Reaction Pathway for Coke-Free Methane Steam Reforming on a Ni/CeO2 Catalyst: Active Sites and the Role of Metal–Support Interactions

Methane steam reforming (MSR) plays a key role in the production of syngas and hydrogen from natural gas. The increasing interest in the use of hydrogen for fuel cell applications demands development of catalysts with high activity at reduced operating temperatures. Ni-based catalysts are promising systems because of their high activity and low cost, but coke formation generally poses a severe problem. Studies of ambient-pressure X-ray photoelectron spectroscopy (AP-XPS) indicate that CH4/H2O gas mixtures react with Ni/CeO2(111) surfaces to form OH, CHx, and CHxO at 300 K. All of these species are easy to form and desorb at temperatures below 700 K when the rate of the MSR process is accelerated. Density functional theory (DFT) modeling of the reaction over ceria-supported small Ni nanoparticles predicts relatively low activation barriers between 0.3 and 0.7 eV for complete dehydrogenation of methane to carbon and the barrierless activation of water at interfacial Ni sites. Hydroxyls resulting from water activation allow for CO formation via a COH intermediate with a barrier of about 0.9 eV, which is much lower than that through a pathway involving lattice oxygen from ceria. Neither methane nor water activation is a rate-determining step, and the OH-assisted CO formation through the COH intermediate constitutes a low-barrier pathway that prevents carbon accumulation. The interactions between Ni and the ceria support and the low metal loading are crucial for the reaction to proceed in a coke-free and efficient way. These results pave the way for further advances in the design of stable and highly active Ni-based catalysts for hydrogen production.


INTRODUCTION
Methane steam reforming (MSR, CH 4 + H 2 O ⇄ 3H 2 + CO) is the main route for the large-scale industrial manufacture of hydrogen, primarily used for the synthesis of ammonia and methanol, among other commodities, 1 as well as the hydrocracking of long-chain hydrocarbons in petroleum refineries. 2 In a typical industrial reformer, the MSR reaction is carried out at 800−1000°C and 14−20 atm, with a H 2 O/ CH 4 ratio of ∼2.5. 1,3 Environmental concerns about air pollution and greenhouse gases have renewed the interest in using hydrogen as a clean energy carrier for automotive applications through its electrochemical conversion in fuel cell systems, which produces water as the only byproduct. However, the severe reaction conditions of industrial MSR result in elevated capital and operating costs, which are prohibitive for small-scale fuel cell applications. Several alternative reactions have been proposed, such as methane dry reforming and partial oxidation, but their lower H 2 /CO ratio compared to that of MSR makes them unfit for fuel cell applications that require high-purity H 2 . 4−6 Therefore, it is necessary to improve MSR technology to reduce heating and steam requirements and achieve cost-efficient H 2 manufacture. In this sense, the capability to operate fuel cells at ambient pressure 7 and the development of hydrogen-selective membrane reactors 8−10 represent an opportunity to increase the thermodynamically limited conversion imposed by the endothermicity of the MSR reaction, 11 allowing for both lower operating temperatures (500−600°C) and lower steamto-methane ratios while maintaining good H 2 yield. Commercial catalysts in industrial reforming units, typically consisting of nickel on magnesium or aluminum oxide supports, are designed to withstand high-temperature operations without losing strength and thus prioritize stability and thermal resistance over surface area. 1 In addition, they are prone to deactivation by coking, sintering, and sulfur poisoning. 12 Noble metals such as Pt, Rh, and Pd are also active for MSR but more expensive than Ni. 11 Therefore, the present challenge is to develop novel Ni-based catalysts to carry out the MSR reaction with high conversion at mild operating conditions for fuel cell applications. Among recently proposed alternatives, low-loaded Ni-impregnated CeO 2 catalysts have shown potential as promising candidates, showing improved coking and sintering resistance and excellent performance in experiments carried out at 600°C and ambient pressure. 13,14 A complete understanding of the MSR reaction mechanism over ceria-supported Ni catalysts, which includes the identification of the active sites and the determination of the relevant reaction pathways, remains elusive, but it is essential to be able to modify the catalyst to enhance activity and selectivity.
