■ INTRODUCTION

Bulk metal oxides are essential to a range of industries where their tunable surface and electronic structure facilitate their use as heterogeneous catalysts, in biomedical devices, and as semiconductors. [1][2][3] In recent years, efforts have shifted to capitalizing on surface atoms and how the change in a chemical environment, i.e., fewer nearest neighboring atoms, can increase their catalytic properties as compared to their bulk counterparts. 4 Such strategies have included isolating a smaller nanooxide size regime to increase the amount of surface atoms or a specific morphology to access desired surface facets with higher surface energies to increase their catalytic activity. 5,6 However, great care is needed to minimize the polydispersity in these systems to confidently assign structure-property relationships given the surface heterogeneity. 7 Other emerging strategies to access nano-sized bulk oxides tailored in size or shape include the growth of nano-sized oxides with templating mesoporous materials 8 or on substrates such as thin films. 9 Another strategy to access discrete nanooxide materials is their incorporation into structural building units in highly ordered hybrid materials, such as metal-organic frameworks (MOFs). 10 Shifting to a smaller size regime with discrete, crystallographically defined clusters enables more confident assignments of structure-activity relationships. 11 Thus, it is imperative to expand and subsequently interrogate the available library of metal-oxo clusters. 12 Of the well-studied metal oxides, metal-oxo clusters that resemble cerium dioxide present a fascinating platform for further research given the facile storage and release of oxygen within ceria to form reactive oxygen vacancy defects. 13 Enabled by the favorable redox couple of Ce 4+ /Ce 3+ , surface O atoms can readily be abstracted depending on reaction conditions and directly participate in reactions through a Mars-Van Krevelen mechanism. 14,15 Moreover, subsurface atoms can facilitate the transport of O vacancies through the lattice, rather than remaining innocent to the chemistry occurring at the surface. 16 The reducibility of Ce 4+ is greatly affected by and can be readily tuned through its local environment 17 and geometry. 18 Thus, ceria-based catalysts and catalyst supports have become indispensable within heterogeneous catalysis. 19 Inspired by their ubiquity, a library of Ce-oxo clusters has been reported in recent years, often focusing on the synthesis to control the size and/or surface accessibility. [20][21][22][23][24] Most recently, the Christou group modified a solvent used to facilitate the growth of intermediate Ce species to access a Ce 100 -based oxo cluster, the highest nuclearity Ce-oxo cluster reported to date. 25 Additionally, work from the Nyman group expanded the metal oxysulfate torous motif, first reported in U and Zr, 26,27 to Ce while providing insights to their assembly. 28,29 In these crystallographically defined systems, every Ce atom is exposed to the surface in a donut-like structure (Figure 1) with a range of monomeric metal linkage moieties interconnecting the clusters to form one dimensional (1D) channels. Recent work from the Farha and Nyman teams further expanded the series to include a Ce 70 torus structure interconnected by Ce monomers and examined the photocatalytic and radical scavenging character of these clusters. 30 To date, most discrete Ce-oxo species have been studied in photocatalytic settings, given that the monodentate alkyl or aromatic capping agents present in other reported Ce-oxo clusters limit their thermal stability. 24,31 However, the limited reports of utilizing discrete Ce-oxo clusters for thermal-based catalysis starkly contrasts the substantial volume of studies investigating bulk ceria as a catalyst and/or catalyst support in applications such as CO oxidation or CO 2 hydrogenation. 32,33 Thus, the Ce 70 torus family, which instead features capping sulfates, is more amenable to interrogate as catalysts for more thermally demanding conditions. The unique surface architecture prompted us to further explore the efficacy of these clusters within gas-phase oxidation reactions.

In this study, we report the oxidation of CO over a range of MCe 70 oxysulfate clusters differing in the linkage cation between the clusters (M = Ce 3+/4+ , Cd 2+ , Co 2+ , Cu 2+ , Fe 2+/3+ , Ni 2+ , Zn 2+ ). We demonstrate that the identity of this interfacial unit can alter the catalytic activity of the MCe 70 cluster family. The higher catalytic activity of CuCe 70 was corroborated with enhanced reducibility as observed through TPR-MS and in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) studies. Furthermore, DRIFTS and CO isothermal physisorption measurements suggested stronger CO binding and enhanced CO uptake, respectively, by CuCe 70 . Following the experimental observation of oxygen vacancy defects through Raman spectroscopy, density functional theory (DFT) calculations indicated that oxygen vacancy defect formation energetics differ among the bridging chain sites, with Cu interfacial sites exhibiting the lowest energy of formation, which further rationalizes its superior catalytic activity.

