Interaction of Hydrogen with Ceria: Hydroxylation, Reduction, and Hydride Formation on the Surface and in the Bulk.

The study reports the first attempt to address the interplay between surface and bulk in hydride formation in ceria (CeO2 ) by combining experiment, using surface sensitive and bulk sensitive spectroscopic techniques on the two sample systems, i.e., CeO2 (111) thin films and CeO2 powders, and theoretical calculations of CeO2 (111) surfaces with oxygen vacancies (Ov ) at the surface and in the bulk. We show that, on a stoichiometric CeO2 (111) surface, H2 dissociates and forms surface hydroxyls (OH). On the pre-reduced CeO2-x samples, both films and powders, hydroxyls and hydrides (Ce-H) are formed on the surface as well as in the bulk, accompanied by the Ce3+ ↔ Ce4+ redox reaction. As the Ov concentration increases, hydroxyl is destabilized and hydride becomes more stable. Surface hydroxyl is more stable than bulk hydroxyl, whereas bulk hydride is more stable than surface hydride. The surface hydride formation is the kinetically favorable process at relatively low temperatures, and the resulting surface hydride may diffuse into the bulk region and be stabilized therein. At higher temperatures, surface hydroxyls can react to produce water and create additional oxygen vacancies, increasing its concentration, which controls the H2 /CeO2 interaction. The results demonstrate a large diversity of reaction pathways, which have to be taken into account for better understanding of reactivity of ceria-based catalysts in a hydrogen-rich atmosphere.

Introduction

Interaction of hydrogen with metal oxides is intimately involved in many important processes, from H2 storage to thermal, photo‐ and electro‐catalysis. It exhibits complex behavior, including homolytic dissociation at two oxygen sites, thermodynamically favored for readily reducible oxides leading to two hydroxyl (OH) groups and the concomitant reduction of two surface metal cations, and heterolytic dissociation at metal and oxygen sites primarily occurring on oxides difficult to reduce. The process may involve the migration of H atoms into the sub‐surface region and the bulk to form hydroxyl and hydride (H−) species therein.

Beyond this, the interaction of H2 with ceria (CeO2) has recently received much attention because ceria, widely used as oxidation catalysts, demonstrate the potential as a selective catalyst for alkyne semi‐hydrogenation reactions. It is assumed that H2 undergoes dissociation forming two surface OH groups. The resulting OH groups react to produce water at elevated temperatures, creating oxygen vacancies (Ov). Density functional theory (DFT) calculations showed that the homolytic dissociation of H2 at two oxygen sites is thermodynamically favored on fully oxidized, stoichiometric CeO2(111) surface. However, the calculations also suggested a heterolytic H2 dissociation to be kinetically favored on this surface to form a Ce−H and a OH species. The resulting Ce−H species further evolve to the thermodynamically more stable OH species, accompanied by the reduction of Ce4+ into Ce3+.

Oxygen vacancies on a ceria surface have been shown, both theoretically and experimentally, to strongly affect the H2–CeO2 interaction. The oxygen vacancies can even be formed in situ during the reaction, e.g., by desorption of water via recombination of surface OHs. Theoretical calculations indicated that the stability of Ce−H species resulting from heterolytic H2 dissociation is enhanced on reduced, i.e., Ov‐containing ceria surfaces. Experimentally, surface oxygen vacancies on CeO2 were observed to affect the reactivity of surface hydroxyls by suppressing the water production pathway and promoting the H2 production pathway. In addition, the processes may be influenced by oxygen vacancy diffusion, as water formation and the accompanied oxygen vacancy formation happen on the surface. According to DFT calculations, both, vacancy formation and diffusion, are rather independent of surface orientation. Using inelastic neutron scattering spectroscopy, Wu et al. first reported the direct spectroscopic evidence for the formation of Ce−H species upon H2 dissociation over ceria powder naturally containing oxygen vacancies. Werner et al. observed that H2 dissociation on a fully oxidized CeO2(111) thin film surface forms hydroxyl species, primarily located at the surface, while on the reduced CeO2−(111) thin films Ce−H species were also found in appreciable quantities below the surface indicating that the absorption of hydrogen leads to the formation of Ov‐stabilized hydride species in the bulk. Furthermore, H2 adsorption at room temperature was shown to lead to the oxidation of Ce3+ in both CeO2−(111) thin films and CeO2− powders to Ce4+, demonstrating the formation of more Ce−H hydride species at the Ce3+‐Ov sites than OH species at the O sites. Very recently, Schweke et al. using thermal gravimetric analysis and thermal desorption spectroscopy of ceria powders studied the influence of temperature, hydrogen partial pressure, and oxygen vacancies on the H2–CeO2 interaction. The presence of oxygen vacancies was found to lead to the penetration of hydrogen into the sub‐surface and bulk, possibly as hydride species.

