Au-Decorated Ce–Ti Mixed Oxides for Efficient CO Preferential Photooxidation

We investigated the photocatalytic behavior of gold nanoparticles supported on CeO2–TiO2 nanostructured matrixes in the CO preferential oxidation in H2-rich stream (photo-CO-PROX), by modifying the electronic band structure of ceria through addition of titania and making it more suitable for interacting with free electrons excited in gold nanoparticles through surface plasmon resonance. CeO2 samples with different TiO2 concentrations (0–20 wt %) were prepared through a slow coprecipitation method in alkaline conditions. The synthetic route is surfactant-free and environmentally friendly. Au nanoparticles (<1.0 wt % loading) were deposited on the surface of the CeO2–TiO2 oxides by deposition–precipitation. A benchmarking sample was also considered, prepared by standard fast coprecipitation, to assess how a peculiar morphology can affect the photocatalytic behavior. The samples appeared organized in a hierarchical needle-like structure, with different morphologies depending on the Ti content and preparation method, with homogeneously distributed Au nanoparticles decorating the Ce–Ti mixed oxides. The morphology influences the preferential photooxidation of CO to CO2 in excess of H2 under simulated solar light irradiation at room temperature and atmospheric pressure. The Au/CeO2–TiO2 systems exhibit much higher activity compared to a benchmark sample with a non-organized structure. The most efficient sample exhibited CO conversions of 52.9 and 80.2%, and CO2 selectivities equal to 95.3 and 59.4%, in the dark and under simulated sunlight, respectively. A clear morphology–functionality correlation was found in our systematic analysis, with CO conversion maximized for a TiO2 content equal to 15 wt %. The outcomes of this study are significant advancements toward the development of an effective strategy for exploitation of hydrogen as a viable clean fuel in stationary, automotive, and portable power generators.


■ INTRODUCTION
Over past decades, the environmental issues related to the use of non-renewable energy sources, the massive emission of greenhouse gases and the exponential growth of world energy consumption, have been fostering the research on renewable and environmentally friendly energy sources. Among the many different strategies developed, hydrogen has turned out to be the ideal clean energy carrier, 1 since it can be employed as feedstock in fuel cells, 2 among which are low temperature polymer electrolyte fuel cells (PEMFCs). Besides the industrially wellestablished process for hydrogen generation by catalytic reforming of hydrocarbons, alternative processes have been proposed to foster the decarbonization of the energy sector, including thermochemical, electrolytic, 3,4 and photolytic ones such as alcohol reforming and autothermal reforming, 5 biomass gasification, 6 and photoelectrochemical (PEC) water splitting. 7 Nevertheless, carbon monoxide is generated as a byproduct and PEMFCs need the use of hydrogen containing less than 50 ppm CO to avoid poisoning of the Pt or Pt−Ru electrodes. CO preferential oxidation in hydrogen-rich stream (CO-PROX) is very promising because this simple and cost-effective reaction, usually performed in the 30−250°C temperature range, 8 can be coupled to PEMFCs. In this configuration, hydrogen storage is not necessary, and it can be used for both automotive applications and stationary and portable power generation. 9 The need to harvest and convert solar energy, an inexhaustible and clean source par excellence, has prompted intensive research efforts in this field, legitimizing the photocatalytic approach as one of the best options for sustainable future energy production.
Nowadays many semiconductors are investigated as photocatalysts, and among them, the most widely studied and applied is titanium dioxide, TiO 2 . There are two well-known drawbacks of titania: its wide band gap, ranging from 3.0 eV of rutile to 3.2 eV of anatase, which limits light absorption to the UV region, and the high recombination rate of electron−hole pairs.
