Oxygen vacancies dependent Au nanoparticle deposition and CO oxidation

Abstract

Oxygen vacancies are critical for both the reactivity and the reaction mechanisms for Au-deposited oxides. The deposition behavior of Au on a CeO₂ support was studied by determining the content of oxygen vacancies induced by ascorbic acid (VC) treatment. Oxygen vacancies were introduced on CeO₂ nanorods (NRs) by VC reduction, and Au nanoparticles (NPs) were loaded by a deposition–precipitation (DP) method. The formation of oxygen vacancies and their effects on Au NP deposition were evaluated by catalytic CO oxidation. The reaction mechanism of Au anchoring was closely linked to the concentration of oxygen vacancies. On the CeO₂ nanorods with an optimal number of oxygen vacancies, the deposition of Au behaved more like a “lattice substitution mechanism”, where charge transfer occurs to form positively charged Au³⁺ species and reduced Ce³⁺, associated with the creation of oxygen vacancies, which are active for CO oxidation. VC treatment induced a large number of oxygen vacancies on CeO₂ NRs, resulting in highly increased reducibility of CeO₂ and strong interaction between Au and CeO₂. Consequently, Au³⁺ cations were reduced directly with a fast reduction rate, instead of undergoing hydrolysis into the hydroxychloro gold(III) complex, [Au(OH)_xCl_4−x]⁻, which was generally generated during the DP procedure. Such strong charge transfer interaction between the oxygen vacancies and Au³⁺ leads to sintering of the reduced Au species to form Au NPs with a larger size and an uneven distribution, and to a decreased Ce³⁺/Ce⁴⁺ ratio, with a decrement of surface oxygen atoms, as well as the reduction of Au³⁺ species to Au⁺; these events are together connected to the activity loss of the catalyst.

---

Introduction

Ceria, a key component in the formulation of catalysts and catalyst supports, has received extensive investigation with various catalytic reactions. The high catalytic activity of CeO₂ originates from the rapid reduction/oxidation shift between Ce³⁺ and Ce⁴⁺, accompanied by the formation and depletion of oxygen vacancies on the ceria surface or in the bulk, providing a high oxygen storage capacity (OSC), which is critical in determining the O₂ diffusion, the rate-controlling step in the CO oxidation reaction.^1,2 The oxygen uptake and release relies on the rates of the redox cycles, where the oxidation of cerium is fast, while reduction is sluggish. From this viewpoint, any processing of CeO₂ that favours the formation of oxygen vacancies will result in an enhanced oxygen diffusion rate, giving rise to an improvement in the catalytic activity of CeO₂. Qu et al.³ used ascorbic acid (VC) and hydrogen peroxide to chemically etch the surface of ceria nanorods, which successfully increased the specific surface area, oxygen vacancies, and surface Ce³⁺ fractions, which steadily increased the CO oxidation reaction activity.

Additionally, CeO₂ has been recognized as a good reducible support for Au nanoparticles (NPs) to perform as a composite catalyst.⁴ The key function of the support is to prevent abnormal growth of active Au particles, and the interaction between gold and the support is the determining factor for catalytic activity. A synergistic effect occurs between the gold particles and CeO₂ at the Au–CeO₂ interface, where redox processes are involved across oxygen vacancies, allowing lattice oxygen atoms from the support to become activated species available for the oxidation reactions.^5,6 Theoretical studies have shown that the highly electronegative Au NPs can bind strongly to the surface of CeO₂, which is enriched with oxygen vacancies. Oxygen vacancies may be the nucleation centres for the Au NPs, allowing the growth of small and well-dispersed Au clusters, which improve the reactivity of the catalyst.^7,8 Therefore, the presence of oxygen vacancies affects both the reactivity of the support and the gold dispersion on the material, modifying the electronic properties of the gold particles. When Au species bind on a highly reduced CeO₂ surface, electron transfer occurs between the Ce³⁺, the oxygen vacancies, and the Au ions, leading to variations in the Au charge. Identifying the chemical states of the active form of gold (i.e., metallic versus ionic) has been a subject of vigorous debate in the literature. The nature of the gold–oxygen vacancy interactions on CeO₂ surfaces has been studied using density functional theory calculations (DFT). If Au was absorbed on a surface with Ce³⁺ and subsurface oxygen atoms, Au will be negatively charged due to the transfer of the localized electron on Ce³⁺ to the Au 6s orbital,^9,10 while the resulting Au^δ− is inert towards CO adsorption.^11,12 Using oxygen vacancy sites as anchoring sites for Au clusters, the Au in contact with surface O atoms is positively charged.^13,14 Fabris et al.¹⁵ have suggested that the CO oxidation activity was associated with the chemical states of Au atoms at the interface. Specifically, Au³⁺ ions dispersed into the ceria lattice as substitutional point defects could sustain a full catalytic cycle. Instead, the supported Au⁺ atoms were readily attracted by oxygen vacancies to be turned into negatively charged Au^δ− species that deactivate the catalyst. Quite recently, Sheu¹⁶ revealed that, without oxygen vacancies, the Au atom adsorbed on a Ce vacancy of CeO₂(111) was highly positively charged. Increasing the number of oxygen vacancies led to a decreased charge on Au from +3 to −1. Furthermore, with an identical number of oxygen vacancies, the Au charge also relies heavily on the locations of the oxygen vacancies, suggesting a potential to adjust the oxidation states of Au atoms absorbed on ceria.

Experimentally, it is generally accepted that the structure and size of supported gold are greatly affected by the nature and concentration of the oxygen vacancies originating from diverse preparation methods and reaction conditions. Nonetheless, relatively little is known about the properties of the deposited Au and Au–CeO₂ interaction resulting from mediating the oxygen vacancies on the support. By modification of the redox properties of the support, the electronic interactions between the metal NPs and the support can be regulated in parallel. Weststrate et al.¹⁷ initially studied the influence of oxygen vacancies on the properties of ceria-supported gold. They deposited Au by evaporation of Au metal on the oxidized and reduced CeO_x substrates, and found that Au particles on the reduced CeO_x with a higher concentration of Ce³⁺ were somewhat larger than on the oxidized substrate. Although they tried to determine the charge state of Au in the same study, no explicit conclusion was obtained. They only demonstrated that small Au particles adsorbed on oxygen vacancies exhibit a significantly higher Au 4f binding energy than Au particles on oxidized CeO₂. With cation doping and substitution, oxygen vacancies can be populated in CeO₂,^18–21 and afterwards can be employed for the preparation of gold catalysts requiring small-sized gold NPs, and to mediate the properties of deposited Au.^21–24 By studying the influence of different dopants on the catalytic performance of nano-Au/CeO₂, Reddy et al.²⁵ found that incorporation of Zr⁴⁺ into the Au/CeO₂ resulted in high CO oxidation activity, attributed to the presence of more Ce³⁺ ions and oxygen vacancies, which had a beneficial effect on the gold dispersion, leading to smaller gold particles. Furthermore, he discovered that the presence of small Au particles is not the only prerequisite for achieving high CO conversion; the oxidation state of gold also plays a crucial role in the CO oxidation, and the metallic Au⁰ was supposed to be the exclusive active species for CO oxidation at quite low temperatures. With the Eu³⁺-doped CeO₂ support, Hernández et al.²⁶ demonstrated that a higher concentration of oxygen vacancies, induced by Eu-doping, interacts directly with gold during deposition–precipitation (DP), leading to a smaller gold particle size with higher gold dispersion, which can promote the reducibility of the catalyst. However, the overall activity of the CO oxidation with Au/CeEu(10) dropped compared to that with Au/CeO₂. Unfortunately, they did not disclose the specific nature of such interactions from the point of view of charge transfer between Au and the support.

