Introduction

Since the exceptional discovery of Haruta who demonstrated that, when smaller than 5 nm, the gold nanoparticules can display very high catalytic CO oxidation activity even at 203 K [1, 2], hundreds of papers on the topic of catalysis by gold have been annually published with a progression which remains exponential. Then, numerous catalytic formulations have been developed, and the efficiency of nanostructured gold deposited on reducible transition metal oxide supports was particularly reported [3]. Among the reducible oxide supports, ceria was particularly investigated because of its capability to act as an oxygen reservoir by storing/releasing oxygen through a redox process which involves the Ce4+/Ce3+ couple. Nanocrystalline ceria-supported gold catalysts were then studied in various catalytic reactions including the water–gas shift reaction [4, 5], the VOC combustion [610], the selective oxidation of alcohol [11], the selective oxidation of CO in excess of H2 (CO-PROX) [12], and the most investigated reaction remaining the low-temperature CO oxidation [13, 14]. The morphology of the ceria crystallites, the size of the gold clusters (particles), and the oxidation state of the active gold atoms are the parameters usually discussed because of their determining importance for the catalytic oxidation activity [15]. The nature of the intermediate species is also studied. For instance, it has been clearly demonstrated that one of the key elements for explaining the high reactivity of nanometer-sized gold nanoparticles in CO oxidation was the abundance of low-coordinated Au atoms in the small particles, where CO can be preferentially adsorbed [16]. On the contrary, the nature of the intermediate species in the oxygen activation step is still a matter of debate [17]. One reason is the fact that the adsorbed oxygen species are not experimentally easy to identify in the conditions of the reaction. Some recent papers showed that the anion photoelectron spectroscopy [18] and infrared multiple photon dissociation [19] allowed determining which oxygenated species are involved in the activation process, but these spectroscopic studies have not been applied to ceria-supported gold catalysts. Computational method is a way to get around the experimental limitations, and very impressive demonstrations in terms of reactive intermediate species and mechanism scheme were reported [17, 20]. Nevertheless, the use of the theoretical modeling did not permit to prevent controversial conclusions.

In this work, we used a combination of characterization techniques involving oxygen storage capacity measurement (OSC), oxygen isotopic exchange (OIE), and Raman and electron paramagnetic resonance (EPR) spectroscopies. These techniques were extensively used on cerium-based oxides to understand the mechanism of dioxygen activation [2123] and to explain the property of these oxides to store and release oxygen [24, 25]. Thus, we proposed to apply the same techniques on ceria-supported gold catalysts. In addition, gold supported on mixed oxide consisting of cerium oxide supported on alumina with two CeO2 loadings (5 to 10 wt.%) were also studied. The latter catalytic formulations showed interesting catalytic performances in the oxidation of two types of VOCs as follows: propene [7] as model of hydrocarbon and 2-propanol [26] as model of alcohol. Ceria–alumina was preferred to ceria support because of the poor resistance of the latter against thermal sintering.

Experimental part

Materials

Alumina, AluC Degussa (110 m2 g−1), was used as the support to load cerium oxide (5 and 10 wt.% with respect to alumina) by impregnation in excess of aqueous solution of Ce(NO3)3, 6H2O (Aldrich, 99.9 %), followed by calcination at 500 °C. One weight percent of gold was loaded by deposition–precipitation with urea on the various CeO2–Al2O3 samples as well as on pure alumina and ceria of high surface area (HSA-5 Rhodia, 200 m2 g−1) and low surface area, sintered HSA-5 by aging at 800 °C under a flow of humid air for 12 h (abbreviated as CeO2 LSA 50 m2 g−1). The catalysts are named as Au/xCeO2/Al2O3 (x = wt% of CeO2), Au/Al2O3, and Au/CeO2. The details of experimental procedures regarding catalyst preparation and characterization by elemental analyses, N2 physisorption at 77 K, XRD, X-ray photoelectron spectroscopy (XPS), and combined TEM–energy-filtered transmission electron microscopy (EFTEM) techniques were largely described in our previous study [7].