The likely net chemical reactions between H2 and ceria include:

homolytic dissociation to OH [Eq. 1]:

oxygen vacancy formation [Eq. 2]:

homolytic dissociation to Ce−H [Eq. 3]:

heterolytic dissociation [Eq. 4]:

It is clear that the H2–CeO2 interaction is complex and quite sensitive to the reaction conditions. Combined studies on model systems with different levels of complexity have been demonstrated to be an effective strategy towards a fundamental understanding of complex systems. Herein, we report a comprehensive experimental study of H2 interaction with well‐defined CeO2(111) thin films and CeO2 powders in wide range of temperatures and pressures, combined with DFT calculations of CeO2(111) surfaces with oxygen vacancies at the surface and in the bulk. We provide spectroscopic evidence for the interplay between hydroxyl and hydride formation within the bulk, and connect those results to the evidence presented for the corresponding species on the surface and in the near surface region, allowing us to reach a consistent picture.

Results and Discussion

Ceria thin films

We first address experimental results obtained on the fully stoichiometric CeO2(111) thin films using infrared reflection absorption spectroscopy (IRAS) and X‐ray photoelectron spectroscopy (XPS). Figure 1 a shows a series of IRAS spectra measured in 10 mbar of D2 at several temperatures in the range of 300–600 K. No reaction occurs at 300 K. On the sample exposed to D2 at 400 K (henceforth denoted as “CeO2‐400”), D2 dissociates forming OD species identified by the ν(OD) band at about 2700 cm−1 depending on exposure (see Figure S1 in Supporting Information). Upon cooling to 235 K in D2 atmosphere, the band increases in intensity and shifts to the higher wavenumbers. Interestingly, this band disappears after pumping D2 out. However, it only decreases in intensity on the “CeO2‐500” sample while the intensity remains on the “CeO2‐600” sample, thus indicating different stability of surface ODs formed at different adsorption temperatures. Concomitantly, the ν(OH) bands at 3660 and 3665 cm−1 and additional bands at 1295 and 1265 cm−1 emerge, which are assigned to adventitious adsorption of residual H2O and CO2 from the vacuum background.

Figure 1

a) IRAS spectra measured on a CeO2(111) film in 10 mbar of D2 at the indicated temperature (i); after sample cooling to 235 K (ii); and after subsequent pumping of the D2 gas phase and heating to 300 K (iii). ν(OH), ν(OD), and νsym(CO2) regions are shown. b) Ce 3d and O 1s XPS spectra measured at grazing emission (60° off‐normal) for CeO2(111) films as grown and after D2 exposure at three different temperatures as indicated. The spectral deconvolution is highlighted.

To examine whether the reaction with hydrogen induced changes in the electronic structure, we measured XPS spectra immediately after the IRAS measurements. The spectra of the Ce 3d and O 1s core levels were recorded both at normal and grazing emissions, the latter are only shown in Figure 1 b. The complex Ce 3d spectra in CeO2 are usually rationalized in terms of the spin orbit pairs of the Ce3d94f0O2p6, Ce3d94f1O2p5 and Ce3d94f2O2p4 final states, labeled in the Figure as 4f′, 4f′′ and 4f′′′, respectively. If present, Ce3+ species manifest themselves by additional signals, which correspond to the 4f′′ and 4f′′′final states. Since the Ce4+ and Ce3+ states overlap, the degree of ceria reduction can only be determined after a relatively complex deconvolution procedure and background subtraction.

The fully stoichiometric CeO2(111) films exhibited a Ce3+/(Ce3++Ce4+) signal ratio (henceforth, Ce3+ concentration) of about 0.01. After reaction with D2 at 300 and 400 K, no additional Ce3+ species were detected, suggesting that surface reduction does not occur at low temperatures, in full agreement with the IRAS results. In contrast, considerable amounts of Ce3+ are observed after reactions at 500 and 600 K, indicating progressive ceria reduction at high temperatures. A well‐resolved O1s signal at the higher binding energy (BE) side of the main peak at 529.4 eV is assigned to OH(OD) and carbonates‐like species detected by IRAS (Figure 1 a).