Meanwhile, it should be noticed that oxygen vacancies are very important in determining the growth mechanism of the Au NPs. For the DP procedure, which is one of the most commonly employed methods for preparing active supported Au catalysts with aqueous HAuCl₄ as the gold precursor, it is generally recognized that the supported Au NPs are obtained from the insoluble Au(OH)₃ intermediate, which is hydrolysed from the adsorbed Au ions, usually AuCl₃, and precipitation of Au(OH)₃ takes place exclusively on defects that act as nucleating sites.²⁷ Therefore, the nature of the solid surfaces is crucial in governing the formation of the Au NPs, and a surface with a high density of oxygen vacancies will lead to stabilized Au NPs with a small size and narrow dispersion. Nevertheless, due to the complexity of the catalyst preparation process, the dependence of the initial nucleation and electronic structure of the supported Au nanoparticles on the structure of the support, specifically the oxygen vacancies, still remains unexplored. When the support with oxygen vacancies has a certain degree of reducibility, direct reduction of Au anions becomes possible. Questions arise: during the DP procedure, which reaction pathway, direct reduction vs. anion hydrolysis, is preferential? How does the reaction pathway affect the morphology, size, and structure of the Au NPs? The influence of the oxygen vacancies on the reaction path and the Au NP formation mechanism has been paid less attention, but is crucial in determining the metal–oxide interactions and the catalytic activity. Aiming at this uncertainty, and to gain a further insight into the electronic interactions occurring between the metal NPs and the support, in this work, we attempt to partially reduce CeO₂ and study the effect of oxygen vacancies on the deposition of Au NPs in the DP procedure. Herein, ascorbic acid (VC) is employed, since it is a well-known reducing reagent for Ce(IV). Previously, Qu et al.³ found that redox cycles between Ce³⁺ and Ce⁴⁺ could be facilely implemented using H₂O₂ as oxidant and VC as reductant, which could be identified from the colour change and photoluminescence switch accordingly. Importantly, the nanorod morphology of the sample was well maintained during the VC reduction, which avoided the influence of the morphological and structural factors but exhibited a direct relationship between the electronic structure (valence state) and the physical properties (photoluminescence). Accordingly, it is expected that oxygen vacancies would be introduced, along with the partial reduction of Ce(IV) in CeO₂ by VC treatment, to afford a platform for investigating the deposition behaviour of the supported Au NPs. In the present work, the deposition behaviour of the Au NPs is compared with the VC-treated CeO₂ NRs and the pristine ones after DP. We found that VC treatment had a great effect on the introduction of oxygen vacancies, and influenced the deposition behaviour of the Au NPs as well as their catalytic activity toward CO oxidation.

Experimental section

Materials

Cerium(III) nitrate hexahydrate (Ce(NO₃)₃·6H₂O), sodium hydroxide (NaOH), hydrogen tetrachloroaurate hydrate (auric acid, HAuCl₄·3H₂O), ammonium carbonate ((NH₄)₂CO₃), and L-VC (C₆H₈O₆) are of analytical grade and are used without further purification.

Preparation of CeO₂ NRs

The CeO₂ NRs were prepared by a hydrothermal method, as described by Yan's group.²⁸ Typically, Ce(NO₃)₃·6H₂O (0.05 M) and NaOH (6 M) were dissolved in 5 and 35 ml of deionized water, respectively. Then, these two solutions were mixed in a Teflon bottle, and this mixture was kept stirring for 30 min, with the formation of a milky slurry. Subsequently, the Teflon bottle with this mixture was held in a stainless steel vessel (autoclave), and the autoclave was sealed tightly. Finally, the sealed autoclave was transferred to an oven at 100 °C, and held for 24 h. After the hydrothermal reaction, the precipitates were filtered, washed with deionized water and dried at 60 °C under vacuum for 12 h.

VC treatment for CeO₂ NRs

The as-prepared CeO₂ NRs (120 mg) were dispersed in 30 ml of water. Upon addition of 88 mg of VC, the mixed solution was stirred for 2 h continuously at room temperature. Then, the mixture was centrifuged, washed with deionized water and dried at 60 °C under vacuum for 12 h.

Preparation of CeO₂ NR–Au catalysts

Pristine or VC-treated CeO₂ NRs (150 mg) was dispersed in 7.5 ml of water whilst stirring, and aqueous (NH4)₂CO₃ solution (3.75 ml; 1 M) was then added. HAuCl₄·3H₂O (99.99%, Alfa; 0.0261 mmol) was dissolved in 3.75 ml of water and added to the above solution dropwise. The pH was kept at 8–9 during the whole process. The resulting precipitate was aged at room temperature for 1 h, then filtered and washed three times with water. The product was dried in vacuum at 60 °C overnight and then calcined in air at 400 °C for 4 h.

Catalytic oxidation for CO

Catalytic activity was measured using a continuous-flow fixed-bed microreactor at atmospheric pressure. In a typical experiment, the system was first purged with high-purity N₂ gas, and then a gas mixture of CO/O₂/N₂ (1 : 10 : 89) was introduced into the reactor, which contained 50 mg of samples. Gas samples were analysed with an online infrared gas analyser (Gasboard-3121, China Wuhan Cubic Co.), which simultaneously detects CO and CO₂ with a resolution of 10 ppm. The results were further confirmed with a Shimadzu Gas Chromatograph (GC-14C).

Characterization

The samples were characterized by X-ray diffraction (XRD) on a Japan Rigaku D/Max-γA rotating anode X-ray diffractometer equipped with graphite-monochromatized Cu Kα radiation (λ = 1.54178 Å) at a scanning rate of 0.02° s⁻¹ in the 2θ range from 10° to 80°. The morphology and structure of the samples were characterized by transmission electron microscopy (TEM, JEOL 6300, 100 kV) and high-resolution TEM (HRTEM, JEM-2100, 200 kV) equipped with energy-dispersive X-ray spectroscopy (EDS, Oxford INCA). X-ray photoelectron spectroscopy (XPS) was performed on an ESCALAB 250 spectrometer with a standard Al Kα source. The charging of the samples was corrected by referencing all of the energies to the C 1s peak energy, set at 285.1 eV, arising from adventitious carbon. Optical absorption spectra were taken using a Varian 5000 spectrophotometer. Micro-Raman spectra were acquired with an RM 1000 Renishaw Raman Microscope System equipped with a laser at 633 nm.