Insights from density functional theory (DFT) calculations on model Ni-based MSR catalysts have so far been limited to extended Ni surfaces. It has been postulated that the activation of CH 4 determines the overall reaction rate 15,16 because its C− H bonds are very stable (440 kJ/mol), 17,18 and pure metal surfaces tend to show low reactivity toward methane. 19 However, experimental and computational studies have shown that the reactions of carbonaceous species with oxygen to form the C−O bond also involve high energy barriers and could therefore be rate-controlling. 20−26 Furthermore, in previous combined computational and in situ spectroscopic studies, it was shown that well-dispersed small Ni nanoparticles supported on a nonreduced CeO 2 surface can in fact activate CH 4 at room temperature, with calculated energy barriers up to 80% lower than those for extended nickel surfaces. 27−31 This highlights the need to consider both the effect of the nature of the support and the metal loading to fully understand the mechanism governing the MSR reaction over supported metal catalysts, which is necessary for the development of improved catalytic systems.
In general, dispersed metal nanoparticles on oxide surfaces tend to be more reactive than the individual components, showing great potential as novel catalytic materials. 32 In lowloaded CeO 2 -supported Ni catalysts, nickel is stabilized as small particles in which the Ni atoms in direct contact with ceria are partially oxidized as a consequence of strong metal− support interactions, 27,33,34 resulting in important changes in the chemical and catalytic properties of these systems, particularly to perform C−H and O−H bond cleavage. 27,28,30,35,36 Furthermore, the easier reducibility of the ceria support allows it to act as an oxygen reservoir, 37 providing unique reaction pathways such as the reverse spillover of oxygen from ceria to metal sites, which has been experimentally observed for a variety of ceria-supported metal catalysts, including Ni/CeO 2 . 38−43 Therefore, the observed superior decoking activity of ceria-supported Ni catalysts for MSR could be ascribed to a mechanism involving the oxygen supply from the support promoting carbon removal as CO, 44,45 in which the role of water as one of the reactants would be the refilling of the oxygen vacancies generated in the reverse spillover step. However, water-mediated carbon removal has also been discussed in the context of steam reforming of CH 4 . 46 Whether carbon removal is assisted by oxygen from the support, from H 2 O, or from both is an essential question in the understanding of the MSR reaction mechanism.
Furthermore, since both reactants in the MSR reaction over Ni/CeO 2 catalysts adsorb at Ni sites and are generally in H 2 O/ CH 4 ratios higher than 1, their potential competition for Ni sites should also be addressed. In this regard, CH 4 conversion in the steam reforming reaction over Ni-impregnated Zr-doped CeO 2 catalysts was found to continuously increase with H 2 O content, suggesting that whatever the competition, it was not detrimental to the reaction. 14 It has been previously found that small coverages of nickel on CeO 2 (111) produce surfaces that are able to catalyze the MSR process at temperatures above 500 K with high activity and low propensity to deactivation by coke deposition. 29 This is a remarkable catalytic performance. Here, using a combination of ambient-pressure X-ray photoelectron spectroscopy (AP-XPS) and molecular modeling based on density functional theory (DFT), we present a comprehensive study of the MSR reaction on the surface of model Ni/CeO 2 (111) catalysts and compare with results reported for the extended Ni(111) surface in the literature. 21 This pathway presents much lower barriers than the one involving C oxidation with lattice oxygen from the ceria support and is promoted by the easy formation of OH groups through the barrierless dissociative adsorption of water at the Ni−CeO 2 interface. The results provide molecular insight into the interplay between C and OH species in the steam reforming of methane on low-loaded Ni/CeO 2 catalysts for which metal−support interactions are crucial to bind and activate methane and water.

METHODS
2.1. Experiments of Ambient-Pressure XPS. The ambient-pressure XPS studies examining the interaction of CH 4 /H 2 O gas mixtures with the Ni/CeO 2 (111) surfaces were performed using instruments located at the Chemistry Division in Brookhaven National Laboratory (BNL) and at the Advanced Light Source (ALS) in Berkeley. 27−30 In both instruments, the Ni/CeO 2 (111) surfaces were prepared and characterized following standard procedures. 27−29 Ce metal was first evaporated onto a Ru(0001) substrate at 700 K under a background pressure of 5 × 10 −7 Torr of O 2 , and then the sample was annealed at 800 K for a period of 10 min at the same O 2 pressure. The CeO 2 (111) films were estimated to be ca. 4 nm thick (≈10 layers of O−Ce−O) based on the attenuation of the Ru 3d XPS signal. Ni was vapor-deposited on the as-prepared ceria films, and the admetal coverage was estimated by the attenuation of the Ce 3d XPS signal. 27−29 The Ni/CeO 2 (111) surfaces were exposed to CH 4 , H 2 O, and CH 4 / H 2 O mixtures at temperatures between 300 and 700 K.
