Operando NAP-XPS Studies of a Ceria-Supported Pd Catalyst for CO Oxidation

Abstract

Supported Pd/CeO₂ catalytic systems have been widely investigated in the low-temperature oxidation of CO (LTO CO) due to the unique oxygen storage capacity and redox properties of the ceria support, which highly influence the structural, chemical and electronic state of Pd species. Herein, operando near-ambient pressure XPS (NAP-XPS) technique has allowed the study of a conventional Pd/CeO₂ catalyst surface during the CO oxidation reaction under experimental conditions closer to the actual catalytic reaction, unfeasible with other surface science techniques that demand UHV conditions. SEM, HRTEM and XRD analyses of the powder catalyst, prepared by conventional incipient wetness impregnation, reveal uniformly CeO₂-loaded Pd NPs of less than 2 nm size, which generated an increase in oxygen vacancies with concomitant ceria reduction, as indicated by H₂-TPR and Raman measurements. Adsorbed peroxide (O₂²⁻) species on the catalyst surface could also be detected by Raman spectra. Operando NAP-XPS results obtained at the ALBA Synchrotron Light Source revealed two kinds of Pd species under reaction conditions, namely PdO_x and Pd^II ions in a Pd_xCe_1−xO_2−δ solution, the latter one appearing to be crucial for the CO oxidation. By means of a non-destructive depth profile analysis using variable synchrotron excitation energies, the location and the role of these palladium species in the CO oxidation reaction could be clarified: PdO_x was found to prevail on the upper surface layers of the metallic Pd supported NPs under CO, while under reaction mixture it was rapidly depleted from the surface, leaving a greater amount in the subsurface layers (7% vs. 12%, respectively). On the contrary, the Pd_xCe_1−xO_2−δ phase, which was created at the Pd–CeO₂ interface in contact with the gas environment, appeared to be predominant on the surface of the catalyst. Its presence was crucial for CO oxidation evolution, acting as a route through which active oxygen species could be transferred from ceria to Pd species for CO oxidation.

1. Introduction

Ceria or cerium dioxide exhibits very interesting redox properties given its electronic configuration ([Xe] 6s²5d¹4f¹). The ability to easily change between Ce^III and Ce^IV oxidation states and the capacity to accommodate oxygen vacancies on its structure (Equation (1)), known as oxygen storage capacity (OSC) [1], makes ceria a ubiquitous constituent in catalytic systems for different applications. For instance, the oxide can be used as a catalytic support for many chemical reactions, such as alcohol oxidation [2], soot combustion [3] or carbon monoxide oxidation [4]:

CeO₂ ⇄ CeO_2–x + x/2 O₂

(1)

Nevertheless, the most known application of CeO₂-based materials is as a component in automotive three-way catalysts (TWCs) [5,6]. Mounted after an internal combustion engine, the TWC’s role is to remove the pollutants produced by fuel combustion, promoting simultaneously the oxidation of CO and unburnt hydrocarbons, as well as the reduction of nitrogen oxides (NO_x). The catalytic material of the early TWCs consisted of a thermally stable support (doped Al₂O₃), and active metallic phase (Pt, Pd, Rh) and a CeO₂ promoter [7]. Nowadays, the so-called advanced TWCs employ CexZr_1–xO₂ solid solutions instead of pure ceria [8]. Besides this well-established use, ceria is emerging as a unique material, holding great promise for future market breakthroughs, such as a component of solid oxide fuel cells (SOFCs), a solid state electrolyte oxygen type sensor and as an antioxidant agent for biomedical applications [9]. As CO oxidation is a primary function of TWCs containing ceria, it has often been used as a probe reaction over ceria-based materials to evaluate different catalyst formulations and their redox properties. It is accepted that CO oxidation proceeds through the Mars–Van Krevelen mechanism over ceria, that is, lattice oxygen atoms of the ceria surface are transferred to CO, and the vacancies created on the oxide are subsequently filled by gaseous oxygen [10]. The formation of oxygen vacancies is the rate determining step, since the replenishment with gas-phase O₂ is favorable for reduced ceria [11].