■ RESULTS AND DISCUSSION

The family of MCe 70 torus clusters (M = Ce 3+/4+ , Cd 2+ , Co 2+ , Cu 2+ , Fe 2+/3+ , Ni 2+ , and Zn 2+ ) was synthesized according to published procedures. 28,30 Notably, the MCe 70 clusters feature the shared formula of repeating Ce 70 toroids as previously reported through single-crystal X-ray diffraction (SC-XRD): [Ce 70 (OH) 36 (O) 64 (SO 4 ) 60 (H 2 O) 10 ] 4-(see Table S1 for detailed formula listings). 28 The toroid clusters are linked through sulfate and monomeric cation-based interfacial units (Figure 1). Powder X-ray diffraction (PXRD) measurements of the clusters resulted in patterns consistent with prior reports and exhibited long-range order (Figure S1). Inductively coupled plasma-optical emission spectrometry (ICP-OES) demonstrated the successful incorporation of the linkage metal with a Ce:M ratio ranging from (6.1:1) to (7.6:1) in addition to a consistent Ce:S ratio throughout all clusters (Table S2). The higher ratio of M in the powder samples as compared to the Ce:M ratio derived from SC-XRD (74:3) has been previously observed and documented. Interestingly, the lower concentration of M used to form MCe 70 clusters increased the rate of assembly and M incorporation. 28 The synthetic procedures used to create microcrystalline MCe 70 samples in this study used a smaller concentration of M as compared to the procedures previously reported to generate MCe 70 single crystals. [28][29][30] Nitrogen physisorption measurements were performed at 77 K to measure the apparent surface areas of all clusters (Figure S3), which are in agreement with previously reported surface areas of MCe 70 clusters. 30 Following the characterization of the MCe 70 cluster family, we explored the clusters' performance in CO oxidation in a packed bed reactor across a selected temperature regime (Figures S5-S11). Additionally, we investigated CeO 2 (commercially obtained) to contextualize the conversions of the MCe 70 materials (Figure S12). CeO 2 achieved a 2.5% conversion at 200 °C, which is in the same range of most of the studied MCe 70 catalysts and ∼2-fold lower than CuCe 70 before increasing to as high as 36% at 300 °C. All MCe 70 catalysts were stable up to 200 °C, so the discussion is restricted to these temperatures. Conversion profiles at 200 °C indicated comparable conversions among all of the clusters, except the CuCe 70 cluster, which exhibited a 3-fold higher conversion than CeCe 70 and nearly 2-fold higher than that of the other MCe 70 clusters (Figure 2). Postcatalysis PXRD patterns demonstrated retention in crystallinity without detectable formation of CeO 2 (Figure S2). Thus, we determined that the family of the clusters remained intact, and there is a difference in catalyst reactivity with ( Next, we investigated the conversion profiles of CeCe 70 and CuCe 70 at 200 °C for 16 h on stream (Figure S13) to probe the catalyst stability. In these studies, CeCe 70 exhibited constant reactivity, suggesting that it does not undergo an initial inductive effect that could have resulted in a lower conversion than CuCe 70 . CuCe 70 suffered a loss of ∼20% catalytic activity after 3 h but remained constantly higher than CeCe 70 . Moreover, through X-ray photoelectron spectroscopy (XPS), we determined that CeCe 70 and CuCe 70 both contained similar amounts of Ce 3+ (45%) post catalysis (Figures S16 andS17). From Auger electron spectroscopy (Figure S19), we determined that the Cu oxidation state of CuCe 70 post catalysis was primarily Cu 2+ with some population of Cu 1+ sites; the CuCe 70 Cu 2p kinetic energy of 919 eV was closer in value to that of CuO (920.2 eV) than that of Cu 2 O (916.9 eV).