Results and discussion

Morphology and surface structure characterization

The as-prepared CeO₂ NRs are reduced by VC in the first step (namely VC-CeO₂) to introduce oxygen vacancies, and then Au NPs are deposited on the oxygen vacancy site surface by a DP method to form the nanocomposite, Au/VC-CeO₂. Controlled experiments are also carried out, where Au NPs deposited on pristine CeO₂ NRs by the DP procedure are also obtained as Au/CeO₂.

Upon addition of VC into the pristine CeO₂ NRs slurry, the colour changes rapidly from light yellow to deep brown, as shown by the insets in Fig. 1a. The significant changes in colour may indicate the VC-induced reduction on CeO₂, which can be verified by the changes in the UV-vis spectra shown in Fig. 1a. The pristine CeO₂ NRs present a typical intense absorption band with a maximum at 310 nm, which is ascribed to the charge transfer absorption from oxygen to cerium.⁵ The charge transfer from O²⁻ to Ce³⁺, and that from O²⁻ to Ce⁴⁺, cannot be clearly distinguished here. However, when adding VC, this absorption shifts to the longer wavelength region, indicating the reduction of Ce⁴⁺ to Ce³⁺, since the electron transition from O²⁻ to Ce³⁺ is a low-energy shift relative to that from O²⁻ to Ce⁴⁺. This transition from Ce⁴⁺ to Ce³⁺ will accompany the formation of the oxygen vacancies, due to a charge compensation mechanism.²⁹ Accordingly, the band gaps (BG) of the samples can be estimated by plotting the square root of the Kubelka–Munk function^30,31 multiplied by (αhν)² versus the photon energy (hν), and extrapolating the linear part of the rising curve to zero (Fig. 1b). It is found that, on account of VC reduction, the BG decreased from 2.8 eV for pristine CeO₂ NRs to 2.7 eV for VC-CeO₂. It is worth mentioning that, according to some recent studies on photocatalysis,³² the BG can be efficiently reduced by the introduction of oxygen vacancy states between the valence band and the conduction band. Herein, the decrement of the BG may suggest the generation of oxygen vacancies associated with reduction.

Fig. 1 (a) Electronic absorption and (b) ​band-gap calculation of CeO₂ NRs, VC-CeO₂, Au/CeO₂, and Au/VC-CeO₂. Insets are photographs of the slurry of CeO₂ NRs (yellow) and VC-CeO₂ (brown).

After the introduction of the gold species, an additional broad absorption band centred at about 560 nm appears (as shown in Fig. 1a), which is ascribed to the surface plasmon (SP) resonance of Au nanoparticles, as a signature for the presence of metallic Au⁰ nanoparticles.⁵ The intensity of the SP band is slightly lowered for Au/VC-CeO₂, which may indicate a decreased Au content on the VC-reduced support. Meanwhile, the absorption corresponding to the charge transfer transitions from O to Ce blue-shifted slightly. Accordingly, the BGs of the Au-supported samples both increased relative to that of the CeO₂ NRs supports, i.e. 3.0 eV for Au/CeO₂, and 2.9 eV for Au/VC-CeO₂ (Fig. 1b). The increased BG of the Au-supported CeO₂ may be indicative of the oxygen vacancies being consumed, i.e., a strong interaction between gold and the CeO₂ NRs, and probably changes in the Ce³⁺/Ce⁴⁺ surface ratio. It was determined that during the DP process, an individual Au atom can occupy an oxygen vacancy, acting as a nucleation site for the growth of a gold cluster.^26,33 The present result agreed well with Zepeda's⁵ investigation on the loading effect of Au on CeO₂, where the author pointed out that BG changes upon Au deposition were caused by changes in the Ce³⁺/Ce⁴⁺ surface ratio due to the interaction with Au species. It is therefore expected that the VC reduction would have a great influence in regulating the metal-support interaction, and the redox properties of Au species and CeO₂, and consequently, the catalytic activity.

TEM and HRTEM (Fig. 2) reveal the morphology of the VC-treated CeO₂ NRs and those loaded with Au nanoparticles (NPs). As shown in Fig. S1,† the as-synthesized pristine CeO₂ NRs have a narrow diameter distribution from 8–10 nm and lengths of 100–400 nm, which resemble the corresponding values quoted in Yan's work.²⁸ As expected, upon VC treatment, no distinctive changes can be observed in either the crystalline structure or the surface morphology of CeO₂ NRs although a dramatic colour change was visualized (inset in Fig. 1). This might also be evidence for the reduction of Ce⁴⁺ and the formation of oxygen vacancies, which contribute to the colour change.

Fig. 2 TEM images of (a) Au/CeO₂, (b) Au/VC-CeO₂ and HRTEM identification of (c) Au/CeO₂, (d) Au/VC-CeO₂.

After the DP procedure, some dispersive dark spots are visualized for both the CeO₂ NRs and VC-CeO₂ (Fig. 2a and b), which are speculated to be the deposited Au NPs. The Au particles are highly dispersed on the surface of the CeO₂ NRs, with a small size of 3–5 nm. Compared with the pristine CeO₂ NRs, the dark spots on the VC-treated sample, VC-CeO₂, have an uneven distribution with an inhomogeneous size. The overall Au particle size increases on the VC-treated support (>5 nm), and some large Au particles with an approximate size of 30 nm are observed, suggesting that VC treatment induces agglomerates of the Au(OH)₃ precipitate during the DP procedure. The presence of the Au NPs can be evidenced from the lattice fringes in the HRTEM images (Fig. 2c and d). After the catalysed CO oxidation, the overall particle size of the Au-deposited samples remains almost unchanged, as visualized in the TEM and HRTEM images (Fig. S2†), indicating the high stability of the samples under the present preparation conditions.

Other than the Au particle size and distribution, the loading capacity of the Au NPs is largely reduced from 0.5% for Au/CeO₂ to 0.2% on the VC-treated CeO₂, as estimated from the EDS measurements, which is consistent with the decreased intensity of the SP absorption in the electronic absorption spectra (Fig. 1). The above results clearly indicate that the VC treatment affects the deposition behaviour of the Au NPs.

