Cerium oxide (CeO₂, Ceria) finds many applications in catalysis as both a catalyst itself, and as a catalyst support. The first and the most well-known catalytic application of ceria is its use as an oxygen storage component in automotive three-way catalysts (TWC), which convert the main pollutants present in vehicle emissions (CO, NO and unburned hydrocarbons) into less harmful compounds [1]. Over the past few decades, ceria and ceria-based materials have found increasing applications in catalysis thanks to their unique structural and electronic properties. The interesting properties of ceria are associated with its ability to cycle between Ce(Ⅲ) and Ce(Ⅳ) states upon interaction with reducing or oxidizing components in reacting mixtures, forming and annihilating oxygen vacancies in a stable fluorite-type structure [2]. Additionally, ceria is characterized by several types of surface and subsurface defects, which can be controlled by adjusting the surface morphology of the ceria itself, or doping the ceria with other promoters to form Ce³⁺ species that may form active sites for heterogeneous catalysis [3-7]. Ceria also has controllable acid-base properties in the form of both Lewis acid and base sites, depending on ceria morphology and pre-treatment procedures [3, 8, 9] and can efficiently catalyze various organic reactions in both gas phase and aqueous phase [3, 10-13].

In addition to its redox and acid-base properties, ceria combined with metals has attracted increasing interest in heterogeneous catalysis for various hydrogenation reactions such as H₂ generation [14-16], CO₂ hydrogenation [17-21], and the hydrogenation of different functional groups [22-27]. Recently, pure ceria demonstrated surprising catalytic ability in hydrogenation reactions [28-31], particularly the selective hydrogenation of alkynes to alkenes in both gas and liquid phases with high selectivity (80%–90%) [26, 31, 32]. Contrary to observed trends for many redox reactions catalyzed by ceria, the formation of surface oxygen vacancies prior to or during reaction was detrimental to the hydrogenation activity [26, 31, 32]. Other recent studies [29, 30, 33] showed that oxygen-vacancy-associated frustrated Lewis pairs (FLPs) promoted the hydrogenation of alkynes over porous ceria rods. The hydrogenation ability of bare ceria, without any metallic promoter, has prompted a renewed strong interest in understanding how ceria dissociates H₂ and catalyzes selective hydrogenation, and has been studied using a number of experimental and computational approaches [22, 23, 25, 33-38].

Many detailed reviews have been written that summarize aspects of ceria or ceria-based materials [3, 4, 8, 30, 39-52], and focus on topics such as redox properties, general oxidation catalysis, and emerging applications including photocatalysis and biomedical technology. There are very limited perspectives [30, 31], however, on hydrogenation reactions catalyzed by ceria-based oxides, and we find no overview focused only on ceria-based materials for hydrogenation without the assistance of a metal component. Therefore, the goal of the present review is to survey the existing understanding of the specific properties of ceria associated with interaction with H₂ and under conditions used for hydrogenation reactions to provide an in-depth understanding of how ceria-based oxides catalyze hydrogenation reactions as well as redox reactions (since reduction of ceria is part of the redox process). We will focus on analyzing experimental and theoretical studies on (1) hydrogen interaction and (2) selective alkyne hydrogenation. Although the focal point is ceria, some relevant cases over doped ceria will also be reviewed. First, we will discuss the experimental aspect concerning the mechanistic understanding and catalytic properties of ceria in interacting with H₂ and selective alkyne hydrogenation. The reductive behaviors of CeO₂ by H₂ and the relation to hydrogenation reactions have been studied in great details by different techniques and the use of these approaches will be surveyed including Fourier transform infrared spectroscopy (FT-IR, DRIFTS) [8, 23, 53], Raman spectroscopy [54], ¹⁷O-NMR [55], X-ray photoelectron spectroscopy (XPS) [56, 57], X-ray diffraction (XRD) [58], and neutron vibrational spectroscopy [38, 53, 59]. After discussing these experimental investigations of ceria, the representative theoretical studies will be reviewed to understand the H₂ dissociation and subsequent hydrogenation mechanisms over ceria at atomic scale. Finally, a summary with the outlook of ceria-based catalysts for hydrogenation reactions is presented.

One of the most straightforward experiments that can be performed to observe the interaction of H₂ with CeO₂ is to heat CeO₂ at a controlled rate under an atmosphere containing H₂. Despite the simplicity of the technique, careful analysis of products or consumption of reactants measured in the effluent using mass spectrometry or thermal conductivity can provide some insight into the dynamic processes that are occurring on the CeO₂ surface as well as in the bulk of the material. The so-called temperature-programmed reduction by H₂ (H₂-TPR) has become a ubiquitous technique used to characterize the surface and bulk properties of CeO₂ [2, 25, 60-68].