We further explored the effect of Cu as a linkage element, hypothesizing that the Cu 2+/1+ redox couple affects the overall reducibility of our catalysts given the literature precedence that the Cu 2+ to Cu 1+ redox couple served as a potent redox promoter on CeO 2 -based catalysts. 34 We investigated the temperature-programmed reduction (TPR) of the CeCe 70 and CuCe 70 catalysts under 5% CO/He while connected to a mass spectrometer (MS). We monitored the formation of detected CO 2 and SO 2 to distinguish between the reduction of Ce 4+ to Ce 3+ , a process that must be coupled with the oxidation of CO to CO 2 , as well as the reduction of surface sulfates on the clusters to SO 2 . For CeCe 70 , we observed thermal conductivity detector (TCD) signals centered at 322 and 498 °C, whereas CuCe 70 exhibited TCD signals at lower temperatures of 288 and 480 °C (Figure 3A,B). We therefore determined that CuCe 70 is more reducible at lower temperatures. Within this temperature regime, TPR-MS traces demonstrated [CO 2 ] ≫ [SO 2 ] for the two sets of lower temperature peaks for both catalysts (Figure 3A,B). However, the low conversion of CO to CO 2 in the TPR-MS precludes definite conclusions about the The third TPR peak that appears ∼650 °C in both catalysts features an equivalent production of CO 2 and SO 2 , which complicates any conclusions that can be made about Ce 4+ reduction at this temperature regime (Figure S20). Nonetheless, the TPR-MS data of the lower temperature TPD signals suggests that the identity of the cation in the bridging unit affects reduction within MCe 70 clusters.

We conducted CO/O 2 TPR experiments in which both CO and O 2 (in ∼1:8 ratio) were flowed over CeCe 70 and CuCe 70 across a temperature range of 40-400 °C to gauge the catalysts' abilities to oxidize CO and form CO 2 (Figure S21). Under these conditions, CuCe 70 achieved a 50% conversion of CO at ∼350 °C while CeCe 70 only achieved up to a 25% conversion of CO at 400 °C, indicating the improved oxidative activity of CuCe 70 as suggested by its earlier reduction in CO only TPR. Furthermore, the more reducible nature of CuCe 70 at lower temperatures coincides with its higher activity in the CO oxidation experiments as compared to CeCe 70 . We were motivated to further investigate the evolution of Ce 3+ within CeCe 70 and CuCe 70 to elucidate the effect of the linking cation on the reduction of Ce 4+ .

An available tool to study Ce 3+ formation in CeO 2 -based catalysts is monitoring the spin-orbit transition peak of Ce 3+ appearing ∼2150 cm -1 , 35,36 as demonstrated by Wu et al. 37 Thus, we conducted in situ DRIFTS measurements of recirculated 13 CO in the presence of an 18 O 2 rich atmosphere flowed over CeCe 70 and CuCe 70 . See Section 2 of the Supporting Information for more experimental details. Labeled 13 CO was used to further separate the spectrum of the reagent gas from the 2150 cm -1 region of interest. 38 We intentionally collected spectra with a resolution lower than that of the rotational coupling constants of CO (2 cm -1 ) to remove the vibrational-rotational peaks of gas-phase CO to enable more facile data fitting. 39 As shown in Figure 4b, we detected a more noticeable shoulder around 2150 cm -1 in the DRIFTS of CuCe 70 , suggesting a greater evolution of Ce 3+ in CuCe 70 as compared to CeCe 70 (Figure 4a). Both catalysts demonstrated the formation of CO 2 with peaks appearing between 2250 and 2320 cm -1 . Interestingly, a strong peak at around 1915 cm -1 and a broader peak around 1955 cm -1 appeared during the DRIFTS experiment in CuCe 70 and not in CeCe 70 .

To identify the contributions of the Ce 3+ peak at 2150 cm -1 and explore the unexpected features between 1915 and 1955 cm -1 , we removed the gas-phase 13 CO signal and deconvoluted the resulting residual spectra between 1875 and 2225 cm -1 . Details of background treatment and the deconvolution procedure are located in the Supporting Information (Section 10). The residual peaks of CeCe 70 and CuCe 70 are plotted in Figure 4c,d. Peaks assigned to adsorbed CO were observed at 2040, 2078-2088, and 2105 cm -1 . In the CuCe 70 system, the second peak appeared slightly red-shifted to 2078 cm -1 , indicating a weaker C-O bond and suggesting a stronger CO interaction as compared to the peak centered around 2088 cm -1 in CeCe 70 . Importantly, with the data deconvolution, we determined the amplitude of the Ce 3+ peak at 2150 cm -1 (Tables S3 andS4). To follow the rate of formation of Ce 3+ in each catalyst, we normalized the 2150 cm -1 peak to the adsorbed CO peak at 2040 cm -1 (Figure S36). With these data, we concluded that there is a faster rate of Ce 3+ formation and likely neighboring oxygen vacancy formation in CuCe 70 despite the ∼5% fitting error associated with fits and surface site fluctuations too fast to appear on this time scale. This more facile Ce 3+ formation aligns with the higher catalytic activity of CuCe 70 in CO oxidation.