It is important to note that, upon increasing the D2 exposure temperature from 500 to 600 K, the intensity of the adsorbates‐related O 1s signal, normalized to the total O 1s signal, only slightly increases, from 0.18 (0.09) to 0.22 (0.1) measured at grazing (normal) emission, whereas the Ce3+ concentration measured from analysis of the Ce 3d region increases substantially, from 0.28 (0.16) to 0.55 (0.42). This finding indicates that, after the reaction at 600 K, considerable amounts of oxygen vacancies are located not only at the surface (where they would be easily filled by residual water as shown by IRAS), but also in the sub‐surface region.

In the next set of experiments, we addressed interaction of D2 with reduced, i.e., CeO2−(111) film surfaces. Figure 2 a summarizes the IRAS results of the exposure experiments at different temperatures. Note that, in contrast to the CeO2(111) surface, the reduced surface may react with traces of water in the UHV background even before the introduction of D2, thus forming OH species as shown in spectra marked by (*) for three different temperatures. The corresponding XPS results are displayed in Figure 2 b.

Figure 2

a) IRAS spectra recorded on reduced CeO2−(111) films in 10 mbar of D2 at the indicated temperature (i); after sample cooling to 235 K in D2 (ii); and after subsequent pumping out of the D2 gas phase and heating to 300 K (iii). Spectra recorded at 300 K prior to the D2 exposure are marked by (*). b) Ce 3d and O 1s XPS spectra measured at grazing emission (60° off‐normal) for reduced CeO2−(111) films as grown and after D2 exposure at two different temperatures indicated. The spectral deconvolution is highlighted.

The results of experiments with D2 at 300 K nicely reproduce our previous results obtained with H2. No surface ODs are observed upon D2 exposure, whereas XPS (not shown here) revealed a strong electronic effect: Almost no Ce3+ remained at the surface after D2 treatment indicating the formation of Ce4+‐D species.

At 400 K, D2 dissociates giving rise to a weak ν(OD) band at 2675 cm−1. Upon cooling in D2, the band gains in intensity together with the OH band (from residual water). After pumping the D2 gas phase out and then heating the sample to 300 K, the OH band increases at the expense of the OD band. However, the measured frequency ratio (ν(OH)/ν(OD) = 1.357) suggests that OH and OD species are identical in nature, thus occupying the same adsorption sites. The interplay between intensities of OH and OD bands can be explained by the H‐D exchange reaction and/or by competition between H2O and D2 for adsorption sites.

Again, after reaction at 400 K, the Ce3+ concentration is dramatically decreased (Figure 2 b) due to the formation of hydride species. Small amounts of Ce3+ originate from surface hydroxyl species, detected by IRAS (Figure 2 a) and O1s XPS. Since the integral intensity of the O1s signal remains unchanged after reaction, the observed oxidation of ceria cannot be explained by migration, if any, of lattice O from deeper layers to the surface. Therefore, the interaction of hydrogen with reduced ceria surfaces at 300 and 400 K primarily results in hydride species.

The same experiment at 500 K, however, revealed a different picture. While also here the adsorption of residual H2O gives rise to a ν(OH) band at 3635 cm−1, surface hydroxylation through D2 dissociation dominates the spectra: A strong ν(OD) band appears at 2693 cm−1. Upon subsequent cooling and pumping, a weak band emerges around 2600 cm−1, which falls in the range of frequencies characteristic for H‐bonded hydroxyls, and additional bands appear in both the ν(OH) and ν(OD) regions, i.e., at 3629 and 2677 cm−1, respectively. These latter values closely resemble those observed in experiments at 400 K, and probably develop during the reaction while sample cooling in D2 atmosphere.

In contrast to the experiments at 300 and 400 K, comparison of the Ce3d spectra before and after reaction with D2 at 500 K revealed similar concentration of Ce3+. However, a significant portion of Ce3+ species may originate from a high density of surface hydroxyls, which obscure XPS detection of hydride species. It may also well be that H(D) atoms migrate into the deeper layers of ceria at 500 K and thus become “invisible” to surface sensitive XPS. Both the Ce3d and O1s spectra after reaction at 500 K look virtually identical to those obtained during the same experiment on the stoichiometric CeO2(111) film (Figure 1 b). Even IRAS spectra measured on CeO2 and CeO2− films at 500 K in D2 atmosphere display almost identical intensity and position of ν(OD) bands (Figure 3 a). Based on these results, one concludes that the initial degree of surface reduction does not play a critical role in the D2–CeO2 interaction at elevated temperatures, since the surface reduction and oxygen vacancy formation occurs in situ on the initially oxidized ceria surface via recombinative desorption of surface hydroxyls formed by hydrogen dissociation.