A number of ways have been pursued to broaden the absorption range and to improve charge separation, such as metal and nonmetal doping, dye sensitization, and coupling with other semiconductors. A growing interest has recently emerged on photoenergy applications of ceria, CeO 2 , and CeO 2 -based catalysts. 10 The wide use of ceria as catalyst or a noninert support for catalysts has been associated in the literature with its unique ability to shuttle between Ce(III) and Ce(IV) states depending on whether it is present in an oxidizing or reducing atmosphere. Non-stoichiometric CeO 2−x compounds can be formed by reduction of Ce(IV) to Ce(III) with oxygen release and the concomitant formation of oxygen vacancies within the crystal structure, leading to a high oxygen mobility without suffering variations from its lattice, even after considerable loss of oxygen. 11 Thanks to this redox capability, known as oxygen storage capacity (OSC), cerium oxides are considered excellent oxygen buffers and are very active in oxidation reactions. 12 The theoretical band gap of CeO 2 is about 6.0 eV between the states of O 2p and Ce 5d; 13 nevertheless the experimental band gap is only around 3.2 eV, a value approaching that of other two semiconductors largely employed in photocatalysis, TiO 2 and ZnO. 14 Most probably this is due to the O 2p → Ce 4f transition, even though the origin of this reduced band gap is still controversial. 15 Introduction of metal ions into the CeO 2 lattices can increase the density of oxygen vacancies and potentially improve the catalytic performance of CeO 2 -supported metal catalysts, through changing the physical and chemical properties of the supports. 16,17 In particular, the preparation of CeO 2 -based heterostructures have been investigated to reduce the band gap of pure CeO 2 , by improving charge separation and/or hindering the recombination due to the novel structures and interfacial charge transfer, resulting in enhanced photocatalytic efficiencies. 18,19 The formation of CeO 2 /TiO 2 composite photocatalysts has been reported, which showed catalytic activity superior to that of single CeO 2 and TiO 2 . 20,21 Similarly, it has been proven that the morphology of CeO 2 -based nanocomposites has a strong effect on the photoactivity of a catalyst. 22 Highly active ceria-based catalysts can be designed by tuning the synthesis process to produce the desired morphologies and microstructures with controlled oxygen vacancies for the targeted reaction. Among others, a synthetic methodology to obtain flower-like ceria-based materials is available. 23−25 During the nucleation phase, under specific reaction conditions and in the absence of glycols or other soft surfactants, rare earths among them ceria can create a hierarchical structure composed of many microsheets, as petals forming a sort of flower.
Introducing metal nanocrystals on the semiconductors can offer unique capabilities to control the light−matter interaction of the support materials by harnessing the light absorption and decay properties of surface plasmon resonance (SPR). 26 As far as catalytic applications are concerned, the use of ceria requires in most cases combination with nanosized metal particles. Since the discovery of the exceptional catalytic properties of supported gold nanoparticles (Au NPs), Au nanoparticle catalysts have been the subject of numerous investigations for total and preferential CO oxidation. 27,28 Gold nanoparticles are the most prominent and investigated class in photocatalysis, since they present efficient absorption in the visible light region through localized surface plasmon resonance (LSPR) 29 and, supported on a semiconductor like titania, give rise to a Schottky barrier. 30 Collective oscillations of hot electrons close to the Au−support interface can occur upon visible light excitation. Hot electrons can overcome the Schottky barrier and be subsequently transferred to the conduction band of the adjacent semiconductor. 31 Furthermore, the presence of Au NPs can extend the lifetime of charge carriers that reach the surface of the semiconductor by enhancing the rates of electron−hole separation at their interface. Formation of highly energetic carriers is the most common mechanism exploited in SPRmediated catalysis, although local heat generation and near-field enhancement are also prominent mechanisms in photocatalysis. 32 The present work (see the scheme in Figure 1a) aims to investigate the photocatalytic behavior of Au NPs supported on CeO 2 −TiO 2 (0−20 wt % TiO 2 loading) nanostructured matrixes in the CO preferential oxidation in H 2 -rich stream (photo-CO-PROX), at room temperature and atmospheric pressure, under simulated solar light irradiation. The target application is H 2 purification in PEMFC technology.
The main concept behind this work is to modify the electronic band structure of ceria through the addition of titania, making it more suitable for interacting with free electrons excited in gold nanoparticles through SPR. The optical characterization demonstrates modulation of the optical band gap in the composite systems, accounting for the effectiveness of the proposed strategy.
Finally, we also assess how a controlled morphology can affect a multicomponent system, improving its photocatalytic performance.

■ EXPERIMENTAL SECTION
All details of the experimental activities are reported in the Supporting Information.