Catalyst design usually includes a support with adequate redox properties loaded with a metallic active phase. Palladium is one of the most active metals interacting with the surface of oxides, and it is characterized by a relatively high affinity with oxygen. In fact, the Pd/CeO₂ catalytic system has been extensively studied and employed as a component in multiple catalytic reactions, such as the low-temperature oxidation of CO (LTO CO) [12], the water–gas shift (WGS) reaction [13,14], and CHx oxidation [15], but the main application of this system is as a component of the aforementioned TWCs’ exhaust gas purification systems. It is well known that the catalytic activity of Pd nanoparticles, as occurs with other metals, is closely related inter alia to their size, shape and valence state [16]. Thus, it is of great importance to stabilize Pd nanoparticles under various reaction conditions, preventing variations on the particle size, shape or physicochemical states that could ultimately affect their catalytic properties. Cerium oxide is widely used as a support to load palladium nanoparticles [13,14], considering its oxygen storage capacity and its redox properties, in order to influence the chemical and electronic states of Pd species [5] and stabilize Pd nanoparticles against sintering [17]. For supported Pd catalysts, the influence of the support nature on their catalytic performance is usually described by the interaction between the support and the Pd or PdO species [18]. Many studies proposed the formation of solid solutions of Pd_xCe_1−xO₂ at the surface and subsurface layers of the ceria lattice [12,19], and quantum-chemical calculations have confirmed the formation of these structures in the Pd/CeO₂ systems [20,21]. In another work, Colussi et al. reported the formation of Pd–O–Ce superstructures on the ceria (110) surface [22]. However, some researchers consider that the oxidation state of Pd highly active in LTO CO is related to PdO or PdO₂ nanoparticles confined inside the support oxide matrix [19,23]. Despite extensive investigations, the exact electronic and geometric structure of the Pd/CeO₂ active centers and the role of the ceria support are still ambiguous.

Additionally, although a large number of catalytic studies in surface chemistry have been performed using different surface science techniques requiring ultrahigh vacuum (UHV) conditions [24], it has been found that under experimental conditions closer to the ones used for the actual catalytic reactions, the reaction mechanisms and catalyst architectures can differ from those under UHV conditions [25]. Indeed, some structures and catalytic active sites may only develop under specific reaction conditions, making the application of in situ or operando techniques indispensable to identify and understand the behavior of these working active species. In the last years, such operando techniques have been increasingly applied. Among them, near ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) has become a key tool to study active sites under more realistic conditions, owing to its sensitivity to the surface and sub-surface of catalysts [26].

In the present study, CO oxidation reaction is investigated by NAP-XPS on a polycrystalline Pd/CeO₂ powder catalyst prepared by conventional incipient wetness impregnation. Since contradictory data concerning the nature of Pd–Ce–O species active in the CO oxidation reaction still persist, this work aims to provide a comprehensive investigation of the interaction between Pd and the support surface and determine its structural and electronic state during LTO CO by using NAP-XPS, as well as other characterization techniques.

2. Materials and Methods

2.1. Catalyst Preparation

Pd/CeO₂ catalyst was prepared by incipient wetness impregnation [27] from a palladium nitrate solution (Alfa Aesar, 99.9%) over CeO₂ polycrystals, synthesized following reported methods [28]. Briefly, 35 mL of an aqueous solution of 6.07 g of Ce(NO₃)₃ (99.5%, Alfa Aesar, Ward Hill, MA, US) was electrosprayed over 245 mL of an aqueous solution containing 1.96 g of NaOH (97%, Fisher Scientific, Waltham, MA, US) under vigorous stirring, using an ultrasonic atomizer (Sonozap HTWS30, Melville, NY, US). The resultant suspension was agitated for 30 min at room temperature and was then transferred into a PTFE-lined stainless-steel autoclave (Reactor Chemipress-500, Trallero And Schlee, La Llagosta, Spain), where it was heated to 150 °C for 24 h. After the hydrothermal treatment, the resulting mixture was cooled and separated by centrifugation (10 min at 6000 rpm), washed three times with deionized water and three times with ethanol. The powder was finally dried at 60 °C in air overnight [29]. As already mentioned, once the support was prepared, a palladium nitrate solution was deposited dropwise by incipient wetness impregnation in order to obtain a Pd loading of 4 wt%. After impregnation, the powder was dried overnight at 100 °C and then calcined in air at 500 °C for 5 h (1 °C·min⁻¹) to generate the Pd nanoparticles.

2.2. Catalyst Characterization

Surface morphology was evaluated using scanning electron microscopy (SEM). The images were acquired with a Zeiss Neon40 Crossbeam Station instrument at 5 kV equipped with a field emission source. The sample was prepared by placing a drop of a suspension of the Pd/CeO₂ powder catalyst in methanol on a silicon wafer and allowed to dry. High resolution transmission electron microscopy (HRTEM) images were obtained by using a field emission gun FEI Tecnai F20 microscope equipped with a field emission source at an accelerating voltage of 200 kV, with a point-to-point resolution of 0.19 nm. The sample was prepared by depositing a drop from a methanol suspension of the catalyst on a holey carbon-coated copper grid and allowing them to evaporate. The average particle diameter was calculated from the mean diameter frequency distribution with the formula: $\mathrm{d}={\displaystyle \sum }{n}_i{d}_i/{\displaystyle \sum }{n}_i$, where $n_i$ is the number of particles with particle diameter $d_i$ in a certain range.

