Highly stable preferential carbon monoxide oxidation by dinuclear heterogeneous catalysts

Significance

We have carried out atomically resolved STEM observations of catalyst behaviors in operando. The data intuitively illustrate how dinuclear catalysts are more stable than single atom ones under the reaction conditions. Ourresults shed new light on the role of different oxygen atoms near the active center. An oxygen atom that links asingle atom catalyst and the supporting substrate often plays multiple roles. Through extensive computationalefforts, in concert with experimental verifications, we learned that bridge oxygen mainly serves to stabilize thecatalyst, while interfacial oxygen acts as the participant in the chemical processes for activity. The knowledgeis expected to be broadly impactful to the community.

Abstract

Atomically dispersed catalysts have been shown highly active for preferential oxidation of carbon monoxide in the presence of excess hydrogen (PROX). However, their stability has been less than ideal. We show here that the introduction of a structural component to minimize diffusion of the active metal center can greatly improve the stability without compromising the activity. Using an Ir dinuclear heterogeneous catalyst (DHC) as a study platform, we identify two types of oxygen species, interfacial and bridge, that work in concert to enable both activity and stability. The work sheds important light on the synergistic effect between the active metal center and the supporting substrate and may find broad applications for the use of atomically dispersed catalysts.

Preferential CO oxidation in the presence of excess H₂ (PROX) promises a route to removing CO as a key contaminant in H₂ for a wide range of applications (1). Significant research has been attracted to carrying out PROX at low temperatures (e.g., <200 °C) so as to maximize the selectivity toward CO removal rather than H₂ oxidation (2). Driven by the understanding that strong CO binding between adjacent metal atoms in metallic nanoparticles (NPs) would block the active site, which leads to inferior activity, much of the recent attention has been directed toward studying atomically dispersed catalysts. Indeed, exciting progress has been made. Outstanding per atom catalytic activity, for example, has been reported on Pt single-atom catalysts (SACs) dispersed on a variety of supports (3, 4). Moving forward, how to maintain the high activity for prolonged operations becomes a critical issue that has received relatively little attention. The handful of studies that address the stability issue of atomically dispersed catalysts describe a strong dependence of such stability on the substrate support (5, 6). That is, the interactions between the metal active center and the supporting substrate (often metal oxides) are critical to the performance of the atomically dispersed catalysts (7, 8). While intuitive, this observation raises critical new questions concerning the mechanisms by which PROX proceeds on a single-atom site (9). A growing body of evidence suggests that the synergistic effect between the active center and the substrate, mediated by the interfacial atoms (often O-based species), is of vital importance (10–14). However, while the pivotal role played by interfacial O atoms toward activity has been inferred in the literature (15), the implications of these species to the stability of the catalyst under operating conditions are not fully understood. More importantly, it still has a knowledge gap on how to capitalize on the existing knowledge to achieve atomically dispersed catalysts that are both active and stable. Here, we report a study aimed at filling this gap. Using a combination of super high-resolution in situ imaging and density functional theory (DFT) calculations, we identified two types of O atoms, bridge and interfacial, in the atomically dispersed dinuclear Ir catalysts (Fig. 1A). It was found that outstanding stability can be afforded by the bridge O atom between two active metal centers, whereas the activity was mainly connected to the interfacial O atom between the active center and the supporting substrate. These results shed important light on the principles governing the stability and activity of atomically dispersed catalysts.

Fig. 1.

Direct characterization of Ir DHC/CeO₂ catalyst. (A) DFT optimized the minimum energy structure of Ir DHC on CeO₂ (side view). The yellow and green balls indicate Ir and Ce; the blue, brown, red, and brandy rose balls represent interfacial OH, bridge O atoms, interfacial O atoms, and O atoms in the bulk and other environments, respectively. (B) A false-colored HAADF-STEM image of Ir DHC/CeO₂ along the [$10\stackrel{-}{1}$] zone axis of CeO₂, where the Ir atoms are highlighted (within yellow dashed rectangles). (C) A magnified view of a single Ir DHC on CeO₂. Normalized intensity profile taken along the line indicated by the yellow arrows. (D) EXAFS experimental and fitted spectra based on the DFT atomic model of Ir DHC on CeO₂ and bulk Ir foil at the Ir L-edge. (E) In situ DRIFTS spectra of Ir DHCs.