■ RESULTS AND DISCUSSION
Thermogravimetric (TG/DTG) analyses were performed on the CTx samples (C = ceria; T = titania; x = 0, 10, 12, 15, 20 wt % TiO 2 ) to determine the optimal calcination conditions required for the stepwise removal of organic residues and byproducts. The curves of C100, T100, and CT15 samples, taken as examples, are reported in Figure 1b. The thermogravimetric curves of the samples C100, CT10, CT12, CT15, and CT20, containing a titania loading between 0 and 20 wt %, exhibit a common trend. A first weight loss occurs, about 5−10 wt %, up to 200°C, attributed to the decomposition and degradation of organic moieties derived from the synthesis (mainly titania precursor, titanium(IV) isopropoxide (TTIP)) and the desorption of physisorbed water.
Then, a second weight loss is recorded, about 25−30 wt %, in the range 280−350°C, ascribed to the thermal decomposition of cerium hydrated carbonates and hydroxycarbonates formed during the synthesis by the reaction between cerium nitrate and potassium carbonate. 36 According to these data, the thermal treatment chosen for all the prepared samples was set at 450°C.
Phase identification was performed by X-ray powder diffraction in the 2θ range 20−80° (Figure 1c). Pure ceria sample, C100, exhibits peaks emerging at 28.56, 33.08, 47.75, and 56.33°attributed to the cubic fluorite structure of cerium dioxide (database PDF 34-394); pure titania sample, T100, displays the characteristic peaks of anatase phase at 25.46, 37.92, 48.11, and 54.68°2θ, in agreement with the values in the standard card (database PDF 73-1764), as expected by considering the calcination temperature used. Mixed oxide samples CT10, CT12, CT15, CT15FP (FP, fast precipitation), and CT20 exhibit the typical XRD profile of cerianite, with characteristic peaks indexed to the cubic fluorite structure. CeO 2 crystallite sizes, estimated from the full width at half-maximum of the 2θ = 28.5°diffraction peak of the CeO 2 (111) face using the Scherrer equation, are summarized in Table 1. Interestingly, all the Ce−Ti mixed oxides do not exhibit peaks related to the presence of titania polymorphs, even in the presence of a high TiO 2 loading (up to 20 wt %), and the crystalline structure of the composite materials is inferred as a CeO 2 cubic fluorite structure. Since the ionic radius of Ti 4+ (0.68 Å) is much smaller than that of Ce 4+ (1.01 Å) and the content of cerium is much higher than that of titanium, it can be suggested that Ti 4+ ions could incorporate into the lattice of CeO 2 to replace Ce 4+ ions, leading to the formation of Ce−Ti solid solution, 37 This is corroborated by the decrease in CeO 2 crystallite size by increasing the TiO 2 loading, passing from 8.3 nm for C100 sample to 6.7 nm for CT20 sample, and by a slight shift toward higher angles of the main reflection peaks of cerium dioxide, suggesting a slight distortion of the ceria lattice due to the incorporation of titanium(IV) ions.
XRD was also carried out after deposition−precipitation of gold nanoparticles on the final AuCTx catalysts (not shown): no modification of the titania−ceria phases was detected, nor were gold-containing phases discerned in the XRD patterns, most probably due to the low Au concentration (0.54−0.82 wt %), below the XRD detection limit.
The porous nature of the synthesized materials, prepared by a controlled and very slow addition of an alkaline solution to a Ti and Ce precursor solution, followed by a long crystallization step, was evaluated by N 2 physisorption at −196°C, and the textural properties are reported in Table 1. The isotherms of the CTx series belong to the type IV profile, according to the IUPAC classification, 38 with a small hysteresis loop suggesting that the very slow coprecipitation led to the formation of mesoporous solids, with a mesoporosity also originated by interparticle voids. Pore size distribution is narrow and centered between 3.5 and 7 nm, which is very interesting, considering that the synthesis was carried out in the absence of any kind of surfactants. In Figure 1d, isotherms and Barrett−Joyner− Halenda (BJH) pore size distributions of samples CT15 and CT15FP, taken as an example, are reported.
Pure ceria, C100, displays a specific surface area of 42 m 2 g −1 and pure titania, T100, has a specific surface area of 52 m 2 g −1 .
The addition of titania to the ceria matrix increases the support surface area by more than double, which is considered to possibly contribute to a higher catalytic activity and can be due to incorporation of Ti into the ceria lattice, with the formation of a solid solution. 39 The change in the Brunauer−Emmett−Teller (BET) surface area due to the addition of Au NPs can be considered as negligible for all the samples (Table 1) and is most probably due to a very slight pore blocking of the support by gold nanoparticles.