Powder X-ray Diffraction (XRD) patterns were recorded with a Siemens D5000 diffractometer using Cu Kα radiation (45 kV, 35 mA) in Bragg–Brentano geometry. The diffraction patterns were recorded in steps of 0.02° and 1 s per step using a 2θ range from 10 to 80°. Crystallite sizes were estimated from the XRD data using the Debye–Scherrer equation.

Micro-Raman spectroscopy was performed using a commercial Renishaw inVia Qontor confocal Raman microscope. The Raman setup consists of a laser (at 532 nm with a nominal 100 mW output power) directed through a microscope (a specially adapted Leica DM2700 M microscope) to the sample, after which the scattered light is collected and directed to a spectrometer with a 2400 lines·mm⁻¹ grating. The laser power was adjusted to 1% of its nominal output power, the exposure time was 0.5 s, and spectra were collected with 50 accumulations.

Temperature programmed reduction with hydrogen (H₂-TPR) was carried out in a quartz U-tube with a Chemstar TPX chemisorption analyzer (Quantachrome Instruments, Boynton Beach, FL, USA) equipped with a Thermal Conductivity Detector (TCD). Initially, 30 mg of sample were pre-treated, heating them from room temperature to 450 °C in argon flow (50 mL·min⁻¹, 10 °C·min⁻¹), over 30 min. The samples were then cooled down to 50 °C under the same atmosphere, and the H₂-TPR experiments were conducted by heating from 50 °C up to 850 °C (10 °C·min⁻¹) under 10% H₂ in Ar (with a total flow of 50 mL·min⁻¹), while monitoring the hydrogen consumption with a TCD detector. The temperature was kept for 30 min at 850 °C.

NAP-XPS was conducted at the NAPP end station of the CIRCE beamline at the ALBA Synchrotron Light Source, following the surface chemical composition under operando conditions at different temperatures, gas environments (H₂, CO and CO + O₂) and pressures. Different excitation photon energies were used to obtain XP spectra of Ce 3d, O 1s, Pd 3d, C 1s and Si 2p from two different sample depths (acquiring photoelectrons with 150 and 450 eV kinetic energies, corresponding approximately to 0.5 and 1.0 nm inelastic mean free paths, respectively). Pd/CeO₂ catalyst powder was pressed with a gold mesh with 0.25 mm of nominal aperture (99.9%, GoodFellow) to form a compact disk and minimize charging effects. NAP-XPS measurements were acquired with a commercial PHOIBOS 150 NAP energy analyzer, equipped with four differentially pumped stages [25]. Hydrogen gas for the preliminary reduction of the catalysts was dosed into the analysis chamber by a leak valve, while the carbon monoxide and the carbon monoxide/oxygen reaction mixture (CO:O₂ = 2:1, 15 mL·min⁻¹ total flow) were introduced by using mass flow controllers and active pumping to control the final pressure (1.5 mbar). The evolution of products was followed by a quadrupole mass spectrometer Microvision Residual Gas Analyzer (MKS Instruments). The sequence of spectra recorded and experiments performed was: (i) sample pre-treatment by reduction under 1 mbar of H₂ from room temperature to 350 °C, (ii) cooling down to 250 °C, H₂ removal and introduction of 10 mL·min⁻¹ of CO at 250 °C and 1 mbar, (iii) CO oxidation at 250 °C by adding 5 mL·min⁻¹ of O₂ to the previous flow of CO at 250 °C (CO oxidation, CO:O₂ = 2:1, total pressure of 1.5 mbar), (iv) CO oxidation at 300 °C, and (v) CO oxidation at 350 °C. The sample was heated using an infrared laser (λ = 808 nm) focused on a stainless-steel plate, on top of which it was mounted. The temperature was monitored with a K-type thermocouple in contact with the samples. Data processing was performed with the CasaXPS program (Casa Software Ltd., Teignmouth, UK). To compensate charging effects, spectra were calibrated considering the C 1s peak at 284.8 eV and the U‴ Ce 3d peak at 916.9 eV. A Shirley-type background was subtracted from each spectrum, and all spectra were fitted using a product of Lorentzian and Gaussian functions, except for the metallic components in the Pd 3d region, which were fitted through an asymmetric line shape. Atomic fractions were calculated using peak areas normalized on the basis of acquisition parameters after background subtraction. Cerium 3d spectra were deconvoluted using six peaks for Ce^IV (V, V″, V‴, U, U″ and U‴), which correspond to three pairs of spin–orbit doublets, and four peaks for Ce^III (V₀, V′, U₀ and U′), corresponding to two doublets, based on the peak positions reported by Mullins et al. [30], where U and V refer to the 3d_3/2 and 3d_5/2 spin–orbit components, respectively.