Figure 3

a) Comparison of IRAS spectra measured on CeO2(111) and CeO2−(111) films in 10 mbar of D2 at 400 K and 500 K. b) Schematic representation of D species formed on oxidized (CeO2) and pre‐reduced (CeO2−) surfaces.

It is interesting that the position of the OD bands observed on CeO2 and CeO2− films at 400 K show considerably different frequencies, i.e., 2696 and 2675 cm−1, respectively (see Figure 3 a). To rationalize this result, we refer to DFT calculations showing that the frequency depends on the local degree of reduction surrounding the surface hydroxyl, i.e., on the number of Ce3+ ions among the three Ce ions the hydroxyl is coordinated to: The higher the Ce3+ ion content, the higher the frequency. Therefore, the OD groups formed on the fully oxidized CeO2 surface (2696 cm−1) are coordinated to more reduced Ce ions than the OD groups formed on the pre‐reduced CeO2− surface (2675 cm−1) (see Figure 3 b for a schematic representation). Even though this might be counterintuitive, the effect can be explained if we recall the XPS results (Figure 2 b) showing that the D2 interaction with the reduced CeO2− surface at 400 K results in hydride species and the oxidation of Ce3+ to Ce4+. Therefore, the low‐frequency band may be assigned to OD which is coordinated to Ce–D hydride species formed on the reduced films (Figure 3 b). Such a red‐shift could even be used as an indication for hydride species at the surface. In the same manner, the low‐frequency “satellite” bands (3629 and 2677 cm−1) found for the CeO2− surface treated at 500 K (Figure 2 a) can also be assigned to OH and OD groups on surface hydride sites, which are formed during sample cooling in D2.

Therefore, combined IRAS and XPS study on well‐defined CeO2(111) films shows that molecular hydrogen dissociates at relatively high pressures and temperatures and forms surface hydroxyls. On pre‐reduced films, however, both surface hydroxyls and hydrides are formed. The hydride formation seems to be the favorable process at relatively low temperatures (<400 K) within the surface layers probed by XPS, while hydroxylation dominates at higher temperatures, which is accompanied by surface reduction through water desorption.

Ceria powders

Now we address experimental results for the H2–CeO2 interaction obtained on ceria powders under realistic pressure conditions, i.e., 1 atm 5 % H2/He, 1 atm H2, and 10 atm H2, in order to investigate the interplay between surface and bulk in hydroxyl and hydride formation. As shown by electron microscopy (Figure 4), conventional CeO2 powders (from Sigma–Aldrich, see Supporting Information) consist of irregularly shaped polyhedral nanoparticles (10–50 nm) which primarily expose the (111) facets and minor (110) and (100) facets.[ , , ] This morphology basically remains after CeO2 powders were treated in 1 atm of H2 at 773 K. Since it has been shown by DFT calculations, that the energies to create oxygen vacancies on the (111), (100) and (110) surfaces are comparable (i.e., 2.0–2.6 eV), and all are much smaller than in the bulk (3.4 eV), the comparison between the powder samples and the CeO2(111) thin films appears to be appropriate.

Figure 4

TEM and HRTEM images and SAED patterns of ceria powders before (a) and after (b) treatment in 1 atm of H2 at 773 K.

According to the temperature programmed reduction (TPR) spectra (Figure S2), surface reduction in 1 atm of 5 % H2/He commences at above 500 K. In pure H2, the reduction reasonably occurs at lower temperatures (≈380 K) and reaches a maximum at 440 K. Hydrogen uptake peak at 615 K is associated with the reduction in subsurface/bulk region. Both processes are accompanied by the formation of water. It appears that the reduction of ceria becomes more difficult with increasing H2 pressure to 10 atm. A broad water signal peaked at 522 K may be attributed to reduction both at surface and in subsurface/bulk layers. Meanwhile, the H2 consumption does not occur concurrently with the water production, beginning already at 360 K and maximizing at 470 K. These observations suggest that the interaction with hydrogen at 10 atm occurs at profoundly lower temperatures that, in turn, influences the subsequent reduction of ceria.