This study was enabled by a unique study platform, the Ir dinuclear heterogeneous catalysts (DHCs) on CeO₂. It features an active center consisting of two Ir atoms, linked and separated by bridge O species. The unit is anchored onto the CeO₂ support by interfacial O species (Fig. 1A and SI Appendix, Figs. S1 and S2). More details about the preparation are provided in the SI Appendix. As shown in Fig. 1B and SI Appendix, Fig. S3, the dinuclear nature is most directly revealed by ex situ aberration-corrected scanning transmission electron microscopy (AC-STEM). The high angle annular dark field-STEM (HAADF-STEM) image as shown in Fig. 1B was taken along the [$10\overline{1}$] zone axis of CeO₂. Notably, under this specific condition with UV treatment for 20 min after immersing the CeO₂ in an Ir dimer precursor solution (16, 17), we observed Ir heterogeneous atoms in pairs. In Fig. 1C, a single pair was examined (Top), where a line scan of the Z-contrast (Bottom) permitted the measurement of the atomic spacing between the two Ir atoms at ca. 3.7 Å. This value is in excellent agreement with the DFT-optimized structure model as shown in Fig. 1A, where Ir atoms are coordinated by O atoms with an average Ir–O distance of 2.0 Å, and the two Ir centers are connected by a mono-µ-oxo bridge with an Ir–Ir distance of 3.7 Å (Ir–O–Ir). The structure of the Ir DHCs was further studied by extended X-ray absorption fine structure spectroscopy (EXAFS) and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS). The EXAFS data shown in Fig. 1D provided strong support to the DFT structure model, which is evidenced by the good match between the simulated and experimental spectra. The presence of Ir–O bonds in the first coordination shell as well as an Ir–O–Ir and Ir–O–Ce interaction in the second coordination shell from the EXAFS simulation (SI Appendix, Fig. S4 and Table S2) also confirmed that there were no significant byproducts such as nanoscale clusters (often referred to as NPs). The homogeneity of the Ir DHCs on CeO₂ was also further supported by the DRIFTS spectra (Fig. 1E). Two characteristic singlet peaks at 2,098 and 2,052 cm⁻¹ can be ascribed to the asymmetric and symmetric stretching of the CO probe on Ir DHC, respectively. Together, this set of data demonstrates that we have successfully obtained monodispersed Ir DHCs on CeO₂, whose structure is resolved as shown in Fig. 1A.

With the structural information confirmed, we next performed PROX using Ir DHCs on CeO₂ at 453 K (Fig. 2A). As an important control, Ir SACs with well-defined structures (SI Appendix, Fig. S5) were also prepared and subjected to similar catalytic conditions. It was observed that while Ir DHCs and SACs exhibited comparable initial activity, the main benefit of Ir DHCs was outstanding durability, with 7% decay for the first 120 h. In contrast, only 75% of the initial activity was measured on Ir SACs at 40 h, which further decreased to 63% after 120 h.

Fig. 2.

Catalytic stability characterization: (A) CO PROX stability test over Ir DHC and SAC on CeO₂. Reaction conditions: 40 mL/min flow rate, 1% CO, 1% O₂, 40% H₂, bal. He, 50 mg catalyst loading, T = 453 K; (B) In situ EXAFS and (C) XANES spectra of Ir DHC and SAC. Reaction conditions: 20 mL/min flow rate, 1% CO, 1% O₂, 40% H₂, bal. He, 473 K, 2 h, ~10 mg catalyst loading.

To help us better understand the stability exhibited by Ir DHCs, we compared our catalyst with various other recently reported catalysts by using 5% decay of CO conversion as a benchmark. It is obvious from SI Appendix, Fig. S6 that our catalyst is among the best (2, 3, 11, 18, 19). It is cautioned that the detailed structures of the reported catalysts, their inherent activities, as well as the test conditions, tend to vary greatly; moreover, a protocol that allows for unbiased, direct comparison is not yet available. Even with these constraints in mind, we can clearly see the outstanding performance exhibited by Ir DHCs in both activity and stability. Most strikingly, the Ir DHCs were much more stable than Ir SACs despite similar synthesis conditions, almost identical loading amounts of active elements, and the same supporting substrate. Characterization of the Ir DHC and SAC after the reaction was carried out. A negligible change was observed for the Ir DHC (SI Appendix, Fig. S7), whereas the features characteristic of Ir NPs can be found in Ir SAC samples after the reaction (SI Appendix, Fig. S8). The result suggests that significant Ir aggregation took place for the Ir SAC but not for the Ir DHC. Intrigued by the observation, we were prompted to ask the following question: What is the key difference between DHCs and SACs that leads to such a stark contrast in their stability?