The surface morphology of the samples, before and after calcination, was investigated by SEM microscopy. No morphological changes were detected after calcination. The materials are characterized by the presence of extended and thin 2D sheets (Figure 2).
The addition of titania drastically reduces the lateral dimensions of the platelets, compared to the C100 sample (see the measured lateral size from SEM in Figure 2). A general view revealed that all the calcined samples of CTx series are made up by microneedles.
For comparison purposes, the sample CT15FP was synthesized by a traditional coprecipitation method, with a faster basification and a consequent lower nucleation time. SEM analysis of this sample revealed a more compact morphology, compared to the CTx series, even if the most peculiar differences were highlighted only through HR-TEM and were not visible through SEM analysis. As expected, pure titania sample, T100, prepared by use of the same synthetic procedure used for the ceria-containing samples, did not exhibit any organized morphology ( Figure S1).
EDS elemental analysis of the CTx series confirmed the presence of titanium in the supports, in a concentration  approaching the nominal amount (even though the presence of TiO 2 did not modify the XRD pattern). For the sake of an example, the EDS analysis of CT15 and CT15 FP samples has been reported ( Figure S2): compared to the nominal Ti/Ce ratio (w/w) (0.13, corresponding to 15 wt % TiO 2 ), the composition measured by EDS is slightly higher (0.14) for CT15 support and lower (0.10) for CT15FP sample, most probably due to the different synthetic methodologies employed.
TEM characterization was carried out on some representative samples to have an insight into the nanostructure composition of the Ce−Ti composite oxide and the nanoscale arrangement of Au nanoparticles. Figure 3 displays the results of elemental maps extracted from STEM−EDX analysis of AuCT15 and AuCT20, in addition to AuC100 and AuT100 as reference samples where Au nanoparticles are supported on either pure Ce-or Ti-based oxide. For all samples, Au-related signal is distributed on the whole supporting material, with no evidence of aggregation or segregation into large-scale nanoparticles.
No difference is observed among the samples in terms of Au distribution, pointing out that Au NPs decoration is independent of the metal oxide composition and shape.
As expected, no specific heterostructure is observed in the mixed oxide samples, where Ti content is distributed quite randomly on the Ce oxide supporting material, even though some small titanium-rich clusters are also present. Figures 4 and 5 report the HR-TEM analysis of AuCT15 and AuCTFP15 samples, chosen as representative samples due to the large surface area and the optimal photocatalytic properties (AuCT15) and as a benchmarking standard (AuCT15FP), as discussed later. AuCT15 sample is characterized by a hierarchical structure composed by compact needlelike microstructures, as previously observed by SEM, fully decorated by 2− 10 nm large nanoparticles with similar densities and crystal structures. In contrast, samples with lower and higher Ti content do not display this complex heterostructure ( Figure S3).
Both the needlelike structure and the nanoparticles covering them display a highly crystalline lattice, whose pattern is compatible with the cubic fluorite CeO 2 structure, also known as cerianite, consistent with XRD analysis results.
SAED analysis on the composite highlights the polycrystalline nature of the material, as well as the perfect matching to the cerianite phase ( Figure S4). No evidence of AuNPs is directly observed on the HR-TEM micrograph due to the superposition of the crystalline nanoparticles and the large thickness of the nanoneedles.
However, STEM-HAADF micrographs (Figure 4c) highlight the presence of denser, high-atomic-weight nanoparticles, whose contrast is compatible with pure Au. Both large (>15 nm) Au NPs and smaller (<10 nm) Au NPs can be noticed, highlighted with green and red arrows, respectively. The composition of these nanoparticles is further confirmed by EDX mapping on the same area. This technique also displays the presence of both Ce and Ti on the large needlelike microcrystals, as well as the decorating nanoparticles. By comparing the relative contents of Ti and Ce on the two different nanostructures, we can observe how the ratio Ti/Ce shifts from 0.14 in the needles, fully fitting with the expected content, to 0.41 in the decorating nanoparticles. Thus, we infer that the Ti-rich CeO 2 nanoparticles are resulting from the segregation of excess Ti in the reaction environment, whose slightly different atomic radius, with respect to Ce, might induce strain in the growing crystallite and limit the size to a few nanometers.