All the samples are also characterized by XRD measurement. The XRD patterns of pristine and VC-treated CeO₂ NRs, and their corresponding samples supporting Au NPs, are given in Fig. S3.† Both pristine and VC-treated CeO₂ NRs display similar diffraction patterns, indicating that the CeO₂ cubic fluoride crystal phase (JCPDS card no. 34-0394) is not changed upon VC treatment. After loading of Au, for either Au/CeO₂ or Au/VC-CeO₂, there are no significant diffraction peaks can be indexed to Au, implying that the Au particles were highly dispersed on the CeO₂ surface and/or with low loading.^34,35

Reducibility evaluation

H₂ temperature-programmed reduction is carried out to characterize the reducibility of the CeO₂ NRs and the Au-containing catalyst in view of the structural discrepancy caused by VC treatment. Reduction patterns are depicted in Fig. 3a and the temperatures of the peak maxima and their respective hydrogen consumption are summarized in Table 1. Both the pristine CeO₂ NRs and VC-CeO₂ show two main reduction peaks ascribed to the typical profiles of ceria supports. The high-temperature peak centered around 800 °C (HT peak) is due to bulk oxygen species, namely the total reduction of Ce⁴⁺ to Ce³⁺.^35–37 The exact HT peak position may be relevant to the reducibility of the bulk material, which is correlated with the oxygen mobility, as proposed by Hernández et al.²⁶ They found that the temperature of the maximum HT peak of Eu-doped CeO₂, namely CeEu(10), was strongly decreased in comparison with that of the bare CeO₂ solid, which was ascribed to the presence of a high concentration of oxygen vacancies. For VC-treated CeO₂, the HT reduction peak shifts slightly toward lower temperature, although marked colour changes are observed, associated with Ce⁴⁺ reduction. The broad peak below 600 °C (LT peak) is attributed to reduction of the oxygen vacancies, including the surface oxygen species adsorbed on the vacancies. It is noteworthy that the shoulder at 385 °C is probably associated with reduction of the surface oxygen of CeO₂, while the peak at about 510 °C is due to the formation of nonstoichiometric Ce oxides, CeO_x (x ranging from 1.9 to 1.7, or the β phase).³⁸ In Hernández's work,²⁶ it was suggested that Eu-doping induced an increased concentration of oxygen defects in this solid. However, CeEu(10) demonstrated LT peaks at a higher temperature than for pure CeO₂. In contrast, no peak shift was observed for the VC-treated CeO₂, in spite of the prompt reduction occurring in the bulk. However, the overall H₂ consumption was relatively reduced after VC reduction. Actually, the H₂ consumption during the surface reduction step was related to the ability of the oxygen vacancies to adsorb and dissociate H₂.¹⁷ From the lower H₂ consumption for VC-CeO₂, we speculate that VC reduction-induced formation of Ce³⁺ occurs mainly in the CeO₂ lattice, while the formation of surface oxygen vacancies is reduced, which will be discussed hereinafter.

Fig. 3 (a) H₂-TPR profiles and (b) CO conversion profiles of CeO₂ NRs, VC-CeO₂, Au/CeO₂, and Au/VC-CeO₂.

Table 1 TPR peak maxima and H₂ consumption of CeO₂ NRs, VC-CeO₂, Au/CeO₂, and Au/VC-CeO₂

Catalysts	Peak position (°C)		H2 consumption (mmol mg^−1 of catalysts)
	TLT	THT	ηLT	ηHT
CeO2 NRs	385; 512	833	56.6	42.2
VC-CeO2	386; 515	827	49.2	37.4
Au/CeO2	149	817	125.5	48.2
Au/VC-CeO2	160; 272	810	107.8	43.1

---

In the case of the reduction profiles of Au-loaded catalysts, the HT peak shifts more than a dozen nanometers toward lower temperature, together with a slightly increased H₂ consumption compared with the respective supports, which may suggest a further improved oxygen mobility and reducibility within the bulk. According to the literature,^33,35 the HT peak of the most reported gold/ceria catalysts remains unchanged upon Au deposition. We tentatively attribute the observed changes to be indicative of the Au–CeO₂ interaction. In line with previous observations in the literature,^26,35,39 the most dramatic change upon Au addition is found in the low-temperature region. The LT section shifts greatly to lower temperature, accompanied by a highly increased H₂ consumption, as an indication of the increased surface oxygen reducibility. Au/CeO₂ shows a single but slightly asymmetric peak centred at about 149 °C, while the LT peak of Au/VC-CeO₂ exhibits two overlapping reduction peaks at 160 °C and 270 °C. Such an overlapped asymmetric LT peak on account of Au deposition has been documented by Hernández et al.²⁶ for the Au/CeEu(10) catalysts. Compared with the single LT reduction peak of Au/CeO₂, peak splitting was observed for Au/CeEu(10) with a higher concentration of oxygen vacancies. It was therefore deduced that the surface oxygen reducibility may be related to interactions between the reducible species and the oxygen vacancies. However, the specific nature of this interaction was never specified in their work. According to the literature concerning the chemical properties of Au species in the gold/ceria, the increased surface oxygen reducibility is ascribed to the pronounced Au–O–Ce interaction,³⁹ resulting from the reduction of both the surface ceria and the ionic gold species.³⁵ As documented previously,^33,35 this LT peak maximum and the H₂ uptake depend heavily on the reaction pathway of Au on ceria. In the case of the sample prepared by DP, a lattice substitution mechanism is proposed for the interactions between Au and the support; namely, the Au⁺ or Au³⁺ ions would fill the vacant Ce⁴⁺ sites with consequent formation of oxygen vacancies and increased oxygen mobility and reducibility. Herein, the H₂ reduction in the low temperature region shows dramatically different behaviour for Au/CeO₂ and Au/VC-CeO₂, suggesting different interactions of Au with the pristine and VC-treated ceria. Overall, the VC-treated samples, both the Au-free CeO₂ NRs and Au/VC-CeO₂, provide depressed reducibility as a consequence of the decreased surface oxygen vacancies.

Like the TPR process, the CO oxidation also relies on the reducibility of the surface oxygen vacancies. Therefore, as expected, VC treatment induces a negative effect on CO oxidation (Fig. 3b). Specifically, both the Au-loaded samples exhibit remarkably higher activity than their corresponding CeO₂ support on account of the synergistic effect between the metal and the support, which is in accordance with the related reports. However, Au/VC-CeO₂ exhibits much more sluggish CO oxidation activity relative to that of Au/CeO₂. Even the CeO₂ support showed slightly decreased reactivity upon VC reduction. Since VC can indeed reduce the Ce⁴⁺ to Ce³⁺, why does the VC-treatment induce deactivation of CO oxidation? And how does the VC-treatment influence the deposition behaviour of Au, as well as the interaction between Au and the support? The key point is the oxygen vacancies. Therefore, the intimate correlation between the VC-reduction and the oxygen vacancy formation will be checked out hereafter. In fact, the same tendency of the reactivity of the Au/CeO₂ catalyst to rely on the concentration of oxygen vacancies has been reported by Hernández et al.²⁶ The Au/CeEu(10) material was less active than the Au/CeO₂, despite the similarity in the gold particle size and the structural and textural properties. Two principal possibilities were proposed to explain the lower activity: one was the loss of active sites from the support (oxygen vacancies); the other was the possible electronic change induced in the gold particles. Both reasons involve the interactions between the metal and the support across the oxygen vacancies.