Additionally, from the deconvolution of the CuCe 70 spectra, we observed a clear increase in the intensity of the lower frequency peaks (1915 and 1955 cm -1 ), while the 2040 and the 2150 cm -1 peaks proportionally decreased. We reasoned that the Cu must be contributing to this interaction given the absence of these low-frequency peaks in CeCe 70 . We determined that the lower frequency peaks are unlikely to result from CO linearly bonding solely to Cu species, as prior reports indicate shifts >2055 cm -1 for Cu x+ -13 CO species (>2100 cm -1 correlates to Cu x+ -12 CO species). 40 Since this region is where bridging CO stretches are expected, 41 the peaks assigned at 1915 and 1955 cm -1 can likely be attributed to (1) bridging CO forming a Cu-CO-Ce dimer and/or (2) bridging CO forming a Ce-CO-Ce dimer that is facilitated by nearby Cu. The ability to form bridging CO in CuCe 70 as opposed to CeCe 70 implied a higher CO uptake in CuCe 70 given this bridging motif is more energetically challenging to access at lower CO concentrations, as previously reported through Monte Carlo simulations. 42 To further explore this hypothesis, we conducted CO isothermal adsorption measurements at 87 K, and we observed an almost 2-fold increase in CO uptake by CuCe 70 as compared to CeCe 70 , consistent with its higher N 2 uptake (Figure S37). The higher CO uptake, which likely enables the unique bridging CO stretches observed in the CuCe 70 DRIFTS spectra, is consistent with the observed superior reactivity of CuCe 70 .

After detecting the evolution of Ce 3+ in our DRIFTS data, we further explored the formation of neighboring O vacancies, species known to directly participate in oxidative or reduction processes, 13,15 using Raman spectroscopy. Measurements under ambient conditions indicated that vacancies are present in the pristine CeCe 70 and CuCe 70 clusters, in agreement with prior studies by Wang et al. (Figure S38). 30 Both catalysts demonstrated a signal around 400 cm -1 that is attributed to surface Ce-O vibrations with a larger intensity than the peak around 465 cm -1 associated with bulk Ce-O vibrations. 43 Herman and co-workers previously demonstrated that a shoulder in Raman spectra of ceria nanoparticles as small as 6 nm at ∼400 cm -1 is due to combined effects of strain and photon confinement. 44 However, more intense surface than bulk features is highly unusual, even among CeO 2 nanoparticles, further highlighting their cluster-like nature as opposed to a nanostructured solid-state character. The intensities found between 500 and 700 cm -1 indicate a rich defect character for both clusters. 45 Given that both clusters are able to accommodate oxygen defects, we conjectured that the ability to form the defects in the Ce 70 toroid system could be affected by the linkage element present.

To further investigate the formation of oxygen vacancy defects within CeCe 70 and CuCe 70 , we calculated and compared the defect formation energy in these systems using density functional theory. Based on previously published structural data of Colliard et al. and Wang et al., we first manually created an idealized cluster model of a Ce 70 toroid with D 10h symmetry. [28][29][30] Due to the missing/disordered hydrogen atoms in the experimental structure solutions, we rigorously tested all reasonable hydrogen placements adhering to the D 10h symmetry. See Section 13 in the Supporting Information for a detailed discussion of possible hydrogen placements. The resulting 22 hydrogen patterns were optimized at the Perdew-Burke-Ernzerhof (PBE) level of theory and the resulting relative energies were compared. We found two hydrogen placements within a range of 9 kJ/mol of each other, while all other placements are found to be more than 450 kJ/mol higher in energy. We therefore chose the hydrogen placement with the lowest energy for our further investigations. The second-lowest energy configuration can also be expected to occur, in line with -O/OH disorder observed in the bond valence sum calculations of Colliard et al. 28 A comparison of all hydrogen placements and their relative energies is given in Section 13 of the Supporting Information.