3. Results and Discussion

3.1. Characterization

The morphology of the CeO₂ powder was analyzed by SEM before the deposition of Pd for the preparation of the Pd/CeO₂ sample. The obtained images showed that ceria polycrystals have an irregular rounded morphology and range from about 5 to 20 nm in size, as illustrated in Figure 1. Palladium deposition modified neither the morphology nor the size distribution of ceria polycrystals. Metallic nanoparticles could not be identified with the backscattered electron signal due to their small size (see Figure 1b).

HRTEM measurements of the Pd/CeO₂ catalyst proved the presence of isolated palladium nanoparticles of less than 2 nm in diameter very well dispersed over ceria polycrystals, as seen in Figure 1c,d. Ceria polycrystals presented a mean size of 7 ± 3 nm, in accordance with SEM results. The Fourier transform images showed spots corresponding to the (111), (200) and (220) planes of the ceria fcc structure at 3.1, 2.7 and 1.9 Å, respectively. As expected, the XRD pattern of CeO₂ polycrystals (Figure 2) showed the characteristic reflections of the CeO₂ fcc phase [23,31]. After Pd deposition, a small peak corresponding to Pd (111) could be discerned, while no sign of PdO phase, easily generated in the calcination process under air, could be identified. Ceria nanoparticles presented an average crystallite size of 14 nm, as estimated from the Debye–Scherrer formula. These results are in good accordance with SEM and HRTEM images.

Figure 3 shows the Raman spectra of the as-synthesized samples collected at room temperature. Generally, four peaks are observed in the Raman spectra of bare CeO₂: the main peak at 462 cm⁻¹ (F_2g) is ascribed to the symmetric stretch mode of oxygen ions in the [CeO₈] cubic subunit characteristic of the ceria fluorite lattice structure, and three weak peaks discerned at approximately 255, 595 and 1180 cm⁻¹. These bands are due to the second-order transverse acoustic (2TA) mode, the defect-induced (D) vibrational mode and the second-order longitudinal optical (2LO) mode, respectively [32]. The disorder-associated D Raman band gives information on ceria intrinsic defect sites, so the more intense it appears, the higher amount of defect sites will have the analyzed ceria. Additionally, an asymmetric band at 832 cm⁻¹ is observed, which corresponds to the O–O stretching vibrations of adsorbed molecular oxygen, suggesting the formation of active peroxide (O₂²⁻) species on its surface [32]. The Raman band at ~ 1075 cm⁻¹ on ceria polycrystals is due to residue nitrate species left from the synthesis process [32] (see band denoted by asterisk in Figure 3b), missing in the Pd/CeO₂ sample because of the involved calcination step. After loading Pd, the disorder-associated band broadens and becomes more intense, and it can be deconvoluted into new defect-induced peaks. A detailed view of this region is given in Figure 3b. As mentioned, the band at 595 cm⁻¹, which from now onwards will be referred to as D₁, is linked to “intrinsic” defects and originates from oxygen vacancies and replacement of Ce^IV atoms by reduced Ce^III cations [33], although other authors have assigned it to Frenkel-type anion defects, i.e., vacancies created when an oxygen atom moves from its lattice position into an octahedral interstitial position [32,34]. This vibrational feature has been also ascribed to defect spaces, including a dopant cation in 8-fold coordination O²⁻ without any O²⁻ vacancy [35,36], but it will be considered anyway “intrinsic”, i.e., already present in pure ceria structure. In addition to the mentioned D₁ band, an intense D₂ component near 540 cm⁻¹ can be detected, attributed to defect spaces involving an oxygen vacancy coupled with Ce^III or other aliovalent dopant cations introduced in the ceria lattice [37,38,39]. Lastly, there is a third D₃ component, centered at 630 cm⁻¹ but weaker than the other two, which is probably associated to “extrinsic” defects caused by the addition of dopants. It has been previously assigned to the presence of MO₈ units that do not contain oxygen vacancies, where M is a foreign cation [36,38]. According to the literature, an estimation of the number of defects in the catalyst can be provided through the intensity ratio between the D and the F_2g bands. Indeed, various reports claim that the tendency of the catalytic activity for CO and other gas phase oxidation reactions can be well evidenced based on the D/F_2g ratio values [40,41]. Although the disorder-associated band clearly broadens and becomes more intense after the Pd loading, D/F_2g ratios have been calculated for both as-synthesized samples from their corresponding spectrum, increasing from 0.10 to 0.75 (variation factor of 7.5) upon palladium loading. It is worth noting the presence of an additional peak near 650 cm⁻¹ for the Pd/CeO₂ catalyst, which corresponds to the B_1g vibrational mode of the Pd–O bond [36,42]. The presence of the latter peak and the new D₃ component could denote a Pd–ceria interaction, possibly entailing the incorporation of some Pd cations into the uppermost ceria layers, without dissolving into the bulk, a fact that could explain the absence of a PdO phase in the XRD pattern. Both Raman spectra show a peak at 350 cm⁻¹ and a shoulder near 420 cm⁻¹ on the left side of the fluorite band. Although no clear assignment could be found in the literature for these signals, they have been observed for ceria nanoparticles [32,43] and could be attributed to distortion in the lattice [34,38]. The presence of all the mentioned bands, some of them not clearly resolved, represents the heterogeneous environment where a specific defect type can be found.