The AP-XPS instrument at BNL was a SPECS AP-XPS chamber equipped with a PHOIBOS 150 EP MCD-9 analyzer. Mg Kα radiation was used to collect the Ni 2p and Ce 3d spectra of the Ni/CeO 2 (111) samples under exposure to the reacting gases. The binding energies in these AP-XPS spectra were calibrated using as a reference the strongest Ce 4+ 3d feature located at 916.9 eV.
At the ALS, the AP-XPS experiments were performed in beamline 9.3.2, which was equipped with a VG Scienta R4000 HiPP analyzer. On exposure of Ni/CeO 2 (111) to the reacting gases, the O 1s region was probed using a photon energy of 650 eV, and the C 1s, Ni 3p, and Ce 4d regions with a photon energy of 490 eV. The energy resolution in the synchrotron experiments was ∼0.2 eV. The Ce 4d photoemission lines were used for binding energy calibration based on the 122.8 eV satellite features. No evidence was found for the existence of beam damage in these AP-XPS studies.
2.2. Models and Computational Details. A Ni 13 cluster adsorbed on the CeO 2 (111) surface with a (3 × 3) periodicity 34 was used as a representative model of low-loaded ceria-supported nickel catalysts, hereafter referred to as Ni 13 · CeO 2 ( Figure S1). The size of the Ni 13 cluster is comparable to that of Ni nanoparticles of model Ni/CeO 2 catalysts in experimental studies 28,29 and has metallic Ni 0 and oxidized Ni 0.55+ sites, both reported to be present at steam reforming conditions. 13 The (3 × 3) CeO 2 (111) surface was modeled using a supercell with the calculated ceria bulk equilibrium lattice parameter of a 0 = 5.485 Å, with six atomic layers (two O−Ce−O trilayers, TLs) separated by at least a 12 Å thick vacuum layer.
Calculations were performed within the spin-polarized density functional theory (DFT) framework as implemented in the Vienna Ab initio Simulation Package (VASP). 55,56 The Kohn−Sham equations were solved within the generalized gradient approximation (GGA), with the Perdew−Burke− Ernzerhof (PBE) exchange−correlation functional. 57 We treated explicitly the Ce(5s 2 5p 6 6s 2 5d 1 4f 1 ), Ni(3p 6 4s 2 3d 8 ), O(2s 2 2p 4 ), and C(2s 2 2p 2 ) valence electrons using a planewave basis with a cutoff energy of 415 eV, whereas the core electrons were represented with the projector-augmented wave (PAW) method. 58,59 Total energies were calculated with a precision of 10 −6 eV. Strong correlation effects due to charge localization were considered with the DFT + U approach within Dudarev's scheme 60 to compensate for the selfinteraction error. 61−65 The U eff parameter was set to 4.5 eV for the Ce(4f) states. 66,67 Long-range dispersion corrections were considered within the DFT-D3 approach. 68,69 The oxidation state of a given Ce ion (Ce 4+ or Ce 3+ ) was determined by considering its local magnetic moment, which can be estimated by integrating the site-and angular momentum projected spin-resolved density of states over spheres with radii chosen as the Wigner−Seitz radii of the PAW potentials. The magnetic moments of the Ce 4+ (4f 0 ) and Ce 3+ (4f 1 ) ions are 0 and ∼1 μB, respectively. As for the oxidation state of the Ni atoms in the supported clusters, using Bader's atom-in-molecule approach, 70,71 we observed that only those Ni atoms bound to surface oxygen from the ceria support are partially oxidized. The average oxidation state of these Ni atoms was calculated as the total number of electrons transferred to the ceria support divided by the number of atoms in direct contact with the support. Full relaxation of atomic coordinates was allowed for both the Ni atoms and the Ce and O ions located in the uppermost TL, and forces were converged to 0.02 eV/Å. The ions in the bottom TL were kept fixed in their bulk positions. The Brillouin zone was sampled with a (2 × 2 × 1) k-point mesh using the Monkhorst−Pack scheme. 72 Transition state (TS) structures were located using the climbing image nudged elastic band (CI-NEB) method 73 with forces converged to 0.05 eV/Å. Harmonic frequencies were calculated for all TS structures using a finite-difference method, with displacements of ±0.015 Å in the coordinates of the adsorbates and the Ni atoms, to verify the existence of a single imaginary frequency.