Using the selected hydrogen placement scheme, we investigated the formation energy (ΔE defect ) for a defect consisting of a single missing neutral oxygen atom using the formula ΔE d e f e c t = E(toroid w. defect) + ( )

. The selected hydrogen placement scheme allows for five distinct neutral oxygen vacancies. The resulting defective toroid was modeled in a triplet spin state, as the spin coupling between Ce atoms adjacent to the defect site is expected to be small and the accuracy of the calculation setups is expected to be lower than the possible effects observable by employing the broken symmetry approach to describe a singlet antiferromagnetic (AFM) spin state. 46 To validate the triplet state, we also calculated the defect formation energies using a singlet spin state of the defective cluster. Absolute energies of the clusters as well as defect formation energies are significantly lower in the triplet state, confirming our spin state choice. See Figure S43 for more details. We note here that the defect formation energies calculated using the singlet state of the defective cluster and bulk ceria are surprisingly close to some values reported in the literature where no spin state is specified. [47][48][49] Our results underline the importance of investigating different spin states (and reporting what spin state was used) when modeling defective ceria using DFT. All calculated defect energies are given in Section 13 of the Supporting Information. We calculated the defect formation energies of the five possible defect sites in the Ce 70 cluster (labeled A-D and F, see Figure 5A,B). The average value is 0.48 eV, with a minimum value of 0.38 eV and one outlier at 0.75 eV (Figure 5C). For comparison, we also calculated the bulk defect energies of bulk CeO 2 and a Ce 6 O 8 (O 2 CH) 8 cluster (an oftenused precursor in Ce-oxo cluster synthesis 20 or building block in metal-organic frameworks 50 and comparable to the Ce 6 octahedron unit in the toroid) using the same methodology to be 2.32 and 0.42 eV, respectively (computational details are given in Section 13 of the Supporting Information). We therefore observe the same trend of decreasing ΔE defect with decreasing particle size that is also observed experimentally. 51,52 Our ΔE defect of bulk ceria is slightly lower than previously reported values calculated using DFT+U (e.g., 3.2 eV by Chen et al. using U f = 6.2 eV). 53 Ziemba et al. calculated ΔE defect for ceria surfaces using DFT+U (U f = 4.5 eV) to be in the range of 1.1-2.1 eV. 54,55 This clearly shows the overall trend of a decreasing ΔE defect for a decreasing dimensionality and extension of the ceria particles and agrees well with our results. Our Ce 70 cluster model furthermore exhibits no significant deviation in ΔE defect from the isolated Ce 6 O 8 (O 2 CH) 8 cluster even though it is an extended system.

To investigate the influence of the bridging Ce and Cu units in the experiment, we extended our model into a onedimensional chain by adding the bridging CeO 4 or CuO 4 unit from the single crystal-derived structure solution to the DFT +U optimized cluster and applying one-dimensional periodic boundary conditions (1D PBC). In the case of Ce, two oxygens were saturated with two H atoms each. For the Cu bridge, two oxygens were saturated with two H atoms each, and the remaining two oxygens were saturated with one H atom each. After the relaxation of the unit cell length and subsequent geometry optimization, the same methodology as for the cluster model was applied to obtain ΔE defect for the five possible defect sites (labeled A-D and F, see Figure 5A,B). The resulting defect formation energies of the Ce-bridged chain (Figure 5C) are in good agreement with the Ce 70 cluster model values. The average ΔE defect is 0.44 eV for the CeCe 70 chain, while the spread is significantly larger than in the Ce 70 cluster model (minimum of 0.13 eV and maximum of 0.67 eV). Figure 5C compares all three model systems and the Ce 6 cluster. More details and a comparison of the effect of varying U d values on the resulting defect formation energies in the Cubridged chains are given in Section 13 of the Supporting Information.

From Figure 5C, it is clearly visible that the substitution of Ce with Cu in the bridging unit results in a significant decrease of ΔE defect . This conclusion agrees well with the experimental observation of increased activity for the Cu system but is still surprising. The chemical nature of the system would, in our opinion, suggest a highly localized nature of the defect formations. The Ce 70 cluster model also suggests a highly localized electron redistribution upon formation of the oxygen vacancy, with the two free electrons localizing in the Ce f orbitals of two adjacent Ce atoms. In the case of the Cebridged periodic model, we also observe no significant charge transfer upon formation of the oxygen vacancy beyond the two adjacent Ce atoms. In the Cu-bridged chain, however, a significant charge transfer toward the Cu atom in the bridge is observed (decrease of Mulliken charge on Cu from 0.47 to 0.18). The previous doublet spin state on Cu transforms into a singlet state in our calculations, suggesting a reduction of the Cu. The defective system, however, still favors the anticipated quartet state and the three unpaired electrons localize on the three Ce atoms adjacent to the oxygen vacancy. This charge transfer is unexpected, and we are at this stage unable to determine whether this is an artifact of the level of theory applied here or if this is indeed close to the true electronic ground state of the system. Further computational and experimental investigations are needed to better model the effects of the bridge on the nearby defect sites. Nonetheless, we are hopeful that our initial systematic study into modeling defect energetics within experimentally derived Ce-oxo clusters will motivate future investigations among other cluster shapes and compositions given the rarity of such inquiries.