Summarizing, the incorporation of Pd to the ceria system induces an increase in the amount of the intrinsic defect sites associated with both oxygen vacancies and reduced Ce^III atoms, reflected by an intense growth of the D₁ component and the appearance of an intense D₂ band. The presence of the new D₃ component and a band near 650 cm⁻¹ suggest a stronger Pd–CeO₂ interaction, possibly involving the incorporation of some Pd cations into the first layers of ceria structure. A variation of lattice parameters caused by Pd incorporation into bulk ceria (with a different atomic radius value) would have been observable in the ceria XRD pattern, which showed no variation upon Pd loading and thus supported the suggested assumptions.

H₂-TPR was used to investigate the reducibility of ceria and Pd/CeO₂ catalysts and the influence of the incorporation of palladium on its redox behavior, and results are reported in Figure 4. It is widely known that reduction of CeO₂ powders occurs at temperatures higher than 400–500 °C [44] and via a stepwise mechanism so that their TPR profiles show a bimodal shape, with a first peak at low temperature attributed to the reduction of the most outer layers of Ce^IV (surface reduction) and a large peak at high temperature characteristic of the reduction of the inner Ce^IV layers (bulk reduction) [45]. As seen in Figure 4, ceria polycrystals show a first wide peak between 400–600 °C, characteristic of surface reduction, and a major peak around 700 °C, attributed to bulk reduction. The incorporation of Pd on CeO₂ nanocrystals promotes significantly the surface reduction of ceria by spillover from Pd species onto the support, which is reflected with a decrease in this reduction temperature from ca. 450 °C to ca. 370 °C.

Interestingly, an additional and negative peak appears near 80 °C, which is commonly assigned to decomposition of the Pd hydride phase that is formed during H₂-TPR at lower temperatures, as already reported [46].

3.2. Operando NAP-XPS

Near ambient pressure X-ray photoelectron spectroscopy was performed to characterize the chemical composition of the surface of both samples during CO oxidation under operando conditions by recording the high-resolution spectra of Ce 3d, O 1s, Pd 3d and C 1s photoelectrons. The spectrum corresponding to the Ce 3d core level region is rather complex and can be fitted with 10 peaks, with two and three doublets assigned to Ce^III and Ce^IV species, respectively. Figure 5a shows a sequence of Ce 3d spectra recorded over Pd/CeO₂ under different experimental conditions. All the contributions are consistent with other literature reports [30,47].

As expected, heating the Pd/CeO₂ catalyst from room temperature to 350 °C under an H₂ rich atmosphere increased the catalyst Ce^III/Ce ratio up to a value of 55%, while the introduction of CO (1 mbar) caused a slight decrease in this ratio, down to around 50%. This effect can be observed through a small drop in intensity of the Ce³⁺ V₀, V′, U₀, and U′ peaks, with binding energies of 880.7, 885.3, 899 and 903.6 eV, respectively. However, the most significant change of the Ce^III amount came when introducing O₂ in the subsequent reaction steps (1.5 mbar of CO:O₂ = 2:1), which caused a decrease in the Ce^III/Ce ratio down to 25.5% (see Figure 6a) due to the exceptional oxygen storage capacity of ceria, able to capture and accommodate oxygen atoms on its structure in an oxidant environment.