Oxygen storage capacity

OSC measurements were carried out at 673 K using a U-form reactor connected to a gas chromatograph equipped with a Porapak column and a thermal conductivity detector. The experimental set-up and the protocol of experiment were reported elsewhere [27]. The samples were placed into the reactor and heated under a continuous flow of helium (30 cm3 min−1) up to 673 K for 30 min before pretreatment with ten pulses (0.246 cm3) of pure O2 at atmospheric pressure followed by a purge with pure He for 10 min. Alternate pulses of CO and O2 were undertaken three times in order to check the reproducibility of the measurement. The amounts of unconverted CO and O2 as well as of produced CO2 were quantified. The OSC was calculated from the average value of CO2 production (after CO pulse) and was expressed in μmol O g−1 or µmol O m−2 taking into account the mass or the BET surface area of the samples.

The number of surface oxygen atoms involved in the OSC process was calculated considering that only ceria participated and assuming a preferential (100) orientation of the ceria surface. Calculations have been already described in a previous publication [27].

Oxygen isotopic exchange

OIE experiments were performed in a set-up already described elsewhere [27, 28]. A U-form reactor was placed in a closed recycle system which was connected on one side to a mass spectrometer (Pfeiffer Vacuum, QMS 200) for the monitoring of the gas-phase composition and on the other side to a vacuum pump. A recycling pump removed limitations due to gas-phase diffusion. OIE experiments were undertaken on 20 mg of catalyst, subjected to a 16O2 activation step at 873 K under atmospheric pressure for 1 h prior to cooling to the desired temperature, at which point the system was degassed and the isotopic mixture charged. The study of homo-exchange (also called equilibration reaction) was performed using an equimolar mixture of 16O2 and 18O2 (99.9 % purity, supplied by Isotec), whereas for the study of the hetero-exchange, the mixture was replaced by pure 18O2. The masses 32, 34, and 36 m/z were monitored as a function of time to follow the exchange. The m/z values of 28 and 44 were also recorded to check the absence of air or CO2. The atomic fraction of 18O in the gas phase (αg), the rate of exchange (Re), and the number of O atoms exchanged (Ne) were calculated as described in previous references [29]. Typically:

$$ {\alpha}_{\mathrm{g}}=\frac{{\mathrm{P}}_{36}+\frac{1}{2}{\mathrm{P}}_{34}}{{\mathrm{P}}_{36}+{\mathrm{P}}_{34}+{\mathrm{P}}_{32}} $$
(1)

Where P36, P34, and P32 were the partial pressures of 18O2, 18O16O, and 16O2, respectively;

$$ {\mathrm{R}}_{\mathrm{e}}=-{\mathrm{N}}_{\mathrm{g}}\frac{{\mathrm{d}\upalpha}_{\mathrm{g}}}{\mathrm{d}\mathrm{t}} $$
(2)

Where Ng was the number of 18O atoms in gas phase at the beginning of the reaction;

$$ {\mathrm{N}}_{\mathrm{e}}={\mathrm{N}}_{\mathrm{g}}\left(1-{\upalpha}_{\mathrm{g}}\right) $$
(3)

Finally, the number of exchangeable atoms could be calculated when equilibrium between the gas-phase and the solid was reached by using:

$$ {\mathrm{N}}_{\mathrm{s}}=\frac{N_e}{\alpha^{*}}={\mathrm{N}}_{\mathrm{g}}\left[\frac{1-{\alpha}^{*}}{\alpha^{*}}\right] $$
(4)

Where α* was the value of αg at equilibrium.

Raman spectroscopy

The Raman study was performed using a KAISER RXN1 spectrometer equipped with a NIR (785 nm) laser diode (25< power <50 mW). The powder sample was put in a built-in cell that allowed heating and spectra acquisition under gas flow. Once in the cell, the sample was flushed under a flow of O2 (50 cm3 min−1) and heated up to 693 K (5 K min−1) for 1 h. Then, the O2 flow was switched to Ar (same flow rate), and the cell was cooled down to room temperature (RT), at which point a spectrum was registered. The sample was finally flushed 30 min with O2 flow (150 cm3 min−1) at RT, then possibly flushed with an Ar flow (150 cm3 min−1).