Oxygen vacancies and Ce³⁺ influenced by VC treatment and Au deposition

Raman spectroscopy is performed to study the electronic defects on the surface of nonstoichiometric CeO₂ in response to VC reduction and Au deposition. Fig. 4 shows the visible Raman spectra of pristine and VC-treated CeO₂ NRs, and the corresponding Au–CeO₂ NRs samples. The Raman spectra of all the samples are dominated by a strong absorption peak at 440–462 cm⁻¹, which is ascribed to the oxygen breathing vibrations (F_2g mode) of the fluorite-type structure of CeO₂.^26,40 The shift to lower frequency compared to the vibration at 466 cm⁻¹ from the CeO₂ single crystal, and the broader and asymmetric character are attributed to the small particle size.⁴¹ Both CeO₂ supports, the original and the VC-treated ones, demonstrate almost identical characters within this energy region. In contrast, Laguna et al.²⁴ observed remarkable changes in this signal for Zr-doped ceria. They found a broadened and higher-energy shift of the F_2g band, along with an increase in the Zr content, as the particle size decreased and structural alteration occurred as a result of the inclusion of Zr into the cubic ceria lattice. It is therefore speculated that VC reduction induced only minor modification in the vibrational structure of CeO₂, as also indicated by the XRD results. One band at 600 cm⁻¹ (yellow stripe) is also observed, which is indexed to a defect-induced mode due to the presence of intrinsic oxygen vacancies generated by the partial reduction of Ce⁴⁺ to Ce³⁺.⁴² By means of in situ Raman analysis, Vindigni et al.⁴¹ proved that this peak was associated with inner unreactive oxygen vacancies in water–gas shift (WGS) catalysts. The intensity of this band relative to that of F_2g and I₆₀₀/I₄₆₅, is generally characterized by the number of inner defects, which is increased after VC treatment compared to that in the pristine CeO₂ NRs, indicating the formation of incremental inner defects accompanied by Ce³⁺ ions. After loading of Au by the DP method, the F_2g peak clearly shows a blue-shift compared with that of pristine CeO₂ NRs, which can be interpreted as a direct gold-support interaction across the oxygen vacancies. The blue-shift in the F_2g signal that indicates the electronic interaction between gold clusters and the support was also demonstrated in an Au-loaded Ce–Zr catalyst.²⁴ Moreover, the most important result is that, as a result of incorporation of Au, a new Raman absorption band at 540 cm⁻¹ (green stripe, D band) can be distinguished, especially for Au/CeO₂. This band is typically observed in cation-doped CeO₂,^26,42 and is attributed to the exogenous oxygen vacancies introduced by the substitution of tetravalent Ce⁴⁺ with trivalent cations. This peak, which can be generated by hydrogen reduction and is eroded upon re-exposure to O₂, manifests the presence of reactive surface peroxides, O₂.^2–43 For the VC-reduced CeO₂ NRs, the intensity of this band is strongly decreased when Au is deposited. Hernández et al.²⁶ have noticed an intensity variation of this band upon interaction with doped Au. When Au was deposited on Au/CeEu(10), the intensity of this band strongly decreased, almost disappearing. They referred such changes to a direct gold-support interaction across the oxygen vacancies by means of a “filling effect”,⁴⁴ where the oxygen vacancies, together with this vibrational mode, were eliminated as a result of their occupation by Au atoms. However, the opposite results were demonstrated by Vindigni et al.⁴¹ for the Au/CeO₂ catalyst. They found the presence of defects on AuCe673O_x, which was evidenced by the appearance of this band, and attributed this to the synergistic contribution of the highly dispersed gold, which strongly interacted with the ceria surface. The formation of oxygen vacancies, rather than elimination, may be ascribed to a lattice substitution mechanism,³³ where Au⁺ or Au³⁺ ions would fill the vacant Ce⁴⁺ sites, with consequent formation of oxygen vacancies and increased oxygen mobility and reducibility.³⁵ Herein, the distinct diversity of the intensity of this band in Au/CeO₂ and Au/VC-CeO₂ may crucially reveal the discrepant reaction mechanism of Au species with the CeO₂ support. Overall, the variation in the relative intensities of the peaks at 600 and 540 cm⁻¹ in the Raman spectra (listed in Table 2) may suggest that, first of all, VC treatment indeed reduced the Ce⁴⁺ to Ce³⁺, together with the formation of oxygen vacancies, which is dominantly associated with the inner defects, as evidenced by the marked colour change shown in Fig. 1a; secondly, the deposition of gold by the DP procedure leads an increase in both the intrinsic and the exogenous oxygen vacancies; last but not least, compared with Au/CeO₂, when Au NPs are deposited on the VC-treated surface of CeO₂, they contribute less to the generation of the surface oxygen vacancies.

Fig. 4 Raman spectra of CeO₂ NRs, VC-CeO₂, Au/CeO₂, and Au/VC-CeO₂.

Table 2 Relative intensity of the peaks I₆₀₀/I₄₆₅ and I_D/I_F2g, indicative of the intrinsic and exogenous numbers of oxygen vacancies, respectively, and the [Ce³⁺]% content of each sample in response to VC treatment and/or Au loading

Catalysts	[Ce^3+]/%	I600/I465	ID/IF2g
CeO2 NRs	25.88	0.089	—
VC-CeO2	30.39	0.104	—
Au/CeO2	33.81	0.222	0.380
Au/VC-CeO2	30.10	0.165	0.190

---

The above two diverse mechanisms, the “filling effect” and the “lattice substitution mechanism”, are related to an Au–CeO₂ interaction that may produce different properties in the metal and support, specifically the Ce³⁺/Ce⁴⁺ and Au⁰/Au^δ+ surface ratios, which are crucial in determining the reducibility. Information on the surface chemical states can be obtained from XP spectra, as shown in Fig. 5. The chemical states of Ce can be determined by analysing the Ce 3d spectra (Fig. 4a), where peaks corresponding to Ce 3d_5/2 and Ce 3d_3/2 spin–orbit contributions are denoted as V and U, respectively. The peaks labelled V⁰, V′, U⁰, and U′represent the Ce(III) state, and those labelled V, V′′, V′′′, U, U′′, and U′′′ are characteristic of the Ce(IV) state.⁵ The relative contents of Ce³⁺ and Ce⁴⁺ in all the samples were obtained by calculating the relative integration areas under the curve of each deconvoluted peak listed in Table 2. With regard to the VC-treated CeO₂ NRs, VC-CeO₂, the intensity of the peaks belonging to Ce(III) increases compared with the pristine CeO₂ NRs, suggesting an increased content of Ce(III) ions created after VC treatment. The reduction of Ce⁴⁺ and increment of Ce³⁺ are as expected, and are associated with the band edge shift in the absorption spectra.

Fig. 5 (a) XP spectra of Ce 3d, (b) O 1s of CeO₂ NRs, VC-CeO₂, Au/CeO₂, and Au/VC-CeO₂, and Au 4f_7/2 levels of (c) Au/CeO₂, and (d) Au/VC-CeO₂.