Answers to this research question have significant implications. While a range of SACs has been shown active toward PROX, a critical question remains unanswered. It concerns the mechanism by which O₂ is activated by the single-atom site where CO is also strongly bonded. The most compelling working hypothesis invokes the participation of various O species such as the interfacial O, which serves to anchor the single-atom site to the supporting oxide or the lattice O that is part of the supporting substrate or both (20–22). The Ir DHCs are different from Ir SACs and, by extension of structural similarities, many reported SACs of other compositions, in that they feature additional O species like the bridge O (Fig. 1A). Moreover, the close vicinity of a second Ir atom that is not directly bonded with Ir (i.e., no metal–metal interactions) might enable synergistic effects for simultaneous O₂ and CO activation. We are, therefore, encouraged to explore two new possibilities: 1) Bridge O may enhance the stability of atomically dispersed active sites and 2) synergistic effects between adjacent Ir atoms may promote PROX. As will be discussed next, our results confirmed the former but ruled out the latter.

To test the first possibility, we have examined the structural changes of Ir SACs and DHCs during the PROX reaction. While no significant difference was observed between the EXAFS spectra for Ir DHCs before (Fig. 1D) and after PROX (Fig. 2B), a rise of the peak that is indicative of Ir–Ir bonding became apparent for Ir SACs under in operando XAS at 473 K (ambient pressure, 1% CO, 1% O₂, 40% H₂ balanced by He). The aggregation of Ir SAC was also validated by X-ray absorption near edge spectroscopy (XANES) spectra (Fig. 2C), which supported that the average valence state of Ir in Ir SACs decreased. This set of data implies that agglomeration of Ir SACs took place during PROX. Under identical conditions, similar changes were not observed for Ir DHCs. From the structural models of Ir DHC and SAC, both Ir SAC and DHC are anchored on the surface by interfacial O species (OH and O), while Ir DHC has one additional bridge O atom to hold the two Ir together (Fig. 1A). During reactions, the formal oxidation state of Ir in SAC is reduced to +1, whereas those in DHC are in mixed valences of +1 and +3, which was also validated by XANES. Of them, the Ir⁺ acts to activate O₂ for CO oxidation, and the Ir³⁺, which is coordinatively stable, provides the structural stability to prevent the DHC from moving around.