The characteristic H₂-TPR profile of CeO₂ generally consists of two main H₂ consumption features near 770 and 1100 K. Consumption of H₂ occurring around 770 K has been ascribed to the reduction of surface oxygen species (forming O vacancies), as the total consumption in this region is well-correlated with the surface area of polycrystalline CeO₂ samples [2]. Consumption of H₂ around 1100 K was ascribed to the reduction of bulk CeO₂, resulting from diffusion of O out of the CeO₂ bulk to the CeO₂ surface where it can react with adsorbed or gas phase H species. Johnson and Mooi proposed a relationship between the total H₂ consumption around 770 K and CeO₂ surface area in CeO₂ samples up to 27 m² g^–1 [60]. Perrichon et al. [61] and Bruce et al. [62] demonstrated that the total consumption of H₂ in this region remains well-correlated with the BET surface area of carbonate-free CeO₂ in samples up to 168 m² g^–1, although there exists some disagreement between these results and those reported by Johnson et al. [60]. Giordano et al. [63] developed a model to describe temperature-programmed consumption of H₂ by CeO₂ with initial surface area up to 44 m² g^–1, and concluded that the nuances of surface reduction kinetics were still not fully understood. Further consideration of the kinetics of morphological changes in the material (e.g., the sintering of high-surface-area particles) was also deemed critical to understanding the H₂-TPR behavior of CeO₂ [64]. As oxygen diffusion in the CeO₂ lattice is expected to be rapid under the conditions required for reduction, the role of different exposed surface planes of CeO₂ was expected to play a significant role in the reduction behavior in the lower-temperature reduction regime.

Indeed, exploration of the role of exposed surface facets using ceria nanoshapes of different morphologies revealed differences in reduction behavior [65-67]. Tana et al. [65] found highly faceted nanoparticles demonstrated a lower oxygen storage capacity than nanorods and nanowires as measured by oxygen titration after reduction. Reversible oxygen storage was attributed to preferential exposure of the (110) surface in rod- and wire-shaped nano-CeO₂ over the less reactive (111) surface exposed on the faceted nanoparticles. Liu et al. [66] observed that surface oxygen was removed by H₂ at lower temperatures as lattice strain was increased, and suggested that catalytic oxidation activity was associated with the reversible oxygen storage capacity of the materials. When Désauney et al. [67] synthesized highly controlled CeO₂ nanocubes, nanorods, and nanooctahedra, a significant difference was observed between the reducibility of the preferentially exposed surface planes. As shown in Fig. 1, the majority of H₂ consumption observed in the H₂-TPR profiles of CeO₂ nanocubes and nanooctahedra took place above 773 K, while H₂ consumption over the nanorods was split more evenly between temperature regimes around 700 and 950 K. Consumption of H₂ resulting from surface reduction occurred at lower temperatures over nanocubes and nanorods than over the nanooctahedra. Reported apparent activation barriers for surface reduction were 70, 130, and 150 kJ mol^-1 over nanocubes, nanorods, and nanooctahedra, respectively, corroborating that the ease of reducibility of CeO₂ surface planes follows the trend (100) > (110) > (111), in agreement with Tana et al.

Analysis of the desorption of H-containing species (using thermal conductivity) from polycrystalline CeO₂ after reduction above 773 K revealed that increasing the reduction temperature decreased the amount of adsorbed H₂ [69]. Below 773 K, analysis of the amount of reversibly adsorbed H₂ must be carried out using mass spectrometry, as surface oxygen was susceptible to reduction by the chemisorbed H-containing species, forming water. After reduction above 773 K, only H₂ desorption was observed by mass spectrometry, indicating results from thermal conductivity could be used for analysis. The total H₂ quantity adsorbed per nm² CeO₂ increased slightly up to reduction temperatures of 873 K and decreased as the reduction temperature was increased thereafter, even after consideration of particle sintering. These results suggested reduction of the CeO₂ surface was not critical for reversible H₂ adsorption to take place. More recent results, however, do indicate that H₂ can be observed in temperature-programmed experiments monitored by mass spectrometry [70, 71]. This observed H₂ is believed to result from the desorption of H- species that are incorporated into the bulk of CeO_2–x, and is directly correlated to the presence of O vacancies, as discussed in the next sections.

Consumption of H₂ observed in the H₂-TPR of CeO₂ was generally attributed to the reduction of CeO₂ lattice oxygen as discussed above. Yet these studies have also given rise to some important questions: What is the effect of H₂ exposure on the bulk structure of CeO₂? Is there hydrogen incorporation into CeO₂ upon reduction with H₂? Whether H atoms enter the bulk or are merely chemisorbed on the surfaces of CeO₂ is the subject of some debate. Some characterization approaches, described below, suggest incorporation of hydrogen into the bulk of CeO₂ takes place upon high temperature H₂ treatment.

Investigation of CeO₂ during in situ reduction in H₂ using XRD found three zones of lattice behavior [72, 73]. Below 600 K, little change in the CeO₂ lattice takes place. From around 600 – 800 K, a slight expansion in the lattice takes place, followed by a much larger expansion above 850 K. These lattice expansions are consistent with the typical H₂-TPR profiles indicating the observed phenomena are related to reduction of the CeO₂, which could be explained by expansion due to the formation of Ce³⁺ species in the reduced suboxide. It has also been argued that a contribution to this expansion could arise from the incorporation of H into the CeO₂ lattice, potentially in the form of a bronze of the form H_xCeO₂ (or non-stoichiometric cerium oxyhydride). The bronze explanation was originally proposed by Fierro et al. [74] and Cunningham et al. [75] on the basis of ¹H nuclear magnetic resonance (¹H-NMR) and electron spin resonance (ESR) spectroscopies. Fierro et al. [74] observed the incorporation of H into the CeO₂ sample by ¹H-NMR concurrent with the obliteration of bulk ESR signals, which suggested H diffusion into the bulk of the CeO₂ during H₂ treatment up to 673 K. A study by Bekheet et al. [58] found no evidence for incorporation of H in to the CeO₂ lattice by ¹H NMR. Matsukawa et al. [76], however, observed the formation of a blue color in a CeO₂ pellet, behavior typical of bronze formation, after brief exposure (30 min) to H₂ at 1073 K in the absence of significant reduction. The formation of the aforementioned blue phase was also accompanied by an expansion of the CeO₂ lattice observed using XRD, consistent with the formation of a H_xCeO₂ bronze phase (x = 0.04 or 0.06). Extended exposure of the sample to the reducing conditions resulted in reduction of the blue proposed bronze to a white CeO_2–x suboxide. The role of treatment time in the transient formation of H_xCeO₂ bronzes has not been extensively studied and may explain the aforementioned contradictory NMR results. Bulk and surface hydride species (characterized by vibrational modes from 700-1100 and 400–600 cm^–1, respectively) were also observed by Wu et al. [38] after partial reduction of CeO₂ nanorods by H₂ exposure using inelastic neutron scattering (INS) spectroscopy. This technique is extremely sensitive to hydrogen-containing species, and suggests that hydride species can exist both on the surface and in the bulk of reduced CeO_x. These results agree with a study that observed desorption of H₂ from CeO₂ attributed to H^- incorporated into bulk and surface O vacancy sites [70]. The existence of H-containing species in the CeO₂ lattice under catalytic conditions, however, remains up for debate.