O 1s spectra can be deconvoluted into multiple components due to the presence of oxygen atoms in different coordination environments [48]. Although the interpretation of the different contributions is complex due to the proximity between the components present, changes in the cerium oxidation state are consistent enough comparing the Ce 3d (Figure 5a) with the O 1s spectra (Figure 5b) recorded under the same experimental conditions, as illustrated in Figure 6a. For the impregnated catalyst, the component at lower binding energy (528.4 eV) corresponds to the oxygen atoms bonded to Ce^IV, while the contribution appearing approximately at 529.9 eV is here related to three-fold oxygen (oxygen atoms bonded to Ce^III or near to oxygen vacancies). Finally, the shoulder at 532 eV can be attributed to surface hydroxyl groups, likely due to the interaction with H₂ gas during the pre-treatment step. It is worth mentioning that the Pd 3p_3/2 contribution, centered at 532.4 and 534.2 eV for metallic palladium and PdO, respectively [49,50], significantly overlaps with O 1s bands. Nevertheless, its signal ought to be much less intense with respect to the corresponding Pd 3d component, a fact that supports the rejection of this such possible contribution. The injection of CO into the analysis chamber resulted in the appearance of its characteristic gas phase band between 536–537 eV [51], emphasizing the powerful capabilities of the NAP-XPS technique that, in addition to the sample surface composition, can also detect the gas molecules surrounding the sample. It is worth mentioning that the position of gas phase species can slightly vary depending on the work function of the sample surface. For this reason, CO contribution appears at lower B_E upon O₂ injection. Two additional peaks at 537.4 and 538.5 eV appeared when dosing oxygen (denoted by red asterisks in Figure 5a), and correspond to contributions of O₂ from the gas phase [51]. CO species can also be confirmed through C 1s spectra (see spectrum ii from Figure 7), where a new feature around 291 eV was observed in addition to the C 1s signal at 284.8 eV, typically assigned to C–C bonds of amorphous carbon contamination [52]. Another contribution could be discerned at 285.6 eV, most probably indicating an interaction of CO molecules with Pd, as will be discussed thereafter.

Similarly, after the exposure to O₂, a new band at 534.5 eV demonstrated the presence of gaseous CO₂ generated through CO oxidation, as displayed in Figure 5b. Indeed, carbon dioxide was immediately formed as a reaction product and its presence could be again proved through both O 1s and C 1s spectra. As observed in Figure 7 (spectrum iii), O₂ oxidized all the existing carbonaceous species, and the characteristic peak of adventitious carbon at 284.8 eV disappeared while an intense band near 293 eV was simultaneously detected next to the CO signal, which is assigned to carbon dioxide gas phase. The presence of carbon dioxide was also confirmed by mass spectroscopy. Figure 6b shows the evolution of O₂, CO and CO₂ partial pressures inside the analysis chamber, recorded by means of a quadrupole mass spectrometer, located in the second stage of the differential pumping system of the NAP-XPS analyzer, which probes the gases flowing from the proximity of the sample surface towards the detector. The shaded areas depict the three different atmospheres (H₂, CO and CO + O₂) generated for the sequence of experiments studied, and a clear variation took place upon O₂ injection, as CO₂ partial pressure rapidly increased at the expense of CO and O₂ consumption by the oxidation reaction.