RESULTS
3.1. Surface Chemistry of the MSR Process on Ni/ CeO 2 (111): An AP-XPS Study. Previous results of AP-XPS indicate that methane dissociates on Ni/CeO 2 (111) surfaces at room temperature (300 K) to yield surface CH x 27,28 and that part of the adsorbed CH 4 undergoes full decomposition that produces C atoms that react with O centers of the support to generate CO x groups (CO 2 or CO 3 species). Maximum reactivity was observed on systems that had Ni coverage below 0.2 monolayer (ML). These systems were able to catalyze the MSR process at temperatures above 500 K with high activity and low propensity to deactivation by coke deposition. 29 Figure S2 shows Ni 2p and Ce 3d XPS spectra collected while exposing a Ni/CeO 2 (111) surface to 20 mTorr of methane at 300 and 700 K. At room temperature, the reaction of methane with the surface does not change the oxidation state of Ni or Ce in the system, but CH x and CO x groups are deposited on the catalyst. 27,28 The reaction is observed only when Ni is added to ceria, but the total coverage of the CH x and CO x groups is larger than that of nickel, suggesting that methane dissociates on Ni or the Ni−ceria interface and then a part of the C-containing species migrate to the ceria. 27,28 These adsorbed species are not stable at temperatures above 500 K, but the reaction with methane is very fast, and at an elevated temperature of 700 K, the decomposition products of methane reduce Ni 2+ to Ni 0 and a part of Ce 4+ to Ce 3+ , which is accompanied by the formation of lattice oxygen vacancies. Therefore, during methane steam reforming over Ni/CeO 2 at T ≥ 700 K, the dissociation of water on the O vacancies closes the catalytic cycle. 16,17 However, as discussed below, the Ni 2+ → Ni 0 and Ce 4+ → Ce 3+ reductions were not observed under a mixture of methane and water.
The bottom traces in Figure 1 show O 1s XPS spectra collected while exposing a Ni/CeO 2 (111) surface to 100 mTorr of H 2 O at different temperatures. The peak at around 535 eV results from H 2 O gas. Features at around 531.8 eV denote the dissociation of the adsorbate and the deposition of OH groups on the surface. 36 At 300 K, the total coverage of OH on the surface was in the range of 0.4−0.6 ML. The OH groups were bound to Ni and Ce cations on the substrate. The formation of Ni−OH bonds leads to a binding energy shift in the position of the Ni 2p core levels, whereas the Ce 3d core levels are not significantly affected by the dissociation of the water molecules ( Figure S3). In Figure 1, there is an attenuation of the signal for surface OH groups when the temperature is increased from 300 to 450 K. Thus, the OH groups are easily formed and they do not bind strongly to the metal/oxide substrate, which are good characteristics for intermediates in a catalytic process. Figure 2 displays C 1s XPS spectra collected while exposing Ni/CeO 2 (111) to a CH 4 /H 2 O mixture at 300−700 K. The pristine surface exhibits a broad feature from 293 to 288 eV attributed to the Ce 4s core level. This feature overlaps with the signal seen for the surface CO x species formed by the full dissociation of methane and the reaction of carbon with surface oxygens. 27,28 The interaction of the CH 4 /H 2 O gas mixture with Ni/CeO 2 (111) yields CO x , CH x O, and CH x groups on the surface. The CH x O species were not observed when the Ni/CeO 2 (111) system was exposed to only methane. 27,28 Therefore, they result from the direct reaction of OH and CH x groups on the surface, pointing to an associative reaction pathway for the MSR process, which is in good agreement with the DFT results described in the next section. At 700 K, the CO x , CH x O, and CH x species disappear from the catalyst surface. Thus, they are reaction intermediates that can be formed and removed easily, and no carbon deposition is observed on the catalyst surface.
An analysis of the O 1s XPS spectra collected under a gas mixture of CH 4 36 In Figure 3, the signal for CH x O/OH is quite strong at 300 K, with the total coverage for the CH x O/OH groups being in the range of 0.6−0.8 ML. But these adsorbed species have limited stability, and their features decrease when the surface is heated to 450 K. At 700 K, CH x O is completely absent (Figure 2), and thus only a very small concentration of OH groups remains on the catalyst surface (Figures 1 and 3). The presence of these adsorbed OH groups is important because any CH x species generated by methane dissociation can react with them to yield the products of the MSR process. Furthermore, Ni 2p and Ce 3d XPS spectra recorded under a mixture of methane and water ( Figure S4) do not show any evidence for Ni 2+ → Ni 0 and Ce 4+ → Ce 3+ reductions, as seen in the case of pure methane ( Figure S2). This is valid for all of the temperatures examined. Therefore, the ceria lattice oxygen is probably not involved in the MSR process on this catalyst, and the AP-XPS results support an associative mechanism that involves the formation of a CH x O intermediate, in agreement with the predictions of the DFT calculations discussed below.