Study of the CeO₂ surface using high-resolution and environmental transmission electron microscopy (TEM) has provided insight into the relationship between H₂ treatment and CeO₂ surface restructuring [59, 77-80]. Oxygen atoms on a CeO₂ (111) surface remained stable under an electron beam [79], while oxygen atoms on the CeO₂ (100) surface were highly mobile under vacuum conditions (5 × 10^–6 mbar, CeO₂ nanocubes) [59, 79]. Mobility in the (110) surface of the CeO₂ nanocubes was significantly reduced under O₂ and CO₂ environments (5 × 10^–2, 2.6 × 10^–2 mbar, respectively) [59, 79]. When the nanocubes were exposed to oxidizing conditions, the CeO₂ nanocube (110) surfaces restructured to form sawtooth CeO₂ (111) nanofacets [77, 80]. Annealing in H₂ ex situ above 773 K prior to observation using TEM resulted in reconstruction of the CeO₂ (110) surface to form larger (111) sawtooth nanofacets [80]. Observation of the reconstruction under environmental TEM conditions (0.7 mbar, up to 1003 K) (Fig. 2), however, revealed that H₂ treatment of the (110) surface results in the smoothing of these (111) nanofacets to form a continuous (110) surface, and suggests that this reconstruction is only stable under reducing atmosphere [77]. Further investigation using the same instrument also revealed a reversible partial phase transformation of CeO₂ from the fluorite structure to a C-type Ce₂O₃ structure above 966 K in 0.7 mbar H₂ [78]. The phase transformation was completely reversed after cooling below 966 K, and reoxidation of Ce³⁺ to Ce⁴⁺ was observed, despite cooling in an H₂ environment. Whether this transformation is associated with H incorporation remains unclear.

A better understanding of the interaction between H₂ and CeO₂ can be achieved using spectroscopic techniques to identify the types of H-containing species on the CeO₂ surface after H₂ exposure. Not all spectroscopic techniques, however, are equal in their ability to detect these H-containing species. While infrared (IR) and Raman spectroscopy have been used to characterize adsorbed OH species, thus far only more H-sensitive techniques such as ¹H-NMR and INS have been used successfully to detect small quantities of H in the CeO₂ bulk or Ce-H surface species as discussed earlier.

Infrared spectroscopy has been extensively used to characterize CeO₂ surfaces and their modification after exposure to H₂ or D₂ [8, 22, 26, 38, 47, 48, 81-88]. Studies by Lamotte et al. [81] and Binet et al. [82] supported the hypothesis that the coordination of methoxy and OH species occurred via one- (Ⅰ), two- (Ⅱ), or three-fold (Ⅲ) binding through a single oxygen to surface cerium atoms [81, 82]. The appearance of these spectra is shown in Fig. 3.

Binet et al. [82] observed that after a 673 K pretreatment in O₂ followed by outgassing of CeO₂ with a surface area of 78 m² g^–1, a mixture of υ(OC) bands consistent with mono- (~1120 cm^–1) and bidentate (~1060 cm^–1) methoxy species were formed after room-temperature adsorption of methanol, and very little contribution was observed from the tri-dentate species (~1010 cm^–1) [8, 82]. A shoulder peak in the bidentate region (Ⅱ', ~1040 cm^–1) was assigned to adsorption of the probe molecule near coordinatively unsaturated Ce⁴⁺ species. As the CeO₂ samples were reduced in H₂ at increasing temperature and subsequently exposed to methanol, the monodentate methoxy species were no longer observed, and a new species was observed at slightly higher wavenumber than the type Ⅱ bidentate methoxy on O₂-treated CeO₂. This new species was also assigned to a bidentate configuration (Ⅱ^*, ~1080 cm^–1), believed to be adsorbed on or near reduced surface sites containing Ce³⁺ (consistent with the formation of surface O vacancies), and became the dominant methoxy species observed after H₂ treatment at temperatures as low as 673 K, consistent with complete surface reduction [83]. An electronic transition, possibly the forbidden ²F_5/2 → ²F_7/2 transition (proposed to no longer be forbidden due to a crystal field effect) or other transition related to Ce³⁺, was also assigned to a feature observed at ~ 2120 cm^–1 [82].