Figure 8 shows the Pd 3d spectra of the Pd/CeO₂ catalyst, which displayed interesting variations of its oxidation states during the different reaction steps. Under H₂ atmosphere, palladium appeared completely reduced, with its Pd 3d_5/2 and Pd 3d_3/2 signals centered at 335.5 and 341 eV, respectively. The position of these signals did not substantially vary upon changing the initial atmosphere for CO, but the asymmetric line shape typical of metallic-like Pd was slightly lost, indicating the presence of an intermediate-B_E component at approximately 336 eV. According to the literature, the latter value is centered between the typical positions of bulk Pd metal (B_E(Pd3d_5/2) = 335.2 ± 0.2 eV) and bulk PdO (B_E(Pd3d_5/2) = 336.8 ± 0.2 eV) [53], and it cannot be attributed to the PdO phase since the characteristic shake-up satellites of the oxide near 341 and 346 eV are not observed in the XP spectra [54]. This component could be assigned to chemisorbed CO molecules but, according to the literature, it could also be attributed to small semi-oxidized palladium clusters, denoting them as PdO_x [19,55,56], that could be formed at the upper atomic layers of Pd-based catalysts by CO dissociation. Indeed, the shift of the photoelectron lines towards higher binding energies (with respect to the typical value of the palladium metal state, B_E(Pd3d_5/2) = 334.8 − 335.4 eV) can be explained by the size effects during the photoionization process of metal clusters. When the size of metal clusters deposited on insulating supports is reduced, the extra-atomic relaxation and charge conservation in the final state is not as effective as for larger particles or bulk samples. In other words, Pd atoms of the cluster boundaries are subjected to an intense polarizing effect from ceria lattice oxygen atoms, which can lead to the transfer of some electron density to the support and, consequently, to the formation of partially charged palladium atoms with an oxidation state value between Pd⁰ and Pd^II [19,55]. The ratio between boundary and the total number of Pd atoms in the cluster raises as its dimensions decrease, so that small Pd clusters probably include a greater number of Pd–O bonds than Pd–Pd ones, modifying the chemical shift of Pd photoelectron lines. For this reason, this state has been referred to as PdO_x, considering both possibilities, i.e., the interaction with adsorbed CO molecules on Pd [57] and the formation of clusters due to CO dissociation reported in the literature, indicating an oxidation state between Pd⁰ and Pd^II. As the component centered at 336 eV appeared with the injection of CO, it could be attributed to the formation of palladium carbide (PdC) surface species [58,59,60], also reported under CO environments by dissociation of part of the CO molecules adsorbed on surface metallic Pd and transient storage of atomic carbon within the palladium nanoparticles, thus generating PdC [59,60]. Nevertheless, no equivalent PdC species peak could be identified in the C 1s core-level spectrum obtained in CO atmosphere (see spectrum ii of Figure 7), usually centered at 283.7 eV. As observed in spectrum ii of Figure 7, a peak at 285.5 eV indicated the interaction of CO with the catalyst surface, attributed to CO molecules adsorbed in hollow/bridge sites on the metallic Pd surface [61,62]. Arrows in Figure 7 indicate two other weak features around 287.4 and 289 eV in C 1s spectra, attributed to oxygen-containing species residing on the surface, specifically assigned to C=O and O=C–O groups [63]. As observed in Figure 8b, the fraction of the PdO_x phase was reduced by more than half (from a value of 74% to 28%) when O₂ was added to CO at 250 °C. Simultaneously, the introduction of O₂ gave rise to the formation of a different Pd species with intense bands at 338.1 and 343.4 eV (Pd3d_5/2 and Pd3d_3/2 respectively, see Figure 8). This Pd3d_5/2 value exceeds by far the typical B_E value of PdO (336.8 ± 0.2 eV), and it could be attributed to the formation of the most oxidized palladium state (IV) [64] or fine PdO nanoparticles with low-coordinated Pd^II ions [65]. Nevertheless, the B_E(Pd3d_5/2) values for all PdO species are in the range of 336.0–337.2 eV and PdO₂ compounds are known to be unstable powerful oxidants [65,66]. Therefore, this increased B_E(Pd3d_5/2) value can be explained by the presence of Pd^II cations in an environment different from the PdO crystalline matrix, or palladium atoms with a chemical state of Pd^δ+ (2 < δ ≤ 4). Actually, Pd/CeO₂ catalysts habitually exhibit such an elevated value of B_E(Pd3d_5/2) [19,23,24,67,68], and multiple studies have attributed it to Pd^II cations in the matrix of a Pd_xCe_1−xO_2−δ solution at the surface layers of the ceria lattice, possibly coordinated to four oxygen ions to form a near square-planar geometry [22,23,24,42,56]. Thus, the positive shift of the Pd3d_5/2 line is ascribed to a decrease in Pd–O distances generated in the Pd_xCe_1−xO_2−δ solution [24,42].

Interestingly, CO oxidation took place with the appearance of this active Pd_xCe_1−xO_2−δ phase, which in turn was generated at the expense of the PdO_x species upon dosing gaseous oxygen at 250 °C. Additionally, an increase in the temperature under the reaction mixture led to a gradual decrease in the PdO_x amount and a simultaneous rise in the Pd_xCe_1–xO_2–δ solution phase, indicated by an intense growth of the component at 338.1 eV (Figure 8a). It is not clear whether the PdO_x phase contributes to the CO oxidation nor to what extent it does so, but three Pd phases were observed through the Pd 3d spectra of the catalyst under reaction mixture, most probably due to being the Pd_xCe_1−xO_2−δ phase, the active form for Pd. A non-destructive depth profile analysis using variable synchrotron excitation energies was performed to elucidate the location and, consequently, the nature and role of these palladium phases in the CO oxidation reaction, and the corresponding Pd 3d spectra are displayed in Figure 9a. Indeed, the use of a lower photon energy, i.e., 485 eV, allows probing of the outermost layers of the sample, since the spectrum obtained at this energy includes Pd 3d photoelectrons with lower kinetic energy, excited from a smaller volume of the sample (more surface-focused) than when the latter is irradiated with 785 eV energy. As observed in Figure 9, the PdO_x component at 336 eV appeared significantly more intense under CO atmosphere (79 vs. 73 %, respectively), suggesting a predominance of such species on the upper surface layers of the metallic Pd supported nanoparticles. The amount of PdO_x phase gradually decreased under reaction mixture and increasing temperature, but at 350 °C PdO_x species were more abundant in the subsurface than in the uppermost surface layers of the catalyst. From such behavior it could be deduced that, although the active form of Pd is the surface Pd_xCe_1−xO_2−δ phase, PdO_x species also play a role on CO oxidation evolution and its consumption (especially at the surface of the catalyst, in contact with the gaseous reactants) is directly related to the appearance and stabilization of the active Pd_xCe_1−xO_2−δ phase. Nevertheless, CO₂ formation has been mainly attributed to the generation of the active Pd_xCe_1−xO_2−δ phase [56,69], which clearly appeared at the Pd–CeO₂ interface in contact with the environment, that is, at the Pd nanoparticle boundaries exposed to both ceria support and gas environment. As observed in Figure 9b, the amount of Pd_xCe_1−xO_2−δ solution phase exceeded the amount of remaining metallic palladium (49 vs. 43%, respectively) under an atmosphere of CO:O₂ = 2:1 at 350 °C, as opposed to the subsurface of the catalyst (see Figure 8b), where Pd⁰ species still prevailed over the other oxidation states (48 vs. 40%, respectively).

