1% by weight is Pt, with the Pt supported on a material that contains 90 wt% TiO 2 and 10 wt% SiO 2 . We use the term as-prepared to emphasise that a catalyst has not undergone any further treatment (e.g. reduction) following calcination, and to distinguish it from the hydrated variants.

Characterisation
The specific surface area of the support materials and catalysts was measured on a Micromeritics Gemini 2360 analyser coupled with a Micromeritics FlowPrep 060. A fivepoint BET method was used, in which nitrogen was adsorbed at − 196 °C, over the standard pressure range 0.05-0.35 p/ p 0 . Before the measurements were made, the samples were outgassed by exposure to N 2 at 120 °C for 45 min.
Pulsed chemisorption of CO was carried out in a flow of He, using a Quantachrome ChemBET TPR/TPD analyser, with the intention of estimating the platinum dispersion in the catalysts. Samples were pre-treated under pure H 2 at 300 °C for 2 h, at a flow rate of 30 cm 3 min − 1 . This pretreatment was expected to reduce any ionic species on the surface of the Pt nanoparticles formed during preparation of the catalysts, without changing the size or morphology of the nanoparticles. The apparent dispersion values were calculated from the total CO uptake at 30 °C, based on the assumption of 1:1 stoichiometry between adsorbed CO molecules and the number of exposed Pt surface atoms.
Powder X-ray diffraction (XRD) patterns were obtained on a PANanalytical X'Pert Pro diffractometer using a Cu Kα X-ray source operating at 40 kV and 40 mA. The signal was recorded for 2θ values between 10° and 80° at increments of 0.02°. The bulk components of the catalytic materials were identified by matching the diffraction patterns with standards in the JCPDS database.
X-ray photoelectron spectroscopy (XPS) analysis of the samples was carried out on a Kratos Axis Ultra DKD photoelectron spectrometer with a monochromatic Al source (1486.6 eV). All spectra were obtained at a pass energy of 40 eV for high resolution scans and 160 eV for survey spectra, over an analysis area of 700 × 300 µm 2 . The measured binding energies were normalised with respect to the C (1s) peak of the surface adventitious carbon (284.7 eV). Qualitative and quantitative interpretation of the data was carried out using CasaXPS software.
Scanning electron microscopy (SEM) images were obtained using a Carl Zeiss EVO 40 microscope operated at 25 kV. Samples were mounted on carbon Leit adhesive discs. Images were collected using backscattered and secondary electron detectors. For analysis by EDX (X-ray energy dispersive spectroscopy), high probe currents (up to 25 nA) were required to allow sufficient generation of X-rays. The data were collected using an Oxford EDX analyser coupled to the microscope.
Surface examination by high resolution transmission electron microscopy (HRTEM) was carried out on a JEOL-2100 JEM microscope, at an accelerating voltage of 300 kV. The samples were dispersed in ethanol and dropped on a copper mesh with a carbon micro-grid. STEM-EDX spectrum imaging experiments were performed on the same instrument.
Data from diffuse reflectance infrared spectroscopy (DRIFTS) were collected on a Bruker Tensor 27 spectrometer fitted with a liquid N 2 -cooled mercury cadmium telluride detector. Samples were placed within a 'praying mantis' high-temperature diffuse reflection environmental reaction chamber (HVC-DRP-4) fitted with zinc selenide windows. Prior to analysis, the samples (as-supplied and hydrated support materials) were ground to a fine powder using an agate mortar and pestle. For background scans, finely ground KBr was used. Spectra were recorded from 64 scans across the frequency range of 600-4000 cm − 1 .

Catalyst testing
Catalyst testing was carried out in a stainless steel fixedbed reactor (internal diameter: 0.5 cm), which was heated in a horizontal tube furnace. The typical catalyst charge was 30 mg of powdered material (particle size: 250-300 µm), packed between two plugs of quartz wool. When assessing alkane-oxidation activity, the gas-feed (flow rate: 200 cm 3 min − 1 ; GHSV: 50,000 h − 1 ) contained 0.5% propane in synthetic air (20% O 2 / 80% He), so that the C 3 H 8 :O 2 ratio was strongly fuel-lean. The reactants and products were analysed by online gas chromatography, using a Varian 3800 gas chromatograph fitted with two columns (HayeSep Q to separate CO 2 and hydrocarbons; molecular sieve 13X to separate CO, O 2 and N 2 ) and with thermal conductivity and flame ionisation detectors. The propane conversion and associated carbon balance were measured at catalyst bed temperatures between 100 and 550 °C, at 50 °C intervals. Several measurements were taken to ensure stable conversion had been reached at each temperature.