Figure 5 a shows in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) spectra of ceria powders exposed to H2 at various conditions. (Henceforth, the notation CeO2−‐X‐Y stands for ceria treated in X atm of H2 and temperature Y (in K).) In the ν(OH) region, a vibrational feature at 3666 cm−1 emerges upon exposure to 1 atm H2 at 473 K, which grows and shifts to 3655 cm−1 at 723 K. In addition, a new broad band centered at 950 cm−1 appears. (Weak vibrational features at 860, 1016 and 1044 cm−1 in this region arise from adventitious CO2 adsorption. ) The spectrum does not change much in 10 atm of H2 at 723 K. Moreover, the observed bands are stable and basically remain after sample cooling to 373 K and subsequent pumping the H2 gas out (Figure S3). Ceria exposed to 10 atm D2 at 723 K revealed a strong OD band at 2691 cm−1, which corresponds to the OH counterpart at 3655 cm−1 in the H2 experiments, and nicely matches the one observed by IRAS on the ceria films at 500–600 K (see Figures 1 a and 2 a). The 950 cm−1 band disappears upon D2 adsorption, indicating that this feature relates to the H‐involving vibrations. This band falls in the range of frequencies characteristic for vibrations of bulk Ce−H species.[ , ] Accordingly, Ce−D species would show up at frequencies strongly red‐shifted as compared to Ce−H, which is below the cut‐off frequency of our setups (both for IRAS studies on thin films and DRIFT studies on powders). Please note, that IRAS spectra on metal supported films obey the well‐known metal selection rules and detect only vibrations associated with dipole changes normal to the surface. Therefore, our present results provide first IR‐based experimental evidence for the formation of bulk hydride species upon interaction with H2. A relatively large bandwidth indicates the presence of Ce−H in different local environments. In addition to bulk species, the surface Ce−H species was previously observed in oxidized ceria rods exposed to 1 bar H2 at 673 K, which exhibited a vibrational feature at around 500 cm−1 in inelastic neutron scattering spectra. Note also that the indications of Ce−H vibrations in a very recent Raman study of cerium‐hydride growth centers on metallic Ce foil.

Figure 5

a) In situ DRIFTS spectra measured for ceria treated in H2 at different conditions and in 10 atm D2 at 723 K. The spectra measured on ceria samples treated in Ar at 723 K and cooled to specified temperatures were used as the references. b) MAS 1H NMR spectra measured for ceria treated in H2 at different conditions and then cooled in H2 to room temperature.

Figure 5 b displays respective MAS 1H NMR spectra of the H2‐treated ceria samples. Unfortunately, Ce−H species cannot be detected. Ceria heated at 773 K in He exhibits signals at 1.14/0.77 and 5.34 ppm arising from surface OH groups and adsorbed H2O, respectively. The signals grow with the H2 pressure and temperature. An additional weak feature emerges at 8.39 ppm upon exposure to 1 atm H2 at 773 K and increases significantly upon exposure to 10 atm H2 at 773 K. This feature can be assigned to bulk OH groups in ceria. Therefore, the NMR results provide clear evidence of the increase of the total OH coverage. The OH formation mainly occurs on the surface at temperatures up to 473 K and extends to the bulk region at 773 K. Comparing the evolutions of bulk Ce−H species derived from the in situ DRIFTS results and bulk OH group from the NMR results, we infer that the formation of bulk Ce−H species is more favored than that of bulk OH groups in ceria exposed to H2 at 773 K, where bulk reduction occurs and bulk oxygen vacancies form. (The full dataset of DRIFTS and NMR results are shown in Figures S4 and S5.)

In order to link these results to the thin film data, the ceria samples were characterized by XPS (Figure 6 a and Figures S6a, S7). Based on XPS data, ceria heated to 773 K in He, in order to create oxygen vacancies, exhibits an initial Ce3+ concentration of 9.2 % within the near surface region determined by the escape depth of the Ce 3d‐electrons (≈3 nm), henceforth referred to as Ce3+ surf. Then, the sample was exposed to 1 atm H2 at three different temperatures (303, 473, and 773 K). Consequently, those samples were annealed at the elevated temperatures, up to 773 K, and XPS spectra were recorded. Exposure to 1 atm of H2 at room temperature (CeO2−‐1–303) decreases the Ce3+ surf concentration, as also observed above for the films, due to OH and hydride formation and its interconversion. Increase in temperature under the same pressure conditions (CeO2−‐1–473, CeO2−‐1–773), or in pressure at 773 K (CeO2−‐10–773), leads to a smaller effect. Upon vacuum annealing of the samples the amount of Ce3+ surf increases due to water desorption and shows the biggest effect for the sample treated at 10 atm.