As already stated, CO oxidation proceeds through the Mars–van Krevelen mechanism over reducible oxide supported metal NP catalysts [10], where lattice oxygen atoms of the ceria surface are transferred to CO_ads, and gaseous oxygen subsequently fills the created vacancies on the oxide. From an electron transfer point of view, CO oxidation process is based on the electron transfer from CO_ads to the catalyst and then to O₂, but the contribution of the palladium phase cannot be ignored in such supported catalytic system. An electron transfer model which correlates electronic Pd–CeO₂ interactions and the Mars–van Krevelen mechanism in CO oxidation was confirmed by Zhao et al. by means of density functional theory calculations [70]. As proposed on Tang’s work [71], the electron transfer occurring at the Pd/CeO₂ interface starts with the creation of surface active oxygen species when the reduced Ce^III is exposed to molecular O₂ and transfers electrons to it, also generating Ce^IV species (Equation (2)). Those active oxygen species then spillover onto the CO molecules adsorbed on Pd^II active sites to produce CO₂ (Equation (3)) and an oxygen vacancy associated with two electrons left, which reduce palladium to its metallic state. Those electrons are eventually transferred from Pd⁰ to Ce^IV species to obtain again active Pd^II and Ce^III (Pd⁰ + 2Ce^IV → Pd^II + 2Ce^III), and thus close the electron transfer cycle from CO_ads to the catalyst and ultimately to O₂, as illustrated in the scheme of Figure 10.

$${\mathrm{O}}_2\to 2{\mathrm{O}}_{\mathrm{ads}}$$

(2)

$${\mathrm{CO}}_{\mathrm{ads}}+{\mathrm{O}}_{\mathrm{ads}}\to {\mathrm{CO}}_2\uparrow$$

(3)

All in all, our results indicate that a PdO_x phase is formed in the presence of CO, which could be somehow related to the appearance of the active Pd_xCe_1−xO_2−δ phase upon introducing O₂. The obtained in situ data should be beneficial and taken into consideration for a full theoretical understanding of the CO oxidation over oxide-supported palladium nanoparticles. The comparison of the presented results with operando NAP-XPS measurements of the Pd/CeO₂ system in other conditions, such as under O₂ at lower temperatures, followed by an atmosphere of CO and, subsequently, a mixture of gaseous CO + O₂, would be certainly helpful to understand both the surface and the subsurface dynamics and thus further elucidate the mechanism of CO oxidation on the Pd/CeO₂ system.

4. Conclusions

This work examines the CO oxidation reaction by operando NAP-XPS on a Pd/CeO₂ powder catalyst prepared by conventional incipient wetness impregnation. SEM, HRTEM and XRD analyses revealed uniformly CeO₂-loaded Pd nanoparticles of less than 2 nm, which generated an increase in oxygen vacancies with concomitant ceria reduction, as indicated by H₂-TPR and Raman measurements. The presence of adsorbed peroxide (O₂²⁻) species on the catalyst surface could also be demonstrated by Raman spectra. Operando NAP-XPS measurements revealed two kinds of surface palladium species under reaction conditions, consisting of electropositive Pd atoms forming a PdO_x phase located at the upper surface layers of the metallic Pd supported NPs, and Pd^II ions interacting with the surface ceria fluorite structure as a Pd_xCe_1–xO_2–δ solution, created at the Pd–CeO₂ interface in contact with the gas environment. The formation of PdO_x took place under a CO atmosphere, most probably by CO dissociation on surface metallic Pd atoms, and the surface Pd_xCe_1−xO_2−δ phase was instantly generated upon injecting gaseous O₂ in the CO environment, becoming the predominant surface species (49.2% at 485 eV) at 350 °C under reaction mixture. The latter species appeared to be very active, and thus crucial, for CO oxidation, acting as a route through which active oxygen species could be transferred from ceria to Pd species for CO oxidation. This work demonstrates the importance of the surface and subsurface chemistry in oxidation processes by operando NAP-XPS measurements.