For NO-oxidation testing, the gas-feed was substituted by 1000 ppm NO in an oxidising atmosphere (7.5% O 2 , He balance), such that the NO:O 2 ratio was 1:75 by volume. The gas composition exiting the reactor was analysed continuously by an FT-IR spectrometer (Gasmet DX-4000) with an integrated 400 cm 3 heated analysis cell. Concentration measurements (in vol-ppm) were calculated using Calc-met™ software with internal calibration standards.

Characterisation and Evidence of Metal-Support Interactions
Bulk characterisation by X-ray diffraction (Fig. 1) showed that the support materials in the series TiO 2 (bi-phasic), TiO 2 -5%SiO 2 and TiO 2 -16%SiO 2 contained mixtures of anatase (≈ 80%) and rutile (≈ 20%); whereas TiO 2 (anataseonly) and TiO 2 -10%SiO 2 , which were both from a different source, contained anatase as the only crystalline phase. The BET surface area measurements were higher when SiO 2 was present and, as expected [18,19], each value decreased (by 1-10%) after deposition of the platinum (Table 1). 1%Pt/ TiO 2 -10%SiO 2 , in which the support was comprised of only one crystalline phase (anatase), had the highest surface area (96 m 2 g − 1 ); 1%Pt/TiO 2 (bi-phasic), which was made from SiO 2 -free bi-phasic TiO 2 had the lowest surface area (50 m 2 g − 1 ). The CO chemisorption values for the catalysts decreased (from a maximum of 0.24 µmol g − 1 for 1%Pt on bi-phasic TiO 2 ) as the proportion of SiO 2 in the support material increased, becoming negligible when the support material contained 16% SiO 2 (Table 1). For catalysts in which there is a weak interaction between the metal nanoparticles and the support material, such a trend would indicate a marked decrease in Pt dispersion (i.e. the Pt nanoparticles become larger) as the SiO 2 content increases. However, as discussed below, this effect on dispersion was not confirmed by electron microscopy. Instead, the trend is consistent with a strong metal-support interaction (SMSI) induced by reduction, resulting in a thin but impervious overlayer of reduced metal oxide forming on the metal nanoparticles, which blocks CO chemisorption. Although, in this study, the reduction temperature (300 °C) prior to chemisorption was lower than that used by Tauster et al. (500 °C) in their classic work on the induction of SMSI in Pt/TiO 2 [20], it has been shown that the presence of SiO 2 can promote SMSI by segregating TiO 2 into nano-domains that interact strongly with Pt nanoparticles [16]. As pointed out by Spencer [21], the threshold temperature for SMSI is not determined by thermodynamic equilibrium, but is dependent on the rate of surface diffusion of Ti-O species, which in turn will depend on the extent of contact between TiO 2 and Pt.
XPS survey spectra indicated the presence of adventitious elements (in their ionic form) on the surface of some of the support materials, which persisted even after Pt was added by impregnation to form the catalysts (Table 1). It is likely that these elements originated during manufacture of the support materials, and hence we deduce that the anataseonly TiO 2 (which contained oxidised S) had been made from the mineral source by a traditional sulphate route [22], whereas the bi-phasic TiO 2 (which contained Cl) had been made via TiCl 4 [23]; we suspect that the TiO 2 -10%SiO 2 (which contained W) had been made in a plant used to produce WO 3 -TiO 2 catalysts for selective catalytic reduction of NO x [24].
From the region for Pt 4f in the high resolution XPS scans (Fig. 2), the spectra could be fitted with two pairs of overlapping Lorentzian curves. The Pt 4f 7/2 and Pt 4f 5/2 lines at 71.4 eV and 74.7 eV are attributed to metallic platinum (Pt 0 ), while the second pair at 72.9 eV and 76.3 eV can be assigned to platinum oxide (PtO). The relative peak areas indicate that the atomic proportion of metallic platinum at the surface increased from 23% in Pt/TiO 2 (bi-phasic) to 80% in Pt/TiO 2 -16%SiO 2 . These results could be indicating the stabilisation of larger metallic nanoparticles when SiO 2 is present in the support material. However, HRTEM analysis of multiple areas of the surface of Pt/TiO 2 -5%SiO 2 and Pt/ TiO 2 -16%SiO 2 ( Fig. 3a and b) showed that the typical Pt particle size was between 1.5 and 5 nm in both catalysts. By contrast, small metal nanoparticles were difficult to find on SiO 2 -free Pt/TiO 2 by HRTEM, but larger particles (≈ 10 nm) could be identified from STEM images and EDX analysis (Fig. 3c). The results show that inclusion of SiO 2 in a TiO 2 support material increases the Pt dispersion, but the size of the Pt nanoparticles does not then greatly change as a function of SiO 2 content.