For the sake of comparison, AuCT15FP sample ( Figure 5) was also analyzed. The fast-precipitated ceria-based microstructures are strongly different from the slow-precipitated sample, with a layered structure forming large platelets, as previously observed by SEM analysis. No evidence of the randomly dispersed Ce/Ti-based nanoparticles is observed. The ceria platelets are clean and display a porous surface in some areas. The HR-TEM micrograph displays the same pattern that was observed in the slow-precipitation process sample, compatible with cerianite phase, as displayed by the fast Fourier transform (FFT) in the insets ( Figure 5). The surface of the sample was clearly irregular and porous, but the size of the crystal was this time exceeding the typical size of the highmagnification image, suggesting a larger domain size. The Au nanoparticles can be clearly recognized on the surface of the layered microcrystal in both HR-TEM and STEM micrographs, but no evidence of small (<10 nm) Au NPs is observed, as confirmed by EDS mapping. Due to the lower thickness of the supporting CeO 2 material, the typical diffraction fringes of cubic Au (2.3 Å) can be observed on the FFT of the HR-TEM micrograph.
Au-based samples were analyzed by X-ray photoelectron spectroscopy (XPS), to obtain information on the surface composition and the oxidation states of gold, titanium, and cerium on the surface. Figure 6 compiles the corresponding core level signals, and Table 2 lists the surface atomic composition and the Au 4f binding energy (BE) values of Au species.
The titanium signal ( Figure 6) presented a doublet, whose Ti 2p 3/2 component is centered at 458.5 eV in AuT100 sample. The BE value is typical of Ti 4+ species in TiO 2 structure. 40 The signal is slightly shifted to higher binding energy values when the titania content is higher.
On the other hand, Ce 3d core level spectra showed, in all cases, four spin−orbit doublets denoted as ν(n) and u(n): three  ascribed to the presence of Ce 4+ (ν (∼882.2 eV) and u (∼900.7 eV); ν″ (∼888.8 eV) and u″ (∼907.2 eV); ν‴ (∼898.2 eV), u‴ (∼916.5 eV)) and one due to Ce 3+ species (ν′ (∼884.5) and u′ (∼903.3 eV)). 41 It is clearly observed that Ce 4+ is the main oxidation state of cerium in these samples. The presence of Ce 3+ species is only evidenced by the contribution at 884.5 eV (ν′),   being a shoulder of the signal at 882.2 eV (ν). The decomposition of Ce 3d core level spectra ( Figure 6) showed as described above eight peaks corresponding to four spin−orbit components (the spin−orbit splitting was about 18.6 eV). The ratio Ce 3d 5/2 /Ce 3d 3/2 was fixed at 1.5 considering the study of Beĉhe et al. 42 It was possible to quantify the surface Ce 4+ /Ce 3+ ratio present for each sample ( Table 2). It is evidenced that titania incorporation improves ceria reducibility by following the order AuCT15 ≈ AuCT15FP > AuCT20 > AuCT12 > AuC100. It is interesting to notice that the preparation method does not alter ceria reducibility, presenting AuCT15 and AuCT15FP samples with similar Ce 4+ /Ce 3+ values. Moreover, if the signal at ca. 882 eV is considered, a slight shift to lower BE is noticed by incorporating Ti into the ceria network. The observed shift in Ce and Ti signals suggests an interaction between both components. Sheerin et al. 43 prepared mixed Ce−Ti oxides containing different Ce and Ti loadings, and a similar shift in Ti signal was observed. XRD and textural properties also pointed in this direction, to the formation of a Ti−Ce solid solution.
Considering the O 1s signal, Ce-containing samples presented two contributions: one at 529.1−529.4 eV, associated with lattice oxygen in ceria (O lat ), and another one at 531.3−531.7 eV, due to the presence of surface oxygen (O surf ), 44 due to the presence of defect sites within the oxide crystal, adsorbed oxygen, or hydroxide species on the surface of the samples. If the spectra are compared, by increasing the titania loading, the main signal hardly changes, being closer to the BE related to CeO 2 (529.1 eV) than to that of TiO 2 (529.9 eV). Instead, AuT100 compound possesses the main signal located at 529.9 eV, typical of TiO 2 species. 45 Considering the surface oxygen signal, it is reported that the area and intensity of this signal provide an indication of oxygen vacancies in the host lattice. 46 The spectra were decomposed (not shown), and the quantification of this contribution has been included in Table 2 as the O surf /O lat ratio. As observed, AuCTx samples present a greater amount of surface oxygen species than pure ceria and titania ones. Nonetheless, the preparation method does not influence this ratio if data related to AuCT15 and AuCT15FP samples are compared.