3.2. Reaction Pathway of the MSR Reaction on Ni/ CeO 2 (111): A DFT Study. Using a Ni 13 cluster supported on a flat CeO 2 (111) surface (Figures 4a and S1), we investigated the surface chemistry of the MSR process on Ni/CeO 2 . The Ni 13 cluster reduces the ceria support upon adsorption with the formation of five Ce 3+ ions. The calculated electronic structure of the Ni 13 ·CeO 2 system shows that the charge transfer by Ni atoms to the support is solely from the nine atoms in the interfacial layer, which are partially oxidized (9× Ni 0.55+ ), whereas four neutral Ni atoms (4× Ni 0 ) are above them (Table S1), in line with previous results. 34 Hence, two types of Ni sites exist for adsorption and activation of reactants on the Ni 13 ·CeO 2 model catalyst, namely, oxidized interfacial sites and metallic terrace sites, hereafter referred to as Ni 13 ·i and Ni 13 ·t, respectively. The reaction pathway for methane steam reforming over the Ni 13 ·CeO 2 model catalysts is discussed below.
3.2.1. CH 4 Activation and Dehydrogenation. The cleavage of the first C−H bond through the dissociative adsorption of CH 4 has generally been considered a rate-controlling step for the MSR reaction on Ni-based catalysts, based on the observed low reactivity of Ni surfaces toward methane 19 and the high activation barrier for the CH 4 → CH 3 + H reaction on Ni(111) obtained in DFT studies. 25,26 However, it was recently shown that methane activation occurs even at 300 K on small ceria-supported Ni particles, 27−30 indicating much  lower activation barriers than on the extended Ni surface. The first step in CH 4 activation involves its molecular adsorption, which is very weak on the CeO 2 (111) surface, 74−77 suggesting that methane should dissociate over Ni sites instead, as shown by XPS spectra of the pristine CeO 2 and the Ni/CeO 2 surfaces under 1 Torr of methane. 27,29 Accordingly, we considered the adsorption and dehydrogenation of CH 4 on Ni sites of the Ni 13 ·CeO 2 system. CH 4 adsorption on the Ni 13 cluster is stronger by about 0.2 eV than that on the extended Ni(111) surface for both the interfacial Ni 13 ·i and terrace Ni 13 ·t sites (Figures S5, S7, and S8). Moreover, the CH 4 molecule comes closer to the surface of the Ni cluster, with C−Ni distances of 218 (Ni 13 ·i) and 228 pm (Ni 13 ·t), compared to 315 pm on Ni(111) 30 (Figure 4b). Inspection of the atom-and orbital-projected density of states (PDOS) onto the d-states of the Ni 13 ·i and Ni 13 ·t sites where CH 4 adsorbs (Table S2) reveals that the d xz states become less occupied upon adsorption of the Ni 13 cluster onto the ceria support. The consequence of such an effect is that the Pauli repulsion to the methane's frontier orbital is reduced, enabling the molecule to come closer to the surface. The states are then occupied upon CH 4 adsorption as measured by the decrease in the number of empty d xz states on both Ni 13 ·i and Ni 13 ·t sites in the CH 4 /Ni 13 ·CeO 2 system. As a result of the close approach of CH 4 , the C−H bond pointing toward the surface becomes preactivated, resulting in an increase in the bond length from 110 pm in the gas-phase CH 4 molecule to 115 and 113 pm at the Ni 13 ·i and Ni 13 ·t sites, respectively (Figure 4b).
Note that upon methane adsorption on the Ni(111) surface, the C−H bond is not stretched, 23,30 and the occupation of d xz states remains unchanged (Table S3).