When CeO₂ with a surface area of 120 m² g^–1 was subjected to a similar treatment, analogous behavior was observed for surface OH species [8, 82, 84]. After initial oxygen treatment at 673 K, however, υ(OH) bands assigned to bidentate (Ⅱ-A ~3635 cm^–1, Ⅱ-B ~3654 cm^–1 analogous to the methoxy Ⅱ and Ⅱ' bands) and hydrogen-bonded (broad, ~3500 cm^–1) hydroxyl species were primarily observed, while monodentate species (Ⅰ, ~3710 cm^–1) and tridentate (Ⅲ, ~3585 cm^–1) were less evident [82]. Treatment in O₂ at 873 K was able to remove nearly all the OH species from the CeO₂ surface [8]. Treatment in H₂ at 673 K produced two distinct bidentate species (Ⅱ^*-A, ~3680 cm^–1, Ⅱ^*-B, ~3650 cm^–1), analogous to methoxy species Ⅱ^* observed after similar treatment over the lower-surface-area CeO₂, and a broad OH stretching feature assigned to hydrogen bonding of residual OH species. Similar distributions of OH species have been observed using Raman spectroscopy, absent from any broad band at ~3520 cm^–1 associated with H-bonded OH species [38]. When H₂ exposure was performed on CeO₂ pretreated in O₂ at 673 K using small doses of H₂ instead of separate heat treatments, the initially visible Ⅱ-B υ(OH) band disappeared while the Ⅱ-A band remained and shifted to higher wavenumber, indicating Ⅱ-B OH species were more reactive with H₂ than Ⅱ-A species [84]. The formation of the Ⅱ^*-B OH species was evidently reversible as evacuation at 673 K after H₂ treatment resulted in the disappearance of the Ⅱ^*-B OH species, and subsequent treatment in H₂ caused the species to return.

Adsorption of a range of acidic and basic molecules, summarized by Binet et al. [8], revealed weakly acidic and strongly basic sites on CeO₂ in both its oxidized and reduced state. Results from adsorption of CO₂, dimethyl ether, and pyridine indicated very little difference in acidity resulting from reduction of CeO₂, and peak shifts from pyridine adsorption indicated weaker acid site strength on CeO₂ than ZrO₂ or TiO₂. Similar conclusions could be drawn from adsorption of CO and acetonitrile, which revealed weaker Lewis acidity on the reduced CeO₂ surface [9]. On the other hand, quantification of the base site strength remains a challenge. Adsorption of a proton donor, pyrrole, resulted in dissociation of the probe molecule and could not be used to quantify the base strength of the CeO₂ surface [88]. Low-temperature adsorption of a weaker H-donor, methane, was employed to probe the CeO₂ surface, revealing coordinatively saturated and unsaturated O^2– surface species but little about the strength of base sites present on the surface [89]. The adsorption of CO₂, meanwhile, is complex, and while a large amount of CO₂ adsorption occurs, it is difficult to clearly identify and assign individual CO₂ adsorption bands without computational support, especially in polycrystalline samples [8, 90]. TPD of CO₂ from shape-controlled nano-CeO₂, however, indicated that nanorods (dominated by (110) and (100) surfaces) contained the highest density of basic sites, followed by nanocubes (predominantly (100) surfaces), then nanooctahedra (predominantly (111) surfaces) [9]. The nanoshape surfaces were also probed by CHCl₃ and CD₃CN adsorption, indicating a similar trend in base site strength.

Raman spectroscopy has also been used to characterize oxygen defect sites on similar shape-controlled CeO₂ nanoparticles via the presence of different adsorbed O₂ species (superoxide and peroxide) after reduction using H₂, which correspond to different types of oxygen vacancies [91]. Reduction at 673 K results in the formation of both one- and two-electron defects on ceria nanorods (dominated by (110) and (100) surfaces), while the same treatment produces isolated or a mixture of isolated and clustered two-electron defects on CeO₂ nanooctahedra (dominated by (111) surfaces) and nanocubes (dominated by (100) surfaces), respectively. Adsorption of methoxy species on shape-controlled nano-CeO₂ at room temperature found predominantly type Ⅰ methoxy species on nanorods, while nanocubes contained a majority of a mixture of type Ⅰ and type Ⅱ', and nanooctahedra contained very few adsorbed methoxy species (only type Ⅰ species were detected) [92]. To date, no comparative study exists to demonstrate the difference between OH or methoxy species formed on oxidized and reduced shape-controlled nano-CeO₂.

The use of H-D exchange has revealed the ability of CeO₂ (and particularly, CeO₂ pre-treated in H₂) to exchange H and D at modest temperatures [22, 84, 86, 87]. When Badri et al. [84] adsorbed D₂ on CeO₂ treated at 673 K in O₂, the formation of υ(OD) bands did not result in any concurrent decrease in observed υ(OH) band intensity until 473 K, indicating that H-D exchange on oxidized CeO₂ did not occur at an appreciable rate until above 423 K. Instead, surface OD species appeared to form in addition to the already present OH species, and converted some of the existing type Ⅱ-B OH species to type I. The new type Ⅱ OD species were assigned to type Ⅱ-A OD species. Further investigation by Werner et al. [22] also found that D₂ dissociation could take place on both stoichiometric CeO₂ and partially reduced CeO_x thin films at pressures above the mbar regime.