Electron paramagnetic resonance spectroscopy

The EPR spectra were recorded on a JEOL FA-300 series EPR spectrometer at ∼9.3 GHz (X-band) using a 100 kHz field modulation and a 2.5 G standard modulation width. The spectra were recorded at 77 K using an insertion Dewar containing liquid nitrogen. Computer simulation of the spectra was performed using the EPRsim32 program [30].

The sample was introduced into a cell consisting of a U-shape reactor with a porous disk for thermal treatment and connected to an EPR tube. The cell can be closed using vacuum valves. After 10 min in dynamic primary vacuum (≈10−2 mbar) at RT, 250 mbar O2 was introduced into the cell, which was then heated at 773 K (5 K min−1) for 2 h. After decreasing the temperature down to 723 K, the sample was evacuated under dynamic primary then secondary vacuum (≈10−4 mbar, 10 min) before rapid cooling to RT. At RT, increasing pressures of O2 were introduced into the cell (from 1.5 up to 50 mbar). After internal transfer of the sample into the EPR tube, the spectra were recorded after 30 min cooling at 77 K (acquisition time: 8 min). The “raw” double integration of the EPR signal (arbitrary units) was performed using the cwEsr software (3.3.36E XB version) provided by JEOL. The double integration per gram of sample, which is proportional to the amount of paramagnetic species in the absence of dipolar interaction, was calculated considering the amount of sample in the EPR tube for each measurement.

Results and discussion

The characterization of the samples studied in this work has been already described in a previous work [7]. The main physicochemical parameters are summarized in Table 1. The BET surface area of pure alumina support slightly decreased as the CeO2 loading increased, while the high BET surface area of commercial ceria was maintained in the Au/CeO2 catalyst to 200 m2 g−1. In Au/xCeO2/Al2O3 (x = 5 and 10 wt.%), mainly 2-D patches and 3-D nanoparticles of CeO2 (ca 8 nm) were detected by XRD and EFTEM in accordance with similar observation made by Martínez-Arias [31]. The average sizes of the gold particles visible by EFTEM also reported in Table 1 did not depend on the support but were generally smaller after reduction than after calcination treatment (note that gold particles are visible only on alumina because of the poor contrast between gold and ceria). Finally, XPS and CO oxidation model reaction showed that whatever the mode of activation (thermal treatment under H2 or O2), all gold was metallic after reduction under H2, but gold remained unreduced on ceria or ceria patches after calcination under O2, while it was metallic on alumina.

Table 1 BET surface area, Au loading, and Au particle size of the catalysts

Oxygen storage capacity

OSC measurements were performed at 673 K. The values obtained for the various samples are reported in Table 2, in which one can distinguish the OSC of the bare supports (denoted “without Au”) and the OSC of the samples containing gold (denoted “with Au”). As it could be anticipated, the Al2O3 support did not display any OSC activity at 673 K. The activity, i.e., the CO2 production, remained null for Au/Al2O3, confirming the reduced state of gold when deposited on alumina after an oxidation stage at 673 K. Indeed, oxidized gold would have been reduced by the CO pulses, and CO2 would have been produced. OSC was detectable for 5CeO2/Al2O3 and 10CeO2/Al2O3 but was not really different for the two samples. The OSC of bare ceria was much higher, 162 μmol O g−1, corresponding to 0.81 μmol O m−2. In order to assess the proportion of oxygen atoms involved in this process, the latter value was compared with the theoretical one. Taking into account that only one oxygen atom out of four is involved in the Ce4+/Ce3+ reduction step, the theoretical OSC for ceria can be estimated as 5.7 μmol O m−2. The ratio between the experimental and theoretical OSC values gives the number of surface oxide layers (NL) participating to the process. For the bare ceria (200 m2 g−1), the results in Table 2 show that at 673 K, NL is restricted to a small fraction of the surface (NL = 0.14). The calculation performed for the ceria of low LSA area shows that NL is the same (NL = 0.15) and therefore that the OSC process is only dependent on the surface area for a given oxide. In contrast to what was expected, the NL values were higher when ceria was supported on alumina.