When Ce(IV) is reduced to Ce(III), oxygen vacancies may arise to maintain the charge neutrality, according to the following reaction mechanism:

2Ce_lattice⁴⁺ + 4O_lattice²⁻ → V_o + O + 2Ce_lattice³⁺ + 3O_lattice²⁻

Therefore, XPS studies concentrating on O species were also conducted. The O 1s signal clearly shows two different surface oxygen species. The low binding energy peak (O_α: 529–530 eV) is ascribed to lattice oxygen, whereas the high-binding-energy peak (O_β: 531–532.8 eV) is assigned to oxygen vacancies, together with the surface-adsorbed oxygen or surface hydroxyl species (O_s + OH_s).³⁹ The peak area ratio of O_β to O_α (in Table S1†) is checked to roughly assess the number of oxygen vacancies in these samples, from which it is apparently found that the concentration of the oxygen vacancies increased as much as the VC reduction. This increment, accompanied by an increase in the Ce³⁺, is mostly ascribed to the intrinsic oxygen vacancies, as indicated by Raman analysis. After loading the Au particles by the DP procedure, the proportion of O_β to O_α for Au/CeO₂ increased compared with the support, indicating the formation of exogenous oxygen vacancies, corresponding to the appearance of the D band in the Raman spectra. Au/VC-CeO₂, on the contrary, demonstrates a dramatically decreased O_β ratio relative to VC-CeO₂. Moreover, attention should be paid to the changes in the Ce³⁺ ratio upon Au loading. Au/CeO₂ shows a markedly increased Ce³⁺ content compared with the bare support. Au/VC-CeO₂ demonstrates a comparable Ce³⁺ content to VC-CeO₂ and a lower content than that of Au/CeO₂ (as shown in Table 2). Considering the significant increase in Ce³⁺ resulting from VC reduction, such a low Ce³⁺ content in Au/VC-CeO₂ may indicate direct consumption of Ce³⁺ by Au loading, which is contrary to the widely accepted mechanism, which can promote Ce³⁺ generation. Here, we notice a discrepancy between the XPS results and the absorption spectra with respect to the changes in Ce³⁺ and the oxygen vacancies on account of Au deposition. This is reasonable, since the band shift in the absorption spectra in response to generation or consumption of Ce³⁺ originates from changes both in the bulk and on the surface, while XPS results are mainly concerned with the changes on the surface. Overall, the above results indicate that VC treatment of CeO₂ causes the reduction of Ce⁴⁺ to Ce³⁺, which is associated with the formation of oxygen vacancies. This reduction occurred both on the surface and in the bulk, mainly resulting in intrinsic oxygen vacancies. Thus, when Au is deposited on the pristine CeO₂ or VC-CeO₂, different reaction mechanisms are expected, leading to the generation of either intrinsic or extrinsic oxygen vacancies, as indicated by Raman analysis.

Interactions between Au species and the support can also be disclosed from changes in the oxidation states of Au species when deposited on surface of CeO₂. As shown in Fig. 5c and d, Au species on the surface of either the pristine CeO₂ or the VC-treated one, carry different oxidation states of Au⁰, Au⁺, and Au³⁺.³⁹ Compared with Au/CeO₂, in Au/VC-CeO₂, the content of Au⁰ is roughly the same, while the relative compositions of Au⁺ and Au³⁺ change considerably (Table S1†). On the VC-reduced surface, the amount of Au³⁺ decreases significantly and Au⁺ becomes the dominant species. As disclosed previously,³⁹ for the water–gas shift reaction, metal nanoparticles (Au⁰) do not participate in the reaction. Non-metallic gold species (Au⁺ and Au³⁺) strongly associated with surface cerium–oxygen groups are responsible for the activity. In revealing the deactivation of Au/CeO₂ catalysts during the CO-PROX reaction, Zepeda et al.⁵ found that activity loss was connected to the fast reduction of Au species and the changes in the redox properties of ceria. Accordingly, we speculate that in this case, the different catalytic activity of Au/CeO₂ and Au/VC-CeO₂ for CO oxidation also concerns mainly Au⁺ and Au³⁺, rather than Au⁰. By theoretical modelling, it is found that the oxidation state of gold is related to CO adsorption energies, which is crucial in determining the catalytic reactivity. The more electrons on Au, the weaker the CO adsorption becomes. Therefore, the present experimental results provide direct evidence for the above theoretical speculation. Further evidence for the activity of Au^δ+ in governing the CO oxidation can be found from the XPS analysis after the catalytic reaction (Fig. S4†). The Au³⁺ for both Au-containing catalysts was depleted after reacting with CO. Au⁺ in Au/VC-CeO₂ was highly consumed in CO oxidation. On the other hand, metallic Au became the dominant species for both of the Au catalysts (Fig. S4 and Table S2†). These changes directly proved the participation of Au^δ+ in catalytic oxidation, where Au^δ+ (Au³⁺ and Au¹⁺) are reduced to Au⁰. Furthermore, it may be speculated that Au³⁺ is superior to Au¹⁺ in oxidation, since Au/CeO₂ shows much higher reducibility than Au/VC-CeO₂ (Fig. 3b). The activity of the Au species in CO oxidation is connected with the adsorption energy of CO, which is beyond the scope of this study. Accompanying the changes in the oxidation states for Au species during CO oxidation, the exogenous oxygen vacancies were also diminished after the oxidation reaction (as shown in Fig. S5†), and it is suggested that these might also be active components in CO oxidation. These findings clearly demonstrate the synergistic effect between the metal and the support.

Herein, we would consider the diverse reaction mechanism of Au with CeO₂ supports, as mentioned above. The discrepancy in the relative proportions of Au⁺ and Au³⁺ definitely indicates different reaction mechanisms of gold with Au/CeO₂ and Au/VC-CeO₂. Density functional theory calculations (DFT) by Fabris et al.¹⁵ revealed that the interaction between the Au atom and the stoichiometric CeO₂ surface involves charge transfer from the adsorbate to the substrate, yielding a positively charged Au^δ+ and reduction of a Ce ion, while binding of Au adatoms at oxygen vacancies of the reduced surface entails strong rearrangement of the Au/oxide contact occurring from the reduced oxide surface to the supported metal atom, leading to partial oxidation of Ce³⁺ to Ce⁴⁺ and reduction of the Au species. In short, two reaction mechanisms for Au loaded on CeO₂ were proposed in previous investigations: the “filling effect” and the “lattice substitution mechanism”. These two mechanisms actually oppose each other and are still an open question for debate. This may simply indicate that in the former, Au interacts with Ce³⁺ (oxygen vacancies) to form metallic or negatively charged Au species as a result of electron transfer from Ce to Au; for the latter, Au interacts with Ce⁴⁺ to form positively charged Au species associated with the generation of Ce³⁺ (oxygen vacancies), due to electron transfer from Au to Ce. Herein, in our present study, we firstly demonstrate that both of the two reaction mechanisms may be competitive and concomitant, relying on the concentration of oxygen vacancies on the support surface.