Figure 6

a) Ce3+ surf concentration derived from XPS and b) changes in the Ce3+ bulk concentration (normalized to pristine ceria) derived from EPR, for ceria samples treated in H2 at different conditions and then cooled in H2 to room temperature followed by vacuum annealing (for 30–60 min) during XPS measurements or purging in He (for 2 hours) during ESR measurements at indicated temperatures.

Now we compare the XPS results with EPR data (spectra are shown in Figures S6b and S8), probing both surface/near surface and bulk species. The relative changes in the Ce3+ concentration (henceforth, Ce3+ bulk), as represented by the EPR intensity normalized to that of pristine ceria (prepared in the same way as the samples for XPS), are shown in Figure 6 b. Except the CeO2−‐1–473 sample, the Ce3+ bulk concentration drops to very low values in CeO2−‐1–773 and CeO2−‐10–773, as well as for the room temperature sample, clearly indicating bulk hydride formation in these samples. Since EPR probes the entire sample, both surface and bulk, the small amount of Ce3+ near the surface (monitored by XPS, Figure 6 a) contributes much less to the signal. The obvious increase of Ce3+ observed by XPS and EPR on the CeO2−‐1–303 sample upon vacuum annealing or purging at 423 K reflects the transformation of surface Ce4+H− into more stable surface OH, accompanied by the reduction of Ce4+ into Ce3+, and the decomposition of surface CeOv 4+H− into hydrogen and CeOv 3+. It is interesting to note, that, after He purging at and above 423 K, the EPR data for the CeO2−‐1–303 almost match the set of data taken on the CeO2−‐1–473 sample. A glance at the water desorption data (Figure S9) tells us that this is due to the onset of water desorption and thus oxygen vacancy formation around 400 K. Both the Ce3+ surf and Ce3+ bulk concentrations in CeO2−‐1–303 and CeO2−‐1–473 and the Ce3+ surf concentrations in CeO2−‐1–773 and CeO2−‐10–773 then increase and exceed the initial concentration of CeO2− upon vacuum annealing or purging at 523 K and above, again, mainly due to the recombinative reaction of surface OH groups to produce water and create Ov and Ce3+. However, the Ce3+ bulk concentrations in the CeO2−‐1–773 and CeO2−‐10–773 samples are always lower than the initial concentration, indicating the presence of bulk Ce−H species. They increase upon annealing at 523–623 K via the reaction 2CeOv 4+H− → H2(g) + 2CeOv 3+, but decrease at higher temperatures (773 K), although the H2 production continues (Figure S9). On one hand, this agrees with the DRIFTS results that the H2 productions above 623 K for CeO2−‐1–773 and CeO2−‐10–773 involve the OH species. On the other hand, this indicates the re‐formation of bulk CeOv 4+H− species most likely by the migration of H atoms in surface OH species to bulk CeOv 3+ sites. Similar phenomena were previously observed to occur on the highly hydrogenated and reduced TiO2 surfaces. These observations demonstrate that the stability of Ce−H species enhances greatly with the Ov concentration in ceria.

In summary, the variations of Ce3+ surf and Ce3+ bulk species (determined by XPS and EPR, respectively) indicate relative amounts of Ce3+ created by the homolytic dissociation to OH and oxygen vacancy formation reactions [Eqs. (1) and (2) in the Introduction], and annihilated by the homolytic dissociation to CeOv 4+H− [Eq. (3)], which, together with the formed H species, can be used to deduce which reactions occur upon H2–CeO2 interaction and where. Certainly, the Ce3+ surf and Ce3+ bulk species overlap in the subsurface region. Compared to ceria pretreated at 773 K in He, the CeO2−‐1–303 sample contains less Ce3+ surf and Ce3+ bulk, due to the kinetically‐favored heterolytic dissociation to OH and Ce4+H− on the ceria surface [Eq. (4)] and homolytic dissociation to CeOv 4+H− at surface oxygen vacancy sites [Eq. (3)]. At 473 and 773 K, H2 undergoes the thermodynamically favored homolytic dissociation to OH species on the ceria surface, the oxygen vacancy formation reaction, and the homolytic dissociation to CeOv 4+H− at initially present and newly formed oxygen vacancies in the sub‐surface region, leading to almost unchanged Ce3+ surf as initially observed in CeO2−‐1–473, CeO2−‐1–773 and CeO2−‐10–773. The only slightly increased Ce3+ bulk concentration observed for CeO2−‐1–473 is likely due to either limited creation of bulk oxygen vacancies or the limited diffusivity of hydride filled oxygen vacancies at that temperature. However, CeO2−‐1–773 and CeO2−‐10–773 with high oxygen vacancy concentrations, arising either from diffusion or creation in the bulk, exhibit almost no Ce3+ bulk, indicating the formation of bulk CeOv 4+H−. Those samples also exhibit bulk OH species arising from the homolytic dissociation to OH or the heterolytic dissociation in the bulk.