Table 2 OSC of xCeO2/Al2O3 and CeO2 at 673 K without or with the presence of 1 wt.% Au

We can notice a beneficial effect of the presence of gold nanoparticles on the OSC values for the various supports (Table 2). On CeO2, after removing the CO2 produced by the reduction of oxidized gold, the OSC value was three times as high in the presence of Au. It is worth noting that on CeO2 LSA, the increase of the OSC by the presence of gold was much lower, showing that the OSC is not only dependent on the ceria surface but also on the Au/CeO2 interaction. Flytzani-Stephanopoulos et al. also reported the improvement of the OSC property of ceria due to the presence of gold [32]. The impact of Au is more pronounced for the ceria–alumina mixed oxide supports, for which the OSC process involves the participation of bulk oxygen atoms (NL >1 indicates that the OSC is not limited to the surface) and for which the OSC seems to depend on the CeO2 content (NL = 1.81 and 2.19 for Au/5CeO2/Al2O3 and Au/10CeO2/Al2O3, respectively).

Isotopic oxygen exchange (18O2/16O2 IE)

We first investigated the homo-exchange reaction (16O2(g) + 18O2(g) ↔ 2 16O18O(g)) by introducing an equimolar mixture of 18O2/16O2 in the Au/CeO2 system, and the oxygen isotopic exchange evolution was followed in a temperature-programmed experiment. When temperature increased, a decrease of the 18O2 partial pressure and an increase of the partial pressures of 16O18O and 16O2 were observed (Fig. 1). A decrease of the 18O atomic fraction (αg) was observed in the gas phase when the exchange was taking place. This behavior is not consistent with a homo-exchange reaction during which αg should remain constant. This result is therefore the indication of a hetero-exchange process in which 16O lattice oxygen atoms, supplied by ceria, participates to the reaction. A very similar evolution of the oxygen isotopomer partial pressures was observed on pure CeO2 (Fig. S1), meaning that the hetero-exchange is due to ceria. Such a result emphasizes the inability of the gold nanoparticles to dissociatively adsorb molecular dioxygen or shows that the dissociation and the exchange with lattice oxygen atoms occur simultaneously.

Fig. 1
figure 1

Evolution of the oxygen isotopomer partial pressures during a tempertaure-programmed homo-exchange experiment on Au/CeO2 catalyst

In the objective to study whether gold nanoparticles influence the mobility of ceria oxygen atoms or not, we performed an isothermal hetero-exchange experiment, introducing pure 18O2 (ca 50 mbar) in the reactor cell at 723 K. The results, in terms of initial rate of exchange (Re) and number of atoms exchanged (Ne) after 20 min reaction, are reported in Table 3. Again, in this set of experiments, we compared the bare supports (columns denoted “without Au”) with the gold catalysts (“with Au”). It clearly appears that the gold nanoparticles exalt the isotopic exchange activity. Since it is well known that Al2O3 support is not able to exchange oxygen at 723 K [33], these results show that the exchange is promoted by the interaction between Au nanoparticles and CeO2. The Re values, which were low for 5CeO2/Al2O3 and 10CeO2/Al2O3, increased in the presence of gold to a value quite similar for both, ca 6–7 × 1017 at O g−1 s−1. The similarity between both solids was confirmed in comparing the number of exchanged atoms (Ne values): 5.43 and 5.53 at O nm-2 were exchanged for Au/5CeO2/Al2O3 and Au/10CeO2/Al2O3, respectively, suggesting that the proportion of gold particles in interaction with ceria was close in the two cases. The Re value normalized per gram of sample was largely higher on bare CeO2 (55.3 × 1017 at O g−1 s−1). Despite this high value, the presence of Au led to an increase of Re (77.8 × 1017 at O g−1 s−1). If one examines the Ne values, their increase was not as drastic as for the Re values (13.7 at O nm−2 and 14.7 at O nm−2 for CeO2 and Au/CeO2, respectively). This can be explained by an equilibrium rapidly reached in the 18O concentration between the gas phase and the solid (depending on the mass of catalyst and the initial pressure of 18O2), which minimized the differences that could be detected after 20 min of exchange. Finally, we studied the effect of the ceria surface area with the results obtained on CeO2 LSA. Contrary to the conclusion previously made regarding the OSC measurement (Table 2), it clearly appears that the oxygen exchange activity is not dependent on the surface area only, since the Ne value, normalized per surface unit, was five times as small for CeO2 LSA as compared to CeO2. The beneficial effect of Au was also noticed on CeO2 LSA, since the Re and Ne values were almost four times higher in the presence of the metal. Nevertheless, the activity remained smaller than that of the high surface area ceria.