When Martin et al. [87] measured the rate of H-D exchange using a temperature-programmed isotopic exchange (TPIE) technique, exchange of H and D reached a maximum rate at 433 K over oxidized CeO₂, but when the CeO₂ was treated in H₂ prior to the TPIE, the exchange reached a maximum rate at ~373 K. The apparent activation energies for H-D exchange over the O₂-treated and H₂-treated (723 K) CeO₂ samples were 59 and 26 kJ mol^–1 respectively. Evidently the H₂ pretreatment produced a group of more reactive H-containing species on the reduced CeO₂ surface, which caused H-D exchange to take place more rapidly than exchange over other oxides such as ZrO₂ or MgO, and doubled the total quantity of exchanged H. Schimming et al. [86] also measured the apparent activation energy of H-D exchange over CeO₂ pretreated at 723 K in H₂, and found a value of 22 kJ mol^–1 in agreement with Martin et al. [87]. When the exchange was carried out at 423 K, IR results indicated that type Ⅰ OH species decreased as type Ⅰ and Ⅱ OD species appeared, suggesting exchange of D₂ with type Ⅰ OH species prior to type Ⅱ-B OH species. In contrast, Badri et al. [83] observed the formation of type Ⅰ OH species during their H-D exchange experiments. As Schimming et al. [86] studied the exchange after H₂ pretreatment, while Badri et al. [83] studied the exchange after O₂ pretreatment, it stands to reason that type Ⅰ OH species are more readily exchanged when the CeO₂ surface is saturated with H-containing species after H₂ pretreatment (including reduction). The H-D exchange measured by Schimming et al. [86] at 623 K was less selective to type Ⅰ OH species, and instead exchange occurred indiscriminately on all site types.

While many studies have focused on the interaction of bulk or polycrystalline CeO₂ with H₂, studies performed under ultra-high vacuum (UHV) offer the possibility of a highly controlled environment to investigate the interaction of H₂ with a well-defined CeO₂ surface. Advances in the growth of bulk-like, ordered CeO₂(111) films with low O-defect contents recently led to studies of pristine and oxygen-deficient CeO₂(111) surfaces after exposure to molecular and atomic hydrogen [93-95]. Investigation of the behavior of CeO₂ under UHV conditions has generated insights into the chemical state and structure of CeO₂(111) surfaces after controlled exposure to these H-containing environments.

Interaction of surface OD species on CeO₂(111) surfaces with atomic D was observed to give rise to D₂O and oxygen vacancies between 150–250 K, resulting in a partial reduction of the surface [96]. As the defect density increased to a critical level, D₂O formation was suppressed, at which point D₂ evolution become the more dominant pathway (at ~ 200–350 K) for elimination of D from the CeO₂ surface. This behavior may be analogous to the observation of reversible H₂ adsorption on CeO_2–x after H₂ pretreatment of a CeO₂ sample [69], and similarly, to the observation of H₂ evolution behavior correlated to O vacancy density in shape-controlled CeO₂ nanoparticles [71]. When the pristine CeO₂(111) surface was exposed to molecular H₂ at room temperature, no reaction was observed, while exposure to atomic H resulted in a mixture of hydroxyl species and oxygen defects [57]. As should be expected, this result implies that the rate of dissociation of H₂ on CeO₂(111) is negligible under the UHV conditions studied. Later study demonstrated that dissociative adsorption of molecular H₂ could not be observed below pressures of 10 mbar [22].

Studies using XPS to probe the Ce 3d photoemission demonstrated that either the formation of oxygen vacancies or the formation of hydroxyl species can give rise to Ce³⁺ species in the lattice. On reduced CeO₂(111), D₂O could also be irreversibly chemisorbed, causing OD groups to fill anion vacancy sites on the defected surface [96]. Investigation of well-defined CeO₂(111) films under vacuum conditions using nuclear reaction analysis (¹⁵N + ¹H → ¹²C + α + γ) indicated that thermally reduced, O-defective CeO₂(111) incorporated H species into the bulk of the film after exposure to 10 mbar H₂ at 300 K, while pristine CeO₂ films stabilized H-containing species only at the film surface after the same exposure (Fig. 4) [22]. These results are consistent with those observed using inelastic neutron scattering spectroscopy, which indicated the formation of bulk Ce-H species after treatment of CeO₂ at 673 K [38], at which temperature surface reduction (i.e. the formation of vacancies) would be expected. A recent study of H₂ adsorption on pre-reduced ceria (both thin film and nanoparticles) using XPS, electron energy-loss spectroscopy, and ESR suggested that H-containing species exist in Ce⁴⁺-H^- pairs, where H^- is stabilized in an O vacancy site, resulting in the re-oxidation of Ce³⁺ species and reforming the atomic structure of the stoichiometric CeO₂ [70].

Conclusions from these studies can be cautiously applied to systems utilizing realistic reaction pressures and treatments. Studies have utilized XPS [22, 57] to determine the oxidation state of H₂-treated CeO₂ particles or surfaces, but care must be taken interpreting these results as partially reduced CeO₂ is sensitive to O₂ exposure, even at low O₂ pressure and typical room temperatures. Furthermore, H₂ treatments performed in UHV systems are often performed at significantly lower pressures than those found in reactors or IR systems, the so-called "pressure gap". As demonstrated by Werner et al. [22], D₂ does not dissociate on CeO₂(111) until pressures well above those typically used in UHV systems, so care must be taken to reproduce experimental work in such controlled environments and the emerging ambient pressure XPS technique may be employed to address this pressure gap [38].