The first dehydrogenation step of these preactivated CH 4 molecules takes place with low activation energy barriers of 0.34 eV at the Ni−CeO 2 interface (Ni 0.55+ ) and 0.36 eV at the Ni terrace (Ni 0 ) (Figure 4c). We note that although interfacial Ni sites are partially oxidized and Ni atoms in the second layer of the cluster have a metallic character, the barriers are comparable. Hence, low-temperature CH 4 activation on lowloaded Ni/CeO 2 systems is expected to take place both at the perimeter of the Ni−CeO 2 interface and on Ni atoms with no direct bonds to the support. The latter, however, is not the same as surface Ni atoms in Ni(111) with an activation barrier for the CH 4 → CH 3 + H reaction that is larger by 0.56 eV (0.90 eV, Figure S5). The combined effects of metal−support interactions and low metal loading contribute to the improved catalytic activity of Ni/CeO 2 compared to Ni(111). Importantly, ceria-deposited small Ni clusters exhibit higher local fluxionality than Ni(111), i.e., Ni−Ni bonds are less rigid for the metal atoms in the clusters and can lead to stronger stabilizing interactions and lower activation energies on catalytic pathways (cf. the change in the average Ni−Ni bond length upon CH 4 adsorption on Ni 13 ·t, +11.7 pm, and on Ni(111), +0.2 pm; Table S4). Further dehydrogenation steps (CH 3 → CH 2 → CH → C) also proceed with relatively low barriers on the supported Ni 13 cluster. The activation barriers (E a ) for the elementary steps involved in CH 4 dehydrogenation are shown in Figure 4c. The corresponding reaction energies (ΔE) and a comparison with previously published values for the Ni(111) surface are shown in Table S5, whereas the structures of the initial, final, and transition states are shown in Figures S7 and S8. The highest energy barrier at the Ni 13 ·t sites corresponds to CH 3 → CH 2 + H dehydrogenation (0.72 eV), whereas at the Ni 13 ·i sites, it is associated with CH → C + H dehydrogenation (0.72 eV). Similar to the abovediscussed case of the first H abstraction from CH 4 , the comparison with the extended Ni(111) surface (Table S5) reveals that the last H abstraction from CH on Ni 13 ·CeO 2 has an activation barrier that is smaller by at least 0.6 eV than that on the extended surface, whereas the barriers for the second and third dehydrogenation steps in both systems are comparable.
The binding of isolated CH x species (x = 0−3) on Ni 13 · CeO 2 is stronger than that on Ni(111) (Table S6), with the largest difference of about 1 eV for the C atom. At both Ni 13 ·i and Ni 13 ·t sites, CH 3 binds on a twofold bridge position and, although CH 2 also binds on a bridge site upon its formation, it changes to a threefold site after the removal of the co-adsorbed H atoms (Figures S7 and S8). Note that on the extended Ni(111) surface, CH 3 and CH 2 adsorb on a threefold facecentered cubic (fcc) site. 53 CH and C species bind to four Ni sites of the Ni 13 cluster producing significant structural distortion ( Figures S7 and S8), which might explain their higher stability compared to the more rigid Ni(111) surface. The fourfold binding is ascribed to the higher degree of unsaturation of the CH and C species, and it is also seen in the Ni(111) surface, where CH and C adsorb on hexagonal closepacked (hcp) hollow sites (instead of fcc), enabling their coordination with an additional Ni atom in the subsurface layer. 53 In spite of the easy formation and increased stability of C atoms on Ni sites of the Ni 13 ·CeO 2 system, a low tendency toward carbon deposition is observed in the AP-XPS experiments performed over model Ni/CeO 2 catalysts (cf. Figure 2), as well as in prior experimental studies. 28 In this regard, it has been argued that carbon deposition on extended Ni surfaces depends strongly on the concentration of oxygen on the catalytic surface. 53 In an oxygen-lacking environment, the interaction between CH x intermediates and oxygen does not occur at a rate sufficient to convert the carbon produced from CH 4 dehydrogenation to CO, thus resulting in carbon accumulation and subsequent deactivation of the catalysts. 53,54 We show below that the Ni 13 ·CeO 2 surface provides unique sites and pathways suitable to convert carbon to CO in the MSR reaction with barriers below 0.9 eV.
3.2.2. H 2 O Dissociative Adsorption. H 2 O dissociates at the Ni−CeO 2 interface through a virtually barrierless process, as previously shown for ceria-supported Ni single atoms and planar Ni 4 clusters. 29,36 The dissociative adsorption involves sites from both the Ni cluster and the CeO 2 surface, with the OH group adsorbing monodentate (OH m ) on Ni 13 ·i and the dissociated proton on lattice oxygen from the ceria support (H s ) ( Figure 5). A hydrogen bond between OH m and H s is formed, stabilizing the structure (d(OH m −H s ) = 179 pm). On the other hand, the dissociation of H 2 O on terrace sites of the Ni 13 cluster does not involve lattice oxygen, producing a bidentate OH species on Ni 13 ·t (OH t ) and a H atom nearby on the cluster, and it is hindered by a barrier of 0.79 eV.
For comparison, the dissociative adsorption of water on Ni(111) is significantly less exothermic with ΔE = −0.41 eV, 24 and it is hindered by a high barrier of 0.90−1.11 eV, 21,23,24,36 whereas on the nondefective CeO 2 (111) surface, no true dissociation occurs and the molecular state coexists with an OH-pair-like configuration that easily recombines and desorbs at the reaction temperature. 36,78 Therefore, these calculations show that H 2 O dissociates preferentially over the Ni−CeO 2 interface, undergoing barrierless activation and easily producing adsorbed OH groups.