To obtain direct evidence that supports this understanding, we carried out in situ AC-STEM observations by placing Ir DHCs and SACs inside a TEM nanoreactor and exposed them to CO gas (760 Torr of 5% CO, SI Appendix, Figs. S9 and S10) and a CO/H₂ gas mixture (400 Torr of 5% H₂ and 10 Torr of 5% CO, Fig. 3 A and B), respectively, at 473 K. In these experiments, internal features of CeO₂ (such as grain boundaries in Fig. 3B) were used as alignment marks for locating the relative positions of the individual Ir DHCs and SACs. A control experiment under pure N₂ showed negligible electron beam effect of image acquisition on the Ir species (SI Appendix, Fig. S11 and associated discussions therein) (23). Under conditions similar to PROX (400 Torr of 5% H₂ and 10 Torr of 5% CO), a diffusion of 15.4 Å was observed for Ir SACs within the first 5 min; then, an additional diffusion distance of 16.2 Å was measured between 5 min and 10 min (Fig. 3B). By contrast, no measurable movement was observed for Ir DHCs under the same conditions for 60 min (Fig. 3A). It is noted that the strong binding of CO has been previously reported to weaken the anchoring of SACs on metal oxide supports (24). Our DFT calculations support the observation, too. It is seen in Fig. 3C and SI Appendix, Figs. S12 and S13 that the diffusion of Ir SACs on the CeO₂ (110) facet is an isothermal process with an estimated activation energy of 1.6 eV, indicating that diffusion is highly possible. The detachment of Ir DHCs on CeO₂, however, incurs a significant thermodynamic penalty of ca. 4.9 eV, making detachment an unlikely pathway (Fig. 3D). Diffusion of Ir DHCs as a whole is highly unlikely due to the requirement of breaking six Ir–O bonds simultaneously. Therefore, we considered a stepwise diffusion process, where the Ir⁺ center diffuses first to lead to the breaking of the Ir–O–Ir structural motif; this process would incur a 1.6 eV increase of the thermodynamic energy with an estimated activation energy of 3.2 eV (Fig. 3E) and is also unlikely. A more likely route by which Ir DHCs can be removed from the surface would be the initial dissociation of the two Ir ions on the surface. Indeed, by exposing the Ir DHC under intentionally strong electron beam irradiation in the vacuum chamber during the AC-STEM observation, we only observed dissociation of Ir DHCs, which led to the formation of Ir SACs that exhibited diffusion behaviors subsequently (SI Appendix, Fig. S14). It is conceivable that the diffusion of Ir SACs would lead to eventual aggregations that form Ir NPs (SI Appendix, Fig. S15), which would help to explain the spectroscopic features observed in Fig. 2B. The understanding is also confirmed by STEM and DRIFTS data for postreaction samples (SI Appendix, Fig. S8). Taken as a whole, this set of data confirms that the dinuclear nature of the Ir DHCs greatly enhances their stability as compared with SACs. With bridge O being the key differentiating feature between Ir DHCs and SACs, it is concluded that the stability directly benefits from the presence of this species. Next, we will demonstrate that bridge O indeed does not participate in the catalytic cycles.

Fig. 3.

Direct stability observation. In situ AC-STEM characterization of Ir DHC/CeO₂ under gas-phase condition. In situ HAADF-STEM images showing the mobility of Ir DHC (A) and SAC (B) with the gas phase (400 Torr of 5% H₂ + 10 Torr of 5% CO, 473 K) to mimic the PROX condition (Scale bar, 2 nm.) The colored circles highlight the same Ir SAC and Ir DHC. Calculated reaction energy changes (ΔE) and associated barrier height (ΔE^‡) for (C) diffusion of Ir SAC, (D) detachment of Ir DHC and (E) diffusion of Ir DHC on CeO₂ surface assisted by one CO.

A growing body of literature on SACs-based PROX or CO oxidation supports that interfacial O plays a critical role in the catalytic cycle. The working hypotheses often involve the combination of interfacial O with adsorbed CO, resulting in CO₂ as a leaving product and an O vacancy to be replenished by adsorbed O₂, which completes the catalytic cycle. When examining the structural details of Ir DHCs and SACs, we propose that the same function could be played by bridge O, as well, in which case a fundamental difference would be expected in the apparent activation energies of the overall reactions. To test this possibility, we next performed CO oxidation at varying temperatures (Fig. 4A). CO was already converted to CO₂ by Ir DHC at temperatures as low as 373 K and was almost depleted at 453 K. A kinetic study was carried out in the kinetically controlled region (<20% conversion) to construct the Arrhenius plots as shown in Fig. 4B. The apparent activation energies for Ir DHCs and SACs were nearly identical. As such, CO oxidation might proceed on the two types of catalysts following similar mechanisms or share a common step that determines the activation energy, which would invoke the participation of interfacial O but not bridge O.

Fig. 4.

Low-temperature activity initiated by interfacial oxygen. (A) CO conversion by Ir DHC for CO PROX reaction at different temperatures. Test conditions: 30 mL/min flow rate, 1% CO, 1% O₂, 40% H₂, bal. He, 250 mg catalyst loading. (B) Arrhenius-type plot of CO PROX reaction over Ir SAC and DHC. Test conditions: 100 mL/min flow rate, 1% CO, 1% O₂, 40% H₂, bal. He, 100 mg catalyst loading, conversion of CO is controlled below 20% to be in the kinetic region. (C) The reaction pathway of catalytic CO oxidation over Ir DHC on CeO₂ suggested by DFT calculations. The vertical axis shows the relative energies of intermediates and transition states in the unit of eV. The DFT-optimized structures of intermediates are shown with labels associated with them.