Table 3 18O2/16O2 isotopic exchange experiments on pre-oxidized samples at 723 K

In order to ascertain the conclusion that the interaction between Au particles and CeO2 surface was the key element to explain the strong activity in oxygen exchange, we prepared two other gold catalysts with higher gold loading, i.e., containing 4 wt.% Au deposited on 10CeO2/Al2O3 and CeO2. Note that the size of the Au particles was maintained around 2.5 nm. The higher gold loading resulted in higher Re and Ne values (see values between parentheses in Table 3). A closer inspection of the evolutions of the 16O2 (m/z = 32) and 18O16O (m/z = 34) partial pressures as a function of time (Fig. 2) provided possible reasons for the beneficial impact of gold in the oxygen exchange reaction. At the beginning of the reaction, both 18O16O and 16O2 molecules appeared simultaneously on 10CeO2/Al2O3 (Fig. 2a) and on ceria support (Fig. 2c). This suggests that simple exchange (Eq. 4) and multiple exchange (Eq. 5) reactions occurred on these supports as already reported for ceria in previous studies [27, 29] as follows:

Fig. 2
figure 2

Evolution of the oxygen isotopomer partial pressures as a function of time during an hetero-exchange experiment at 723 K on a 10CeO2/Al2O3, b Au(4 wt.%)/10CeO2/Al2O3, c CeO2, and d Au(4 wt.%)/CeO2 catalysts

$$ {}^{18}{\mathrm{O}}_{2\left(\mathrm{g}\right)}{+}^{16}{\mathrm{O}}_{\left(\mathrm{s}\right)}{\to}^{18}{\mathrm{O}}^{16}{\mathrm{O}}_{\left(\mathrm{g}\right)}{+}^{18}{\mathrm{O}}_{\left(\mathrm{s}\right)} $$
(4)
$$ {}^{18}{\mathrm{O}}_{2\left(\mathrm{g}\right)}+{2}^{16}{\mathrm{O}}_{\left(\mathrm{s}\right)}{\to}^{16}{\mathrm{O}}_{2\left(\mathrm{g}\right)}+{2}^{18}{\mathrm{O}}_{\left(\mathrm{s}\right)} $$
(5)

The presence of gold made the 16O2 isotopomer the main molecule present at the very beginning of the experiments (Fig. 2b, d), indicating that Au favored the multiple exchange. The promotion of the multiple hetero-exchange mechanism by gold when supported on Ce-doped ordered mesoporous alumina materials has already been reported by Fonseca et al. [34]. Taking into account that the multiple exchange must involve binuclear oxygen species as intermediate, it can be inferred from Fig. 2 that the interfaces between the gold nanoparticles and the ceria crystallites are the preferential location for dioxygen activation, in the form of either diatomic peroxide or superoxide species. The correlation between multiple exchange and peroxide or superoxide intermediate species was suggested by Winter to explain the complex exchange mechanism observed over basic oxides [35]. Corma et al. provided spectroscopic evidences that superoxide species and peroxide adspecies at the one-electron defect site were the active species in the CO oxidation reaction for gold supported on nanocrystalline CeO2 [36].