Finally, the Au 4f signal presented the most intense Au 4f 7/2 and Au 4f 5/2 binding energy peaks at ∼83.4−87.3 eV, corresponding to Au 0 . 47 A less intense contribution located at ∼85.0 and 88.6 eV was also observed and assigned to partially reduced surface gold species, Au δ+ . 48,49 On the basis of this, it can be stated that gold exists mainly in a metallic state, but there is a proportion of Au δ+ on the surface. The support composition does not alter the Au 4f BE value but does alter the percentage of each species, as observed in Table 2.
The surface atomic composition indicated that, although there is not a clear trend between the Au/(Ce + Ti) ratio and the support composition, AuCT20 sample presented the highest Au/(Ce + Ti) ratio, therefore indicating a greater Au surface exposure. The AuCT15 sample, instead, presents the highest amount of Au δ+ species on the surface. The literature reported that the presence of these partially reduced gold species is highly beneficial for the catalytic activity in the CO oxidation reaction. Escamilla-Perea et al. 49 studied the CO oxidation on SBA-15supported gold nanoparticles decorated by CeO 2 . These authors explained the CO oxidation by considering the collaborative effects of the high surface exposure of Au δ+ and Ce 3+ species, derived from XPS and the IR band assigned to CO on Au δ+ . A similar dependence was reported previously by Minicòet al. 50 Moreover, we 51 also observed that the presence of Ce 3+ on the catalyst surface improved the activity of zirconia-supported CuO−CeO 2 catalysts in the carbon monoxide oxidation reaction. Therefore, the presence of Au δ+ species and Ce 3+ ions on the surface seems to improve the catalytic activity of these systems.
The photoresponsive behavior of the samples was explored by DRUV−vis spectroscopy to determine the energy gap value of the supports (Figure 7 and Table 1), estimated from the linear extrapolation of (αhν) 1/2 versus photon energy.
Titania sample, T100, shows an E g value of 3.10 eV, typical of pure titania, slightly lower than that of anatase polymorph. Pure ceria, C100, displays a much lower band gap value (2.79 eV)  with respect to that usually reported in the literature for bulk CeO 2 , with a red shift attributed to the presence of Ce 3+ at the grain boundaries, which forms some localized gap states in the band gap. 52 As for ceria−titania mixed oxides, it has been reported 53 that the migration of Ti 4+ ions into the ceria lattice at the oxide− oxide interface of CeO 2 −TiO 2 favors the formation of Ce 3+ sites, which can be very useful in catalytic processes. For example, the presence of Ce 3+ ions in M/CeO x /TiO 2 (M = Au, Cu, Pt) systems seems to favor the water dissociation, making these systems excellent catalysts for the H 2 production through the water gas shift reaction. 54,55 Moreover, in ceria−titania mixed oxides, the Ce 3+ species seem to introduce additional states in the band gap, corresponding to the partially occupied 4f levels, and thereby reduce the band gap up to 2.2 eV. 56 All the prepared CTx systems show energy gap values lower than those of pure oxides, ranging between 2.34 and 2.57 eV (Figure 7b), with a red shift that could be ascribed to the presence of additional localized energy levels (oxygen vacancies) within the band gap, absorbing relatively low energy photons.
In Figure 7c the Tauc plots of the sample with 15 wt % TiO 2 loading, before and after Au sensitization, are reported. The observed LSPR broad band centered at ca. 2.20 eV (absorption at about 560 nm) can be attributed to the presence of Au NPs deposited on the supports by deposition−precipitation method. The low intensity in signal could be ascribed to the very low amount of gold (0.54−0.82 wt %), as confirmed by ICP-OES analysis ( Table 1).
The AuCTx catalysts were finally tested in the CO preferential oxidation in excess of hydrogen at room temper-ature and atmospheric pressure under dark and simulated solar light irradiation, and their performances are shown in Figure 8. TOF values (mmol·g cat −1 ·h −1 ) of CO oxidation are reported in Table 1.