Hydrogenation of unsaturated C=C and C≡C bonds is well-known to take place over a variety of metal surfaces, especially precious metals such as Pd, which can easily adsorb dissociated H species. Exploration of the ability of metal oxides to perform hydrogenation reactions, however, resulted in the discovery of the ability of CeO₂ to hydrogenate unsaturated C≡C bonds [26]. This unexpected result, and the resulting highly selective hydrogenation behavior, have added significant value to understanding the interaction of H₂ with CeO₂ surfaces and bulk.

Vilé et al. initially reported very high selectivity for the gas-phase hydrogenation of propyne to propene (91% selectivity at 96% conversion) and ethyne to ethene (81% selectivity at 86% conversion) at 523 K in an excess of H₂ (30:1 H₂:alkyne ratio) without significant conversion of the alkene product to alkane. It was noted, however, that increasing the reaction temperature above 523 K resulted in a sharp increase in isomerization products, and increased oligomerization products were observed as catalyst contact times were increased. In contrast to the results from H-D exchange over CeO₂ reported previously, the hydrogenation activity of CeO₂ decreased as the temperature of H₂ pretreatment increased [25, 26]. This result suggests that oxygen vacancies are not favored for hydrogenation over ceria. It remains unclear if the exchangeable H-containing species on the reduced ceria after high-temperature H₂ treatment can be utilized for the selective hydrogenation reaction.

A study of the hydrogenation of a wider range of acetylenic compounds found that CeO₂ supported on TiO₂ is selective for the hydrogenation of the completely unsaturated C≡C bond in the presence of functional groups including olefinic bonds, benzene rings, alcohols, and sulfonates [32]. Results from microcalorimetry of 1-hexyne adsorption on 20% CeO₂ supported on TiO₂ indicated a very large initial heat of adsorption of 900 kJ mol^-1, which was ascribed to multiple dehydrogenation steps during adsorption of the 1-hexyne on the clean CeO₂ surface. At higher coverage, the measured adsorption energy was ~200 kJ mol^-1, more consistent with a single dehydrogenation step upon chemisorption. Studies using IR spectroscopy have demonstrated the presence of acetylide species on the CeO₂ surface under reaction conditions [23, 26], but the relevance of these species to catalysis is not consistent with comparable activity of CeO₂ for the hydrogenation of internal C≡C bonds. Instead, a proposed mechanism for hydrogenation involves π- or di-σ-bonding of the associatively adsorbed C≡C to the active site, where the resulting adsorbate can be easily hydrogenated [23].

The gas phase (but not liquid phase) hydrogenation rate of acetylenic compounds over these supported CeO₂ catalysts was fairly independent of CeO₂ dispersion, proposed to result from an increase in catalytically inactive surface O-defects (measured using XPS) as the dispersion was increased. Investigation of oriented CeO₂ surfaces for acetylene hydrogenation revealed that the (111) surface of octahedron-like nanoparticles were more active for ethylene hydrogenation at 473 K than nanocubes containing (100) surfaces, which are known to contain higher proportions of O-defected sites [97]. Aging the CeO₂ nanoparticles to produce more (110) surfaces only served to decrease the observed acetylene hydrogenation activity, and was consistent with an inverse relationship between hydrogenation rate and oxygen defect density on the surface of the catalyst (Fig. 5). While the presence of subsurface H has been reported to play a role in hydrogenation over noble metals in the past [98], the role of subsurface H species associated with O-defects in ceria remains to be revealed in the hydrogenation reactions.

Titania-supported ceria is very stereo-selective to cis- isomers of internally unsaturated products (e.g., the hydrogenation of 2-pentyne produced 93% cis-2-pentene at 100% conversion) during the hydrogenation of acetylenic compounds [32]. This behavior, and the observation of a parahydrogen-induced polarization (PHIP) effect in the hydrogenation products of alkynes over CeO₂, raises the question of a pairwise H₂-addition mechanism during hydrogenation by oxides [99]. Study of the PHIP effect in hydrogenation reactions over controlled CeO₂ nanoshapes reveals that pairwise addition of H₂ to propylene over CeO₂ is a thermally activated effect, which begins to take place at temperatures above 573 K independent of the CeO₂ facets exposed during hydrogenation of propene [100]. During the hydrogenation of propyne over CeO₂, a significant dependence of the hydrogenation rate on the CeO₂ nanoshape (resulting from variation in the exposed surface) was observed (Fig. 6). Nanorods dominated by (110) and (001) surfaces demonstrated a higher pairwise addition selectivity than nanooctahedra dominated by (111) surfaces, which are also known to be the more active surface for the hydrogenation reaction [101]. Nanocubes dominated by (001) surfaces demonstrated a pairwise addition selectivity that was similar to the nanooctahedra, but were less active for the hydrogenation of propylene. This result suggests that the mechanism for H₂ addition is dependent on the exposed surface and is expected to be dominated by a non-pairwise H-addition mechanism over CeO₂ (111) under conditions required to produce high selectivity to the unsaturated hydrogenation product.