To investigate the nature of the binuclear oxygen species adsorbed on the catalytic surface, characterizations by Raman and electron paramagnetic resonance spectroscopies were performed in focusing the study on the impact of the presence of gold nanoparticles.

Identification of oxygen active species

In an attempt to detect undissociated oxygen species by Raman spectroscopy, the catalyst pretreatments were performed in conditions as close as possible as those used for the isotopic exchange study (see Experimental part). Raman spectroscopy is a sensitive and specific technique capable to discriminate oxygen species (O2, O2 or O2 2−) adsorbed on CeO2, as reported in earlier works [22, 37]. The vibrational frequencies of these adsorbed oxygen species strongly decrease with their electronic charge; in the 820–890 cm−1 range for O2 2− (peroxo), 1,120–1,140 cm−1 for O2 (superoxo), and 1,480–1,570 cm−1 for physically adsorbed O2 [22, 37]. Moreover, the resulting bands are sharp compared to the ones related to ceria lattice vibrations, which make them quite easy to identify.

First of all, due to the strong fluorescence of the Al2O3-containing materials, only CeO2 and Au/CeO2 materials could be studied. Under O2 flow at RT, apart from the vibration bands characteristic of ceria (Fig. S2), i.e., an intense band at 458 cm−1 attributed to 1st order F 2g lattice vibrational mode of O h and weaker bands at 253 and 1,176 cm−1 attributed to second order modes, arising from a mixing of A1g, Eg, and F2g lattice vibrational modes [38] (additional bands were visible at 1,357, 1,404, and 1,380 cm−1, assigned to residual carbonate/carboxylate species), the spectra of CeO2 and Au/CeO2 (Fig. 3) revealed the presence of a sharp band at 1,551 cm−1 that grew under O2 flow. It can be reasonably attributed to physically adsorbed O2, since it easily disappeared by flushing with inert gas. One can note that apart from the vibration bands characteristic of ceria described above, a shoulder around 590 cm−1 was also visible for both CeO2 and Au/CeO2 (Fig. S2). It is related to the presence of oxygen vacancies [39]. The intensity of the 590 cm−1 shoulder relative to the F2g mode band (458 cm−1) is about twice as high for Au/CeO2 (9 × 10−2) as for CeO2 (4 × 10−2), revealing a much more oxygen-defective material in the presence of gold. These additional vacancies are probably located at the perimeter of the interface between gold and ceria and would explain the exaltation of the multiple exchange mechanism observed by isotopic exchange technique for gold-containing catalysts [40]. In addition, tiny bands at 810 and 826 cm−1 could be observed on Au/CeO2 only (not on CeO2), in the frequency range of adsorbed O2 2− peroxo species. No O2 (superoxo) species could be observed. It may be noted that the observation of such superoxo species often requires low temperature (93 K), which could not be reached with our spectrometer. In order to ascertain the presence of O2 (superoxo) species, EPR experiments were also performed.

Fig. 3
figure 3

Raman spectra after calcination at 693 K Ar, then O2 flow of a CeO2, and b Au/CeO2