It was also apparent that the intrinsic activity was not primarily influenced by the anatase/rutile ratio (compare the activity plots in Fig. 4 for 1%Pt supported on anataseonly and the bi-phasic supports), the BET surface area (see values listed in Table 1) or the Pt particle size (see HRTEM and STEM data in Fig. 3). It did, however, show a dependence on the mean oxidation state of the surface platinum. With the exception of 1%Pt/TiO 2 (bi-phasic), which contained adventitious Cl, there was a near-linear inverse correlation between NO conversion (at 350 °C) and Pt 2+ /Pt 0 ratio (Fig. 5). This is consistent with the active sites for NO oxidation being exposed metal atoms at the surface of the platinum nanoparticles [27][28][29]. However, our results appear to conflict with the classic work by Ross and co-workers [30] on structure sensitivity of Pt/SiO 2 and Pt/Al 2 O 3 catalysts, which showed that the specific-activity for NO oxidation is influenced by the size of the Pt nanoparticles (with larger Pt nanoparticles resulting in higher specific-activity). The model proposed by Olsson and Fridell [31] allows our observations to be reconciled with those of Ross and coworkers [30]. In the Olsson-Fridell model, NO-oxidation activity primarily depends on the metallic character of the Pt nanoparticles, which in turn may depend on the size of the nanoparticles (e.g. when the support is SiO 2 or Al 2 O 3 , as used by Ross and co-workers [30].) The outlier in Fig. 5, 1%Pt/TiO 2 (bi-phasic), had the highest Pt 2+ /Pt 0 ratio (3.36) and hence the lowest surface concentration of metallic platinum, and yet its activity exceeded that of 1%Pt/TiO 2 -10%SiO 2 (Pt 2+ /Pt 0 = 1.3). Our surface characterisation of 1%Pt/TiO 2 (bi-phasic) suggests that the adventitious Cl − from the support material had stabilised most of the platinum in its ionic form, with the small proportion of Pt 0 forming relatively large particles (≈ 10 nm). Although there is scant literature on the effects of chloride on NO oxidation over Pt catalysts, work on NOx reduction by hydrocarbons [32] indicates that Cl − ions inhibit dissociation of chemisorbed NO on Pt. The relatively high NOoxidation activity that we observe for 1%Pt/TiO 2 (bi-phasic), despite the predominance of Pt 2+ at the surface, may reflect the role of Cl − ions in preventing active sites on the Pt becoming blocked by dissociated NO.

Propane oxidation
Propane is not a typical component of exhaust gas from an internal combustion engine, but as one of the most difficult alkanes to oxidise completely [33], it is often used to assess the ability of potential aftertreatment catalysts to eliminate the alkane component of hydrocarbon emissions. The propane oxidation activity of the SiO 2 -containing 1%Pt/TiO 2 catalysts gave rise to a succession of light-off curves, all of which were shifted to lower temperatures relative to the trace for 1%Pt supported on bi-phasic TiO 2 (Fig. 6). In light of published data [34] which show that the crystal form of TiO 2 does not have a marked effect on the VOC oxidation activity of Pt/TiO 2 , it was surprising that 1%Pt/TiO 2 (anatase-only) was much more active than 1%Pt/ TiO 2 (bi-phasic). It is also notable that the activity of each of the SiO 2 -containing catalysts was substantially greater than that of 2%Pt/Al 2 O 3 , despite having only 50% of the platinum loading.
When compared to NO oxidation (Fig. 5), the propaneoxidation activity of the TiO 2 -containing catalysts did not show a similar dependence on the mean oxidation state of the surface platinum (Fig. 7). Furthermore, although the presence of SiO 2 in the TiO 2 support material altered the textural properties of the catalysts, the changes in propaneoxidation activity could not be accounted for simply in terms of an increase in BET surface area (see Supplementary  Information Fig. 1) or by changes in Pt particle size. The most likely explanation for the non-linearity observed in Fig. 7 is that the activity was being affected by the adventitious ions present in three of the catalysts (Table 1), i.e. Cl in 1%Pt/TiO 2 (bi-phasic), S in 1%Pt/TiO 2 (anatase-only) and W in 1%Pt/TiO 2 -10%SiO 2 . It is known that propane oxidation over Pt catalysts is promoted by oxidised sulphur species [35][36][37] and by WO 3 [34,38], but inhibited by chloride ions [39,40].
Significantly, the propane oxidation performance of the least active catalyst (1%Pt on bi-phasic TiO 2 ) could be promoted to the level of the catalyst with the highest SiO 2 content (1%Pt/TiO 2 -16%SiO 2 ) by subjecting the TiO 2 to basic hydration before impregnation with Pt (Fig. 8). Even simple hydration of the bi-phasic TiO 2 improved the activity of the resultant catalyst (Fig. 8).