Understanding hydrogenation of unsaturated C-C bonds over CeO₂ might provide insights leading to improved selective hydrogenation catalysts. Some studies have suggested the role of so-called "frustrated-Lewis-pairs" (FLPs) consisting of adjacent Ce³⁺ surface species, which lead to a very low barrier for H₂ dissociation and potentially hydrogenation activity [29, 33]. Other studies indicate hydrogenation rates over CeO₂ can be improved by doping other metals into the CeO₂ surface, perhaps modifying these FLPs. Recent studies suggest the potential to improve on hydrogenation activity by doping CeO₂ with Ga to create more hydride species during the dissociation of H₂ [85, 102], while others have directly demonstrated improved hydrogenation rates can be achieved with Ni-doped CeO₂ [27]. Further understanding of how these modifications drive faster hydrogenation over CeO₂ might be applied to produce even greater CeO₂-based selective hydrogenation catalysts in the future.

As discussed earlier, ceria itself has demonstrated superior selectivity for selective alkyne hydrogenation. In situ techniques (XPS, XAS, DRIFTS, STM) can be combined to obtain fundamental information about ceria-based catalysts under controlled and reaction conditions. However, the hydrogenation mechanism including the dissociation of H₂ and the reaction pathway is still incomplete. In this part, we are going to focus on computational modeling which can give an alternative view of the chemistry taking place over ceria surface and offer inaccessible information on the atomic scale starting from reviewing H₂ interaction, and then the alkyne hydrogenation reactions.

Understanding the interaction of hydrogen with ceria is important to develop further insights into different types of hydrogenation reactions. Two H₂ dissociation mechanisms over CeO₂ have been proposed from DFT calculations: homolytic and heterolytic cleavage. Homolytic dissociation of H₂ produces two O-H groups on the surface by binding with two surface oxygen sites, while heterolytic cleavage produces one Ce-H and one O-H species [103]. So far, hydrogen dissociation mechanisms have been proposed based on experimental evidence from infrared spectra that showed O-H and M-H bands for heterolytic and O-H bands for the homolytic cleavage. Hence, heterolytic cleavage has been believed to occur on nonreducible oxide such as MgO [104] or γ-Al₂O₃ [105], while homolytic cleavage has been occurred for reducible oxides such as TiO₂ [106] or CeO₂ [8, 36]. However, García-Melchor et al. [36] demonstrated that polar M-O bond from (111) CeO₂ surface can also cause heterolytic dissociation through DFT calculation using PBE+U method. As shown in Fig. 7, The homolytic product is thermodynamically stable but heterolytic products are kinetically easier to access. Moreover, the resulting hydride species from heterolytic cleavage can transfer and finally yields the homolytic products. This DFT prediction was later experimentally demonstrated by neutron spectroscopy that hydroxyl groups were observed on oxidized ceria while cerium hydride was present only after oxygen vacancies were created [38].

Through many studies aimed to understand the H-ceria interaction over a stoichiometric (111) surface, experimental studies suggested that oxygen vacancies play a pivotal role in the reactivity of hydroxyls [96]. Lu et al. [107] and Watkins et al. [108] adapted the surface including oxygen vacancies to investigate the surface reaction with H including H diffusion and H₂O formation. Lu et al. [107] demonstrated that H coupling, surface diffusion and bulk diffusion of H were all promoted at the reduced surface, while H₂O formation was less affected by oxygen vacancies by using on-site Coulomb interaction (DFT+U). On the other hand, Watkins et al. [108] adapted GGA and GGA+U function that incorporated with a local on-site Coulomb repulsion term to cover the questions that whether individual Ce⁴⁺ ions are reduced or whether electrons are delocalized over many metal ions. This study demonstrated a strongly exothermic character of H₂ chemisorption on the stoichiometric (111) surface and rapid water dissociation were occurred at vacancy sites compared with defect-free ceria surface.

As we mentioned earlier, not only oxygen vacancies but also different morphologies and crystal planes play an important role in their catalytic reactivity and selectivity. It is noted, however, that extensive modeling studies were focused on (111) facet due to its lowest surface energy and stable structure. Recently, Zhang et al. [109] studied the formation of oxygen vacancies via hydrogen reduction on (111), (110), (100) surfaces to reveal the effect of crystal-planes (Fig. 8). The heterolytic dissociation of H₂ was favored over the homolytic dissociation on each surface and, more importantly, (100) surface showed the best performance for H₂ activation. In addition, the role of frustrated Lewis pair over (110) and (100) plane were proposed by the same group regulating the oxygen vacancies [29]. When the oxygen vacancies were created on CeO₂ (110) and (100), two adjacent surface Ce cations proximal to the oxygen vacancy and one neighboring oxygen anion build the FLP sites. The heterolytic cleavage of H₂ is accelerated at FLP sites with a decrease of the activation barrier on CeO₂ (110) and (100) surface, which is attributed to the elongated distance and enhanced acidity and basicity of the Lewis pairs. At the FLP sites, H resource for hydrogenation could originate from hydride at the Ce site which is more active than a proton anchored on an oxygen atom on the surface.

A substantial number of DFT studies have been devoted to investigating reaction intermediates and mechanism of alkyne hydrogenation using different approaches and methods. Carrasco et al. [25] studied the acetylene hydrogenation mechanism over CeO₂ (111) with a DFT+U approach. The proposed mechanism was that acetylene adsorption leads to highly active β-C₂H₂ radical species that are hydrogenated to form C₂H₃ with a low energy barrier leading to superior performance for acetylene hydrogenation. Furthermore, the total hydrogenation over CeO₂ (111) from C₂H₅ required a large activation energy barrier which suggests total hydrogenation is an unfavorable reaction. In this study, however, the hydrogen dissociation step was proposed as the rate-limiting step. The intermediate species (β-C₂H₂) formation step is about 4 eV below than the reactant energy and the energy barrier for the final ethylene formation step is about 3 eV, which is kinetically infeasible.