The photocatalytic behavior of the bare supports, both pure and mixed oxides, was investigated, and the response was found negligible (few percentage units, not shown) both in dark mode and under solar light irradiation. Gold-based samples, on the contrary, showed a very interesting catalytic behavior, both in dark conditions and under sunlight irradiation. On the other hand, since the discovery of the exceptional catalytic properties of nanosized supported gold particles, it is well-known that Au catalysts are active in the total and preferential CO oxidation reactions. 27, 28 As shown in Figure 8, the CO conversion versus TiO 2 content plot of AuCTx samples shows a trend depending on the increasing amount of TiO 2 , which seems to have a significant effect on the catalytic activity, improving the conversion passing from 0 wt % TiO 2 loading (AuC100) up to 15 wt % TiO 2 loading (AuCT15). Conversely, an excessive amount of titania (AuCT20 sample) not only does not further improve the catalytic activity but also diminishes the CO conversion.
AuCT15 is, by far, the most active sample, with a CO conversion of 52.9% in dark mode and 80.2% under sunlight, respectively, as also highlighted by the TOF values of CO oxidation (Table 1). This catalyst also exhibits a much higher activity in comparison with both pure oxide systems, AuC100 and AuT100. In detail, the CO conversion values in dark mode are 12.1, 29.1, and 52.9% for AuC100, AuT100, and AuCT15, respectively. Under simulated solar light irradiation, the CO  conversion values are 40.5, 51.3, and 80.2% for AuC100, AuT100, and AuCT15, respectively. Cerium oxide has been proven as an excellent catalyst support to be used in in the CO-PROX reaction, and mechanistic studies have shown that the lattice oxygen plays a crucial role, since the reaction follows a Mars−van Krevelen mechanism, 57,58 according to which CO is chemisorbed on Au NPs 59 and oxygen is supplied by the support, generating reduced Ce 3+ species that are oxidized back with oxygen from the feed.
The incorporation of a second metal into the CeO 2 lattice, as TiO 2 , seems to enhance ceria reducibility, participating in the Mars−van Krevelen mechanism and eventually improving the catalytic performance both in the dark and under sunlight irradiation. This could explain the better performance of AuCT15 sample with respect to AuC100 and AuT100. Moreover, as emerged from XPS analyses (Table 2), the incorporation of titania into the CeO 2 lattice improved the reducibility of ceria by increasing the amount of Ce 3+ species and altered the Au oxidation state on the surface favoring the presence of partially reduced gold. The most active sample, AuCT15, displays not only the highest percentage of Ce 3+ , but also by far the largest amount of partially reduced Au species on the surface, indicating that those species might be directly involved in the reaction mechanism. Further addition of TiO 2 up to 20 wt % in CeO 2 (AuCT20) gave rise to a decrease in the textural properties (i.e., specific surface area), in Ce 3+ and Au δ+ amounts on the surface, and eventually in the CO conversion, in both dark and simulated solar light modes.
Along with the CO conversion efficiency, selectivity toward CO 2 was also investigated, and the results are reported in Figure  8b. Selectivity values in the CO-PROX are the result of the simultaneous CO and H 2 oxidation reactions, both occurring in the presence of molecular oxygen catalyzed by metal oxides. The desired reaction is clearly the CO oxidation, whereas H 2 oxidation is highly undesirable due to H 2 consumption (needed to feed a PEMFC). Nevertheless, as 100% of selectivity is usually not attainable, H 2 oxidation usually takes place and H 2 O forms, reducing the activity and selectivity of the catalyst. It should be noted that, considering the reaction stoichiometry, the oxygen excess factor used (λ), 34,35 and the CO conversion values, the O 2 consumption is attributable to both oxidation reactions. Nevertheless, under both modes all the catalysts exhibit higher selectivities for CO oxidation compared to the undesired H 2 oxidation, as is typical in Au-based systems. 60,61 However, the observed CO 2 selectivity values are lower under simulated light irradiation. This is probably due to the simultaneous increase of the hydrogen oxidation due to the light irradiation, even though the factors governing photocatalytic activity in the CO-PROX reaction are not yet fully understood.
A durability test was performed on AuCT15 sample, and the CO conversion trend is reported in Figure S5. The reaction was monitored for 24 h, using GHSV = 22 000 h −1 ; λ = 2; and 1.2% CO, 1.2% O 2 , 50% H 2 , and He balance (vol %). After switching on of the simulated solar light, the catalyst showed a diminution of CO conversion from 82 to 78% within the first 6 h, remaining nearly stable (74.4%) for the further operation time and suggesting a very good physicochemical stability of this multicomponent system.