On account of the above XPS and Raman analyses, the dependence of the gold-supporting interaction on oxygen vacancies can be clearly found. The pristine CeO₂ NRs, bearing the optimal number of oxygen vacancies, will interact with Au by the lattice substitution mechanism, leading to an increase in positively charged Au³⁺ species, together with an increased amount of oxygen vacancies, which are beneficial for oxidation. When Au particles are deposited on the surface of VC-CeO₂ with an excess amount of oxygen vacancies, a filling effect occurs, which induces the reduction of Au³⁺ to Au⁺, accompanied by the oxidation of Ce³⁺ to Ce⁴⁺ and consequently a loss of reactivity. Moreover, the size effect originating from different reaction mechanisms should be considered. Highly dispersed Au nanoparticles with a smaller size are obtained with Au/CeO₂. However, larger Au NPs with a lower loading amount are obtained for Au/VC-CeO₂. In agreement with our finding, Venezia et al.³⁵ have disclosed a spontaneous tendency of gold to reduce its oxidation states by segregating out of the ceria lattice. Taking into account the fact that the size and distribution of Au NPs on the surfaces of pristine and VC-treated CeO₂ supports are extremely discrepant, different reaction pathways of the Au species on CeO₂ supports may be expected with a controlled amount of oxygen vacancies.

Reaction pathway of Au on CeO₂ mediated by VC reduction

During the VC treatment, the yellow color of CeO₂ NRs, originating from Ce(IV)–O(II) charge transfer, drastically changes to a deep brown color due to the formation of a nonstoichiometric structure (CeO_2−x), leading to an increase in defects in treated CeO₂ NRs. The relative intensity ratio, I₆₀₀/I_F2g, in the Raman spectra discussed above provides direct evidence in support of an increase in oxygen vacancies after treatment. However, after DP of Au NPs, the changes in the oxygen vacancies display different trends, where the number of oxygen vacancies decreased sharply in Au/VC-CeO₂ compared with Au/CeO₂, as observed in Raman and XPS spectra.

With regard to Au deposition by the DP method, there is a well-established mechanism, in which a hydroxychloro gold(III) complex [Au(OH)_xCl_4−x]⁻ is formed as a result of the consecutive exchange of Cl⁻, originating from HAuCl₄, with OH⁻ as the pH value increases during the DP process. However, kinetically, the exchange rate of the Cl⁻ with the OH⁻ ligand on gold(III) complexes is sluggish. Therefore, when HAuCl₄ is mixed with VC-treated CeO₂ NRs with a high content of oxygen vacancies, there is a high possibility that Au ions would be reduced directly, instead of the formation of the [Au(OH)_xCl_4−x]⁻ complex. To verify this possibility, pristine and VC-treated CeO₂ NRs were dispersed in water and HAuCl₄ aqueous solution was added drop by drop. Interestingly, for the VC-treated CeO₂ NRs, the color changed into pink within a few minutes, while for the pristine version, no color change was observed. After centrifugation and thorough washing, these two samples were further characterized by TEM (Fig. S6†). Spherical Au NPs are clearly observed supported on the VC-treated CeO₂ NRs (Fig. S56†). However, neither Au NPs nor an Au signal in the EDX spectrum is observed for the pristine CeO₂ NRs (Fig. S6a†). This result undoubtedly demonstrates the strong reducing ability of VC-CeO₂ NRs, which can quickly reduce Au ions to Au NPs. In comparison with pristine CeO₂ NRs, the increased reducing ability of VC-CeO₂ should originate from the large number of generated oxygen vacancies. During the reduction procedure, VC-CeO₂ with a high content of oxygen vacancies transfers the electrons to Au ions, resulting in a sharp decrease in oxygen vacancies, and the oxidation of Ce³⁺ to Ce⁴⁺, together with the reduction of Au³⁺ to Au⁺, as demonstrated by XPS.

Overall, although VC is supposed to be a strong reductant and reduce the CeO₂ to generate a higher number of oxygen species, the excess amount of oxygen species leads to several negative effects when loading Au by the DP procedure, which results in a tendency toward CO oxidation. The excess amount of oxygen species has strong reducing ability, reducing the HAuCl₄ promptly, resulting in Au NPs with a large size; such strong electron transfer from the oxygen species (Ce³⁺) to Au³⁺ induces an increased number of Au⁺. The Au⁺ is inclined to diffuse into the vacancies and transforms into negatively charged Au^δ− adspecies, which would inhibit the adsorption of molecular CO or O₂, leading to the deactivation of the catalyst.¹⁵ The decrease in oxygen species as a consequence of such strong charge transfer will also diminish the ability to weaken the C–O bond,⁴⁵ thus leading to a decrease in catalytic activity.

Conclusions

VC treatment induces the reduction of Ce⁴⁺ to Ce³⁺ and creates oxygen vacancies in CeO₂ NRs, which are predominantly intrinsic oxygen vacancies that are inactive for CO oxidation. However, these increased oxygen vacancies endow the VC-treated CeO₂ NRs with a certain reduction ability, allowing them to reduce AuCl⁴⁻ ions to Au NPs directly, rather than forming the hydroxychloro gold(III) complex, [Au(OH)_xCl_4−x]⁻. Such a reaction pathway, accompanied by strong charge transfer between the Au species and oxygen vacancies on CeO₂, results in rapid growth of Au NPs with a large size. Meanwhile, the Ce³⁺ generated by VC reduction is oxidized by Au³⁺, leading to a decreased Ce³⁺ content, a decrease in surface oxygen vacancies, and the reduction of Au³⁺ to Au⁺. As a consequence, the overall CO oxidation is deactivated. This work proves that the Au deposition during the DP procedure relies strongly on the content of oxygen vacancies in CeO₂, while the charge transfer between the Au species and oxygen vacancies has double-edged effects on the catalytic activity. On one hand, oxygen vacancies are binding sites for Au atoms to form highly dispersed small Au NPs. Charge transfer between Au and the support induces positively charged Au³⁺ species and Ce³⁺, associated with the generation of oxygen vacancies. “A synergistic effect” occurs at the gold-support interface, which can activate the lattice oxygen atoms from the support to the surface for the oxidation reactions. On the other hand, an excess amount of oxygen vacancies induces a strong reducing ability and strong charge transfer from the support to Au, leading to agglomerated Au NPs and rapid reduction of Au³⁺ reactive species, therefore resulting in catalyst deactivation. Moreover, this work demonstrates that a high degree of oxygen vacancies changes the Au NP formation pathway during the DP procedure, which may provide some enlightenment for the preparation of reducible metal oxide-supported Au composite catalysts.

Acknowledgements

The authors acknowledge financial support from the National Natural Science Foundation of China (21371109), the ‘‘Taishan Scholar’’ project of Shandong Province, and ​the Shandong Provincial Natural Science Foundation, China (Grant No. ZR2015BM008). Ke Tang is acknowledged for his help with H₂-TPR measurements.