While Cu/CeO 2 -based catalysts have been well-studied for CO oxidation, they often are in the form of polydisperse oxides, hindering the understanding of the real active sites. 56 Our inquiry into highly crystalline, discrete clusters provides an alternative approach to derive meaningful insights relating to structure and reactivity with increased confidence of the atomic-level structure. Our results indicated that even a single ion-based linkage unit between neighboring clusters can serve as an active interface that readily tunes the reducibility and defect formation energetics of the Ce 70 ring and consequently the catalytic activity. We envision that our study can further motivate the creation of single/few ion 57 or ultrathin 58 interfaces within ceria-containing nanostructures.

■ CONCLUSIONS

In summary, we have established a correlation between higher catalytic activity for CO oxidation and the identity of the bridging cationic units linking the Ce 70 toroids within the MCe 70 cluster assemblies. We observed the highest CO conversion with CuCe 70 , containing Cu in the bridging units. Through CO-TPR studies, we observed that CuCe 70 is more reducible than CeCe 70 . In situ DRIFTS data suggested a faster formation of Ce 3+ and identified unique CO bridging adsorption bands within CuCe 70 , while CO isotherms confirmed higher CO uptake in CuCe 70 as compared to CeCe 70 . Lastly, calculations determined that the presence of Cu in the bridging unit can substantially lower the energetic requirements to form a single oxygen vacancy in CuCe 70 compared to CeCe 70 . Thus, we conclude that the incorporation of Cu in the bridging chain of the Ce 70 torus family results in superior catalytic activity for CO oxidation due to its higher reducibility, enhanced CO binding and uptake, and more favorable oxygen vacancy defect formation. This investigation into gas-phase reactivity and cluster chemistry demonstrates that subtle atomic changes can affect the assemblies' properties. We are hopeful that this study will inspire further inquiry into this fascinating array of species as well as the expansion of further metal-oxo clusters or bulk metal oxide architectures with unique interfaces.

■ EXPERIMENTAL SECTION

MCe 70 Syntheses. CeCe 70 , CoCe 70 , CuCe 70 , FeCe 70 , and NiCe 70 were synthesized to published procedures for microcrystalline samples. 30 CdCe 70 was synthesized according to published procedure for CdCe 70 �standard, yet was left to sit undisturbed for 5 days. 28 ZnCe 70 was synthesized according to modifications from published procedure and is detailed below. 28 87 mg of Zn(NO 3 ) 2 and 110 mg of Ce(SO 4 ) 2 were placed in 2 mL of water with 20 μL of a 1.0 M aqueous solution of tetrabutylammonium chloride. The clusters were left for 36 h undisturbed at room temperature.

After synthesis, all clusters underwent the following washing procedure:

The as-synthesized cluster solutions were centrifuged to decant the collected solid from the mother solution. The isolated cluster was then washed twice more with deionised (DI) water, followed by centrifugation, and then twice with ethanol. After the second ethanol wash and centrifugation, the isolated powder dried overnight on the benchtop.