To avoid the interaction between the radical oxygen species and the gas phase O2 which is known to induce signal broadening, it was necessary to perform the EPR experiments at low temperature, i.e., at the liquid nitrogen temperature, 77 K, which is the usual temperature used for the identification of the oxygen radicals. The EPR study was first performed over CeO2 and Au/CeO2. Again, the pretreatments had to be adapted to the technical constraints, but were performed in conditions as close as possible to those of isotopic exchange (see Experimental part). After treatment under vacuum at 723 K, the EPR spectra of CeO2 and Au/CeO2 samples revealed a weak anisotropic signal at g ≈ 1.97 (Figs. 4 and S3) composed of two species (g⊥1,2 ≈ 1.97, g||1 = 1.948, and g||2 = 1.941, insert of Fig. 4). The signal obtained for Au/CeO2 was slightly broader because of the presence of a third species (1.98 < g⊥3 < 1.96 and g||3 = 1.937–1.936) (Fig. S3). Its intensity (double integration) was similar for CeO2 and Au/CeO2 samples. Though this signal at g ≈ 1.97 was currently observed on ceria-based samples, its attribution is still under debate. It was tempting to assign it to Ce3+; however, according to the literature, such paramagnetic ion could be observed at very low temperature only (T <20 K) [41, 42]. Some authors proposed that this signal originated from Ce3+ in peculiar geometry/symmetry [43, 44], while other authors proposed that it arises from quasi-free electrons with some orbital mixing with the empty f-orbitals of Ce4+ ions [23, 45].

Fig. 4
figure 4

EPR spectra of CeO2 sample (recorded after 30 min at 77 K) under increasing PO2

Upon addition of O2, the signal at g ≈ 1.97 remained unchanged, which further strengthened the hypothesis of a paramagnetic species trapped in the bulk of the CeO2 particles (no possible dipolar interaction with O2 that could affect the signal). Moreover, a complex new signal of high intensity appears in the 3,150 G < B < 3300 G region, typical of O2 species (Fig. 4). It had almost the same shape for CeO2 and Au/CeO2 (Fig. S3), and the simulation of the signal (Figs. 5 and S4) indicated the presence of five different O2 species in both samples with slightly different proportions (Table 4). The OI's and OII's species have been previously observed on similar materials [4648]. These species were attributed to different adsorption modes of O2 on the surface, OI with EPR-equivalent oxygen atoms (adsorption parallel to the surface), and OII with nonequivalent oxygen atoms (end-on or other asymmetric adsorption).

Fig. 5
figure 5

Experimental (recorded after 30 min at 77 K) and simulated EPR spectra of CeO2 sample, under 1.5 mbar O2

Table 4 Simulated EPR parameters of O2 species adsorbed on CeO2, Au/CeO2, 10CeO2/Al2O3, and Au/10CeO2/Al2O3 under 1.5 and 50 mbar O2

Upon increasing O2 pressure, the shape of the O2 signal was barely affected, but its intensity increased, then decreased when PO2 > 8 mbar and 15 mbar for CeO2 and Au/CeO2, respectively, as attested by the values of the double integration reported in Table 5. The decreasing intensity above a critical O2 pressure probably resulted from dipolar interactions between O2 and nearby O2 molecules [49, 50]. This points out that the O2 signal intensity depends not only on the number of paramagnetic species, but also on this broadening effect, thus comparisons of the intensities are meaningful for the smallest O2 pressures, only. The interesting point regarding these experiments is that the intensity (double integration) of the overall O2 signal is smaller for Au/CeO2 than for CeO2 (Table 4), indicating that no extra O2 species were generated by the presence of gold.

Table 5 Overall intensity (double integration) of O2 species adsorbed on CeO2, Au/CeO2, 10CeO2/Al2O3, and Au/10CeO2/Al2O3 upon increasing O2 pressures