Notes and references

1. D. R. Mullins, S. H. Overbury and D. R. Huntley, Surf. Sci., 1998, 409, 307.
2. L. J. Wu, H. J. Wiesmann, A. R. Moodenbaugh, R. F. Klie, Y. M. Zhu, D. O. Welch and M. Suenaga, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 69, 125415.
3. W. Gao, Z. Zhang, J. Li, Y. Ma and Y. Qu, Nanoscale, 2015, 7, 11686.
4. C. J. Zhang, A. Michaelides and S. J. Jenkins, Phys. Chem. Chem. Phys., 2011, 13, 22.
5. J. A. Hernández, S. A. Gómez, T. A. Zepeda, J. C. Fierro-González and G. A. Fuentes, ACS Catal., 2015, 5, 4003.
6. F. F. Zhu, G. Z. Chen, S. X. Sun and X. Sun, J. Mater. Chem. A, 2013, 1, 288.
7. E. Wahlström, N. Lopez, R. Schaub, P. Thostrup, A. Rønnau, C. Africh, E. Lægsgaard, J. K. Nørskov and F. Besenbacher, Phys. Rev. Lett., 2003, 90, 026101.
8. C. Zhang, A. Michaelides, D. A. King and S. J. Jenkins, J. Phys. Chem. C, 2009, 113, 6411.
9. K. Kośmider, V. Brázdová, M. V. Ganduglia-Pirovano and R. Pérez, J. Phys. Chem. C, 2016, 120, 927.
10. J.-F. Jerratsch, X. Shao, N. Nilius, H.-J. Freund, C. Popa, M. V. Ganduglia-Pirovano, A. M. Burow and J. Sauer, Phys. Rev. Lett., 2011, 106, 246801.
11. W. Song and E. J. M. Hensen, J. Phys. Chem. C, 2013, 117, 7721.
12. P. Ghosh, M. Farnesi Camellone and S. Fabris, J. Phys. Chem. Lett., 2013, 4, 2256.
13. Z.-P. Liu, S. J. Jenkins and D. A. King, Phys. Rev. Lett., 2005, 94, 196102.
14. C. Zhang, A. Michaelides, D. A. King and S. J. Jenkins, J. Am. Chem. Soc., 2010, 132, 2175.
15. M. F. Camellone and S. Fabris, J. Am. Chem. Soc., 2009, 131, 10473.
16. M.-W. Chang and W.-S. Sheu, Phys. Chem. Chem. Phys., 2016, 18, 15884.
17. C. J. Weststrate, R. Westerström, E. Lundgren, A. Mikkelsen and J. N. Andersen, J. Phys. Chem. C, 2009, 113, 724.
18. W. Y. Hernández, M. A. Centeno, F. Romero-Sarria and J. A. Odriozola, J. Phys. Chem. C, 2009, 113, 5629.
19. E. D. O. Jardim, S. Rico-Francés, F. Coloma, E. V. Ramos-Fernández, J. Silvestre-Albero and A. Sepúlveda-Escribano, Appl. Catal., A, 2014, 487, 119.
20. O. H. Laguna, F. Romero Sarria, M. A. Centeno and J. A. Odriozola, J. Catal., 2010, 276, 360.
21. T. R. Reina, S. Ivanova, M. I. Domínguez, M. A. Centeno and J. A. Odriozola, Appl. Catal., A, 2012, 419–420, 58.
22. T. R. Reina, E. Papadopoulou, S. Palma, S. Ivanova, M. A. Centeno, T. Ioannides and J. A. Odriozola, Appl. Catal., B, 2014, 150–151, 554.
23. T. R. Reina, W. Xu, S. Ivanova, M. Á. Centeno, J. Hanson, J. A. Rodriguez and J. A. Odriozola, Catal. Today, 2013, 205, 41.
24. O. H. Laguna, A. Pérez, M. A. Centeno and J. A. Odriozola, Appl. Catal., B, 2015, 176–177, 385.
25. P. Sudarsanam, B. Mallesham, P. S. Reddy, D. Großmann, W. Grünert and B. M. Reddy, Appl. Catal., B, 2014, 144, 900.
26. W. Y. Hernández, F. Romero-Sarria, M. A. Centeno and J. A. Odriozola, J. Phys. Chem. C, 2010, 114, 10857.
27. A. Corma and H. Garcia, Chem. Soc. Rev., 2008, 37, 2096.
28. H. X. Mai, L. D. Sun, Y. W. Zhang, R. Si, W. Feng, H. P. Zhang, H. C. Liu and C. H. Yan, J. Phys. Chem. B, 2005, 109, 24380.
29. J. Wang, D. N. Tafen, J. P. Lewis, Z. L. Hong, A. Manivannan, M. J. Zhi, M. Li and N. Q. Wu, J. Am. Chem. Soc., 2009, 131, 12290.
30. J. Ng, S. Xu, X. Zhang, H. Y. Yang and D. D. Sun, Adv. Funct. Mater., 2010, 20, 4287.
31. N. Serpone, D. Lawless and R. Khairutdinov, J. Phys. Chem., 1995, 99, 16646.
32. J. Gan, X. Lu, J. Wu, S. Xie, T. Zhai, M. Yu, Z. Zhang, Y. Mao, S. C. I. Wang, Y. Shen and Y. Tong, Sci. Rep., 2013, 3, 1021.
33. Q. Fu, H. Saltsburg and M. Flytzani-Stephanopoulos, Science, 2003, 301, 935.
34. H. Zhang, Y. Xie, Z. Sun, R. Tao, C. Huang, Y. Zhao and Z. Liu, Langmuir, 2011, 27, 1152.
35. A. M. Venezia, G. Pantaleo, A. Longo, G. Di Carlo, M. Casaletto, F. L. Liotta and G. Deganello, J. Phys. Chem. B, 2005, 109, 2821.
36. H. C. Yao and Y. F. Yu Yao, J. Catal., 1984, 86, 254.
37. S.-Y. Lai, Y. Qiu and S. Wang, J. Catal., 2006, 237, 303.
38. J. Z. Shyu, W. H. Weber and H. S. Gandhi, J. Phys. Chem., 1988, 92, 4964.
39. S. Zhang, X. S. Li, B. B. Chen, X. B. Zhu, C. Shi and A. M. Zhu, ACS Catal., 2014, 4, 3481.
40. V. G. Keramidas and W. B. White, J. Chem. Phys., 1973, 59, 1561.
41. F. Vindigni, M. Manzoli, A. Damin, T. Tabakova and A. Zecchina, Chem.–Eur. J., 2011, 17, 4356.
42. A. D. Liyanage, S. D. Perera, K. Tan, Y. Chabal and K. J. Balkus, ACS Catal., 2014, 4, 577.
43. V. V. Pushkarev, V. I. Kovalchuk and J. L. d'Itri, J. Phys. Chem. B, 2004, 108, 5341.
44. M. G. Sanchez and J. L. Gazquez, J. Catal., 1987, 104, 120.
45. L.-S. Zhong, J.-S. Hu, A.-M. Cao, Q. Liu, W.-G. Song and L.-J. Wan, Chem. Mater., 2007, 19, 1648.

---

Footnote

† Electronic supplementary information (ESI) available: TEM images of the supporters and the Au–CeO₂ with their EDS, XRD pattern of the as-prepared samples. See DOI: 10.1039/c6ra14778j

---