3.2.3. CO Formation and Carbon Removal. Since chemisorbed CH 4 on Ni/CeO 2 easily loses all its hydrogens (Figure 4c), we first explore the oxidation of carbon on interfacial Ni 13 ·i sites via its direct reaction with surface lattice oxygen (O s ), resulting in a CO molecule adsorbed on the Ni cluster and an oxygen vacancy on the ceria support, which could later be reoxidized by water. This type of Mars−van Krevelen redox cycle has been suggested to be the route for many catalytic reactions involving CeO 2 . 79−81 We note that C atoms adsorbed on terrace sites can easily migrate to the Ni− CeO 2 interface with a barrier of 0.37 eV ( Figure S8) and therefore they could be available for oxidation by lattice oxygen, even if CH 4 activation and dehydrogenation take place on terrace sites. The formation of CO through the direct reaction of C with lattice oxygen has a very high barrier of 2.17 eV (Figure 6), and thus this pathway is deemed unlikely to take place. Instead, the adsorbed carbon atom could react with O or OH species chemisorbed on the Ni cluster to directly form CO from C + O or an oxidized COH intermediate that could then dehydrogenate to CO, which would be in line with the results of the AP-XPS study. Therefore, we investigated next the energy barriers involved in the formation of CO through these pathways.
Regarding the existence of chemisorbed O species, as discussed above, H 2 O dissociates (OH m + H s , Figure 5) at the Ni−CeO 2 interface through a practically barrierless process. The monodentate OH m species can migrate to a bidentate position OH b (I → II in Figure 7, cf. Figure S9) to then dissociate into O and H species on the Ni cluster with a barrier of 1.33 eV (II → III), which is close to that reported for the Ni(111) surface (1.16−1.31 eV). 21,23,24 We note that the possibility of forming OH and O species at interfacial Ni sites by migration of lattice O from the support to the Ni cluster (oxygen reverse spillover) has also been considered (red pathway in Figure 7). In this pathway, a surface lattice oxygen ion migrates to the cluster, leaving an oxygen vacancy on the   form O species chemisorbed on the Ni cluster and thus they are not easily available for the direct oxidation of C atoms. Moreover, the direct formation of CO from C and O atoms on the Ni cluster is hindered by a high barrier of 1.47 eV (cf. R7 in Table S7), further discouraging a pathway involving the direct oxidation of carbon with chemisorbed oxygen.
However, the reaction of C with OH groups readily available from the dissociation of H 2 O at the Ni−CeO 2 interface produces the COH intermediate with an energy barrier of 0.89 eV, which is significantly lower than that of the abovementioned reaction of C with chemisorbed O (1.47 eV). This may be related to the significantly lower binding of the OH species compared to the O species (−3.97 and −5.86 eV, respectively; Table S6). The COH formation on the Ni 13 cluster has also a lower barrier than those reported in the literature for Ni(111) (1.14−1.46 eV). 21,24,26 Overall, these results allow us to propose a reaction pathway for the production of CO via the direct reaction of C with OH groups through a COH intermediate (Figure 8), and thus O species chemisorbed on the Ni cluster would not be required to oxidize carbon. The first step (I in Figure 8) corresponds to the barrierless activation of water at the Ni−CeO 2 interface near a C atom on Ni 13 ·i, with ΔE = −1.75 eV. Next, C and OH react to form the COH intermediate in an endothermic step (ΔE = 0.43 eV) with an energy barrier of 0.89 eV (I → II in Figure 8). Finally, a similar barrier of 0.88 eV must be overcome to dehydrogenate the COH intermediate and produce CO (II → III in Figure 8).
Structures and energies of all of the states involved in the COH intermediate pathway are detailed in Figure S10. Reaction and activation energies are summarized and compared with literature values for Ni(111) in Table S7. We note that the reaction of CH x (x = 1−3) and OH to form CH x OH intermediates was also considered, but the barriers (E a ≥ 0.81 eV) are larger than those of the dissociation of the CH x species (E a ≤ 0.72 eV), as shown in Figure S11. This indicates that CH x could preferentially dehydrogenate fully to C and then react with OH, in line with the COH pathway presented above. However, it should be noted that adsorbate coverage effects can slow down the rate of CH x dehydrogen-ation steps, particularly at low temperatures (≤450 K) at which a higher coverage of adsorbates (CH x , OH, H) is expected, reducing the availability of free active sites for the decomposition of methyl species. Thus, at such temperatures, various CH x (x = 1−3) species can coexist on the catalyst surface (cf. Figure 2).