Catalyst Reactivity Studies. In a typical experiment, ∼40 mg of catalyst (based on ∼0.3 mmol of Ce 70 ) was diluted with 1 g of quartz sand (trace-metal grade) and packed in a quartz tube plugged with quartz wool. The height of the bed was approximately 1 in. The sample was pretreated at 200 °C for 2 h with a flow rate of 20 sccm of 20% O 2 /He. After pretreatment, the catalyst was immediately exposed to reactant gases of 20 sccm of 0.5% CO/20% O 2 /He and 20 sccm of 20% O 2 /He co-fed with 50 sccm of N 2 . For a typical temperature ramp-up and ramp-down profile, the reactor was first cooled to 150 °C at a rate of 10 °C/min and held for 2 h, then ramped to 200 °C at a ramp rate of 10 °C/min and held for 2 h, then ramped to 250 °C at a ramp rate of 10 °C/min and held for 2 h, then ramped to 300 °C at a ramp rate of 10 °C/min and held for 4 h, then cooled to 250 °C at a rate of 10 °C/min and held for 2 h, then cooled to 200 °C at a rate of 10 °C/min and held for 2 h, and then cooled to 150 °C at a rate of 10 °C/min and held for 2 h before a final cool down to 40 °C. To generate the 200 °C conversion data to gather triplicate data points (Figure 1 in MS), a typical experiment repeated the pretreatment protocol of heating the reactor tube at 200 °C for 2 h with a flow rate of 20 sccm of 20% O 2 /He. To remain consistent with the prior temperature ramp profile described above, the catalyst was first cooled to 150 °C and held for 2 h under the reactant conditions described above before ramping to 200 °C at 10 °C/min and held for 2 h to generate the additional trials used for catalyst comparisons. For the stability studies of CeCe 70 and CuCe 70 , each catalyst was again pretreated at 200 °C for 2 h with a flow rate of 20 sccm of 20% O 2 / He. Then, the catalyst was heated at 200 °C for over 16 h under the reactant conditions mentioned above to remain consistent with the other studies.

CO Temperature-Programmed Reduction Experiments. CO TPR. Approximately 70 mg of material was placed in a quartz reaction tube plugged between quartz wool and quartz sand. Temperatureprogrammed experiments were performed on an Altamira AMI-200 instrument equipped with a thermal conductivity detector (TCD). In a typical experiment, catalysts were pretreated at 200 °C for 2 h under 30 cc/min under 10% O 2 /He and then cooled to 40 °C in 30 sccm He. Then, the system was heated to 900 °C at a ramp rate of 10 °C/ min under 25 sccm of 5% CO/He with a carrier gas flow of 5 cc/min of He. Then, the system was cooled to 40 °C at a ramp rate of 30 °C/ min under 30 cc/min of N 2 .

CO/O 2 TPR. Approximately 30 mg of material was placed in a quartz reaction tube mixed with white quartz sand (Aldrich) to a height of 1 cm and plugged between layers of quartz wool and quartz sand. Temperature-programmed reaction experiments were performed on an Altamira AMI-200 instrument equipped with an inline atmospheric mass spectrometer (an SRS RGA100 quadruple mass spectrometer with a custom-built sampling manifold). In a typical experiment, the sample was reduced in 10% H 2 /N 2 (Airgas, certified) at 200 °C for 2 h prior to the CO/O 2 TPR. The sample was then cooled down to 40 °C under He (Airgas, UHP) flowing at 30 sccm and finally heated to 400 °C at a ramp rate of 10 °C/min under 10 sccm of 5% CO/He (Airgas, certified) and 40 sccm of 10% O 2 /He (Airgas, certified). Ionic masses, m/z, of 28 (CO), 44 (CO 2 ), 32 (O 2 ), and 4 (He) were continuously monitored via atmospheric mass spectrometer. Molar fractions given in the plots were corrected for ionic sensitivities and fragmentation patterns and all normalized including He (not shown).

In Situ DRIFTS Measurements. In situ 13 CO DRIFTS measurements were carried out using a Nicolet iS50 Fourier-transform infrared (FTIR) spectrometer equipped with a Harrick Scientific Praying Mantis DRIFTS accessory and a high-temperature reaction chamber. Each sample (12 mg of each) was crushed using a mortar and pestle and placed in the sample cup of the DRIFT reaction chamber on top of a bed of KBr. The temperature was raised to 200 °C under 10% O 2 /Ar (Airgas, UHP grade) flowing at 50 sccm. A background spectrum was collected after 200 °C for 2 h and used as a reference for the subsequent spectra. Then, a gaseous mixture of 13 CO (20.0 mbar) and 18 O 2 (50 mbar) in He (1000 mbar) was recirculated through a custom-built system equilibrated at 1070 mbar. Infrared (IR) spectra were acquired over the course of ∼2 h until spectral changes in the CO and CO 2 stretch region (∼2200-1900 cm -1 ) were negligible. The spectra resolution was 4 cm -1 . The system was evacuated under Ar as the reaction chamber cooled back down.