Considering the reaction carried out in dark mode operating conditions, the "traditional" CO-PROX, and comparing the results obtained for the AuCT15 sample with the literature data, a maximum of 85% CO conversion at 100°C with a selectivity to CO 2 of about 50% was obtained by Rossignol et al. 62 Table 3. Considering that most of the reference catalysts show a much higher Au loading and/or have been tested under visible light irradiation 65,66 (a solar light source is by far less energetic), it can be concluded that the sample AuCT15 displays a very good catalytic performance also in the photo-PROX reaction if compared to the literature data.
A comparison was finally made between the samples containing about 15 wt % titania and prepared by two different coprecipitation methods: a slow basification for the AuCT15 sample and a much faster coprecipitation for the AuCT15FP system, leading to very different morphological and catalytic properties. In fact, as evidenced by SEM and HR-TEM analysis, AuCT15 showed a hierarchical structure formed by compact needle-like microstructures decorated by Ti-rich nanoparticles. On the contrary, AuCT15FP did not show any self-organized structure, presenting a layered structure formed by large platelets, with no evidence of the randomly dispersed Ce/Tibased nanoparticles.
By comparing the CO conversion values of these two catalysts (Figure 8a), the one prepared by a slow coprecipitation method displays a much higher catalytic activity both in dark conditions and under simulated sunlight irradiation. The CO conversion value is 52.9% for the sample AuCT15 and less than half, 24.1%, for AuCT15FP, with a gap of 28.8%. Under solar light irradiation, the CO conversion values are 80.2% for AuCT15 and 56.1% for AuCT15FP, with a difference of 24.1%.
The highest catalytic performance exhibited by the sample AuCT15 could be attributed to several factors, including the high specific surface area and total pore volume, the higher concentration of Ce 3+ species, and the lower size of Au NPs.
Finally, the synergistic effect of gold NPs deposited on the highly porous ceria−titania network has to be taken into

■ CONCLUSIONS
In summary, we investigated the catalytic properties toward the photo-CO-PROX reaction of CeO 2 −TiO 2 mixed oxides obtained through a slow and surfactant-free coprecipitation method and decorated by Au nanoparticles deposited by deposition−precipitation. The variation of the TiO 2 content in the mixed oxide results in significant textural and morphological changes of the supports, which display a specific surface area ranging between 81 and 132 m 2 g −1 and are mainly composed of micrometer sized thin platelets. The dimension of the platelets decreases with the increase of TiO 2 content. For a specific composition (15 wt %) a hierarchical structure is formed, in which the platelets are conformal decorated by small, Ti-rich oxide nanoparticles. Au nanoparticles are finely and homogeneously dispersed at the surface of the oxides and present a quite broad size distribution. It is evidenced that the reducibility of ceria and gold oxidation state is altered by the incorporation of Ti into the ceria lattice. In this regard, the amount of Ce 3+ and partially reduced gold Au δ+ species is improved in the CTx samples in comparison with pure oxides, with the AuCT15 sample displaying, not only the highest percentage of Ce 3+ , but also the largest amount of partially reduced Au species on the surface and reported as catalytically active in the CO oxidation reaction.
The mixed oxides present an optical energy gap far below the theoretical one for pure ceria and titania, enabling direct exploitation of solar light for photocatalysis. The presence of Au NPs further broadens the absorbance to lower energy photons, close to the yellow region of the solar spectrum, thanks to LSPR.
We tested the compounds toward the preferential photooxidation of CO to CO 2 in excess of H 2 under simulated solar light irradiation at room temperature and atmospheric pressure, resulting much more active than a sample with a nonorganized structure prepared, as a benchmark, by a fast coprecipitation method. The most efficient sample exhibits CO conversions of 52.9 and 80.2% and CO 2 selectivities equal to 95.3 and 59.4% in dark and simulated solar light modes, respectively. We found a clear morphology−functionality correlation in our systematic analysis, with CO conversion maximized for a TiO 2 content equal to 15 wt %.
These results represent a promising step toward the preparation of an effective system for CO preferential oxidation in H 2 -rich stream. The synthetic route is environmentally friendly and low cost, providing a potential solution for the open issue of H 2 stream purification in PEMFCs.