In the case of 10CeO2/Al2O3 and Au/10CeO2/Al2O3, upon O2 addition, the signal of O2 was very different and much more intense at low PO2 (more than five times) (Fig. 6) than that for CeO2 and Au/CeO2 samples (Figs. 4 and S3). Again, the shape of the O2 signal was similar for 10CeO2/Al2O3 and Au/10CeO2/Al2O3 (Fig. S5). It was also much less complex than the former ones (Figs. 4 and S3), and could be accurately simulated considering only one O2 species, thereafter referred as OCA in the literature [31] (Table 4). It has been proposed that the OI and OII species formed on 3-D CeO2 nanoparticles, while the OCA species formed on 2-D platelets of CeO2 [31]. This assignment fits with the fact that CeO2 consists of 3-D particles, and that, as mentioned at the beginning of the section “Results and discussion”, previous EFTEM images of xCeO2/Al2O3 revealed the presence of both 3-D and 2-D (platelets) CeO2 particles on alumina [7]. The intensity of the OCA species strongly decreased and broadened for O2 pressures >1.5 mbar, and almost disappeared at PO2 ≈ 50 and 15 mbar, for 10CeO2/Al2O3 and Au/10CeO2/Al2O3, respectively, which was not observed with CeO2 and Au/CeO2. Such different behavior towards O2 dipolar broadening has been reported earlier [51] on CeO2/SiO2 and CeO2/Al2O3 materials, and the authors assigned it to the difference in O2 accessibility to O2 species, thus locating O2 in the bulk (CeO2/SiO2: little to no broadening effect) or on the surface (CeO2/Al2O3: strong broadening effect) of CeO2. Again, the double integration of the O2 signal was somewhat smaller for Au/10CeO2/Al2O3 than for 10CeO2/Al2O3, whatever the pressure of O2 (Table 5).

Fig. 6
figure 6

EPR spectra of 10CeO2/Al2O3 sample (recorded after 30 min at 77 K) under increasing PO2

To summarize, the comparison of the intensity of EPR signals of O2 species in CeO2 and Au/CeO2 pointed out a lower amount of O2 species in the presence of Au, for which the OSC property and the oxygen isotopic exchange activity was exalted, which ruled out the hypothesis that O2 species would play an active role in the corresponding processes. However, the formation of a noticeable amount of O2 2− species adsorbed on Au/CeO2 and not on CeO2 strengthened the hypothesis of an activation of dioxygen molecule via Au nanoparticles through the formation of a peroxo species. This interpretation matches the conclusion brought by Pal et al. [18] who reported the results obtained by combining photoelectron spectroscopy and computational method, showing that the O–O bond was more activated (more elongated) in the peroxo form than that in the superoxo one, and suggesting that peroxo mode of chemisorptions plays a crucial role in the dioxygen activation.

Conclusion

In this work, we studied the oxygen mobility and the nature of the oxygenated species responsible for the chemisorption of dioxygen on Au/CeO2 and Au/xCeO2/Al2O3 (x = 5 and 10 wt.%) catalysts. We stressed the role of the Au nanoparticles in the step of dioxygen activation. To achieve this goal, oxygen storage capacity measurements and 18O/16O isotopic exchange reaction were undertaken to study the oxygen mobility, while Raman spectroscopy and electron paramagnetic resonance spectroscopies were employed to determine the nature and the location of the adsorbed oxygen species. All the characterizations were performed on catalysts previously activated in conditions as close as possible to the thermal pretreatment used prior to the reactions of oxidation studied in our previous work [7].

Au nanoparticles were shown to exalt both the OSC property of CeO2 and the exchange rate of oxygen between the gas phase and the lattice oxygen atoms of CeO2, whether the Au nanoparticles were supported on CeO2 or on CeO2/Al2O3. Complementary results obtained by modifying the gold loading and the ceria surface indicated that the improvement in the oxygen mobility was dependent on the Au/CeO2 interfacial perimeter. Moreover, the close inspection of the evolutions of the isotopomer partial pressures during the isotopic exchange experiments led to the conclusion that the interfaces between the gold nanoparticles and the ceria crystallites were the preferential location for dioxygen activation via adsorption of binuclear species.

Further characterization by Raman and EPR spectroscopies were performed to determine the nature of the dioxygen species. The study of CeO2 and Au/CeO2 by in situ Raman under oxygen revealed the presence of peroxo species on Au/CeO2 and of additional vacancies probably located at the perimeter of the interface between gold and ceria. EPR revealed the presence of superoxo species on both samples, but their concentration was lower in the presence of gold. The same observation was made for 10CeO2/Al2O3 and Au/10CeO2/Al2O3. As a consequence, the efficiency of Au/CeO2 and Au/xCeO2/Al2O3 catalysts for oxidation reaction could be explained by the activation of dioxygen molecule at the ceria/Au nanoparticles interfacial perimeter involving peroxo species.