Theoretical calculations

To study the relative stabilities of proton (H+−O) and hydride (H−−Ce) species at CeO2(111) surfaces with different degrees of reduction, we constructed CeO2(111)‐nOVsurf [n refers to the number of surface oxygen vacancies ranging from 1 to 4, and n = 4 corresponds to a Ov coverage of 0.16 monolayer (ML)] based on the “hydroxyl‐vacancy model”. Note that for the cases of n = 2, 3, and 4, the surface oxygen vacancies form an energetically favorable (linear) surface oxygen vacancy dimer, trimer, and tetramer, respectively (Figure 7 a). It can be seen from Figure 8 a that, by increasing the concentration of surface oxygen vacancies (n = 1 → 4), the adsorption energy (strength) of proton decreases significantly (1.40 → 0.61 eV), which agrees well with the previous theoretical study. In contrast, the adsorption energy of hydride shows an opposite trend. More specifically, the adsorption energy of hydride drastically changes from −0.71 to 0.41 eV, when increasing the concentration of surface oxygen vacancies (n = 1 → 4), indicating that surface hydride can be stabilized by oxygen vacancies. Presumably, this is because the electron‐donating character of CeO2(111) is enhanced, and, hence, surface hydride gets stabilized. Therefore, it may be expected that surface hydride is more stable than surface hydrogen atom at heavily reduced CeO2−(111) surfaces with n > 4, although we were not able to simulate such surfaces due to the high computational cost.

Figure 7

a) Optimized structures (top view) of CeO2(111)‐nOVsurf (n = 1, 2, 3, or 4). Surface cerium, hydrogen, surface oxygen, and subsurface oxygen atoms are represented by light yellow, white, red, and pink balls, respectively; the bottom two O‐Ce‐O trilayers are drawn with lines for clarity. b) Optimized structures (side view) of CeO2(111)‐1OVbulk and CeO2(111)‐4OVbulk. Cerium and oxygen atoms are in light yellow and red, respectively. c) Scaled DFT‐calculated vibrational frequencies of surface and bulk hydrides based on the selected models (i.e., CeO2(111)‐4OVsurf, CeO2(111)‐1OVbulk, and CeO2(111)‐4OVbulk) and schematic views of the corresponding vibrational modes. The top, middle, and bottom O‐Ce‐O trilayers of the CeO2(111) slab are simplified by light red, light blue, and gray sheets, respectively. The purple arrows represent the directions of the vibrations of the hydrides (green balls). In a, b, and c, dashed black circles represent oxygen vacancies.

Figure 8

Calculated adsorption energies ( ) of hydride and proton as a function of the number of oxygen vacancies based on a) CeO2(111)‐nOVsurf (n = 1, 2, 3, or 4) and b) CeO2(111)‐nOVbulk (n = 1 or 4). Calculated energy profiles for c) H2 dissociation on CeO2(111)‐4OVsurf through the homolytic‐1, homolytic‐2, and heterolytic pathways and d) the transformation of surface hydride to surface hydrogen atom on CeO2(111)‐4OVsurf. For c, H2(g), H2(ads), TSH‐H, 2H(ads) represent, respectively, the state with gas‐phase H2, the surface adsorption state of H2, the transition state of H2 dissociation, and the surface adsorption state of two H (which can be hydride or hydrogen atom). For d, H−(ads), TSH, and H+(ads) represent, respectively the surface adsorption state of hydride, the transition state of transformation of hydride to proton, the surface adsorption state of proton. Corresponding structures are shown in Figures S10–S13.

We also evaluated the relative stabilities of hydrogen atom and hydride species in the bulk of ceria by constructing the slab models of CeO2(111) with one and four aggregated oxygen vacancies in the middle O‐Ce‐O trilayer (denoted as CeO2(111)‐1OVbulk and CeO2(111)‐4OVbulk, respectively, see Figure 7 b). Not surprisingly, compared with the surface case, oxygen vacancies in the “bulk” have similar effects on the stabilities of surrounding hydrogen atom and hydride species (see Figure 8 b). It is interesting to note that, for the vacancy‐rich case of CeO2(111)‐4OVbulk, the bulk hydride is ≈1.4 eV more stable than a corresponding hydrogen atom, suggesting that hydride is the dominant species in the bulk region of heavily reduced ceria. Meanwhile, these results also indicate, that bulk hydride is more stable than surface hydride, whereas a surface hydrogen atom is more stable than a bulk hydrogen atom.