In a final step, CO and H 2 must desorb to close the catalytic cycle of the endothermic MSR reaction. We observe that the XPS spectra do not show adsorbed CO ( Figure 2); therefore, the desorption of CO should not be too difficult. It must be noted that gradient-corrected exchange−correlation functionals, such as PBE, overestimate the binding of CO on metal surfaces; 86−88 therefore, desorption of the molecule is predicted to be more difficult than it actually is. For instance, the calculated CO adsorption on Ni(111) (Table S6) is overestimated by about 0.5 eV compared to the experimental value. 89 On the other hand, H 2 molecules can be easily formed from the bonding of two H species chemisorbed on the Ni 13 cluster, with ΔE = 0.47 eV and a barrier E a = 0.62 eV (Figure  Energies of all states are referred to those of the clean surface and gas-phase species according to the stoichiometry of the MSR reaction. S12). As for the H species adsorbed on the surface lattice oxygen (H s ), which are formed by water dissociation at the Ni−CeO 2 interface, they would have to migrate from the support to the cluster before reacting with other H species to form H 2 . Direct migration is hindered by a high energy barrier of 1.48 eV, but the process becomes easier when assisted by additional water dissociated at the terrace sites of the Ni 13 cluster, providing a pathway for which the highest barrier is E a = 0.75 eV ( Figure S12).
The results presented above reveal that the activation of CH 4 and H 2 O and the formation of H 2 occur with relatively small energy barriers of about 0.7 eV on Ni/CeO 2 , and the oxidation of carbon through an associative pathway involving the COH intermediate takes place with activation energy below 0.9 eV. This is quite different from the case of the extended Ni(111) surface, for which both the activation of methane and water and the oxidation of carbonaceous intermediates to form CO involve high energy barriers (≥1 eV, cf. Table S7). 21,25 Furthermore, the barrierless activation of water at the Ni−CeO 2 interface allows for a higher supply of OH species and, consequently, lower steam-to-methane ratios are required to achieve the same OH formation rate as that for extended Ni surfaces and traditional Ni catalysts supported on aluminum or magnesium oxides, for which H 2 O dissociation is not easy.

CONCLUSIONS
We conclude that the selectivity of the MSR reaction can be steered to prevent coke formation by choosing the "right" metal−oxide combination and controlling the effects of metal loading. Well-dispersed Ni nanoparticles supported on ceria are active and efficient MSR catalysts. The interactions between the reducible support and the small-sized nanoparticles are crucial for facile methane dehydrogenation and water dissociation at the Ni−CeO 2 interface. Studies of AP-XPS indicate that CH 4 /H 2 O gas mixtures react with Ni/ CeO 2 (111) surfaces to form OH, CH x , and CH x O at 300 K. All of these species are easily formed and desorb at temperatures below 700 K when the rate of the MSR process is accelerated. In line with the experiments, DFT calculations reveal a MSR reaction pathway with barriers below 1 eV that would enable reduced operating temperatures. The path proceeds via the formation of a COH intermediate species from chemisorbed C atoms and OH groups, hindering carbon accumulation and catalyst deactivation even with a low steamto-methane ratio in the reactant feed. Water also facilitates the removal of hydrogen from the support at the Ni−CeO 2 interface.
In summary, when undertaking the rational design and improvement of novel ceria-supported metal catalysts for the MSR reaction, it has to be taken into account that both CH 4 and H 2 O activation steps occur very easily on low-loaded Ni/ CeO 2 , and therefore the goal should be to modify the catalyst to decrease the barrier for the oxidation steps to form CO, for which one possibility may be to use Ni-based bimetallic catalysts. 24 The properties of the ceria support material may also be chemically modified by, for example, doping with zirconium to improve its oxygen storage/transport characteristics that promote MSR pathways involving the participation of lattice oxygen, which was found to be unlikely in this study using pure ceria. This is in line with the recently observed promising performance of Ni catalysts supported on Zr-doped ceria for MSR at low temperatures. 14, 90 We anticipate that these strategies could represent an opportunity to further improve Ni/CeO 2 catalysts and to guide the design of novel catalysts with lower kinetic barriers for the MSR reaction.
Ni 2p and Ce 3d spectra of AP-XPS for the interaction of methane, water, and methane/water mixtures with Ni/CeO 2 (111) catalysts; Ni 13  ■ ACKNOWLEDGMENTS