The understanding is strongly supported by computational studies of the catalytic cycle (25, 26), the results of which are detailed in Fig. 4C and SI Appendix, Figs. S16–S21 and Tables S3–S7. Our proposed catalytic mechanism starts with a di-carbonyl intermediate I that features 2 Ir³⁺ centers connected by a bridge O atom, as suggested by in situ DRIFTS spectra and computational study of Ir SAC and DHC under H₂ and CO atmosphere. As shown in Fig. 4C, there are interfacial O species (O and OH) and a bridge O atom in intermediate I. Therefore, there are three possible pathways for CO to react with the nearby O atoms. As shown in SI Appendix, Fig. S19, only when the CO couples with an interfacial O atom, the reaction is exothermic, forming a five-coordinated Ir center with a chemically adsorbed CO₂ (intermediate II). Intermediate II releases a CO₂ molecule to form a mixed-valent intermediate III with one Ir in oxidation state (III) and one Ir in oxidation state (I). The four-coordinated Ir⁺ center has a square-planar geometry, a typical geometry for an Ir⁺ complex (25). The overall process from intermediate I to intermediate III is a typical reductive elimination reaction of Ir³⁺ complexes. As highlighted in Fig. 4C, intermediate III has an oxygen vacancy near the Ir⁺ center which was previously occupied by an interfacial O atom. The oxygen vacancy in intermediate III can bind to an O₂ molecule at the Ir⁺ center with an energy change of −1.02 eV to form intermediate IV, which undergoes O–O cleavage to form intermediate V with an Ir⁴⁺–O unit. A CO molecule from the gas phase can be physically adsorbed by intermediate V, forming intermediate VI and then oxidized to CO₂ and intermediate VII (27) with a large energy change of −2.06 eV. The release of the second CO₂ in intermediate VII generates intermediate VIII with a coordinatively unsaturated Ir³⁺ center, which can be replenished by a CO molecule from the gas phase to complete the catalytic cycle. In the proposed catalytical cycle, we were able to locate two transition states: one for CO reaction with an interfacial O atom to form CO₂ (TS1 between intermediates I and II) and one for the O–O cleavage to activate O₂ (TS2 connects intermediates IV and V). The oxidation of the first CO molecule is the rate-determining step (rds) with a calculated activation energy of 0.78 eV, consistent with the experimentally measured apparent activation energy.

In the proposed catalytic cycle, an O vacancy is generated in situ through the oxidation of the first CO molecule by an interfacial O atom, which activates gas phase O₂ between the Ir⁺ center and CeO₂ surface to oxidize the second CO molecule and replenish the interfacial oxygen. Intermediate III is a mixed-valent Ir dimer with one Ir in oxidation state 3+ and one in oxidation state 1+. It is highly unlikely to have both Ir³⁺ being reduced to Ir⁺ under reaction conditions since the replenishment of the O vacancy by O₂ is highly exothermic and thermodynamically favorable (Fig. 4C). More discussions about the mechanism are provided in the SI Appendix. It is emphasized here that only one Ir center in the DHC and the interfacial O are involved in the catalytic cycle, accounting for the high activity of both Ir SAC and DHC and their similar energy profile along the reaction path (SI Appendix, Fig. S20), whereas the bridging O atom serves to hold the two Ir centers together and the other Ir provides strong anchoring to the substrate. Together, the DHC affords superior stability without compromising activity.

In conclusion, two types of O species are identified for atomically dispersed Ir catalysts. The first is interfacial O which connects the active metal center to the substrate. It participates in the PROX reaction by adsorbing O₂ and oxidizing CO and may be considered a critical component for the reactivity. Similar O species are abundant in SACs, which may help explain the high activity that has been reported on these catalysts. The second type of O species bridges two metal active centers and is independent of the metal–substrate interactions. There is no evidence that bridge O participates in the catalytic cycles of PROX. Rather, bridge O mainly helps to minimize diffusion of the metal active centers to afford outstanding stability. Building on the recent successes in achieving high activity on atomically dispersed catalysts, our results imply that introducing structural components may contribute significantly to the eventual goal of achieving catalysts that are both active and stable.

Code Availability Statement.

The computer code and algorithm generated during and/or analyzed during the current study have been included in the SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The datasets generated during and/or analyzed during the current study have been included in the manuscript or the SI Appendix.