Based on the representative CeO2(111)‐4OVsurf model, we studied the following three possible dissociation pathways of H2: (i) homolytic dissociation of H2 to two surface hydrogen atoms (denoted as the homolytic‐1 pathway); (ii) homolytic dissociation of H2 to two surface hydrides (the homolytic‐2 pathway); (iii) heterolytic dissociation of H2 to one surface proton and one surface hydride (the heterolytic pathway). The calculated energy profiles and the relevant structures are presented in Figure 8 c and Figure S12, respectively. The three dissociation pathways all start from a weakly adsorbed H2 ( = 0.02 eV) above a Ce atom. We found that the heterolytic pathway gives the lowest activation energy barrier of 0.56 eV, followed by the homolytic‐2 pathway (barrier: 1.15 eV). Although the homolytic‐1 pathway is thermodynamically most favorable, it has a highest barrier of 1.75 eV, which is consistent with the previous study. Therefore, formation of surface hydride through the heterolytic and homolytic‐2 pathways is kinetically favorable on the CeO2(111)‐4OVsurf surface. We also found that the transformation of surface hydride to surface proton is extremely difficult at CeO2(111)‐4OVsurf since it needs to overcome a high barrier of 2.67 eV (Figure 8 d and Figure S13). Accordingly, hydride is kinetically stable at such surface. Meanwhile, the recombinative desorption of surface hydrides to H2 is most favorable both, thermodynamically and kinetically, among all H2 production pathways from H species. The results based on the CeO2(111)‐4OVsurf and CeO2(111)‐4OVbulk models also suggest that surface hydrides, once formed, can readily diffuse into the bulk region and be stabilized therein at heavily reduced conditions. Therefore, the DFT results fully agree with the above experimental observations for H2–CeO2− powder systems.

The scaled DFT‐calculated vibrational frequencies of surface and bulk hydrides based on the selected models (i.e., CeO2(111)‐4OVsurf, CeO2(111)‐1OVbulk, and CeO2(111)‐4OVbulk) are summarized in Table S1. The vibrations of bulk hydride depend strongly on the configuration of oxygen vacancies, but they cover the experimentally observed broad vibrational band centered at 950 cm−1 (Figure 7 c). For surface hydride, the vibration at 963 cm−1 (Figure 7 c) exhibits a dipole moment parallel to the surface, which rationalizes its absence in IRAS experiments with CeO2−(111) thin film due to the well‐known metal surface selection rule applied in IRAS.

Summary

Our combined experimental and computational results provide a comprehensive picture of the H2–ceria interaction and highlight the vital role of oxygen vacancies in this process as well as the important interplay between surface and bulk. On a stoichiometric CeO2 surface, H2 only dissociates to form surface hydroxyls at elevated temperatures. On reduced CeO2−, hydroxyls and hydrides are formed on the surface and in the bulk, depending on Ov concentration and location. The general trends are that hydroxyl is destabilized as the Ov coverage increases, whereas hydride is stabilized, and that surface hydroxyl is more stable than bulk hydroxyl, whereas bulk hydride is more stable than surface hydride. The surface hydride formation is the kinetically favorable process at relatively low temperatures, and the resulting surface hydride can easily diffuse into the bulk region and get stabilized therein. At higher temperatures, surface hydroxyls can react to produce water and create additional oxygen vacancies, which in turn control the H2–ceria interaction.

The results demonstrate the facile formation of hydride (CeOv 4+H−) at the CeOv 3+ site, leading to the oxidation of CeOv 3+. Ce3+ sites associated with oxygen vacancies play an important role in determining the CeO2− reactivity, therefore, their annihilation via the CeOv 4+H− formation will affect the CeO2− reactivity. The results also demonstrate the hydride formation at the in situ created oxygen vacancy sites of ceria during the reduction process, suggesting that the ceria reduction by H2 should follow the reaction equation CeO2 + (x+y)H2 → CeO2− + x H2O + 2yCeOv 4+H− + (x−y)Ov rather than CeO2 + x H2 → CeO2− + x H2O + x Ov.

We believe that our results showing the diversity of reaction pathways taking place upon interaction of H2 with ceria will aid in full understanding of the reactivity of ceria‐based catalysts in a hydrogen‐rich atmosphere.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file.