Both simple and basic hydration of the bi-phasic TiO 2 increased its bulk density by > 50%, but did not measurably alter either its bulk phase composition or its surface area (see Supplementary Information Table 2; Fig. 2). The most Fig. 6 Propane-oxidation activity of as-prepared catalysts as a function of temperature Fig. 7 Propane-oxidation activity of as-prepared catalysts at 300 °C as a function of mean oxidation state of surface Pt Fig. 8 Propane-oxidation activity as a function of temperature of asprepared 1%Pt/TiO 2 (bi-phasic), 1%Pt/TiO 2 after simple hydration of the support material and 1%Pt/TiO 2 (bi-phasic) after basic hydration of the support material. Compared to as-prepared 1%Pt/TiO 2 -16%SiO 2 obvious change was the absence of the chloride signal during XPS analysis of the hydrated materials, indicating that the adventitious Cl − ions had been removed by the hydrothermal pre-treatment. However, another perceptible change became apparent from characterisation by DRIFTS, but this was not in the regions of the spectrum where we were expecting so see the effects of chloride displacement by hydroxyl species on the surface of TiO 2 [41], namely between 3640 and 2500 cm − 1 (O-H stretching vibrations) and at 1630 cm − 1 (O-H bending). As can be seen in Fig. 9, the spectrum for the highly active variant of 1%Pt/TiO 2 (bi-phasic), formed following basic hydration of the support, showed the appearance of a small but broad peak at 1450 cm − 1 . This peak was also present in the spectrum of the highly active SiO 2 -free catalyst in which 1%Pt was supported on anatase-only TiO 2 , but not in the spectrum of a high SiO 2 -content catalyst (Fig. 9iii). Although the other peaks at lower wavenumbers (1050−800 cm − 1 ) can be assigned to the vibration modes for Ti-O-Ti [41][42][43], the peak at 1450 cm − 1 is not readily recognisable from published data for TiO 2 . However, a similar feature, which has been detected at 1449 cm − 1 by in-situ FTIR of Pt/Al 2 O 3 during propane and O 2 co-adsorption [38], has been attributed to the surface water that forms as the propane oxidises [38].
Overall, these results showed that a highly active propane-oxidation catalyst, with equivalent performance to a catalyst with high SiO 2 content, could be prepared from SiO 2 -free TiO 2 by subjecting it to hydrothermal pre-treatment. In addition to removing chloride ions from the catalyst surface, the effect of the pre-treatment was to alter the state of hydration of the TiO 2 in the support, which appears to be a key parameter in determining the intrinsic activity of Pt catalysts for the deep oxidation of alkanes. In light of the work of Caporali et al. [44], which showed that H 2 O-derived oxygen was active in the deep oxidation of hydrocarbons over a Pd diesel oxidation catalyst, we propose that hydration of TiO 2 provides a similar hydroxyl-mediated pathway to CO 2 , in addition to the reaction between adsorbed alkane and dissociated O 2 on Pt 0 active sites. One of the limitations of our study, however, is that it did not allow us to determine the relative extent to which propane oxidation over 1%Pt on bi-phasic TiO 2 is enhanced by (i) removal of the chloride ions, compared to (ii) increasing the degree of hydration, both of which occur during hydrothermal pre-treatment.

Conclusions
In this study, oxidation catalysts comprising 1%Pt supported on TiO 2 and TiO 2 -SiO 2 were prepared from a chloride-free Pt precursor using a non-aqueous impregnation method, so eliminating the metal precursor as a source of any adventitious chloride ions in the catalysts. Characterisation of these catalysts was complicated by the well-known SMSI effect in Pt/TiO 2 [20], the onset of which is known to be enhanced by the presence of SiO 2 in the support material [16]. The resultant suppression of CO-chemisorption gave rise to anomalously large values for the estimated Pt crystallite size with increasing SiO 2 content, whereas electron microscopy showed that, in fact, relatively small nanoparticles (typically 2-5 nm diameter) were being stabilised by the silica.
The NO oxidation activity of the catalysts showed an inverse dependence on the mean oxidation state of the exposed Pt. This is consistent with the accepted model of structure-sensitivity for NO oxidation over Pt [31], in which a higher specific rate of reaction results from a greater number of Pt 0 surface sites (usually associated with larger metal nanoparticles [32]), whereas more ionic nanoparticles (which are often smaller) are less active. As the oxidation of alkanes is known to exhibit the same type of structure-sensitivity [45], in which the Pt oxidation state [46,47] rather than its morphology is the determining factor, a similar inverse correlation would be expected between propane-oxidation activity and the Pt 2+ /Pt 0 ratio. However, we have observed that the correlation breaks down when other factors exert an influence, including the presence of adventitious ions (which can arise from the routes taken during manufacture of the support materials) and the state of hydration of the catalysts.