An alternative mechanism, which involved concerted reaction between co-adsorbed H₂ and alkyne, was suggested by García-Melchor et al. [36]. This mechanism involved an adsorbed H atom as a hydrogen source for hydrogenation and suggested the formation of oligomers was prevented by restricting the adsorption and diffusion of alkyne molecules over the fully hydroxylated or reduced surface. It was noted, however, that this mechanism also involves a high barrier (~2 eV) and carries a heavy entropic penalty because of the weak adsorption energy of both alkyne and hydrogen species on the surface.

It is noteworthy that the experimental results indicate that the (111) face of CeO₂ is more active than the (110) and (100) faces for acetylene hydrogenation, but there is very little evidence presented to support the importance of oxygen vacancies on the ceria surface in the hydrogenation reaction. Recently, a hydrogenation mechanism over oxygen vacancies of CeO₂ (111) surface was suggested by Riley et al. [27]. The proposed mechanism involves the heterolytic dissociation of molecular hydrogen at oxygen vacancies. This causes the exposure of Ce sites followed by Ce-H formation on the surface. The hydride formation prevents the formation of over-stabilized β-C₂H₃ species, and the oxygen vacancy attributes to a decrease in the large kinetic barrier associated with its hydrogenation on the oxygen sites.

Given experimental results and the large formation energy for oxygen vacancy on stoichiometric CeO₂ (111) surface (2.57 eV), the oxygen vacancies are almost certainly involved in hydrogenation, but it is hard to generate oxygen vacancies on a defect-free ceria surface model. A great deal of work has been devoted to examining how the reactivity can be improved by enhancing oxygen vacancy formation. As a result, earlier theoretical works from Nolan et al. [110] and Watson et al. [111] have suggested the doping of divalent ions (e.g. Ni, Cu, Pt, Pd, and Zn) helps to form oxygen vacancies over ceria (111) and (110) surface. The work from Riley et al. [27] also studied the mechanism of acetylene hydrogenation over Ni-doped ceria. It was demonstrated that Ni doping lowered the activation energy barrier of the overall reaction. The doping of divalent metal plays the role of enhancing the formation of oxygen vacancies on ceria surface. It is notable that the Ni dopant was not suggested to directly participate in the hydrogenation reaction but instead to create O-vacancies. In addition to Ni-doped ceria, DFT study for hydrogenation of acetylene over Ga-doped CeO₂ was conducted by Guo et al. [34]. The Ga dopant was shown to promote the formation of surface oxygen vacancies, which create the frustrated Lewis pairs and accelerate heterolytic H₂ cleavage. However, the energy barrier of rate-limiting step in Ga-doped CeO₂ with higher Ga concentration was 0.21 eV lower than undoped CeO₂ (111), which suggests that Ga is not only working as a promoter to form oxygen vacancy, but also participating in the hydrogenation reaction with the formed Ga-H species.

Ceria-based materials are among the most interesting and widely investigated catalysts, which can be attributed to their excellent redox behavior as well as their versatile acid-base properties. This review surveyed specifically the recently discovered hydrogenation performance of ceria and the fundamental understanding of this unusual property of ceria through both experimental and computational studies. Regarding H₂ adsorption and activation, it is becoming clear that H₂ can dissociate over ceria surface to yield either homolytic (two OHs) or heterolytic products (one OH and one Ce-H) depending on the density of surface O-vacancies. Although the heterolytic dissociation mechanism has a lower activation barrier, the homolytic products are more thermodynamically stable and thus found on stoichiometric ceria surfaces. When used to catalyze alkyne hydrogenation, the reaction was proposed generally limited by the H₂ activation step while the subsequent hydrogenation of adsorbed alkyne species is facile by either the hydride or the hydroxyl species over ceria. As to the role of surface O-vacancies on ceria for the hydrogenation of alkynes, opposing views remain, and more recent computational work supports the active role of surface O-vacancies in activating H₂ and facilitating the hydrogenation pathways. It is expected that continued efforts are needed to further elucidate the reaction mechanisms of selective hydrogenation over ceria and the catalytic role of different hydrogen species as well as O-vacancies.

Despite the interesting hydrogenation performance of ceria, its reaction rate is fairly low, and the reaction temperature is still much higher (> 100 K) than that of typical metal catalysts. To improve the activity and lower the reaction temperature for hydrogenation reactions, further tuning of ceria is certainly needed. Lessons can be learned from the vast number of studies controlling the redox and acid-base properties of ceria as well as the recent investigations of hydrogenation reactions. It is evident that doping of ceria, a large body of work has been done in the past to improve the redox property, is one promising route to enhance the hydrogenation activity of ceria at lower temperatures. The FLPs concept introduced recently in ceria catalysis is certainly interesting and further modification and construction of different types of FLPs via dopants and surface modifications (hybrid systems for example) is expected to lead to more facile H₂ dissociation and hydrogenation steps over ceria. Experimental studies, especially operando studies, and theoretical modeling are paramount for mechanistic understanding of how these modifications of ceria can lead to facile hydrogenation under milder conditions such that practical ceria-based oxides may be designed for hydrogenation catalysis without the use of noble metals.

Notice

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).

Acknowledgments

This research was sponsored by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Chemical Science, Geosciences, and Biosciences Division, Catalysis Science program.